CHAPTER 5 GIRAFFE User Manual V4.2
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 292
| Download | |
| Open PDF In Browser | View PDF |
GIRAFFE USER’S MANUAL
Version 4.2
January 2008
Cornell University
School of Civil & Environmental Engineering
Ithaca, NY
TABLE OF CONTENTS
Page
TABLE OF CONTENTS ................................................................................................................. i
LIST OF TABLES ......................................................................................................................... iv
LIST OF FIGURES ...................................................................................................................... vii
CHAPTER 1
INTRODUCTION
1
1.1
Background
1
1.2
Scope
2
CHAPTER 2
OVERVIEW OF GIRAFFE SIMULATIONS
3
2.1
Introduction
3
2.2
System Definition
3
2.3
System Damage
5
2.4
Earthquake Demand Simulation
5
2.5
Hydraulic Network Analysis
5
2.6
Compilation of Results
6
2.6.1
Hydraulic Network Analysis Results
7
2.6.2
Performance Index
7
HYDRAULIC NETWORK ANALYSIS
9
CHAPTER 3
3.1
Introduction
9
3.2
Components in Hydraulic Networks
9
3.3
Governing Laws
3.4
10
3.3.1
Equation of Continuity
11
3.3.2
Bernoulli Equation
11
Energy Losses
3.4.1
12
Frictional Loss
12
i
3.4.2
Minor Loss
13
3.5
Energy Gains
15
3.6
Flow Equations
17
3.7
EPANET
18
3.8
3.7.1
EPANET Hydraulic Network Components
19
3.7.2
EPANET Input File
23
3.7.3
EPANET Hydraulic Simulation Methodology
24
3.7.4
EPANET Output File
25
Negative Pressure Treatment
CHAPTER 4
25
PIPE DAMAGE MODELING
28
4.1
Introduction
28
4.2
Definitions
29
4.3
Pipe Leak Simulation
29
4.3.1
Hydraulic Model
29
4.3.2
Leak Classification
31
4.3.3
Probability of Leak Types
38
4.4.
Pipe Break Simulation
38
4.5
Implementation of Pipe Damage Models
39
4.5.1
Deterministic Implementation
40
4.5.2
Probabilistic Implementation
40
EARTHQUAKE DEMAND SIMUALTION
44
CHAPTER 5
5.1
Introduction
44
5.2
Methodology
45
CHAPTER 6
GIRAFFE INPUTS AND OUTPUTS
49
6.1
Introduction
49
6.2
Inputs
49
6.2.1
Control Parameters
49
6.2.2
Deterministic Simulations
50
ii
6.2.3
Monte Carlo with Fixed Simulation Runs
53
6.2.4
Monte Carlo with Flexible Simulation Runs
56
6.3
Definition Parameters
57
6.4
Outputs
61
6.4.1
Deterministic Simulations
61
6.4.2
Monte Carlo Simulations
61
GIRAFFE SIMULATION EXAMPLES
63
CHAPTER 7
7.1
Introduction
63
7.2
Hydraulic Network Model
63
7.3
Deterministic Simulations
68
7.3.1
Inputs
68
7.3.2
Simulation Procedures
70
7.3.3
Outputs
71
7.4
Monte Carlo with Fixed Simulation Runs
79
7.5
Monte Carlo with Flexible Simulation Runs
85
REFERENCES
88
APPENDIX A
GIRAFFE QUICK START TUTORIAL
APPENDIX B
GIRAFFE INPUT PREPARATION
APPENDIX C
GIRAFFE INPUT PREPARATION AND OUTPUT VISUALIZATION
USING MANIFOLD GIS
APPENDIX D
FRAGILITY MODULE
APPENDIX E
FLOW AND NETWORK NONLINEARITIES
iii
LIST OF TABLES
Table
No.
Page
No.
3.1
Frictional Head Loss Evaluation Formulas
14
3.2
Summary Table for Physical Components in an EPANET Hydraulic
21
Network Model
3.3
Sections in an EPANET Input File
24
4.1
Probability of Leak Types for Different Pipelines
39
6.1
GIRAFFE Control Parameters
50
6.2
Input Parameter for Pipe Damage Generation File for Deterministic
51
Simulations
6.3
Input File for Pipe Damage Generation for Deterministic Simulations
51
6.4
Description of Columns in Pipe Break Section
52
6.5
Description of Columns in Pipe Leak Section
53
6.6
Input Parameters for Monte Carlo Simulations with Fixed Simulation Runs
54
6.7
Input File for Probabilistic Pipe Damage Generation
55
6.8
Description of Columns in Probabilistic Pipe Damage Input File
55
6.9
Input File for Earthquake Demand Simulation
56
6.10
Description of Columns in Earthquake Demand Simulation Input File
56
iv
6.11
Input Parameters for Monte Carlo Simulations with Flexible Simulation
57
Runs
Table
No.
Page
No.
6.12
Parameter Definition File
58
7.1
EPANET Format System Definitions File
66
7.2
Input File for Pipe Damage Information for Deterministic Simulation
70
7.3
Damaged System at Time 0
73
7.4
Junction Results at Time 0
76
7.5
Tank Results at Time 0
77
7.6
Pipe Results at Time 0
77
7.7
Pump Results at Time 0
77
7.8
Valve Results at Time 0
77
7.9
Serviceability at Time 0
77
7.10
Junction Results at Time 24
79
7.11
Tank Results at Time 24
79
7.12
Pipe Results at Time 24
80
7.13
Pump Results at Time 24
80
7.14
Valve Results at Time 24
80
7.15
Serviceability at Time 24
80
v
7.16
Pipe Damage Input File for Monte Carlo Simulation with Fixed Simulation
82
Runs
Table
No.
7.17
Page
No.
Input File for Simulating Earthquake Demand for Monte Carlo Simulation
82
with Fixed Simulation Runs
7.18
Damaged System for the Last Run of Monte Carlo Simulation
83
7.19
Serviceability of Monte Carlo Simulation with Fixed Simulation Runs at
86
Time 0
7.20
Serviceability of Monte Carlo Simulation with Fixed Simulation Runs at
86
Time 24
7.21
Serviceability of Monte Carlo Simulation with Flexible Simulation Runs at
87
Time 0
7.22
Serviceability of Monte Carlo Simulation with Flexible Simulation Runs at
Time 24
vi
87
LIST OF FIGURES
Figure
No.
Page
No.
2.1
GIRAFFE Simulation Flow Chart
4
3.1
General Shape of Pump Characteristic Curve
16
3.2
Pump Operation Point
17
3.3
Physical Components in an EPANET Hydraulic Network
20
3.4
Negative Pressure Node Demonstration (after Markov, et al., 1994)
27
4.1
Comparison Between Model Predictions and Sprinkler Data
30
4.2
Hydraulic Model for Pipe Leak
31
4.3
Schematic Drawing of Annular Disengagement
33
4.4
Schematic Drawing of Round Crack
34
4.5
Schematic Drawing of Longitudinal Crack
35
4.6
Schematic Drawing of Local Loss of Pipe Wall
36
4.7
Schematic Drawing of Local Tear of Pipe Wall
38
4.8
Hydraulic Model for Pipe Break
40
4.9
Poisson Process for Pipe Damage Generation
42
5.1
LADWP Water Supply System
46
5.2
Overlay of Distribution and Trunk system
46
5.3
Prediction of Normalized Demand
46
vii
Figure
No.
Page
No.
6.1
Configuration Window for System Options
58
7.1
Hydraulic Network Model Constructed by H2ONET
65
7.2
Hydraulic Simulation Results for Undamaged System from EPANET
68
7.3
Inputs for Deterministic Simulation
70
7.4
Damaged System at Time 0
75
7.5
Simulation Results at Time 0
76
7.6
GIS Map for GIRAFFE Simulation Results at Time 0
78
7.7
Simulation Results at Time 24
79
7.8
Inputs for Monte Carlo Simulation with Fixed Simulation Runs
81
7.9
Inputs for Monte Carlo Simulation with Flexible Simulation Runs
87
7.10
Pop-Up Window with Results
87
viii
CHAPTER 1
INTRODUCTION
1.1
BACKGROUND
Water supplies constitute a key component of critical civil infrastructure that supports fire
protection and provides water for potable household consumption as well as industrial and
commercial uses.
Water is conveyed mostly in underground pipelines.
Thus, ground
movements triggered by earthquakes have a direct effect on the integrity and reliability of water
distribution networks. Water supplies are vulnerable to earthquakes. This vulnerability has been
demonstrated by extensive damage sustained during previous earthquakes, such as the 1906 San
Francisco (e.g., Schussler, 1906; Manson, 1908; Lawson, 1908), 1971 San Fernando (e.g.,
Steinbrugge, et al., 1971; Eguchi, 1982), and 1994 Northridge (e.g., Lund and Cooper, 1995;
Hall, 1995; Eguchi and Chung, 1995; O’Rourke, et al., 2001) earthquakes. Earthquake damage
to water supply systems may disrupt residential, commercial, and industrial activities; impair
fire-fighting capacities; and prolong local community recovery in the aftermath of earthquakes.
It is very important, therefore, to model the earthquake performance of water supply systems in a
robust and reliable way for emergency planning, community restoration, and assessment of
regional economic impacts.
Earthquake performance of a water supply system depends on the available flows and
pressures in the damaged system. The flows and pressures can be predicted using hydraulic
network analysis, which involves solving a set of linear and/or nonlinear algebraic equations,
normally by means of computer programs. Commercial hydraulic network analysis software
packages are designed for undamaged systems, and may predict unrealistically high negative
pressures when used for damaged systems. Hydraulic network analysis results with negative
pressures are inaccurate. Real water supply systems are not air tight, and thus their ability to
support negative pressures is limited. To simulate the seismic performance of water supply
systems, earthquake damage to pipelines needs to be added to the network and then hydraulic
simulation performed using the damaged network. There are no pipe break or leak simulation
1
algorithms in commercial software packages. It is therefore important to develop an algorithm to
model pipe breaks and leaks, and integrate this algorithm into an analysis program for simulation
purposes.
A computer program, GIRAFFE, has been developed for the hydraulic network
simulations of heavily damaged water supply systems. GIRAFFE stands for Graphical Iterative
Response Analysis for Flow Following Earthquakes. It involves over 10,000 lines of C++ code
and works iteratively with the EPANET hydraulic network analysis engine. GIRAFFE embodies
an iterative procedure for negative pressure elimination, methods for simulating pipeline breaks
and leaks, and the simulation of earthquake demands associated with distribution networks.
GIRAFFE can perform both deterministic and probabilistic simulations, and provides results
which can be directly linked to GIS to conduct spatial analysis and map presentations.
This manual is written to provide users with a tool for understanding the main features,
modeling methodology, and input and output parameters and data files for GIRAFFE
simulations.
Selected examples are presented to help users to understand the GIRAFFE
simulation procedures.
1.2
SCOPE
This manual is divided into 7 chapters. The first chapter provides the background of
GIRAFFE. Chapter 2 presents an overview of GIRAFFE simulations.
Chapters 3 to 5 present the methodologies applied in GIRAFFE simulations. Chapter 3
provides an introduction of hydraulic network analyses and negative pressure treatment. Chapter
4 describes the pipe damage modeling methodology applied in GIRAFFE. Chapter 5 presents
the methodology used for earthquake demand simulations associated with distribution networks.
Chapter 6 provides a detailed description of the GIRAFFE input and output parameters
and data files. Chapter 7 provides three examples associated with the three GIRAFFE simulation
2
options, which are deterministic, Monte Carlo with fixed simulation runs, and Monte Carlo with
flexible simulation runs.
CHAPTER 2
OVERVIEW OF GIRAFFE SIMULATIONS
2.1
INTRODUCTION
The first version of GIRAFFE was designed to work in an MS-DOS environment.
GIRAFFE versions 2 and 3 are equipped with a graphical user interface (GUI) to provide a better
user experience. Version 3 is installed by opening the file, Giraffe_Install.exe, on the installation
disc. The installation procedure will also allow you to install EPANET2.0 and the necessary
Microsoft .NET Framework 1.1 Package. It is recommended that users install the EPANET2.0
software as it provides a GUI to help users visualize the hydraulic network and GIRAFFE
simulation results. EPANET can be downloaded from the installation disk or from the EPANET
website: http://www.epa.gov/ORD/NRMRL/wswrd/epanet.html#Downloads.
GIRAFFE can perform both deterministic and Monte Carlo simulations.
For a
deterministic simulation, GIRAFFE adds damage to the network deterministically and then
performs a hydraulic analysis on the damaged network. For Monte Carlo simulation, users can
either specify the number of Monte Carlo simulation runs or let the code decide the simulation
runs automatically using the built-in self-termination algorithm.
For each Monte Carlo
simulation, GIRAFFE damages the system probabilistically and then analyzes the damaged
network.
A complete GIRAFFE simulation includes five major modules, which are system
definition, seismic damage, earthquake demand simulation, hydraulic network analysis, and
compilation of results. A flow chat of a GIRAFFE simulation is shown in Figure 2.1. The major
functions of each module are introduced in the following sections.
3
Defining a water system
Damaging the water system
Simulating earthquake demand
Performing hydraulic
analysis using EPANET
Modifying the
water system
Yes
Connectivity
error?
No
Assessing EPANET results
Yes
Pi < Plimit exists
for any node i?
No
Compiling final results
Figure 2.1 GIRAFFE Simulation Flow Chart
4
Monte Carlo
Simulation Loop
2.2
SYSTEM DEFINITION
The system definition module defines the hydraulic network being analyzed. It provides
information on the physical and operational properties, topology, and demands of a system.
Users can use the GUI of EPANET for system definition and then export the system definition
file. The hydraulic network model of the water supply system operated by the Los Angeles
Department of Water and Power (LADWP) is developed based on the software, H2ONET
(LADWP, 2002), which uses EPANET as the analysis engine and an AutoCAD platform for
network visualization and the presentation of results. The LADWP hydraulic network model can
be exported from H2ONET into EPANET format, which can then be analyzed by GIRAFFE.
Therefore, when GIRAFFE is used to analyze the LADWP hydraulic network, users do not need
to define the system and only need to export the H2ONET hydraulic model into the EPANET
format. Chapter 7 provides an example on how to export a hydraulic network model developed
using H2ONET into the EPANET format.
2.3
SYSTEM DAMAGE
The damage module adds damage to pipelines. The detailed modeling methodology for
pipe damage is described in Chapter 4. In general, pipe breaks and leaks can be modeled. A
pipe leak can be classified as five different types: annular disengagement, round crack,
longitudinal crack, local loss of pipe wall or local tear of pipe wall. One pipe can have multiple
breaks and leaks. Two simulation options, deterministic and probabilistic, are provided for pipe
damage.
GIRAFFE also incorporates the earthquake performance of tanks by accounting for water
losses with time from damaged pipelines. When considering the tank performance, hydraulic
simulation is divided into different time steps, which are set by users. Within each time step,
GIRAFFE performs a steady state hydraulic simulation for a fixed set of tank levels. From one
time step to the next, the tank levels are updated based on the current tank water levels, tank
outflows, and tank cross-sectional areas.
5
2.4
EARTHQUAKE DEMAND SIMULATION
The earthquake demand simulation module implicitly considers the effects of damage to
small diameter distribution pipelines, which are not included in the hydraulic network model, by
increasing nodal demands. The increase of nodal demands is determined by fragility curves,
which relate demand to pipe repair rate. The fragility curves are developed on the basis of
Monte Carlo simulations of the LADWP distribution networks. The detailed methodology for
earthquake demand simulation is provided in Chapter 5. Because the earthquake demand is
simulated probabilistically by fragility curves, this module only works for probabilistic
simulations.
2.5
HYDRAULIC NETWORK ANALYSIS
This module uses the EPANET hydraulic network engine iteratively to solve the
damaged hydraulic network and eliminate negative pressures. As shown in Figure 2.1, the
damaged system is sent to the EPANET engine for hydraulic network analysis. It is possible that
the damaged system cannot be solved because some elements may not have connectivity with the
main system due to earthquake damage. In this case, the EPANET engine gives error messages,
which tell the user the ID of each element disconnected from the main system. GIRAFFE reads
the error messages and fixes the errors by eliminating the disconnected elements from the
database. GIRAFFE then checks the nodal pressures, and identifies the lowest nodal pressure in
the system. If the lowest pressure is higher than the preset pressure limit, which is zero for
negative pressure elimination, the hydraulic analysis stops. If the lowest pressure is lower than
the pressure limit, the program eliminates the node, the links connected to this node, and the
operational parameters associated with the node and links. After each step of elimination,
GIRAFFE performs a hydraulic network analysis again, and this process continues until there is
no pressure lower than the pressure limit in the system. GIRAFFE requires the user to set the
pressure limit to increase the flexibility of the program. For example, areas with inadequate
pressures for fire fighting can be identified by setting a pressure limit required for fire fighting
purposes.
6
2.6
COMPILATION OF RESULTS
This module compiles the hydraulic analysis results into a format compatible with GIS.
It also provides a performance index to measure the system serviceability.
2.6.1
Hydraulic Network Analysis Results
The H2ONET LADWP hydraulic network model database can be exported as GIS data,
in which junctions, pipes, pumps, valves, and tanks are exported into separate shapefiles. The
hydraulic analysis results are thus compiled for these five types of elements. Please note,
reservoirs are treated as a special type of tank which have a fixed grade. A user may classify a
reservoir as a tank to allow water levels to vary dynamically (See Appendix D for special notes
regarding this case). The major outputs for pipes, valves, and pumps are their respective flow
rates. The major outputs for junctions and tanks are their respective pressures and grades. For
the components that are eliminated from the main system due to either negative pressure or
connectivity problems, their results are set to zero to represent the isolation of these components.
For a deterministic simulation, the outputs for the five types of components are reported. For a
probabilistic simulation, the outputs for the five types of components are reported for each run of
the Monte Carlo simulation. The flow rates in pipes and pressures at junctions, which are the
key outputs, are reported for each Monte Carlo simulation run. The mean, standard deviation,
and coefficient of variation (COV) of the flow rate in each pipe and pressure at each node for all
Monte Carlo simulation runs are also calculated and reported. If users perform time-history
simulation to consider the tank performance, the outputs are reported for each time step.
2.6.2
Performance Index
This module provides an index for measuring the seismic serviceability of a damaged
water supply system. The serviceability is defined as the ratio of the available demand to
required demand corresponding to a seismic damage scenario,
7
Ss =
QT
QT*
(2.1)
where S s is the serviceability, Q T is the available demand, and QT* is the required demand. The
serviceability can be calculated for each demand node and for the entire system.
For a
deterministic simulation, the serviceability for each demand node is either 0, if this demand node
is isolated due to the negative pressure or connectivity problems, or 1, if this demand can be
satisfied. The serviceability for the entire system is the sum of the demands that can be satisfied
over the sum of the total required demands.
For a probabilistic simulation, the system
serviceability is reported in a matrix format.
For each Monte Carlo simulation run, the
serviceability is reported for each demand node and for the entire system. The mean of the nodal
and system serviceability for all Monte Carlo simulation runs is also calculated and reported. If
users perform time-history simulation to consider the tank performance, the outputs are reported
for each time step.
GIRAFFE provides a simulation option, in which the program will determine how many
Monte Carlo simulation runs are needed to have statistically significant results using the system
serviceability as an index. In this simulation option, GIRAFFE calculates the mean and COV of
the system serviceability, starting from ten simulation runs. Then after every five simulation
runs, GIRAFFE calculates the mean and COV of the system serviceability of all the simulations
and compares the current mean and COV of the system serviceability with the previous ones. If
the difference of both mean and COV of the system serviceability from the two sets of results is
smaller than a user defined percentage (the default is set to 2%), the simulation is terminated,
otherwise, the simulation continues.
8
CHAPTER 3
HYDRAULIC NETWORK ANALYSIS
3.1
INTRODUCTION
The basic function of a water supply system is to deliver water from sources to customers.
Moving water from source to customer requires a network of pipes, pumps, valves, and other
appurtenances. Storing water to accommodate fluctuations in demand due to varying rates of
usage or fire protection requires storage facilities, such as tanks and reservoirs. Pipes, pumps,
valves, storage, and the supporting infrastructure together comprise a water supply system. A
hydraulic network model is a mathematical model of a water supply system in which the
physical components of the system are represented as nodes and links. Hydraulic network
analysis utilizes the physical and operational properties, topology, and demands of a water
supply system as basic input data, and calculates pressures at nodes and flows in links.
Hydraulic network analysis can be used to predict pressure and flow conditions in a water supply
system under different operational scenarios to ensure that sound, cost-effective engineering
solutions can be implemented in the design, planning, and functioning of the water supply
system.
This section provides a brief introduction of hydraulic network analysis. The basic
methodology for hydraulic network analysis is introduced. The EPANET hydraulic simulation
models are described. The negative pressure treatment for simulating heavily damaged water
supply systems is discussed.
3.2
COMPONENTS IN HYDRAULIC NETWORKS
In general, a hydraulic network model consists of two basic classes of elements: nodes
and links. The nodes represent facilities at specific locations in a water supply system, and the
links define relationships between nodes. Typical nodal elements include junctions and storage
9
nodes, and typical link elements are pipes. Other components, such as valves and pumps, can be
modeled as either links or nodes, depending on different modeling techniques. The primary
modeling purpose of each physical element is briefly described below.
1. Junctions: represent locations where links intersect and where water enters or leaves the
network.
2. Storage nodes: represent locations of storage reservoirs and tanks. The pressures at
storage nodes are known and treated as boundary conditions to solve flow equations. In
contrast to tanks, which have limited storage capacity and for which the volume of stored
water varies with simulation time, reservoirs represent external water sources with
unlimited storage capacity, such as sources from lakes, rivers, or ground aquifers.
3. Pipes: represent links conveying water from one node to another.
4. Pumps: represent elements adding energy to flowing water in the form of an increased
hydraulic grade. A pump can be modeled as either a node or link.
5. Valves: represent elements controlling water flow or pressure from one node to another.
A valve can be modeled as either a node or link. There are different types of valves with
different functions, such as check, pressure reducing, flow control, throttle control, air
release, and vacuum breaking valves.
These physical components are interconnected to form a network and operate together
under some operational rules. Typical operational rules include the change of the status of pipes,
pumps, and valves under certain conditions. For example, the status of a pump is typically
controlled by the water level of the tank it serves. When water in the tank is lower than a certain
level, the pump is opened to boost water to the tank. When water in the tank is higher than a
certain level, the pump is closed and the tank supplies water to customers. The operational rules
give a water supply system the ability to work efficiently under different operation scenarios.
3.3
GOVERNING LAWS
Hydraulic network analysis assumes that a pipeline network is always full and
pressurized with water, and steady state flow condition is reached for every pipeline.
10
Incompressible flow in a pipeline network is then governed by two principle laws: the laws of
mass and energy conservation.
3.3.1
Equation of Continuity
In hydraulic network analyses, conservation of mass is typically expressed as equation of
continuity, which simply states that the algebraic sum of flows into and out of any node should
be zero (Jeppson, 1976). Consider a node i, for which the continuity equation can be expressed
as
n pi
∑Q
k =1
ik
~
= Qi
(3.1)
~
in which Qi is the external flow at node i, (normally called demand), n pi is the number of pipes
connected to node i, k is an index for pipes, and Qik is the flow rate in pipe k to node i. Typically,
~
Qik is positive for flows coming into the node and negative going out. In contrast, Qi is positive
for flows going out of the node and negative coming into.
3.3.2
Bernoulli Equation
The conservation of energy between two cross-sections, i and j, within a flow is
expressed by the Bernoulli equation (Jeppson, 1976) in the form of hydraulic head as
2
2
pj v j
v
zi +
+ i + hp = z j +
+
+ hf
γ w 2g
γ w 2g
pi
(3.2)
where z is the elevation head, p is the internal pressure measured from atmospheric levels, γ w is
the unit weight of water, p γ w is the pressure head, v is the flow velocity, g is the gravitational
11
acceleration, v 2 2 g is the velocity head, h p is the head gain from external mechanical energy,
such as pumps, and h f is the head losses including frictional and minor losses.
A fundamental aspect of the Bernoulli equation is that there is only one hydraulic head at
each node in a hydraulic network. The algebraic sum of the head losses and gains around any
closed loop should be zero, which is expressed as
nL
∑h
k =1
k
=0
(3.3)
where n L is the number of pipes in the loop and hk is the head gain or loss in pipe k .
3.4
ENERGY LOSSES
Whenever water flow passes a fixed wall or boundary, friction exists due to the viscosity
of water. The friction transforms part of the useful energy into heat or other forms of nonrecoverable energy, which results in frictional head losses. A number of appurtenances, such as
inlets, bends, elbows, contractions, expansions, valves, meters, and pipe fittings are commonly
included in water supply systems. These devices alter the flow pattern in pipes by creating
additional turbulence, which leads to head losses in excess of frictional head losses. These
additional head losses are called minor or local losses.
3.4.1
Frictional Loss
Frictional loss results from the shear stress developed between water and the pipe wall.
Its magnitude depends on the density, viscosity, and velocity of water, as well as the internal
roughness, length, and size of the pipe (Jeppson, 1976). There are various formulations to
evaluate frictional head losses, and all formulations can be generalized into the following form
(Walski, et al., 2001):
12
h fk = K fk Qknk
(3.4)
in which h fk is the frictional head loss along pipe k, Qk is the flow rate through the pipe, K fk is a
resistance coefficient, and nk is a constant flow exponent.
The most widely used formulations to calculate frictional head losses in hydraulic
network analysis are the Darcy-Weisbach, Hazen-Williams, and Chezy-Manning equations. The
resistance coefficient, K fk , and flow exponent, nk, associated with each formulation are listed in
Table 3.1. The Darcy-Weisbach equation is physically-based, as it is derived from the basic
equations of Newton’s Second Law.
The main disadvantage associated with the Darcy-
Weisbach equation is that the frictional factor, f, and thus the resistance coefficient, Kfk, is a
function of flow rate, Qk. When Equation 3.4 is used to solve flow rate, Qk, with known head
loss, h fk , the equation is an implicit expression of the flow rate. Trial-and-error or numerical
methods must be applied to solve it.
The Hazen-Williams and Manning formulas are
empirically-based expressions developed from experimental data. The Hazen-Williams formula
is the most frequently used formulation for hydraulic network analysis in the U.S. Jeppson
(1976) provides a detailed discussion of the three formulas.
3.4.2
Minor Loss
Minor losses (also called local losses) are induced by local turbulence. The importance
of such losses depends on the geometric dimension of the hydraulic network and the required
simulation accuracy. If pipelines are relatively long, these minor losses may be truly minor
compared with frictional losses and can be neglected. In contrast, if pipelines are short, the
minor losses may be large and should be considered. If devices, such as a partly closed valve,
cause large losses, the minor losses can have an important influence on the flow rate. In practice,
some engineering judgment is required to decide if the minor losses need to be considered.
13
Table 3.1 Frictional Head Loss Evaluation Formulas
Equation
Resistance Coefficient K fk
Flow Exponent nk
Darcy-Weisbach
8 fl k
gd k5π 2
2
Hazen-Williams
Bl k
C d k4.87
1.852
Chezy-Manning
Al k µ 2
d k5.333
2
1.852
Notes:
g: Acceleration of gravity
f: Friction factor in Darcy-Weisbach formulation, a function of the flow rate and physical
properties of the pipeline. The friction factor, f, can be determined using the Colebrook-White
equation (Jeppson, 1976), Moody diagram (Moody, 1944), or Swamee-Jian formula (Swamee
and Jian, 1976).
l k : Length of pipe
d k : Diameter of pipe
B: Dimensional constant in Hazen-Williams formulation, equal to 4.73 and 10.70 in British and
SI units, respectively.
C: Hazen-Williams roughness coefficient, a function of the pipe physical properties. The values
of C for different types of pipeline are available in the literature (e.g., Jeppson, 1976;
Armando, 1987; Walski, et al., 2001).
A: Dimensional constant in Chezy-Manning formulation, equal to 4.64 and 10.29 in British and
SI units, respectively.
µ: Manning roughness coefficient, a function of the pipe physical properties. The values of µ
for different types of pipeline are available in the literature (e.g., Jeppson, 1976; Armando,
1987; Walski, et al., 2001).
14
The minor losses are generally expressed as
hmk =
K mk 2
'
Qk = K mk
Qk2
2
2 gAk
(3.5)
'
in which K mk
= K mk (2 gAk2 ) , g is the acceleration of gravity, Qk is the flow rate, K mk is the
minor loss coefficient, and Ak is the pipe cross-sectional area. The values of K mk for different
types of minor losses have been determined from experiments, and are available in the literature
(e.g., Crane Company, 1972; Miller, 1978; Armando, 1987; Idelchik, 1999; Waskli, et al., 2001).
Sometimes, it is more convenient to equate the minor losses to frictional losses caused by a
fictitious length of pipe, known as an equivalent pipe length. This length can be derived from
Equations 3.4 and 3.5, with the substitution of the selected resistance coefficient K fk and flow
exponent nk.
3.5
ENERGY GAINS
There are many occasions when energy needs to be added into a hydraulic system to
overcome elevation difference, as well as frictional and minor losses. A pump is a device to
which mechanical energy is applied and transferred to water as hydraulic head. The head added
to water is called pump head, and is a function of discharge through the pump. The relationship
between pump head and discharge rate is called a pump head characteristic curve, as shown in
Figure 3.1. The pump characteristic curve is nonlinear, and as expected, the more water that
passes through the pump, the less head it can add.
The head that is plotted in the head characteristic curve is the head difference across the
pump, called the total dynamic head. This curve needs to be described as a mathematical
equation to be used in hydraulic simulation. Some models fit a polynomial curve to selected data
points, but a more common approach is to describe the curve using a power function in the form
of
hP = ho − cQPm
15
(3.6)
70
60
Shutoff Head
Design Point
Head (m)
50
40
30
Maximum Flow
20
10
0
0
20
40
60
80
100
120
Flow (L/s)
Figure 3.1 General Shape of Pump Characteristic Curve
where hP is the pump head, ho is the cutoff (shutoff) head (pump head at zero flow), QP is the
pump discharge, and c and m are the coefficients describing the curve shape.
The purpose of a pump is to overcome elevation differences and head losses due to pipe
friction and obstructed flow at fittings. The amount of head which a pump must add to overcome
elevation differences is referred to as static head or static lift and is dependent on system
topology but independent of the pump discharge. Frictional and minor losses, however, are
highly dependent on the pump discharge rate. When these losses are added to the static head for
a series of discharge rates, the resulting plot is called a system head curve.
The pump
characteristic curve is a function of the pump and independent of the system, while the system
head curve is dependent on the system and independent of the pump.
When a pump
characteristic curve and a system head curve are plotted on the same axes, there is only one point
that lies on both of them. This intersection, as shown in Figure 3.2, defines the pump operation
point, which represents the discharge that passes through the pump and the head that the pump
adds in hydraulic network simulations.
16
70
60
Pump Head Cur
ve
Head (m)
50
40
Pump Operation Point
30
20
He
S stem
10
rve
ad Cu
Head Loss
Static Lift
0
0
30
60
90
120
Flow (L/s)
Figure 3.2 Pump Operation Point
3.6
FLOW EQUATIONS
Hydraulic network analysis is governed by the laws of mass (continuity equation) and
energy (Bernoulli equation) conservation. The major unknowns that need to be determined are
flows in links and hydraulic head at nodes. The flows and hydraulic heads are linked with each
other by the head loss equations, Equations 3.4 and 3.5. Based on different primary unknowns
used in the equations, four types of flow equations can be developed, which are Q-, H-, ∆Q - and
hybrid equations (Jeppson, 1976), to express the laws of mass and energy conservation. The four
types of equations set flow rates in links, hydraulic heads at nodes, corrective flows in network
loops, and mixture of flow rates in links and hydraulic heads at nodes as primary unknowns,
respectively. Shi (2006) provides a detailed description of the four types of flow equations. The
solution to these equations requires solving a set of linear and/or nonlinear equations. For
networks with a large number of components, numerical methods must be used. Four widely
used numerical methods are Hardy-Cross, Newton-Rapshon, linear theory, and the gradient
method. References on the detailed procedures of applying the four numerical methods to flow
equations are included in Shi (2006).
17
3.7
EPANET
Many commercial software packages are available in the market for hydraulic network
analysis.
Among them, EPANET, developed and distributed by the US Environmental
Protection Agency (EPA) (Rossman, 2000), is one of the earliest and most widely used. Because
EPANET contains a state-of-the-art hydraulic analysis engine and its source code is freely
available to the public, a family of software packages including WaterCAD (WaterCAD, 2005),
MIKENET (MIKENET, 2005), H2ONET (MWH Soft Inc., 1999), and others, use the EPANET
analysis engine and develop their own products around it.
EPANET was designed to be a research tool for improving the understanding of the
movement and fate of drinking water constituents in water distribution systems (Rossman, 2000).
It has two major capabilities: hydraulic and quality modeling for water in a pressurized pipeline
network. The water quality modeling is beyond the scope of this study, and therefore, only the
hydraulic modeling capabilities of EPANET are discussed. The following discussion is based on,
but not limited to, the information provided in the EPANET user manual by Rossman (2000).
EPA released a DOS and Windows version of EPANET. The DOS version is an
analysis engine that is coded in the C language. The Windows version includes the analysis
engine with a GUI written with the Daphi language. To run EPANET in the DOS environment,
all network input data are stored in an input text file and analysis results are written into an
output text file. To run EPANET in the Windows environment, users can use the GUI to
construct a hydraulic network model and input network attributes graphically.
The GUI
compiles the input information into a text file, and calls on the engine to do the analysis. After
finishing the analysis, the GUI retrieves data from the text output file generated by the engine
and displays the results graphically for visualization. The source codes and executable files of
both the analysis engine and GUI are available from the Internet free of charge. Thus, users can
use EPANET to perform hydraulic network analyses, and can also modify the source codes for
their own product development.
18
3.7.1
EPANET Hydraulic Network Components
An EPANET hydraulic network model consists of various physical components, which
are the mathematical representations of physical objects in a real water supply system.
Mathematical representations are also used for operational components that control the behavior
and operational properties of the physical components.
3.7.1.1
Physical Components
EPANET models a water supply system as a collection of links connected to nodes. The
nodes represent junctions and storage nodes, including tanks and reservoirs. The links represent
pipes, pumps, and control valves. Figure 3.3 illustrates how these objects can be connected to
one another to form a network. Each reservoir, tank, pump, and valve, because of its different
physical properties and/or functions, can have different modeling options. Table 3.2 lists all the
physical components that EPANET can model. In total, there are 17 different components,
including 1 junction, 4 storage nodes, 1 pipe, 4 pumps, and 7 valves. Table 3.2 provides a brief
description of the functions and basic input and output parameters associated with hydraulic
simulations of each physical component.
3.7.1.2
Operational Components
In addition to the physical components, EPANET employs three types of operational
components: curves, patterns, and controls that describe the operational aspects of the physical
components.
Curves
Curves are objects that contain data pairs representing a relationship between two
quantities. An EPANET model can utilize four types of curves, which are pump characteristic,
efficiency, volume, and head loss curves.
A pump characteristic curve represents the
relationship between the head and flow rate that a pump can deliver. EPANET can model three
19
Tank
Demand
Reservoir
Junction
Pipe
Junction Pipe
Junction
Pipe
Demand
Junction
Junction
Pipe
Demand
Valve
Demand
Demand
Pipe
Junction Pipe
Pipe
Pump
Demand
Demand
Demand
Junction
Pipe
Junction
Figure 3.3 Physical Components in an EPANET Hydraulic Network
different shapes of pump curves: single-, three-, and multi-point curves, dependent on the
number of points used to calibrate the pump characteristics. An efficiency curve describes pump
efficiency as a function of pump flow rate and is used for determining energy consumption and
calculating costs associated with pump operations. These calculations are not considered in this
study. A volume curve describes how storage tank volume varies as a function of water level. It
is used when it is necessary to accurately represent tanks, for which the cross-sectional area
varies with water height. A head loss curve is used to describe the head loss through a general
purpose valve as a function of flow rate. It provides the capability to model devices and
situations with unique head loss-flow relationships, such as reduced flow-backflow prevention
valves, turbines, and well draw-down behavior.
Time Patterns
A time pattern is a collection of multipliers that can be applied to a quantity to allow it to
vary over simulation time. Nodal demands, reservoir heads, and pump schedules can all have
time patterns associated with them. When applying time pattern to a quantity, the hydraulic
simulation time is divided into different time intervals, which are set by users. Within each time
20
Table 3.2 Summary Table for Physical Components in an EPANET Hydraulic Network Model
Components
Descriptions
Inputs
Junction
Points where links join together and where Coordinates; elevation;
water enters or leaves the network
demand
Constant
Level
Unlimited capacity water sources with Coordinates; hydraulic head ( a
constant value)
constant water level during simulation time
Variable
Level
Unlimited capacity water sources with water Coordinates; hydraulic head curve
level varying with simulation time
(hydraulic head vs. time)
Cylindrical
Limited capacity
cylindrical shape
Variable
Area
Limited capacity water sources with variable Coordinates; volume vs. hydraulic
cross-sectional area
grade curve
water
sources
Tank
Pipe
Constant Power
Pump
Hydraulic head;
pressure
Hydraulic head
Reservoir
Storage
Node
Outputs
One-Point
Three-Point
Multiple-Point
with
Coordinates; bottom elevation;
diameter; initial, minimum, and
maximum water level
Start and end node; diameter;
Links conveying water from one node in the length; roughness and minor loss
network to another
coefficients; status (open, closed,
or containing check valve)
Pumps which a supply constant amount of Start and end node; diameter;
energy to water
energy; status (open or closed)
Start and end node; diameter;
Pumps with characteristic curves defined by
operation flow and head gain;
one point
status (open or closed)
Pumps with characteristic curves defined by Start and end node; diameter; pump
three points
curve; status (open or closed)
Pumps with characteristic curves defined by Start and end node; diameter; pump
multiple points
curve; status (open or closed)
21
Hydraulic head
Hydraulic head
Flow rate;
head loss
Flow rate;
head gain
Component
Check (CVs)
Pressure Reducing
Valves (PRVs)
Pressure Sustaining
Valves (PSVs)
Valve
Pressure Breaker
Valves (PBVs)
Flow Control
Valves (FCVs)
Throttle Control
Valves (TCVs)
General Purpose
Valves (GPVs)
Table 3.2 (Continued)
Description
Allow water through one direction (built in pipe)
PRVs limit the pressure on their downstream end
to not exceed a pre-set value when the upstream
pressure is above the setting. If the upstream
pressure is below the setting, then flow through
the valve is unrestricted. If the downstream
pressure exceeds the upstream pressure, the
valve closes to prevent reverse flow.
PSVs attempt to maintain a minimum pressure
on their upstream end when the downstream
pressure is below the setting. If the downstream
pressure is above the setting, then flow through
the valve is unrestricted. If the downstream
pressure exceeds the upstream pressure, the
valve closes to prevent reverse flow.
PBVs force a specified pressure loss to occur
across the valve. Flow through the valve can be
in either direction.
FCVs limit the flow to a specified amount.
TCVs simulate a partially closed valve by
adjusting the minor head loss coefficient of the
valve.
GPVs are used to represent a link where the user
supplies a special flow-head loss relationship
instead of following one of the standard
hydraulic formulas.
22
Input
Output
None (presence is indicated by a
“CV” at the end of a pipe
definition line)
Start and end node; diameter;
minor loss coefficient;
downstream pressure setting;
status (open or closed)
Start and end node; diameter;
minor loss coefficient;
downstream pressure setting;
status (open or closed)
Flow rate;
head loss
Start and end node; diameter;
minor loss coefficient; pressure
setting; status (open or closed)
Start and end node; diameter;
minor loss coefficient; flow
setting; status (open or closed)
Start and end node; diameter;
minor loss coefficient; status (open
or closed)
Start and end node; diameter; head
loss vs. flow rate curve; status
(open or closed)
interval the quantity remains at a constant level, equal to the product of its nominal value and the
pattern multiplier for that time period.
Controls
Controls are statements that determine how the network is operated over time. They
specify the status of selected links as a function of time, tank water levels, and pressures at select
junctions within the network. There are two types of controls in EPANET hydraulic network
simulations: simple and rule-based. Simple controls change the status or setting of a link based
on one control condition, such as water level in a tank, pressure at a junction, time into the
simulation, or the time of day. Rule-based controls change the link status or settings based on a
combination of conditions that might exist in the network.
3.7.2
EPANET Input File
EPANET stores all input data in a text file with the file extension, .inp. The inp file is
organized into sections with each section beginning with a key word enclosed in brackets. The
various sections are listed in Table 3.3. Detailed examples of the input file can be found in
Chapter 7. In general these sections can be classified into five categories; Network Components,
System Operation, Water Quality, Options and Reporting, and Network Map/Tags.
The Network Components category stores information about the hydraulic properties of
network physical components including junctions, reservoirs, tanks, pipes, pumps, and valves.
The System Operation category stores information of system operational properties such as
curves, patterns, initial status, controls, rules, and demand. The Water Quality category stores
information for water quality simulation.
The Options and Reporting category stores
information of simulation and report options, and times for extended period simulation. The
Network Map/Tags category stores information on the coordinates of each node and coordinates
of each vertex of links.
23
Table 3.3 Sections in an EPANET Input File
Network
Components
System
Operation
Water
Quality
Options and
Reporting
Network
Map/Tags
[TITLE]
[DEMANDS]
[SOURCES]
[ENERGY]
[COORDINATES]
[JUNCTIONS]
[CURVES]
[QUALITY]
[OPTIONS]
[VERTICES]
[RESERVOIRS]
[PATTERNS]
[REACTIONS]
[TIMES]
[END]
[TANKS]
[ENERGY]
[PIPES]
[STATUS]
[PUMPS]
[CONTROLS]
[REPORT]
[VALVES]
Users can use the GUI provided by EPANET to construct a hydraulic model and export
the inp file. Because EPANET is one of the most widely used hydraulic software programs,
most of the commercial hydraulic network analysis software packages can export EPANET input
files for data exchange. For example, a network model constructed by H2ONET can be directly
exported with the EPANET input file format and analyzed by the EPANET engine. Furthermore,
because the EPANET input file is well organized with different sections, portions can be easily
modified via a text editor.
3.7.3
EPANET Hydraulic Simulation Methodology
The EPANET hydraulic engine can perform either steady state or extended period
simulation. During a steady state simulation, EPANET computes junction heads and link flows
for a fixed set of reservoir levels, tank levels, and water demands at a fixed point of time. For
extended period simulation, EPANET computes junction heads and link flows for a fixed set of
reservoir levels, tank levels, and water demands over a succession of points in time. From one
time step to the next, reservoir levels and junction demands are updated according to their
prescribed time patterns while tank levels are updated using the current flow solution. The
solution for head and flow at a particular time involves simultaneously solving a set of hybrid
equations using the gradient method (Todini and Pilati, 1987).
24
3.7.4
EPANET Output File
The outputs from the EPANET engine are generated in a text file with the extension of
file name, .rpt. An output file can contain four sections: Status, Energy, Nodes, and Links.
Users can apply the control parameters in the input file to specify the interested sections and the
quantities associated with each section to be reported.
The Status section lists the initial status of all reservoirs, tanks, pumps, valves, and pipes,
as well as any changes in the status of these components as they occur over time in an extended
period simulation. The Energy section lists the energy consumption and cost for the operation of
each pump in the network.
The Nodes section lists simulation results for nodes with the
quantities specified by the user. The default quantities reported for each node include demand,
hydraulic head, and pressure. Results are listed for each reporting time step of an extended
period simulation. The Links section lists simulation results for links with quantities specified
by the user. The default quantities reported for each link include flow, velocity, and head loss.
Diameter, length, water quality, status, setting, reaction rate, and friction factor can also be
reported if required by the user.
3.8
NEGATIVE PRESSURE TREATMENT
Hydraulic network analysis solves for incompressible water flow in a pressurized
pipeline network based on two principle laws: the laws of mass and energy conservation. The
law of mass conservation can be expressed as the equation of continuity, which assumes that all
demands in a system must be satisfied. The law of energy conservation indicates that water can
only flow from nodes with high energy to nodes with low energy. The energy of water is
expressed as hydraulic head, which is the summation of elevation and pressure heads. Hydraulic
head neglects velocity, which is typically small and does not contribute significantly to the
energy balance. The conventional hydraulic network analysis algorithm does not differentiate
positive and negative pressures, and only uses the total head difference to drive water flow to
satisfy demands. The forced satisfaction of all demands, with no differentiation of positive and
negative pressures, may lead to the prediction of unrealistically high negative pressures at some
25
nodes. This outcome is especially true in an earthquake-damaged system, in which demands due
to water losses from pipeline breaks and leaks may be much higher than the supply from
reservoir and transmission pipeline sources.
To account more accurately for flows and pressures, hydraulic network analysis in a
damaged system should be based on the assumption that a water distribution network is not air
tight when internal pressures fall below atmospheric levels (Markov, et al., 1994). Consider node
i, shown in Figure 3.4, of a water supply system with pressure pi < 0, where zero stands for the
atmospheric pressure. The hydraulic head at node i is Hi = HiE + pi/γw, in which HiE is the
elevation head and γw is the unit weight of water. Since the physical system is not air tight, air
enters it through node i, causing the pressure at node i to become equal to the atmospheric
pressure so that pi = 0 and Hi = HiE. Let Qk be the flow in pipe k connected with nodes i and j.
Qk will be zero if the hydraulic head at node i is higher than that at node j (HiE = Hi > Hj ). If this
is the case for all pipes connected with node i, the node is considered as a no-flow node through
which no water can pass. If there are pipes where this condition is not satisfied, the node is
considered as a partial flow node, through which water can pass with reduced flow rates
compared with those predicted by conventional hydraulic network analysis with negative
pressures. By admitting air into the system, flow conditions around the partial flow node become
complicated. They may involve pressurized flow, transition from pressurized flow to openchannel flow, and open-channel flow (Shi, 2006). Open-channel flow is characterized by the
existence of a free water surface in the flow profile and is more difficult to solve than pressurized
flow. Currently, commercial software packages are not configured to solve the flow conditions
around partial flow nodes.
In GIRAFFE, an isolation approach is applied to treat the negative pressures. This
isolation approach works with EPANET hydraulic network engine iteratively. After hydraulic
network analysis of the damaged system using the EPANET engine, nodes with negative
pressures are identified and isolated step by step, starting with the node of highest negative
pressure. The isolation is simulated by eliminating the node, all connected links, and control
parameters associated with the node and links from the *.inp system definition file. After each
26
elimination, network connectivity is checked. If part of the system is isolated from the main
system without
j=1
k
i
j=2
j = ···
j=n
Figure 3.4 Negative Pressure Node Demonstration (after Markov, et al., 1994)
water sources, it is taken out of the system. The flow analysis and the elimination process
continue until no negative pressure nodes exist in the system.
By discounting water conveyance through partial-flow nodes, the approach adopted in
GIRAFFE removes flow under atmospheric conditions as well as transitional pressures
approaching atmospheric. Such flow will generally occur at relatively low rates and is not
reliable for fire protection after an earthquake.
Hence, the model eliminates piping with
uncertain and/or unreliable flows, thus concentrating on those parts of the system that can be
effective during emergency response.
The modeling approach adopted in GIRAFFE, in effect, expresses a damage state as an
operational state by converting the damaged network into one that meets the requirements of
positive pressure and flow in all pipes. By eliminating pipelines with unreliable flow, it has the
practical advantage of showing the system operator what parts of the network are no longer
functional, and thus provides information about the most vulnerable distribution sectors and
potential strategies for mitigation. The model does not account explicitly for water delivery and
pressure losses associated with unsteady flow because accurate network analyses for this
condition are not available. Instead, the model removes the unreliable portions of the system to
27
display the remaining part of the network that meets threshold serviceability requirements for
positive pressure.
CHAPTER 4
PIPE DAMAGE MODELING
4.1
INTRODUCTION
To predict the flow and pressure conditions in a damaged water supply system using
hydraulic network analysis, pipeline damage, including leaks and breaks, needs to be added into
the network, followed by the hydraulic simulation of the damaged system. GIRAFFE provides
comprehensive methods for pipe damage modeling. This chapter presents the methodology for
pipe damage simulation used in GIRAFFE. It begins with the definition of pipe leaks and breaks.
The hydraulic models of leaks and breaks are discussed with special attention given to leak
simulation.
A classification for leak types is proposed and mathematical formulations are
developed to determine the opening area of each leak type. Finally, the implementation of the
pipeline break and leak models in association with Monte Carlo simulation is described.
4.2
DEFINITIONS
Following the seismic guidelines for water pipelines by the American Lifelines Alliance
(2005), “a break is defined as the complete separation of a pipeline, such that no flow will pass
between the two adjacent sections of the broken pipe; and a leak is defined as a small leak in a
pipeline, such that water will continue to flow through the pipeline, albeit at some loss of
pressure and flow rate being delivered, with some flow being lost through the leak”. Leaks can
include pin holes in pipe barrels, very minor joint separations on segmented pipelines, and very
small splits in large diameter steel transmission pipelines. A pipe with a break loses its water
transportation function totally, and a pipe with a leak loses its function partially.
28
4.3
PIPE LEAK SIMULATION
This section provides the methodology of leak simulation used in GIRAFFE.
The
hydraulic model of a leak is discussed. Leaks are classified into five different types and the leak
area is simulated as a function of pipe diameter.
4.3.1
Hydraulic Model
A pipe leak is essentially an orifice in the pipe wall or at a pipe joint, which allows water
to be discharged into the surrounding soil. In GIRAFFE, a pipe leak is simulated as a fictitious
pipe with one end connected to the leaking pipe and the other end open to the atmosphere,
simulated as an empty reservoir. A check valve is built into the fictitious pipe, only allowing
water to flow from the leaking pipe to the reservoir but not in the reverse direction. The
roughness and minor loss coefficients of the fictitious pipe are taken as infinite and 1,
respectively, such that all energy loss from the leak is related to the minor loss. The minor loss
results from flow turbulence created by the sudden expansion of water passing through the flow
area of the orifice to an infinite area external to the pipe (Jeppson, 1976). The diameter of the
fictitious pipe is determined by the leak area.
Based on this hydraulic model, water loss from a leak can be calculated as
Q = [2g/(Kγw)] 0.5Ap0.5 = (2g/γw)0.5Ap0.5 = CD p0.5
(4.1)
in which Q is the flow rate, g is the gravitational acceleration, γw is the unit weight of water, K is
the minor loss coefficient equal to 1, A is the orifice area, p is the pipe internal pressure, and CD
is the discharge coefficient equal to (2g/γw)0.5A. The pipe leak can be considered as analogous to
a sprinkler used in fire protection, from which water flow is governed by the same flow equation
as Equation 4.1 (Puchovsky, 1999). To validate the model, a set of sprinkler data with discharge
coefficient, CD, and orifice area, A, are used to test the theoretical relationship between CD and A
29
from Equation 4.1. The comparison in Figure 4.1 shows that the theoretical predictions and real
data follow closely spaced, parallel trends. The CD of the real sprinklers is roughly 10% lower
0
0.1
0.2
0.3
0.4
(in2)
16
Discharge Coefficient, CD, [(L/s)kPa0.5]
Sprinkler Data from
Puchovsky (1999)
Theoretical Prediction (K=1)
Linear Regression of
Sprinkler Data Points
0.3
12
y =0.0014 x
0.2
8
y = 0.0013x - 0.0151
R2 = 0.98
0.1
4
0.0
0
50
100
150
200
250
0.5
Discharge Coefficient, CD, [(gpm)/(psi) ]
0.4
0
2
300 (mm )
Orifice Area, A
Figure 4.1 Comparison Between Model Predictions and Sprinkler Data
than the theoretical CD. The small difference results from the frictional loss of real sprinklers
that have a short length; while the model ignores all frictional loss and leads to more water loss.
Figure 4.2 shows the implementation of the pipe leak model in GIRAFFE. It is assumed
that a leak occurs in the pipe ab, which is connected to the upstream node A and downstream
node B. The length of pipe ab is L and the leak occurs at a distance λL, measured from the
upstream node A along the longitudinal direction of pipe ab, in which λ is a constant, called the
length ratio in this study. GIRAFFE simulates the leak by: 1) eliminating pipe ab from the
network; 2) adding a new junction, A1Jab, at the leak location, of which the elevation is
determined by the linear interpolation of the elevations of nodes A and B; 3) adding two pipes,
A1Oab and A2Oab, which have the same diameter and roughness as pipe ab, to replace the
original pipe ab. Pipe A1Oab is connected to node A and junction A1Jab, and pipe A2Oab is
connected to junction A1Jab and node B; 3) adding an empty reservoir A1Rab, which has the
same elevation as the newly added junction A1Jab; and 4) adding a pipe, A1Lab, to connect the
30
newly added junction, A1Jab, and reservoir, A1Rab. The length of pipe A1Lab is set to 0.5 feet,
roughness is 1,000,000, and minor loss coefficient is 1, such that all energy loss from pipe A1Lab
A1Rab
A
A1Lab
B
λL
A
(1-λ)L
Pipe ab
Length L
Pipe leaking
Empty reservoir
Check valve
A1Jab
A1Oab
A2Oab
λL
(1-λ)L
B
Hydraulic model
Figure 4.2 Hydraulic Model for Leak Simulation
is accounted for as minor loss. A check valve is built in pipe A1Lab such that water can only
flow from the leaking pipe to the empty reservoir.
In general, to simulate a pipe leak, GIRAFFE deletes one pipe and adds three new pipes,
one junction, and one empty reservoir. To ensure that each new element has a unique ID, all
new elements are assigned to an ID starting with a letter A and ending with the ID of original
pipe. The third character in the ID of the new elements is either an O indicating this pipe is used
to replace the original pipe, R indicating the element is a newly added reservoir, J indicating the
element is a newly added junction, or L indicating the element is a newly added pipe to model
leak. The second character is a number to indicate the order of this type of elements. For
example, the number 1 in A1Oab indicates that this pipe is the first section from the upstream
node of the original pipe and number 2 in A2Oab indicates that this pipe is the second section of
the original pipe.
4.3.2
Leak Classification
Using the leak simulation model, a key input parameter is the orifice area, which depends
on pipe material and joint properties, as well as seismic loading characteristics. To develop a
rational basis for estimating the orifice area, a detailed study has been performed by Shi (2006)
on the material properties, joint characteristics, and damage mechanisms of five of the most
commonly used types of pipelines in North America, including cast iron, ductile iron, concrete,
31
steel with riveted joints, and steel with welded slip joints. Based on this study, leaks are
classified into five different types A described in the following sections.
4.3.2.1
Annular Disengagement
Annular disengagement refers to joint looseness of segmented pipelines resulting from
joint axial pullout movement during seismic loading.
A schematic drawing of annular
disengagement is shown in Figure 4.3. This leak type may occur in cast iron, ductile iron,
jointed concrete cylinder, and riveted steel pipelines. The opening from annular disengagement
occurs in the circumferential direction, and its area is determined by the joint configuration,
relative pullout movement, and condition of the gasket seal or caulking material.
To estimate the opening area of an annular disengagement, the opening area, called
equivalent orifice area (EOA) in this study, is correlated to an area index and the maximum
possible annular area, and calculated as
A = k × Amax
(4.2)
where A is the EOA, Amax is the maximum annular area, and k is a constant. The Amax is
determined by the configuration of the joint and can be estimated as
Amax ≈ tDπ
(4.3)
where D is the pipeline diameter and t is the thickness of maximum possible annular space.
Substituting Equation 4.3 into 4.2 results in
A = k × Amax = tkDπ
32
(4.4)
Since a leak is modeled as a fictitious pipeline in hydraulic network analyses, the orifice
needs to be converted into a pipe with a cross-sectional area equal to the EOA. The diameter of
the fictitious pipe, called equivalent orifice diameter (EOD) in this study, can be calculated as
d = 4 A / π = 2 tkD
(4.5)
Gasket Lost
Compressive Seal
D
t
Figure 4.3 Schematic Drawing of Annular Disengagement
In GIRAFFE, the maximum possible annular space, t, is taken as 10 mm (0.4 in.) based
on the studies conducted by Shi (2006) on the configurations of joints for the four types of
pipelines in which annular disengagement may occur. As for the ratio, k, of the actual leak area
to the maximum possible leak area, a default value of 0.3 is proposed on the basis of field
observations (O’Rourke, 2005) from previous earthquakes. Users may change the default values
for t and k through the Options menu within GIRAFFE (Click on Options | Configuration | Pipe
Leakage Model).
4.3.2.2
Round Crack
The second leak type is a round crack, which refers to the circumferential cracking of the
pipe barrel or joint under the effects of bending or the combination of bending and tensile forces.
A schematic drawing of a round crack is shown in Figure 4.4. Round cracks occur in pipes
33
composed of brittle material and joints, such as cast iron pipes with lead caulked joints. The
EOA is determined by the opening angle of the crack and pipe diameter, and can be calculated as
A = 0.5πD × (θD) = 0.5πθD 2
(4.6)
where θ is the open angle of the crack and D is the pipe diameter.
θ
D
Figure 4.4 Schematic Drawing of Round Crack
The EOD of a round crack can be calculated as
d = 4 A / π = 4(0.5πθD 2 ) / π = 2θ D
(4.7)
Based on field observations (O’Rourke, 2005), a default value of 0.5o is proposed for the
opening angle in GIRAFFE. Users may change the default value for the opening angle, θ ,
through the Options menu within GIRAFFE (Click on Options | Configuration | Pipe Leakage
Model).
4.3.2.3
Longitudinal Crack
The third leak type is a longitudinal crack, which refers to the cracking of the pipe barrel
or seam along the length of the pipe (longitudinal direction) caused by the external loading
and/or high internal pressures during earthquakes. A schematic drawing of a longitudinal crack
34
is shown in Figure 4.5. The longitudinal cracking may occur in metal pipes, which include cast
iron, ductile iron, and riveted steel pipes.
The EOA of a longitudinal crack can be calculated as
A = L ×W
(4.8)
where L and W are the length and width of the crack, respectively. The length, L, is in the pipe
longitudinal direction and can be taken as the length of a pipe section. The width, W, is in the
pipe circumferential direction and can be estimated as a function of the opening angle, θ , of
the crack and pipe diameter, D. The width, W, can be calculated as
A
W
W
W
θ
LL
A’
D
A – A’ Section
Figure 4.5 Schematic Drawing of Longitudinal Crack
W = Dθ
(4.9)
Substituting Equation 4.9 into Equation 4.8 results in
A = W × L = LDθ
(4.10)
The EOD of a longitudinal crack can be calculated as
d = 4 A / π = 2 LDθ / π
35
(4.11)
The default value for length of the longitudinal crack is taken as thirteen meters (40 ft), which
provides a reasonable, but conservative estimate of the length of metal pipe sections. The
opening angle of the longitudinal crack is estimated as 0.1o from field observations (O’Rourke,
2005). Users may change the default values for the opening angle, θ , and longitudinal crack
length, L, through the Options menu within GIRAFFE (Click on Options | Configuration | Pipe
Leakage Model).
4.3.2.4
Local Loss of Pipe Wall
The fourth leak type is the local loss of pipe wall. This leak type is caused by the loss of
a small portion of pipe wall, which is deteriorated by corrosion, under the earthquake loading
effects. A schematic drawing of a local loss of pipe wall is shown in Figure 4.6. The EOA of a
local loss of pipe wall can be calculated as
L
W
D
Figure 4.6 Schematic Drawing of Local Loss of Pipe Wall
A = L ×W
(4.12)
where L and W are the length and width of the orifice. The length, L, is along the pipe
longitudinal direction and can be estimated as a ratio, k1, of pipe diameter as
L = k1 × D
(4.13)
The width, W, is along the pipe circumferential direction and can be estimated as a ratio, k2, of
the pipe circumferential length to yield
36
W = k2πD
(4.14)
Substituting Equations 4.14 and 4.13 into 4.12 results in
A = πk1k2 D 2
(4.15)
The EOD of a local loss of pipe wall can be calculated as
d = 4 A / π = 4(πk1 k 2 D 2 ) / π = 2 k1 k 2 D
(4.16)
The loss of pipe wall due to corrosion is usually small. Five percent is proposed as a rough
estimate of the parameters, k1 and k2, in GIRAFFE. Users may change the default value for k1
and k2 through the Options menu within GIRAFFE (Click on Options | Configuration | Pipe
Leakage Model). However, it is always assumed that k1 = k2.
4.3.2.5
Local Tear of Pipe Wall
The fifth leak type is local tear of a pipe wall, which typically occurs as a rupture in the
bell casing of a wrinkled welded slip joint and is induced by compressive forces. A schematic
drawing of a local tear of a steel pipeline with welded slip joint is shown in Figure 4.7.
The EOA of a local tear of pipe wall can be calculated as
A = L ×W
(4.17)
in which, L and W are the length and width of the split, respectively. The length, L, is along the
pipe circumferential direction and can be estimated with a ratio, k, of the pipe circumferential
length,
L = kπD
37
(4.18)
Substituting Equation 4.18 into Equation 4.17 results in
A = W × L = kπDW
(4.19)
The EOD of a local tear of pipe wall can be calculated as
d = 4 A / π = 4(kπD * W ) / π
= 2 kWD
(4.20)
In GIRAFFE, the default value for length of a local tear is taken as 30% of the pipe
circumferential length, and the width is assumed to be 12 mm (0.5 in.) based on the data from
Northridge earthquake (Shi, 2006). Users may change the default values for k and W through the
Options menu within GIRAFFE (Options | Configuration | Pipe Leakage Model).
W
L
D
Figure 4.7 Schematic Drawing of Local Tear of Pipe Wall
4.3.3
Probability of Leak Types
Since each type of pipeline can have multiple types of leaks, the relative likelihood of
each leak type has to be estimated for each type of pipeline to model pipe leaks using Monte
Carlo simulation. Based on pipeline material and joint properties, as well as limited field data, a
probability table shown in Table 4.1 is proposed for the five leak types associated with various
38
types of pipelines. It should be noted that the default probabilities associated with the leak types
under Options | Configuration | Pipe Leak Model do not match the values listed in Table 4.1 due
to the way Monte Carlo simulation calculations are performed in the GIRAFFE code. Users
interested in understanding this process should refer to Chapter 6, Section 2.3 in Wang (2006).
It should be noted that the only leak type for welded steel pipelines is the local tear of
pipe wall resulting from compressive buckling. The majority of locations of local buckling,
although they need to be repaired after earthquakes, are not severe enough to tear the pipe wall
and cause leakage. A conservative estimate adopted in this work is that 80% of repairs from
local buckling would not cause leakage, and 20% of repairs would cause leakage. Therefore, in
GIRAFFE, the repair rate is discounted to 20% when using the repair rate to estimate the number
of leaks for steel pipeline performance simulation after earthquakes.
Table 4.1 Probability of Leak Types for Different Pipelines
Cast Iron
Type 1
Annular
Disengagement
0.3
Type 2
Round
Crack
0.5
Type 3
Longitudinal
Crack
0.1
Type 4
Local Loss
of Pipe Wall
0.1
Type 5
Local Tear of
Pipe Wall
N/A1
Ductile Iron
0.8
N/A1
0.1
0.1
N/A1
Riveted Steel
0.6
N/A1
0.3
0.1
N/A1
Welded Steel
N/A1
N/A1
N/A1
N/A1
1.0
Jointed
Concrete
1.0
N/A1
N/A1
N/A1
N/A1
Pipe Material
1: Not Applicable
4.4
PIPE BREAK SIMULATION
Following the definition of pipe breaks used in this study, a break is a complete
disconnection of the original pipeline. Water can flow from the two broken ends into the
surrounding soil. Figure 4.8 shows the hydraulic model of a pipe break in GIRAFFE. It is
assumed that a break occurs in the pipe ab, which is connected to the upstream node A and
downstream node B. The length of pipe ab is L and the break occurs at a distance λL measured
39
from the upstream node A along the pipe ab. GIRAFFE simulates the break by: 1) eliminating
pipe ab from the network; 2) adding two new empty reservoirs, A1Rab and A2Rab, of which the
elevation is determined by the linear interpolation of the elevations of nodes A and B; and 3)
adding two pipes, A1Oab and A2Oab, which have the same diameter and roughness as pipe ab.
Pipe A1Oab is connected to node A and junction A1Rab, and pipe A2Oab is connected to node B
and junction A1Rab. A minor loss coefficient of 1 and a check valve are added to each of pipes
A1Oab and A2Oab to represent the energy loss and to prevent water from flowing back into the
broken pipeline. In general, to simulate a pipe break, GIRAFFE deletes one pipe and adds two
new pipes and two empty reservoirs. The rules to assign IDs to the new elements are the same as
those used in leak simulations.
4.5
IMPLEMENTATION OF PIPE DAMAGE MODELS
To simulate the earthquake performance of a water supply system, pipe damage
including breaks and leaks needs to be added into the network. Hydraulic simulation is then
performed on the damaged network to predict the flow and pressure distributions. The pipeline
break and leak models can be implemented into a hydraulic network model both
deterministically and probabilistically.
B
A
λL
(1-λ)L
Pipe ab
Length L
Pipe Break
A1Rab
A2Rab
λL
A1Oab
(1-λ)L
A
Hydraulic Model
40
A2Oab
B
Figure 4.8 Hydraulic Model for Pipe Break
4.5.1
Deterministic Implementation
The deterministic implementation specifies the number and location of leaks and breaks,
and the orifice area of each leak, occurring in a pipeline network. Pipe leaks and breaks are then
added in the network using the models shown in Figures 4.2 and 4.8, respectively.
The
deterministic implementation can be used to simulate the performance of a water supply system
under a specific damage scenario.
4.5.2
Probabilistic Implementation
The probabilistic implementation generates randomly distributed pipeline breaks and
leaks in the system according to pipeline repair rate, RR , length, L, and the conditional
probability of pipe break, Pbk , given that damage occurs.
In addition, the probabilistic
implementation determines the type of each leak probabilistically.
The probabilistic
implementation includes three steps: generating pipe damage, deciding on damage states (leak or
break), and determining leak type.
4.5.2.1
Generating Pipe Damage
To generate the locations of pipe damage probabilistically, it is assumed that pipe damage
follows a Poisson process with a mean arrival rate equal to repair rate, RR. The repair rate is
correlated with the seismic hazard parameters, such as peak ground velocity (PGV) and
permanent ground deformation (PGD).
The determination of repair rate for each pipeline
involves spatial manipulation which is performed by GIS.
41
For a Poisson process with a mean arrival rate RR, let L1 be the first location of damage,
which is measured from the upstream node of the pipeline along its longitudinal direction. Let Lk
be the distance between the (k-1)th and kth locations of damage. The {L1, L2,···, Lk,···} is called
the sequence of interarrival distances in Poisson processes (Sheldon, 2000). The actual distance
of the kth location of damage measured from the pipe upstream node is the cumulative distance
from L1 to Lk. For instance, if L1 = 0.1L and L2 = 0.5L, where L is the length of the original
pipeline, then the first location of damage occurs at 0.1 of pipe length measured from the pipe
upstream node, and the second location of damage occurs at 0.1 + 0.5 = 0.6 of pipe length.
The L1, L2, ···, Lk can be simulated as independent exponential random variables with a
mean equal to 1/ RR (Sheldon, 2000) and generated using the Monte Carlo simulation algorithm
Lk = −
1
ln(1 − u1 )
RR
(4.21)
where u1 is a random variable which is uniformly distributed between 0 and 1. By generating the
interarrival distance Lk repeatedly until the cumulative length exceeds the pipe length, L, it is
able to determine the locations of damage in the pipeline. Figure 4.9 provides an illustration of
the pipe damage generation. In this example, a total of three locations of damage are generated
at points A, B and C, in the pipeline, because the cumulative length of the fourth location of
damage exceeds the pipe length.
L2
L1
L3 L4
B
A
C
L
L1, L2, L3, and L4 are Interarrival Lengths
Figure 4.9 Poisson Process for Pipe Damage Generation
4.5.2.2
Deciding on Damage State
42
After generating pipe damage for each location of damage, a uniformly distributed
random number µ 2 over (0, 1) is generated and compared with the conditional probability of pipe
break, Pbk , given that damage occurs. The damage is treated as a break if µ 2 exceeds Pbk , and a
leak otherwise.
The current version of GIRAFFE focuses on PGV-related pipe damage and assigns a
default value of 0.2 for the conditional probability of pipe break, Pbk , for cast iron, ductile iron,
steel with riveted joints, concrete, and other material of pipelines. For steel pipelines with
welded joints, previous data show that they are unlikely to break. Thus a default value of 0 is
assigned to Pbk for steel pipelines with welded joints. If better information is available, users can
change these default values under the Options menu in GIRAFFE by clicking on Options |
Configuration | Pipe Damage Probability.
4.5.2.3
Determining Leak Type
The third step determines the type of each leak probabilistically and calculates the orifice
area of each leak using the equations developed in Section 4.3.2. The default probabilities of
each type of leak, corresponding to various types of pipeline, are listed in Table 4.1. To
determine the type of each leak, a uniformly distributed random number, µ3 , over (0, 1) is
generated and compared with the cumulative probability of the leak types associated with the
pipeline. For example, assume the probability that a leak in a cast iron pipeline is an annular
disengagement is 0.3, round crack is 0.5, longitudinal crack is 0.1, and local loss of pipe wall is
0.1. The leak is classified as an annular disengagement if the uniformly distributed random
number is within the range between 0 and 0.3; round crack if within the range between 0.3 and
0.8, longitudinal crack if within the range between 0.8 and 0.9, and local loss of pipe wall if
within the range between 0.9 and 1.0. After deciding on the leak type, the EOA and EOD can be
calculated for each leak.
43
CHAPTER 5
EARTHQUAKE DEMAND SIMULATION
5.1
INTRODUCTION
Water supply systems are characterized by broad coverage and a high level of detail. The
broad coverage is associated with a large service area. The high level of detail is related to the
large amount of different pipelines and facilities in the system. A hydraulic network model,
which models both broad coverage and component details, will be difficult to manage and
troubleshoot. One technique for simulating a complex system is to decouple various parts of the
system, apply models with appropriate levels of complexity to each part, and integrate the
decoupled analyses to show system performance.
A water supply system typically consists of trunk and distribution systems. The trunk
system consists of large diameter trunk lines, which serve as the backbone of the system by
transporting water from sources to local areas.
The distribution system consists of small
diameter distribution lines which receive water from trunk lines and distribute it to customers.
One technique to simulate a complex water supply system is therefore to decouple the trunk and
distribution systems. The response of the system can be simulated with a system-wide trunk line
model which covers the entire service area but includes only large diameter trunk lines. In the
trunk line model, the small diameter distribution lines are replaced with demand nodes. The
local response of the system can be simulated using distribution network models, which cover a
small local area but include small diameter distribution lines. Using multi-scale modeling, a
complex water supply system can be decoupled into several systems which have manageable
complexity.
The H2ONET LADWP hydraulic network model is a trunk line model that includes 2200
km of pipelines from the LADWP trunk line system ranging in diameter from 300 to 3850 mm
and replaces the remaining 9800 km of distribution pipelines as demand nodes. The trunk line
model can give an accurate prediction of flows and pressures in the trunk system if the nodal
44
demands can be simulated accurately. These demands represent the aggregated demands from
the downstream distribution networks. In normal operations, the demands from the distribution
networks are known values that are relatively easy to simulate. The demands are much more
difficult to simulate after the system has sustained earthquake-induced damage.
GIRAFFE provides a method to simulate the earthquake demand associated with
distribution networks. The earthquake demands are simulated by means of fragility curves
relating demand to repair rate in local distribution networks. The repair rate is correlated with
seismic hazard parameters including peak ground velocity and permanent ground deformation.
The fragility curves were developed using distribution network simulations of the LADWP water
supply system.
5.2
METHODOLOGY
A detailed description of the development of the fragility curves for earthquake demand
simulation is provided by Shi (2006). To develop the fragility curves, five distribution networks
were selected to be representative of the roughly 30 LADWP distribution network models used
for local flow and pressure analyses. Each distribution network model covers one large pressure
zone or several small pressure zones.
Figure 5.1 shows the locations of the five chosen
distribution networks. Figure 5.2 provides an expanded view of the distribution network in
pressure zone 1000, superimposed on the trunk system model. The distribution network includes
both large diameter trunk lines and small diameter distribution lines. The smallest pipelines in
the distribution networks have a diameter of 100 (4 in.) or 150 mm (6 in.), and the majority of
pipelines have diameters smaller than 300 mm (12 in.).
In distribution network simulations, pipe damage is evaluated only in the distribution
lines since trunk line damage is accounted for explicitly in the trunk system model. The pipe
damage is assumed to follow a Poisson process with a mean arrival rate equal to repair rate, RR,
and is generated using Monte Carlo simulation. Flow analysis is performed for the damaged
system and negative pressures are eliminated using the iterative approach described in Chapter 3.
Flows in trunk lines before and after damage to distribution lines are monitored, and the flows
45
after damage are normalized to the flows before damage. The normalized flows provide a proxy
Distribution
System 1000
Figure 5.2 Overlay of Distribution and
Trunk system.
Normalized Demand
6
Prediction (Mean)
Prediction (Upper 68% Confidence Interval)
Prediction (Upper 90% Confidence Interval)
Theoretical Lower Bound
5
4
3
m(p)
2
c(p)
1
ND = c(p) + m(p)RR
0
0
0.2
0.4
0.6
Repair Rate (No/km)
0.8
1
Figure 5.3 Prediction of Normalized Demand
Figure 5.1 LADWP Water Supply System.
for the normalized demands since water from the trunk lines is distributed by means of nodal
demands. Monte Carlo simulations were performed for RR ranging from 0.02 to 100 repairs/km
and statistical analysis is performed for RR ranging from 0.02 to 1 repairs/km, which is a typical
range for PGV-related pipeline damage.
The normalized demands, representing the increase of demands from damage to
distribution lines, are expressed as fragility curves in the format
ND = I + S x RR
(5.1)
in which ND is the normalized demand, I and S are the intercept and slope of the linear
regression, respectively, and RR is the repair rate. The intercept, I, and slope, S, are further
correlated with the mean pressure, MP, of the distribution network and statistically estimated
46
from the simulation data in the five distribution networks (Shi, 2006). Estimates with different
confidence levels can be obtained for the intercept, I, and slope, S.
Two simulation options are provided in GIRAFFE, mean prediction with noise terms and
90% confidence level prediction. The equation for the mean prediction with noise terms is in a
format of
ND = I + S × RR
= [II + IS × MP + N (0, σ I )] + [SI + SS × MP + N (0, σ S )]× RR
= {0.9012 + 0.0036 MP + N [0, (−0.0198 + 0.0015MP)]}
+ {− 0.877 + 0.0248MP + N [0, (−0.351 + 0.0094 MP)]}RR
(5.2)
in which, II and IS are the intercept and slope of the intercept term, I, N (0, σ I ) is a Gaussian
random variable with zero mean and standard deviation of σ I , SI and SS are the intercept and
slope of the slope term, S, and N (0, σ S ) is a Gaussian random variable with zero mean and
standard deviation σ S . The default values of II, IS, SI and SS used in GIRAFFE are determined
from the mean regressions on the basis of the simulation data from the five representative
distribution networks (Shi, 2006). Users have the option of changing these parameters under the
Options menu in GIRAFFE (Options | Configuration | Nodal Demand Calibration). The σ I and
σ S are used to simulate the variation of the mean values of the intercept, I, and slope, S, with
respect to their mean values. The σ I and σ S are also correlated with mean pressure, MP, and
their values are evaluated using regressions of the simulation data from the five representative
distribution networks (Shi, 2006). Users may change the default values for σ I and σ S under the
Options menu in GIRAFFE (Options | Configuration | Nodal Demand Calibration).
The equation for the 90% confidence level prediction is in a format of
ND = I + S × RR
= [II + IS × MP ] + [SI + SS × MP ]× RR
= (1.1412 + 0.0055MP ) + (− 0.0514 + 0.0347 MP )RR
47
(5.3)
in which, II and IS are the intercept and slope of the intercept term, I, and SI and SS are the
intercept and slope of the slope term, S. The default values of II, IS, SI and SS in GIRAFFE are
determined from the 90% confidence level regressions of the simulation data from the five
representative distribution networks (Shi, 2006). Users may change the default values for II, IS,
SI and SS under the Options menu in GIRAFFE (Options | Configuration | Nodal Demand
Calibration).
Figure 5.3 shows the prediction of normalized demands in a pressure zone with a mean
pressure of 0.69 MPa (100 psi). From this figure, the mean estimate of demand including postearthquake demand from leaks and breaks for a RR equal to 1 repair/km is approximately 2.5
times the design demand, while the upper 90% confidence level estimate is roughly 5 times the
design demand.
The earthquake demand simulation is pressure zone based. The basic input parameters
are MP and RR associated with each demand node. For a demand node, MP is the average nodal
pressure in the pressure zone in which the demand node is located before system damage. The
MP can be obtained by performing a hydraulic network analysis on the undamaged system and
then conducting a statistical analysis on the nodal pressures with respect to pressure zones. The
RR represents the repair rate of the distribution lines around the demand node. For PGV-related
pipe damage, the RR is calculated using regression relationships between PGV and RR
developed from previous investigations (e.g., Jeon, 2002; Jeon and O’Rourke, 2005). The
determination of RR for a given earthquake scenario involves spatial manipulation and is
performed using GIS, which gives the RR related to each demand node as input to GIRAFFE.
The GIS procedures for determining the RR is explained in Shi (2006). Users may change the
value for RR under the Options menu in GIRAFFE (Options | Configuration | Nodal Demand
Calibration). After the determination of ND for each demand node, GIRAFFE then calculates
the demands after the earthquake by multiplying the ND by the original demands, and modifies
the system definition file by replacing the original demands with the post-earthquake demands.
48
CHAPTER 6
GIRAFFE INPUTS AND OUTPUTS
6.1
INTRODUCTION
The input for GIRAFFE simulations includes control parameters and data files. The
control parameters specify the lowest pressure to be eliminated, the time length and time step to
update tank water levels, and simulation options. The input data includes files for system
definition, pipe damage generation, and earthquake demand simulation. The major outputs from
GIRAFFE simulations are hydraulic analysis results of network physical components, including
junctions, tanks, pipes, pumps, and valves, and the serviceability of the damaged system. The
input parameters and data files and the output files are introduced in this chapter.
6.2
INPUT
GIRAFFE can perform both deterministic and probabilistic simulations. For probabilistic
simulations, users can either specify the number of Monte Carlo simulation runs or let the
program determine the number of simulation runs using the self-termination algorithm built into
the code. For both deterministic and probabilistic simulations, users need to input some common
control parameters to specify the system definition file, lowest pressure to be eliminated, total
length of time to update tank water levels, and time step to update tank water levels.
6.2.1
Control Parameters
Upon starting the GIRAFFE program, a window appears prompting the user to select a
simulation option: Deterministic, Monte Carlo Fixed or Monte Carlo Flexible. Users may also
select the simulation type by clicking on the Simulations drop down menu in the toolbar. Table
6.1 lists the input control parameters that are required for each of the 3 simulation options.
49
Table 6.1 GIRAFFE Control Parameters
Name
Description
System
Definition File
Name of the EPANET system definition file with the extension
of .inp.
File name may have a maximum length of 80
characters.
Minimum
Pressure limit, in psi, below which GIRAFFE eliminates the
Pressure to
node and connected links from the system. The typical input is
Eliminate
0 psi for negative pressure elimination.
Simulation Time
Total length of simulation time in hours to update tank water
levels. 0 for steady state simulation.
Simulation Time The time step in hours to update tank water levels. 1 for steady
Step
6.2.2
state simulation.
Deterministic Simulations
If the user selects a deterministic simulation, the GIRAFFE GUI window that appears
asks the user to input the name of the file in which the pipe damage information is stored. An
example of the GUI window is shown in Figure 7.3. Table 6.2 shows the name and descriptions
of the parameter for specifying the pipe damage file. Table 6.3 shows an example of the pipe
damage file.
The pipe damage file is a tab-delimited text file. Users can use Microsoft Word, Excel,
or Notepad to construct the file and save it with the typical extension of text files, such as .inp.
Users may also create a pipe damage file via the GIRAFFE GUI for a deterministic simulation as
shown in Appendix A. The input file consists of two blocks with one storing pipe break
information and the other storing pipe leak information.
The block storing pipe break
information starts with the line [Pipe_Break_Information]. Users need to copy this exact line
into their input file, and not leave any space before [Pipe_Break_Information], otherwise the
program will not run correctly.
50
Table 6.2 Input Parameter for Pipe Damage Generation File for Deterministic Simulations
Name
Type
PipeBreak
char
Description
Name of input file for pipe damage generation. File name may
have a maximum length of 80 characters.
Table 6.3 Input File for Pipe Damage Generation for Deterministic Simulations
[Pipe_Break_Information]
PipeID
PreRatio
BreakRatio
22
22
12
0
0.6
0
2
BreakNo
LeakNo
PreIndex
3
3
1
1
2
1
0
1
0
0
0
0
0.3
0.9
0.5
[Pipe_Leak_Information]
PipeID
LeakD
PreRatio
22
RepairNo
0.3
LeakRatio
RepairNo
0.6
3
BreakNo
LeakNo
PreIndex
1
1
1
The second line of the pipe damage file is a headline describing the type of values in each
column in the pipe break records that follow. It is recommended that users copy the headline
into their input file. The headline terms in the pipe break records are explained in Table 6.4.
The block storing pipe leak information starts with a line with [Pipe_Leak_Information].
Users need to copy this exact line into their input file and not leave any space before
[Pipe_Leak_Information], otherwise the program will not run correctly. The next line is a
headline describing the type of values in each column in the pipe leak records that follow. It is
recommended that users copy the headline into their input file. The headline terms in the pipe
leak records are explained in Table 6.5.
51
Table 6.4 Description of Columns in Pipe Break Section
Name
Type
PipeID
char
Explanation
The ID of the pipe which users want to break. Maximum length
of 30 characters.
The length ratio of the previous location of pipe damage, either
PreRatio
float
break or leak, in the same pipeline. If the current break is the first
location of damage in the pipeline, then the PreRatio is set to 0.
BreakRatio
float
RepairNo
int
The length ratio of the location of the current pipe break.
The total number of locations of pipe damage, including breaks
and leaks, in the pipeline.
The number of locations of breaks in the upstream of the current
BreakNo
int
location of pipe break. The current location of pipe break is
counted.
LeakNo
int
The number of locations of leaks in the upstream of the current
location of pipe break.
The type of the previous location of pipe damage immediately
PreIndex
int
upstream of the current break: 0 for leak and 1 for break. If the
current break is the first location of pipe damage in the pipeline.
The PreIndex is set to 0.
52
Table 6.5 Description of Columns in Pipe Leak Section
Name
Type
PipeID
char
LeakD
float
Explanation
The ID of the pipe which users want to add the leak. Maximum
length of 30 characters.
Equivalent orifice diameter of the leak with the units of inches.
The length ratio of the previous location of pipe damage, either break
PreRatio
float
or leak, in the same pipeline. If the current leak is the first location of
damage in the pipeline, then the PreRatio is set to 0.
LeakRatio
float
RepairNo
int
BreakNo
int
LeakNo
int
The length ratio of the location of the current leak.
The total number of locations of pipe damage, including breaks and
leaks, in the pipeline.
The number of locations of breaks in the upstream of the current
location of pipe leak.
The number of locations of leaks in the upstream of the current
location of pipe leak. The current location of pipe leak is counted.
The type of the previous location of pipe damage immediately
PreIndex
int
upstream of the current leak: 0 for leak and 1 for break. If the current
leak is the first location of pipe damage in the pipeline. The Preindex
is set to 0.
6.2.3
Monte Carlo with Fixed Simulation Runs
If the user selects a “Monte Carlo Fixed Number” simulation, GIRAFFE will perform a
Monte Carlo simulation with a number of simulation runs specified by the user. GIRAFFE will
ask the user to input the name of the file storing information for probabilistic pipe damage
generation. A user has the option to choose to perform the earthquake demand simulation or not.
If users choose to perform the earthquake demand simulation, they need to choose between the
simulation options of mean prediction with noise terms or 90% confidence level prediction. An
example of the GUI window with inputs is shown in Figure 7.8. The parameters users need to
input (in addition to the control parameters) are listed in Table 6.6 in sequence.
53
Table 6.6 Input Parameters for Monte Carlo Simulations with Fixed Simulation Times
Name
Description
Name of the input file for probabilistic pipe damage
Pipe Repair Rate File
generation. File name may have a maximum length of
80 characters.
Number of Simulations
Monte Carlo simulation time ranging from 1 to 100
Random Seed
Seed for random number generation.
Nodal Demand
Option to choose to simulate the earthquake demand or
Calibration
not: “Yes” for simulated and “No” for not simulated.
(If “Yes” was selected for “Nodal Demand Calibration”,
Regression Equation
this value is required.) Options for earthquake demand
simulation: “Mean Prediction Plus Noise Terms” or
“90% Confidence Level Prediction”.
(If “Yes” was selected for “Nodal Demand Calibration”,
Mean Pressure File
this value is required.) Name of the input file for
earthquake demand assessment. File name may have a
maximum length of 80 characters.
One example of an input file for probabilistic pipe damage generation is given in Table
6.7. This file is a tab-delimited text file and users can use Microsoft Word, Excel, or Notepad to
construct the input file and save it with the extension .inp. The probabilistic pipe damage input
file starts with a headline, followed by the record of each pipeline. It is recommended that users
copy the headline to their own files. The headline terms in the pipe damage generation input file
are explained in Table 6.8.
An example of an input file for earthquake demand simulation is shown in Table 6.9.
This is also a tab-delimited text file which users can create using Microsoft Word, Excel, or
Notepad, and save as a text file with the extension .inp. The input file starts with a headline,
followed by the record of each demand node. The headline terms in the earthquake demand
simulation input file are explained in Table 6.10.
54
Table 6.7 Input File for Probabilistic Pipe Damage Generation
PipeID
10
12
16
18
20
22
4
6
8
Length_km
RR
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Material
CI
CI
DI
DI
CON
CON
RV
RV
STL
Table 6.8 Description of Columns in Probabilistic Pipe Damage Input File
Name
Type
Description
The ID of the pipe which users want to damage. Users have to make
PipeID
char
sure this pipe is in the system definition file otherwise the program
cannot run correctly. Maximum length 30 characters
Length
float
The length of the pipe in km. The length of each pipe can be obtained
from the system definition file.
Pipe repair rate in repairs per kilometer of pipe length, which is
correlated with seismic hazard parameters, such as peak ground
RR
float
velocity and permanent ground deformation. The determination of
repair rate for each pipeline involves spatial manipulation, which is
conducted using GIS.
The material of the pipeline. CI: cast iron pipeline; DI: ductile iron
Material
char
pipeline, RS: riveted steel pipeline; CON: concrete pipeline; STL:
welded steel pipeline, and N/A: other types of pipelines beside the
abovementioned five types of pipeline.
55
Table 6.9 Input File for Earthquake Demand Simulation
ID
G_RR
Ave_PRESSURE
CC1007
0.15160
77.1906
CC1043
0.13645
59.7064
CC1053
0.11148
77.1906
Table 6.10 Description of Columns in Earthquake Demand Simulation Input File
Name
Type
Description
The ID of the demand node. Users have to make sure this
ID
char
demand node is in the system definition file otherwise the
program cannot run correctly.
Maximum length of 30
characters.
Pipe repair rate in repairs per kilometer of pipe length, which is
correlated with seismic hazard parameters, such as peak ground
G_RR
float
velocity and permanent ground deformation. The determination
of repair rate for each pipeline involves spatial manipulation,
which is conducted using GIS.
Ave_PRESSURE float
6.2.4
The average nodal pressure of the pressure zone, in which the
demand node is located.
Monte Carlo with Flexible Simulation Runs
If the user selects a “Monte Carlo Flexible Number” simulation, GIRAFFE will perform
a Monte Carlo simulation in which the program will automatically determine how many
simulation runs are needed as per default or user-specified convergence criteria. An example of
a GUI window with “Monte Carlo Flexible” inputs is shown in Figure 7.9. The input parameters
are similar to those for the “Monte Carlo Fixed” simulation and are shown in Table 6.11. The
pipe damage generation and demand simulation input files have the same formats as those used
56
for a “Monte Carlo Fixed Number” simulation. Users can refer to Tables 6.7 to 6.10 for the
format of these input files.
Table 6.11 Input Parameters for Monte Carlo Simulations with Flexible Simulation Runs
Variable Name
Pipe Repair Rate File
Random Seed
Calibrate Nodal Demand
Explanation
Name of input file for pipe damage generation. File
name may have a maximum length of 80 characters.
Seed of random number generation.
Options to choose to simulate the earthquake demand or
not: “Yes” for simulated and “No” for not simulated.
(If “Yes” was selected for “Nodal Demand Calibration”
Regression Equation
this value is required.) Options for earthquake demand
simulation: “Mean Prediction Plus Noise Terms” or
“90% Confidence Level Prediction”.
(If “Yes” was selected for “Nodal Demand Calibration”
Mean Pressure File
this value is required.) Name of the input file for
earthquake demand assessment. File name may have a
maximum length of 80 characters.
6.3
Definition Parameters
Besides the parameters and data files described above, the GIRAFFE code includes a
parameter definition file named as parameter_definition.h, which defines the simulation capacity,
parameters for leak simulations, and parameters for earthquake demand simulations. GIRAFFE
is configured to work with the LADWP water supply system, which represents one of the largest
water supply systems in the world. As such, GIRAFFE should have enough capacity to simulate
other water supply systems but doing so may require a change to the definition parameters. The
default parameters for leak and earthquake demand simulations are based on the best data
currently available. The parameter definition file is shown in Table 6.12 with comments on each
defined parameter. Users generally do not need to change the values of the parameters in
57
parameter_definition.h. As such, these parameters are defined in the code to avoid too many
input parameters from users. Users have the ability to change many of the default parameters by
clicking on Options | Configuration in the GIRAFFE toolbar menu.
Alternatively, a user can
modify the file, parameter_definition.h, by changing the number after each variable definition
and rebuilding the code to generate a new executable file.
Users may change any of these default parameters located in the toolbar menu under Options |
Configuration and then save the new system configuration. The default parameter configuration
is saved as Default.txt in the “Configuration” folder that exists in the same directory where the
GIRAFFE application is installed. To save a new configuration, click on Options | Configuration
| System Options in the GIRAFFE toolbar and a window as shown in Figure 6.1 appears. From
this window, the user can change the output folder, load an existing configuration or save an
existing configuration. Clicking on “Load Existing Configuration” will take the user to the
“Configurations” folder where they can select any saved configuration files to load. Clicking on
“Save Existing Configuration” will allow the user to save the current set of parameters that can
be defined under the Options | Configuration menu. This option allows the user to switch
between different parameter configurations quickly and easily and thus avoid having to change
parameter values between simulation runs.
Figure 6.1 Configuration Window for System Options
58
Table 6.12 Parameter Definition File
//*********************************Defining Constants*************************************//
#define NJunction 10000 //Maximum number of junctions in a network//
#define NPipe 10000 //Maximum number of pipes in a network//
#define NLink 20000 //Maximum number of links in a network//
#define NNode 20000 //Maximum number of nodes in a network//
#define NDemandNode 1200 //Maximum number of demand nodes in a network//
#define LID 20 //Maximum number of characters for network component IDs//
#define MaxNSimu 100 //Maximum number of Monte Carlo simulations//
#define LFileName 80 //Maximum number of characters for file name and directory//
#define LLine 255 //Maximum number of characters in a line in input text files//
#define MLinktoNode 15 //Maximum number of links connected to the same node//
#define MaxNDamage 100 //Maximum number of locations of damage in a pipeline//
#define MaxLMat 3 //Maximum number of characters used to define pipe material//
#define MaxTime 10 //Maximum times to update the tank water level in a simulation//
#define MaxBreak 200 // Maximum number of breaks occurring in a network//
#define MaxLeak 1000 // Maximum number of leaks occurring in a network//
//***************************Defining Parameters for Modeling Leakage*************************//
#define DLLeak 0.5 //Default length of an added pipe for simulating leaks//
#define DCLeak 1000000 //Default roughness coefficient of an added pipe for simulating leaks//
#define DMLeak 1 //Default minor loss coefficient of an added pipe for simulating leak//
#define BreakProCI 0.2 //Probability of pipe break, given pipe damage occurs, for cast iron pipes//
#define BreakProDI 0.2 //Probability of pipe break, given pipe damage occurs, for ductile iron pipes//
#define BreakProRS 0.2 //Probability of pipe break, given pipe damage occurs, for riveted steel pipes//
#define BreakProCON 0.2 //Probability of pipe break, given pipe damage occurs, for concrete pipes//
#define BreakProSTL 0 //Probability of pipe break, given pipe damage occurs, for welded steel pipes//
#define STLLeakRatio 0.2 //Probability of pipe leak, given pipe damage occurs, for welded steel pipes//
#define Type1tD 0.4 //Thickness of annular space for leak type 1, annular disengagement, in the units //
// of inches; t in Eqn. 4.5//
#define Type1kD 0.3 //Ratio of actual leak area to the maximum possible leak area for leak type 1, annular//
//disengagement; k in
Eqn. 4.5//
Table
6.12
Continued
#define Type2aD 0.5 //Opening angle of leak type 2, round crack, in the units of degrees; θ in Eqn. 4.7//
#define Type3kD 480 //Length of leak type 3, longitudinal crack, in the units of inches; L in Eqn. 4.11//
59
Table 6.12 Continued
#define Type3aD 0.1 //Opening angle of leak type 3, longitudinal crack, in the units of degrees; θ in //
//Eqn. 4.11//
#define Type4kD 0.05 //Ratio of the length and width of leak type 4, local loss of pipe wall, to the pipe //
//diameter and circumferential length, respectively; k1 and k2 in Eqn. 4.15//
#define Type5kD 0.3 //Ratio of the length of leak type 5, local tear of pipe wall, to the pipe circumferential//
//length; k in Eqn. 4.19//
#define Type5wD 0.5 //Width of the leak type 5, local tear of pipe wall, in the units of inches; w in Eqn. 4.19//
#define CIType1D 0.3 //Probability of leak type 1 for cast iron pipelines//
#define CIType2D 0.8 //Cumulative probability of leak types 1 to 2 for cast iron pipelines//
#define CIType3D 0.9 //Cumulative probability of leak types 1 to 3 for cast iron pipelines//
#define CIType4D 1.0 //Cumulative probability of leak types 1 to 4 for cast iron pipelines//
#define CIType5D 1.0 //Cumulative probability of leak types 1 to 5 for cast iron pipelines//
#define RSType1D 0.6 //Probability of leak type 1 for riveted steel pipelines//
#define RSType2D 0.6 //Cumulative probability of leak types 1 to 2 for riveted steel pipelines//
#define RSType3D 0.9 //Cumulative probability of leak types 1 to 3 for riveted steel pipelines//
#define RSType4D 1.0 //Cumulative probability of leak types 1 to 4 for riveted steel pipelines//
#define RSType5D 1.0 //Cumulative probability of leak types 1 to 5 for riveted steel pipelines//
#define CONType1D 1.0 //Probability of leak type 1 for concrete pipelines//
#define CONType2D 1.0 //Cumulative probability of leak types 1 to 2 for concrete pipelines//
#define CONType3D 1.0 //Cumulative probability of leak types 1 to 3 for concrete pipelines//
#define CONType4D 1.0 //Cumulative probability of leak types 1 to 4 for concrete pipelines//
#define CONType5D 1.0 //Cumulative probability of leak types 1 to 5 concrete pipelines//
#define DIType1D 0.8 //Probability of leak type 1 for ductile iron pipelines//
#define DIType2D 0.8 //Cumulative probability of leak types 1 to 2 for ductile iron pipelines//
#define DIType3D 0.9 //Cumulative probability of leak types 1 to 3 for ductile iron pipelines//
#define DIType4D 1.0 //Cumulative probability of leak types 1 to 4 for ductile iron pipelines//
#define DIType5D 1.0 //Cumulative probability of leak types 1 to 5 for ductile iron pipelines//
#define STLType1D 0.0 //Probability of leak type 1 for welded steel pipelines//
Table 6.12 Continued
#define STLType2D 0.0 //Cumulative probability of leak types 1 to 2 for welded steel pipelines//
#define STLType3D 0.0 //Cumulative probability of leak types 1 to 3 for welded steel pipelines//
#define STLType4D 0.0 //Cumulative probability of leak types 1 to 4 for welded steel pipelines//
#define STLType5D 1 //Cumulative probability of leak types 1 to 5 for welded steel pipelines//
60
Table 6.12 (Continued)
#define OtherType1D 0.2 //Probability of leak type 1 for welded pipelines with other materials//
#define OtherType2D 0.4 //Cumulative probability of leak types 1 to 2 for pipelines with other materials//
#define OtherType3D 0.6 //Cumulative probability of leak types 1 to 3 for pipelines with other materials//
#define OtherType4D 0.8 //Cumulative probability of leak types 1 to 4 for pipelines with other materials//
#define OtherType5D 1.0 //Cumulative probability of leak types 1 to 5 for pipelines with other materials//
//*******************Defining Parameters for Earthquake Demand Simulation*********************//
#define MiiMP 0.9012 //Intercept of the intercept term of the linear regression between normalized //
//demand and repair rate for mean regression; II in Eqn. 5.2//
#define MisMP 0.0036 //Slope of the intercept term of the linear regression between normalized //
//demand and repair rate for mean regression; IS in Eqn. 5.2//
#define MsiMP 0.877 //Intercept of the slope term of the linear regression between normalized //
//demand and repair rate for mean regression; SI in Eqn. 5.2//
#define MssMP 0.0248 //Slope of the slope term of the linear regression between normalized //
//demand and repair rate for mean regression; SS in Eqn. 5.2//
#define MiiSD -0.0198 //Intercept of the linear regression between the standard deviation of mean intercept//
//and mean pressure; see Eqn. 5.2//
#define MisSD 0.0015 //Slope of the linear regression between the standard deviation of mean intercept//
//and mean pressure; see Eqn. 5.2//
#define MsiSD -0.351 //Intercept of the linear regression between the standard deviation of mean slope//
//and mean pressure; see Eqn. 5.2//
#define MssSD 0.0094 //slope of the linear regression between the standard deviation of mean slope//
//and mean pressure; see Eqn. 5.2//
#define UiiMP 1.1412 //Intercept of the intercept term of the linear regression between normalized //
//demand and repair rate for 90% confidence level regression; II in Eqn. 5.3//
#define UisMP 0.0055 //Slope of the intercept term of the linear regression between normalized //
//demand and repair rate for 90% confidence level regression; IS in Eqn. 5.3//
#define UsiMP -0.0514 //Intercept of the intercept term of the linear regression between normalized//
//demand and repair rate for 90% confidence level regression; SI in Eqn. 5.3//
#define UssMP 0.0347 //Slope of the intercept term of the linear regression between normalized //
//demand and repair rate for 90% confidence level regression; SS in Eqn. 5.3//
#define pi 3.1415926 //Constant pi//
#define mRRCap 0.02 //Lower bound of repair rate for Monte Carlo simulation, below which it is assumed //
//that no pipe damage occurs. The lower bound is to avoid the numerical stability //
//problems when using the Eqn. 4.20 to generation locations of pipe damage.//
61
6.4
Outputs
The major outputs for GIRAFFE simulations are the hydraulic analysis results for each
type of network physical component, including junctions, tanks, pipes, pumps, and valves.
GIRAFFE also reports the serviceability of the damaged system.
6.4.1
Deterministic Simulations
The main outputs of deterministic simulations are hydraulic analysis results for junctions,
tanks, pipes, pumps, and valves, which are reported in the text files, JunctionResults_Time*.out,
TankResults_Time*.out,
PipeResults_Time*.out,
PumpResults_Time*.out,
and
ValveResults_Time*.out, respectively. GIRAFFE also reports the serviceability of each demand
node and the entire system in the text file, Serviceability*.out. GIRAFFE saves the damaged
system in Damage_System_Time*.inp, and modified system in Modified_System_Time*.inp, for
users to visualize the damaged and modified systems. In these files, the character * represents
the simulation time in the units of hours. For example, a simulation including tank level change
over 24 hours where the tank level update is set at 24 hours would have two sets of results
generated, one at time 0 and one at time 24. The Damage_System_Time0.inp represents the
system immediately after pipeline damage is added in the network. No hydraulic simulation and
negative pressure elimination are performed to the Damage_System_Time0.inp.
The
Modified_System_Time0.inp represents the system after hydraulic simulation and negative
pressure elimination of the Damage_System_Time0.inp. The Damage_System_Time24.inp is the
Modified_System_Time0.inp with tank water levels updated according to the simulation results at
time 0 and the time step, 24 hours. The Modified_System_Time24.inp represents the system after
hydraulic simulation and negative pressure elimination to the Damage_System_Time24.inp. The
detailed formats of these files can be found in the examples presented in Chapter 7.
6.4.2
Monte Carlo Simulations
The main outputs of the Monte Carlo simulation are system serviceability. The system
serviceability information is reported in the file, Serviceability*.out.
62
The serviceability is
reported in a matrix format. For each Monte Carlo simulation run, the serviceability is reported
for each demand node and for the entire system.
The mean of the nodal and system
serviceability for all Monte Carlo simulation runs is also calculated and reported. GIRAFFE also
reports the results of junctions, tanks, pipes, pumps, and valves, in JunctionResults_Time*.out,
TankResults_Time*.out,
PipeResults_Time*.out,
ValveResults_Time*.out for each run of simulation.
63
PumpResults_Time*.out,
and
CHAPTER 7
GIRAFFE SIMULATION EXAMPLES
7.1
INTRODUCTION
This chapter provides an example associated with the three GIRAFFE simulation options,
which are deterministic, Monte Carlo with fixed simulation runs, and Monte Carlo simulation
with flexible runs. The water supply system used in the example is introduced in the first
subheading. The inputs and outputs associated with each of the three simulation options are
explained in the three subheadings that follow.
7.2
HYDRAULIC NETWORK MODEL
Since the LADWP hydraulic network model works with the H2ONET software, this
example applies H2ONET to construct the hydraulic network model. Detailed procedures for
constructing a hydraulic network model using H2ONET can be found in the H2ONET users
manual (MWH Soft Inc., 1999). The H2ONET hydraulic network model is then exported
directly from H2ONET to EPANET input file format. To export the H2ONET model, users
need to go to the Exchange dropdown menu in the H2ONET GUI, find the EPANET v2.0 menu,
click the Export button, and specify the directory and name of the export file.
Figure 7.1 shows the hydraulic network model with the H2ONET GUI. The menu used
to export the H2ONET hydraulic model to an EPANET input file is also shown in this figure.
The network contains 1 reservoir with ID 1, 1 tank with ID 7, 1 pump with ID 2, 1 PRV with ID
14, and 9 pipes. Eight demand nodes are distributed around the network. Each demand node has
a demand of 100 gpm. In general, water flows from the tank and reservoir in the northwest
towards the southeast to satisfy the demands. The EPANET input file exported from H2ONET
is shown in Table 7.1. Detailed descriptions of the EPANET input file can be found in the
64
Figure 7.1 Hydraulic Network Model Constructed by H2ONET
EPANET Users Manual (Rossman, 2000). The hydraulic network exported from H2ONET can
be analyzed by the EPANET engine and the analysis results can be visualized using the GUI of
EPANET shown in Figure 7.2. In this figure, the node and link IDs are shown as black numbers.
The link flows in units of gpm and nodal pressures in units of foot of water height are coded
using the colors indicated in the legends.
65
Table 7.1 EPANET Format System Definition File
[TITLE]
[JUNCTIONS]
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
200.000000
17
100.000000
19
200.000000
[RESERVOIRS]
1
450.000000
[TANKS]
7
450.000000
120.000000
[PIPES]
10
12
16
18
20
22
4
6
8
9
11
15
17
19
19
5
9
11
7
9
13
13
15
17
3
3
5
0.000000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
[PUMPS]
2
1
3
POWER
10.000000
[VALVES]
14
9
13
4.000000
PRV
1.000000
1.000000
1.000000
1.000000
120.000000
30.000000
0.000000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
100.000000
0.000000
[DEMANDS]
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
100.000000
17
100.000000
19
100.000000
[CURVES]
[PATTERNS]
PATN1 1.000000
PATN1 1.000000
`
[STATUS]
1.000000
1.000000
Table 7.1 Continued
66
1.000000
1.000000
1.000000
Table 7.1 Continued
[CONTROLS]
[SOURCES]
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.000000
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
[COORDINATES]
1
140.726688
3
169.667221
5
169.576993
7
207.220708
9
207.220708
11
207.252760
13
241.998111
15
241.998111
17
280.016223
19
280.016223
174.581772
174.431972
130.595466
199.588372
174.450158
130.579090
174.517132
129.944774
174.565669
130.044299
[VERTICES]
[End]
67
Figure 7.2 Hydraulic Simulation Results for Undamaged System from EPANET
68
7.3
DETERMINISTIC SIMULATIONS
The hydraulic network was first analyzed deterministically by GIRAFFE. The input
parameters, data files, and output files for this deterministic simulation are described below.
7.3.1
Inputs
Figure 7.3 shows the GIRAFFE GUI window with inputs for a deterministic simulation.
The hydraulic network model, which is defined in the EPANET system definition file,
Example_1.inp, was analyzed by GIRAFFE. The simulation time is 24 hours and the time step
to update the tank water levels is also 24 hours such that the tank water levels are updated once
after 24 hours of running.
Table 7.2 shows the input file for pipe damage generation,
Pipe_Damage.inp. Three breaks occurred in this network with two breaks occurring in pipe 22
and one in pipe 12. The two breaks occurred in pipe 22 are differentiated by their different
length ratios, 0.3 and 0.9, respectively. The one break in pipe 12 occurred at the middle point of
pipe 12 with a length ratio of 0.5. One leak occurred in pipe 22 with a length ratio of 0.6 and
leak diameter of 2 inches.
69
Figure 7.3 Inputs for Deterministic Simulation
Table 7.2 Input File for Pipe Damage Information for Deterministic Simulation
(Pipe_Damage.inp)
[Pipe_Break_Information]
PipeID
PreRatio
BreakRatio
22
22
12
0
0.6
0
7.3.2
2
BreakNo
LeakNo
PreIndex
3
3
1
1
2
1
0
1
0
0
0
0
0.3
0.9
0.5
[Pipe_Leak_Information]
PipeID
LeakD
PreRatio
22
RepairNo
0.3
LeakRatio
RepairNo
0.6
3
Simulation Procedures
70
BreakNo
LeakNo
PreIndex
1
1
1
After GIRAFFE receives the inputs, it performs the deterministic simulation according to
the following procedures.
1) Damage the network and output the damaged system, Damage_System_Time01.inp.
2) Apply the EPANET engine to perform hydraulic network analysis to the damaged
system and the iterative approach to eliminate negative pressures or pressures below
the set threshold pressure.
The elimination process continues until no negative
pressures exist in the network.
3) Output the system definition file, Modified_System_Time01.inp, and report the results
of each type of physical component in the files, JunctionResults_Time0.out,
TankResults_Time0.out,
PipeResults_Time0.out,
PumpResults_Time0.out,
and
ValveResults_Time0.out.
4) Calculate the system serviceability at time 0 and report the system serviceability in
the file, Serviceability0.out.
5) Read the TankResults_Time0.out, determine the outflow of each tank, and update the
tank water levels according to the initial tank water levels, tank cross-section areas,
tank outflows, and the time step. In this example, GIRAFFE updates the water level
of tank with ID 7 once after 24 hours of tank running.
6) Output the damaged system, Damage_System_Time241.inp.
7) Apply the EPANET engine to perform hydraulic network analysis to the system with
tank water level updated, and the iterative approach to eliminate negative pressures.
The elimination process continues until no negative pressures exist in the network.
8) Output the system definition file, Modified_System_Time241.inp, and report the
hydraulic simulation results of each type of physical component in the files,
JunctionResults_Time24.out,
TankResults_Time24.out,
PumpResults_Time24.out, and ValveResults_Time24.out.
71
PipeResults_Time24.out,
9) Calculate the system serviceability at time 24 and report the system serviceability in
the file, Serviceability24.out.
7.3.3
Outputs
GIRAFFE reports two sets of simulation results, with one at time 0 and the other at time
24.
7.3.3.1
Outputs at Time 0
The Damage_System_Time01.inp, shown in Table 7.3, stores the system definition
information immediately after the system damage. Comparing Tables 7.1 and 7.3 show that 1) 1
junction with ID A1J22 is added in the [JUNCTION] section to model the pipe leak in pipe 22; 2)
7 reservoirs, with IDs A1R22, A2R22, A3R22, A4R22, A5R22, A1R12, and A2R12 are added in
the [RESERVOIR] section to model the two breaks in pipe 22, 1 leak in pipe 22, and 1 break in
pipe 12; 3) the original pipe 22 in the [PIPES] section is replaced with pipes A1O22, A2O22,
A3O22, and A4O22 because of the three locations of damage, including two breaks and one leak,
occurred in this pipe; 4) the original pipe 12 in the [PIPES] section is replaced with pipes A1O12
and A2O12 because one break occurred in the pipeline; and 5) One pipe A1L22 is added in the
[PIPES] section to model the leak occurred in pipe 22. Users can use the EPANET GUI to
visualize the damaged system as shown in Figure 7.4.
The Modified_System_Time01.inp stores system definition information after the
GIRAFFE analysis of the Damage_System_Time01.inp. In this system, the negative pressure
nodes and connected links have been eliminated in sequence. This system can be visualized
using the EPANET GUI, as shown in Figure 7.5. This figure shows that node 19 and the
connected pipes, 20 and A4O22, are eliminated because of the negative pressure. Node A1J22
and the connected pipes, A2O22, A3O22, and A1L22, are also eliminated.
72
Table 7.3 Damaged System at Time 0 (Damage_System_Time01.inp)
[TITLE]
[JUNCTIONS]
A1J22 160
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
200.000000
17
100.000000
19
200.000000
[RESERVOIRS]
A1R22 130
A2R22 130
A4R22 190
A5R22 190
A1R12 100
A2R12 100
A3R22 160
1
450.000000
[TANKS]
7
450.000000
[PIPES]
A1O22
A3O22
A4O22
A1O12
A2O12
A2O22
A1L22
10
16
18
20
4
6
8
17
A1J22
19
9
11
A1J22
A1J22
7
13
13
15
3
3
5
120.000000
A1R22
A4R22
A5R22
A1R12
A2R12
A2R22
A3R22
9
15
17
19
5
9
11
0.000000
914.4
914.4
304.8
1524
1524
914.4
0.5
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
120.000000
12
12
12
12
12
12
2
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
30.000000
100
100
100
100
100
100
1e+006
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
[PUMPS]
2
1
[VALVES]
14
9
3
POWER
13
4.000000
10.000000
PRV
100.000000
Table 7.3 Continued
[DEMANDS]
3
100.000000
5
100.000000
9
100.000000
73
0.000000
0.000000
1
1
1
1
1
1
1
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
CV
CV
CV
CV
CV
CV
Table 7.3 Continued
11
13
15
17
19
100.000000
100.000000
100.000000
100.000000
100.000000
[CURVES]
[PATTERNS]
PATN1 1.000000
PATN1 1.000000
1.000000
1.000000
1.000000
1.000000
1.000000
1.000000
[STATUS]
A3O22 Closed
[CONTROLS]
[SOURCES]
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.00000011
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
[COORDINATES]
A1R22 284.468 163.435
A2R22 284.468 158.983
A4R22 284.468 136.723
A5R22 284.468 132.27
A1R12 211.622 154.711
A2R12 211.625 150.324
A1J22 280.016 147.853
A3R22 284.468 147.853
1
140.726688
174.581772
74
1.000000
1.000000
1.000000
Table 7.3 Continued
3
5
7
9
11
13
15
17
19
169.667221
169.576993
207.220708
207.220708
207.252760
241.998111
241.998111
280.016223
280.016223
[VERTICES]
A1O22 280.016
A3O22 280.016
A4O22 280.016
A1O12 207.235
A2O12 207.238
A2O22 280.016
174.431972
130.595466
199.588372
174.450158
130.579090
174.517132
129.944774
174.565669
130.044299
163.435
136.723
132.27
154.708
150.321
158.983
[End]
Figure 7.4 Damaged System at Time 0
75
Figure 7.5 Simulation Results at Time 0
The detailed hydraulic simulation results associated with each type of component,
including junctions, tanks, pipes, pumps, and valves, are shown in Tables 7.4 to 7.8. In these
tables, only the results for the components in the original system are listed such that these data
files can be linked into GIS for map presentations. The results of the eliminated components due
to negative pressures or connectivity problems are set to 0. Figure 7.6 shows the simulation
results in a GIS map. The GIS shapefiles of junctions, tanks, pipes, pumps, and valves are
directly exported from the H2ONET software. By linking the simulation results for each type of
the physical component with the corresponding GIS shapefile, it is possible to visualize the
unsatisfied demands and the pipes unable to transport water.
76
Table 7.4 Junction Results at Time 0
Node_ID
3
5
9
11
13
15
17
19
Demand_gpm
100
100
100
100
100
100
100
0
Head_ft
454.92
307.92
278.8
168.05
278.8
278.62
163.86
0
Pressure_psi
153.79
90.09
77.47
29.49
77.47
34.07
27.67
0
Table 7.5 Tank Results at Time 0
Tank_ID
1
7
Demand_gpm
-8036.31
-5459.8
Head_ft
450
570
Pressure_psi
0
52
Table 7.6 Pipe Results at Time 0
Pipe_ID
10
12
16
18
20
22
4
6
8
Flow_gpm
5459.8
0
100
3305.09
0
0
3774.68
4161.63
3674.68
Velocity_fps
15.49
0
0.28
9.38
0
0
10.71
11.81
10.42
Headloss_/1000ft
95.54
0
0.06
37.71
0
0
48.23
57.78
45.89
Table 7.7 Pump Results at Time 0
Pump_ID
2
Flow_gpm
8036.31
Velocity_fps
0
Headloss_/1000ft
-4.92
Table 7.8 Valve Results at Time 0
Valve_ID
14
Flow_gpm
3505.09
Velocity_fps
9.94
Headloss_/1000ft
0
Table 7.9 Serviceability at Time 0
Node_ID
3
5
9
11
13
15
17
19
Demand
1
Node_Serviceability
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
0
1
1
1
1
1
1
1
0
0.875
0.875
Sum
77
0
0.25 0.5
1
km
Figure 7.6 GIS Map for GIRAFFE Simulation Results at Time 0
7.3.3.2
Outputs at Time 24
The Damage_System_Time241.inp stores the system definition information at a time 24
hours after updating the tank water level. The Modified_System_Time241.inp stores the system
definition information after GIRAFFE simulation of the Damage_System_Time241.inp. The
final simulation results at time 24 can be visualized using the EPANET GUI as shown in Figure
7.7. This figure shows that tank 7 is depleted after 24 hours of running and therefore, there is no
water flowing from this tank. All water flow in this network is supplied by reservoir 1. After the
depletion of tank 7, negative pressure occurred at node15 and thus this node and the connected
pipe 16 were eliminated. The system serviceability dropped from 0.875 to 0.75 due to the
unsatisfied demand at node 16. The simulation results associated with each component, and the
serviceability of each demand node and the entire system are shown in Tables 7.10 to 7.15.
These simulation results can also be linked into a GIS.
78
Figure 7.7 Simulation Results at Time 24
Table 7.10 Junction Results at Time 24
Node_ID
3
5
9
11
13
15
17
19
Demand_gpm
100
100
100
100
100
0
100
0
Head_ft
454.27
307.53
168.69
167.92
168.69
0
138.36
0
Pressure_psi
153.5
89.92
29.76
29.43
29.76
0
16.62
0
Table 7.11 Tank Results at Time 24
Tank_ID
1
7
Demand_gpm
-9273.55
0
Head_ft
450
450
79
Pressure_psi
0
0
Table 7.12 Pipe Results at Time 24
Pipe_ID
10
12
16
18
20
22
4
6
8
Flow_gpm
0
0
0
1609.71
0
0
3770.99
5402.57
3670.99
Velocity_fps
0
0
0
4.57
0
0
10.7
15.33
10.41
Headloss_/1000ft
0
0
0
9.95
0
0
48.14
93.69
45.8
Table 7.13 Pump Results at Time 24
Pump_ID
2
Flow_gpm
9273.55
Velocity_fps
0
Headloss_/1000ft
-4.27
Table 7.14 Valve Results at Time 24
Valve_ID
14
Flow_gpm
1709.71
Velocity_fps
4.85
Headloss_/1000ft
0
Table 7.15 Serviceability at Time 24
Node_ID
3
5
9
11
13
15
17
19
Demand
1
Node_Serviceability
100
100
100
100
100
100
100
100
100
100
100
100
100
0
100
0
1
1
1
1
1
0
1
0
Sum
7.4
0.75
0.75
Monte Carlo with Fixed Simulation Runs
Figure 7.8 shows the GIRAFFE GUI window with inputs for the Monte Carlo simulation
with fixed simulation times. The same hydraulic network as shown in last section was analyzed
by GIRAFFE.
Ten Monte Carlo simulations were performed.
The earthquake demand
associated with the distribution network damage is simulated using the 90% confidence level
prediction. The input file for pipe damage generation, rr.inp, is shown in Table 7.16. It is
assumed that each pipe has a RR = 1 repair/km in this example.
80
The input file,
Node_Pressure.inp, for earthquake demand simulation is shown in Table 7.17. It is assumed that
the distribution network has a RR =1 repair/km around each demand node.
Figure 7.8 Inputs for Monte Carlo Simulation with Fixed Simulation Runs
It is further assumed that the network is divided into two pressure zones, one upstream of
pressure reducing valve 14, including junctions 3, 5, 9, and 11, and the other downstream of
pressure reducing valve 14, including junctions, 13, 15, 17, and 19. The mean pressure of each
pressure zone is calculated by averaging the pressures at the junctions inside the pressure zone
for the undamaged system and then the mean pressure is assigned to each demand node inside
the pressure zone. The pressure at each junction for the undamaged system is shown in Figure
7.2.
GIRAFFE analyzes the network following to the same procedures described in Section
7.3.2 for ten simulation runs.
GIRAFFE saves the damaged system definition file,
Damage_Info_Dert*.inp, and the component results for each Monte Carlo simulation run. The
81
files associated with each simulation run are bundled in separate folders and saved with a similar
naming convention as that used in the deterministic simulation. The damaged system and
modified system files are appended with a number indicating which simulation run they are
associated with, e.g. Damage_System_Time09.inp is the damaged system file at time 0 for
simulation run 9, and Modified_System_Time245.inp is the modified system file at time 24 for
simulation run 5. Table 7.18 shows the damaged system for the last Monte Carlo simulation at
time 0. Comparing Tables 7.3 and 7.18 shows that the demand in Table 7.18 is different. The
demands at nodes 3, 5, 9, and 11, are changed from 100 gpm to 921 gpm and the demands at
nodes 13, 15, 17, and 19, are changed from 100 gpm to 422 gpm. The increased demands are
associated with water loss from damage to distribution networks around the demand nodes.
Table 7.16 Pipe Damage Input File for Monte Carlo Simulation with Fixed Simulation Runs
(rr.inp)
PipeID
10
12
16
18
20
22
4
6
8
Length_km
RR
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Material
CI
CI
DI
DI
CON
CON
RV
RV
STL
Table 7.17 Input File for Simulating Earthquake Demand for Monte Carlo Simulation with
Fixed Simulation Runs (Node_Pressure.inp)
ID
3
5
9
11
13
15
17
19
G_RR
1
1
1
1
1
1
1
1
Ave_PRESSURE
202
202
202
202
78
78
78
78
The increased demands are calculated using Eqn. 5.3 and the appropriate values for RR and MP.
Because the mean pressure of nodes 3, 5, 9, and 11 is much higher than that of nodes 13, 15, 17
and 19, the post-earthquake demands at nodes 3, 5, 9, and 11 are much higher than that at nodes,
82
13, 15, 17 and 19. GIRAFFE reports the system serviceability at two time points, times 0 and 24,
in files Serviceability0.out and Serviceability24.out, respectively.
These files are shown in
Tables 7.19 and 7.20. These tables show that the system serviceability is reported in a matrix
format. For each Monte Carlo simulation, the serviceability is reported for each demand node
and for the entire system. The mean of the nodal and system serviceability for all Monte Carlo
simulations is also calculated and reported.
Table 7.18 Damaged System for the Last Run of Monte Carlo Simulation at Time 0.
[TITLE]
[JUNCTIONS]
3
100.00000
5
100.00000
9
100.00000
11
100.00000
13
100.00000
15
200.00000
17
100.00000
19
200.00000
A1J6 100.00000
A1J10 445.01920
A1J12 100.00000
A2J12 100.00000
[RESERVOIRS]
1
450.00000
A1R6 100.00000
A1R10 445.01920
A1R12 100.00000
A1R18 100.00000
A2R12 100.00000
A2R18 100.00000
[TANKS]
7
450.000000
120.000000
0.000000
83
120.000000
30.000000
0.000000
Table 7.18 Continued
[PIPES]
4
3
8
5
16
13
20
15
22
17
A106 3
A1L6 A1J6
A206 A1J6
A1010 7
A1012 9
A1018 13
A1L10 A1J10
A1L12 A1J12
A2010 A1J10
A2012 A1J12
A2018 17
A2L12 A2J12
A3012 A2J12
5
11
15
19
19
A1J6
A1R6
9
A1J10
A1J12
A1R18
A1R10
A1R12
9
A2J12
A2R18
A2R12
11
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
963.56244
0.50000 1.58533
2084.43750
43.37563
2369.45337
2387.31592
0.50000 1.58533
0.50000 1.20000
3004.62427
646.33081
660.68408
0.50000 1.58533
32.21569
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
1000000.00000
12.00000
12.00000
12.00000
12.00000
1000000.00000
1000000.00000
12.00000
12.00000
12.00000
1000000.00000
12.00000
3
POWER 10.000000
13
12.000000
PRV
1.000000
1.000000
1.000000
1.000000
100.00000
100.00000
100.00000
100.00000
100.00000
100.00000
1.00000 CV
100.00000
100.00000
100.00000
100.00000
1.00000 CV
1.00000 CV
100.00000
100.00000
100.00000
1.00000 CV
100.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
1.00000 CV
0.00000
0.00000
1.00000 CV
0.00000
[PUMPS]
2
1
[VALVES]
14
9
100.000000
0.000000
[DEMANDS]
3
921.01996
5
921.01996
9
921.01996
11
921.01996
13
422.53998
15
422.53998
17
422.53998
19
422.53998
[CURVES]
[PATTERNS]
PATN1 1.000000
PATN1 1.000000
[STATUS]
[CONTROLS]
[SOURCES]
84
1.000000
1.000000
1.000000
1.000000
1.000000
Table 7.18 Continued
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.000000
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
[COORDINATES]
1
140.7267
3
169.6672
5
169.5770
7
207.2207
9
207.2207
11
207.2528
13
241.9981
15
241.9981
17
280.0162
19
280.0162
A1J6 181.5390
A1R6 181.5372
A1J10 207.2207
A1J12 207.2456
A1R10 207.2207
A1R12 211.6327
A1R18 269.8697
A2J12 207.2524
A2R12 211.6395
A2R18 273.6715
174.5818
174.4320
130.5955
199.5884
174.4502
130.5791
174.5171
129.9448
174.5657
130.0443
174.4377
181.9484
199.2306
140.3457
204.2583
140.3521
178.3545
131.0428
131.0492
178.3594
[VERTICES]
A1018 269.8745
A2018 273.6763
174.5527
174.5576
[END]
85
Table 7.19 Serviceability of Monte Carlo Simulation with Fixed Simulation Times at Time 0
(Serviceability0.out)
Node_ID Demand
1
2
3
4
5
6
7
8
9
10
Node_Serviceability
3
5
9
11
13
15
17
19
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
0
0
1
0.9
1
1
1
0.3
0.7
0.2
0.5
0.75
1
0.625 1
0.75
0.75
0.75
0.75
100
100
100
100
100
100
100
100
Sum
0.75
0.7625
Table 7.20 Serviceability of Monte Carlo Simulation with Fixed Simulation Times at Time 24
(Serviceability24.out)
Node_ID Demand
1
2
3
4
5
6
7
8
9
10
Node_Serviceability
3
5
9
11
13
15
17
19
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
0
0
1
0.9
1
1
1
0.2
0.6
0.1
0.5
0.75
1
0.625 0.875 0.75 0.75
0.75
0.625 0.625
100
100
100
100
100
100
100
100
Sum
7.5
0.725
Monte Carlo with Flexible Simulation Runs
Figure 7.9 shows the GIRAFFE GUI window with inputs for the Monte Carlo with
flexible simulation runs. The same hydraulic network as shown in the previous section was
analyzed by GIRAFFE.
The earthquake demand associated with the distribution network
damage is simulated using the 90% confidence level prediction. The input files, rr.inp, for pipe
damage generation and, Node_Pressure.inp, for earthquake demand simulation are the same as
those shown in Tables 7.16 and 7.17, respectively.
86
Figure 7.9 Inputs for Monte Carlo Simulation with Flexible Simulation Times
Figure 7.10 Pop-Up Window with Results
The system serviceability is reported in Tables 7.21 and 7.22 for times 0 and 24,
respectively.
These tables show that 20 Monte Carlo simulations were performed.
The
calculations of the mean and COV of the system serviceability for the first 15 and the total 20
simulations show that the difference of the mean and COV from the two sets of simulations is
less than 0.02. Thus the program terminated after 20 simulations using its self-termination
87
algorithm. As shown in Figure 7.10, the number of simulations and associated convergence
criteria appears in a pop-up window upon completion of the GIRAFFE run. By comparing
Tables 7.19 with 7.21, and Table 7.20 with 7.22, it is found that the system serviceability of the
first ten simulations is same for the two simulation options.
simulations used the same seed to generate random numbers.
88
This is because these two
Table 7.21 Serviceability of Monte Carlo Simulation with Flexible Simulation Runs at Time 0
Node_ID Demand 1
3
5
9
11
13
15
17
19
100
100
100
100
100
100
100
100
Sum
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
0
0
100
0
100
0
100
0
100
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
0.75 0.75 0.75 0.75 0.75 0.5
0.5
0.625 0.75 0.75
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
0.5
0.75 1
0.625 1
0.625 0.5
0.625 0.75 0.75
Node_Serviceability
1
0.8
1
0.95
1
0.15
0.6
0.1
0.7
Table 7.22 Serviceability of Monte Carlo Simulation with Flexible Simulation Runs at Time 24
Node_ID Demand 1
2
3
5
9
11
13
15
17
19
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
0.5
0.75 1
Sum
100
100
100
100
100
100
100
100
3
4
5
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
0
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
0
100
0
100
0
100
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
0
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
0.625 0.875 0.75 0.75 0.75 0.625 0.625 0.5
87
0.5
0.625 0.625 0.75 0.625 0.5
0.625 0.75 0.75
Node_Serviceability
1
0.75
1
0.95
1
0.1
0.55
0.05
0.675
REFERENCES
American Lifelines Alliance (2005). Seismic Guidelines for Water
Pipelines. http://www.americanlifelinesalliance.org.
Armando, L. (1987). Handbook of Hydraulic Engineering. John Wiley & Sons, NY.
Crane Company (1972). Flow of Fluid through Valves and Fittings. Crane Company, New York,
NY.
Eguchi, R.T. (1982). “Earthquake Performance of Water Supply Components during the 1971
San Fernando Earthquake.” Technical Report 82-1396-2a. J.H. Wiggins Company,
Redondo Beach, CA.
Eguchi, R.T. and Chung, R.M. (1995). “Performance of Lifelines during the January 17, 1994
Northridge Earthquake.” Lifeline Earthquake Engineering, TCLEE Monograph No. 6.
O’Rourke, M.J. ed., ASCE, Reston, VA, 120-127.
Hall, J.F. (1995). “Northridge Earthquake of January 17, 1994, Reconnaissance Report.”
Earthquake Spectra, EERI, Oakland, CA, April.
Idelchik, I.E. (1999). Handbook of Hydraulic Resistance. 3rd Edition, Begell House, New York,
NY.
Jeon, S.-S. (2002). “Earthquake Performance of Pipelines and Residential Buildings and
Rehabilitation with Cast-in-Place Pipe Lining Systems.” Ph.D. Dissertation, School of
Civil & Environmental Engineering, Cornell University, Ithaca, NY.
Jeon, S.-S. and O’Rourke, T.D. (2005). “Northridge Earthquake Effects on Pipelines and
Residential Buildings.” Bulletin of the Seismological Society of America, Vol. 95, No.1,
294-318.
Jeppson, R.W. (1976). Analysis of Flow in Pipe Network. Ann Arbor Science Publisher, Ann
Arbor, MI.
Kershaw D. (1978). “The Incomplete Choleski-Conjugate Gradient Method for the Iterative
Solution of Systems of Linear Equations.” Journal of Computation Physics, Vol. 26, 4365.
Lawson, A. C. (1908). “The California Earthquake of April 18, 1906.” Report of the State
Earthquake Investigation Commission. Carnegie Institute of Washington, No. 87, Vol. I,
451p.
88
Los Angeles Department of Water and Power (LADWP) (2002). H2ONET Hydraulic Model of
the LADWP Water Supply System. Los Angeles, CA.
Lund, L. and Cooper, T. (1995). “Water System.” Northridge Earthquake: Lifeline Performance
and Post-Earthquake Response, Technical Council on Lifeline Earthquake Engineering
Monograph No. 8, Schiff, A.J., ed., ASCE, New York, NY, 96-131.
Manson, M. (1908). Reports on an Auxiliary Water Supply System for Fire Protection for San
Francisco, California. Board of Public Works, San Francisco, CA.
Markov, I., Grigoriu, M.D., and O’Rourke, T.D. (1994). “An Evaluation of Seismic
Serviceability Water Supply Networks with Application to the San Francisco Auxiliary
Water Supply System.” Technical Report NCEER-94-0001, National Center for
Earthquake Engineering Research, Buffalo, NY.
Miller, D.S. (1978). Internal Flow System. BHRA Fluid Engineering, Bedford, UK.
MIKENET (2005), http://www.dhisoftware.com/mikenet/
Moody, L.E. (1944). “Friction Factors of Pipe Flow.” Transactions of the American Society of
Mechanical Engineers, ASME, Vol. 16, New York, NY
MWH Soft, Inc. (1999). H2ONET Users Guide. Pasadena, CA.
O’Rourke, T.D. (2005). Personal Communications.
O’Rourke, T.D., Stewart, H.E., and Jeon, S-S. (2001). “Geotechnical Aspect of Lifeline
Engineering.” Geotechnical Engineering, ICE, Vol. 149, No. 1, 13-26.
Puchovsky, M.T. (1999). Automatic Sprinkler Systems Handbooks, National Fire Protection
Association (NFPA).
Rossman, L.A. (2000). EPANET 2 Users Manual. National Risk Management Research
Laboratory, Office of Research and Development, U.S. Environmental Protection
Agency, Cincinnati, OH.
Schussler, H. (1906). The Water Supply of San Francisco, California. Martin B. Brown Press,
New York, NY.
Sheldon, M.R., (2000), Introduction to Probability and Statistics for Engineers and Scientists.
2nd Edition, Harcourt and Technology Company, San Diego, CA.
Shi, P. (2006). “Seismic Response Modeling of Water Supply Systems.” Ph.D. Dissertation,
School of Civil & Environmental Engineering, Cornell University, Ithaca, NY.
89
Steinbrugge, K.V., Schader, E.E., Bigglestone, H.C., and Weers, C.A. (1971). San Fernando
Earthquake, February, 9, 1971. Pacific Fire Rating Bureau.
Swamee, P.K. and Jian, A.K. (1976). “Explicit Equations for Pipe Flow Problems.” Journal of
Hydraulic Engineering, ASCE, Vol. 102, No. 5, 657p.
Todini, E. and Pilati, S. (1987). “A Gradient Method for the Analysis of Pipe Networks.”
Proceedings of the International Conference on Computer Application for Water Supply
and Distribution. Leicester, UK.
Walski, T.M., Chase, D.V., and Savic, D.A. (2001). Water Distribution Modeling. Hastad Press,
CT, USA.
Wang, Yu. (2006). “Seismic Performance Evaluation of Water Supply Systems.” Ph.D.
Dissertation, School of Civil & Environmental Engineering, Cornell University, Ithaca,
NY.
WaterCAD (2005), http://www.hasetad.com/software/watercad/
90
APPENDIX A
GIRAFFE QUICK START TUTORIAL
A.1
INTRODUCTION
This appendix provides a quick start tutorial on how to use GIRAFFE. The quick start
tutorial will help first-time users become familiar with the core set of GIRAFFE features and
should be used as a launching point to a more comprehensive understanding of GIRAFFE. Users
are expected to have some knowledge of how to use the hydraulic network analysis software
packages H2ONET and EPANET before starting to use GIRAFFE.
Users can obtain this
knowledge from the H2ONET Users Manual (MWH Soft, Inc., 1999) and the EPANET Users
Manual (Rossman, 2000).
A.2
INSTALLING SOFTWARE
If the Microsoft Install Wizard does not automatically start upon inserting the GIRAFFE
installation CD into your computer’s CD drive, open the CD folder in Windows Explorer and
double click on Install_GiraffeV4.1.exe. This will automatically install GIRAFFE Version 4.1,
EPANET 2.0, Microsoft.NET Version 2.1, and a Matlab component. The default folder for
GIRAFFE is C:\Program Files\Cornell University\Giraffe. After installing GIRAFFE, select
this item from the Start menu and then double click on the Giraffe.exe icon in the GIRAFFE
program folder to launch the program. The current version of GIRAFFE operates via a graphical
user interface (GUI). When the user clicks the “Generate Pipe Repair Rate and Mean Pressure
Files” button (see Appendix C, Section 2) in the Monte Carlo Fixed or Flexible simulations in
GIRAFFE interface, the stochastic damage tool will be installed in C:\Program Files\Cornell
University\Appendix B.
For Windows XP Professional x64 Edition, the GIRAFFE application will be installed in
the 32-bit directory, C:\Program Files (x86)\Cornell University\Giraffe, and the stochastic
damage tool (see Appendix C, Section 2) will be installed in the 64-bit directory, C:\Program
Files\Cornell University\Appendix B, when the user clicks the “Generate Pipe Repair Rate and
Mean Pressure Files” button in the GIRAFFE interface.
1
A.3
EXAMPLE NETWORK
Since the LADWP hydraulic network model works with the H2ONET software, this
example also uses H2ONET to construct the hydraulic network model. Detailed procedures for
constructing a hydraulic network model using H2ONET can be found in H2ONET Users Manual
(MWH Soft, Inc., 1999). Figure A.1 shows the hydraulic network model constructed using
H2ONET with component identifications (IDs) indicated as black characters and nodal demands
indicated as red numbers.
The network contains 8 junctions with IDs 3, 5, 9, 11, 13, 15, 17, and 19, respectively.
All the junctions have an elevation of 100 ft, except junctions 15 and 19, which have an
elevation of 200 ft. The network contains 9 pipes with IDs, 4, 6, 8, 10, 12, 16, 18, 20, and 22,
respectively. All pipes have a length of 3048 ft, diameter of 12 in., and roughness coefficient of
100. There is one reservoir with ID 1, one tank with ID 7, one pump with ID 2, and one PRV
with ID 14 in the network. The reservoir has a hydraulic grade of 450 ft. The tank is a cylinder
tank with a diameter of 30 ft, maximum water level of 120 ft, minimum water level of 0 ft, and
bottom elevation of 450 ft from the datum. The tank is assumed to be full at the beginning of
simulation time. The pump is a constant power pump, which supplies a power of 10 kw-hours.
The valve is a pressure reducing valve with a pressure setting of 100 psi. Eight demand nodes
are distributed around the network. Each demand node has a demand of 100 gpm.
Three simulations are performed to the network using GIRAFFE: deterministic, Monte
Carlo with a fixed number of simulation runs, and Monte Carlo with a flexible number of
simulation runs.
For a deterministic simulation, GIRAFFE adds damage to the network
deterministically and then performs a hydraulic analysis on the damaged network. For Monte
Carlo with fixed simulation runs, users specify the number of Monte Carlo simulations to be
performed. For Monte Carlo with flexible simulation runs, GIRAFFE determines how many
2
Figure A.1 Hydraulic Network Model Constructed by H2ONET
Monte Carlo simulations need to be performed to have statistically significant simulation results
using a built-in self-termination algorithm. The self-termination algorithm is explained in the
main text of the GIRAFFE Users Manual and Shi (2006). In each Monte Carlo simulation,
GIRAFFE damages the system probabilistically and then analyzes the damaged network.
A.4
DETERMINISTIC SIMULATIONS
Step 1: Export EPANET File
GIRAFFE works with the EPANET format system definition file, which can be exported
from H2ONET directly. To export the H2ONET model into EPANET format file:
3
Report Option Button
Run Manager Button
Figure A.2 Export H2ONET Model to EPANET Format File
• Click the Run Manager button at the bottom of the H2ONET GUI as shown in Figure
A.2. A Run Manager dialogue box will appear.
• Click the Report Option button in the Run Manager dialogue box. A Report Option
dialogue box will appear.
• Uncheck the No Node Report and No Link Report text boxes and then click the OK
buttons to close the Run Manager and Report Option dialogue boxes. If these two
boxes have already been unchecked, leave them unchecked and close the Run Manager
dialogue box and Report Option dialogue box.
• Go to the Exchange | EPANET v2.0 | Export menu in the top of H2ONET GUI, specify
the directory and name of the export file, and then click Export button. In this example,
the EPANET file is saved as Example_1.inp in the accompanying Users Manual CD.
4
Step 2: Check EPANET File
It is important to double check that the EPANET format file can be analyzed by the
EPANET engine. The EPANET format file, Example_1.inp, is shown in Table A.1. In general,
the EPANET file needs to include the following sections.
•
Section [TITLE]
•
Sections
defining physical
components
in
the hydraulic network,
including
[JUNCTIONS], [RESERVOIRS], [TANKS], [PIPES], [PUMPS], and [VALVES].
•
Sections defining operational components, including [DEMANDS], [CURVES],
[PATTERNS], [STATUS], and [CONTROLS]
•
Sections defining water quality simulation parameters, including [SOURCES],
[QUALITY], and [REACTIONS].
•
Sections defining simulation and report options, including [ENERGY], [OPTIONS], and
[REPORT]. Users need to make sure that the report option for node and link is “ALL”.
•
Sections defining the locations of network components, including [COORDINATES] and
[VERTICES].
•
Section [End]
It is possible that there are no records in some sections. In this case, users still need to
keep the title of that section. It is recommended that users load the EPANET file into EPANET
to verify that the file can be analyzed by EPANET. To load the file into EPANET and run the
file:
•
Go to File | Import | Network menu in the EPANET GUI as shown in Figure A.3.
•
Browse to the file, Example_1.inp, and click it to load it into EPANET.
•
Click Project | Run Analysis menu.
A message box will pop up to report the Run Status. If it reports “Run was successful” as
shown in Figure A.3, then EPANET could analyze the file and it can be analyzed further in
GIRAFFE. If the run was unsuccessful, EPANET will report error messages. Users need to
5
correct the errors in the .inp file following directions given by the error messages. Due to several
incompatibilities between EPANET and H20NET, the input file may have errors regarding
H20NET control features that are not supported by EPANET. Modifications should be made
within the H20NET model to rectify these errors, and then the new file can be exported as an
input file.
Note to LADWP:
Adjustments were made in the LADWP-Cornell model to eliminate approximately 400 negative
pressure nodes. This was accomplished by decreasing the node elevations such that the nodal
pressures were increased to 5psi. Most elevation adjustments were less than 10 ft. From a
comparison standpoint, it is recommended you make these adjustments for consistency between
models. After performing a simulation within H20Net, look at the output pressures for all nodes,
and copy all negative pressure nodes into excel (copy NodeID, Output Elevation, Output Grade
and Output Pressure for each). For each negative pressure node create a column called “New
Elevation” and perform the following calculation to find what the new elevation would need to
be to create a nodal pressure of 5psi:
Output Elevation (ft) – [5 – Ouptut Pressure (psi)] * 2.3067 ft/psi
Within the H20Net model, go to “Edit Database Tables” and replace node elevations with these
new elevation values. Rerun the simulation and double check that there are no longer any
negative pressure nodes.
Table A.0 lists modifications that were made to various control features that either contained
typographical errors or were not compatible with EPANET. All pipe flows and node pressures
were checked after these modifications and results were either identical to or within 1% of the
original values.
6
Table A.0. Control Modifications
FCV
CC7090
TCV
Control
CC6420
CC7110
CC7230
CC7210
CC7240
Curves
35
Control
CC7380
CC7290
GH7250
CC7300
Curves
10000
GH7310
CC7330
Curves
4.2209091
GH7320
CC7340
Curves
H6170
CC7350
Curves
10000
HP6060
Control
GH7020
MW6140
Control
GH7040
SY6330
261.818 Disable Control
GH7370
VF6270
Control
HH6200
VF6280
Control
HH6210
VF6290
Control
HH6280
Curves
0.25
MW6070
Curves
0.25 Setting 556
VF6390
MW6410
Control+Curves
VF6400
MW6420
Curves
Curves
656.92 Setting 0
Curves
888 Setting 0
WS7100
Curves
597.2 Setting 0
WS7120
Curves
WS7150
Curves
VF6380
VF6830
Control
MW6430
VF6840
Control
VF6580
VF6910
VF6730
VF6930
VF6850
VF6940
VF7102
VF6960
Control
VF6970
WS6960
63.360108
0.25
VF7112
VF7122
Control
WS7210
Curves
VF7132
0.25
597.2 Setting 0
VF101
delete last row
WS7170
Curves
0.25
MW17
delete first row
WS7180
Curves
0.25
Link
EH656
initial status
WS7190
Curves
WS7250
Curves
597.2 Setting 0
WS7260
Curves
597.2 Setting 0
GH824
Tanks
Fixed Head Reservoir
CC4220
Curves: starting from 0
MW4100
Curves: starting from 0
VF5690
VF4180
HP4030
Curves: starting from 0
VF4010
HP4060
Curves: starting from 0
Misc.
Remove Secondary demand pattern at HH775
7
0.25
Table A.1 EPANET Format File Exported from H2ONET
[TITLE]
[JUNCTIONS]
//ID
Elevation(ft)
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
200.000000
17
100.000000
19
200.000000
Headline added by the author to help users
understand the meaning of the parameters. Must
not have headlines when loading the file into
EPANET or analyzing it using GIRAFFE
Pattern//
Headline
[RESERVOIRS]
//ID
Head(ft)//
1
450.000000
Headline
[TANKS]
//ID
Elev(ft)
InitialLevel(ft)
7
450.000000 120.000000
[PIPES]
//ID FromNode
10
7
12
9
16
13
18
13
20
15
22
17
4
3
6
3
8
5
[PUMPS]
//ID FromNode
2
1
[VALVES]
//ID
FromNode
14
9
MinLevel(ft)
0.000000
MaxLevel(ft)
120.000000
Dia.(ft)
MinVol(ft3)
30.000000 0.000000
VolCurve//
Headline
ToNode
9
11
15
17
19
19
5
9
11
Length(ft)
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
ToNode
3
Diameter(in)
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
Parameter(kw-hr)//
POWER 10.000000
Roughness
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
MinorLoss
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
CheckValve//
Headline
Headline
ToNode
13
[DEMANDS]
//ID
Demand(gpm)//
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
100.000000
17
100.000000
19
100.000000
Diameter(in)
4.000000
Type
PRV
Headline
8
Setting(psi)
100.000000
MinorLoss//
0.000000
Table A.1 Continued
[CURVES]
[PATTERNS]
PATN1 1.000000
PATN1 1.000000
1.000000
1.000000
1.000000
1.000000
1.000000
1.000000
[STATUS]
[CONTROLS]
[SOURCES]
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.000000
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
[COORDINATES]
//ID
x(ft)
1
140.726688
3
169.667221
5
169.576993
7
207.220708
9
207.220708
11
207.252760
13
241.998111
15
241.998111
17
280.016223
19
280.016223
Must use
ALL
Headline
y(ft)//
174.581772
174.431972
130.595466
199.588372
174.450158
130.579090
174.517132
129.944774
174.565669
130.044299
[VERTICES]
[End]
9
1.000000
1.000000
1.000000
Figure A.3 Loading EPANET File into EPANET
10
Step 3: Construct Pipe Damage Input File
For a deterministic simulation, users need to specify the location of each pipe break and
leak, as well as the opening area of each leak. In this example, it is assumed that damage occurs
in two pipes, 12 and 22, as illustrated in Figure A.4. One break occurs at point D in pipe 12.
The location of the damage is defined by a length ratio, which is the ratio of the pipe length,
measured from the pipe upstream node to the damage location, to the original pipe length. It is
assumed that point D is at the middle of pipe 12, as such the length ratio for the break in pipe 12
is 0.5. Three locations of damage occurs in pipe 22: one break at point A with a length ratio of
0.3, one leak at point B with a length ratio of 0.6 and leak diameter of 2 inch, and one break at
point C with a length ratio of 0.9.
The pipe damage input file is a text file, which can be constructed by entering the
parameters by hand, by using the GIRAFFE GUI input window, or by opening a Manifold
System project and selecting the desired pipes. The following section describes how to create the
pipe damage input file using each of the three methods. If creating the file by hand, users can use
Microsoft Word, Excel, or Notepad to construct the file and save it as a tab-delimited text file
with the extension .inp. Users may also create a pipe damage file via the GIRAFFE GUI input
window for a deterministic simulation. All three methods for creating the file are discussed in
this section. In this example, the pipe damage input file is saved as Pipe_damage.inp, which is
shown in Table A.2.
The input file consists of two blocks with one storing pipe break
information and the other storing pipe leak information.
Constructing Pipe Damage File using Microsoft Word, Excel or Notepad
The block storing pipe break information starts with the line [Pipe_Break_Information].
Users need to copy this exact line into their input file and not leave any space before
[Pipe_Break_Information], otherwise the program will not run correctly. The second line is a
headline describing the type of values in each column in the pipe break records that follow. It is
recommended that users copy the headline into their input file. The headline terms in the pipe
break records are explained in Table A.3. Following the headline and a blank line are the three
11
records for the breaks at points A and C in pipe 22, and the break at point D in pipe 12,
respectively.
0.3L22
Break
A
0.5L12
D
Break
L12
0.3L22
L22
Leak
B
0.3L22
Break
C
Figure A.4 Illustration for Pipe Damage
Table A.2 Pipe Damage Input File for Deterministic Simulation
[Pipe_Break_Information]
PipeID
PreRatio
BreakRatio
22
22
12
0
0.6
0
2
BreakNo
LeakNo
PreIndex
3
3
1
1
2
1
0
1
0
0
0
0
0.3
0.9
0.5
[Pipe_Leak_Information]
PipeID
LeakD
PreRatio
22
RepairNo
0.3
LeakRatio
RepairNo
0.6
3
BreakNo
LeakNo
PreIndex
1
1
1
The block storing pipe leak information starts with the line [Pipe_Leak_Information].
Users need to copy this exact line into their input file and not leave any space before
[Pipe_Leak_Information], otherwise the program will not run correctly. The next line is a
headline describing the type of values in each column in the pipe leak records that follow. It is
recommended that users copy the headline into their input file. The headline terms in the pipe
12
leak records are explained in Table A.4. Following the headline and a blank line is the record for
the leak at point B in pipe 22.
Table A.3 Description of Columns in Pipe Break Section
Name
Type
Explanation
PipeID
char
The ID of the pipe which users want to break.
The length ratio of the previous location of pipe damage, either break or
PreRatio
float
leak, in the same pipeline. If the current break is the first location of
damage in the pipeline, then the PreRatio is set to 0.
BreakRatio
float
The length ratio of the location of the current pipe break.
The total number of locations of pipe damage, including breaks and
leaks, in the pipeline. For example, there are three locations of damage
RepairNo
int
in pipe 22, including two breaks and one leak. As such, the RepairNo is
3 for all the records associated with pipe 22. There is one location of
damage, which is a break, in pipe 12. As such, the RepairNo is 1 for the
break record associated with pipe 12.
The number of locations of breaks in the upstream of the current
location of pipe break in the same pipeline. The current location of pipe
break is counted. For example, for the first pipe break record, which is
BreakNo
int
for the break at point A in pipe 22, the BreakNo is 1 because it is the
first break in pipe 22. For the second pipe break record, which is for the
break at point C in pipe 22, the BreakNo is 2 because it is the second
break in pipe 22.
The number of locations of leaks in the upstream of the current location
of pipe break in the same pipeline. For example, for the first pipe break
record, which is for the break at point A in pipe 22, the LeakNo is 0
LeakNo
int
because there is no leak upstream of point A in pipe 22. For the second
pipe break record, which is for the break at point C in pipe 22, the
LeakNo is 1 because there is 1 leak at point B, which is located
upstream of point C in pipe 22.
PreIndex
int
The type of the previous location of pipe damage immediately upstream
13
of the current break in the same pipeline: 0 for leak and 1 for break. If
the current break is the first location of pipe damage in the pipeline.
The PreIndex is set to 0.
Table A.4 Description of Columns in Pipe Leak Section
Name
Type
PipeID
char
LeakD
float
Explanation
The ID of the pipe which users want to add the leak. Maximum
length 30 characters
Equivalent orifice diameter of the leak in inches.
The length ratio of the previous location of pipe damage, either break
PreRatio
float
or leak, in the same pipeline. If the current leak is the first location of
damage in the pipeline, then the PreRatio is set to 0.
LeakRatio
float
RepairNo
int
The length ratio of the location of the current leak.
The total number of locations of pipe damage, including breaks and
leaks, in the pipeline.
The number of locations of breaks in the upstream of the current
location of pipe leak in the same pipeline. For example, for the leak
BreakNo
int
record in Table A.2, which is for the leak at point B in pipe 22, the
BreakNo is 1 because there is one break at point A, which is located
in the upstream of point B in pipe 22.
The number of locations of leaks in the upstream of the current
location of pipe leak in the same pipeline. The current location of
LeakNo
int
pipe leak is counted. For example, for the leak record in Table A.2,
which is for the leak at point B in pipe 22, the LeakNo is 1 because it
is the first leak in pipe 22.
The type of the previous location of pipe damage immediately
PreIndex
int
upstream of the current leak in the same pipeline: 0 for leak and 1 for
break. If the current leak is the first location of pipe damage in the
pipeline. The Preindex is set to 0.
14
Constructing a Pipe Damage File using GIRAFFE GUI
Users can be guided through the creation of a pipe damage file by the GIRAFFE GUI. To
create pipe damage in GIRAFFE, select “Deterministic” from the screen that appears when
GIRAFFE is first opened, or go to Simulations | Deterministic in the main GIRAFFE toolbar.
There are two alternatives for creating pipe damage in the deterministic GUI: using only
GIRAFFE to assign damage based on pipe IDs (“Create Pipe Damage”) or using Manifold
System to spatially assign damage (“Create Pipe Damage Using Manifold GIS”). Both of these
methods will be discussed in this section.
The first method, which only uses GIRAFFE to assign damage, is shown in Figure A.5.
After loading the system definition file (in this case, Example1.inp) and clicking on the “Create
Pipe Damage” button in the deterministic GUI, a pop-up window appears so that the user can
select a pipe from a drop down menu and enter the number of breaks and leaks associated with
that pipe. To create the same example pipe damage file used previously in this section, the user
should select Pipe ID 22 from the drop down menu. There are 2 breaks and 1 leak associated
with this pipe, so the user should enter 2 for “No. of Pipe Breaks” and 1 for “No. of Pipe Leaks”
and then click the “Add Damage” button. Figure A.6 shows the “Pipe Details” window that
opens, prompting the user to enter in the break ratio for the first pipe break. After entering 0.3
for the break ratio and selecting the “Save” button, the user is prompted to enter the break ratio
for the second pipe break. The user should enter 0.9 for the second break ratio and hit “Save”.
The next prompt, shown in Figure A.7, asks the user to enter the leak ratio and leak
diameter, in inches, for the pipe leak associated with pipe 22. Entering the values and hitting
“Save” will take the user back to the original “Create Pipe Damage” window (shown in Figure
A.5) where another pipe ID can be selected to repeat the process and add additional damage to
the system. Once all pipe breaks and leaks have been entered, the user simply closes the “Create
Pipe Damage” window by clicking on the X at the top of the window The pipe damage file is
automatically saved as Pipe_Damage_temp.inp in the GIRAFFE program folder once the
15
“Create Pipe Damage” window is closed. The newly created pipe damage file automatically
populates the “Pipe Damage File” box in the GIRAFFE GUI. This file can be used to view the
breaks and leaks entered via the GUI, but any changes to this text file will not be recognized by
the GIRAFFE engine when it performs the simulation because the Pipe_Damage_temp.inp file is
only a temporary file for viewing. If changes need to be made to values already entered, the user
must re-enter all of the pipe breaks and leaks via the GUI. (To avoid repeating the entire GUI
process when an entry mistake has occurred, the user may copy, rename and alter the
Pipe_Damage_temp.inp file and then select this new file from the “Browse” button by the “Pipe
Damage File” input box.)
Figure A.5 Creating a Pipe Damage File via the GIRAFFE GUI – Entering the number of pipe
breaks and leaks associated with a Pipe ID
16
Figure A.6 Creating a Pipe Damage File via the GIRAFFE GUI – Entering the break ratio
associated with each pipe break
Figure A.7 Creating a Pipe Damage File via the GIRAFFE GUI – Entering the leak ratio and
leak diameter associated with each pipe leak
The second method for assigning pipe damage uses the Manifold System application
(Figure A.8). Note that this example uses the entire LADWP water distribution system rather
than the small system example. This tool will open a Manifold project and allow the user to
select the pipes from a spatial representation of the pipe network. However, before clicking the
“Create Pipe Damage Using Manifold GIS”, the user must first add the tool to Manifold’s
custom controls. To do this, go to the Manifold Tools folder in the GIRAFFE program folder
17
and copy the contents of the folder (two folders, LADWP and Damage, and an .xml file). Then
paste the files in the Config folder hierarchy for Manifold (normally C:\Program
Files\Manifold System\Config).
After clicking on the “Create Pipe Damage Using Manifold GIS” button in the
deterministic GUI, a pop-up window appears asking the user to select a saved Manifold project
(in this example, Deterministic_damage.map) as shown in Figure A.9. After pressing OK, the
project will open and the pipes can be selected (Figure A.10). The saved project should contain
a shapefile representing all pipes in the system (epa_pipes.shp). It is also helpful to create ID
labels for the pipes and overlay them in a map as was done in the example map (right-click in the
Project pane and select Create | Labels). A sample Manifold project (Deterministic_damage.map)
has been included in the GIRAFFE program file in the folder Example_Files | Appendix A,
which includes the required files in the appropriate format. Before running the tool in Manifold,
the user should select the desired pipes, either specific pipes or a large section of pipes. Note that
the layer in which the pipes are being selected must be named Epa_pipes and it must contain a
column called [ID 2], which contains the unique identification numbers for each pipe in the
system (see Figure C.2). After the pipes have been selected, click on the pipe damage tool in the
toolbar
. If the toolbar is not visible, go to Tools | Add-Ins | Add-In Manager and check the
box next to “Create Pipe Damage.” Restart Manifold as directed.
Clicking the Create Pipe Damage tool will cause a pop-up window to appear, which asks
for the output file location (Figure A.11). The next window asks for the type and amount of
damage to each of the selected pipes (Figure A.12). After these data have been entered, the
window then asks for the break ratio or leak ratio and leak diameter for each incidence of
damage (Figures A.13(a) and (b)). When the last incidence of damage has been added, a pop-up
window will identify the output location specified in the beginning of the process and then
Manifold will close automatically. The pipe damage file automatically populates the “Pipe
Damage File” box in the GIRAFFE GUI.
18
Figure A.8. Creating a Pipe Damage File via the GIRAFFE-Manifold GUI
Figure A.9. Creating a Pipe Damage File via the GIRAFFE-Manifold GUI – Selecting the
Manifold Project.
19
Figure A.10 Creating a Pipe Damage File via the GIRAFFE-Manifold GUI – Selecting the Pipes.
Figure A.11 Creating a Pipe Damage File via the GIRAFFE-Manifold GUI – Using the Pipe
Damage Tool and Selecting the Output File Location.
20
Figure A.12 Creating a Pipe Damage File via the GIRAFFE-Manifold GUI – Assigning Damage
to Selected Pipes.
Figure A.13(a) Creating a Pipe Damage File via the GIRAFFE-Manifold GUI – Specifying
Amount of Damage to Selected Pipes.
21
Figure A.13(b) Creating a Pipe Damage File via the GIRAFFE-Manifold GUI – Specifying
Amount of Damage to Selected Pipes.
22
Step 4: Input Parameters in GIRAFFE GUI Window
Figure A.14 shows the GUI window with the required inputs for a deterministic
simulation. The meaning of each entry for a deterministic simulation is explained in Table A.5.
Figure A.14 GUI Window with Inputs for Deterministic Simulation
23
Table A.5 GIRAFFE Input Parameters for Deterministic Simulation
Name
Description
Name of the EPANET system definition file with the extension
System
Definition File
of .inp, .txt or .dat.. File name may have a maximum length of
80 characters.
Minimum
Pressure limit, in psi, below which GIRAFFE eliminates the
Pressure to
node and connected links from the system. Typically 0 for
Eliminate
negative pressure elimination.
Simulation Time
Total length of simulation time in hours to update tank water
levels. 0 for steady state simulation.
Simulation Time The time step in hours to update tank water levels. 1 for steady
Step
state simulation.
Pipe Damage File Name of input file for pipe damage generation. The file name
may have a maximum length of 80 characters.
Step 5: Perform Simulation
After GIRAFFE receives the inputs, it performs the deterministic simulation according to
the following procedures:
1) Damage the network and output the damaged system, Damage_System_Time0.inp.
2) Apply the EPANET engine to perform hydraulic network analysis to the damaged
system and an iterative approach to eliminate negative pressures. The elimination
process continues until no negative pressures exist in the network.
3) Output the system definition file, Modified_System_Time0.inp, and report the results
of each type of physical component in the files, JunctionResults_Time0.out,
TankResults_Time0.out,
PipeResults_Time0.out,
ValveResults_Time0.out.
24
PumpResults_Time0.out,
and
4) Calculate the system serviceability at time 0 and report the system serviceability in
the file, Serviceability0.out.
5) Read the TankResults_Time0.out, determine the outflow of each tank, and update the
tank water levels according to the initial tank water levels, tank cross-sectional areas,
tank outflows, and the time step. In this example, GIRAFFE updates the water level
of tank with ID 7 once after 24 hours of tank running.
6) Output the damaged system, Damage_System_Time24.inp.
7) Apply the EPANET engine to perform hydraulic network analysis to the system with
tank water level updated, and the iterative approach to eliminate negative pressures.
The elimination process continues until no negative pressures exist in the network.
8) Output the system definition file, Modified_System_Time24.inp, and report the
hydraulic simulation results of each type of physical component in the files,
JunctionResults_Time24.out,
TankResults_Time24.out,
PipeResults_Time24.out,
PumpResults_Time24.out, and ValveResults_Time24.out.
9) Calculate the system serviceability at time 24 and report the system serviceability in
the file, Serviceability24.out.
Step 6: View Simulation Results
After the GIRAFFE simulation, the result files can be viewed and checked.
The
simulation results are saved in the Giraffe_Output folder which is located in the same location
as the GIRAFFE application.
•
View damaged system at time 0.
The damaged system is saved in the file,
Damage_System_Time0.inp, as shown for the small system example in Table A.6. The
added components associated with pipeline damage are described in red text boxes. The
Damage_System_Time0.inp can be loaded into EPANET and can be visualized using the
EPANET GUI, as shown in Figure A.15. Users need to check if GIRAFFE adds the
pipeline damage correctly.
25
•
View simulation results at time 0. The hydraulic simulation results associated with
each type of component including junctions, tanks, pipes, pumps, and valves, are shown
in Tables A.7 to A.11. The system serviceability at time 0 is shown in Table A.12. The
simulation results can be visualized using the EPANET GUI as shown in Figure A.16 by
loading the Modified_System_Time0.inp into EPANET and running the simulation.
•
View damaged system at time 24. The damaged system at time 24 is saved in the file,
Damage_System_Time24.inp, shown in Table A.13. The Damage_System_Time24.inp
can be loaded into EPANET to be visualized as shown in Figure A.17.
•
View simulation results at time 24. The hydraulic simulation results associated with
each type of component, including junctions, tanks, pipes, pumps, and valves, are shown
in Tables A.14 to A.18. The system serviceability at time 24 is shown in Table A.19.
The simulation results can be visualized using the EPANET GUI as shown in Figure
A.18 by loading the Modified_System_Time24.inp into EPANET and running the
simulation.
26
Table A.6 Damaged System at Time 0
[TITLE]
[JUNCTIONS]
A1J22 160
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
200.000000
17
100.000000
19
200.000000
Added junction to model leak in pipe 22
Added empty reservoirs to model two
breaks in pipe 22, 1 leak in pipe 22, and 1
break in pipe
[RESERVOIRS]
A1R22 130
A2R22 130
A4R22 190
A5R22 190
A1R12 100
A2R12 100
A3R22 160
1
450.000000
[TANKS]
7
450.000000
[PIPES]
A1O22
A3O22
A4O22
A1O12
A2O12
A2O22
A1L22
10
16
18
20
4
6
8
17
A1J22
19
9
11
A1J22
A1J22
7
13
13
15
3
3
5
Original pipe 22 is replaced with pipes A1O22, A2O22,
A3O22, and A4O22. Original pipe 12 is replaced with
pipes A1O12 and A2O12. One pipe A1L22 is added to
model the leak occurred in pipe 22
120.000000
A1R22
A4R22
A5R22
A1R12
A2R12
A2R22
A3R22
9
15
17
19
5
9
11
0.000000
914.4
914.4
304.8
1524
1524
914.4
0.5
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
120.000000
12
12
12
12
12
12
2
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
30.000000
100
100
100
100
100
100
1e+006
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
[PUMPS]
2
1
[VALVES]
14
9
3
POWER
10.000000
13
4.000000
PRV
100.000000
27
0.000000
0.000000
1
1
1
1
1
1
1
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
CV
CV
CV
CV
CV
CV
Table A.6 Continued
[DEMANDS]
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
100.000000
17
100.000000
19
100.000000
[CURVES]
[PATTERNS]
PATN1 1.000000
PATN1 1.000000
1.000000
1.000000
1.000000
1.000000
1.000000
1.000000
[STATUS]
A3O22 Closed
[CONTROLS]
[SOURCES]
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.00000011
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
28
1.000000
1.000000
1.000000
Table A.6 Continued
[COORDINATES]
A1R22 284.468 163.435
A2R22 284.468 158.983
A4R22 284.468 136.723
A5R22 284.468 132.27
A1R12 211.622 154.711
A2R12 211.625 150.324
A1J22 280.016 147.853
A3R22 284.468 147.853
1
140.726688
3
169.667221
5
169.576993
7
207.220708
9
207.220708
11
207.252760
13
241.998111
15
241.998111
17
280.016223
19
280.016223
[VERTICES]
A1O22 280.016
A3O22 280.016
A4O22 280.016
A1O12 207.235
A2O12 207.238
A2O22 280.016
163.435
136.723
132.27
154.708
150.321
158.983
Added coordinates for new
reservoirs and junctions
174.581772
174.431972
130.595466
199.588372
174.450158
130.579090
174.517132
129.944774
174.565669
130.044299
Added vertices for new
pipes
[End]
29
Figure A.15 Damaged System with Node and Link IDs at Time 0
Table A.7 Junction Results at Time 0
Node_ID
3
5
9
11
13
15
17
19
Demand_gpm
100
100
100
100
100
100
100
0
Head_ft
454.92
307.92
278.8
168.05
278.8
278.62
163.86
0
Pressure_psi
153.79
90.09
77.47
29.49
77.47
34.07
27.67
0
Table A.8 Tank Results at Time 0
Tank_ID
1
7
Demand_gpm
-8036.31
-5459.8
Head_ft
450
570
30
Pressure_psi
0
52
Table A.9 Pipe Results at Time 0
Pipe_ID
10
12
16
18
20
22
4
6
8
Flow_gpm
5459.8
0
100
3305.09
0
0
3774.68
4161.63
3674.68
Velocity_fps
15.49
0
0.28
9.38
0
0
10.71
11.81
10.42
Headloss_/1000ft
95.54
0
0.06
37.71
0
0
48.23
57.78
45.89
Table A.10 Pump Results at Time 0
Pump_ID
2
Flow_gpm
8036.31
Velocity_fps
0
Headloss_/1000ft
-4.92
Table A.11 Valve Results at Time 0
Valve_ID
14
Flow_gpm
3505.09
Velocity_fps
9.94
Headloss_/1000ft
0
Table A.12 Serviceability at Time 0
Node_ID
3
5
9
11
13
15
17
19
Demand
1
Node_Serviceability
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
0
1
1
1
1
1
1
1
0
0.875
0.875
Sum
31
Figure A.16 Simulation Results at Time 0
32
Table A.13 Damaged System at Time 24
[TITLE]
[JUNCTIONS]
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
200.000000
17
100.000000
[RESERVOIRS]
A1R22 130
A2R22 130
A4R22 190
A5R22 190
A1R12 100
A2R12 100
A3R22 160
1
450.000000
[TANKS]
7
[PIPES]
A1O22
A1O12
A2O12
10
16
18
4
6
8
17
9
11
7
13
13
3
3
5
450
Tank water level is updated.
Tank is empty in this example.
0
A1R22
A1R12
A2R12
9
15
17
5
9
11
0
120
914.4
1524
1524
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
3048.00000
[PUMPS]
2
1
3
POWER 10.000000
[VALVES]
14
9
13
12.000000
PRV
30
0
12
12
12
12.00000
12.00000
12.00000
12.00000
12.00000
12.00000
100
100
100
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
100.000000
0.000000
[DEMANDS]
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
100.000000
17
100.000000
[CURVES]
33
1
CV
1
CV
1
CV
0.000000
0.000000
0.000000
0.000000
0.000000
0.000000
Table A.13 Continued
[STATUS]
[CONTROLS]
[SOURCES]
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.000000
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
[COORDINATES]
A1R22 284.468 163.435
A2R22 284.468 158.983
A4R22 284.468 136.723
A5R22 284.468 132.27
A1R12 211.622 154.711
A2R12 211.625 150.324
A3R22 284.468 147.853
1
140.726688
3
169.667221
5
169.576993
7
207.220708
9
207.220708
11
207.252760
13
241.998111
15
241.998111
17
280.016223
174.581772
174.431972
130.595466
199.588372
174.450158
130.579090
174.517132
129.944774
174.565669
[VERTICES]
A1O22 280.016 163.435
A1O12 207.235 154.708
A2O12 207.238 150.321
[End]
34
Tank is empty
Figure A.17 Damaged System at Time 24
35
Table A.14 Junction Results at Time 24
Node_ID
3
5
9
11
13
15
17
19
Demand_gpm
100
100
100
100
100
0
100
0
Head_ft
454.27
307.53
168.69
167.92
168.69
0
138.36
0
Pressure_psi
153.5
89.92
29.76
29.43
29.76
0
16.62
0
Table A.15 Tank Results at Time 24
Tank_ID
1
7
Demand_gpm
-9273.55
0
Head_ft
450
450
Pressure_psi
0
0
Table A.16 Pipe Results at Time 24
Pipe_ID
10
12
16
18
20
22
4
6
8
Flow_gpm
0
0
0
1609.71
0
0
3770.99
5402.57
3670.99
Velocity_fps
0
0
0
4.57
0
0
10.7
15.33
10.41
Headloss_/1000ft
0
0
0
9.95
0
0
48.14
93.69
45.8
Table A.17 Pump Results at Time 24
Pump_ID
2
Flow_gpm
9273.55
Velocity_fps
0
Headloss_/1000ft
-4.27
Table A.18 Valve Results at Time 24
Valve_ID
14
Flow_gpm
1709.71
Velocity_fps
4.85
Headloss_/1000ft
0
Table A.19 Serviceability at Time 24
Node_ID
3
5
9
11
13
15
17
19
Demand
1
Node_Serviceability
100
100
100
100
100
100
100
100
100
100
100
100
100
0
100
0
1
1
1
1
1
0
1
0
Sum
0.75
36
0.75
Figure A.18 Simulation Results at Time 24
A.5 MONTE CARLO WITH FIXED SIMULATION RUNS
Step 1: Export EPANET Format File
Export the hydraulic network model from H2ONET to the EPANET file format
following the same approach as described in Section A.4, Step 1.
The exported file,
Example_1.inp, is installed along with the GIRAFFE program and resides in the folder: Example
Files\Appendix A.
Step 2: Review EPANET File
Check the EPANET format file following the same procedures as described in Section
A.4, Step 2.
37
Step 3: Construct Files for Pipe Damage Generation and Earthquake Demand Simulation
The probabilistic implementation generates randomly distributed pipeline breaks and
leaks in the system according to pipeline repair rate, RR, length, L, and the conditional
probability of pipe break, Pbk , given that damage occurs.
In addition, the probabilistic
implementation determines the type of each leak according to the probability of that leak type for
different types of pipelines. The probabilistic implementation includes three steps: generating
pipe damage, deciding on damage states (leak or break), and determining leak type. Its detailed
methodology can be found in the GIRAFFE Users Manual main text and Shi (2006). The main
inputs from users for probabilistic pipe damage generation are the repair rate (RR), length (L),
and material of each pipeline. The conditional probability of pipe break, Pbk , and the probability
for each leak type for different types of pipelines have default values that can be changed by
clicking on Options | Configuration | Pipe Damage Probability in the GIRAFFE toolbar.
Figures A.19 and A.20 show the default values for the Pipe Damage Probability and the Pipe
Leakage Model. These default values of 20% breaks and 80% leaks are based on pipeline
damage repair data from a seismic event in the Seattle area. Data associated with the 1994
Northridge earthquake, however, seems to suggest a 5% break rate and 95% leak occurrence is
better suited to a Los Angeles area seismic event. Therefore, the user may decide to change the
default values in order to better model the characteristics of the study area.
38
Figure A.19 Default values for Pipe Damage Probability
39
Figure A.20 Default values for Pipe Leakage Model
The input file for probabilistic pipe damage generation is shown in Table A.20. This file
is a text file and users can use Microsoft Word, Excel, or Notepad to construct it and save it as a
tab-delimited text file with the extension .inp. The probabilistic pipe damage input file starts with
a headline, followed by the record of each pipeline. It is recommended that users copy the
headline to their own files. The headline terms in the pipe damage generation input file are
explained in Table A.21.
It is assumed that each pipe has a repair rate, RR, equal to repair/km in this example. The
determination of RR for each pipeline for a given earthquake scenario involves spatial
manipulation which is performed by GIS (see Appendices B and C for detailed methodology and
explanations). The pipe length and material information can be obtained from the hydraulic
network model database.
40
The input file for earthquake demand simulation is shown in Table A.22. This is also a
text file which users can create using Microsoft Word, Excel, or Notepad, and save as a tabdelimited text file with the extension .inp. The input file starts with a headline, followed by the
record of each demand node. The headline terms in the earthquake demand simulation input file
are explained in Table A.23.
It is assumed that each demand has a RR = 1 repair/km in this example.
The
determination of RR for each demand node for a given earthquake scenario involves in spatial
manipulation which is performed by GIS. It is further assumed that the network is divided into
two pressure zones, one upstream from pressure reducing valve 14, including junctions 3, 5, 9,
and 11, and the other downstream from pressure reducing valve 14, including junctions, 13, 15,
17, and 19. The mean pressure of each pressure zone is calculated by averaging the pressures at
the junctions inside the pressure zone for the undamaged system.
Then the mean pressure is
assigned to each demand node inside the pressure zone. The pressure at each junction for the
undamaged system is shown in Figure A.2.
41
Table A.20 Pipe Damage Input File for Monte Carlo Simulation with Fixed Simulation Times
(rr.inp)
PipeID
10
12
16
18
20
22
4
6
8
Length_km
RR
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Material
CI
CI
DI
DI
CON
CON
RV
RV
STL
Table A.21 Description of Columns in Probabilistic Pipe Damage Input File
Name
Type
Description
The ID of the pipe which users want to damage. Users have to make
PipeID
char
sure this pipe is in the system definition file otherwise the program
cannot run correctly. Maximum length 30 characters
Length
float
The length of the pipe in km. The length of each pipe can be obtained
from the system definition file.
Pipe repair rate in repairs per kilometer of pipe length, which is
correlated with seismic hazard parameters, such as peak ground
RR
float
velocity and permanent ground deformation. The determination of
repair rate for each pipeline involves spatial manipulation, which is
conducted using GIS.
The material of the pipeline. CI: cast iron pipeline; DI: ductile iron
Material
char
pipeline, RS: riveted steel pipeline; CON: concrete pipeline; STL:
welded steel pipeline, and N/A: other types of pipelines beside the
abovementioned five types of pipeline.
42
Table A.22 Input File for Earthquake Demand Simulation
ID
3
5
9
11
13
15
17
19
G_RR
1
1
1
1
1
1
1
1
Ave_PRESSURE
202
202
202
202
78
78
78
78
Table A.23 Description of Columns in Earthquake Demand Simulation Input File
Name
Type
Description
The ID of the demand node. Users have to make sure this
ID
char
demand node is in the system definition file otherwise the
program cannot run correctly. Maximum length 30 characters.
Pipe repair rate in repairs per kilometer of pipe length, which is
correlated with seismic hazard parameters, such as peak ground
G_RR
float
velocity and permanent ground deformation. The determination
of repair rate for each pipeline involves spatial manipulation,
which is conducted using GIS.
Ave_PRESSURE float
The average nodal pressure of the pressure zone, in which the
demand node is located.
43
Step 4: Input Parameters in GUI Window
Figure A.15 shows the GUI window with the required inputs for a Monte Carlo
simulation with a fixed number of simulation runs. Users may select “Monte Carlo Fixed” from
the screen that appears when first opening GIRAFFE, or by going to Simulations | Monte Carlo
Fixed in the main GIRAFFE toolbar. The meaning of each entry for a deterministic simulation is
explained in Table A.24.
Figure A.21 GUI Window with Input for a Monte Carlo Simulation with Fixed Simulation Runs
44
Table A.24 Input Parameters for Monte Carlo Fixed Simulation
Name
Description
Name of the EPANET system definition file with the
System Definition File
extension of .inp.. File name may have a maximum
length of 80 characters.
Pressure limit, in psi, below which GIRAFFE
Minimum Pressure to
eliminates the node and connected links from the
Eliminate
system.
Typically input 0 for negative pressure
elimination.
Simulation Time
Simulation Time Step
Total length of simulation time in hours to update tank
water levels. 0 for steady state simulation.
The time step in hours to update tank water levels. 1 for
steady state simulation.
Name of the input file for probabilistic pipe damage
Pipe Repair Rate File
generation. File name may have a maximum length of
80 characters.
Number of Simulations
Monte Carlo simulation time ranging from 1 to 100
Random Seed
Seed for random number generation.
Nodal Demand
Options to choose to simulate the earthquake demand or
Calibration
not: “Yes” for simulated and “No” for not simulated.
(If “Yes” was selected for “Nodal Demand Calibration”,
Regression Equation
this value is required.) Options for earthquake demand
simulation: “Mean Prediction Plus Noise Terms” or
“90% Confidence Level Prediction”.
(If “Yes” was selected for “Nodal Demand Calibration”,
Mean Pressure File
this value is required.) Name of the input file for
earthquake demand assessment. File name may have a
maximum length of 80 characters.
45
Step 5: Perform Simulation
GIRAFFE analyzes the network following the same procedures described in the
deterministic simulation for 10 Monte Carlo runs in the User Manual, Section 7.4.
Step 6: View Results
GIRAFFE saves the damaged system definition file, Damage_Info_Dert*.inp, and the
component results for each Monte Carlo simulation run. The Damage_Info_Dert*.inp files
contain the pipe break and leak information for each simulation run and have the same format as
the input file for deterministic pipeline damage generation, as shown Table A.2. The files
associated with each simulation run are bundled in separate folders and saved with a similar
naming convention as that used in the deterministic simulation. These files can be found in the
“Giraffe_Output” folder that exists in the same directory where the GIRAFFE application is
installed. The damaged system and modified system files are appended with a number indicating
which simulation run they are associated with, e.g. Damage_System_Time09.inp is the damaged
system file at time 0 for simulation run 9, and Modified_System_Time245.inp is the modified
system file at time 24 for simulation run 5.
Besides the results for each simulation run,
GIRAFFE reports the serviceability at times 0 and 24 for all simulation runs. The simulation
results users need to check are:
•
Damaged system at time 0, Damage_System_Time0.inp, as shown in Table A.25, for the
10th simulation run. Users need to check if the demands are updated if they choose the
simulation option to perform earthquake demand simulation.
•
System
serviceability
at
times
0
and
24,
in
files
Serviceability0.out
and
Serviceability24.out, respectively. These files are shown in Tables 7.19 and 7.20. In
these tables, the system serviceability is reported in a matrix format. For each Monte
Carlo simulation, the serviceability is reported for each demand node and for the entire
system. The mean of the nodal and system serviceability for all Monte Carlo simulations
is also calculated and reported.
46
Table A.25 Damaged System for the Last Run of Monte Carlo Simulation
[TITLE]
[JUNCTIONS]
A1J10 440.746
A2J10 344.961
A1J12 100
A1J16 105.151
A2J16 126.669
A1J22 147.075
A1J4 100
A2J4 100
3
100.000000
5
100.000000
9
100.000000
11
100.000000
13
100.000000
15
200.000000
17
100.000000
19
200.000000
[RESERVOIRS]
A2R12 100
A3R12 100
A2R22 153.127
A3R22 153.127
A1R10 440.746
A2R10 344.961
A1R12 100
A1R16 105.151
A2R16 126.669
A1R22 147.075
A1R4 100
A2R4 100
1
450.000000
[TANKS]
7
450.000000
[PIPES]
A2O12
A3O12
A2O22
A3O22
A1O10
A1L10
A2O10
A2L10
A3O10
A1O12
A1L12
A1O16
A1L16
A2O16
A2L16
A1J12
11
A1J22
19
7
A1J10
A1J10
A2J10
A2J10
9
A1J12
13
A1J16
A1J16
A2J16
120.000000
A2R12
A3R12
A2R22
A3R22
A1J10
A1R10
A2J10
A2R10
9 2133.26
A1J12
A1R12
A1J16
A1R16
A2J16
A2R16
4.28195
1874.56
184.441
1428.7
80.5885
0.5
834.149
0.5
12
1169.15
0.5
156.993
0.5
655.89
0.5
0.000000
120.000000
12
12
12
12
12
1.58533
12
1.2
100
12
2.4
12
2.4
12
2.4
1
1
1
1
0
1
0
1
100
100
100
100
100
1e+006
100
1e+006
0
100
1e+006
100
1e+006
100
1e+006
47
0
1
0
1
0
1
CV
CV
CV
CV
CV
CV
CV
CV
CV
30.000000
0.000000
Table A.25 Continued
A3O16
A1O22
A1L22
A1O4
A1L4
A2O4
A2L4
A3O4
18
20
6
8
A2J16
17
A1J22
3
A1J4
A1J4
A2J4
A2J4
13
15
3
5
15
A1J22
A1R22
A1J4
A1R4
A2J4
A2R4
5
17
19
9
11
2235.12
1434.86
0.5
737.49
0.5
1120.94
0.5
1189.57
3048.00000
3048.00000
3048.00000
3048.00000
12
12
2.4
12
3.57771
12
3.57771
12
12.00000
12.00000
12.00000
12.00000
100
100
1e+006
100
1e+006
100
1e+006
100
100.000000
100.000000
100.000000
100.000000
100.000000
0.000000
0
0
1
0
1
0
1
0
0.000000
0.000000
0.000000
0.000000
CV
CV
CV
[PUMPS]
2
1
[VALVES]
14
9
3
POWER 10.000000
13
12.000000
[DEMANDS]
3
5
9
11
13
15
17
19
921.02
921.02
921.02
921.02
422.54
422.54
422.54
422.54
PRV
Demands are changed
to consider the effects
of earthquake damage
to distribution network
[CURVES]
[PATTERNS]
PATN1 1.000000
PATN1 1.000000
1.000000
1.000000
1.000000
1.000000
[STATUS]
[CONTROLS]
[SOURCES]
[QUALITY]
[REACTIONS]
GLOBAL BULK 0.000000
GLOBAL WALL 0.000000
[ENERGY]
[OPTIONS]
UNITS GPM
HEADLOSS H-W
48
1.000000
1.000000
1.000000
1.000000
1.000000
Table A.25 Continued
VISCOSITY 1.1e-005
DIFFUSIVITY 1.3e-008
SPECIFIC GRAVITY 1.000000
TRIALS 40
ACCURACY 0.001
DEMAND Multiplier 1.000000
[REPORT]
PAGESIZE 30
STATUS NO
NODE ALL
LINK ALL
[COORDINATES]
A2R12 211.619 159.757
A3R12 211.622 155.37
A2R22 284.468 153.139
A3R22 284.468 148.687
A1J10 207.221 198.924
A1R10 209.735 198.924
A2J10 207.221 192.044
A2R10 209.735 192.044
A1J12 207.233 157.622
A1R12 211.62 157.628
A1J16 241.998 172.221
A1R16 246.455 172.221
A2J16 241.998 162.63
A2R16 246.455 162.63
A1J22 280.016 153.607
A1R22 284.468 153.607
A1J4 169.645 163.825
A1R4 165.262 163.843
A2J4 169.612 147.704
A2R4 165.229 147.722
1
140.726688
3
169.667221
5
169.576993
7
207.220708
9
207.220708
11
207.252760
13
241.998111
15
241.998111
17
280.016223
19
280.016223
[VERTICES]
A2O12 207.231
A3O12 207.235
A2O22 280.016
A3O22 280.016
174.581772
174.431972
130.595466
199.588372
174.450158
130.579090
174.517132
129.944774
174.565669
130.044299
159.754
155.367
153.139
148.687
[End]
49
Table A.26 Serviceability of Monte Carlo Simulation with Fixed Simulation Times at Time 0
Node_ID Demand
1
2
3
4
5
6
7
8
9
10
Node_Serviceability
3
5
9
11
13
15
17
19
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
0
0
1
0.9
1
1
1
0.3
0.7
0.2
0.5
0.75
1
0.625 1
0.75
0.75
0.75
0.75
100
100
100
100
100
100
100
100
Sum
Mean Node
Serviceability for All
Monte Carlo Runs
System
Serviceability of
Each Monte Carlo
0.75
0.7625
Mean System
Serviceability for All
Monte Carlo Runs
Table A.27 Serviceability of Monte Carlo Simulation with Fixed Simulation Times at Time 24
Node_ID Demand
1
2
3
4
5
6
7
8
9
10
Node_Serviceability
3
5
9
11
13
15
17
19
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
0
0
1
0.9
1
1
1
0.2
0.6
0.1
0.5
0.75
1
0.625 0.875 0.75 0.75
0.75
0.625 0.625
Sum
100
100
100
100
100
100
100
100
50
0.725
A.6 MONTE CARLO SIMULATION WITH FLEXIBLE SIMULATION RUNS
Step 1: Export EPANET Format File
Export the hydraulic network model from H2ONET to EPANET format file,
Example_1.inp, following the same approach as described in Section A.4, Step 1.
Step 2: Check EPANET File
Check the EPANET format file, Example_1.inp, following the same procedures
described in Section A.4, Step 2.
Step 3: Construct Files for Pipe Damage Generation and Earthquake Demand Simulation
Construct the input files, rr.inp, for pipe damage generation and, Node_Pressure.inp, for
earthquake demand simulation, using the same format as shown in Tables A.20 and A.22,
respectively. The files must be in tab-delimited format.
Step 4: Input Parameters
Selecting a Monte Carlo Flexible simulation in GIRAFFE produces a GUI window as
shown in Figure A.16. All the entries in this GUI window have the same meaning and format as
those shown in Figure A.15 and described in Table A. 24.
Step 5: GIRAFFE Performs Simulation
GIRAFFE analyzes the network following the same processes described in the
deterministic simulation and determines how many Monte Carlo simulation runs are needed to
have statistically significant results using the built-in algorithm.
Step 6: View Results
GIRAFFE saves the damaged system definition file and the component results for each
Monte Carlo simulation and saves the serviceability at times 0 and 24 for all simulation runs.
The system serviceability is reported in Tables A.28 and A.29 for times 0 and 24, respectively.
In total, 20 runs of Monte Carlo simulations are performed in this example.
51
Figure A.22 Inputs for Monte Carlo Simulation with Flexible Simulation Runs
52
Table A.28 Serviceability of Monte Carlo Simulation with Flexible Simulation Runs at Time 0
Node_ID Demand 1
3
5
9
11
13
15
17
19
100
100
100
100
100
100
100
100
Sum
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
100
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
100
0
0
100
0
100
0
100
0
100
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
0.75 0.75 0.75 0.75 0.75 0.5
0.5
0.625 0.75 0.75
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
0.5
0.75 1
0.625 1
0.625 0.5
0.625 0.75 0.75
Node_Serviceability
1
0.8
1
0.95
1
0.15
0.6
0.1
0.7
Table A.29 Serviceability of Monte Carlo Simulation with Flexible Simulation Runs at Time 24
Node_ID Demand 1
2
3
5
9
11
13
15
17
19
100
0
100
100
100
0
0
0
100
100
100
100
100
0
100
0
0.5
0.75 1
Sum
100
100
100
100
100
100
100
100
3
4
5
100
100
100
100
100
100
100
100
100
100
100
100
100
0
0
0
100
100
100
100
100
100
100
0
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
0
100
0
100
0
100
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
0
100
100
100
0
100
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
0
0
100
0
100
100
100
0
0
0
100
100
100
100
100
0
0
0
100
100
100
100
100
0
100
0
100
100
100
100
100
0
100
0
0.625 0.875 0.75 0.75 0.75 0.625 0.625 0.5
53
0.5
0.625 0.625 0.75 0.625 0.5
0.625 0.75 0.75
Node_Serviceability
1
0.75
1
0.95
1
0.1
0.55
0.05
0.675
A.7 USING DETERMINISTIC RESULTS AS INPUT FOR MONTE CARLO SIMULATION
RUNS
The primary utility of running a deterministic simulation is that it allows the user to
explicitly specify damage to the system and then view the effects of that damage. This is
especially useful when experts can predict regions of permanent ground deformation where
damage is likely to occur. A useful application of GIRAFFE is the ability to combine a stochastic
simulation and a deterministic simulation. The deterministic simulation allows users to specify
damage due to permanent ground deformation and results from this simulation can be fed into a
stochastic simulation where additional damage will be applied to the system based on the
selected earthquake scenario. Note that this can be done with both fixed and flexible Monte
Carlo simulations.
Step 1: Start a Deterministic Simulation
Run the deterministic simulation, adding damage to pipes as described in Section A.4
DETERMINISTIC SIMULATIONS. When the simulation is complete, there will be a file
called Damage_System_Time01.inp in the “Giraffe_Output” folder of results. This file is in the
same format as the original system definition file, but it includes damage to the system that was
specified in the deterministic simulation.
Damage_System_Time01
file
and
not
(Please note that the user should use the
the
Modified_System_Time01
file.
The
Modified_System_Time01 file will have been through hydraulic analysis and negative pressures
and corresponding model elements will have been removed. For accurate results, the user needs
to use the Damage_System_Time01 file for the next step, so all model elements are present at the
start of the stochastic simulation). Using the Damage_System_Time01.inp file as the system
definition file for a Monte Carlo simulation, the user can have additional damage (created by the
stochastic simulation) imposed with the damage explicitly applied via the deterministic
simulation module.
Step 2: Run Monte Carlo Simulation Using Deterministic Results
Open up the GIRAFFE GUI to begin a Monte Carlo Simulation (either Flexible or Fixed).
Use the Damage_System_Time01.inp file as the system definition file (you may want to rename
54
this file as something more descriptive). At this point, the user can choose to incur additional
trunk line damage via the Monte Carlo simulation by selecting a pipe repair rate file
(SM_rrout.inp) for a particular scenario earthquake. Damage to the distribution network will be
simulated if the user chooses “Yes” for “Calibrate Nodal Demand” and then selects the
appropriate mean pressure file. If desired, the tank fragility module can be applied. Figure A.23
shows what the inputs would look like for this type of a simulation.
In some cases, a user will not want to incur any additional trunk line damage (i.e. only
include the damage that has been applied deterministically), but would still like to have damage
simulated for the distribution network. This can be accomplished by creating a pipe repair rate
file with zeros for all of the pipe repair rates (as shown in Figure A.24). The user can then select
“Yes” for “Calibrate Nodal Demand” and selects a mean pressure file and distribution line
damage will occur. If desired, the tank fragility module can be applied.
Once the inputs are in place, the user can run a Monte Carlo simulation and view results.
The output files will include the effects of both deterministic and stochastic damage to the
system.
Figure A.23. Example of Monte Carlo Simulation Using a Deterministic SimulationGenerated System Definition File.
55
Figure A.24 Example of a Pipe Repair Rate file for No Trunk Line Damage
56
REFERENCES
MWH Soft, Inc. (1999). H2ONET Users Guide. Pasadena, CA.
Rossman, L.A. (2000). EPANET 2 Users Manual. National Risk Management Research
Laboratory, Office of Research and Development, U.S. Environmental Protection
Agency, Cincinnati, OH.
Shi, P. (2006). “Seismic Response Modeling of Water Supply Systems.” Ph.D. Dissertation,
School of Civil & Environmental Engineering, Cornell University, Ithaca, NY.
57
APPENDIX B
GIRAFFE INPUT PREPARATION
B.1
INTRODUCTION
This appendix provides a demonstration on how to prepare input files for GIRAFFE
simulations. To simulate the performance of a damaged water system, GIRAFFE needs two
types of input files: a system definition file in EPANET format defining the intact water system
and some system damage files describing damage scenarios. GIRAFFE can perform both
deterministic and stochastic simulations. Depending on whether the simulation is deterministic or
stochastic, GIRAFFE asks for different system damage files. This appendix focuses on preparing
system damage files for stochastic simulations by a series of manipulations and spatial analyses
using H2ONET, ESRI ArcGIS, and Microsoft Excel. The preparation of the system definition
file and system damage file for deterministic simulations are described in detail in Section A.4
DETERMINISTIC SIMULATIONS of Appendix A. Although the thought process of the
input file preparation applies to other GIS and hydraulic analysis software, this demonstration
uses ESRI ArcGIS 8.3 and H2ONET 3.5 in particular, and usage of other similar GIS and/or
hydraulic analysis software may lead to slight variations of the procedures. Users should be
familiar with GIRAFFE, EPANET, H2ONET, ESRI ArcGIS, and Microsoft Excel before the
GIRAFFE input file preparation. For more information on EPANET, H2ONET, and ESRI
ArcGIS, users can refer to the EPANET User Manual (Rossman, 2000), H2ONET User Manual
(MWH Soft, Inc., 1999), and ESRI User Manual (Booth and Mitchell, 2001).
This appendix is tailored to the seismic performance evaluation of the Los Angeles water
supply system, which is operated by Los Angeles Department of Water and Power (LADWP).
For the details of the evaluation process, please refer to Wang (2006) and Shi (2006). The
appendix starts with a brief description of LADWP seismic hazard characterization, followed by
H2ONET analysis of the LADWP water supply system. Then, it proceeds to the GIS
manipulations and spatial analysis and Excel spreadsheet calculations for the preparation of input
files in the GIRAFFE format.
1
B.2
LADWP SEISMIC HAZARD CHARACTERIZATION
The seismic hazard characterization for the LADWP water supply system was developed
by approximating the aggregate seismic hazard in the area that takes into account all currently
identified, potential seismic sources in a probabilistic context. This was accomplished by
examining 59 scenario earthquakes that were selected to provide probability of exceedance
characteristics for strong ground motion similar to those for all currently identified potential
seismic sources in the area (Lee et al. 2005; and Wang 2006). Table B.1 summarizes information
about the 59 scenario earthquakes.
For each of the 59 scenario earthquakes, several strong ground motion parameters at
equivalent rock sites, i.e., peak ground acceleration (PGA), peak ground velocity (PGV), and
spectral acceleration with 5% damping at T = 0.2 sec (SA0.2), and T = 1.0 sec (SA1), respectively,
are generated at 572 points in a grid with uniform separation of points and interval of 0.03°
longitude and latitude covering the LADWP water supply system. The grid is shown in Figure
B.1, superimposed by the LADWP trunk line system.
For each strong ground motion parameter at the 572 grid points, strong motion data are
generated corresponding to both the mean and mean ± σ, where σ is the total standard error for
the strong ground motion. Table B.2 shows an illustration of strong ground motion for scenario
earthquake 175 (Verdugo, Mw = 6.9). The first column indicates the scenario ID, which is
defined in Table B.1. The second and third columns define the geographic coordinates for the
grid points. The fourth, fifth, and sixth columns show the mean, mean + σ, and mean - σ PGA,
respectively. In a similar fashion, the remaining columns show the mean, mean + σ, and mean σ PGV, SA0.2, and SA1, respectively. Please note: due to limited space, only 41 of 572 grid points
are shown in Table B.2.
2
Table B.1. Characteristics of 59 Scenario Earthquakes
Scenario
Scenario
ID
Name
12
el15
18
SAF - Mojave
19
SAF - Carrizo
21
SAF-All southern segments
22
SAF - 1857
23
SAF - Southern 2 segments
118
Holser
119
Hollywood
120
Raymond
122
Clamshell-Sawpit
141
Newport-Inglewood offshore
145
Coronado Bank
159
Newport-Inglewood
160
Newport-Inglewood
161
Newport-Inglewood
162
Newport-Inglewood
166
Sierra Madre
167
Sierra Madre
168
Sierra Madre
169
San Gabriel
170
San Gabriel
171
San Gabriel
173
Malibu Coast
174
Santa Monica
175
Verdugo
176
Verdugo
177
Verdugo
189
Oak Ridge-onshore
191
Oak Ridge-onshore
195
San Cayetano
196
San Cayetano
198
Santa Susana
202
Simi-Santa Rosa
203
Simi-Santa Rosa
219
Anacapa-Dume
Magnitude
Mw
6.8
7.3
7.4
8.1
7.8
7.7
6.5
6.4
6.5
6.5
7.1
7.6
7.1
6.6
6.6
6.6
7.2
6.7
6.7
7.2
6.7
6.7
6.7
6.6
6.9
6.4
6.4
7
6.5
7
6.5
6.7
7
6.5
7.5
3
Annual Occurrence
Frequency
3.60E-03
4.13E-03
2.28E-03
3.00E-03
9.61E-03
3.37E-03
1.66E-04
6.64E-06
7.41E-04
1.06E-03
2.56E-03
1.75E-03
8.10E-04
2.37E-03
5.58E-04
1.50E-04
7.45E-04
4.40E-03
2.21E-04
1.53E-03
9.97E-05
1.27E-03
2.70E-06
5.23E-04
9.65E-04
1.57E-05
2.84E-06
4.13E-03
3.86E-03
6.86E-03
6.03E-03
3.01E-03
6.35E-04
2.87E-04
9.36E-04
Table B.1. (Continued)
Scenario
ID
220
221
222
370
371
372
378
388
397
398
399
440
443
444
446
447
451
452
453
454
559
560
561
562
Scenario
Name
Anacapa-Dume
Anacapa-Dume
Anacapa-Dume
Northridge
Northridge
Northridge
Channel Island Thrust
Upper Elysian Park
Puente Hills blind thrust
Puente Hills blind thrust
Puente Hills blind thrust
Cucamonga
Sierra Madre-San Fernando
Palos Verdes
Palos Verdes
Palos Verdes
Palos Verdes
Palos Verdes
Palos Verdes
Palos Verdes
Background Source
Background Source
Background Source
Background Source
Magnitude
Mw
7
7
6.5
7
6.5
6.5
7.5
6.4
7.1
6.6
6.6
6.9
6.7
7.3
6.8
6.8
6.3
6.3
6.3
6.3
7
7
7
7
Annual Occurrence
Frequency
5.70E-04
9.43E-04
1.29E-06
1.43E-03
2.88E-04
2.37E-05
5.12E-04
6.13E-05
8.63E-04
1.04E-05
8.21E-05
6.18E-03
9.41E-04
1.05E-03
8.20E-04
6.24E-04
3.27E-03
1.44E-03
2.07E-03
2.17E-03
1.05E-03
7.75E-04
1.29E-03
7.63E-04
The standard error, σinter-event, associated with inter-event variability to account for only
the “source” effects (Wang 2006) is estimated as 0.31 for PGA, PGV, and SA1, and 0.35 for SA0.2
(Lee et al. 2005). Since the σinter-event is the standard deviation of the natural log of the strong
ground motion, the strong motion data corresponding to mean ± σinter-event, can be calculated from
the mean strong motion data by:
mean ± σ int er − event = mean × exp(±σ int er − event )
4
(B.1)
Figure B.1. Spatial Distribution of 572 Grid Points for Strong Motion Data
5
Table B.2. Illustration of Strong Ground Motion Data for Scenario Earthquake 175
6
B.3
H2ONET ANALYSIS OF LADWP WATER SUPPLY SYSTEM
The system characteristics of the LADWP water supply system have been consolidated
into a hydraulic network model (LADWP 2002) by LADWP engineers using a commercial
software, H2ONET (MWH Soft, Inc. 1999). H2ONET is an interactive, multi-application
software program for the modeling of water distribution piping systems. It combines a point and
click interface for network construction, drawing, and database management. It contains highly
advanced and computationally efficient hydraulic and water quality simulation modules based on
EPANET (EPA 2005), and a graphical interface running within AutoCAD (Autodesk 2005) for
the Windows environment. H2ONET not only is capable of construction and maintenance of the
water supply system data inventory with reference to spatial coordinates, but also offers flexible
data exchange with other software, such as EPANET and GIS, enabling integration with other
relevant information and data.
The components in the H2ONET hydraulic network analysis can be divided into two
broad categories: link-type components, such as pipelines, and node-type components, such as
demand nodes. For stochastic simulations, GIRAFFE needs damage information for both the
link-type and node-type components. This section describes how to export data from H2ONET
to GIS for both the link-type and node-type components, after a brief description of the LADWP
water supply system.
B.3.1 System Description
Figure B.2 shows the LADWP water supply system in H2ONET. The system provides
water to about 3.8 million people in a service area of approximately 1,200 km2. The total water
consumption of the LADWP system in a typical summer and winter day is about 2.5×106 and
1.2×106 m3, respectively. The water is distributed primarily by gravity flow from north to south
throughout the LADWP service area. The H2ONET hydraulic network model contains 9,287
nodes and 10,665 links, representing about 2,186 km of pipelines, 1,052 demand nodes, 591
control valves, 110 tanks and reservoirs, 151 local groundwater wells, and 284 pumps.
7
Figure B.2. Overview of LADWP Water Supply System in H2ONET
B.3.2 Exporting Link-type Component Data to GIS
The procedures to export link-type component data from H2ONET to GIS are illustrated
in this section using the pipe data as an example. To export the data from H2ONET to GIS,
•
go to the H2ONET drop-down menu Exchange | Export Manager…and click on
it, as shown in Figure B.3
•
a H2ONET Export window will open, select Pipe in Export Source, Domain in
Element Scope, Shapefile in Format, name your file and specify your
destination to store the file.
•
Click Next button, and another window will open
•
Click Next button, and another window will open
•
Click Finish button
8
(a) Step 1
(b) Step 2
Figure B.3. Exporting Pipe Data from H2ONET to GIS
9
(c) Step 3
(d) Step 4
Figure B.3. Exporting Pipe Data from H2ONET to GIS (Continued)
B.3.3 Exporting Node-type Component Data to GIS
The procedures to export node-type component data from H2ONET to GIS are illustrated
in this section using the junction data as an example. To export the data from H2ONET to GIS,
•
go to the H2ONET drop-down menu Exchange | Export Manager…and click on
it, as shown in Figure B.4
10
(a) Step 1
(b) Step 2
Figure B.4. Exporting Junction Data from H2ONET to GIS
11
(c) Step 3
(d) Step 4
Figure B.4. Exporting Junction Data from H2ONET to GIS (Continued)
•
a H2ONET Export window will open, select Junction in Export Source,
Domain in Element Scope, Shapefile in Format, name your file and specify
your destination to store the file.
•
Click Next button, and another window will open
•
Click Next button, and another window will open
•
Click Finish button
12
GIRAFFE accounts for the damage to distribution pipelines implicitly by adjusting the
nodal demands and requires mean pressures of the undamaged, local distribution systems to
facilitate such adjustment (Shi 2006 and Wang 2006). To obtain the mean pressures of the local
distribution systems, the pressures at each node of the system must be first exported to GIS by
the following procedures:
•
Go to the H2ONET drop-down menu Tools | Run Manager…and click on it, as
shown in Figure B.5
•
A Run Manager Window will open, and click Run button after choosing the
appropriate settings and model.
•
After the simulation is finished, go to the H2ONET drop-down menu Tools |
Output Report/Graph…and click on it.
•
An Output Report Manager Window will open, and click on the New button
•
An Output Report & Graph Window will open, select Junction Report, and click
OK.
•
An window will open showing the hydraulic analysis results at all nodes
•
Select all the results and copy them.
•
Open Microsoft Excel, paste all the node results to a new spreadsheet, and save it
as Node_Pressure.dbf in a dBASE IV (*.dbf) format, which can be directly linked
to ESRI ArcGIS or Manifold System.
13
(a) Step 1
Run Button
(b) Step 2
Figure B.5. Exporting Node Pressure to GIS
14
(c) Step 3
(d) Step 4
Figure B.5. Exporting Node Pressure to GIS (Continued)
15
(e) Step 5
(f) Step 6
Figure B.5. Exporting Node Pressure to GIS (Continued)
16
(g) Step 7
Figure B.5. Exporting Node Pressure to GIS (Continued)
17
B.4
GIS SPATIAL ANALYSIS AND EXCEL SPREADSHEET CALCULATION
After the seismic hazard in the LADWP water supply system is characterized by the 59
scenario earthquakes and the system characteristics of the LADWP system are exported from
H2ONET, GIS spatial analysis and Excel spreadsheet calculations are followed to generate the
system damage files in GIRAFFE format. This section first describes how to calculate the mean
pressures at each demand node, followed by interpolation of strong ground motion data and site
condition correction. Then it demonstrates the procedures to assign strong ground motion
demands to both link-type and node-type components of water system. This section uses the
Scenario 175 Verdugo earthquake as an illustration, and the same procedures apply to each of the
59 scenario earthquakes.
B.4.1. Importing LADWP Water Supply System in GIS
To import the LADWP water supply system into a GIS, launch ESRI ArcGIS or
Manifold System 7.x and add the pipe (epa_pipes in this example) and junction data
(epa_junctions in this example) in ArcMap or in your Manifold project, as shown in Figure B.6(a)
(ArcGIS) and B.6(b) (Manifold System). After importing both shapefiles, assign their projection
to be State Plane – California 5, North American Datum (NAD) of 1983, with units of feet. To
do this in Manifold, right click on each file and select Assign Projection. In the pop-up window,
go to National Grids | State Plane (NAD83, feet) | State Plane – California 5, and make sure
that the center latitude/longitude, offset, scale, false easting/northing, and units match what is
shown in Figure B.6(c). To project the shapefiles in ArcGIS, open ArcToolbox and go to Data
Management Tools | Projections | Project Wizard (shapefiles, geodatabases). Follow the
instructions in the Projection Wizard and in the Spatial Reference Properties window click the
Select button to select a predefined coordinate system. In the Browse for Coordinate System
window, go to Projected Coordinate Systems | State Plane | NAD83 (Feet) | NAD 1983
StatePlane California V FIPS 0405 (Feet).prj and click the Add button.
18
Add Data Button
Figure B.6(a). Importing Pipe and Junction Data in ESRI’s ArcGIS
19
Figure B.6(b). Importing Pipe and Junction Data in Manifold System GIS.
Figure B.6(c). Projecting Pipe and Junction Data in Manifold System GIS.
20
After projecting the shapefiles, calculate the mean pressures at local distribution systems:
•
Right-click on epa_junctions and select Joins and Relates | Join… from the popup menu, as shown in Figure B.7(a).
•
A Join Data window will open. Assign the parameters as shown in Figure B.7(b)
to join the node pressure data, Node_Pressure.dbf, to the epa_junctions shapefile.
•
Right-click on epa_junctions and click on Open Attribute Table in the pop-up
menu. The attribute table of epa_junctions will open.
•
Find the attribute column named epa_junctions.Zone, and right-click the header
of the column.
•
Click on Summarize… in the pop-up menu.
•
A window will open. Assign the parameters as shown in Figure B. 7(e) to
generate a file named MeanPressure.dbf containing the mean pressures for each
local pressure zone
•
Right-click on epa_junctions and go to the pop-up menu Joins and Relates |
Join…, as shown in Figure B. 7(a).
•
The Join Data window will open. Assign the parameters as shown in Figure B.7(f)
to join MeanPressure.dbf to epa_junctions.
21
(a) Step 1
Figure B.7. Calculating Mean Pressure for Each Local Pressure Zone
22
(b) Step 2
Figure B.7. Calculating Mean Pressure for Each Local Pressure Zone (Continued)
23
(c) Step 3
Figure B.7. Calculating Mean Pressure for Each Local Pressure Zone (Continued)
24
(d) Step 4
Figure B.7. Calculating Mean Pressure for Each Local Pressure Zone (Continued)
25
(e) Step 5
(f) Step 6
Figure B.7. Calculating Mean Pressure for Each Local Pressure Zone (Continued)
26
GIRAFFE considers demand nodes in the LADWP trunk line system as an approximation
of local distribution systems and adjusts the nodal demands to simulate the local distribution
pipeline damage indirectly. Therefore, only the information regarding the demand nodes is
required in GIRAFFE simulations. To obtain the GIS data containing the demand nodes only:
•
open the attribute table of epa_junctions, as shown in Figure B. 8(a).
•
Click on the Option button at the bottom of the attribute table.
•
Click on Select by Attributes… in the pop-up menu.
•
In the Select by Attributes window, key in the syntax as shown in Figure B. 8(b),
and click Apply button.
•
Right-click on epa_junctions and go to the pop-up menu Data | Export Data…,
as shown in Figure B. 8(c)
•
An Export Data window will open. Name the data DemandNode, as shown in
Figure B. 8(d)
B.4.2. Strong Ground Motion Data Interpolation
As described in Section B.2, strong ground motion data are generated at 572 points for
each of the 59 scenario earthquakes. The data are provided in a *.txt format and cannot be
directly used by ESRI GIS. To convert the file format, open the strong ground motion data file
(e.g. 175.txt for the scenario 175 Verdugo earthquake) in Microsoft Excel, and save it in dBASE
IV (*.dbf) format (e.g. 175.dbf). To add the strong ground motion data to GIS,
•
go to ArcMap drop-down menu Tools | Add XY Data…and click on it, as shown
in Figure B.9(a)
•
In the open window, assign the parameters as shown in Figure B.9(b), and click
the Edit... button.
•
A window will open. Click the Select… button as shown in Figure B.9(c).
•
Another window will open. Go to Geographic Coordinate Systems | North
America | North America Datum 1983.prj, and click the Add button.
27
•
Right-click on 175 Events, and go to Data | Export Data… in the pop-up menu,
as shown in Figure B.9(g).
•
A window will open. Generate strong motion data in GIS format (*.shp) named
175_Data as shown in Figure B.9(h).
(a) Step 1
Figure B.8. Generating GIS Data for Demand Nodes
28
(b) Step 2
Figure B.8. Generating GIS Data for Demand Nodes (Continued)
29
(c) Step 3
(d) Step 4
Figure B.8. Generating GIS Data for Demand Nodes (Continued)
30
(a) Step 1
Figure B.9. Importing Strong Ground Motion Data in GIS
31
(b) Step 2
(c) Step 3
Figure B.9. Importing Strong Ground Motion Data in GIS (Continued)
32
(d) Step 4
(e) Step 5
(f) Step 6
Figure B.9. Importing Strong Ground Motion Data in GIS (Continued)
33
(g) Step 7
(h) Step 8
Figure B.9. Importing Strong Ground Motion Data in GIS (Continued)
34
As described in Section B.2, the mean, mean + σ, and mean - σ value of the ground
motion are provided, as well as an estimate of σinter-event. In the seismic performance evaluation of
the LADWP water supply system, the mean + σinter-event PGV are used (Wang 2006). To calculate
the mean + σinter-event PGV,
•
open attribute table of 175_data, as shown in Figure B.10(a).
•
Click the Option button at the bottom of the attribute table and select Add
Field… in the pop-up menu.
•
Create a new attribute column named SM_PGV with the parameters specified in
Figure B.10(c).
•
Right-click the header of the SM_PGV column and select Calculate Values… in
the pop-up menu, as shown in Figure B.10(d).
•
Ignore the warning message by clicking the Yes button.
•
Key in the syntax in the Field Calculator window and click OK, as shown in
Figure B.10(e).
The PGV contours based on the mean + σinter-event value are then generated in GIS using
the Geostatistical Analyst module by the following procedures:
•
go to Geostatistical Analyst | Geostatistical Wizard…, as shown in Figure
B.11(a).
•
The Geostatistical Wizard window will open. Assign the parameters as shown in
Figure B.11(b), and click Finish button.
•
A color contour surface will be generated as shown in Figure B.11(c).
•
Right-click the contour surface, and select Properties… in the pop-up menu, as
shown in Figure B.11(d).
•
A layer properties window will open. Check Filled contours, click the
Symbology tab, and then click the Classify… button, as shown in Figure
B.11(e).
•
A classification window will open, as shown in Figure B.11(f).
35
•
Change the parameters in the classification window as shown in Figure B.11(g),
and click OK.
•
The color contour surface changes the intervals to 5 cm/sec, as shown in Figure
B.11(h).
•
Right-click the contour surface and go to Data | Export to Vector… in the popup menu, as shown in Figure B.11(i).
•
A window will open. Assign the parameters to generate a GIS file name
SM_PGV_B, as shown in Figure B.11(j).
•
Add SM_PGV_B to the ArcMap, as shown in Figure B.11(k).
(a) Step 1
Figure B.10. Calculating Mean + σinter-event PGV
36
(b) Step 2
(c) Step 3
Figure B.10. Calculating Mean + σinter-event PGV (Continued)
37
(d) Step 4
(e) Step 5
Figure B.10. Calculating Mean + σinter-event PGV (Continued)
38
(a) Step 1
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites
39
(b) Step 2
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
40
(c) Step 3
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
41
(d) Step 4
(e) Step 5
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
42
(f) Step 6
(g) Step 7
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
43
(h) Step 8
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
44
(i) Step 9
(j) Step 10
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
45
(k) Step 11
Figure B.11. Generating Contour Surfaces for PGV at Rock Sites (Continued)
The strong ground motion data described above are generated for the rock site conditions,
i.e., NEHRP B or BC category site conditions (FEMA, 2003). However, the site conditions in the
LADWP water system service areas do not necessarily fall into the NEHRP B or BC site
categories. The NEHRP site conditions are divided into 6 categories, from A to F, representing
the site conditions from hard rock to soft soils, to soils requiring site specific evaluation.
Intermediate categories, such as BC, CD, and DE, can also be assigned to accommodate the site
conditions that fall close to the category boundary.
Wills et al. (2000) developed a site-condition map for California based on geologic units
and the average shear wave velocity in the upper 30-m subsurface layer. The GIS data for the site
conditions used in this study are provided by California Geological Survey, and the effects of site
46
amplification are accounted using the NEHRP-HAZUS approach (Wang 2006). The PGV for
category site conditions (other than B and B/C) can be calculated by
V pi = FPGViV pB
(B.2)
where Vpi is the PGV for category site condition i (i.e., site conditions corresponding to A, C, D,
or E), VpB is the PGV for site category B, and FPGVi is the correction factor for site condition i,
given by Table B.3.
The site condition data are added to ArcMap, as shown in Figure B.12(a). To make the
correction for site conditions, the following procedures are utilized:
•
Go to the drop-down menu Tools | GeoProcessing Wizard…, as shown in Figure
B.12(b)
•
In the GeoProcessing window check Intersect two layers, and click the Next
button, as shown in Figure B.12(c).
•
A window will open. Specify the parameters as shown in Figure B.12(d) to
generate a new GIS file named SM_PGV_Soils, and click the Finish button
•
Add SM_PGV_Soils to ArcMap, as shown in Figure B.12(e).
•
Open the attribute table of SM_PGV_Soils and click the Option button at the
bottom of the attribute table window. Go to Add Field… in the pop-up menu, as
shown in Figure B.12(f).
•
Create an attribute column named PGV, and specify the parameters as shown in
Figure B.12(g).
•
Right-click on the header of PGV column with and go to Calculate Values… in
the pop-up menu, as shown in Figure B.12(h).
•
A field calculator window will open. Key in the syntax as shown in Figure B.12(i),
and click OK.
•
Click the Option button at the bottom of the attribute table window, and go to
Add Field… in the pop-up menu, as shown in Figure B.12(j).
47
•
Create an attribute column named PGV_Soils, and specify the parameters as
shown in Figure B.12(k).
•
Click the Option button at the bottom of the attribute table window, and go to
Select by Attributes… in the pop-up menu, as shown in Figure B.12(l).
•
A window will open. Key in the syntax to select site condition CD as shown in
Figure B.12(m), and click the Apply button.
•
Right-click on the header of PGV_Soils column and go to Calculate Values… in
the pop-up menu, as shown in Figure B.12(n).
•
A field calculator window will open. Key in the syntax to make correction for site
condition CD as shown in Figure B.12(o), and click OK button.
•
Repeat the last four steps for correction of other site conditions listed in Table B.3.
Table B.3. Site Condition Correction Factor FPGV for PGV
Site Class
PGV ≤ 14
cm/sec
14 cm/sec
< PGV ≤
23.67 cm/sec
23.67 cm/sec
< PGV ≤
33.13 cm/sec
33.13 cm/sec
< PGV ≤
42.5 cm/sec
PGV > 42.5
cm/sec
A
0.8
0.8
0.8
0.8
0.8
B
1.0
1.0
1.0
1.0
1.0
C
1.7
1.6
1.5
1.4
1.3
D
2.4
2.0
1.8
1.6
1.5
E
3.5
3.2
2.8
2.4
2.4
F
---a
---a
---a
---a
---a
AB
0.9
0.9
0.9
0.9
0.9
BC
1.0
1.0
1.0
1.0
1.0
CD
2.05
1.8
1.65
1.5
1.4
DE
2.95
2.6
2.3
2.0
1.95
Note: a: Site-specific geotechnical investigation and dynamic site response analyses should be performed.
b: Use straight line interpolation for intermediate values of PGV.
48
(a) Step 1
Figure B.12. Site Condition Correction
49
(b) Step 2
(c) Step 3
Figure B.12. Site Condition Correction (Continued)
50
(d) Step 4
(e) Step 5
Figure B.12. Site Condition Correction (Continued)
51
(f) Step 6
(g) Step 7
Figure B.12. Site Condition Correction (Continued)
52
(h) Step 8
(i) Step 9
Figure B.12. Site Condition Correction (Continued)
53
(j) Step 10
(k) Step 11
Figure B.12. Site Condition Correction (Continued)
54
(l) Step 12
Figure B.12. Site Condition Correction (Continued)
55
(m) Step 13
Figure B.12. Site Condition Correction (Continued)
56
(n) Step 14
(o) Step 15
Figure B.12. Site Condition Correction (Continued)
57
B.4.3. Seismic Demands on Link-type Components
To determine the PGV that each pipeline is subjected to, the LADWP pipeline data layer
is combined with the corrected PGV contour surfaces in the ArcGIS using the “Intersect”
function. The “Intersect” function in ArcGIS not only combines the information from both input
data layers into an output layer, but also divides the pipelines according to the PGV contour
interval they fall into. Consider, for example, a pipeline that is so long that extends over three
PGV contour intervals, saying 40-45, 45-50, and 50-55 cm/sec intervals. The ArcGIS “Intersect”
function automatically divides the long pipeline into three new short pipelines and assigns a PGV
interval of 40-45, 45-50, or 50-55 cm/sec to each of them according to their locations,
respectively. A relatively small PGV interval of 5 cm/sec is utilized when developing the contour
surfaces, intending to determine the PGV values to each system component with relatively high
accuracy. The mean of the PGV interval (e.g., 42.5 cm/sec for 40-45 cm/sec interval) is taken as
the seismic demand for the system components located within the PGV interval. The detailed
procedures in ArcGIS are as follows:
•
go to the drop-down menu Tools | GeoProcessing Wizard… and click on it, as
shown in Figure B13(a).
•
A window will open, check Intersect two layers, and click Next button, as
shown in Figure B13(b).
•
A window will open, assign the parameters as in shown in Figure B13(c) to
generate a new GIS file, SM_PGV_Pipes, and click Finish button.
58
(a) Step 1
(b) Step 2
Figure B.13. Assigning PGV to Pipelines
59
(c) Step 3
Figure B.13. Assigning PGV to Pipelines (Continued)
Because the pipes may be divided by the “Intersect” function in ArcGIS, the length for
each divided pipe needs to be re-calculated. GIRAFFE can simulate damage to pipelines
composed of five different materials, cast iron (CI), ductile iron (DI), riveted steel (RS), steel
(STL), and concrete (CON). However, the H2ONET database contains more material types than
the five specified. For example, there are several different steel pipes in the H2ONET database.
Also, there are some pipes for which information about composition in the H2ONET database is
lacking. Therefore, additional GIS spatial analysis is needed to adjust the data. The detailed
procedures in ArcGIS are as follows:
•
Open attribute table of SM_PGV_Pipes, add a new attribute column PipeLength
as shown in Figure B.14(a).
•
Calculate the length of each divided pipe in the PipeLength column by keying in
the syntax in the field calculator window, as shown in Figure B.14(b), and
clicking OK button.
•
Add a new attribute column Material1 as shown in Figure B.14(c).
•
Click Option button at the bottom of attribute table, and click on Select by
Attributes… in the pop-up menu, as shown in Figure B.14(d).
60
•
A window will open. Key in the syntax as shown in Figure B.14(e) to select cast
iron (CI) pipes, and click Apply button.
•
Right-click on the header of the Material1 column, and select Calculate
Values… in the pop-up menu, as shown in Figure B.14(f).
•
A window will open, and key in the syntax as shown in Figure B.14(g) to classify
cast iron pipes as CI.
•
Repeat the last four steps to classify ductile pipes, concrete pipes, riveted steel
pipes, steel pipes, and pipes without material information as DI, CON, RS, STL,
and N/A, respectively. See Table B.4 for a complete description of the material
reclassification scheme.
(a) Step 1
Figure B.14. Adjusting GIS Data
61
(b) Step 2
(c) Step 3
Figure B.14. Adjusting GIS Data
62
(d) Step 4
Figure B.14. Adjusting GIS Data (Continued)
63
(e) Step 5
Figure B.14. Adjusting GIS Data (Continued)
64
Table B.4. Pipe Material Reclassification Scheme
(f) Step 6
(g) Step 7
Figure B.14. Adjusting GIS Data (Continued)
65
Original Classes
CI DI CON
CI DI CON
COP
New Classes
N/A RS
STL
N/A RS
STL
AC
RIV B&S WS
MANN
MATH
Molox Bell & Ball Joint
ST
STD
steel
Steel
STL GALV
VICT
WCJ
WRG
WS
WSJ
WWJ
The GIS data are then exported to Microsoft Excel for spreadsheet calculation of repair
rate in each pipe. To export the GIS data, click on the Option button at the bottom of attribute
table, as shown in Figure B.15(a) and go to Export… in the pop-up menu. A window will open,
and a spreadsheet file named Pipes.dbf in dBASE IV format is generated, as shown in Figure
B.15(b). Open the Pipes.dbf in Excel, and delete all other columns except ID, PGV_Soils,
PipeLength, and Material1. Then, the repair rate for each pipe can be calculated using
regressions (Wang 2006). Figure B.16 shows the regressions used in this study, which are based
on the performance of water supply systems in the 1994 Northridge earthquake (Jeon 2002 and
Wang 2006). The repair rates for each section of pipe (after having divided the pipes using the
GIS “Intersect” function) were calculated individually, using the repair rate vs. PGV regressions.
These repair rate values were then integrated by a weighted average (relating the divided pipe
lengths to the original pipe length) to obtain one repair rate for the original long pipe. An equalweight average of five repair rates using the five regressions in the figures was applied to the
pipelines (about 7% of total length in the LADWP system) without composition information
available in the H2ONET database (e.g. MATERIAL1 = “N/A”). After the calculation of repair
rates, the Excel spreadsheet was saved in MS-DOS text format (*.txt) and then renamed
following the input file name convention of GIRAFFE (*.inp). Figure B.17 shows an illustration
66
of the pipe damage file in GIRAFFE format. Be sure that the first line of the file matches exactly
what is shown. Note that length is in units of kilometers.
(a) Step 1
(b) Step 2
Figure B.15. Exporting Pipeline Damage GIS Data to Spreadsheet Calculation
67
10.000
10.000
Fit Equation (Concrete):
Ln(Y)=2.59Ln(X) - 12.11
Cast Iron
Ductile Iron
2
r = 0.83
r2 = 0.85
1.000
Repair Rate (No. of Repairs/km)
Repair Rate (No. of Repairs/km)
Fit Equation (Cast Iron):
Ln(Y)=1.21Ln(X) - 6.81
0.100
0.010
1.000
0.100
Fit Equation (Riveted Steel):
Ln(Y)=1.41Ln(PGV) - 8.19
0.010
r2 = 0.84
Fit Equation (Steel):
Ln(Y)=2.59Ln(X) -14.16
Fit Equation (Ductile Iron):
Ln(Y)=1.84Ln(X) -9.40
r2 = 0.74
r2 = 0.76
0.001
0.001
10
100
PGV (cm/sec)
10
1000
Concrete
Riveted Steel
Steel
100
PGV (cm/sec)
(a)
(b)
Figure B.16. Regressions of Pipeline Repair Rate vs. PGV (Wang 2006)
PipeID
H26
H28
H1210
H30
H670
H668
H1196
H152
H154
H1188
H68
H58
…
…
…
Length
0.0813727
0.428971
0.254347
0.23626
0.178226
0.125398
0.156046
0.15661
0.0964659
0.602962
0.439224
0.108612
…
…
…
RR
Material
0.00618749
0.0631566
0.0631566
0.0521385
0.0867362
0.0576887
0.074455
0.00780834
0.00780834
0.0683708
0.0631566
0.00780834
…
…
…
STL
DI
DI
DI
CI
CI
CI
STL
STL
CI
DI
STL
…
…
…
Figure B.17. Illustration of GIRAFFE Pipeline Damage File
68
1000
B.4.4. Seismic Demands on Node-type Components
The PGVs that the demand nodes are subjected to are determined by an ArcGIS function,
“Spatial Join”, which combines the information in the two input data layers (i.e. the demand
node layer and PGV contour surface layer) into an output data layer according to their spatial
positions. The detailed procedures in ArcGIS are as follows:
•
right-click on DemandNode, and go to Joins and Relates | Join… in the pop-up
menu, as shown in Figure B. 18(a).
•
A window will open, specify the parameters as shown in Figure B.18(b), and click
OK.
(a) Step 1
Figure B.18. Assigning PGVs to Demand Nodes
69
(b) Step 2
Figure B.18. Assigning PGVs to Demand Nodes (Continued)
The GIS data are then exported to Microsoft Excel for spreadsheet calculation of repair
rate at each demand node. To export the GIS data, click on the Option button at the bottom of
attribute table, as shown in Figure B.19(a), and go to Export… in the pop-up menu. A window
will appear, and a spreadsheet file named DistributionD.dbf in dBASE IV format is generated,
as shown in Figure B.19(b). Open the DistributionD.dbf in Excel, and delete all other columns
except ID, PGV_Soils, and Ave_Pressure. Then, the repair rates for the local distribution
pipelines are calculated using the regression for cast iron pipelines (Figure B.16) because the
majority of the local distribution lines are composed of cast iron (Wang 2006). After the
calculation of repair rates, the Excel spreadsheet is saved in MS-DOS text format (*.txt) and then
renamed following the input file name convention of GIRAFFE (*.inp). Figure B.20 shows an
illustration of local distribution pipeline damage file in GIRAFFE format. Be sure that the first
line in the file matches exactly what is shown.
70
(a) Step 1
(b) Step 2
Figure B.19. Exporting Demand Node GIS Data to Spreadsheet Calculation
71
ID
H41
H473
H71
H83
H87
H121
H231
H253
H293
H1303
H1327
H867
H1211
…
…
…
G_RR
0.086736
0.086736
0.086736
0.086736
0.086736
0.086736
0.086736
0.086736
0.086736
0.086736
0.086736
0.067573
0.067573
…
…
…
Ave_PRESSURE
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
84.1035
…
…
…
Figure B.20. Illustration of GIRAFFE Local Distribution Pipeline Damage File
72
REFERENCES
Autodesk (2005). http://usa.autodesk.com/adsk/servlet/home?siteID=123112&id=129446
Booth, B. and Mitchell, A. (2001). Getting Started with ArcGIS. Redlands, CA.
Environmental Protection Agency, USA (EPA, 2005).
http://www.epa.gov/ORD/NRMRL/wswrd/epanet.html
Federal Emergency Management Agency (FEMA, 2003). FEMA-450: NEHRP Recommended
Provisions for Seismic Regulations for New Buildings and Other Structures. 2003 Edition,
Washington, D. C., Developed by the Building Seismic Safety Council (BSSC) for
FEMA.
Jeon, S-S. (2002). “Earthquake Performance of Pipelines and Residential Buildings and
Rehabilitation with Cast-in-place Pipe Lining Systems.” Ph.D. Dissertation. Cornell
University, Ithaca, NY.
Los Angeles Department of Water and Power (LADWP) (2002). H2ONET Hydraulic Model of
the LADWP Water Supply System. Los Angeles, CA.
MWH Soft, Inc. (1999). H2ONET User Guide. Pasadena, CA.
Rossman, L.A. (2000). EPANET 2 User Manual. National Risk Management Research
Laboratory, Office of Research and Development, U.S. Environmental Protection
Agency, Cincinnati, OH.
Shi, P. (2006). “Seismic Response Modeling of Water Supply Systems.” Ph.D. Dissertation,
Cornell University, Ithaca, NY.
Wang, Y. (2006). “Seismic Performance Evaluation of Water Supply Systems.” Ph.D.
Dissertation, Cornell University, Ithaca, NY.
Wills, C. J., Petersen, M., Bryant, W. A., Reichle, M., Saucedo, G. J., Tan, S., Taylor, G., and
Treiman, J. (2000). “A Site-Conditions Map for California Based on Geology and ShearWave Velocity.” Bulletin of the Seismological Society of America, Vol. 90, No. 6B,
S187-S208.
73
APPENDIX C
GIRAFFE INPUT PREPARATION AND OUTPUT VISUALIZATION
USING MANIFOLD GIS
C.1
INTRODUCTION
This appendix provides a demonstration of how to prepare system damage input files for
GIRAFFE simulations using Manifold GIS, and how to use Manifold to create a cartographic
representation of GIRAFFE results. The system damage products of this appendix are identical
to those generated in Appendix B; therefore, the user may choose which method to use when
generating the system damage input files. The method presented in this Appendix is much more
efficient than the method used in Appendix B, though the user has less control over the results
using this method. This appendix focuses on preparing system damage files for stochastic
simulations by a series of manipulations and spatial analyses using Manifold GIS. The
preparation of the system damage file for deterministic simulations using Manifold GIS is
described in Section A.4 DETERMINISTIC SIMULATIONS of Appendix A. Although the
thought process of the input file preparation applies to other GIS and hydraulic analysis software,
this demonstration uses Manifold System 7.1. Users should be familiar with GIRAFFE,
EPANET, H2ONET, and Manifold before preparing GIRAFFE input files. For more information
on EPANET, H2ONET, and Manifold, users can refer to the EPANET User Manual (Rossman,
2000), H2ONET User Manual (MWH Soft, Inc., 1999), and Manifold System User Manual
(Manifold Net Ltd, 2006).
This appendix is tailored to the seismic performance evaluation of the Los Angeles water
supply system, which is operated by Los Angeles Department of Water and Power (LADWP).
For the details of the evaluation process, please refer to Wang (2006) and Shi (2006). For a brief
description of LADWP seismic hazard characterization and H2ONET analysis of the LADWP
water
supply
system,
refer
to
Section
B.2
LADWP
SEISMIC
HAZARD
CHARACTERIZATION and Section B.3 H2ONET ANALYSIS OF LADWP WATER
SUPPLY SYSTEM of Appendix B.
1
C.2
GIS SPATIAL ANALYSIS AND CALCULATION USING MANIFOLD
After the system characteristics of the LADWP system are exported from H2ONET, GIS
spatial analysis is used to generate the system damage files in GIRAFFE format. This section
describes how to use the GIRAFFE system damage tool, which is linked to Manifold in order to
perform the spatial analyses and calculations necessary to generate the damage files. This section
uses the Scenario 175 Verdugo earthquake as an illustration, and the same procedures apply to
each of the 59 scenario earthquakes.
This section also describes how to use the Manifold System Damage Add-In, which
formats the pipe data for use with the GIRAFFE system damage tool and performs the same
function as the GIRAFFE system damage tool.
C.2.1. Importing and Formatting Required Data in Manifold GIS
GIRAFFE requires specific GIS files to calculate the system damage. The following files
should be imported into a Manifold project and named exactly as shown unless otherwise
indicated:
•
epa_junctions.shp – Shapefile containing points that represent junction data for
the entire system. Obtained by exporting data from H20NET model as explained
in Section B.3.3 Exporting Node-type Component Data to GIS. See Figure C.1
for table format.
•
Epa_pipes.shp – Shapefile containing lines that represent pipe data for the entire
system. Pipes have been segmented into 1000’ pieces and segmented at the soil
boundaries as described in Section C.2.3, Table C.1. Pipes must be segmented
before running the tool, but this must only be done when the system is updated.
Obtained by exporting data from H20NET model as explained in Section B.3.2
Exporting Link-type Component Data to GIS. See Figure C.2 for table format.
•
Node_Pressure.dbf – Table containing pressure values for each node in the
system. Created by exporting node pressure data from H20NET system model
(see Section B.4, Figure B.5 on pages 14-17). See Figure C.3 for table format.
There must be a column named node_pressure (all lowercase letters).
2
•
Soil_Category_NAD83 Drawing.shp – Shapefile containing polygons that
represent the soil types in the area. Provided by California Geological Survey. See
Figure C.4 for table format.
•
Pgvsurf – User generated, empty surface in which scenario peak ground velocity
(PGV) values are interpolated.
Figure C.1. Illustration of epa_junctions table.
3
Figure C.2. Illustration of Epa_pipes table.
Figure C.3. Illustration of Node_Pressure table.
4
Figure C.4. Illustration of Soil_Category_NAD83 Drawing table.
A sample Manifold project (manifold_giraffe.map) has been included in the GIRAFFE
program file in the folder Example_Files | Appendix C, which also includes the required files in
the appropriate format. However, the user can also create a Manifold project to perform the same
function. To do this, import and format the required data:
•
Launch Manifold. Go to File | Import | Drawing to add the pipe (Epa_pipes) and
junction data (epa_junctions), as shown in Figure C. 5. All shapefiles, including
Soil_Category_NAD83 can be imported in this way.
•
Go to File | Import | Table to add the node pressure data (Node_Pressure), as
shown in Figure C.6.
•
Go to File | Import | Surface to add the surface to which PGV will be
interpolated (pgvsurf.grd), as shown in Figure C.7.
•
Alternatively to importing pgvsurf, the user can also create it (Figure C.8). Import
a scenario shapefile (175_data.shp in this example) and then,
o Right-click to copy the scenario drawing.
o Right-click in the project pane and select Paste As | Surface.
o Enter the parameters shown in Figure C.8(b).
5
o Rename the surface pgvsurf by right-clicking on the layer and selecting
Rename.
•
Format Node_Pressure.dbf by opening the table and right-clicking the header of
the column PRESSURE_(. Select Rename.
o In the pop-up input box, rename the column node_pressure (all lowercase
letters for the column header is important). The column containing the
node IDs should be named ID as shown in Figure C.3.
•
Open the attribute tables of epa_junctions and Epa_pipes, and check that the
columns containing the pipe and node IDs are named ID 2 as shown in Figures
C.1 and C.2. If the columns are named otherwise, rename them.
•
Next, check that each file contains the necessary columns. In the epa_junctions
attribute table, there should be a column named Zone. In the Epa_pipes attribute
table, there should be a column named Material. In the Soil_Category_NAD83
Drawing there should be a column named VSCAT, which describes the soil
categories (Figure C.4). These columns must exist and contain valid data values
for the tool to run correctly.
•
The pipe objects in Epa_pipes.shp must also be segmented into 1000’ pieces and
broken at the soil boundaries before running the system damage tool in GIRAFFE.
If this has not already been done, use the Segment Pipes for Repair Rate
Calculation tool in the Manifold System Damage Add-In as explained in Section
C.2.3 Using the Manifold System Damage Add-In. After running the segment
tool, the Epa_pipes file is ready to be used with the inbuilt GIRAFFE damage tool.
6
Figure C.5. Importing Pipe and Junction Data in GIS
Figure C.6. Importing Node Pressure Data in GIS
7
Figure C.7. Importing Pgvsurf in GIS
(a) Step 1
Figure C.8. Creating pgvsurf
8
(b) Step 2
Figure C.8. Creating pgvsurf
(c) Step 3
Figure C.8. Creating pgvsurf
C.2.2. Running the System Damage Tool in GIRAFFE
Once the Manifold project has been created, open GIRAFFE and select either Monte
Carlo Fixed or Flexible Simulations. To generate the system damage input files,
•
click Generate Pipe Repair Rate and Mean Pressure Files, as shown in Figure
C.9(a)
•
A pop-up window will occur. Navigate to and select the Manifold project that
contains the necessary data (in this example, manifold_giraffe.map)
9
•
In the next pop-up window, select the appropriate scenario shapefile
(175_data.shp). This scenario should correspond to the scenario for which the
simulation is being run. Selecting Open will begin the process.
•
When the process is completed, a pop-up notice will occur. Click OK. The
system damage files have now been generated for pipes and nodes, RRInput.inp
and LocalDemandInput.inp, respectively, and can be found in the GIRAFFE
program folder: C:\Program Files\Cornell University\GIRAFFE\AppendixB*.
The files are automatically placed the pipe repair rate and nodal demand boxes (if
Calibrate Nodal Demand is set to Yes) in the GIRAFFE interface.
* For 64-bit users, the files will be outputted to the 64-bit directory (C:\Program
Files\Cornell University\Appendix B), though this is not the location of the
GIRAFFE program folder (32-bit directory: C:\Program Files (x86)\Cornell
University\GIRAFFE).
(a) Step 1
Figure C.9. Generating System Damage Using Manifold-Linked Tool in GIRAFFE
10
(b) Step 2
Figure C.9. Generating System Damage Using Manifold-Linked Tool in GIRAFFE
(c) Step 3
Figure C.9. Generating System Damage Using Manifold-Linked Tool in GIRAFFE
(d) Step 4
Figure C.9. Generating System Damage Using Manifold-Linked Tool in GIRAFFE
11
(e) Step 5
Figure C.9. Generating System Damage Using Manifold-Linked Tool in GIRAFFE
C.2.3. Using the Manifold System Damage Add-In
Alternatively to creating the system damage files in GIRAFFE, the user may create the
files using the Manifold system damage add-in. The same files are required as for the system
damage tool in GIRAFFE (Epa_pipes, epa_junctions, pgvsurf, Soil_Category_NAD83 Drawing,
Node_Pressure.dbf). Using the add-in allows the user to view intermediate files generated
during the system damage preparation. Also, the pipe repair rate input file and the nodal demand
file can be prepared separately, which is useful if the user does not wish to calibrate nodal
demand in a Monte Carlo simulation.
Additionally, if the water supply system is updated (i.e. pipes or nodes are added or
removed) the user can prepare the data for use with the add-in or with the GIRAFFE system
damage tool. Note that if the system is updated, the pipes file should be segmented before it is
12
used in the GIRAFFE system damage tool because this tool does not segment the pipes. To use
the system damage add-in,
•
copy the LADWP folder from Manifold Tools in the GIRAFFE program folder
(normally C:\Program Files\Cornell University\GIRAFFE\Manifold Tools)
and paste it in the Config folder hierarchy for Manifold (normally C:\Program
Files\Manifold System\Config).
•
Launch Manifold and open the necessary files or a saved map containing the files.
If the toolbar does not automatically appear, go to Tools | Add-Ins | Add-In
Manager as shown in Figure C.10(a).
o A window will pop up, and check the box next to LADWP GIRAFFE
System Damage Preparation. Restart Manifold as directed.
o The custom system damage add-in toolbar should now be visible.
•
Before using the toolbar, import the scenario shapefile (for example, 175_data)
into the Manifold project.
13
(a) Step 1
Figure C.10. Using the Manifold System Damage Add-In
(b) Step 2
Figure C.10. Using the Manifold System Damage Add-In
14
(c) Step 3
Figure C.10. Using the Manifold System Damage Add-In
Each of the four buttons in the custom toolbar performs a different function in the system
damage preparation. Table C.1 describes the function of each button and the files required to run
each button. The system damage files generated are located in C:\temp, and they can be directly
inserted into a GIRAFFE Monte Carlo simulation.
15
Table C.1 Using the Manifold System Damage Custom Toolbar
Icon
Name
Function
Segment Pipes for Repair This tool segments, or breaks, the pipes
Rate Calculation
into 1000’ segments, and segments the
pipes at the soil boundaries. This only
needs to be run whenever a new,
unbroken pipes shapefile is imported into
the project or when the system is
updated.
Prepare Scenario Data
for Repair Rate
Calculation
This tool formats the scenario data PGV
values and interpolates the PGV values
onto a surface. This only needs to be run
once per scenario, and must be done
before using the following two tools.*
Prepare Pipe Repair Rate
Data
This tool calculates the soil-corrected
PGV values and the repair rate for each
pipe in the system, and then exports the
results as RRInput.inp.† This tool should
only be used after preparing the scenario
data and segmenting the pipes.
Prepare Earthquake
Demand Simulation Data
This tool calculates the mean pressure
and repair rate for each demand node in
the system, and then exports the results
as LocalDemandInput.inp.† This tool
should only be used after preparing the
scenario data.
Required Files
• Soil_Category_NAD83
Drawing
• Epa_pipes
• Pgvsurf
• Scenario shapefile
• Soil_Category_NAD83
Drawing
• Epa_pipes
• Pgvsurf
• Soil_Category_NAD83
Drawing
• epa_junctions
• Node_Pressure.dbf
• Pgvsurf
*Before using the Prepare Scenario Data for Repair Rate Calculation tool, the user must import
the scenario data. This can be done by going to File | Import | Drawing and navigating to the
appropriate file. Scenario data must be a shapefile.
† Files generated are located in C:\temp. These files can be used directly as input in a GIRAFFE
Monte Carlo simulation.
16
C.3
VISUALIZING GIRAFFE RESULTS IN MANIFOLD
After running a GIRAFFE simulation, the results can be visualized in Manifold. This
allows the user to create maps for illustration purposes and to observe spatial patterns that may
not be obvious in data tables.
C.3.1. Using the Manifold System Visualization Add-In
An additional custom Manifold add-in exists to help the user visualize GIRAFFE outputs
in a meaningful way. To use the visualization add-in,
•
copy the LADWP folder from Manifold Tools in the GIRAFFE program folder
(normally C:\Program Files\Cornell University\GIRAFFE\Manifold Tools)
and paste it in the Config folder hierarchy for Manifold (normally C:\Program
Files\Manifold System\Config).
•
Launch Manifold and open the necessary files or a saved map containing the files.
If the toolbar does not automatically appear, go to Tools | Add-Ins | Add-In
Manager as shown in Figure C.11(a).
o A window will pop up, and check the box next to LADWP GIRAFFE
Output Visualization. Restart Manifold as directed.
•
The custom output visualization add-in toolbar should now be visible.
17
(a) Step 1
Figure C.11. Using the Manifold Visualization Add-In
(b) Step 2
Figure C.11. Using the Manifold Visualization Add-In
18
(c) Step 3
Figure C.11. Using the Manifold Visualization Add-In
19
Each of the eight buttons in the custom toolbar performs a different function for
visualizing the data. Table C.2 describes what each button does and the files required to run each
tool.
Table C.2 Using the Manifold Visualization Custom Toolbar
Icon
Name
Function
Import GIRAFFE Files
Imports GIRAFFE output files with
*.out extension. This tool can only
import one file at a time. Default folder
is C:\Program Files\GIRAFFE.
View Pipe Flow at Time Categorizes pipes by flow: unknown
0
flow, no flow before and after
earthquake, and flow. A new attribute,
Flow_Category_0, is added to
Epa_pipes.
View Pipe Flow at Time Categorizes pipes by flow: unknown
24
flow, no flow before and after
earthquake, and flow. A new attribute,
Flow_Category_24, is added to
Epa_pipes.
View Unsatisfied
Selects unsatisfied demands at demand
Demands at Time 0
nodes immediately after earthquake
damage. The user can then copy
selected nodes into a new drawing.
View Unsatisfied
Demands at Time 24
Node Serviceability at
Time 0
Node Serviceability at
Time 24
View Pipe Damage
Selects unsatisfied demands at demand
nodes 24 hours after earthquake
damage. The user can then copy
selected nodes into a new drawing.
Creates a new layer containing
serviceability data for all demand nodes
immediately after earthquake damage
(time 0).
Creates a new layer containing
serviceability data for all demand nodes
24 hours after earthquake damage (time
24).
Creates two new layers showing pipes
with breaks and leaks. The points are
located at the centroid of each pipe and
don’t represent the exact location of
damage.
20
Required Files
None required.
• LinkResults_Time0.out
• Epa_pipes
• Pipes_NoFlow_BeforeEQ
• LinkResults_Time24.out
• Epa_pipes
• Pipes_NoFlow_BeforeE
Q
• Serviceability0.out
• epa_junctions
• Serviceability24.out
• epa_junctions
• Serviceability0.out
• epa_junctions
• Serviceability24.out
• epa_junctions
• Epa_pipes
• Damage_Info_Dert*.txt
A sample Manifold project (manifold_vis.map) has been included on the GIRAFFE
installation CD in the Example Data folder, which includes the required files in the appropriate
format. Each tool requires different files, which should be imported into a Manifold project and
named exactly as shown unless otherwise indicated. Explanations of GIRAFFE output files can
be found in Appendix A and in the User Manual.
•
Epa_pipes.shp – represents pipe data for entire system. See Figure C.2 for table
format and Section C.2.1. Importing and Formatting Required Data in Manifold
GIS for detailed explanation of file contents. Note that pipes should not be
segmented as for the system damage toolbar.
•
epa_junctions.shp – represents junction data for entire system. See Figure C.1 for
table format and Section C.2.1. Importing and Formatting Required Data in
Manifold GIS for detailed explanation of file contents.
•
Pipes_NoFlow_BeforeEQ.shp – represents pipes that had no flow before the
earthquake. Table format should be the same as Epa_pipes.shp and the layer name
should be exactly as shown. This layer is created by running a stochastic
hydraulic simulation without any damage to pipes (i.e. all pipe repair rates are
equal to zero in the pipe damage input file) and without nodal calibration, and
then using flow values from LinkResults_Time0.out to select pipes with flow
equal to zero.
When using the visualization tools, all GIRAFFE output files should be from the same
simulation folder of the GIRAFFE results or, if the results are not from the same simulation
folder, the user should clearly state this on the map. To assist the user in using only results from
the same simulation folder, the View Unsatisfied Demands and View Pipe Damage tools both
prompt the user to input the simulation number (i.e. folder number from the GIRAFFE results).
Therefore, when importing GIRAFFE results into a Manifold project it is vital that the user
ensure that all output files are from the same simulation.
When using the visualization tools for analysis, it is important to keep in mind that the
maps will only show a snapshot of what the system could look like after an earthquake. Due to
21
the stochastic nature of the process, there will be some variation in results depending on which
simulation is viewed and the parameters specified for the GIRAFFE simulation run.
Before using any of the pipe flow or damage visualization tools, you must first change
the settings in Manifold so that length and area are reported in feet. This is necessary only if the
drawings’ projection is in units of feet (such as State Plane). If the drawings’ projection is in
units of meters (such as Universal Transverse Mercator) then this additional step is unnecessary.
If you do not know the units of the projected drawings, right-click on any of the drawings and go
to Assign Projection. In the pop-up window there is a drop-down menu that displays the units of
the projection. If the units are feet, then go to Tools | Options and, under the Miscellaneous
heading, check the box next to Use English measurements units. According to the Manifold
Help file, if this option is checked, units for projected drawings will be reported in feet. If it is
not checked, units for projected drawings will be reported in meters.
When creating a map of pipe flow it is very important to distinguish between pipes that
had no flow before the earthquake and pipes that have no flow because of the earthquake.
Therefore, the pipe flow visualization tools require a shapefile, Pipes_NoFlow_BeforeEQ,
showing pipes with no flow before the earthquake. To create this file:
•
Run a stochastic hydraulic simulation without any earthquake damage. In the pipe
repair rate input file, set all of the repair rate values equal to zero. Select “No” for
Nodal Demand Calibration.
•
After running the simulation, import LinkResults_Time0.out from the last
simulation. Make a copy of Epa_pipes and name it Pipes_NoFlow_BeforeEQ.
Open Pipes_NoFlow_BeforeEQ Table and go to Table | Relations in the main
Manifold toolbar to create a relation between LinkResults and the layer.
•
Click on the Add Relation button
. In the pop-up window enter the parameters
shown in Figure C.13(a) to match the IDs of the pipes in each file. Click ‘OK’.
You should now see the new relation in the Table Relations window. Put a check
next to the column Flow_gpm as shown in Figure C.13(b). Click ‘OK’.
•
The column Flow_gpm should now be in Pipes_NoFlow_BeforeEQ Table. In the
Query Toolbar, enter the parameters shown in Figure C.13(c) so that pipes with
flow not equal to zero will be selected. Click ‘Select’ then press the Delete key on
your keyboard to delete the selected pipes.
22
•
There should be approximately 2000 pipes left in the drawing. This file only has
to be created whenever pipes are added or removed from the system, or when
flow in pipes is manually shut off.
(a) Step 1
Figure C.13. Create Pipes_NoFlow_BeforeEQ layer.
(b) Step 2
Figure C.13. Create Pipes_NoFlow_BeforeEQ layer.
(c) Step 3
Figure C.13. Create Pipes_NoFlow_BeforeEQ layer.
23
C.3.2 Creating Maps in Manifold System Using Outputs from the Visualization Add-In
Although the visualization tools are generally self-explanatory and easy to use, the pipe
flow tool requires three files and obliges the user to color the pipes based on their flow category.
The four flow categories are:
o Flow – pipes with flow greater than zero after the earthquake.
o Removed – that were removed from the results due to negative pressures.
o Damaged – pipes with damage (leaks or breaks). Flow data can be found
in LinkResults.
o No Flow Before – pipes with no flow before the earthquake.
o No Flow After – pipes with no flow because of the earthquake.
For the user’s convenience, several legends have been saved for easy display of pipe flow
data. The legends are saved as *.xml files, and can be found in the GIRAFFE program folder
under Example_Files | Appendix C. After running the pipe flow tools in the visualization
toolbar, open Epa_pipes and then:
•
Click on the background color well for lines and choose theme from the pulldown color choice menu, as shown in Figure C.14 (a).
•
Choose Flow_Category_0 (or Flow_Category_24) as the Field. Then click on the
Load from File button (Figure C.14(b)).
•
In the pop-up window, navigate to the GIRAFFE program folder and then to
Example_Files | Appendix C. Select Flow0Legend.xml (or Flow24Legend.xml)
and click Open, as shown in Figure C.14(c).
•
Click OK in the thematic formatting dialogue, or change the colors as desired.
The pipes should now be colored according to flow at time 0.
•
Additionally, a legend can be added to the map by going to View | Legend and
checking the Show Legend box. The legend can be customized to suit the user
(Figure C.14(e)). The final map is shown in Figure C.14(f).
•
When visually examining pipe flow, the map should also display valves, pumps,
tanks, and reservoirs in the system to completely and accurately depict pipe flow
in the system. These shapefiles can be found in the GIRAFFE directory under
Example_Files/Appendix D.
24
(a) Step 1
Figure C.14 Displaying Pipe Flow Data
(b) Step 2
Figure C.14 Displaying Pipe Flow Data
25
(c) Step 3
Figure C.14 Displaying Pipe Flow Data
(d) Step 4
Figure C.14 Displaying Pipe Flow Data
26
(e) Step 5
Figure C.14 Displaying Pipe Flow Data
27
(f) Step 6
Figure C.14 Displaying Pipe Flow Data
As noted in Table C.2 Using the Manifold Visualization Custom Toolbar, when the
nodes have been selected using the unsatisfied demands tools, the selected nodes can be copied
into a new layer for use in a map. To do this,
•
Run any of the two unsatisfied demands tools
•
Make sure that epa_junctions is open. Go to Edit | Copy, as shown in Figure
C.15(a), to copy all of the selected objects in the layer.
•
Right-click in the project pane and select Create | Drawing as shown in Figure
C.15(b). Name the drawing. The new drawing should appear in the project pane.
•
Double-click on the new drawing (UnsatisfiedDemands_Time0) to open the
layer. Go to Edit | Paste. Select OK in the Paste Objects pop-up window. The
objects copied from the original file in Step 1 should now appear in the layer as
shown in Figure C.15(h).
28
(a) Step 1
Figure C.15. Creating a new node layer.
(b) Step 2
Figure C.15. Creating a new node layer.
29
(c)
Step 3
Figure C.15. Creating a new node layer.
For the user’s convenience, several legends have been saved for easy display of
serviceability, pipe damage, and PGV data. The legends are saved as *.xml files, and can be
found in the GIRAFFE program folder under Example_Files | Appendix C. After creating node
serviceability layers using the visualization toolbar, open the layer and then:
•
Click on the background color well for points and choose theme from the pulldown color choice menu.
•
Choose Serviceability0 (or Serviceability24) as the Field. Then click on the Load
from File button.
•
In the pop-up window, navigate to the GIRAFFE program folder and then to
Example_Files | Appendix C. Select Serv0Legend.xml (or Serv24Legend.xml)
and click Open.
•
Click OK in the thematic formatting dialogue, or change the colors/intervals as
desired. The nodes should now be colored according to serviceability at time 0.
30
Similarly, the results from the View Pipe Damage tool can also be viewed using premade color schemes. However, in this case legends exist for the shape of the point as well as the
color. After running the pipe damage visualization tool, there should be two new layers called
Breaks Drawing and Leaks Drawing. These show which pipes in the system have breaks and
leaks, and the extent of the damage. Changing the colors and symbols of the two damage layers
is very similar to changing the color theme of the Serviceability and Pipe Flow layers (Figure
C.15):
•
In Manifold, open either Breaks Drawing or Leaks Drawing. Click on the symbol
button (next to the color wells) in the color toolbar and go to Theme.
•
Choose BreakNo or LeakNo as the Field. Then click on the Load from File
button.
•
In the pop-up window, navigate to the GIRAFFE program folder and then to
Example_Files | Appendix C. Select BreakSym.xml (or LeakSym.xml) and click
Open.
•
Click OK in the symbol formatting dialogue, or change the symbols as desired.
•
This can also be done to change the color theme of the two layers, using the saved
legends BreakCol.xml and LeakCol.xml.
•
To create a map of the breaks and leaks, right-click in the project pane and select
Create | Map. Check the boxes next to Breaks Drawing, Leaks Drawing, and
Epa_pipes, and click OK. To create a legend, go to View | Legend in the main
Manifold toolbar, and select the box next to Show legend. The final result should
look similar to the map in Figure C.16.
31
Figure C.16 Map of Pipe Breaks and Leaks
A saved legend also exists for the peak ground velocity (PGV) surface described in
Section C.2 GIS Spatial Analysis and Calculation Using Manifold GIS. This surface
(pgvsurf) can be used as a background for maps in order to demonstrate the location of the
earthquake’s epicenter. If pgvsurf has not already been created for the scenario, refer to Sections
C.2.1 and C.2.3 for how to create it in Manifold. For this example, refer to the example project
32
manifold_giraffe.map, which is located in the GIRAFFE program folder under Example_Files |
Appendix C. To change the colors of pgvsurf:
•
Go to View | Display Options. In the pop-up window, click on the Load from
File button (Figure C.17(b)).
•
Navigate to the GIRAFFE program folder and go to Example_Files | Appendix
C. Select PGVLegend.xml and press Open, as shown in Figure C.17(c).
•
After pressing OK in the thematic color dialogue box, pgvsurf should now appear
as a gradient of oranges and reds (Figure C.17(d)).
(a) Step 1
Figure C.17 Displaying pgvsurf.
33
(b) Step 2
Figure C.17 Displaying pgvsurf.
(c) Step 3
Figure C.17 Displaying pgvsurf.
34
(d) Step 4
Figure C.17 Displaying pgvsurf.
(e) Step 5
Figure C.17 Displaying pgvsurf.
35
Alternatively, the PGV surface can be visualized as contours. These are easy to make in
Manifold:
•
With the pgvsurf layer open, go to Surface | Contours as shown in Figure
C.17(a).
•
Select OK in the pop-up window, or change the intervals if desired (Figure
C.17(b)).
•
Pgvsurf Contours should appear in the project pane. Double-click to open this
layer. Click on the areas color well and select Theme to change the colors.
(a) Step 1
Figure C.17 Creating PGV Contours.
36
(b) Step 2
Figure C.17 Creating PGV Contours.
(c) Step 3
Figure C.17 Creating PGV Contours
37
REFERENCES
Manifold Net Ltd, 2006. Manifold System 7x User Manual. Carson City, NV
Federal Emergency Management Agency (FEMA, 2003). FEMA-450: NEHRP Recommended
Provisions for Seismic Regulations for New Buildings and Other Structures. 2003 Edition,
Washington, D. C., Developed by the Building Seismic Safety Council (BSSC) for
FEMA.
MWH Soft, Inc. (1999). H2ONET User Guide. Pasadena, CA.
Rossman, L.A. (2000). EPANET 2 User Manual. National Risk Management Research
Laboratory, Office of Research and Development, U.S. Environmental Protection
Agency, Cincinnati, OH.
Shi, P. (2006). “Seismic Response Modeling of Water Supply Systems.” Ph.D. Dissertation,
Cornell University, Ithaca, NY.
Wang, Y. (2006). “Seismic Performance Evaluation of Water Supply Systems.” Ph.D.
Dissertation, Cornell University, Ithaca, NY.
38
APPENDIX D
FRAGILITY MODULE
D.1
INTRODUCTION
This appendix provides information on the Fragility Module, which is a part of GIRAFFE
version 4.0. The Fragility Module is available for the “Deterministic”, “Monte Carlo Fixed” and
“Monte Carlo Flexible” options in GIRAFFE (Figure D.1).
Figure D.1. GIRAFFE Version 4 Main Window.
Figure D.2 shows the user interface for the “Monte Carlo Fixed” option, and the check
box for activating the Fragility Module.
Figure D.2. Fragility Module Activation.
1
D.2
INPUT FILES
Once the Fragility Module is activated the user should supply four files delivering the
seismic intensities (peak ground accelerations in units of g) at the locations of system
components including tanks, reservoirs, pumps and valves. Figure D.3 shows the user interface
of the Fragility Module.
Figure D.3. Fragility Module User Interface.
The four files containing the peak ground accelerations (in g units) at the locations of the
tanks, reservoirs, pumps and valves are different for each scenario earthquake. Fragility Module
provides these four files for all of the 59 scenario earthquakes. The user can browse to locate the
files corresponding to the selected scenario event. Figure D.4 shows complete input for the
Verdugo Earthquake scenario.
Figure D.4. Fragility Module User Interface for the Verdugo Earthquake.
2
The format of these input files is predetermined. Figure D.5 presents an example file with
the name Tanks_PGA_175.inp, for the peak ground accelerations in units of g at the location of
the tanks using the Verdugo Earthquake scenario (scenario 175).
Scenario 175
ID
PGA_Soil_g
H4050 0.20774774
H4040 0.20774774
H4030 0.197544669
H4020 0.20777472
H4090 0.237838532
H4010 0.23248291
H4160 0.23248291
CC4040 0.399021763
...
Figure D.5. Tanks Seismic Intensity File for the Verdugo Earthquake.
The first line contains the scenario ID for the selected earthquake. The second line
contains the column names. From the third line on, the first column contains the identification
numbers (IDs) of all the tanks in the H2ONet database and the second column contains the peak
ground accelerations (PGA, in g units) corresponding to each tank location. The format of the
pump, reservoir and valve files is similar to that of the tank file.
After providing the necessary input for the Fragility Module the user goes back to the
main window of the “Monte Carlo Fixed” option by clicking the “OK” button (Figure D.4) and
can start the simulation by just clicking the “Run Simulation” button (Figure D.7).
Figure D.7. Monte Carlo Fixed Option with Activated Fragility Module.
3
D.3
OUTPUT FILE
After reading in the necessary files, the Fragility Module performs fragility analyses for
each component and delivers the damaged components (details are given in Section D.5). The
damaged components are passed to GIRAFFE through a file called Fragility_output.inp. This
file is created in the directory where the Fragility Module is located. The Fragility_output.inp
file is also saved in the directory where all other output files are created for the current
simulation given a scenario earthquake. The name and the format of the file are predetermined.
An example for the Fragility_output.inp file for the Verdugo Earthquake scenario is shown in
Figure D.6.
[SCENARIO_NO]
175
[SIMULATION_NO]
1
[PUMPS]
[VALVES]
[TANKS]
SM4010
HH4180
MW4130
MW4140
HP4010
HP4050
GH4000
EH4050
ST4020
ST4060
GH4020
FH4030
FH4040
[RESERVOIRS]
Figure D.6. Fragility Module Output File.
IMPORTANT NOTE: If a component in the system definition input file is not present in the
provided database files (PGA files or tank type file), then the worst case is assumed, that is, the
PGA at the location of this component is assumed to be equal to the highest PGA in all
component PGA databases, and if this component is a tank then it is assumed that this tank is an
4
unanchored concrete tank (resulting in largest failure probability). This situation will happen
when a new component is added to the LADWP 2002 H2ONet model.
D.4
INTEGRATION BETWEEN GIRAFFE AND FRAGILTY MODULE (DETAILS)
The Fragility Module consists of fragility.exe, fragility.ctf and the files installed by the
MCRInstaller.exe. The main file is fragility.exe and is called within GIRAFFE.
D.4.1 Input for the Fragility Module: FragilityReport.fra
The FragilityReport.fra file is created automatically by GIRAFFE and is located in the
same directory with the fragility module (fragility.exe and fragility.ctf files). The file contains the
location and the filenames for the peak ground accelerations (in g units) for each component of
the system (tanks, reservoirs, pumps, valves), the location and the filename of the random
numbers for damage assessment for the scenario earthquake under consideration, and the
location and the filename of the component identity numbers (IDs) in the current system file. The
locations and the filenames for the peak ground accelerations for tanks, reservoirs, pumps and
valves are provided by the user through the user interface of GIRAFFE. The location and the
filename of the random numbers file (RandomFile.fra) and the component ID file
(ComponentFile.fra) are predetermined. The RandomFile.fra and the ComponentFile.fra are
located in the same directory with the fragility module and is recreated by GIRAFFE for each
simulation.
The order of the filenames in the FragilityReport.fra file is fixed, that is, lines 1, 2, 3 and
4 should contain the location and name of files containing ground motion intensities
(accelerations in g units) for pumps, valves, tanks and reservoirs, respectively, line 5 should
contain the location and name of the random number file, and line 6 should contain the location
and name of the component ID file. The fragility module locates these files and reads the
content. Therefore, the format of these files are predetermined.
Following are examples providing the format of the files FragilityReport.fra,
Tanks_PGA_175.inp (the format of the files for pumps, valves and reservoirs is similar to that of
the tanks), RandomFile.fra and ComponentFile.fra for Verdugo Earthquake scenario (scenario
175).
5
FragilityReport.fra:
E:\...\Giraffe\Example files\PGA_Fragility\175\Pumps_PGA_175.inp
E:\...\Giraffe\Example files\PGA_Fragility\175\Valves_PGA_175.inp
E:\...\Giraffe\Example files\PGA_Fragility\175\Tanks_PGA_175.inp
E:\...\Giraffe\Example files\PGA_Fragility\175\Reservoirs_PGA_175.inp
E:\...\Giraffe\Fragility\RandomFile.fra
E:\...\Giraffe\Fragility\ComponentFile.fra
Figure D.8. FragilityReport.fra File for Verdugo Earthquake Scenario.
Tanks_PGA_175.inp: (PGA’s should be in g units)
Scenario 175
ID
PGA_Soil(g)
H4050 0.20774774
H4040 0.20774774
H4030 0.197544669
H4020 0.20777472
H4090 0.237838532
H4010 0.23248291
H4160 0.23248291
CC4040 0.399021763
…
Figure D.9. Tanks_PGA_175.inp File for Verdugo Earthquake Scenario.
RandomFile.fra:
[SIMULATION_NO]
1
[PUMPS]
0.8626
…
0.3562
[VALVES]
0.8230
…
0.0253
[TANKS]
0.7531
…
0.5625
[RESERVOIRS]
0.3339
…
0.5363
6
Figure D.10. RandomFile.fra File for Verdugo Earthquake Scenario.
ComponentFile.fra:
[PUMPS]
H5000
H5140
H5130
…
FH5080
[VALVES]
H6060
H6050
CC6010
…
GH7050
[TANKS]
H4050
H4040
H4030
…
FH4050
[RESERVOIRS]
H4000
H4180
H4060
…
VF5880
Figure D.11. ComponentFile.fra File for Verdugo Earthquake Scenario.
D.4.2 Output of the Fragility Module: Fragility_output.inp
After reading in the necessary files, the fragility module performs fragility analysis for
each component and delivers the damaged components. The damaged components are passed to
GIRAFFE through a file called Fragility_output.inp. This file is created in the directory where
Fragility Module is located. The Fragility_output.inp file is also saved in the directory where all
other output files are created for the current simulation given a scenario earthquake. The name
and the format of the file are predetermined. An example file is provided below.
7
Fragility_output.inp:
[SCENARIO_NO]
175
[SIMULATION_NO]
1
[PUMPS]
[VALVES]
[TANKS]
MW4130
MW4140
HH4190
SM4120
HP4140
SM4010
HP4010
HP4070
ST4020
VF5540
FH4030
FH4040
[RESERVOIRS]
Figure D.12. Fragility_output.inp File for Verdugo Earthquake Scenario.
8
D.5
FRAGILITY INFORMATION FOR TANKS, RESERVOIRS, PUMPS AND
VALVES IN THE 2002 LADWP H2ONET MODEL
D.5.1 Tank types
The water tanks in the 2002 LADWP H2ONet Model are divided into 4 groups according
to their type:
– Group-1: Steel
– Group-2: Concrete, anchored
– Group-3: Concrete, unanchored
– Group-4: Buried
All welded, riveted and bolted steel tanks are assigned to Group-1. Prestressed concrete
tanks and reinforced concrete tanks built after 1950 are assigned to Group-2. Reinforced concrete
tanks built during and before 1950 are assigned to Group-3. All buried tanks are assigned to
Group-4. Section D.5.5 summarizes the name, type, date of construction and Group ID for all
tanks and some in-ground storage facilities (reservoirs, buried tanks and pipes) in the 2002
LADWP H2ONet Model. Please note that another file for reservoirs also exists in the 2002
LADWP H2ONet Model.
D.5.2 Tank damage states
Two damage states are defined, in accordance with the seismic performance analysis of
the whole water system, to characterize the seismic performance of water tanks:
– DS-1: A tank is hydraulically 100% functional during the first 24 hours after an event.
– DS-2: A tank is non-functional during the first 24 hours after a seismic event.
Damaged water tanks can be incorporated in the recovery/restoration model accordingly.
D.5.3 Tank fragility curves
Group-1: The fragility information of the tanks in Group-1 was obtained from O’Rourke
and So (2000). The paper uses one of the five HAZUS damage states to define the seismic
performance of a tank. It is assumed that a tank is in DS-1 (defined above) for HAZUS damage
states 1, 2, 3, and that it is in DS-2 (defined above) for HAZUS damage states 4 and 5. A
9
lognormal curve is fitted to the data (all tanks with HAZUS damage state ≥ 4) pro vided in Table
4 from O’Rourke and So (2000). Figure D.13 shows the data points from the paper, fitted
lognormal fragility curve, 90% confidence bounds and a pseudo-R2 measuring the goodness of
the fit, for steel tanks in DS-2. Please note that defining a lognormal fragility curve for the
HAZUS damage state 4 sets the maximum threshold for damage that impairs functionality, and
is thus associated with the DS-2 defined above. Defining DS-2 in this manner automatically
covers HAZUS damage states 4 and 5.
Figure D.13. Fragility Curve for Steel Tanks (Group-1).
Group-2 and 3: The fragility information of the tanks in Groups 2 and 3 are obtained
from the HAZUS-MH Technical Manual (FEMA, 2006). The HAZUS manual provides the
parameters of lognormal fragility curves for five damage states for several types of water tanks
(Table 8.9). Group-2 and Group-3 defined above correspond to classifications PST1 and PST2 in
the HAZUS-MH Technical Manual, respectively. It is assumed that a tank is in DS-1 for
HAZUS damage states 1 (no damage), 2 (slight/minor) and 3 (moderate), and that it is in DS-2
for HAZUS damage states 4 (extensive) and 5 (collapse). Figure D.14 shows the lognormal
fragility curves for anchored and unanchored concrete tanks in DS-2. These fragility curves are
identical to those for HAZUS damage state ≥ 4 associated with PST1 and PST2.
10
Figure D.14. Fragility Curves for Concrete Tanks (Group-2 and 3).
Group-4: LADWP experience indicates that in-ground storage facilities perform well
under seismic events or need to be characterized on a more detailed, site-specific basis.
Accordingly, it is assumed that a Group-4 tank or reservoir would not fail due to seismic ground
motions. We have used a dummy lognormal fragility curve resulting in zero failure probabilities
for all possible seismic intensity levels. The dummy lognormal fragility curve used in the
fragility module can be easily replaced in the future when and if new data are acquired that can
help delineate a more appropriate fragility curve. Figure D.15 illustrates the dummy lognormal
fragility curve for in-ground storage facilities.
Figure D.15. Fragility Curve for Buried Tanks (Group-4).
11
Table D.1 shows the parameters of the lognormal fragility curves for the 4 tank and
reservoir types defined above. Figure D.16 shows the lognormal fragility curves for the 4 tank
types.
Table D.1. Parameters of Lognormal Fragility Curves for Water Tanks.
Group
1 (steel)
2 (anchored concrete)
3 (unanchored concrete)
4 (buried)
Median (in g units)
1.294
0.950
0.700
10.000
Dispersion
0.387
0.600
0.550
0.001
Figure D.16. Fragility Curves for Water Tanks.
Note: A tank that is not in the 2002 LADWP H2ONet Model is assumed to be in Group-3, which
results in the largest failure probability (hence a conservative assumption).
D.5.4 Modeling Reservoirs as Tanks
In some cases, it is necessary to model a reservoir as a “tank” such that the level of the reservoir
could vary with time. In such a case, the reservoir will have a tank ID that will be subject to the
tank fragility code, and since this tank ID did not appear at the time tank IDs were assigned to
Groups (this is hard-coded as a part of the tank fragility module) it will be assumed to be in
Group-3, which results in the largest failure probability (the most conservative assumption) . In
reality, reservoirs are large, buried structures and should not be subject to tank damage. A trick
12
can be used to prevent these reservoirs (classified as tanks) from being subject to the tank
fragility module.
Open the Tanks_PGA_Scenario#.inp file and add the tank IDs for the
reservoirs being classified as tanks and give each an extremely low PGA value of 0.000001g.
The PGA value cannot be zero due to calculation restrictions. It is not necessary, but would be
good practice to then remove the reservoir IDs from the Reservoirs_PGA_Scenario#.inp file
since it now appears in the Tanks_PGA_Scenario#.inp file. Before each simulation, be sure you
are using the appropriate tank and reservoir PGA files that capture what you are actually
modeling.
During
each
simulation
using the tank
fragility module,
check
the
Fragility_output.inp file created to see which tank IDs are being damaged. If a reservoir you
were trying to prevent from being damaged is appearing as damaged with each simulation, you
have a clue to double check the PGA input files you are using.
D.5.4 Other components (pumps, valves, reservoirs)
Experience indicates that pumps, valves and reservoirs perform very well under seismic
events. Hence the dummy lognormal fragility curve used for the buried tanks, resulting in no
damage, is used for these components as well. Similarly, if new data are acquired that can help
delineate a more appropriate fragility curve, the new curve can be easily implemented in the
module.
13
D.5.5 A summary of LADWP tanks and in-ground storage facilities
H2ONet ID
H4040
H4030
H4010
H4160
CC4040
SY4010
WS4010
SM4030
SM4040
SM4090
SM4100
SM4230
SM4050
SM4080
HH4000
EH4010
SY4040
SY4050
SY4060
SM4060
SM4110
SM4120
SM4150
SM4280
HH4010
HH4050
HH4060
HH4090
MW4070
MW4080
SM4000
SM4010
SM4020
EH4000
EH4040
EH4080
CC4750
HH4170
HH4180
HH4190
MW4000
HP4040
HP4140
HP4150
MW4130
MW4140
MW4150
HP4010
HP4020
HP4050
HP4070
GH4000
Name
summerland tank no2
9th street tank
harbor heights tank no2
harbor hills tank
baldwin hills tank
paseo miramar tank
sawtelle tank
blue jay tank
briarcrest tank
lookout mountain tank no1
lookout mountain tank no2
cyprean tank
coldwater canyon tank no2
firenze tank no2
griffith park tank no2
corbin tank
temescal tank
trailer tank
marquez knolls tank
eastridge tank no2
roscomare tank no1
roscomare tank no2
summitridge tank
mountain gate tank
hollywood knolls tank no2
innsdale tank
mulholland tank
tyrolean tank no2
elysian park tank
edendale tank
alta view tank
beverly glen tank no2
beverly ann tank
calneva tank
sepulveda tank
zelzah tank
rowena tank
los feliz tank
toyon tank (north)
toyon tank (south)
ascot tank
lomitas tank
bairdstown tank no1
bairdstown tank no2
mount washington tank no1
mount washington tank no2
meridian tank
highland park tank
verdugo tank
kulli tank
hillmont tank no2
donick tank
Type
welded steel covered
riveted steel covered
welded steel covered
steel
welded steel elevated
riveted steel pipe inclined, buried
welded steel covered
welded steel covered
welded steel covered
welded steel dome covered
welded steel covered
welded steel covered
prestressed concrete covered
prestressed concrete
welded steel covered
welded steel covered
prestressed concrete covered
prestressed concrete covered
welded steel covered
welded steel covered
hewitt type circular conrete
prestressed concrete covered
welded steel
welded steel covered
welded steel covered
riveted steel covered
riveted steel covered
welded steel covered
riveted steel covered
concrete encased steel covered
welded steel covered
prestressed concrete covered
welded steel covered
welded steel covered
welded steel
welded steel covered
prestressed concrete covered
prestressed concrete covered
prestressed concrete covered
prestressed concrete covered
prestressed concrete covered
bolted steel covered
riveted steel covered
welded steel covered
concrete circular covered
reinforced concrete covered
welded steel
hewitt type circular conrete
welded steel covered
concrete covered
welded steel covered
welded steel covered
14
Date
1968
1926
1962
1952
1941
1986
1961
1938
1933
1960
1958
2000
1975
1987
1992
1985
1963
1970
1941
1956
1962
1983
1970
1931
1931
1961
1926
1906
1964
2000
1966
1959
1966
1948
2000
2000
2000
2000
1990
1929
1923-1930
1948
1948
1954
1996
1937
1939
1923
1980
1982
Group
1
1
1
1
1
4
1
1
1
1
1
1
2
2
1
1
2
2
1
1
3
2
1
1
1
1
1
1
1
3
1
2
1
1
1
1
2
2
2
2
2
1
1
1
3
2
1
3
1
3
1
1
H2ONet ID
GH4070
GH4080
EH4050
ST4020
ST4040
ST4060
VF5540
ST4010
ST4030
ST4080
VF5070
GH4060
ST4070
GH4020
FH4030
FH4040
FH4010
FH4020
HH4110
H4050
H4020
H4090
VF4140
SM4070
SM4160
SM4220
SM4140
EH4060
EH4070
VF4270
ST4050
SY4000
CC4250
SY4030
CC4220
CC4230
CC4240
MW4100
WS4020
WS4030
VF5550
CC4280
HP4030
HP4060
FH4000
GH4150
FH4050
Name
kittridge tank no3
kittridge tank no4
topanga tank
higway highlands tank
estepa tank no2
sunland tank
tujunga tank
apperson tank
irma tank
rim canyon tank
clear well tank
lakeside tank no2
sister elsie tank
susana tank
maclay tank no1
maclay tank no2
alta vista tank no1
alta vista tank no2
wonderview tank
summerland res. no1
18th street res.
harbor city res.
laurel canyon res.
firenze res. no1
woodrow wilson res.
eastridge res. (no1)
mandeville res.
winnetka res. no2
winnetka res. no1
north hollywood forebay
redmont res.
pacific palisades res.
franklin res. no2 (lower)
santa ynez canyon res.
ivanhoe res.
silver lake res.
elysian res.
solano res.
stone canyon res. (lower)
stone canyon res. (upper)
encino res.
hazard res.
garvanza res.
eagle rock res.
green verdugo res.
de soto res.
maclay res.
Type
welded steel covered
welded steel covered
welded steel covered
welded steel covered
welded steel covered
riveted steel covered
prestressed concrete covered
reinforced concrete covered
welded steel covered
welded steel covered
welded steel covered
bolted steel covered
welded steel covered
prestressed concrete covered
prestressed concrete covered
prestressed concrete covered
riveted steel covered
welded steel covered
horizontal welded steel pipe, buried
concrete covered sunken, buried
concrete covered, buried
concrete covered, buried
circular concrete sunken covered, buried
concrete covered, buried
sunken concrete circular, buried
concrete covered, buried
concrete covered, buried
concrete covered, buried
concrete covered, buried
concrete sump
excavated concrete lined covered, buried
concrete covered, buried
earth fill dam, A. C.
earth fill dam, asphalt
earth reservoir
earth fill dam, A. C.
earth fill dam, A. C.
concrete lined covered, buried
earth fill dam, natural
earth fill dam, A. C.
earth fill dam, natural
excavated concrete lined covered, buried
concrete lined covered, buried
earth fill dam, A. C.
earth fill dam, A. C.
cut and fill res. covered, buried
concrete lined covered, buried
Date
1973
1973
1936
1958
1964
1938
1993
1929-1951
1953
1956
1986
1954
1956
1990
1992
1992
1929
1954
1941
1934
1921
1929
1931
1941
1931
1950
1950
1957
1950
1920-1951
1929
1982
1970
1906-1952
1908-1953
1903-1943
1904
1921-1956
1954
1921
1902-1918
1902-1907
1953
1953
1941
1917
Group
1
1
1
1
1
1
2
3
1
1
1
1
1
2
2
2
1
1
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
Table D.2. A Summary of LADWP Tanks and In-Ground Storage Facilities.
Note: Another file for reservoirs also exists in the 2002 LADWP H2ONet Model. The reservoir
file includes all the reservoirs in the table above, as well as other reservoirs in the system.
15
D.5.6 Seismic hazard at the location of tanks, pumps, valves and reservoirs
The peak ground accelerations at the location of tanks, pumps, valves and reservoirs are
obtained by scaling the peak ground accelerations at the level of bedrocks (provided by the URS)
by factors depending on the local soil conditions. The scaling factors that are used for this
purpose for different NEHRP site classes are provided in Table D.3.
Correction Factor
Site
Class
A
B
C
D
E
AB
BC
CD
DE
(PGA<=0.15)
0.8
1
1.2
1.6
2.5
0.9
1
1.4
2.05
(PGA >0.15 AND (PGA >0.25 AND (PGA >0.35 AND
PGA <= 0.25)
PGA <= 0.35)
PGA <= 0.45) (PGA > 0.45)
0.8
0.8
0.8
0.8
1
1
1
1
1.2
1.1
1
1
1.4
1.2
1.1
1
1.7
1.2
0.9
0.9*
0.9
0.9
0.9
0.9
1
1
1
1
1.3
1.15
1.05
1
1.55
1.2
1
0.95*
Table D.3. Site Amplification Factors for PGA
Values shown with an asterisk (*) in Table D.3 were not provided in NEHRP Provisions and are
based on judgment (Source: HAZUS – MH Technical Manual).
16
REFERENCES
Federal Emergency Management Agency (FEMA), Multi-Hazard Loss Estimation Methodology
- Earthquake Model. Department of Homeland Security, Federal Emergency Management
Agency, Mitigation Division, HAZUS-MH MR2 Technical Manual, 2006.
O’Rourke, M. J. and So, P., Seismic Fragility Curves for On-Grade Steel Tanks. Earthquake
Spectra, Vol. 16, No. 4, 2000, pp. 801-815.
17
APPENDIX E
FLOW AND NETWORK NONLINEARITIES
E.1
INTRODUCTION
This appendix highlights examples of flow and network nonlinearities that are time
dependent are affected by network modeling choices. Network models that include tanks with
the ability to drain and fill over time can display nonlinear flow characteristics.
Special
consideration should be given to how these types of tanks interact with others nearby, as well as
the selection of the size of time step for simulation. Examples illustrating the importance of
these factors are presented in the following sections.
E.2
VARYING TANK LEVELS
The model created for the 2007 LADWP water system contains two types of tanks which
allow tank levels to vary with time: cylindrical tanks and variable area tanks. A cylindrical tank
is defined by a bottom elevation (feet), initial height of water above the bottom elevation (feet)
and tank diameter (in feet). The tank is assumed to have a constant diameter and the volume of
water (cubic feet) in the tank is calculated at each time step based on the current height of water
and the diameter of the tank. As an alternative to a constant diameter tank, a user may specify a
variable area tank where the volume of water is defined by a curve that relates water volume
(cubic feet) to the height of water in the tank (feet). EPANET linearly interpolates the water
volume based on the user defined curve.
An example of a variable area tank curve for the Los Angeles Reservoir is shown in
Figure E.1(a).
It should be noted that when viewing a variable area tank curve in
AutoCad/H20Net, the x-axis represents the water volume in cubic feet, and the y-axis represents
the water height in feet. In the GIRAFFE input file, the values for the curves are listed under the
heading [CURVES] and an example is shown in Figure E.2(b). The values are listed following
the Curve ID (in this case, “VF04”) with the first number representing the height of water in feet,
and the second value representing water volume in cubic feet.
1
A user may also specify a minimum and maximum height of water (feet) or a minimum
volume of water (cubic feet), and the level of the tank will vary within these boundaries. If the
water height in a tank drops below the minimum level (the tank is empty), the outgoing pipe is
automatically closed and no further water can exit the tank. Similarly, if the water height reaches
the maximum level (the tank is full), the incoming pipe is closed and no further water can enter
the tank.
Figure E.1(a). Variable Area Tank Curve for Los Angeles Reservoir in Autocad/H20Net.
2
Curve
Water Water Volume
ID Height (ft)
(cu. ft.)
Figure E.1(b). Variable Area Tank Curve for Los Angeles Reservoir in GIRAFFE input file.
E.3
SELECTION OF TIME STEP
3
The architecture of the GIRAFFE program makes the size of the simulation time step
critical to obtaining accurate results. GIRAFFE assumes all demands and tank levels remain
constant for the duration of the time step, and only updates these values at the start of the next
time step. Thus, choosing a large time step may obscure some network subtleties that would
otherwise be observed with a smaller time step.
Consider the simple network model shown in Figure E.2. Tank 1 is modeled as a Fixed
Head Reservoir which means the level of water in the tank remains constant. Tank 2 is modeled
as a Cylindrical Tank which allows the water level to vary dynamically. Tanks 1 and 2 are
connected by a pipe with a check valve such that water can only flow from Tank 1 to Tank 2
(water will never flow from Tank 2 towards Tank 1).
At the start of a simulation, Tanks 1 and 2 have the same elevation head, thus no flow
occurs between them for the first time step. As Tank 2 drains with time, the elevation head will
drop below that of the Fixed Head Reservoir, Tank 1, and flow will be induced from Tank 1 to
Tank 2. To illustrate the importance of time step selection, consider two 24-hour simulations for
this network with different time step increments: time step of 6 hours, and time step of 1 hour.
For the purposes of this example, assume Tank 2 starts with 4 million gallons of water, and
supplies a demand of 56,000 gpm.
Tank 1
Tank 2
Check valve
Demand
Demand
4
Demand
Figure E.2. Example Network Model.
24 Hour Simulation with 6 Hour Time Step
GIRAFFE assumes that all tank levels and demand values remain constant for the duration of a
time step. To calculate how much water is lost from Tank 2 during the first 6 hour time step,
GIRAFFE computes:
Tank Volume – [(Demand on tank) x (No. of Hours in Time Step)]
4 mil. Gal. – [(56,000gpm) x (6 hrs)] = -16,160,000 Gal.
Clearly, the tank cannot have a negative amount of water – the above computation shows that the
tank goes dry during the 6 hour time step. The tank goes dry after only 71 minutes in this
example, and as the GIRAFFE simulation continues negative pressure nodes develop around this
dry tank as the demand remains at 56,000gpm and the tank and connection pipes are removed
from the system.
24 Hour Simulation with 1 Hour Time Step
To calculate how much water is lost from Tank 2 during the first 1 hour time step, GIRAFFE
computes:
Tank Volume – [(Demand on tank) x (No. of Hours in Time Step)]
4 mil. Gal. – [(56,000gpm) x (1 hrs)] = 640,000 Gal.
Tank 2 has lost 84% of its original volume, but has not been removed due to negative pressure
occurrences as seen with the 6 hour time step example. At the end of this time step, GIRAFFE
updates all tank levels and demand values. Since Tank 2 has drained it now sits at a lower
elevation head than Tank 1, and in the next time step water will flow through the pipe connecting
the two tanks and Tank 1 will be replenished and remain in-service for the duration of the next
5
time step.
Over the full 24 hours, this interaction continues and Tank 2 is continually
replenished by Tank 1 and all demands remained satisfied.
This example illustrates how important time step selection can be. In large-scale
modeling it is difficult to predict and catch where interactions such as these occur and the
ramifications of not capturing these subtle types of network behavior can be widespread. It is
recommended that the user always select the smallest possible time increment (1 hour) so as not
to miss any important network interactions.
6
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.6 Linearized : Yes Author : yu Comments : Company : Cornell University Create Date : 2011:06:17 08:50:39-04:00 Modify Date : 2011:06:17 08:58:33-04:00 Source Modified : D:20080107193218 Subject : Tagged PDF : Yes XMP Toolkit : Adobe XMP Core 4.2.1-c043 52.372728, 2009/01/18-15:08:04 Metadata Date : 2011:06:17 08:58:33-04:00 Creator Tool : Acrobat PDFMaker 9.1 for Word Document ID : uuid:843f2bbf-2f92-456e-ad07-4da6794b203f Instance ID : uuid:dce94c86-4a9a-4e64-8071-728bde817a2b Format : application/pdf Title : CHAPTER 5 Description : Creator : yu Producer : Adobe PDF Library 9.0 Keywords : Page Layout : OneColumn Page Count : 292EXIF Metadata provided by EXIF.tools