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