1 A Systematic Exploration of Uncertainty and Convergence of Inverse Transient Calibration for WDSs B. S. Jung 1 and B. W. Karney 2 1 PhD Candidate, Department of Civil Engineering, University of Toronto, 35 St. George St., Toronto, ON, M5S 1A4, Canada; email : jung@ecf.utoronto.ca 2 Professor, Department of Civil Engineering, University of Toronto, 35 St. George St., Toronto, ON, M5S 1A4, Canada; email: karney@ecf.utoronto.ca Abstract Despite over ten years of research into ITC techniques for water distribution systems, many problems remain. One reason for these difficulties is that real water distribution systems invariably have many other uncertainties in addition to the leakage rates and friction factors that are conventionally considered as unknowns. For example, properties such as pipe diameter, wave speed, the possible presence of air, the value of the water demand at the time of the tests, and uncertain measurement accuracy, all add to the complexity and difficulty of obtaining a reliable calibration. The current paper investigates quantitatively how several of these uncertainties deteriorate system calibration, and thus the paper generally considers the necessity of a systematic calibration approach to explicitly include these additional uncertainties during the ITC process. To this end, two evolutionary optimizations, namely Genetic Algorithms and Particle Swarm Optimization, are compared and contrasted during the ITC iterations. The advantage of the evolutionary algorithms is that they help the search to escape from poor local optima in multifaceted and complicate problems and thus to locate a good global (or near-global) optimum. However, even these approaches can often be expected to converge poorly when the full scale of the field problem is reflected in the search space. Introduction Distribution networks are an essential part of all water supply systems. The construction and maintenance of these pipelines system cost many millions of dollars every year. One of the challenging and difficult (but important) issues for water distribution systems is the accurate estimation of the pipe internal roughness (or friction factor) and the detection of both normal demands and leakage in the pipeline system. Finding leaks and calibrating the friction factor is, at least, theoretically possible using inverse methods where the results of measurements are known but parameters of physical system are unknown. In essence, inverse methods seek to determine the physical data that, when input into a simulation model, will yield a reasonable match between measured and predicted pressures and flows in the pipe network. Recently, a calibration method called “Inverse Transient Calibration” (ITC) has been widely investigated and advocated (Liggett and Chen, 1994; Nash and Karney, 1999; Vitkovsky et al., 2000; Copyright ASCE 2005 EWRI 2005 Downloaded 11 Jan 2010 to 142.150.190.39. Redistribution subject to ASCE license or copyright; see http://www.ascelibrary.org