Muti-Optimization Study On Rehabilitation Of Water Mains

Water distribution systems (WDS) as vital urban facilities play an important rolein the modern life and urban developments. In many cities in China, the operation ofmains suffers from such problems as leakage, pipe failure, pressure fluctuation,contamination and so on, which results in reduced level of service. Research thesubject of rehabilitation strategies for mains do really make sense. A decision supportsystem for rehabilitation policy-making is formulated in this thesis, with operationsresearch, system engineering, reliability theory, optimization theory, mathematics andcomputer technology involved flexibly.Optimal sort of pipes to be restored in mains is illustrated. In this section,pipesare divided into two sets. Primary set is retrieved from the candidate set by topologyreliability, and the rest part is so called secondary set. Then total influenced flow andweighted sum of pressure and quality reliability are separately defined as failure strikelevel for the two sets here and programed with EPANET engine. The probability ofpipe failure and the probability of accident are formulated. The two sets are sortedbased on Bayesian decision method. The event score is failure strike level and theprobability is formulated by reliability theory. Finally, optimal rehabilitation sort areformulated based on some provisions such as threshold break rate and corrosion leveland models above. Positive result was derived here.Optimal rehabilitation scheduling for the whole network is illustrated, aiming toget optimal pipe set and size choice. As a muti-objective optimal problem, trade-offsamong total cost, failure strike level of the mains, pressure reliability and qualityreliability are considerd as four independent goals here. GPU parallel computingbased numeration method is structured here first to solve the muti-objective problem.NSGA-II and GPU parallel computing are both used to solve the example problem forcomparison. It was shown that GPU parallel computing shared same good solutionwith NSGA-II but, ran faster than NSGA-II.