Research On Vulnerability Assessment And Disaster-bearing Capacity Optimization Of Water Supply Networks

With the increasing development of Chinese modernization construction,a large quantity of population,wealth and infrastructure are concentrated in cities.The quantity of injuries and deaths and property loss would be huge when suffered from extreme damage such as tsunami and earthquake.The water supply networks system is a crucial part of city lifeline engineering.The loss of water supply function when it is damaged not only influences people’s basic life needs,but also has negative impacts on the post-disaster reconstruction and social production.Therefore,the security and reliability of water supply networks is an important reflection of urban disaster bearing capacity.Assessing the vulnerability of water supply networks,finding out the vulnerable parts and further reconstructing these parts have essential significance for the work of disaster prevention and reduction.This paper,based on the research results of relative fields at home and abroad,in the methods of hydraulic calculation of water supply networks,catastrophe theory,complex networks theory,clustering algorithm and optimization algorithm etc.makes deep research towards the vulnerability analysis and optimization of water supply network under earthquake.The major research contents are as follows:First of all,vulnerability refers to the state that the system and its components are vulnerable to damage or exposure.Vulnerability is determined by different elements together.This paper analyzes the source of vulnerability of water supply networks from two perspectives:function and structure.The former one chooses the hydraulic performance index of networks and the damage degree index of components(nodes and pipes)under the effects of earthquake.The latter one chooses the threshold of damaged components and the water supply hierarchical level.Hydraulic performance index is based on the basic hydraulic equation,applying the EPANET to be the calculation engine to find the node water pressure and pipe flow rate.The damage degree index of components applies the graph theory to calculate the non-connection rate and uses catastrophe theory to calculate the damaged pipes index.The critical threshold of nodes and pipes are calculated by the cascading failure models of complex networks theory,and the water supply hierarchical level of nodes and pipes are calculated by breadth-first search algorithm according to the influencing ability of components in networks.Secondly,due to the fact that the source elements of vulnerability are numerous and complicated formed,this paper applies clustering algorithm to analyze vulnerability from multiple perspective based on similarity and classify the components in networks.Then,sequencing different categories from the perspective of vulnerability according to the statistics features of clustering centers.Taking a water supply networks in Zhejiang Province as an example,the paper applies the fuzzy c-means clustering algorithm to classify and sequence the nodes and pipes,which provides basis for the subsequent seismic optimization of the network.The results of fuzzy c-means clustering are compared with the system clustering method and K-means clustering method,and the accuracy of the results is verified.Thirdly,on the basis of finding the weak parts of network,the network design will be optimized.The optimized model takes the pipe diameter as the parameter to be optimized,the cost minimization and the vulnerability improvement maximization as the objective function.The advantages and disadvantages of the optimization methods are compared on two levels:first,genetic algorithm is superior to the traditional optimization method represented by orthogonal test method,and the latter can be regarded as a special case of the former after simplification;second,the fast non-dominated sorting genetic algorithm based on Pareto optimal idea(NSGA-Ⅱ)is superior to the basic genetic algorithm,because when the genetic algorithm sets the fitness function,it often transforms multiple objective functions into single objective function,which may cause single result and deviation.Finally,in order to solve the problem that GA is easy to fall into premature convergence,the method of generating initial population is improved by using orthogonal table which features "balanced disperser and neat comparison",so that the initial population is evenly distributed and covers the solution space as much as possible.Through the comparison between the random initial population and the orthogonal designed initial population,it is found that the performance of the latter is better than the former.