Health Assessment Of Urban Water Supply Network Based On GA-BP Neural Network

In recent years,water supply pipeline explosion accidents occur frequently all over the country,which not only causes a waste of water resources,but also has a great impact on people’s production and life.Due to the invisibility of underground pipe network,the actual water supply pipe network system is complex.Therefore,how to scientifically and effectively predict the health state and pipe burst probability of water supply network is very important.There are complex nonlinear relationships among the factors of water supply network.Traditional prediction methods have the problems of strong subjectivity and high data accuracy,resulting in great limitations of the existing prediction models.Because BP neural network has significant advantages in dealing with nonlinear problems,its application background is highly consistent with the nonlinear system of water supply network health assessment system.Therefore,taking Bengbu water supply network system as the research object,this paper first puts forward the health evaluation index system of urban water supply network,then establishes the health evaluation model of water supply network based on BP neural network.Finally,aiming at the problems that BP algorithm is easy to fall into local optimization and depends too much on the initial value,genetic algorithm is further used to optimize BP neural network.The main research contents are as follows:(1)According to the health status evaluation index system,combined with SPSS statistical software,the influencing factors of the health status of urban water supply network are analyzed in detail.Through feature selection,the mapping relationship between the six input indexes of pipe,pipe diameter,pipe age,water pressure,pipe buried depth and season and the output index of water supply network health state is obtained.(2)Combined with MATLAB simulation software,the health status evaluation model of urban water supply network is established based on BP neural network.The sample set is designed,and 400 pieces of data are used as the sample set according to historical maintenance records and other relevant data.The model function,number of neurons and training step size are selected.By comparing the mean square error and step size,the optimal model function is established.The training function trainlm,learning function learngdm and hidden layer transfer function logsig,the number of neurons in 10 optimal hidden layers and the optimal training step size of 200 steps are selected.The experimental results show that the mean square error is 0.0537 and the prediction accuracy of pipe burst is 83.5%.(3)In order to solve the problems of BP falling into local optimization and relying too much on the initial value,the BP neural network is optimized by genetic algorithm.The optimal weight and threshold are selected through coding,selection,crossover and mutation,and assigned with the original BP neural network.The experimental results show that the mean square error is reduced to 0.0379 and the prediction accuracy is improved to91.3%.Experiments show that GA-BP neural network model is feasible and reliable for predicting the health state of underground water supply network in the city,and provides a guiding basis for subsequent reconstruction projects.