Establishment method of water supply pipeline damage prediction model coupling BPNN regression and BN classification

Abstract

The invention discloses a method for establishing a water supply pipeline damage prediction model coupling BPNN regression and BN classification, and belongs to the field of urban water supply pipe networks. Firstly, pipeline damage record data are processed, influence factors of pipeline damage events are determined, then a pipeline damage number regression model is established, pipelines are grouped, and the damage number regression model is established based on BPNN with the damage number of pipeline groups as a dependent variable; meanwhile, a pipeline damage event classification model is established, independent variables are divided into discrete states, whether each pipeline is damaged or not serves as a dependent variable, and the damage event classification model is established based on BN. And finally, coupling the BPNN model and the BN model to carry out pipeline damage probability prediction, and predicting the damage probability of each of the pipelines. The damaged pipeline prediction accuracy is improved, and updating and reconstruction of high-risk pipelines are guided.

Description

technical field [0001] The invention relates to a water supply pipe breakage prediction method by coupling an Error Back Propagation Neural Network (BPNN) regression model and a Bayesian Network (BN) classification model, belonging to the field of urban water supply pipe networks. Background technique [0002] The urban water supply network is the infrastructure for conveying water resources, and the stable and efficient operation of the water supply network provides guarantees for people's production and life. Due to the aging of pipelines and the influence of environmental factors, pipeline damage accidents occur frequently, which not only causes a lot of waste of water resources, but also brings safety problems such as water quality risks and road collapse. However, it is difficult to directly detect the status of water supply pipelines. Therefore, it is of great significance to develop a water supply pipeline damage prediction model, predict the future damage status of t...

AI Technical Summary

Problems solved by technology

[0003] In the research on the prediction of pipeline damage probability based on the classification model, in the research on the damage prediction of water supply pipelines with the damage probability as the output, although the AUC of the model is above 0.7, the false alarm rate of non-damaged pipelines is not high, but in the actual pipeline network The number of non-damaged pipelines is large, and the number of pipelines that lead to false alarms is still large, and it is difficult to effectively guide the actual pipeline transformation

[0004] In terms of the prediction of the number of pipeline damage based on the regression model, the prediction object of the regression model with the number of damage as the output is mostly a group of pipelines, and the prediction results of all pipelines in the same group are the same. High-risk pipelines are grouped. When there are many pipelines in the group, the range of high-risk pipelines will be large, and it is difficult to provide effective guidance for pipeline maintenance.

Images