A new evaluation model of a water conveyance channel based on Bayesian theory by integrating monitoring and detection information

Channels are commonly used in long-distance water transfer projects, where landslides, collapses, or erosion may occur in its course of operation; thus, safety evaluation is conducted through monitoring and detection in its key and potentially hazardous areas. However, monitoring and detection information cannot comprehensively reflect the prominent problems of the safety state of the channel in terms of time and space. Therefore, studying how to realize the integration of monitoring and detection information is an important task for the safety evaluations of channels. In this paper, a method of integrating monitoring and detection information based on Bayesian theory is presented. The research shows that the fusion method of gathering monitoring and detection information based on Bayesian theory successfully captures the safety state of high-filling channels, and it can quantify and reduce uncertainty compared with fuzzy theory and the GA-BP neural network. By studying the influence of monitoring information on the safety of the channel, it is found that the horizontal displacement has a greater impact on the safety of the channel than the vertical displacement. A comparison of the results of fusing seven different monitoring points shows that the comprehensive utilization of horizontal and vertical displacement can improve the accuracy of the evaluation results. Compared to the safety coefficient calculated by the actual exploration, the error rate of the GA-BP neural network is 42.7%, and the fusion method based on Bayesian theory is 2.9%. The proposed method based on Bayesian theory can better use the detection information to recognize and understand the rock and soil in advance; hence, the evaluation results are more reliable and consistent with the actual engineering state.