Research On Safety State Evaluation Of Steel Radial Gate Based On HDE And Improved BP Neural Network

As an important infrastructure in water conservancy project,the safety state of steel radial gate directly affects the realization of normal function of water conservancy project.Due to the early construction time and bad working conditions,most of the steel radial gates in service in China are damaged to different degrees,and even can not be used normally.Therefore,it is urgent to systematically evaluate the safety state of the steel radial gate in order to make a reasonable maintenance plan for the gate.For this purpose,the following research works have been done in this paper :1.The safety state evaluation index system of steel radial gate is constructed,and 6 first-level indexes such as strength,stiffness and stability,manufacturing and installation quality,working performance,operation management and corrosion situation are determined.According to each first-level index,18 second-level indexes are divided down.The four safety status grades of the steel radial gate and each index are set up,namely A,B,C and D.2.In order to provide data sample support for real-time safety status assessment of steel radial gate,an off-line safety status assessment model for gate was established in view of the characteristics of periodic gate inspection.The structural entropy weight method is used to determine the index weight coefficient in the model,which combines the advantages of subjective weight and objective weight.Considering the universality of the model,the harmony degree equation(HDE)is used to quantify the indexes and determine the safety state of the steel radial gate.3.Aiming at the problem that the standard BP neural network model has low reliability in the safety state evaluation of steel radial gate,an online safety state evaluation model of gate based on IG-SFPA-BP network was established.In the model,the input characteristics of BP neural network were determined by calculating the information gain(IG)of each index in the safety state evaluation index system of steel radial gate,so as to reduce the influence of redundant features.The self-adaptive flower pollination algorithm(SFPA)is used to optimize the initial weights and thresholds of BP neural network,prevent BP neural network from falling into local optimum,and improve the applicability of BP neural network to identify the safety state of steel radial gate.4.Combined with engineering examples,the off-line safety state evaluation model of steel radial gate is analyzed.The results of multiple evaluation groups are consistent with the actual inspection results,which verifies the reliability of the off-line evaluation model.According to the actual engineering data and the results obtained from the offline evaluation,after many experiments and comparisons,the IG-SFPA-BP network model has higher accuracy,shorter running time and better stability than other network models,which verifies the accuracy and applicability of the online evaluation model.On the basis of the above experimental results,the remote operation and maintenance system of hydraulic mechanical equipment realizes the gate safety state evaluation function according to the two safety state evaluation models.