Research On Fault Diagnosis Of Safety Braking System Of High-Speed Railway Bridge Erector Based On Optimized Deep Network Model

With the development of China’s railroad transportation network,high-speed rail bridge erector plays a vital role in the process of line construction.As a large bridge transportation equipment,its safety brake system is extremely complex and has many monitoring points,coupled with the different operating conditions,external environment and operating time,the brief manual inspection strategy is often too simplistic and one-sided,which is not conducive to the correct and comprehensive assessment of the operating status and reliability of the bridge erector safety brake system.In view of this,this thesis takes SLJ900/32 integrated bridge erector as the research background,and conducts a systematic review,analysis and summary around its safety inspection technology,in view of the lack of existing research and existing problems,this thesis establishes a continuous delay hidden deep belief network(CDHDBN)based on the continuous delay hidden deep belief network.Network(CDHDBN)fault diagnosis method,and embed the method into the cloud server of the bridge machine safety brake system test bed,and realize the online fault monitoring and diagnosis of the safety brake system.At the same time,it provides certain ideas for the subsequent intelligent,integrated and comprehensive monitoring platform of such bridge erector dynamic system.The specific research work is as follows:(1)The safety braking system of SLJ900/32 integrated transport-frame bridge erector is taken as the research object,the three-dimensional model of test bench is established by Solid Works software,the static analysis of important components is done by ANSYS software,and the test platform of bridge erector safety braking system is built,and the important parameters of the test bench are selected by using the fault tree analysis method.(2)Introduce chaotic algorithm and whale behavior mechanism to improve the Snake Optimization(SO)algorithm and establish an Improved Snake Optimization(ISO)algorithm.The improved algorithm first uses the principle of logistic chaos to initialize the snake swarm with chaotic mapping;then incorporates the behavior of whale bubble network mechanism to update the position of the improved snake swarm;on this basis,adaptive weight factors are added,and the network performance is compared with the swarm intelligence algorithm proposed in recent years to further verify the superiority of the ISO algorithm and to provide an algorithmic basis for the subsequent fault diagnosis model.The algorithmic foundation is laid for the optimization of important parameters.(3)To address the problem of poor accuracy of Deep Belief Network(DBN)in learning continuous real-time data,a continuous delayed hidden layer deep belief network(CDHDBN)model is developed.Firstly,the Continuous Deep Belief Network(CDBN)is introduced to make the discontinuous hidden layer and explicit layer values continuous by adding Gaussian random noise,thus improving the network model’s ability to handle complex data.This makes it easier to detect numerical anomalies when monitoring real-time data.On this basis,the important parameters of the CDHDBN model are optimized to finally obtain the ISO-CDHDBN model,and the network performance is tested with the common data set to further verify the applicability of the ISO-CDHDBN model.(4)Based on the ISO-CDHDBN model,a fault diagnosis method is designed for the test bench of bridge machine safety brake system,and the method is embedded into the data monitoring platform of the upper computer server system.Firstly,the data is transmitted to the server through the lower computer and fault monitoring is performed;then the fault discrimination is performed through the human-computer interface for abnormal values,and the fault results and possible causes of the faults are given.In addition,the network model can be restored or replaced for problems such as missing or abnormal network models.