Research On Fault Diagnosis Of Valve Remote Transmission Mechanism In Nuclear Power Plant

At the valve with high nuclear radiation density in the nuclear power plant,the valve remote transmission mechanism acts as a torque transmission mechanical equipment for opening and closing the valve.When it fails,it will affect the normal opening and closing of the valve,and seriously lead to the shutdown and maintenance of nuclear power plant equipment.Cause economic losses.In this thesis,with the support of the scientific research project of Liaoning Provincial Department of Education ’ Nuclear-grade Mechanical Damper Dynamics Simulation Analysis and Research(LYB201702)’,the steering gear box,bearing and shaft of the valve remote transmission mechanism are taken as the research object.The prototype of the valve remote transmission mechanism produced by Shenyang Xintong Power Station Equipment Manufacturing Co.,Ltd.is used to build a test platform,and the research work on the fault diagnosis of the valve remote transmission mechanism is carried out.The accurate identification and diagnosis of the fault type are realized,and the maintenance suggestions are put forward.In this thesis,the test platform of valve remote transmission mechanism is built based on Lab Windows/CVI software.The fault vibration signals of valve remote transmission mechanism are collected,processed and analyzed.The vibration signal denoising,fault feature extraction and fault type identification in the fault diagnosis process of valve remote transmission mechanism are mainly studied.By collecting a large number of test data and conducting comparative analysis and verification,the identification of five fault types is realized on the basis of time domain analysis,frequency domain analysis,wavelet packet algorithm,BP neural network and genetic algorithm.The effectiveness and feasibility of the method are further verified by MATLAB simulation.Starting from the fault mechanism and signal processing,a series of fault diagnosis studies have been carried out on the entire valve remote transmission mechanism.The specific contents are as follows :(1)The fault mechanism of the valve remote transmission mechanism is explored,and the mechanical structure and transmission efficiency of the valve remote transmission mechanism are elaborated.On this basis,the main forms of typical faults of the key components of the valve remote transmission mechanism,such as steering gear box,bearing and shaft,are analyzed in depth,and the vibration mechanism of gear,bearing and shaft is studied.(2)The fault feature extraction is carried out after denoising the vibration signal of the valve remote transmission mechanism,and the time domain analysis and frequency domain analysis are carried out with reference to the analysis indexes of time domain and frequency domain.In the time domain,the most representative and sensitive time domain feature vectors are determined by comparing and analyzing the time domain feature parameters under different transmission efficiencies.In the frequency domain,the characteristic frequency of the fault signal is obtained by Fourier transform,and the possibility of using it as the fault feature vector of BP neural network is excluded after analysis.In order to reduce the interference of noise on fault feature information,the signal is decomposed and reconstructed by wavelet packet,and the wavelet packet energy entropy feature vector is further obtained.(3)Based on the theory of BP neural network,the fault diagnosis model of valve remote transmission mechanism is established.After setting the basic parameters of BP neural network,the five fault types of valve remote transmission mechanism are coded,and the judgment rules are given.500 groups of time domain feature vector and wavelet packet energy entropy feature vector are respectively input into BP neural network for sample training.After BP neural network test,the results show that the fault type of valve remote transmission mechanism can be judged accurately and quickly.(4)Based on the genetic algorithm,the weights and thresholds of the BP neural network fault diagnosis model are optimized.When the input is the time domain feature vector and the wavelet packet energy entropy feature vector respectively,the fault diagnosis rate is increased to 97.3% and 98.7%,and the optimized average absolute error is small.It has good practical application value.