Research On Fault Diagnosis Of Electric Gate Valve In Nuclear Power Plant Based On CEEMDAN And Bi-LSTM

As an important equipment in nuclear power plants,valves are mainly used to open and close pipelines,control flow direction,adjust and control parameters of conveying medium,etc.Due to the harsh working environment and high frequency of use of valves in nuclear power plants,valves have a higher probability of failure compared with other equipment.According to statistics,valve failure occupies a large proportion of nuclear power plant shutdown factors,which not only limits the economic benefits of nuclear power plants,but also greatly increases the threat of radioactive leakage from nuclear power plants.Therefore,the electric gate valve is taken as the research object in this paper,and the electric gate valve experimental bench is built to collect the experimental data of the electric gate valve.On the basis of the above work,the key technologies of electric gate valve signal processing,fault diagnosis and fault degree evaluation are studied,and a set of electric gate valve fault diagnosis and fault degree evaluation system is developed,and the following research work is carried out:(1)An electric gate valve test bench is built,and three types of electric gate valve faults are set up: electric gate valve three-phase imbalance fault,electric gate valve sealing packing damage fault,electric gate valve internal leakage fault.According to various fault occurrence mechanisms and fault characteristics,the sensor is arranged to complete three types of fault data and normal state experimental data collection.(2)The Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)algorithm is used to reduce the noise of the vibration signal of the electric gate valve,and on this basis,complete the extraction of the time-frequency characteristic parameters of the vibration signal;use SAEU3 H A series of digital acoustic emission detection systems perform acoustic emission signal noise reduction processing and characteristic parameter extraction.(3)A variety of cyclic neural network models are established: Long Short Term Memory(LSTM)deep neural network,Gated Recurrent Unit(GRU)deep neural network,Bidirectional Long Short-Term Memory(Bi-LSTM)deep neural network.Optimize the abovementioned network weights and parameters.On this basis,three deep neural networks are used to realize the fault diagnosis function of electric gate valve,and the Bi-LSTM deep neural network is used to realize the function of evaluating the degree of internal leakage fault of electric gate valve.(4)Taking the electric gate valve as the development object,using the Py Charm platform in the Python 3.6 environment,a system with the functions of extracting the signal characteristic parameters of the electric gate valve,the fault diagnosis of the electric gate valve and the evaluation of the fault degree of the electric gate valve is developed.And verify the accuracy and effectiveness of the system through the offline data of the electric gate valve.The experimental results show that the system developed in this paper has a high accuracy in the classification of electric gate valve faults.In terms of evaluating the degree of internal leakage faults of electric gate valves,the error between the predicted internal leakage flow and the actual internal leakage flow is less than 2%.