Research On Status Monitor And Fault Diagnosis Of Electric Gate Valve Based On AE And XGBoost

In industry,valves are often used as mechanical equipment to control the transport of working fluids.Nuclear power plants have complex structures and numerous equipment.There are a large number of different types of valves in nuclear power plants,which have functions such as controlling fluid transport and protecting system safety.Due to the harsh working environment and frequent switching of the valve,it is easy to cause the performance of the valve to degrade,and the corresponding task cannot be completed well.According to statistics,in nuclear power plants,the cost of valve maintenance accounts for about half of the total cost,of which shut-off valve failures such as gate valves account for about half of the total valve failures.The performance of the valve will affect the safety of nuclear power plant operation to a certain extent.The electric gate valve is used as the research object in this paper,some common faults of the electric gate valve are simulated by building an experimental platform.On this basis,the experimental platform is applied to collect the normal state and fault state signals of the electric gate valve,and the collected signals are denoised and feature extracted,and then complete the research of state monitoring,fault diagnosis and other technologies.The main contents of this paper are as follows:In this paper,an experimental bench is built to simulate the three-phase unbalance of the electric gate valve,the loose packing bonnet and internal leakage,etc,and the acceleration sensor and the acoustic emission sensor are used to collect the vibration and acoustic emission signals of the electric valve under normal and fault conditions.The wavelet packet method is applied to de-noise and feature extraction of the collected vibration signal,and the SWAE4 acoustic emission system is used to complete the feature extraction of the acoustic emission signal.In this paper,the fault detection of the valve can be divided into three parts.First,the automatic encoder(AE)is applied to monitor the state of the electric gate valve,the autoencoder is trained using the experimental data of the normal state electric gate valve,and the output of the autoencoder intermediate layer is used as the final state monitoring indicator;then,in view of the imbalance of different types of samples in the industry,this paper uses a generative adversarial neural network to generate sample data and increases the number of minority samples to solve the problem of sample imbalance.Finally,the classification model of the extreme gradient boosting method is adopted to diagnose the fault of the electric gate valve,and the XGBoost regression model is used to evaluate the fault degree of the internal leakage fault of the electric gate valve.The experimental results show that the research method can accurately identify the state of the electric gate valve,and can make a better prediction result for the sample imbalance fault diagnosis and the fault degree evaluation of the electric gate valve.The research content of this paper is the state monitoring and fault diagnosis of the electric gate valve further research on other technologies has laid the foundation.