Online Prediction And Impact Wear Analysis Of The Dynamic Contact Of The Nuclear Main Pump Heat Exchange Tube Based On Deep Learning

In nuclear power system,the heat exchanger of nuclear main pump is one of the key equipment.Due to the heat conduction in the primary and secondary circuits of the heat exchanger and the vibration caused by high temperature and high pressure medium flow,fretting wear occurs between the heat exchange tube and the supporting parts,resulting in local damage or even rupture of the heat exchange tube,reducing the service life of the heat exchanger,endangering the safety of nuclear power.Therefore,to prevent the damage of heat exchange tube and improve the safety and service life of nuclear power equipment is a major issue of nuclear power engineering.The simulation study of fretting wear under different normal alternating loads is not only of great significance to explore the mechanism of complex fretting damage under special working conditions,but also can provide theoretical support and engineering practice guidance for anti fretting damage design and operation safety of nuclear power equipment.In this paper,the numerical simulation and vibration experiment are combined,and the neural network prediction model is constructed to predict and analyze the vibration impact force.In the simulation software,the simulation is carried out according to the actual experimental conditions,and the stress,strain and wear of the heat exchange tube at the impact are analyzed.When the heat exchanger is in normal operation,the vibration displacement signal of the heat exchange tube can be obtained by measurement,and the collision force and clearance between the tube and plate will change with the change of wear,so it is impossible to accurately estimate the wear condition in real time,which leads to the occurrence of safety accidents.Due to the uncertainty of the clearance between the tube and plate and the stiffness of the constraint end in the process of impact wear,this paper studies the following two main aspectsFirst,the vibration model of the heat transfer tube is tested and simulated.The heat transfer tube and the baffle are installed on the test bench of the vibration exciter.The exciting force is applied to the tube to make it collide with the baffle.The displacement and exciting force data collected by the sensor are compared with the calculation program in MATLAB to verify the reliability of the calculation program.The neural network model for the analysis and prediction of the collision force between tubesheets is constructed,and the collision force is predicted according to the vibration displacement signal,so as to realize the online prediction of wear.Taking the impact displacement as the input and the average,maximum and minimum impact force as the output,different neural network structures are used to analyze the accuracy of the model during the training process.The results show that the network model can effectively and accurately predict the impact force between the heat exchange tube and the baffle,and different network structures have different prediction performance,And the prediction accuracy of samples is also different.For the sample types in this paper,convolutional neural network has higher prediction accuracy than fully connected network.Secondly,the finite element method is used to simulate the impact wear in the workbench software,and the deformation,stress and strain distribution and numerical value of the impact position between the heat exchange tube and the support plate under different loads are obtained.Based on the Archard wear model,the impact wear of the heat exchange tube and the baffle plate under different conditions is calculated by using the simulation software,The accuracy and reliability of the finite element model are verified by comparing the simulation results with the experimental results.