Failure Model And Evaluation Of Electrochemical Migration Of Circuit Board Based On VSG

The vast and boundless sea is a rich energy treasure trove endowed by nature to humanity.When ships work in a marine climate environment for a long time,the circuit board in electronic equipment in wet and salt spray corrosion environments will undergo electrochemical migration,resulting in a decrease in the insulation resistance of the circuit board,insulation failure,affecting the reliability of electronic equipment,and even causing fires and other accidents.Therefore,it is necessary to predict the electrochemical migration failure time to assist the staff in making maintenance or replacement decisions to ensure the reliable operation of electronic equipment.In order to predict the failure of circuit boards electrochemical migration,an in-depth search was carried out based on the machine learning model.In this thesis,aiming at the small sample problem of circuit board failure data acquisition and the defects of the machine learning model itself,how to establish a more accurate machine learning prediction model from data and algorithms was studied.The main work of this thesis is as follows:1)Design circuit board electrochemical migration failure modeling scheme.Firstly,the temperature,humidity,salt spray concentration and the working voltage of the circuit board were taken as the influencing factors and the accelerated aging experiment was designed to collect the electrochemical migration failure time of the circuit board as the data basis for modeling.Secondly,starting from the node design and gradient descent method of BP neural network,the working mechanism and principle of BP neural network were studied,and the network was used as the basic model of the full text.Finally,the evaluation method of circuit board electrochemical migration failure was designed.Mean absolute error,mean square error,root mean square error and coefficient of determination were used as the evaluation indexes of the model.2)A BP neural network prediction model based on virtual sample generation technology was established.The electrochemical migration failure cycle of circuit board is long and the collected data is limited.For this small sample problem,firstly,the multi-distribution overall trend diffusion technique and the VSG technique based on the hypersphere characteristic equation were used to expand the sample size.Secondly,the BP neural network prediction model was established by combining the original sample and the virtual sample.Finally,by analyzing the simulation results,the VSG based on the hypersphere characteristic equation has better performance.Compared to the original BP neural network,the average relative error of this model has been reduced by more than 34.7%.3)A BP neural network prediction model based on sparrow search algorithm was established.In order to solve the problem that BP neural network is easy to fall into local optimum in its own training process,SSA was used to optimize BP neural network.Firstly,particle swarm optimization,firefly algorithm,grey wolf algorithm,butterfly optimization algorithm and SSA were compared on the standard test function,and SSA was selected to optimize BP neural network.Secondly,a BP neural network prediction model optimized by SSA was established.The simulation results show that compared to the original BP neural network,the MAE,MSE,and RMSE of the model are reduced by 34.2%,53.6%,and 31.9%,respectively,and the R~2 is increased by 12.4%..4)Design circuit board electrochemical migration failure prediction software.The BP neural network prediction model based on VSG technology and the BP neural network prediction model optimized by SSA were integrated into the software.The data,simulation results and evaluation indicators were presented in the form of tables and graphics,which provides users with a good human-computer interaction experience.