Cavity Failure Prediction Of CCGA Solder Joint In Special Vehicle Control System Based On BP Neural Network

As the core component of some special vehicle control systems,the reliability of CCGA(Ceramic Column Grid Array)will be greatly challenged when special vehicles work in various harsh environments.As a common welding problem in CCGA,solder joint void has a significant impact on its reliability.In order to reduce the void rate of solder joints and improve the reliability of products,it is necessary to study the failure prediction of the void rate of solder joints in CCGA,so as to keep the void rate below 15%.As a working mode formed by simulating intelligent animal nervous system,neural network has good intelligence and inherent advantages in product failure prediction.At present,the research on solder joint void rate of array packaging components is mainly focused on BGA(Ball Grid Array,spherical grid array).For CCGA solder joint void rate,the method of combining experiment with experience is mostly used.The use of neural network model to predict the failure of solder joint void rate is innovative,which expands the thinking of solving such problems.It can be seen that the application of neural network to the prediction of the void rate failure of CGA solder joints has high theoretical research value and engineering application value.In this paper,by analyzing the failure mechanism of solder joint voids,two factors which have great influence on the void rate of CGA solder joints are extracted:solder paste thickness and temperature curve.The optimum thickness of solder paste can be obtained by experiment.Considering the influence of preheating time and reflux time in temperature curve on solder joint failure,a two-parameter artificial neural network failure prediction model is established to form a universally applicable prediction method of CCGA solder joint void rate based on neural network.In the field of solder joint failure test,a 5-DOF spatial mechanism scheme is adopted to optimize the test mechanism.The simulation and feasibility demonstration of the 5-DOF spatial mechanism scheme are completed.The CCGA solder joint failure test scheme is designed,and the preliminary distribution characteristics and statistical characteristics of the test data are analyzed.Combining with the data obtained,the training of the two-parameter feature of the neural network model is carried out,and the prediction results of the model are compared with the single-variable feature prediction method which separately analyses the preheating time,reflux time and peak temperature.The validity and superiority of the prediction model are verified.