Failure Diagnosis Of Board-level Solder Joints Based On Digital Twin Under Multi-physics Fields

With the rapid development of high-tech,technological innovation is driving electronic devices towards intelligent and precision-based directions.In electronic packaging devices,chips are mechanically fixed and electrically interconnected with the circuit board through solder joints.Therefore,real-time monitoring of the service status of solder joints and providing predictive maintenance for electronic packaging devices is of great significance in ensuring the normal operation of electronic equipment.During the actual service life of electronic devices,solder joints often undergo the influence of force and thermal coupling,which can cause different failure modes of solder joints under different loading coupling conditions.This paper takes board-level solder joints as carriers and uses data-driven digital twin technology to construct a failure solder joint diagnosis system,which provides online intelligent diagnosis and life prediction for solder joints under force and thermal coupling environments.To build a complete failure solder joint diagnosis system and implement functions such as online diagnosis and prediction,based on the concept and operating logic of the digital twin five-dimensional model,a five-dimensional model of the diagnosis system is established,and a system of diagnosis construction is designed to guide the specific process of online diagnosis of failure solder joints.A vibration-temperature coupling test is designed to simulate the force and thermal service environment of the solder joint,providing a data source for collecting training samples for the diagnosis system and verifying the effectiveness of the diagnosis model.When diagnosing the failure mode of solder joints,a problem of complex data types in the sample set is occurred.To solve this problem,a failure diagnosis model based on the idea of semi-supervised learning algorithm is proposed according to the characteristics of data types.Based on historical sample data,the model uses the idea of semi-supervised learning and the strategy of label propagation to fully utilize the mutual dependence of sample labels under similar feature distributions,generalize the limited label information to the maximum extent,and predict the label information of unknown data,that is,diagnose the failure mode of solder joints.In addition,a life prediction model is constructed to induce and statistically diagnose the derivative data generated during the operation of the diagnosis model,providing life prediction function for the solder joints in service.Among them,the diagnosis results of the failure diagnosis model reveal the failure modes of solder joints in different packaging positions and different load environments,combined with material characteristics,verifying the reasons for the change of failure mode of solder joints.In the mean while,under the same vibration-temperature coupling test environment,different alloy composition solder joint life comparison groups are designed to verify the influence of different alloy compositions on solder joint life.The predicted results of the diagnosis model and the life model are compared with the experimental results,the accuracy of diagnosis reaches 83.3% which shows that the diagnosis model has accurate prediction.An analysis and research were conducted on the key technologies and diagnostic mechanisms in digital twin based failure diagnosis systems.Response data generated from solder joints during coupling tests were collected by sensors and uploaded to the failure diagnosis system.The system was able to monitor the real-time loading environment and the operational status of the solder joints,and drive the diagnostic model to dynamically run,performing real-time diagnosis of failure modes and predicting their lifespan.The prediction results were visualized through the user interaction platform of the diagnostic system.The proposed method provides an online intelligent diagnosis and lifespan prediction system platform for entities operating under loading environments,guiding test personnel in making informed decisions.