A Machine Learning Model for the Detection of Solder Voids with Adjacent Sensors

In this work, a machine-learning model is trained for the detection of voids in the solder layer underneath a power semiconductor based on the readings of temperature sensors adjacent to the device. For this, multiple configurations of a nearest-neighbor classification model are trained and tested with data obtained from the measurement and simulation of a test arrangement with artificial faults. A grid search with cross-validation is used to propose an optimized model for the given task, based on different scoring metrics. Additionally, the success of the classification is evaluated in the context of the application case.