AI-assisted structural defect detection with high resolution using acoustic data (LIUD_U26EMP)

Reliability analysis and failure modelling is crucial and very challenging for modern electronics, especially safety critical electronics-based systems working under harsh environmental conditions. To establish a reliable failure model for solder joints which is urgently needed for prognostic and health management of electronic packaging, it is necessary to identify structural defects at the micro level scale so that the accurate life expectancy of solder joints can be predicted. However, it is challenging to inspect the defect at its early stage for examples, within 20um, due to physical limitation of non-destructive testing.

Therefore, an effective approach to defect detection with a high level of resolution is a critical technique to be developed for reliable life prediction. In this project, AI empowered framework for high-resolution defect inspection of solder joints is proposed using experimental acoustic data and finite element simulations. As time constraints for cycle life prediction of soler joints in the experiments, a physics-informed neural network model will be developed to leverage the acoustic information of other solder joints as data argmentation for the joint of the interest from an experimental point of view.

Meanwhile, fine finite element models for joints will be used to identify the latent relationship between the one of the interest and other joints in the same Printed Circuit Boards (PCB) so that experimental time and costs will be significantly reduc