Damage detection in integrated circuit packages using ultrasonic guided wave and clustering algorithm

Reliable detection of interface delamination is important to ensure the reliability of integrated circuit packages with an efficient heat management. In this paper, a robust detection framework is developed to capture sealant delamination between the integrated heat sink and substrate. Ultrasonic guided wave is utilized as an interrogator because of its ability of detecting damages in small and complex structures. In order to efficiently interpret the ultrasonic profile, the noise is reduced through matching pursuit decomposition while minimizing unnecessary loss of information. Then, an automatic feature extraction algorithm is implemented to extract features including amplitude and time-of-arrival. Furthermore, an unsupervised machine learning algorithm, density-based spatial clustering, is utilized to interpret the extracted features and capture the delamination under different experimental condition. The detection accuracies are compared with existing inspection methods to validate the developed method, and the results from the developed method show good agreements with existing methods while maintaining a relatively high inspection efficiency.