Quantitative Assessment of the Porosity in Ag-sintered Joints by non-destructive acoustic Inspection supplemented with a Deep Leaning assisted Signal Analysis

Owed to adjusted concepts in the fields of electricity generation, mobility and supply, power electronics started to receive increasing interest in the previous decade. This led to developments addressing sustainability, efficiency, power density and reliability, such as novel die-attach technologies like Ag-sintering. The present paper describes the application of non-destructive acoustic inspection supplemented by Deep-Learning assisted signal analysis for quantitatively assessing the porosity in sintered silver layers. Acoustic and reference data were acquired from two independent sets of samples that exhibited a broad range in pore content. For data labelling and verification of the methods performance the physically present porosities were obtained through a quantitative analysis of SEM images recorded at physical cross-sections. With the method described here a prediction accuracy (RMSE) of 0.68% relative porosity was obtained.