Visual Detection of Typical Process Defects in Power Modules and Intelligent Quality Assessment via Electroluminescence Imaging

Power electronic systems demand high reliability, yet process defects within direct bonded copper (DBC) substrates significantly compromise insulation integrity under coupled electro-thermal stress. This paper introduces a novel visual intelligent detection method based on the electroluminescence (EL) effect that employs a “working condition separation strategy” to quantitatively assess this risk. First, the electric field distribution is mapped under no-load/room temperature conditions to identify static stress concentrations caused by geometric features and defects. Subsequently, the surface temperature distribution is characterized under a representative operational load (0.8 kV DC bus, 100 Ω load resistance), where our EL-based thermometry model achieves a Mean Absolute Error (MAE) of approximately ±3.2°C, with peak module temperatures reaching 90°C. By integrating these distinct electro-thermal datasets, a fuzzy logic-based model generates a quantitative, 2D insulation risk map. Experimental validation on a commercial 1.2 kV/600 A Insulated-Gate Bipolar Transistor (IGBT) module demonstrates the framework's efficacy. The results precisely identify and differentiate high-risk locations, such as a 63% probability of insulation weakness at sharp corners versus a 23% probability along straight trace edges, and quantify their combined stress state. This visual, spatially resolved approach offers a powerful tool for refined quality control, in-depth reliability analysis, and design validation for power modules.