Digital twin of power modules based on physics informed machine learning

This paper presents a novel digital twin for power modules in electric vehicles (EVs), utilizing physics-informed Artificial Intelligence (AI) to predict thermomechanical stresses. Traditional finite element analysis (FEA) simulations, known for their high computational demands, are replaced with a more efficient AI-based model. By analyzing operational data from EV road tests, the study showcases the ability of a Long Short-Term Memory (LSTM) network to accurately forecast the behavior of power modules under varying conditions. This approach promises significant advancements in power module design and reliability, enabling faster development cycles and real-time health monitoring capabilities.