Modelling the fatigue damage in power components using machine learning technology

Stoyanov, Stoyan ORCID: 0000-0001-6091-1226, Sulthana, Razia, Tilford, Tim, Zhang, Xiaotian, Hu, Yihua, Yang, Xingyu ORCID: 0000-0003-2871-8025, Shen, Yaochun ORCID: 0000-0002-8915-1993 and Wang, Yangang (2025) Modelling the fatigue damage in power components using machine learning technology Power Electronic Devices and Components, 10. p. 100079. ISSN 2772-3704, 2772-3704

PDF 2025-Power Electronic Devices and Components.pdf - Open Access published version Download (7MB) | Preview

Abstract

Thermo-mechanical finite element (FE)-based simulation technology has been used extensively for virtual prototyping and to predict material degradation and thermal fatigue damage in electronics assembly materials. However, from an end-user point of view, the deployment of such high-fidelity modelling is not straightforward as it requires comprehensive device and material characterisation data that is not readily available through technical datasheets and must be gathered using costly and time-consuming bespoke characterisation tests and access to metrology instruments. In addition to that, FE modelling requires access to advanced software and specialised FE skill sets. Here, a novel physics-informed Machine Learning (ML) approach for developing computationally fast metamodels for predicting fatigue damage and its spatial distribution at common failure sites of power electronics components is developed, validated and demonstrated. The significance of this work is in the attributes and the capabilities of the proposed modelling technology that enable the end-users of power components to perform insightful model-based assessments of the thermal fatigue damage in the assembly materials due to different application-specific, qualification and user-defined load conditions, removing current requirements for comprehensive device characterisations and deploying complex finite element models. The proposed methodology is demonstrated with two different metamodel structures, a regression decision tree and a neural network, for the problem of predicting the thermal fatigue damage in wire bonds of insulated-gate bipolar transistor (IGBT) power electronics modules (PEMs) exposed to passive temperature cycling loads. The results confirmed that the proposed approach and the modelling technology could offer FE model substitution and the capability to spatially map highly nonlinear three-dimensional spatial distributions of the damage parameter over local sub-domains associated with material fatigue degradation and failure.

Item Type:	Article
Uncontrolled Keywords:	3403 Macromolecular and Materials Chemistry, 40 Engineering, 34 Chemical Sciences, Networking and Information Technology R&D (NITRD), Machine Learning and Artificial Intelligence
Divisions:	Faculty of Science & Engineering Faculty of Science & Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User:	Symplectic Admin
Date Deposited:	17 Jan 2025 15:51
Last Modified:	02 Jan 2026 15:24
DOI:	10.1016/j.pedc.2025.100079
Related Websites:	Publisher
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3189758
Disclaimer:	The University of Liverpool is not responsible for content contained on other websites from links within repository metadata. Please contact us if you notice anything that appears incorrect or inappropriate.