open access

https://hdl.handle.net/2115/100447

Bayesian Model Calibration Using Temporal Process Features and Streamlined Active Learning

英語

Electro-thermomechnical model

2026-07-01

Proceedings of the 10th International Symposium on Reliability Engineering and Risk Management (ISRERM 2026)

Accurate multi-physics finite element modelling of resistance spot welding (RSW) remains challenging due to the uncertainties in contact and material parameters that shape the coupled thermal–electrical–mechanical process. This study introduces a Bayesian calibration framework, in contrast to the deterministic parameter optimisation approaches commonly adopted in existing RSW literature, which neglect the inherent variability of the welding pro-cess. The proposed framework performs calibration using multivariate model outputs, combining nugget geometry with geometrical and temporal attributes [1] extracted from process signals such as dynamic resistance and electrode force. These features provide physically interpretable information on process evolution while reducing redundancy and correlation relative to full time-series data, enabling more effective parameter identifiability than nugget-only calibration. The methodology is first assessed on an analytical model using artificial data to evaluate inference accu-racy and computational efficiency. Bayesian inference is carried out using the Streamlined Bayesian Active Learning Cubature (SBALC) method [2], which infers posterior distributions of uncertain model parameters (such as contact thickness, contact resistance, and temperature-dependent material properties) while requiring only a limited number of forward model evaluations.

会議発表論文

Version of Record