Title: A new Bayesian model averaging ensemble modelling method for surrogate-assisted reliability-based design optimisation of FBTAM

Authors: Liangqi Wan; Yumeng Wei; Qiaoke Zhang

Addresses: School of Management Science and Engineering, Nanjing University of Finance and Economics, Nanjing, 210023, China; College of Economics and Management, Nanjing University of Aeronautics and Astronautics, Nanjing, 210016, China ' School of Management Science and Engineering, Nanjing University of Finance and Economics, Nanjing, 210023, China ' School of Marketing and Logistic Management, Nanjing University of Finance and Economics, Nanjing, 210023, China

Abstract: A reliable design of the flexure-based bridge-type amplification mechanism (FBTAM) using surrogate-assisted reliability-based design optimisation (RBDO) greatly relies on a robust surrogate model accounting for the model form uncertainty. At present, surrogate modelling methods have attracted widespread attention in RBDO framework, especially for the ensemble modelling method. However, they suffer from model-form uncertainty in ensemble modelling process due to the lack of information and knowledge, as well as assumptions and simplifications made in models, which leads to an inaccurate estimation. To enhance the robustness of an ensemble modelling method of surrogate-assisted RBDO for the FBTAM, a new Bayesian model averaging (BMA) ensemble modelling method is proposed. The BMA strategy is adopted where the model-form uncertainty is well considered in the model structure before constructing an ensemble model. A typical FBTAM is utilised to verify the effectiveness of the proposed method. Comparison results revealed that the proposed method has a higher robustness and accuracy compared to the existing modelling method, and thus reached a better RBDO design.

Keywords: Bayesian model averaging; BMA; reliability-based design optimisation; RBDO; desirability function; compliant mechanism; model-form uncertainty.

DOI: 10.1504/IJPQM.2026.153637

International Journal of Productivity and Quality Management, 2026 Vol.48 No.1, pp.1 - 19

Received: 13 Oct 2023
Accepted: 24 Jan 2024
Published online: 19 May 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article