Surrogate Modeling of Multistage Assembly Processes Using Integrated Emulation

Abstract

For the design and optimization of multistage assembly processes, a computationally cheap mathematical model that links design parameters with the final product dimensional quality is highly desirable. We propose a systematic approach to building a surrogate model of simulations of multistage assembly processes. At the heart of this approach is a multiple-input-multiple-output surrogate modeling framework that uses a recently developed integrated emulation technique. The unique feature of this technique is that the surrogate models for multiple outputs are fitted simultaneously. The corresponding experimental design issues are also addressed. The proposed method provides good prediction accuracy and requires minimal physical knowledge of the underlying system. The effectiveness of the method is demonstrated through a case study.