A Surrogate‐Assisted Hybrid Uncertainty Framework for Robust Quality Design of Multi‐Bolt Rotor Assembly

Abstract

Precision multi‐bolt rotor assembly is governed by coupled geometric deviations, friction‐dependent preload scatter, contact interaction, and tightening‐path dependence. This study develops a surrogate‐assisted hybrid uncertainty framework for robust numerical design of this assembly process. The framework adopts an assembly‐specific two‐stage decision structure in which path‐dependent tightening sequences are first screened, followed by robust optimization of six bolt‐specific torques under probabilistically represented dimensional deviations and interval‐valued friction‐related uncertainties. An SVR surrogate is used to reduce repeated finite element evaluations, while a bi‐level PSO‐MCS implementation evaluates midpoint and radius measures of the coaxiality response. For the representative two‐stage aero‐engine rotor, optimization across 50 samples under the prescribed normal dimensional‐error scenario yielded an average assembly‐accuracy improvement of 16.0%. The framework provides a process‐oriented numerical design route for multi‐bolt assemblies under mixed uncertainty.