A novel computational framework for efficient assembly accuracy prediction based on structure partition and stiffness identification

Assembly accuracy is a critical metric for evaluating the performance of complex mechanical systems. Traditional kinematic tolerance analysis often neglects contact-induced deformations, while monolithic finite element analysis struggles with prohibitive computational costs and frequent convergence failures when handling multi-surface nonlinear contacts in over-constrained structures. This study proposes a computational framework that manages complex system coupling by explicitly decoupling microscopic nonlinear contact mechanics from macroscopic structural analysis through a two-stage linearization strategy. Initially, the assembly is partitioned into discrete components to isolate localized nonlinear behaviors caused by non-ideal surface morphology and material elasticity, which are then encapsulated into compact equivalent stiffness matrices via targeted virtual experiments. These matrices are subsequently integrated into a global matrix structural analysis model where geometric deviations are treated as equivalent nodal loads, thereby transforming the highly coupled nonlinear assembly problem into an efficient linear system for rapid equilibrium resolution. Validation through numerical simulations on rabbet joints and empirical measurements on spindle assemblies demonstrates that this framework achieves prediction accuracy comparable to high-fidelity finite element analysis while reducing computation times from minutes to milliseconds and entirely circumventing the convergence instability typical of multi-contact simulations.