A novel multi-stage joint inversion approach for mechanical parameter determination of complex load-bearing structures

Complex load-bearing structures often exhibit load-response heterogeneity, where different structural parts are governed by distinct dominant loads, posing significant challenges to the reliable inversion of mechanical parameters. In this study, a novel multi-stage joint inversion strategy integrated with an Improved Particle Swarm Optimization-Backpropagation Neural Network (IPSO-BPNN) model is proposed to invert the mechanical parameters of complex load-bearing structures. The IPSO-BPNN model enhances global search capability and inversion stability by introducing a stochastic cognition term and a linearly decreasing inertia weight into the conventional PSO framework. To account for the heterogeneous deformation mechanism, multiple inversion stages dominated by different primary controlling loads are identified, and the corresponding stage-wise inversion results are synthesized through a weighted averaging scheme. The proposed methodology is validated by benchmark function tests and an engineering-scale case study, demonstrating its optimization efficiency and inversion accuracy. The results indicated that the proposed framework provides an effective solution for mechanical parameter inversion for complex load-bearing structures with heterogeneous load responses.