Digital twin-based multi-objective parameter identification for a cable-stayed bridge

Accurate parameter identification is essential for ensuring the structural safety and operational performance of long-span bridges. This paper presents a digital twin-based multi-objective optimization framework for identifying structural parameters of a cable-stayed bridge by matching its frequencies, mode shapes, and deflections. The framework integrates three techniques: (1) adaptive reference point relocation to maintain solution diversity across irregular Pareto fronts; (2) sensitivity-informed variation operators that concentrate exploration on parameters with significant influence; and (3) hybrid local search incorporating quasi-Newton refinement for accelerated convergence. The framework was evaluated through implementation on a case study of a 580-meter-long cable-stayed bridge. Experimental results demonstrate up to a 52% reduction in parameter identification error and 25–35% reductions in total runtime compared with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) and standard NSGA-III. Robustness analysis confirms the framework’s reliability with errors consistently below 3%. This research contributes to advancing bridge condition assessment capabilities, thereby enhancing structural monitoring and optimizing maintenance strategies for critical infrastructure.