A Two-Stage Structural Identification Method Using Jaya Algorithm and Gradient-Based Local Search

Numerous heuristic intelligence algorithms have been proposed to solve the optimization-based structural identification, while they suffer from slow computational efficiency and poor identification accuracy with predefined wide search space limits of unknown parameters. To deal with this issue, a two-stage structural identification method, initially implementing global search with Jaya algorithm and search space reduction (JSSR) method, subsequently identifying high-quality potential solution by implementing intensive local search starting from the previous identified best solution, is proposed and utilized in this paper. More specifically, in the former stage, Hammersley sequence sampling generates uniform initial population to preliminarily explore the whole solution space, and then the neighborhood of the global optimum is coarsely determined with Jaya algorithm in the reduced search range by search space reduction method. In the latter stage, the identified best solution is fine-tuned to the optimum with gradient-based sequential quadratic programming (SQP) method. The performance of genetic algorithm, butterfly optimization algorithm, improved butterfly optimization algorithm and proposed two-stage method (JSSR-SQP) are investigated by numerical case on a 16-element simply-supported beam structure for comparative study. The final identification results demonstrate that the proposed two-stage identification method can achieve much better identification accuracy with less computational resources than other three methods in the relatively large search space.