Hybrid structural identification strategy suitable for distributed computing

System identification is a process of determining parameters of a dynamic system based on given input and output signals. This serves as a useful tool in structural health monitoring and damage assessment in a non-destructive way. A soft computing approach known as Genetic Algorithm (GA) is employed. Through numerical simulation study, GA-based identification method is proven to be efficient and robust with no requirement of good initial guess. Nevertheless, GA is not efficient in fine-tuning, i.e. the convergence becomes slow when near the optimal solution due to the stochastic nature of the algorithm. Hence it is necessary to embed a local search operator in GA in order to enhance convergence. A new local search operator called the extrapolation-interpolation operator (EIO), which conducts multidirectional search locally, is proposed in this study. This is a useful plug-in operator without compromising the advantages of GA. The operator is simple to implement and yet improves the convergence significantly. A good balance between convergence through interpolation and diversity through extrapolation can be achieved, thereby reducing the risk of premature convergence. Furthermore, the proposed hybrid strategy is suitable for distributed computing which can be used to expedite the structural identification process. Numerical examples are presented to illustrate the performance of the proposed strategy.