Metamodel-assisted hybrid optimization strategy for model updating using vibration response data

In this study, an effective and novel method, termed Metamodel Assisted Hybrid of Particle Swarm Optimization with Genetic Algorithm (MA-HPSOGA), is developed to identify unknown structural dynamic parameters. The method first constructs four popular metamodels to substitute the computationally expensive numerical analysis based on the Latin hypercube sampling method and probabilistic finite element analysis, and their accuracy is assessed by R-squared. Subsequently, a suitable and low-cost metamodel is selected in combination with a hybrid optimization strategy by incorporating Genetic Algorithm (GA) into Particle Swarm Optimization (PSO). Two examples with measured vibration response data and different levels of complexity are used to verify the effectiveness and practicality of the presented method. The results showed that polynomial chaos expansion assisted HPSOGA has the highest computational efficiency and accuracy in the four coupled methods. Besides, compared to the conventional iteration-based dynamic parameter identification methods, the presented method shows an overwhelming advantage in terms of computational efficiency. Furthermore, the performance of HPSOGA is compared with its sub-algorithms, showing that the hybrid strategy offers faster convergence and stronger robustness. Our findings reveal that the MA-HPSOGA may be used as a promising method for achieving high-efficiency model updating in large-scale complex structures.