Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : September 08, 2024 - September 11, 2024
As an approach to demonstrate the seismic resilience of structures, seismic response analysis has been increasingly used to evaluate structural integrity. Generally, analytical models capable of expressing intricate nonlinear behavior tend to become complex, requiring numerous variables to be set. It has been confirmed that employing metaheuristic optimization methods, such as genetic algorithms, is effective for identifying these variables. However, optimization methods like genetic algorithms necessitate parameter settings, which significantly impact the search accuracy (convergence accuracy of the objective function). In response, we have developed an optimization method (FHPFO) that does not require parameter settings. Its effectiveness has been verified using benchmark functions and engineering benchmarks. However, it remains unverified for the hysteresis models of dampers. Therefore, this paper demonstrates variable identification for a steel damper which exhibits strong nonlinearity and cannot be represented by simple bilinear models. Moreover, the FHPFO was compared with other optimization methods using the highly multimodal CEC2017 benchmark functions. These analytical results reveal that even for variable identifications with strong nonlinearity, appropriate variables can be derived, and the effectiveness of FHPFO was confirmed in the high-dimensional and highly multimodal CEC2017 benchmarks.
