Implementation of Evolutionary Algorithms to Parametric Identification of Gradient Flexible Plate Structure

This paper focused on modelling of a gradient flexible plate system utilizing an evolutionary algorithm, namely particle swarm optimization (PSO) and cuckoo search (CS) algorithm. A square aluminium plate experimental rig with a gradient of 30° and all edges clamped were designed and fabricated to acquire input-output vibration data experimentally. This input-output data was then applied in a system identification method, which used an evolutionary algorithm with a linear autoregressive with exogenous (ARX) model structure to generate a dynamic model of the system. The obtained results were then compared with the conventional method that is recursive least square (RLS). The developed models were evaluated based on the lowest mean square error (MSE), within the 95% confidence level of both auto and cross-correlation tests as well as high stability in the pole-zero diagram. Investigation of results indicates that both evolutionary algorithms provide lower MSE than RLS. It is demonstrated that intelligence algorithms, PSO and CS outperformed the conventional algorithm by 85% and 89%, respectively. However, in terms of the overall assessment, model order 4 by the CS algorithm was selected to be the ideal model in representing the dynamic modelling of the system since it had the lowest MSE value, which fell inside the 95% confidence threshold, indicating unbiasedness and stability.