Particle swarm optimization for NARX structure selection: application on DC motor model / Mohd Ikhwan Abdullah

This thesis was presents the nonlinear identification of a DC motor using Binary Particle Swarm Optimization (BPSO) algorithm, as a model structure selection method, replacing the typical Orthogonal Least Squares (OLS) used in system identification. The BPSO algorithm is an evolutionary computing technique put forward by (Kennedy and Eberhart, 1997). By representing its particles technique as probabilities of change (bit flip) of a binary string, the binary string was then used to select a set of repressor as the model structure, and the parameter estimated using QR decomposition. The DC motor dataset was simulated to test the performance of the new model structure selection approach. The findings indicate that the BPSO-based selection method has the potential to become an excellent and effective method to determine parsimonious NARX model structure in the system identification model. The NARX model structure was used to model the system dynamic. The several optimizations on the neural network training have been performed and several regularizations of the training have been considered. Similar investigate in nonlinear identification as done in its linear counterpart. The NARX coupled with suitable regularisations have outperformed the linear models. Even though the linear models are sufficient for this system, the nonlinear model can represent the system better.