Design considerations for piecewise affine system identification of nonlinear systems

A PieceWise AutoRegressive eXogenous (PWARX) model for the AMIRA DR300 DC motor is identified using clustering techniques available in the Hybrid Identification Toolbox (HIT). The choice of design parameters like magnitude of the noise variance and size of the local dataset are discussed. These parameters influence the quality and performance of the PWARX model. The performance of the PWARX model is compared with that of a linear model. The results show superior performance of the PWARX model especially in the nonlinear regions.