A Nonlinear Model for Online Identifying a High-Speed Bidirectional DC Motor

Abstract

The modeling system is a process to define the real physical system mathematically, and the input/output data are responsible for configuring the relation between them as a mathematical model. Most of the actual systems have nonlinear performance, and this nonlinear behavior is the inherent feature for those systems; Mechatronic systems are not an exception. Transforming the electrical energy to mechanical one or vice versa has not been done entirely. There are usually losses as heat, or due to reverse mechanical, electrical, or magnetic energy, takes irregular shapes, and they are concerned as the significant resource of that nonlinear behavior. The article introduces a nonlinear online Identification of a high-speed bidirectional DC motor with dead zone and Coulomb friction effect, which represent a primary nonlinear source, as well as viscosity forces. The Wiener block-oriented nonlinear system with neural networks are implemented to identify the nonlinear dynamic, mechatronic system. Online identification is adopted using the recursive weighted least squares(RWLS) method, which depends on the current and (to some extent) previous data. The identification fitness is found for various configurations with different polynomial orders, and the best model fitness is obtained about 98% according to normalized root mean square criterion for a third order polynomial.