Intelligent modeling of a piezoelectric tube actuator

Various model-based control methods are currently used in control of piezoelectric tubes, others such as internal model control and model predictive control are anticipated to be employed soon. All these control systems are designed based on black box models. However, systematic black box modeling of piezoelectric tubes has been overlooked in the literature to a large extent or has been presented in a too brief and faulty way. In this article, a novel structure of artificial neural networks is used to model and to assess the nonlinearity of piezoelectric actuators. Apart from nonlinearity, other features of the achieved models like delay time, sampling time, orders as well as system identification process are clearly stated, and more importantly, it is clarified that different definitions of accuracy are needed for different purposes of black box modeling, with change in model features, the accuracy may decrease for one purpose (e.g. predictive control) and increase for another one (e.g. simulation). This highly critical point has never been raised and addressed in modeling of piezoelectric tubes, and a definition of accuracy which suits static systems/models has been widely used in the past to assess models of piezoelectric tubes which are obviously dynamic. Experimental results support the proposed modeling ideas.