MRAC Combined Neural Networks for Ultra-Sonic Motor

Abstract

It is difficult for an ultra-sonic motor (USM) to derive a plant model based on the physical analysis. It is well-known that PID control can be constructed even if there is no plant model. In practice, many PID controllers for USM have been proposed. However, there are limitations of control performance on the conventional fixed-gain type PID control because USM causes serious characteristic changes of the plant during operation and contains non-linearity caused by frictions. It is well-known that a model reference adaptive control (MRAC) is very effective to compensate characteristic changes of the plant. However it is not useful for non-linearity of the plant. Then we propose an improved design scheme of MRAC combined with neural networks (NN). The feature of the proposed design scheme is that an improved architecture of the NN is adopted, as a result a simple calculation expression of the Jacobian is derived.