Piezomechanics using intelligent variable-structure control

The so-called piezomechanics contain three parts: piezoelectric translator, carriage mechanism, and control system. It is well known that piezomechanics have three drawbacks: (1) it should only be loaded axially; (2) it contains a hysteresis feature; and (3) its expansion is dependent on temperature. The first drawback is tackled by the design of the carriage mechanism. This paper focuses on dealing with the second and third drawbacks by using an intelligent variable-structure control. First, a neural network is employed to learn the dynamics of the piezomechanism. Second, a novel forward control based on the learned model is employed to achieve an acceptable tracking result. Because the tracking performance by a forward control cannot be guaranteed as the system is subject to uncertainties, a discrete-time variable-structure control is synthesized to improve the performance. No state estimator is required for the proposed control. The stability of the overall system is verified via the Lyapunov analysis. Experiments are also presented to confirm the effectiveness of the proposed control