Nonlinear modeling of the forward path in a mechanical active control system

Abstract

In most active control applications the estimate of the forward path is closely related to the performance of the controlled system. Nonlinearities in the forward path are most likely to degrade the performance. In the active control of machine tool vibration in a lathe, the plant under control principally consists of two physical parts: amplifier and a tool holder construction with integrated actuators. One applicable actuator type was used and it was based on highly magnetostrictive material. However, these types of actuators generally have nonlinear behavior. To obtain deeper understanding on the dynamic properties of the plant under control and thereby obtain a solid base for the feedback control of the machine tool vibration, nonlinear modeling of the forward path was proposed. Two methods were used: The approach of neural networks and the reverse multiple-input single-output by Bandat have been used in order to describe the forward path as accurately as possible. In the active control system of machine tool vibration in a lathe the forward path is the transfer from the actuator amplifier to the output from the accelerometer mounted on the toolholder shank in the toolholder construction.