Noise Tolerance of Online Self-Tuning Mechanism for Model-Free Vibration Controller Based on a Virtual Controlled Object

Abstract

This study presents an adaptive self-tuning vibration control scheme. This method can be constructed without controlled object modelling and the self-tuning scheme has strong noise tolerance. A state equation to design a model-free controller is derived by introducing a virtual controlled object (VCO), which is a single-degree-of-freedom system, between an actuator model and an actual controlled object. A VCO-based model-free linear quadratic regulator (VCO-LQR) is designed using the state equation. The VCO-based adaptive control is established by combining the VCO-LQR with an online self-tuning mechanism based on measured outputs and the simultaneous perturbation stochastic approximation (SPSA). The VCO- and the SPSA-based adaptive vibration suppression method is extremely simple due to the VCO-based controller design and the SPSA- and measurement-based self-tuning scheme. Moreover, this self-tuning mechanism has strong noise tolerance because the SPSA can be driven by noisy loss function measurements. The damping performance and noise tolerance of the self-tuning scheme are demonstrated via vibration control simulations.