Self-tuning MIMO disturbance feedforward control for active hard-mounted vibration isolators

Abstract

This paper proposes a multi-input multi-output (MIMO) disturbance feedforward controller to improve the rejection of floor vibrations in active vibration isolation systems for high-precision machinery. To minimize loss of performance due to model uncertainties, the feedforward controller is implemented as a self-tuning generalized FIR filter. This filter contains a priori knowledge of the poles, such that relatively few parameters have to be estimated which makes the algorithm computationally efficient. The zeros of the filter are estimated using the filtered-error least mean squares (FeLMS) algorithm. Residual noise shaping is used to reduce bias. Conditions on convergence speed, stability, bias, and the effects of sensor noise on the self-tuning algorithm are discussed in detail. The combined control strategy is validated on a 6-DOF Stewart platform, which serves as a multi-axis and hard-mounted vibration isolation system, and shows significant improvement in the rejection of floor vibrations.

Disturbance feedforward control for active vibration isolation systems with internal isolator dynamics

“…However, since the input to C 0 is a measured acceleration signal, pure integrators are generally undesired, because they may induce the amplification of low-frequency sensor noise and signal drift. To circumvent this problem, the pure integrators will be replaced by n th -order weak integrators H (α,n) as proposed in [17], or…”

“…This section summarizes a MIMO self-tuning control approach presented in [17] that was developed for spring-damper compensation. The block diagram corresponding to this approach is shown in Figure 5.…”

“…According to the frequency range of interest in this paper, n perf is designed as a fifth-order band-pass filter from 1 to 100 Hz. A side effect of the large roll-off in n perf is that high-frequency content is strongly filtered, which prevents convergence problems of w(k) under the presence of non-modeled plant dynamics at high frequencies [17].…”

“…However, to provide an effective feedforward control force an accurate model of the vibration isolator is needed. In [16,17] feedforward controllers are presented which rely on a spring-damper model for the isolator. This model may be suitable for isolators with a relatively high suspension frequency and as such having internal isolator dynamics occurring well beyond the frequency range of interest.…”

“…This paper has two main contributions. The first contribution is an extension of the Multi-Input Multi-Output (MIMO) feedforward controller design in [17]. The extension involves the inclusion of an arbitrary number of pre-selected poles from FRF measurements, while the zeros are obtained from data-based selftuning.…”