Filtered-error RLS for self-tuning disturbance feedforward control with application to a multi-axis vibration isolator

Abstract

High performance vibration isolation can be realized by disturbance feedforward control with a self-tuning generalized FIR filter and residual noise shaping. For this application, filtered-error recursive least squares (FeRLS) self-tuning is proposed in a multi-input multi-output context. In comparison to filtered-error least mean squares (FeLMS), FeRLS achieves faster and more uniform parameter convergence without the need of pre-whitening. Efficient implementation is realized by exploiting sparsity in the involved matrices. Feasibility of implementation is demonstrated on a multi-axis hard-mount vibration isolation setup. Experimental results show the better parameter convergence and the ability to track changes in the floor vibration spectrum. A reduction of the transmissibility of floor vibrations up to 40 dB in the frequency range of interest is obtained, reducing vibration power by 90–94% in the 1–300 Hz frequency band in multiple directions.