DATA-BASED PRE-COMPENSATION OF BALL-SCREW FEED DRIVE DYNAMICS WITH LIMITED MODEL KNOWLEDGE

抄録

This paper presents a novel trajectory prefiltering technique to mitigate unwanted residual vibrations experienced in modern high-speed machine tools. Residual vibrations are triggered when machine axes undergo large accelerations which induce inertial forces that excite the lightly damped structural modes of the machine tool structure. This paper presents design of an IIR (infinite impulse response) filtering technique that filters reference axis motion commands to shape their frequency spectrum and avoid residual vibrations. To facilitate a practical auto-tuning strategy, the trajectory pre-filters are designed based on summation of set of 2^nd order basis functions with complex poles and zeros that are tuned based on the vibration data collected on the machine. Poles of the filter are first assigned approximately with limited knowledge of vibration modes. The numerator of the pre-filters is tuned iteratively through closed-loop tracking experiments via 2^nd order newton iteration and pole location are then fine-tuned through 1^st order iterations. Experimental validations are performed on a cartesian motion system. It is shown that proposed 2-step strategy enables auto-tuning of the pre-filter through closed-loop experiments with minimal human interaction. Residual vibrations originating from ball-screw vibrations are suppressed and trajectory induced command tracking errors are reduced from ~30 to ~10microns.