Linear parameter-varying system identification of an industrial ball screw setup

Ball screws are high-precision positioning systems that translate rotational motion of an electric motor into translational motion of a slide. Due to the distributed stiffness and inertia of the ball screw, the resonance frequencies of the system vary with the position of the slide. This paper discusses the experimental identification of an industrial ball screw setup by treating it as a linear parameter-varying (LPV) system. Application of a novel nonlinear least-squares LPV system identification method with l 2,1 -norm regularization yields a sparse model with satisfactory accuracy both in time and frequency domain and over the whole operating range of the setup.