Frequency-domain-based nonlinear normalized iterative learning control for three-dimensional ball screw drive systems

Abstract

Iterative learning control (ILC) is a well-established method for achieving precise tracking in repetitive tasks. However, most ILC algorithms rely on a nominal plant model, making them susceptible to model mismatches. This paper introduces a novel normalization concept, developed from a frequency-domain perspective using a data-driven approach, thus eliminating the need for system model information. The proposed method is designed specifically for unknown, nonrepetitive discrete-time systems, enhancing their transient tracking performance. By normalizing the input–output ratio, the method prevents excessive amplification of the system input and reduces computational complexity. Notably, this data-driven approach is effective for both iteration-invariant and iteration-varying trajectory tracking tasks. Two examples demonstrate the performance and potential advantages of the proposed method in a three-dimensional ball screw drive system.