Comparing different approaches for model parameters identification in short time

Model Parameter values of machine tools change during machining. An optimal process control needs precise knowledge of the actual parameter values. Three different algorithms are introduced to estimate the modal parameter values of system in a short time window with high resolution: least squares estimation (LSE), estimation of signal parameters via rotational invariance (ESPRIT) and orthogonal matching pursuits (OMP) algorithm. These algorithms are augmented with a sliding-window operation to reveal the actual system dynamic behavior at every time instance. This paper focuses on comparing the performance and the identification accuracy of the proposed methods and the influence of the applied window size and noise content using numerical examinations. The results show that the sliding-window LSE can estimate transient parameters accurately and suits realtime control processes.