Resonant frequency estimation for adaptive notch filters in industrial servo systems

This paper proposes a new algorithm for estimating the resonant frequency of adaptive notch filters used in servo systems. Notch filters and adaptive notch filters are widely used in commercial servo systems for suppressing a resonance which is a major obstacle in improving their performance. However, the conventional frequency estimation algorithm gives a dynamic behavior that is proportional to the difference between the square of the estimated frequency and the square of the actual frequency. This can cause the estimation dynamics to be too slow for low-frequency resonances, if both low- and high-frequency resonances are present and if the estimator gain is designed for a high frequency. This paper develops a new algorithm to give a dynamic behavior that is proportional to the difference in the estimated frequency and actual frequency. This allows selecting the estimation parameters independent from the value of resonant frequencies. The developed algorithm is implemented to a production servo controller and applied to a production printed-circuit-board inspection system. The experimental results show that the developed algorithm is much more effective in suppressing the resonances of both low- and high-frequencies, compared to the pervious algorithm. Furthermore, the ability to suppress vibration allows increasing the feedback gain, which in turn allows improving the tack time performance from 190 ms to 100 ms. All experiments reported in this paper were performed in an actual industrial environment using a production system, and the developed algorithms are applied to the production system.