カルマン・フィルタを用いたねじり軸共振系の位置決め制御

Abstract

In a magnetic tape unit, high speed and precise positioning control should be achieved. This paper demonstrates that a Kalman filter can precisely estimate the state variables for the capstan motor, combined with its torsion resonance characteristics modeling, and widen the loop bandwidth by the feedback from estimated state variables.The capstan motor is modeled as a two freedom resonant system that consists of two parts, a torque generator part and a capstan part. The Kalman filter is constructed to estimate rotation angle and rotation angle velocity for the two parts, from the motor driving current and the capstan rotation angle.The comparison between experimental frequency response for estimated state variables and calculated values confirmed that the two freedom model for the capstan motor was able to describe its actual resonant characteristics and that the Kalman filter circuit operated up to several kHz beyond the first resonant frequency for the system. The actual capstan motor was controlled by feedback from these estimated signals, based on the optimal control theory. The result shows that the loop bandwidth for the positioning control was extended about twice as wide as the conventional method, where a tachogenerator and band reject filters were used. The motor was started up for about 0.4ms.