Application of Multilayer Kalman Filter to a Flexible Drive System

Abstract

The paper proposes a novel estimation method for a mechanical two-mass system. The concept of a multi-layer estimator is proposed to improve the estimation quality of the system states, especially for unknown initial conditions of the drive. The proposed estimator has two layers. The first layer consists of two individual Kalman filters, whereas the second layer is based on aggregation mechanism to calculate the final states of the plant. The proposed layers comprise a multi-layer Kalman filter (MLK). The investigated drive has a changeable value of the load-side inertia. To ensure the desired responses of the drive, an adaptive-control structure based on a proportional-integral (PI) controller with an additional feedback loop is implemented. The simulation and experimental results illustrating the effectiveness of the proposed MLK in an open- and closed-loop configuration are presented. The MLK guarantees a considerably more accurate estimation quality than a classical single estimator.