Model-based health monitoring of rotate-vector reducers in robot manipulators

Health monitoring of robot manipulators is crucial for their long-term reliable operation. Since the feedback data provided by the controller of robot manipulators are most often from the actuator’s side, health monitoring of rotate-vector (RV) reducers housed on the output of actuators is challenging. To avoid the increased cost and complexity of adding sensors to robot manipulators, this paper investigates the long-term health monitoring of RV reducers, using only feedback data from the controller. A generic model for RV reducers is first developed to characterize their nonlinear behaviours and performance degradation. The model is then extended to robot manipulators to analyse their dynamic response under the time-varying excitation generated by healthy and faulty RV reducers. Next, a new motor-torque-based residual together with an adaptive threshold is proposed for the health monitoring of RV reducers, taking payload, measurement noise and friction disturbance into account. The proposed threshold depends only on the variance of measurement noise, and is therefore easy to implement in practice. Extensive simulations and experiments are conducted on two robot manipulators to validate the proposed model, residual and threshold. The results show that the proposed model is able to account for the effects of faults and nonlinear behaviours of RV reducers on the motor performance. The proposed residual with an adaptive threshold is sensitive to faults in RV reducers and robust to friction disturbance and time-varying payload, and it outperforms conventional momentum-based residuals.