Estimation of Torque Variation due to Torsional Vibration in a Rotating System Using a Kalman Filter-Based Approach

Purpose It is observed that the motor-running machines that the fluctuation in power supply causes torsional vibration in the rotating systems. Estimation of torque helps in efficient operation of machines with limited sensors. Measurement noise and the effect of exogenous inputs in torque estimation are main problems. Methods Here, authors have used a Kalman filter-based input estimation technique for indirect estimation of torque variation in a rotating system. Two mathematical models of the rotating system are used in the present work. The estimation becomes unstable when degrees of freedom is more than the number of parameters to be estimated. For such scenario a modified approach, that consists of using a penalty is proposed. Experimental measurements are performed on a drivetrain to verify the proposed technique. Results Results are presented for different braking torques. It is found that the estimated torque is close to the measured torque. The effect of the penalty on estimation is studied. A stability analysis is performed with different values of penalty values. Finally, a sensitivity analysis is performed to check for the robustness of the algorithm. Conclusions It is found that the estimated torque converges to the measured torque with a higher penalty value. The estimator is found robust as the estimated torque matches well with the measured torque for different test cases. Sensitivity analysis shows that for more accurate estimation, accuracy of model stiffness is of the essence.

Automated summary

In this study, a Kalman filter-based input estimation technique is used to indirectly estimate torque variation in a rotating system. The estimated torque closely matches the measured torque and the algorithm demonstrates robustness in different test cases. The accuracy of the model stiffness is found to be crucial for more accurate estimation.