Utilizing Data from a Sensorless AC Variable Speed Drive for Detecting Mechanical Misalignments

Conventional condition monitoring techniques such as vibration, acoustic, ultrasonic and thermal techniques require additional equipment such as sensors, data acquisition and data processing systems which are expensive and complicated. In the meantime modern sensorless flux vector controlled drives can provide many different data accessible for machine control which has not been explored fully for the purpose of condition monitoring. In this paper polynomial models are employed to describe nonlinear relationships of variables available from such drives and to generate residuals for real time fault detection and performance comparisons. Both transient and steady state system behaviours have been investigated for optimal detection performance. Amongst 27 variables available from the drive, the torque related variables including motor current, Id, Iq currents and torque signals show changes due to mechanical misalignments. So only these variables are explored for developing and optimising detection schemes. Preliminary results obtained based on a motor gearbox system show that the torque feedback signal, in both the steady and transient operations, has the highest detection capability whereas the field current signal shows the least sensitivity to such faults.