Performance Evaluation of a State and Input Observer for Machine Diagnostics

In this paper, the performance of a state and input observer is evaluated on a DC spindle drive system designed and built for this purpose. The observer estimates the spindle drive load torque (a disturbance input) based on easy-to-access signals: armature voltage, current and speed. The results show that the bandwidth and signal-to-noise ratio of the load torque estimate is as good as that obtained with an in-line torque transducer for the conditions tested. The use of these observers for detecting machine tool malfunctions and abnormal process conditions related to load torque (such as tool breakage, tool wear, slide stick/slip, and bearing preload) is discussed. State and input observers appear to have good potential for machine diagnostics.