On-line preload monitoring for high-speed anti-friction spindle bearings using robust state observers.

Abstract

Catastrophic and premature bearing failure caused by excessive thermally-induced bearing preload is a major design problem for spindle bearings of high-speed machine tools. Due to a lack of a low cost and easy to maintain on-line preload measuring technique, the traditional solution is to limit the maximum spindle speed and the initial bearing preload. This solution is incompatible with the need to increase machining productivity which requires increasing the spindle speed and product quality (surface finish, dimensional accuracy), which requires increasing (or at least not decreasing) the preload to keep the spindle system stiff. This thesis proposes a dynamic mathematical model of the spindle bearing system and a monitoring strategy based on the model to estimate spindle bearing preload dynamically. The model is derived from physical laws of heat transfer and thermoelasticity and expressed in a state-space form. The state variables defined in the model are the temperatures of the outer ring/housing, the rolling elements, and the inner ring, while the induced preload is an algebraic function of these states. The monitoring strategy is based on observer techniques from modern control theory. The spindle bearing preload is estimated by means of low cost thermocouples attached to the bearing out ring and spindle housing. A systematic procedure to determine the observer gains is developed to account for modeling errors, unknown parameters, nonlinearities, and measurement noise. The modeling errors are particularly significant in this problem because factors like bearing skidding, wear, and lubricant deterioration are not easily modeled and the observer design must be insensitive to them. The preload observer has been successfully validated through two different bearing configurations for a wide range of speeds and running conditions. The results show that the model-based monitoring technique which combines the measurement of outer ring temperature and robust state observer can be an effective and low cost solution for preload monitoring in high speed machine tools.