A Novel Indicator for Mechanical Failure and Life Prediction Based on Debris Monitoring

Since failure of mechanical components can lead to catastrophic failure of the entire system, significant efforts have been made to monitor system behavior and try to predict the end of useful life of a component. A method to assess the process of mechanical wear in real time is based on monitoring the amount of debris in the lubricant. Although this approach has shown some potential in application, the nonlinearly cumulative damage in the late stage of mechanical life presents significant challenge to early prediction of the Remaining Useful Life. This paper considers continuous wear (devoid of sudden large particle dislodging or catastrophic failure) and assumes that it is a positive feedback physical process. This assumption serves as a basis of a dynamic model developed to describe the nonlinear behavior of wear in the late stage of useful mechanical life. Based on this model, it was discovered that the inflection point in the cumulative debris, during continuous wear process, presents a more accurate indicator of pending mechanical failure compared to the existing indicators. The peak in generation rate is considered as the end of useful mechanical life. The model is validated based on data from four wind turbine gearboxes. The results indicate that the proposed model and associated indicator can be used with significant confidence to predict the Remaining Useful Life during the early stages of operation and have distinct advantages over the standard linear model (Moving Average Model), both in terms of accuracy and robustness.