A Unified System Residual Life Prediction Method Based on Selected Tribodiagnostic Data

Abstract

This paper proposes a new systematic method for assessing system material wear to build a system degradation model and estimate residual technical life. Selected metal wear debris from lubricating oil, which contains information about the lubricant conditions and system conditions, is analyzed. We focus on the iron (Fe) and copper (Cu) debris, which we (and other researchers) consider to be valuable, of the contact degradation and wear failure systems. By monitoring the changes in debris content in the lubricating oil, we build a system degradation model and further predict the moment when the system no longer fulfills its functions; the residual life might then be set as the time reference to implement preventive maintenance. The degradation model is founded on the specific characteristics of a stochastic diffusion process with bivariable, using the bivariate Wiener process with a time scale transformation. An inference function to describe the dependency among the selected wear debris was also applied because the oil field data exhibit some uncertainty and correlation. Based on the degradation modeling results, the system reliability curve and the failure probability density curve predict the MTBF value and the expected mean residual life can be obtained, and provide the foundations for the condition-based maintenance of the system. However, the potential applications of the results are much broader. For instance, the results can be used as inputs to mission plan optimization and further reduce system maintenance costs.