A CBM Model Based on VAR Modeling of Oil Data and SPE

摘要

In this paper, we present a preventive replacement model for a truck transmission subject to condition monitoring atrnregular inspection epochs. Transmission oil samples are taken roughly every 600 hours and the metal readingsrnobtained from the spectrometric analysis in ppm are recorded. First, the portions of the histories when transmissionrnwas in a healthy state are identified by plotting Hotelling’s s T 2 statistic. A vector AR model is then fitted to the oilrndata using healthy portions of the histories. The squared prediction error (SPE) vector calculated for each sample isrnused as a covariate vector for a proportional hazards-based CBM model which is built using EXAKT software. Thernmodel is compared with the previously built CBM model using raw oil data as covariates.