Combining Canonical Variate Analysis, Probability Approach and Support Vector Regression for Failure Time Prediction

Reciprocating compressors are widely used in oil and gas industry for as transport, lift and injection. Critical reciprocating compressors that operate under high-speed conditions and compress hazardous gases are target equipment on maintenance improvement lists due to downtime risks and safety hazards. Estimating performance deterioration and failure time for reciprocating compressors could potentially reduce downtime and maintenance costs, and improve safety and availability. This study presents an application of Canonical Variate Analysis (CVA), Cox Proportional Hazard (CPHM) and Support Vector Regression (SVR) models to estimate failure degradation and remaining useful life based on sensory data acquired from an operational industrial reciprocating compressor. CVA was used to extract a one-dimensional health indicator from the multivariate data sets, thereby reducing the dimensionality of the original data matrix. The failure rate was obtained by using the CPHM based on historical failure times. Furthermore, a SVR model was used as a prognostic tool following training with failure rate vectors obtained from the CPHM and the one-dimensional performance measures obtained from the CVA model. The trained SVR model was then utilized to estimate the failure degradation rate and remaining useful life. The results indicate that the proposed method can be effectively used in real industrial processes to predict performance degradation and failure time.