Machine prognostics based on health state estimation using SVM

摘要

The ability to accurately predict the remaining useful life of machine components is critical for continuous operations in machines which can also improve productivity and enhance system safety. In condition-based maintenance (CBM), effective diagnostics and prognostics are important aspects of CBM which provide sufficient time for maintenance engineers to schedule a repair and acquire replacement components before the components finally fail. All machine components have certain characteristics of failure patterns and are subjected to degradation processes in real environments. This paper describes a technique for accurate assessment of the remnant life of machines based on prior expert knowledge embedded in closed loop prognostics systems. The technique uses Support Vector Machines (SVM) for classification of faults and evaluation of health for six stages of bearing degradation. To validate the feasibility of the proposed model, several fault historical data from High Pressure Liquefied Natural Gas (LNG) pumps were analysed to obtain their failure patterns. The results obtained were very encouraging and the prediction closely matched the real life particularly at the end of term of the bearings.