Towards Enhanced Prognostics with Advanced Data-Driven Modelling

A considerable amount of prognostics research has been conducted to improve the remaining useful life prediction of engineering assets. Advantages such as lowering sustainment costs and improving maintenance decision making, are significant motivations to enhance the prognostics capability. Sensor selection, data pre-processing, knowledge elicitation and the mathematical techniques are some of the elements required of prognostics research to enhance capability.

This paper takes a broad view of prognostics and explores techniques available from a variety of research and application disciplines. A prognostics dataflow diagram illustrates the complete prognostics process and the paper discusses the impact of improvements in each process step to enhance the prognostics performance. The mathematical approach to prognostics is a crucial issue. Exploring cross-disciplinary prognostic approaches is helpful to extract useful techniques from different domains and to fuse the strengths of each discipline.

A case study of fatigue induced crack-growth using Bayesian approaches is used to illustrate that data-driven prognostics can deliver benefits to the industry.