Fault Prediction Using Supervised and Unsupervised Learning Algorithms in Cyber Physical Systems

In the last decade, industry has become highly dependent on smart systems which enable the physical world to merge with the virtual one. This development led to the emergence of Cyber Physical Systems (CPS). In this environment, services and resources must be always available to support the continuity of systems operation. Indeed, CPSs are intended to be flexible systems that can decide automatically how to adapt their internal behavior in response to the dynamics of the environment. The ability to, automatically, recognize and predict any fault or failure, that occurs while delivering services, is a step towards realizing such systems. We present in this paper an approach to early fault prediction using machine learning algorithms. The viability of the proposed solution is confirmed by a real world application in an industrial CPS.