ABSTRACT
This chapter introduces the existing data-driven techniques for determining remaining useful life (RUL). It outlines current data-driven prognosis techniques for ball bearing, together with the bearing dataset selected for this case study. The chapter examines which features can be extracted from the raw signal, the role of the frequency domain in sampling vibrational phenomena, how to filter raw signals and how these filtering techniques can be useful for the feature extraction. It explains the chosen degradation model and its parameters, and the theoretical background of RUL estimation and its assumptions. The chapter reports on and discusses the results of the RUL estimation. It reviews several commonly used techniques to analyze vibration signals. The chapter describes hidden Markov model (HMMs) to model the operating conditions for RUL estimation. It contains the RUL estimation results, comparing the HMM outcomes with actual RUL values.
