ABSTRACT
Data-driven techniques are applicable in situations when sufficient run-to-failure data are available to map out the damage space and when it is not feasible to perform cost/benefit analyses using physics-based damage propagation algorithms. Yet run-to-failure data can be difficult to find.
Moreover, although data-driven techniques or methods can provide remaining useful life (RUL) estimates, their outcomes (damage estimates) may differ, making performance metrics able to provide a comprehensive and objective assessment of the performance of prognostics algorithms extremely attractive. Yet few public repositories allow a comparative analysis of different prognostics algorithms.
The chapter covers many of the different strategies for RUL estimation using data-driven methods.
