Artificial immune system-based diagnostics and prognostics scheme and its experimental verification

Abstract

In this paper, a novel fault diagnostics and prediction (FDP) scheme is introduced by using artificial immune system (AIS) as an online approximator for a class of nonlinear discrete-time systems. Traditionally, AIS is considered as an offline tool for fault detection (FD). However, in this paper, AIS is utilized as an online approximator in discrete-time (OLAD) along with a robust adaptive term in the proposed fault diagnostics observer. Using the fact that the system outputs are alone measurable, an output residual is determined by comparing the observer and system outputs and a fault is detected if this output residual exceeds a predefined threshold. Upon detection, the OLADs are initiated to learn the unknown fault dynamics while the robust adaptive term ensure asymptotic convergence of the output residual for a state fault whereas a bounded result for an output fault. Additionally, for prognostics purposes, the parameter update law for AIS is used to estimate the time-to-failure (TTF). Finally, the performance of the proposed FDP scheme is demonstrated experimentally on an axial piston pump test-bed for two failure modes.