Adaptive Neuro-Fuzzy Inference System for mid term prognostic error stabilization

Abstract

Abstract: The high costs in maintaining complex equipments make necessary to enhance maintenance support systems and industrial and research communities take a growing interest in the “prognostic process”. However, this activity is still not well bounded and real prog-nostic systems are scarce. Thus, the general purpose of the paper is to explore the way of performing failure prognostics so that manager can act consequently. The prognostic process is discussed from different points of view (concept, metrics, approaches and tools) in order to point out the pragmatic challenges of this activity. Assuming that maintenance decisions follow from a prediction step, the stabilization of mid term prediction errors appears to be essential. For that purpose a neuro-fuzzy predictor based on the ANFIS model is proposed to perform prognostic.