A neuro-fuzzy self built system for prognostics: a way to ensure good prediction accuracy by balancing complexity and generalization

El-Koujok, Mohamed; Gouriveau, Rafael; Zerhouni, Noureddine

doi:10.1109/phm.2010.5413348

Cited by 6 publications

(1 citation statement)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(Adeline, Gouriveau, & Zerhouni, 2008) tests and compares different methods based on neural networks in terms of prediction precision, computation cost and requirements related to the implementation. Fuzzy neural networks combines neural networks and fuzzy logic to deal with ambiguous, inaccurate, noisy or incomplete data (El-Koujok, Gouriveau, & Zerhouni, 2010). Fuzzy systems use knowledge as expert rules.…”

Section: Data-driven Prognosismentioning

confidence: 99%

A generic ageing model for prognosis - Application to Permanent Magnet Synchronous Machines

Vinson,

Ribot,

Combacau

2014

PHME_CONF

View full text Add to dashboard Cite

In the context of more electrical aircrafts, Permanent Magnet Synchronous Machines are used in a more and more aggressive environment. It becomes necessary to supervise their health state and to predict their future evolution and remaining useful life in order to anticipate any requested maintenance operation. Model-based prognosis is a solution to this issue. Any prognosis method must rely on knowledge about the system ageing. A review of existing ageing laws is presented. The generic ageing model proposed in (Vinson, Ribot, Prado, & Combacau, 2013) is extended in this paper. It allows representing the ageing of any equipment and the impact of this ageing on its environment. The model includes the possible retroaction of the system health state to itself through stress increase in case of damage. The proposed ageing model is then illustrated with Permanent Magnet Synchronous Machines (PMSM). Two critical faults are characterized and modeled : inter-turns short-circuits and rotor demagnetization. Stator and rotor ageing are well represented by the proposed ageing model. The prognosis method developed in (Vinson et al., 2013) is extended to consider this new generic ageing model. In order to test the prognosis algorithm, ageing data are needed Since no real measurements are available, a virtual prototype of PMSM is developed. It is a realistic model which allows running a fictive but realistic scenario of stator ageing. The scenario comprises apparition and progression of an inter-turns short-circuit and its impact on stator temperature, which value has an impact on the ageing speed. The prognosis method is applied successfully to the PMSM during this scenario and allows estimating the RemainingUseful Life (RUL) of the stator and the machine.

show abstract

Section: Data-driven Prognosismentioning

confidence: 99%