A Model-Based Fault Prognostics Scheme for Uncertain Non-linear Discrete-time Systems with Multiple Distinct Faults

Abstract

In this paper, an online prognostics framework is proposed for a class of uncertain non-linear discrete-time systems with multiple faults affecting the system state with all the states being considered measurable. Multiple faults imply that each system state is affected by several faults at the same time provided the faults are separable. In this framework, multiple faults (incipient or a combination of incipient faults) are detected by using the proposed fault detection (FD) estimator, which consists of an online approximator in discrete time and a robust adaptive term. Subsequently, the fault isolation (FI) module is initiated such that each state of the FI observer corresponds to a particular fault type in the case of single fault or fault combination in the case of multiple faults. The faults will be isolated successfully when the corresponding FI state residuals converge to zero in contrast with other FI schemes where they guarantee only boundedness. In addition, multiple isolation estimators are not required here since a decision scheme is utilized by using FD and FI estimators to determine the fault location, type and number of faults that occurred. Suitable mathematical conditions are derived to show the class of faults that could be isolated. Time to failure is determined by using the parameter update law of the FI estimator and the failure thresholds. Finally, a simulation example is used to demonstrate the proposed prognostics scheme. © The Author(s) 2013.

Department(s)

Electrical and Computer Engineering

Second Department

Computer Science

Keywords and Phrases

Fault detection; fault isolation; multiple faults; non-linear systems; observer; prognostics

International Standard Serial Number (ISSN)

0142-3312

Document Type

Article - Journal

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2024 SAGE Publications, All rights reserved.

Publication Date

01 Jan 2014

Link to Full Text

Share

 
COinS