A Robust Fault Detection and Prediction Scheme for Nonlinear Discrete Time Input-Output Systems

Abstract

Model-based fault detection (MFD) techniques are preferred over hardware based schemes due to low cost and minimal changes to the system when the system states are available. However, one of the major challenges in model based monitoring, diagnosis and prognosis (MDP) approach was to develop a detection and prognosis (DP) scheme in discretetime in the presence of partial state information since discrete-time schemes are normally preferred for ease of implementation. Therefore, in this paper, we propose a unified fault detection and prediction (FDP) scheme for a nonlinear discrete-time input-output system in the presence of modeling uncertainties when certain states are not available for measurement. A nonlinear estimator with an online tunable approximator and a robust term is introduced to monitor the system. A residual is generated by comparing the output of the system with that of the estimator. A unknown fault is detected when the generated residual exceeds a mathematically derived threshold. Subsequently, the online approximator and the robust terms are initiated. The approximator uses the system input and output measurements while its own parameters are tuned online using a novel update law. Additionally, robustness, sensitivity, and the stability of the fault detection scheme are rigorously examined. The proposed scheme is guaranteed to be asymptotically stable due to the introduction of the robust term and using some mild assumption on the system uncertainty. Subsequently the process of determining the time to failure (TTF) is introduced. Finally, the FDP scheme is simulated on a magnetic suspension system.

Department(s)

Electrical and Computer Engineering

Second Department

Computer Science

Keywords and Phrases

Fault detection and prognostics; Lyapunov stability; Nonlinear discrete time system; Online approximator

International Standard Serial Number (ISSN)

0974-8571

Document Type

Article - Journal

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2024 The Muk Publications and Distributions, All rights reserved.

Publication Date

01 Jan 2019

This document is currently not available here.

Share

 
COinS