Adaptive actuator fault detection, isolation and accommodation in uncertain systems

Document Type

Article

Publication Date

1-1-2007

Publication Title

International Journal of Control

Volume

80

Issue

1

First Page

45

Last Page

63

Abstract

Adaptive actuator fault detection, isolation, and accommodation problems for linear multi-input single-output (MISO) systems with unknown system parameters are investigated. To solve the detection problem, we construct an adaptive estimate of the output signal. By comparing it with the output of the system, any type of actuator faults can be detected. However, the fault isolation problem is much more complicated. In order to solve it using an adaptive approach, the article considers constant actuator faults, which arise when the actuator output (such as a valve) is stuck at some fixed value. A novel idea which entails controller design for fault isolation is proposed. Thus, the controller in this case is not only designed to meet the control objective, but also to help with fault isolation, in case of an actuator failure. To accomplish this, assuming that there are m inputs, a group of additive functions, called fault isolation design functions in m∈-∈1 inputs, solely used for fault isolation, are introduced. Assuming that less than m∈-∈1 faults can occur, adaptive estimates of the output need to be defined to isolate the faults. Isolation is accomplished by comparing these estimates with the output of the actual system. If there exists only one estimate that matches the output of the system, it is concluded that the combination corresponding to the estimate is the faulty combination. This determines the number of actuator faults and isolates the faulty actuators. Once the faults are detected and isolated, the adaptive fault accommodation problem is accomplished by simply turning off the faulty actuators and using the remaining normal actuators. Finally, an illustrative example with simulation results is provided to support the theoretical results.

DOI

10.1080/00207170600921011

ISSN

00207179

E-ISSN

13665820

Share

COinS