Abstract
In this paper, an algorithm for the detection and diagnosis of faults in sensors and actuators in unknown non-linear systems is presented. Firstly, the normal operation of the non-linear system is modelled via a feedforward neural network in order to produce a healthy model which is used later as a redundant relation. It is assumed that every faulty sensor or actuator can be modelled by a single parameter f. Such a parameter is defined to be f = 1 when the system is healthy and 0 < f < 1 otherwise. A residual signal is evaluated from the difference of the output of the healthy model and that of the system. A simple threshold function is used to detect the faulty behaviour of the system. The estimation of f is then computed by minimizing the differences between the output of the system and that of the healthy neuro model with the help of the gradient descent rule. It has been shown that the algorithm is able to estimate the fault with large magnitude accurately. However, if the deviation of f from its healthy value is small, a linearized healthy model, together with the recursive least-squares algorithm, can be used to estimate the fault. As a result, a combination of both methods will provide a powerful technique for accurate fault diagnosis.
Original language | English |
---|---|
Pages (from-to) | 261-278 |
Number of pages | 18 |
Journal | Proceedings of the Institution of Mechanical Engineers. Part I: Journal of Systems and Control Engineering |
Volume | 215 |
Issue number | 3 |
DOIs | |
State | Published - 1 Dec 2001 |
Externally published | Yes |
Keywords
- Actuators and sensors
- Fault detection and diagnosis
- Least-squares estimation
- Neural networks
- Non-linear autoregressive moving-average model with exogenous input (NARMAX model)
- Non-linear dynamic systems