Dynamic threshold generators for robust fault detection in linear systems with parameter uncertainty

Sensor Fault Detection and Isolation Using System Dynamics Identification Techniques.

Detection of faults, such as clogged valves, broken bearings or biased sensors, has been brought more and more into focus during the last few decades. There are two main reasons why faults are impo ...

Dynamic threshold generators for robust fault detection

Shift-invariant signal norms for fault detection and control

The methods of data validation which were developed these last years largely call for the redundancy resulting from models. The case of models with certain parameters (static and/or dynamic) was analyzed and received many solutions. However, there is relatively few work concerning the data validation in the presence of model uncertainties. The aim of this communication is to present a method of data validation for dynamic linear systems, which is able to take into account the uncertainties of the model parameters. Firstly we represent the dynamic model of the system in a static form by piling the state and measurement vectors on an observation window. Secondly, the elementary operations relating to the intervals make it possible to propose a state estimation of the system taking into account the parameters uncertainties. As the uncertainties are supposed to be bounded, the estimation's result is provided in an interval form. A sequential algorithm is used to obtain the state estimation by carrying out the intersection between the estimation resulting from three methods (Gauss elimination, Gauss-Seidel iteration and Krawczyk iteration). By analyzing this estimation, we can detect and isolate the data which are affected by gross errors as biases and propose a correction to make these data coherent with the model of the system.

/pdf/data-validation-of-uncertain-dynamic-systems-4qh42866da.pdf

Data validation of uncertain dynamic systems

A method for detecting sensor faults in a turbofan engine is presented. The proposed method consists of an observer with integral action and an adaptive detection threshold. The threshold is computed with the assumption of parametric uncertainty in the process model. Successful simulations with sensor data from an RM12 jet engine shows that the method is capable of detecting even a very small increase in sensor noise promptly without generating false alarms.

/pdf/parametric-uncertainty-in-sensor-fault-detection-for-7rvinslsrj.pdf

T. Norlander

Papers

Parametric uncertainty in sensor fault detection for turbofan jet engine