There are many faults that can occur in a technical system such as a process industry. Valves and hoses can leak, bearings can jam, sensors can be in error and so on. Common for all these faults is, however, that if they are not discovered in time, they may cause problems in the production and, eventually, accidents.
Many kinds of faults are simply the result of wear and tear and occur inevitably in any facility. This is the background to why fault detection is a hot topic for research since several decades. The idea is that if a mathematical model for the plant is available then available measurements may be compared to the predictions of this model. If they differ significantly then it may be concluded that a fault has occurred.
The problem with this approach is that mathematical models are never exact and therefore there is always a difference between the measurements and the predictions of the model. In consequence, a decision on whether a fault has occurred or not must be based on the magnitude of this difference. In this project, we explore the possibilities of automating this decision process. If one can quantify how large the uncertainty of the process model is, then this may be utilized for calculating how much the measurements must differ from the model prediction in order to conclude that a fault has occurred.
That the applications are broad is illustrated by the two systems that the results have been applied to, detection of sensor faults in the jet engine of the Gripen fighter aircraft, and the detection of clogged valves of a flotation process in the mining industry. Our partners regarding these applications are Volvo Aero and Boliden, respectively. The project has, among other things, resulted in a PhD title for Michael Bask with the thesis ”Dynamic threshold generators for robust fault detection”.
Johansson, A, Bask, M & Norlander, T 2006, 'Dynamic threshold generators for robust fault detection in linear systems with parameter uncertainty', Automatica, vol. 42, nr. 7, s. 1095-1106.
Bask, M 2005, Dynamic threshold generators for robust fault detection, Luleå tekniska universitet, Luleå (Doctoral thesis / Luleå University of Technology; 2005:11).
Johansson, A & Bask, M 2005, 'Dynamic threshold generators for fault detection in uncertain systems', I Proceedings of the 16th IFAC World Congress, International Federation of Automatic Control.
Bask, M & Johansson, A 2005, 'Linear optimization of parameters in dynamic threshold generators', I Proceedings of 16th IFAC World Congress, International Federation of Automatic Control.
Bask, M & Johansson, A 2005, 'Realizable upper bounds for the modulus of impulse responses', I Proceedings of the SICE Annual Conference: SICE 2005 Annual Conference in Okayama - Proceedings, 2005, Society of Instrument and Control Engineers s. 2045-2050.
Bask, M & Johansson, A 2004, 'Robust time-varying thresholds for supervision of valves in a flotation process', I Proceedings of the IEEE Conference on Decision and Control: CDC 2004, s. 4305-4310.
Bask, M & Johansson, A 2003, 'Model-based supervision of valves in a flotation process', I 42nd IEEE International Conference on Decision and Control, IEEE s. 744 - 749.