Qualitative model based fault diagnosis using a threshold level

Sobhi Baniardalani, Javad Askari, Jan Lunze
2010 International Journal of Control, Automation and Systems  
This paper deals with the effect of using a threshold level in qualitative model based fault diagnosis algorithm. By introducing the concept of α-consistency, a fault diagnosis algorithm is presented in this paper. In this method for each measured input and output signal, a measure of consistency is computed for each fault. If this measure is less than a threshold level α, then the observed input and output are not consistent with the fault. Therefore the fault is excluded from the set of the
more » ... ssible faults. In order to illustrate the proposed method, this algorithm is applied on a 2-tank system. The obtained results show that the faults can be isolated faster. Furthermore to illustrate the diagnosis reliability, a confidence interval is defined. This interval determines the probability of correct fault diagnosis. Keywords: α-consistency, qualitative model, qualitative model based fault diagnosis, stochastic automaton, threshold level. In which [u(k)] and [y(k)] represent qualitative values of u(k) and y(k) respectively. Some different solutions for this problem have been presented in [1, 4, 7, 11, 13] . These methods consist of two general steps. First, the quantized system is represented by a discrete event model [12, 18, 19] . Second a diagnostic algorithm uses this model and observed input and output sequences to find the faults. Models such as nondeterministic automata [1,11], stochastic automata [5, 6, 13, 14] , Petri nets [10], timed event graphs [20], semi-Markov models or timed © ICROS, KIEE and Springer 2010 __________ Manuscript
doi:10.1007/s12555-010-0323-4 fatcat:pkhk6ry5qrddzouikrhkskfvze