Diagnosis of quantized systems based on a timed discrete-event model

The paper deals with the diagnosis of continuous-variable systems whose sta te can be measured only by means of a quantizer Hence, the on-line informat ion used in the diagnosis is given by the sequences of input and output eve nts. The diagnostic algorithm uses a representation of the quantized system by means of a semi-Markov process, which provides a timed description of t he event sequences generated by the quantized system. The paper describes h ow the probability density function of the semi-Markov process can be deter mined for a given quantized system. The diagnostic algorithm can be applied on-line, because it determines the fault probabilities recursively for the sequence of input and output events. The result is illustrated by a numeri cal example which concerns a part of a batch process.