Estimation of the firing frequencies in discrete and continuous Petri nets models

Production frequencies provide useful information for performance analysis, monitoring and control applications of manufacturing systems. However, such frequencies are not always measurable. In this case, they must be estimated from the observation of other variables. This paper deals with production frequencies estimation for systems modelled by discrete or continuous Petri nets. For this purpose, the content of the buffers is measured on-line with a sampling period, and the evolution equations of the model are reversed. In the case of timed Petri nets, the production frequencies estimation is obtained from the approximation of the firing sequences and in the case of continuous Petri nets it results directly from the variations of the marking vector. The exact and approximated solutions of the estimation problem are described. Accuracy of the estimation is related to measurement error, Petri net structure and sampling period. Moreover, when the estimation provides several solutions, the Petri net model is extended such that a unique solution is obtained.