Redundant Information Analysis in a Decomposed Petri Net Model

Many works propose to use decentralized models to control and supervise physical systems. The decomposition tasks are, often, executed by an expert on the physical application. Recent works developed at LAAS-CNRS propose method to systematically decompose centralised models based on Petri nets. This method generates redundant information between the sub-modules. In some cases, these redundancies are necessary to guarantee the dependability of the system (by duplicating critical information). It can also be used by the diagnosis algorithm in order to improve the fault isolation (by pointing out the relation between sub-models) and to reduce the maintenance costs (by reducing the number of sub-models that must be modified to update the system). However, the duplication of information introduce extra costs that can be observed by the complexity of keeping the data coherence and by the surcharge of the communication network coming from the data exchange between interdependent sub-models. This paper presents a set of indicators that can be used by the designer´s team to evaluate the consequences of redundant information introduction and to help them to decide the most appropriated configuration of the system