An approach to incremental fuzzy modelling of dependencies in complex systems

To build a computer human support system to monitor complex systems we always have to deal with the concern of knowledge representation. Modularization and hierarchy seem good functional modelling approaches to handle complex systems with several subsystems and many interdependent functions and structures. However, modelling the dynamic behaviour of these dependant functions and structures, is not well handled by the available approaches. This problem is very relevant in real time monitoring. An approach based on fuzzy logic is proposed to model system behaviour. Fuzzy systems are sets of rules governing the system. They are built using human expert knowledge, and system data. By using system data, we propose an incremental approach to modelling the system's dependencies. The approach uses a clustering method to group data. Fuzzy systems are created based on the significance and position of clusters. Clusters ineffective to the fuzzy system are ignored. Thus, an optimised fuzzy system is built. Optimisation means smaller fuzzy systems with better accuracy.