A fuzzy pattern recognition approach for dynamic systems diagnosis. Application to a model of the French telephone network

Diagnostic methods for the functional state of a static system are well-known. However, the diagnosis of a dynamic process is more difficult to manage because the system state evolves in time. In this paper, a complex system is assumed to evolve from one functional state to another by passing through intermediate states distributed according to a specific path in a multidimensional space. This space is defined from the relevant parameters observed in the system. In order not to create a copious number of intermediate functional states, a two-step decision process based upon fuzzy pattern recognition is proposed. It consists of building membership functions along the path according to which the system state evolves from one known functional state to another. These multidimensional membership functions are used to diagnose the system state. As an example, an application of this method to a model of the French long distance telephone network is illustrated.