A tool for system monitoring based on artificial neural networks

A research has been carried out finalized to the definition of a methodology useful for the diagnosis and prediction of the correct evolution state of physical systems. In this paper we present a related model and a specific network topology for the considered problem. In particular, the prediction procedure is based on a “Self Organizing Map”(SOM) and an “Error Back-Propagation”(EBP) networks combined to form a hierarchical architecture. The network system has been developed and tested using data furnished by Alenia and consisting in sensorial data (FBG, Fiber Bragg Grating) and multi-format descriptive data regarding evaluation (SB). The obtained results have shown that the developed methodology is a promising tool for the diagnosis activity. Key-Words: Artificial Neural Networks, Self-Organizing Maps, Prediction Systems, Life cycle monitoring