Search

A recurrent neural network model for process monitoring is proposed. Fault detection, isolation and diagnosis are introduced in the context of production systems. Sequences and temporal constraints between events are considered as essential for process monitoring. The difficulty that have artificial neural networks to deal with time is highlighted. Self-recurrent connections are proved to be a simple mechanism to treat time and sequences and their basic properties are given. A modular architecture is presented and applied to a flexible assembly cell.

The aim of this paper is to present a software and hardware architecture for real time control of traffic light signals, and prediction of congestion in traffic flow. The system presented here is under development within the framework ofa European Esprit Project: ASTORIA.