NeuroControl of Temperatures with OWE Neural Network Architecture in Continuous Steel Annealing Furnace

Artificial Neural Network (ANN) is an attractive solution for the identification and control of complex systems. This technique can be very heavy for processes depending on many parameters like, for example, in steel-making. In such cases, solutions such as Orthogonal Weight Estimator (OWE) architectures are recommanded. They make it possible the definition of a generic neuronal architecture, adapted to the main parameters of the problem and a dynamic estimation of the weights of this architecture with regard to contextual parameters. This application to a furnace shows that these contextual parameters can correspond to the temporal evolution of a temperature and tackles the difficult issue of transient states.