Intelligent process modelling using Feed-Forward Neural Networks

Abstract

A supervised Feed-Forward Neural Network (FFNN) is developed. Since Neural Networks (NN) are expensive techniques, Design of Experiments and statistical techniques are employed to offset this expense. Sometimes information is not available, in such a case, the modeller can compromise accuracy for the experimental cost. Results show that each model has an approximation capability. One or more models, once added results in enhanced modelling capacity. Different models are developed and their convergence are investigated. Conclusions indicate that neural networks are valid modelling techniques. Cost of developed models is high and can be offset with approximation tools such as design of experiments.