Computers & Chemical Engineering, Vol.18, No.11-12, 1157-1170, 1994
Some Practical Insights into Neural-Network Implementation in Metallurgical Industry
A neural network has been used to study steel transformation curves. Some problems occurred like non-linearity of the phenomena, measurements uncertainties, lack of examples in the database or variables interdependencies. Solutions for these problems are successively presented, leading to a fractal neural network able to learn differential equations, qualitative information and to treat noisy experimental measurements. Some validation results are presented.