Expert neural network with fuzzy coding for process control

This paper introduces a new intelligent method using neural networks and fuzzy logic to control a hot coil strip process having a look-up table. An expert neural network with a gating network made by multilayer neural networks is used for model a rule based look-up table that has a quantization error caused by an offset of the input variables. Also, a new fuzzy coding scheme is introduced to scale an input variable that has the non-stationary characteristics. The table modeling performance of the proposed methods is verified by the actual data obtained from a hot coil strip process of Pohang steel company in Korea.