Feedforward Neural Networks In The The Pangang Tar Production Control System

It has great significance that improving the level of automation during the tar distillation process of production, for reducing the consumption of raw materials as well as energy consumption and improving product quality. At the same time, it is conducive to reduce the work intensity of the labor, raise the labor productivity and enhance the safety of production capability. Clearly, it is the key initiative that implements on-line automatic control in the use of neural network model in the tar distillation process of production control. It analyses the process path and characteristics in the production process, combining with the artificial neural network(the former feed-back propagation neural network).There are eight mathematical models established, such as the export of the furnace temperature control model, evaporator Top temperature control model, the top oil and gas fraction temperature control model, naphthalene early furnace temperature control exports model, the top naphthalene fine temperature control, the naphthalene fine furnace temperature control model, the export methylnaphthalene furnace temperature control model etc. It analyzes the input and output parameters of each model as well as the set of guiding principles. It discusses the collection of offline data,online learning and the whole process as well as the realization principles of the online control.At the end of the paper, the results achieved are gave, including some of the collected data,some results of the offline learning and online control.