Modeling And Control Of Neural Network For Cupola

Artificial Neural Network (ANN) is interconnected widely by a mass of processing unit, it was proposed in the foundation of modern neuroscience research results. The neural network gives first place to simulating parallel processing in large scale, with strong character of robustness, fault tolerance and studies ability independently. The neural network has the capability of random approaches and learning to the random nonlinear function, so it make the neural network has a good effect to solute the control of the nonlinear system and the system which is difficult to modeling, moreover it has great potential aspect to solute the control of the highly misalignment and the serious uncertainty systems. The neural network's application has penetrated to every aspect of automatic control field, including system identification, systems control, optimized computation as well as control system's failure diagnosis and fault-tolerant control and so on.Copula is one of the most basic melting equipment for molten iron in the production of foundry. The process of cupola melting not only has the character of nonlinear, strong perturbation and time lag, but also has three complex processes such as burning of coke at bottom, heat transferring and metallurgy responded. It's a typical complex commercial process. So it's difficult to model the cupola's modeling in traditional way. From control side, it's strict for the control of melting in the copula as a process of production. In view of its complexity, before the copula was improved fundamental and the smelting work standardization, people's early thought of using the classic and modern control theory has no ideal effect.First of all, this article introduced the background of this topic and the significance for selecting the topic, and the present research situation of the control technology for melting in cupola, and introduced the developing process of the artificial neural networks simply; Elaborated artificial neural networks' essential feature, the way of studying and the rule of studying, have studied two kind of forward feed neural network: BP and RBF neural network, and their learning algorithm. Then modeled the copula with the neural network, and emulated cupola's actual data by using MATLAB. At last proposed control cupola iron liquor temperature by using the neural network adaptive control method aims at the complex nonlinear system of the cupola, takes the BP neural network as NNI and NNC. First carries on the off-line training to NNI, when the NNI training achieves the ideal effect, then carries on the study to NNC, by using the deviation which obtains from NNI, NNC may follow system's change quickly, enables the control to meet the ideal requirements rapidly. The results of simulation in MATLAB showed that this kind of control plan is feasible, and it has laid the foundation for it's application in the actual production.