The Research Of The Application Of Immunity-based Hybrid Learning Algorithm In Static Model Of The Refining Of Vanadium

Vanadium is a kind of important and valuable mental that has wild application in steel making, electronic production and national defense industry. At present, our country's refining of vanadium is based on human's experience. The automatic degree is so low that the quality of the products fluctuates greatly. With the "Information technology drive industrialization" has become the key strategy of the every enterprise, the application of computer in refining of vanadium is an important direction. An important way of using computer is constructing mathematical model of refining of vanadium. Mathematical model can describe the rules of system's movement in quantity, so it has important meaning to scientific research and practical manufacturing. At present, in many fields of steel-making, there have constructed mathematical model that have positive effect on raising technical level and stabilizing technical guide line.This dissertation has made deep research on the "static model of refining of vanadium and the research of its computable intelligence of control strategy" project It makes some innovative discussion on the modeling mechanism, model algorithm, and model optimization from the point of view of immunity theory. It then uses the immunity-based hybrid learning method in the modeling of the RBF neural network, which has enriched the modeling method, and made deeper understanding of the project.In the discussion of the immunity-based hybrid learning method, we make detailed explain about the background of immunology, the theory of immunity network. From the point of view of the immunity system, we make a new explain of the RBF neural network and raise the viewpoint that RBF neural network is an artificial immunity network. We make detailed explain about the two immunity-based algorithm (learning from clonal selecting and learning from internal affinity) from the point of the view of the principle, step and validation. At last, we use the algorithm in the modeling of refining of vanadium: use the learning from clonal selection to construct the model that have the enough predictive accuracy, then use the decreasing algorithm (learning from internal affinity method) to adjust the model to get the simplest architecture of the model. Through the simulation, it approved that the immunity-based hybrid learning method has obvious advantage over the traditional algorithm.