Study On Electric Arc Furnace Endpoint Prediction Method Based On Neural Network

Electric arc furnace(EAF) steelmaking utilizes electrical energy as heat source to smelt. It can smelt the grades which are strict with mechanics capability and chemical component, such as special tool steel, air craft quality steel and stainless steel. Compare with basic oxygen furnace(BOF) steelmaking, the EAF have advantages such as investment is small, capital construction and capital recovery is quick, thermal efficiency is high, temperature of steel liquid can be raise flexibly, component of steel liquid can be adjusted and controlled, have high adaptability. In addition, EAF can use scrap totally, it is competitive in steelmaking.The effectively operation, high efficiency and high quality, saving energy and reducing the cost are essential to the steel and iron company. The automatic control system influences the quality and output of the products directly. The endpoint prediction of EAF steelmaking is an important part of its optimal control. It is helpful for the operator to choose the most effective control method. The paper, based on referring to the large numbers of literature, make particular study to endpoint prediction of electric arc furnace. Knowing different prediction method, the paper have neural network as the basal prediction method, which considering practical process and actual research condition. The main idea of neural network is, based on the analysis of the past spot data and the mastery of the previous neural network model, which consider practical characteristic of steelmaking process, calculate the parameter of the model, then predict the other spot data according the confirmed model.The artificial neural network is capable of dealing with non-linear problems. It can be used for on-line response. Because EAF steelmaking process is complicated, it has many influence factors of endpoint, and it is hard to measure continually and precisely. I adopt an RBF neural network, a mathematical model of endpoint and correlative factors is developed to predict the endpoint of EAF steelmaking. It is proved by emulation that the model has comparatively strong ability of self-studying with better convergence character than conventional algorithm. The precision of the results is comparatively high.