Greedy Kernel Components Acting on ANFIS to Predict BOF Steelmaking Endpoint

In order to make overall consideration of the information from the original variables in the basic oxygen furnace (BOF) steelmaking dynamic process, an adaptive neural network fuzzy inference system (ANFIS) model based on kernel and greedy components is proposed. This kind of model can improve the endpoint predicting precision of the steel carbon contents and temperature. After hidden information is exposed in the high feature space through the kernel function transformation, greedy algorithm is used to remove redundant information and reduce the dimensions. The extracted components are used as the new inputs of ANFIS, and the implication relation among the inputs is reflected by rules, which simulate the operators experience, and consequently reduce the influence resulted from different operators. When the practical data are simulated, the simulated results are close to the practical values. The method is effective.