A Method of Impurities Classification Used for Multispectral Molten Steel Based on Self-organizing Feature Map Neural Network

摘要

Through the measurement of molten steel and implanting equivalent blackbody, the characteristics of impurities in molten steel is extracted by analyzing the relationship between wavelength and spectral emissivity. The molten steel which contains impurities is separated by using self-organizing feature map neural network. Design consideration of the molten impurities filter system is then presented, including the principle analysis, the simulation model as well as the corresponding neural network. Simulation results indicate that the system can work effectively in molten impurities category recognition.