Application of Genetic Algorithm Combined with BP Neural Network in Soft Sensor of Molten Steel Temperature

In this paper, a new soft sensor model of LF molten steel temperature is introduced after analyzing the process of LF refining and the factors of influencing molten steel temperature. A genetic algorithm combined with improved back-propagation (BP) neural network is used in this model. Genetic algorithm is used to optimize weight and bias values of BP network. The simulation studies show this new combined algorithm has global searching abilities and can improve the prediction model precision and convergence speed. Based on the results the model is programmed by C language to use in real process. The heats that error between the temperature predicted by the combined algorithm model and real temperature dose not go beyond ± 5 degree is more than 85% of all.