Prediction Model of Nickel Converter Based on Neural Network Algorithm

Document Type


Publication Date



Department of Materials Science and Engineering


Based on the vital position of the endpoint prediction of converting, and the fact that the determination of its endpoint still depends on workers’ experience, this paper proposes a prediction model based on BP neural networks and extreme learning machines. Five models are included: (1) back propagation (BP) neural network, (2) genetic algorithm-optimized back propagation neural network (GABP), (3) extreme learning machines (ELM), (4) genetic algorithm-optimized extreme learning machines (GAELM), and (5) particle swarm optimization of extreme learning machine (PSOELM). Based on the historical production data of the Jinchuan Group, these models have been applied for the prediction of air delivery volume and air delivery time of the converter. By comparing the root mean square error (RMSE), goodness of fit (R 2) and hit rate within ± 10% relative error range of each model, GAELM has the highest accuracy with a 93.12% hit rate for air delivery volume and a 93.65% hit rate for air delivery time, indicating that GAELM can effectively predict the end of the converter blowing, and provide a good reference for endpoint control and the determination of the nickel converter.

Publication Title