基于BP神经网络的钢液终点磷含量预测模型

Prediction model of phosphorus content in molten steel based on BP neural network

磷含量是描述钢液质量的一个重要的含量。结合遗传算法(GA)和误差反馈型神经网络(BP),建立了优化的GA-BP神经网络预测模型,预测转炉炼钢过程钢液终点磷含量。对现场收集的数据进行仿真学习,结果表明,该预测模型收敛速度快,具有较高的预测精度,平均绝对误差可达到0.002 7%。随着训练样本的增加和模型结构的进一步优化和完善,将具有很好的应用前景。

Phosphorus is an important element to describe molten steel quality. Building a model is necessary to predict phosphorus content. Combined Genetic Algorithms（GA） and back-propagation neural network （BP）, an optimized GA-BP model was established to predict phosphorus content. Some data were chosen to train the network model. The results show that the convergence rate was faster, the model had higher accuracy, the average absolute error can reach 0.002 7 %. Witb the inerease of training samples and optimization of model structure, this model will have a good applicaton prospect.