铁矿石烧结性能预报模型

Predictive model of iron ore sintering capabilities

研究了铁矿石烧结性能的评价指标及其主要影响因素,提出了误差修正的带动量项的线性再励自适应变步长BP神经网络算法,建立了铁矿石烧结性能预报模型。模型预报结果表明,用拓扑结构为12—34—4的BP神经网络训练6700次后,神经网络训练误差为0.000187,模型预报命中率均达83.5%以上,模型具有很好的泛化能力和自适应能力。

The valuing indexes and some main influencing factors in iron ore sintering capabilities were investigated in this paper. Based on the research, a BP neural network learning algorithm with amending error, appending momentum and adaptive variable step size linear reinforcement was presented, and a predictive model of iron ore sintering capabilities was established. By adopting the BP neural network with the 12-34-4 structure and after 6 700 times train, the predictive result of model of iron ore sintering capabilities is satisfying, the neural network training error is 0.000 187, and the predictive hit-ratio of random samples is over 83.5%. It can be concluded that the predictive model is generally applicable and has self-adaptability.