基于过程参数控制的烧结矿质量预测模型

A Prediction Model for Sintering Quality Based on Control of Process Parameters

针对烧结过程非线性、强耦合性和大时滞的特点,从过程参数控制的角度对烧结工艺进行了总体分析,确定了烧结矿性能评价指标及其主要影响参数,在此基础上提出了一种带动量项和变学习率的BP神经网络算法,建立了烧结矿质量预测模型。仿真实验结果表明,模型具有较强的自学习功能和较高的预测精度,用拓扑结构为15-25-4的BP神经网络和0.65×10-3的网络误差进行训练,模型的预报命中率在81.25%以上,充分验证了基于过程参数控制的烧结矿质量预测模型的准确性和有效性。

Aiming at characteristics of non-linearity, strong coupling and long time-delay in the sintering process, an overall analysis was conducted for the sintering process from the perspective of process parameter control. Thus, the sintered ore properties evaluation index and the main impacting parameters for it were all obtained. Then, a BP neural network algorithm with momentum and variable learning rate was proposed, with which a quality prediction model for the sintered ore established. The following simulation experimental results showed that the model has a higher prediction precision and stronger self-learning ability. The predictive hit-ratio of random samples was over 81.25% by adopting BP neural network with the structure of 15-25-4 and network error of 0. 65 × 10^-3, which verified the accuracy and effectiveness of this quality prediction model on the basis of process parameter control.