论文部分内容阅读
针对现场烧结终点控制复杂与难度大的实际,开发了神经网路预报系统。预报系统采用4层前向神经网络,进行多因素输入建模,输出采用具有极值特性的二次曲线计算的烧结终点与实际最高废气温度,预报烧结终点与最高废气温度,为现场终点控制的最新可行方法。网络结构设计先进合理、精度高、泛化能力强,训练方差为0.00001814,用训练样本集测试输出,烧结终点绝对平均误差为0.04,终点废气温度绝对平均误差为4.57℃。采用训练后网络预报,烧结终点(风箱号)绝对误差最大仅为0.09,终点废气温度绝对误差最大为3.57℃,命中率100%。用预报结果有针对性调节烧结参数可收到明显效果。
In view of the fact that the control of the final sintering site is complicated and difficult, a neural network forecasting system has been developed. The forecasting system uses a four-layer feedforward neural network to model the input of multi-factor inputs. The output is calculated from the end point of sintering and the actual maximum exhaust temperature calculated by the quadratic curve with extremum characteristics. The final sintering end point and the maximum exhaust gas temperature are predicted. The latest available method. The network structure is advanced and reasonable in design, with high precision and strong generalization ability. The training variance is 0.00001814. The output of the training sample set is tested. The absolute mean error of the final sintering point is 0.04 and the absolute mean error of the final exhaust gas temperature is 4.57 ℃. After training network prediction, the absolute error of the sintering end (bellows) is only 0.09, the absolute error of the final exhaust temperature is 3.57 ℃, and the hit rate is 100%. With the forecast results targeted sintering parameters can receive significant results.