从辅料运行特性的角度对烧结工艺做了总体分析,并对烧结矿的质量指标及其影响因素做了研究,在此基础上运用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。模型预报结果表明,用拓扑结构为15-20-4的BP神经网络和0.000 191的网络误差进行训练后,模型的命中率在83.3%以上,充分展示了基于辅料运行特性的烧结矿质量预测模型的准确性和有效性。
A general analysis of sinter process from the aspect of characteristic of subsidiary material′s movement was finished,and the quality index and its influence factors were investigated.The quality prediction model for the sinter was built using a transformed BP neural network algorithm with momentum and variable learning rate.The prediction results show that the predictive hit-ratio of random samples is over 83.3% through adopting the BP neural network with the 15-20-4 structure and network err of 0.000 191.It can be concluded that the model have higher accuracy and effectiveness of prediction on the basis of characteristic of subsidiary material.
参考文献
| [1] | 孙文东 .烧结生产系统的优化与控制研究[D].华中科技大学,2004. |
| [2] | 刘竹林.炼铁原料[M].北京:化学工业出版社,2007 |
| [3] | 季斐.现代综合评价主要方法的比较概述[J].学术探讨,2010(08):296. |
| [4] | 徐春晖,徐向东.前馈型神经网络新学习算法的研究[J].清华大学学报（自然科学版）,1999(03):1-3. |
| [5] | Jacobs R A .Increased Rates of Convenee Through Learning Rate Adaption[J].Neural Networks,1998,1(04):295. |
| [6] | Rumelhart D E.Learning Representation by BP Errors[J].Nature(London),1986(07):149. |
| [7] | 邓志东;孙增所 .利用线性再励的自适应变步长快速BP算法[J].模式识别与人工智能,1993,6(04):319. |
| [8] | 阅平凡.人工神经网络模型一分析与应用[M].合肥:安徽教育出版社,1991 |
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%