东北大学学报(自然科学版) ›› 2008, Vol. 29 ›› Issue (1): 33-36.DOI: -

• 论著 • 上一篇    下一篇

基于ELM新方法的LF终点温度软测量混合模型

田慧欣;毛志忠;王嘉铮;   

  1. 东北大学信息科学与工程学院;东北大学信息科学与工程学院;太阳锻造实业有限公司 辽宁沈阳110004;辽宁沈阳110004;辽宁鞍山114000
  • 收稿日期:2013-06-22 修回日期:2013-06-22 出版日期:2008-01-15 发布日期:2013-06-22
  • 通讯作者: Tian, H.-X.
  • 作者简介:-
  • 基金资助:
    国家自然科学基金资助项目(60374003)

Hybrid modeling based on ELM for soft sensing of end temperature of molten steel in LF

Tian, Hui-Xin (1); Mao, Zhi-Zhong (1); Wang, Jia-Zheng (2)   

  1. (1) School of Information Science and Engineering, Northeastern University, Shenyang 110004, China; (2) Sun Smithing Industrial Co. Ltd., Anshan 114000, China
  • Received:2013-06-22 Revised:2013-06-22 Online:2008-01-15 Published:2013-06-22
  • Contact: Tian, H.-X.
  • About author:-
  • Supported by:
    -

摘要: 针对LF精炼炉冶炼过程中物理化学反应过程及传热过程的复杂性,采用混合模型对钢水温度进行软测量,将传统的机理模型与ELM新方法相结合,利用ELM智能算法校正机理模型中难以准确获得的参数,再用机理模型进行预测.这种混合模型既克服了传统机理模型难以准确实现的不足也避免了"黑箱"模型过分依赖数据的缺陷.同时ELM新方法也克服了传统BP算法的不足,使预测精度得到了提高.仿真结果表明,此混合模型具有较好的预测结果,终点温度预测误差不大于±5℃的炉次大于90%.

关键词: LF炉, ELM, 混合建模, 软测量, 钢水温度

Abstract: Combining the conventional mechanism model with the newly developed ELM (extreme learning machine), a hybrid model was developed for soft sensing of molten steel temperature with the aim to rise above the complexities of the physico-chemical reaction process and heat transfer process during LF (ladle furnace) refining. The intelligent ELM algorithm can correct the unacquainted parameters in mechanism model and then predict exactly them with the mechanism model. The hybrid model thus has the advantages not only overcoming the difficulty that the mechanism model is hard to implement exactly but also getting rid of the over reliance upon available data in the Black Box model. In addition, it can make up for the defect of conventional BP algorithm, thus improving the prediction accuracy. Simulation results showed its favorable predictability, i.e., the number of heats of which the predictive errors of end temperature of molten steel in LF are all not over ± 5°C is greater than 90%.

中图分类号: