摘要
文章应用BP神经网络技术,探讨辊底式连续热处理加热炉钢板温度监测系统方案的可行性,提出串行网络设计思想和方法,在对各炉膛温度与被加热钢板温度之间预先寻求权值函数的基础上,将各炉膛BP神经网络串联构成热处理加热炉的BP神经网络监测系统,进一步寻求整个热处理炉各炉膛温度与钢板最终温度之间的权值函数(数学模型),作为热处理炉钢板温度检测系统的传递函数。该方案的特点是系统权值函数的逼近速度更加快捷,稳定性更好,系统硬件构成简单,加之引用加热温度的梯度预知限定条件,可防止权值函数掉入局部最优缺陷的学习失败。
Adopting BP Neuro-network technology, this paper aims to research on the feasibility of roller-type heat treatment furnace steel board temperature monitoring system. Based on precognitive weight function built on temperatures between every firepot and the heated steel board, a serial network is designed to connect all firepots into a complete heat treatment furnace BP monitoring system. Wight function relationship is further investigated between the temperature of the separate firepots of the heat treatment furnace and the outlet temperature of the steel board temperature monitoring system. This research helps to reach better approximating quickness of the systemic weight function, greater stability, and a simpler systemic hardware composition. The adoption of precognitive prescriptive condition of heating temperature gradient will help the weight function avoid study failure of partial optimal flaw.
出处
《四川理工学院学报（自然科学版）》
CAS
2007年第1期1-5,共5页
Journal of Sichuan University of Science & Engineering(Natural Science Edition)
基金
重庆市科学技术委员会重点攻关项目项目(2001-6698#)
关键词
加热炉
温度检测
BP神经网络
heat treatment furnace
temperature monitoring
BP Neuro-network
