摘要
针对包含多道加工工序、输入变量很多的复杂工业系统建模精度难以提高的问题,提出一种改进的前馈神经网络结构,输入变量不是由同一层输入,而是根据变量起作用的前后次序分别在网络的不同层输入,真实反映了大工业过程的各生产工序中的参数发生作用的时间顺序。同时由于输入变量在适当的时候输入网络,从而使网络的规模减小。该神经网络是处理高维问题,尤其是建立包含多道加工工序的大工业过程模型问题的强有力工具。将该神经网络用于热连轧产品质量建模,经过实测数据拟合与检验,仿真结果表明:提出的小波神经网络结构是可行的而且有很好的应用前景。
Aiming at the course of dimensionality and low accuracy during modeling complex industrial systems ,a mod-ified architecture of feed-forward neural network is proposed.The input variables are not input in one layer,but in dif-ferent layers according them to make action being early or late;it can show the sequence of work procedure in complex industrial production processes and the real complicated relationship between input and output.Meanwhile the variables are input to neural network in necessity,the scale of neural network is relative small.The proposed neural network is a powerful tool for solving high-dimension problems ,especially for modeling large-scale industrial processes with several work procedures.The developed methodology is tested by means of modeling the product quality of hot rolling mill with engineering data.Simulation results show that the proposed method is suitable for this set of problems and has good prospects for further use.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第3期206-208,232,共4页
Computer Engineering and Applications
基金
国家863高技术研究发展计划资助项目(编号:863-51-011)
西安交通大学自然科学基金项目(编号:0900-573024)资助
关键词
前馈神经网络
小波分析
复杂系统
建模
大工业过程
加工工序
质量模型
热连轧机
Neural network,Wavelet,High-dimension,Three input-layer,Large-scale industrial process,Work procedure,Quality model,Hot rolling mill
