摘要
针对焦炉生产过程中直接检测火道温度成本高、精度低等问题,提出运用自适应神经网络模糊推理系统理论(ANFIS)建立焦炉火道温度预报模型,模型采用模糊减法聚类方法选取模糊规则数目,大大减少规则冗余量;结合最小二乘和误差反向传播混合算法对神经网络参数进行优化,采用现场的热工数据作为输入,将获得的模型与传统的线性回归模型和BP神经网络模型进行了比较,数值仿真结果表明所建立的模型具有学习速度快、预报精度高、泛化能力强等优点。
The temperature predictive model for coke oven flow was proposed based on ANFIS theory, which tries to conquer the high cost and low precision with detecting directly. The structure of ANFIS was initialized by the Subtractive Fuzzy--clustering algorithm, whichreduced the number of the rules greatly. The synthesis of LMS and BP algorithm were used for the training and optimization of the neural networks. The numerical simulations based on the sampling data demonstrated that the learning speed, predictive precision and fitting level had been greatly improved compared to the linear regressive model and the BP neural networks.
出处
《计算机测量与控制》
CSCD
2007年第4期462-463,473,共3页
Computer Measurement &Control
基金
湖北省国际科技合作重点项目基金(2006CA025)
湖北省教育厅研究项目(2001A19006)。