火灾模拟实验炉温软测量建模的研究  

Study on Modeling of Furnace Temperature Soft Measurement of Fire Simulated Experiment

作　　者：赵望达[1] 徐志胜[1] 吴敏[2] 

机构地区：[1]中南大学土木建筑学院,湖南长沙410075 [2]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《计算机测量与控制》2006年第10期1296-1298,共3页Computer Measurement &Control

基　　金：国家"十五"科技攻关项目(2002BA806B03);湖南省科技厅计划项目(02SSY3027)

摘　　要：火灾模拟实验炉是对建筑构件进行模拟测试实验的装置,其升温必须按ISO834国际标准执行,炉温的估计是实验的关键,采用了主元分析和RBF神经网络相结合的模型构成火灾模拟实验炉温软测量;主元分析(PCA)实现输入变量的降维,RBF神经网络采用K-均值聚类算法进行隐层中心和连接权调节的学习,实现快速收敛;该融合模型使炉温估计精度比常规的最小二乘方法拟合精度提高2倍以上,保证了升温过程温度的精确控制。The fire simulated experiment furnace is a device to testing building eomponent, and the furnace temperature must be strictly according to ISO834 international standard. A furnace temperature soft measurement method is adopted based on a data fusion model with a method combining pritzciple component analysis （PCA） and radial basis function neural network （RBFNN）. PeA can reduce the dimen- sions of input variable. RBFNN may quickly reach convergence of learning algorithm by adjusting the center of hide layer and the connective weigh with K-mean cluster algorithm. This model improves the precision of furnace temperature two times more than least square method, then, furnace temperalure is accurately controlled.

关 键 词：火灾模拟实验 炉温软测量 主元分析 RBF神经网络 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]