融合机理与数据的篦冷机温度软测量模型  

Soft-testing model for grate cooler temperature measurement with mechanism and data fusion

作　　者：王美琪 陈恩利 刘鹏飞 戚壮 侯爽 Wang Meiqi;Chen Enli;Liu Pengfei;Qi Zhuang;Hou Shuang(School of Mechanical Engineering,Shijiazhuang Tiedao University,Shijiazhuang 050043,China;Hebei Extension Center of Water Resources Research and Hydrological Technology Experiment,Shijiazhuang 050000,)

机构地区：[1]石家庄铁道大学机械工程学院,石家庄050043 [2]河北省水资源研究与水利技术试验推广中心,石家庄050000

出　　处：《仪器仪表学报》2018年第6期182-188,共7页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(51641609);河北省高等学校科学技术研究(BJ2016047,BJ2017001);河北省教育厅青年基金(QN2016182);国家重点研发计划(2017YFC0704000)项目资助

摘　　要：针对篦冷机内部温度难以在线测量的问题,提出一种融合渗流换热机理与神经网络数据辨识的篦冷机温度软测量模型。首先考虑局部非热平衡效应,并结合多孔介质渗流换热理论建立床层内气固换热的机理模型;其次以床层边界条件与机理模型的温度预测值作为网络输入量,机理模型的温度预测偏差作为输出量,建立双并联前馈神经网络,并采用快速学习网算法训练网络参数;将辨识后的人工神经网络作为篦冷机气固渗流换热机理的补偿器,建立混合温度软测量模型。利用篦冷机生产运行数据进行了二次风温的软测量实验,实验结果表明,与机理模型相比混合软测量模型测量精度有较大提高,并且与实验数据的误差较小,混合软测量模型对于篦冷机温度具有较好的测量精度。The temperature inside the grate cooler is difficult to be measured on line. To solve this problem,one kind of soft-sensing model of integration seepage heat transfer mechanism and neural network data identification is proposed. Firstly,the non-equilibrium of local thermal is considered. The mechanism model of gas-solid heat exchange based on the porous medium seepage heat transfer theory is formulated. Then,the boundary conditions and temperature prediction of the mechanism model is utilized as the network input. The temperature prediction deviation of the mechanism model as output a double-parallel feedforward neural network is established. The fast learning network algorithm is adopted to obtain the parameters of the double-parallel feedforward neural network. Finally,the neural network is used as the error compensator for the seepage heat transfer mechanism model. In this way,a mixed temperature soft measurement model is established. The soft measurement experiment of secondary wind temperature is carried out by using the production and operation data of grate cooler. Compared with the mechanism model,experimental results show that the measurement precision of the mixed soft measurement model is improved obviously. The error of the experimental data is small. The mixed soft-sensing model has the advantage of high accuracy for the temperature measurement inside the grate cooler.

关 键 词：篦冷机 熟料温度 神经网络 软测量模型 

分 类 号：TH701[机械工程—仪器科学与技术]