Application of artificial neural network and PCA to predict the thermal conductivities of nanofluids

详细信息    查看全文

• 作者：Fakhri Yousefi ; Somayeh Mohammadiyan ; Hajir Karimi
• 刊名：Heat and Mass Transfer
• 出版年：2016
• 出版时间：October 2016
• 年：2016
• 卷：52
• 期：10
• 页码：2141-2154
• 全文大小：1,403 KB
• 刊物类别：Engineering
• 刊物主题：Engineering Thermodynamics and Transport Phenomena
  Industrial Chemistry and Chemical Engineering
  Thermodynamics
  Physics and Applied Physics in Engineering
  Theoretical and Applied Mechanics
  Engineering Fluid Dynamics
  
• 出版者：Springer Berlin / Heidelberg
• ISSN：1432-1181
• 卷排序：52

文摘

This paper applies a model including back-propagation network (BPN) and principal component analysis (PCA) to compute the effective thermal conductivities of nanofluids such as Al₂O₃/(60:40)EG:H₂O, Al₂O₃/W, Al₂O₃/(20:80)EG:W, Al₂O₃/(50:50)EG:W, ZnO/(60:40) EG:W, CuO/(60:40)EG:W, CuO/W, CuO/(50:50)EG:W, TiO₂/W, TiO₂/(20:80)EG:W, Fe₃O₄/(20:80) EG:W, Fe₃O₄/(60:40) EG:W, Fe₃O₄/(40:60) EG:W and Fe₃O₄/W, as a function of the temperature, thermal conductivity of nano particle, volume fraction of nanoparticle, diameter of nanoparticle and the thermal conductivity of base fluids. The obtained results by BPN–PCA model have good agreement with the experimental data with absolute average deviation and high correlation coefficients 1.47 % and 0.9942, respectively.List of symbolsKThermal conductivitySWCNTSinglewall carbon nanotubeMWCNTMultiwall carbon nanotubePCAPrinciples component analysisANNArtificial neural networkBPBack propagationMLPMultilayer percepetronRBFRadial basis function networksGRGhost recon networksCFBCascade forward back propagationSSESquare errorAADAbsolute average deviationBPEBack propagation errorMSEMean square errorR2Correlation coefficientMAREMean absolute relative errorEGEthylene glycolWWater