Prediction of pistachio thermal conductivity using artificial neural network approach
2007
Chayjan, R.A. (Bu-Ali Sina Univ., Hamedan (Iran). Agricultural Machinery Engineering Dept.) | Montazer, G.A. (Tarbiat Modares Univ., Tehran (Iran). IT Engineering Dept.) | Hashjin, T.T. (Tarbiat Modares Univ., Tehran (Iran). Mechanics of Agricultural Machinery Engineering Dept.) | Khoshtaghaza, M.H. (Tarbiat Modares Univ., Tehran (Iran). Mechanics of Agricultural Machinery Engineering Dept.) | Ghobadian, B. (Tarbiat Modares Univ., Tehran (Iran). Mechanics of Agricultural Machinery Engineering Dept.)
One of the most important thermo-physical characteristics of pistachio is thermal conductivity, which was predicted at a range of temperatures (50 to 95 degree C) and moisture contents (3.8 to 52.15% dry basis; d.b.) in this species using line heat source method and artificial neural networks (ANNs). Two independent variables of temperature and moisture content were considered as inputs of ANNs and thermal conductivity was considered as an output. Radial basis function network (RBFN) with 112 entities and four neurons in second hidden layer with linear threshold function and learning rule of delta was selected to be the best ANN architecture that had the least mean square error 0.0045 and R2 = 0.99. Decreasing moisture content to 10.8% (d.b.) reduced thermal conductivity, but decreasing it further to 3.8% (d.b.) caused proportionate increase in thermal conductivity of the samples. Prediction accuracy of thermal conductivity by designed ANN was 3.24% better than statistical results. Thus the RBFN is recommended for predicting the pistachio thermal conductivity.
显示更多 [+] 显示较少 [-]