Estimation of thermal conductivity of pure gases by using artificial neural networks

A feedforward three-layer neural network is proposed to predict conductivity (k) of pure gases at atmospheric pressure and a wide range of temperatures based on their critical temperature (Tc), critical pressure (Pc) and molecular weight (MW). The accuracy of the method is evaluated and tested by its application to experimental conductivities of various gases which some of them are not used in the network training. Furthermore, the performance of the proposed technique is compared with that of conventional recommended models in the literature. The results of this comparison show that the proposed neural network outperforms other alternative methods, with respect to accuracy as well as extrapolation capabilities. Besides, conventional conductivity correlations are usually used for a limited range of temperature and components while the network method is able to cover a wide range of temperatures and substances.