An Evaluation of Thermosyphon Heat Pipe Charged R134a and R600a Using Machine-Learning Algorithms

Abstract

Thermosyphon Heat Pipe (THP) is a heat transfer device that consists of three sections with working fluid inside, including evaporator, adiabatic, and condenser. The refrigerants in thermosyphons are commonly used for low-temperature applications. In this article, two higher heat transfer refrigerants, namely R600a and R134a, are used as the working fluid in thermosyphon to compare their performance. The total thermal resistance of the refrigerant R600a and R134a charged thermosyphon was calculated using experimental results, and it shows that R600a provides lower thermal resistance than R134a. Experimental results show that the heat transfer of R600a increases by 12% than R134a. The temperature at the evaporator and condenser section provides the performance of the thermosyphon. Therefore, in addition to the experimental results, three different machine-learning methods such as support vector machine supporting vector machine (SVM), artificial neural network artificial neural networks (ANN) and convolutional neural network Convolution Neural Network (CNN) are deployed to classify the efficiency of the experimental process and find the accuracy of the output using these methods. These approaches are used to diagnose the temperature with the input parameters such as the ratio between length to diameter, condenser to evaporator length, thermal conductivity of working fluid, boiling point of working fluid, temperature of the cooling water, applied heat flux and heat input. The neural schema applied over the measured temperature obtained from the experimental results was optimized with the algorithmic approach which provides the relative grade. In this article, the classifiers are used to compare the performance of the refrigerants and measure the accuracy of the evaluation for R600a and R134a.