Thermal resistance modeling of oscillating heat pipes filled with acetone by using artificial neural network

Abstract

Oscillating heat pipes (OHPs) are applicable in different energy systems such as solar collectors, desalinations and fuel cells as thermal mediums or thermal management devices. Thermal resistance of OHPs is affected by different elements including the architecture of OHP, filling ratio, heat load and operating fluid. Depending on the operating temperature range, appropriate working fluid is used. Acetone is one of the fluids that is applicable for the OHPs working in low or medium temperatures. Performance modeling of OHPs can be conducted by employing numerical simulation and data-driven methods. Utilizing powerful data-driven methods can be preferred in terms of time consumption and computational cost. In the present article, two types of artificial neural network, multilayer perceptron (MLP) and group method of data handling (GMDH), are utilized for thermal resistance modeling of OHPs filled with acetone under various working conditions. To find the models with the highest possible accuracy, different architectures are considered in this work. Models’ outputs revealed that the thermal resistance of the aforementioned OHPs can be reliably predicted by using the abovementioned methods. R-squared of the models proposed by using MLP and GMDH are 0.989 and 0.965, respectively. Values of mean squared errors for the mentioned methods are around 0.0045 and 0.0144, respectively.