Model Evaluation of Various Thermo-Physical Properties of Nanofluids and ANN Modelling for 10kWe Integrated Reactor

Abstract

A 10kWe integrated reactor with Stirling generator is in design, to satisfy China’s power demand for both Earth orbit and deep space exploration in the next two decades. The integration of the core and the thermoelectric conversion system, reduces the number of tubes and pump structures, which leads to a higher energy conversion rate, fewer failure risks and coolant leaks. The waste heat of this reactor would be transferred through its heat pipes to its radiators, then to the space. Applying nanofluids would help reduce the heat pipe sizes, because nanofluids have great alterations in their thermo-physical properties with a small fraction of nanoparticles. Numerous models have been proposed to characterize the thermo-physical properties of nanofluids. However, it is found that researchers have different, sometimes even contradictory conclusions about some of the properties. At the same time, these properties could be affected by various aspects, the simple models are not sufficient for the reference. This work focuses on evaluating the models of density, specific heat capacity, thermal conductivity, viscosity, and Nusselt number of nanofluids with statistical methods, and provides reference thermo-physical properties for the design of the heat pipes in the space reactor. For this reason, a great amount of experimental data is collected. Profiting from the collected data, artificial neural network (ANN) models based on Pytorch are trained and compared with the other models.