Assessment of nucleate pool boiling heat transfer and critical heat flux for power electronics cooling with a low-GWP refrigerant

To combat climate change, greenhouse gas emissions have to decrease across all sectors. In the transport sector, several options to reduce the emission of carbon dioxide are considered, such as hydrogen and bio-fuel powered engines. For vehicles, the most adopted strategy is to convert to battery electric vehicles, which can be charged using renewable electrical energy. Currently, electric vehicles have several drawbacks compared to conventional internal combustion engine vehicles, such as lower driving range and higher cost. These drawbacks can be combated by making the electric drivetrain (motor, power electronics and battery) more compact and therefore more power-dense. Further increasing the power-density requires more effective cooling systems. The power modules in the power electronics of the drivetrain feature the highest heat fluxes and are therefore one of the most challenging components to cool. State-of-the-art thermal management systems of electric drivetrains utilize liquid cooling with water-glycol mixtures. Two-phase cooling, with a coolant that starts to boil as it comes into contact with heat generating components, has the potential of acquiring higher heat transfer rates and thus more effective cooling. Two-phase cooling does come with a major risk. If the heat flux from a drivetrain component to the coolant increases above a maximal value, the critical heat flux, a continuous vapour film will be formed on that surface. This layer acts as a thermal insulator and thereby reduces the heat transfer rates dramatically, which results in overheating and failure of the component. In this study, the fluid chosen for the two-phase cooling system is FK-649, a fluoroketone. It has a low global warming potential equal to 1, which means that sporadic leaks of the fluid will not contribute significantly to global warming. Boiling heat transfer is a complex phenomenon with a strong interaction of heat and mass transfer. Although research on boiling heat transfer has been conducted for almost a century, there is still no consensus on which heat transfer mechanisms are dominant. This lack of fundamental understanding renders it difficult to predict heat transfer rates. Several heat transfer correlations have been proposed by different authors, taking into account to various degrees the effects of heat flux, fluid properties and boiling surface properties. Several correlations have also been proposed for the critical heat flux and these show a larger agreement with each other than the heat transfer correlations. Limited data is available for boiling heat transfer of FK-649. It is also unclear which correlations are best suited to predict heat transfer rates and critical heat flux for FK-649. To determine the heat transfer rates associated with pool boiling of FK-649, an experimental setup is designed and constructed. The setup consists of a sealed reservoir containing FK-649. Boiling occurs on a horizontal boiling surface at the bottom of the reservoir, while the generated vapour is condensed by a spiral condenser at the top. Polycarbonate windows are added for visual access to the boiling phenomenon. The type of boiling surface can be altered and the saturation temperature of FK-649 can be controlled to the desired value. The saturation temperature during the experiments was varied in three levels: 36 °C, 41 °C and 46 °C. In the power module under test, IGBTs are the highest heat dissipating components. These components are placed on a baseplate which has a larger area than those of the components. This allows to spread the heat generated in the components with the goal of having a larger heat transfer area. As the heat is not generated uniformly over the entire internal baseplate area, this may result in a non-uniform heat flux at the boiling surface. To analyse this effect, measurements of pool boiling on the power module are compared to measurements with a boiling surface that is uniformly heated by an electrical resistive heater. The measured heat transfer rates matched within 10%. Also visually, no evidence of non-uniform heat fluxes were perceived. Bubble nucleation occurred uniformly across the boiling surface and was not correlated with the location of the IGBTs. It is concluded that the baseplate effectively spreads the heat and that a quasi-uniform heat flux is achieved at the boiling surface. These results indicate that the correlations developed for surfaces with uniform heat fluxes can be directly applied to power module baseplate cooling. Thirteen correlations that predict nucleate pool boiling heat transfer are assessed. In order of the date of publication, these are the correlations of Kruzhilin, Rohsenow, Forster-Zuber, Kutateladze-Borishanskii, Borishanskii-Mostinksi, Shekriladze-Ratiani, Labuntsov, Gorenflo, Stephan-Abdelsalam, Cooper, Kutateladze, Leiner and Pioro. Of these correlations, the Rohsenow and Pioro correlations require fitting one or more constants to experimental data for each specific surface-fluid combination. The predicted thermal performance is compared to the experimental results. As expected, the correlations with fitting constant perform best. The Rohsenow correlation is the preferred correlation, it has an average deviation of 7.5% and predicts all measurement points within 25.6%. Of the non-fitted correlations, the Labuntsov correlation is best at predicting heat transfer rates with FK-649. The average deviation of the prediction from the measurements is equal to 7.5% and all data is predicted within 29.1%. Remarkably, the average error of the Labuntsov correlation is equal to that of the fitted Rohsenow correlation. Both correlations take into account accurately the effect of saturation temperature on heat transfer. The Labuntsov correlation is thus preferred for predicting heat transfer rates without available measurement data for fitting. Caution should however be exercised if the correlation is used for non-metallic surfaces or surfaces with a non-standard surface roughness, as the Labuntsov correlation does not take into account influences of the boiling surface. The experimental results showed that the effect of the heat flux on the nucleate boiling heat transfer rates can be divided into three separate regimes. These regimes were also perceived visually. From low to high heat flux, the three regimes are the partial nucleate boiling, fully developed nucleate boiling and partial dryout boiling regimes. These regimes are also reported by several authors in literature. However, none of the heat transfer correlations take into account the different nucleate boiling regimes and they therefore do not predict adequately the variation of the heat transfer rates with heat flux. The measurements in this study were correlated using a specific power-law relation for each of the three regimes. For more accurate predictions, future heat transfer correlations should take into account the differences between the three regimes. Four critical heat flux correlations are compared to the experimental data: those of Kutateladze, Zuber, Lienhard-Dhir and Mudawar et al. Both the Zuber as the Lienhard-Dhir correlations predict all data within 10%. It is concluded that the critical heat flux correlations perform much better than the heat transfer correlations. To assess the feasibility of two-phase power electronics cooling with FK-649, the measured thermal performance is compared to that of water-glycol cooling. The maximal heat transfer coefficient measured in this study was 4965 W/m²K and the highest critical heat flux was 146 kW/m². The heat transfer coefficients are similar to those of water-glycol cooling of a flat baseplate. However, the maximal attainable heat fluxes with water-glycol are more than two times higher than the critical heat flux of FK-649 boiling. For two-phase cooling of power modules with FK-649 to outperform water-glycol cooling, strategies for increasing the critical heat flux such as increasing the saturation temperature, using subcooled boiling or using flow boiling should be investigated