Search

• An experimental dataset of flow boiling heat transfer in mini channels with pin fins was built. • A transfer learning framework was proposed to predict the heat transfer coefficient. • The effects of different factors on heat transfer coefficient prediction accuracy was identified. Flow boiling in mini channels with micro pin fins is a promising heat sink technique to achieve high-efficient aircraft thermal management. The accurate prediction of its heat transfer coefficient is critical for the practical design of two-phase heat exchanger based on mini channels with micro pin fins. Previous investigation shows that heat transfer coefficient prediction accuracy of the machine learning method is generally better than that of conventional empirical correlations. However, the machine learning method cannot guarantee its prediction accuracy among different data domains. To extend the application region of the conventional machine learning method, a transfer learning framework was proposed in present study. First of all, an experimental system was built to acquire test data from different sample domains (i.e., the diamond pin fins with different geometries). Then a conventional machine learning model was developed based on the deep learning method. Furthermore, the developed deep learning model was adjusted with transfer learning process, and the performance of these two kinds of models (i.e., conventional machine learning model and transfer learning model) was comprehensively evaluated. Results showed that the conventional machine learning model had a good prediction accuracy with an overall deviation of 4.11 % but was only confined in the same data domain as training data. Differently, the transfer learning model with 70 % data set from the new domain can achieve appropriate prediction in the new domains with deviation of 4.28 %. The results of this paper demonstrated that the conventional machine learning model can extend into different domains with reasonable prediction accuracy through transfer learning frameworks. It will benefit the practical design of high-efficiency heat exchangers for aircraft thermal management. [ABSTRACT FROM AUTHOR]

• The flow and thermal performance of ZnO/water nanofluid were investigated. • Impacts of channel size, concentration and Reynolds number were identified. • A novel GA-optimized BP-ANN was proposed for Nusselt number prediction. • CFD studies with single-phase, mixture and various turbulence models were conducted. The thermal and flow characteristics of ZnO/water nanofluid in two multiport mini-channels were experimentally and numerically studied. Nanofluids with volumetric concentrations of 0.75 and 1.5% were used. Experimentally, the influences of concentration, Reynolds number (100–3750) and channel size (1.22 and 1.42 mm) on performance were identified. A novel Genetic Algorithm-optimized Backpropagation-Artificial Neural Network (GA-optimized BP-ANN) was proposed for Nusselt number prediction. Numerically, the performance using single-phase and mixture model with different turbulence models were evaluated. Results reveal that the nanofluids show better heat transfer performance and higher pressure drop than that of water. Additionally, the improvement is more obvious in laminar/turbulent transition region at a higher concentration. A better heat transfer performance is observed in a smaller channel after laminar flow region. For thermal performance factor, enhancement only appears at higher Reynolds numbers after flow transition. The most remarkable enhancements are nearly 1.3 and 1.48 for the two channels at the Reynolds numbers of 1600 and 1430, respectively. The developed GA-optimized BP-ANN shows an extremely high prediction accuracy with a Mean Absolute Relative Deviation (MARD) of 2.70%. Numerically, the single-phase model combined with the Lam-Bremhorst model exhibits better simulation results for nanofluid in mini-channel with a MARD of 9.0% than the mixture model. [ABSTRACT FROM AUTHOR]

• Both flow boiling and condensation in mini channels with segregated fins were investigated; • Machine learning based performance prediction models were developed; • The influential factors of different dimensionless parameters were analyzed. Two phase flow boiling and condensation heat transfer in mini channels are promising technologies in handling the challenges of modern thermal management in various fields. In this paper, experimental systems were developed to generate data for flow boiling and condensation of refrigerant R134a in two mini channels with serrated fins under various working conditions. Based on these data, machine learning based Artificial Neural Network (ANN) models were trained to predict the heat transfer performance for both flow boiling and condensation. The proposed ANN models predict the heat transfer coefficient satisfactorily with a Mean Absolute Relative Deviation (MARD) of 11.41% and 6.06% for boiling and condensation respectively compared with experimental data. The test results also indicate that using the most important parameters rather than all available parameters is able to formulate a reasonable ANN model. It also reveals that the developed ANN model can be transferred in a different dataset from the same domains but shows an unacceptable performance for different domains. The proposed ANN model can illuminate future study on two phase flow and be used in practical heat exchanger design, which will benefit the advanced thermal management system. [ABSTRACT FROM AUTHOR]

• Local condensation heat transfer in smooth channels and pin fin array is presented. • The effect of operating parameters on the heat transfer characteristics was analyzed. • Sensitivity analysis is performed of heat transfer to identify the critical parameters. • Condensation heat transfer increases with heat flux in the range of vapor quality above 0.4. In this paper, an experimental study of flow condensation was carried out of R134a in a multiport mini channel and a mini channel with pin fin array. The former comprised 20 parallel rectangular channels with an equivalent diameter of 0.64 mm. The latter is a narrow rectangular mini channel containing 10 rows of diamond pin fins with a staggered configuration, and height and longitudinal/transverse pitch of 0.5 mm and 2 mm respectively. To acquire the local value of heat flux and heat transfer coefficient, air jet impingement cooling devices were adopted to condense the vapor of refrigerants. The effects of vapor quality, mass flux, heat flux, and saturation pressure on flow condensation heat transfer coefficient were investigated with the operating conditions: vapor quality from 1 to 0, mass flux from 160 to 450 kg/(m2s), heat flux from 10.0 to 39.8 kW/m2, and saturation pressure from 600 to 1500 kPa. The experimental results indicated that the pin fin array significantly improves the condensation heat transfer coefficient up to nearly 186 % compared to the multiport mini channel in the high vapor quality region. For both channels, the local heat transfer coefficient increases with an increase in vapor quality, mass flux, and heat flux whereas decreases with increase in saturation pressure. The influence of heat flux and mass flux on heat transfer coefficient was more pronounced in the high vapor quality region than in the low vapor quality region. A sensitivity analysis was performed at different vapor quality levels. The most influential parameters in the smooth channel and the pin fin array are saturation pressure and mass flux, respectively. The relative contribution of heat flux is more obvious in the high vapor quality region. Some of the existing correlations have good prediction performance for the smooth channel experimental data in this paper, and the corresponding MAD is within 20 %. However, their deviations are much greater for the applications of pin fin array. [ABSTRACT FROM AUTHOR]