In this study, the flow boiling heat transfer and pressure drop of R245fa in a dimpled flat tube were experimentally measured. The effects of vapor quality and heat flux were analyzed accordingly. The experimental results showed that with an increase in vapor quality, the surface coefficient of heat transfer first increased and then decreased, and the corresponding vapor quality at the maximum surface coefficient of heat transfer was affected by the heat flux. When the heat fluxes were 2 kW/m², 5 kW/m², and 10 kW/m², the surface coefficient of heat transfer reached its maximum value at vapor qualities of 0.35, 0.40, and 0.45, respectively, and then gradually decreased. Increasing the heat flux had little effect on the frictional pressure gradient, but it was beneficial for improving the surface coefficient of heat transfer. The decrease in the surface coefficient of heat transfer with an increase in vapor quality was due to the suppression of nucleate boiling. A comparison of the experimental results for R245fa and R134a indicated that vapor quality and heat flux have different effects on the surface coefficient of heat transfer. Under the same conditions, the ratio of the surface coefficient of heat transfer between R245fa and R134a ranged from 0.64 to 1.31.

Owing to the transformation of data centers toward intelligent computing, the power density per cabinet has increased, leading to an increment in the proportion of cooling-energy consumption. Reducing the power-usage effectiveness (PUE) of data centers is crucial. This study proposes a cooling system that combines natural water sources with liquid-cooling technology. Considering a data center in Laibin, Guangxi as the research object, a system model is constructed to analyze the system performance under different control methods and operating conditions. Subsequently, it is compared with conventional water-cooled chiller cooling systems. The result shows that controlling the supply-air temperature through the chilled-water pump reduces energy consumption by approximately 10% energy and facilitates waste-heat utilization. Additionally, controlling the chilled water-supply temperature through the cooling water pump is more energy efficient, although the energy-saving ratio decreases as the reservoir water temperature increases. The natural water-source cooling system achieves an annual energy-saving rate of 73.5% with a cooling load factor of 0.05, which is lower than that of the conventional system. When the supply-air temperature setpoint is higher, the system energy consumption fluctuates less. Based on a comprehensive consideration, 28 ℃ is a suitable temperature. This study provides a reference for the design of green and energy-efficient data-center cooling systems.

Implementing efficient, clean, and energy-saving cooling solutions is important to decarbonize data centers. This study investigates a variable-spacing multijet direct-chip cooling device that increases the heat-transfer coefficient and improves temperature uniformity. The effects of the coolant flow rate, inlet temperature, pin-fin design parameters, and jet hole spacings on the thermal resistance, pressure drop, standard deviation of temperature, and Nusselt number are investigated. The dataset for the surrogate model construction is obtained based on computational fluid dynamics and Latin hypercube-sampling experimental designs. An artificial neural network model is developed with structural and thermal parameters as inputs and thermal resistance, pressure drop, and temperature uniformity as outputs. An algorithm called constrained multi-objective optimization based on the even search is used to find the solution. The optimization results show that the optimal design outperforms the initial design and has better performance specifications than those of existing studies for both thermal and hydraulic performances. Thus, it has good prospects for engineering applications. Herein, the optimization of variable-spacing multijet direct-chip cooling is investigated, enabling the chip to operate at a higher performance level.