Thermalhydraulic assessment and performance prediction of supercritical minichannel heat sink with airfoil-shaped obstructions using GA-tuned neural network

Abstract

Performance enhancement of minichannel heat sinks is an active domain of research and the use of supercritical fluids as coolant is only a recent conception. While the minimization of pressure losses is an important design consideration, supercritical fluid poses the additional challenge of heat transfer deterioration when the fluid temperature crosses the pseudocritical value. Present study computationally explores the option of installing airfoil-shaped flow obstructions in a CO2-driven heated minichannel with square cross-section. Four different designs have been conceived, along with the plain minichannel without any obstruction, and each of the design parameters has been varied over a wide range for all the configurations. A performance evaluation criteria has been defined, which combines the heat transport characteristics with pressure losses. Use of obstructions is observed to be advantageous only a limited range of Reynolds number for single and two pairs of airfoils, but expedient over the entire range of supply flow rate with three or more pairs. Increase in distance between neighboring airfoils and width of individual obstruction, along with a vertical orientation, improves the overall performance. Thermalhydraulic data is employed to train a neural network model, using parameters optimized through a Genetic algorithm, for performance forecasting of two different configurations, and the model is able to predict the performance for several test cases with satisfactory level of accuracy.