Fluid flow and heat transfer characteristics of a vertical channel with detached pin-fin arrays arranged in staggered manner on two opposite endwalls

Abstract

This work investigates the fluid flow and the heat transfer characteristics of a vertical parallel plate channel filled with protruding partial length, high density meshing fins on the two inside walls, considering electronic equipment cooling as one of the possible applications. The goal is to understand the flow physics and quantify the overall heat transfer enhancement achieved. Air is used as a coolant and is assumed to have constant thermo-physical properties. Full three dimensional Reynolds-averaged Navier–Stokes equations and the energy equation are solved using finite volume method for the conjugate heat transfer problem in hand. The standard k-є turbulence model with enhanced wall treatment is used to model turbulence closure, in a Reynolds number range of 2200–23,000. The performance of the system is evaluated in terms of the thermal resistance, separately defined for two plates and the pumping power. The numerical results are validated against experimental results, obtained by performing steady state forced convection heat transfer experiments in a vertical wind tunnel facility. A detailed parametric study considering the clearance distance between the plates, pin-fin diameter, transverse pin fin pitch and axial pin fin pitch as the design variables has been undertaken. From the numerical results, it is seen that the pin-fin diameter is the most dominating variable for both the pressure drop and the thermal resistance. The results of the parametric study are used to train an Artificial Neural Network using the Levenberg–Marquardt algorithm to obtain surrogate models for the objective functions. A multi-objective optimization, based on geometrical parameters, is performed using multi-objective evolutionary algorithm, NSGA-II to minimize the thermal resistance and the pumping power simultaneously. A global Pareto optimal set consisting of non-dominated solutions is then obtained using NSGA-II and is divided into five clusters by k-means clustering method to arrive at representative solutions for each cluster.