Multi-Objective Optimization of Complex Two-Phase Flow Channels Using Genetic Algorithms

Modern advances in metal additive manufacturing (AM) techniques have given engineers greater design freedom than ever before. In the field of thermal-fluids engineering, optimized and complex flow channels and heat transfer devices can be easily fabricated, improving the heat transfer performance in a number of applications, such as electronics cooling. Two-phase flows are particularly useful in effectively dissipating heat in these high heat flux applications. Unfortunately, few modern design tools for two-phase thermal-fluids systems exist, chiefly because of the extreme complexity of two-phase flow behavior. Therefore, engineers are forced to rely on traditional designs, such as those with straight tubes and fins, to improve the thermal performance of these two-phase systems. This research proposes the use of the Homogenous Equilibrium Model (HEM), boiling heat transfer correlations, and multi-objective genetic algorithms (GAs) to automate the design optimization of complex flow channels, such as those with expanding and contracting radii defined by linear splines and Bézier curves. The HEM is a simplifying two-phase flow model and is useful under specific assumptions and flow conditions, and as such, is only valid for flows undergoing those conditions. Use of other two-phase simplifying models is better suited for other two-phase flow conditions and is of interest for future research. Multi-objective GAs are an optimization technique mimicking natural selection and are best suited for optimization problems with few independent variables and competing objective functions. In the optimization of two-phase flow channel designs presented in this study, these independent variables define the geometry of the channels, and the competing objective functions are minimizing coolant pressure loss and maximizing heat transfer performance. The NSGA-II, a specific multi-objective genetic algorithm, was used with the prior mentioned thermal-fluids models and correlations in MATLAB to optimize channels defined with increasing complexity. When applied to problems with many independent variables, the NSGA-II greatly outperforms exhaustive searches, in which every possible channel design is evaluated. As the number of independent variables defining the flow channels increases, which is necessary to fully capitalize on AM capabilities, the number of possible designs for a given set of design constraints increases exponentially. As a result, the usefulness of GAs becomes increasingly evident with increasing channel complexity. The findings of this study, which are presented in a series of case studies, suggest that as design freedom and channel complexity increase, so do the performance of the optimal channels, which are measured by the coolant pressure loss and overall heat transfer coefficient.