The objective of this study was to develop an automatic, self-sufficient, preliminary design algorithm for optimization of topologies of branching networks of internal cooling passages. The software package includes a random branches generator, a quasi 1-D thermo-fluid analysis code COOLNET, and multi-objective hybrid optimizer. COOLNET analysis software has the same trends as shown in an earlier publication depicting the results of a similar analysis code used by Pratt & Whitney. The hybrid multi-objective optimization code was verified against classical test cases involving multiple objectives. The number of branches per level was optimized in order to minimize coolant mass flow rate, total pressure drop, and maximize total heat removed. Optimization with four levels of fractal branching channel networks was tested. This optimization varied the number of branching channels extending from each single channel. COOLNET needed approximately forty iterations on average to analyze each configuration. The number of iterations necessary for each geometry depended on the number of branches per configuration. The hybrid multi-objective optimizer needed 500 iterations to create a converged Pareto front of optimized branching network configurations for the case of four branching levels. A population of 60 designs was used. The total number of function evluations analyzed was 30,000. The entire design optimization process takes approximately 3 hours on a single 3.0 GHz Pentium IV processor. In this work the total number of Pareto-optimal designs was 100. After finding the Pareto front points, the user has to decide which optimized cooling network configuration is the best for the desired application. It was demonstrated that this can be accomplished by utilizing Pareto-optimal solutions to create a curve representing pumping power vs. total heat removed and by observing which designs provide favorable break-even energy transfer. The magnitude of the ratio of heat transferred to total pressure drop and ratio of heat transfer to pumping power could be further increased by incorporating the channel’s hydraulic diameter, cross sectional area, lengths, and wall roughness as optimization variables.

This content is only available via PDF.
You do not currently have access to this content.