Thermal optimization of a single inlet T-junction

We present the thermal optimization study of a single inlet 90 T-junction. Head losses in association with the lumped method are used to calculate the 2D temperature field from a specific geometrical configuration, heat sources, boundary conditions and available ventilation power. This simple case is used as a benchmark to show a possible optimization methodology and to test 4 optimization algorithms suitable to non-linear problems; they are the Active-Set, Interior Point, Genetic Algorithm and Particle Swarm Optimization. The optimization problem is defined by an Aggregated Objective Function (AOF) that depends on three geometrical parameters which are subjected to linear and non-linear constraints. In order to consider not only the global thermal behavior, but also local hotspots, the AOF combines mean solid temperature pondered by its volume and the maximal temperature. Five weights for these criteria were tested. The physical interpretation of different points helped to identify the optimal geometrical characteristics. All optimization algorithms demonstrated to be able to avoid local minima with this benchmark, if properly used. (C) 2011 Elsevier Masson SAS. All rights reserved.