Optimization of the rib structure inside cooling channels with genetic algorithms

Abstract

To increase the thermal efficiency of industrial gas turbines, the performance of the internal cooling channels is of utmost importance. These cooling channels are often equipped with ribs on the top and bottom walls to increasetheturbulenceandthemixingoftheflow.Dependingonthepositionofthecoolingchannelintheturbine blade and the external heat load, the ribs have to be designed differently. Until now, the decision regarding the optimal design parameters was mainly based on the knowledge of the engineer. The present work investigates genetic algorithms as an optimization method to find optimal designs for the rib configuration in internal cooling channels. The main optimization targets are the enhancement of the heat transfer with a low increase in pressure drop. Results are presented for simplified 2D as well as realistic 3D cooling channel configurations. In the present study, a numerical method is investigated for the optimization of the rib structure inside cooling channels. Cooling channels are used for the internal convective cooling of gas turbine blades. The channels are mostly designed as multi-pass systems through the blade. In the main part of the blade, these channels have square or rectangular cross sections, while in the leading and trailing edge region the channels have triangular cross sections. Such channels normally have roughness elements, for example ribs, along two walls. Due to the ribs, the turbulence and the mixing of the flow is enhanced.To increase the thermal efficiency, different rib designs are used, as can be seen in the wide rangeofribangles α,ribheights eandinnerpitchdistances pconsideredinmanystudies.Atypicalgas turbine blade with cooling channels is shown in Fig. 1.The present work investigates an optimization method based on genetic algorithms (GAs) to find optimal designs for the rib configuration in an internal cooling channel. Until now, the choice of the rib angle, rib height and inner pitch distance was mainly based on the engineers' knowledge, obtained from experiments and sometimes numerical computations. Finding an adequate design is a time-consuming task due to the numerous parameters influencing the performance of a cooling channel. Even small changes can influence the turbulent flow and the heat transfer and pressure loss in the channel. One can find a large number of experimental investigations about the heat transfer performance and friction factor characteristics in cooling channels. Some of the important experimental works are summarized in (1-3). Numerical calculations offer a cheap alternative to experimental investigations and have shown good results in previous studies that dealt with heat transfer in internal cooling channels (4, 5). In future, when turbulence models are further developed, they might have the potential to replace more and more experimental work. GAs are an abstraction of natural phenomena. They try to simulate the mechanisms of natural selection or evolution copied from biological organisms. Due to the high complexity of the turbulent flow and heat transfer problems appearing in cooling channels, traditional search algorithms are often not applicable. Furthermore, as gradient-based methods, they often result in local minima or maxima. This problem cannot appear in GAs due to the random search procedure. The whole optimization process is based on the automatic coupling of a GA with other software tools (for the geometry generation ProEngineer (6), for the grid generation CENTAUR (7) and for the calculation of the