Adjoint-based optimization of a rib roughened internal cooling channel for gas turbine applications

Nowadays, gas turbines are among the most widespread energy sources for both energy production and aircraft propulsion. Over the last decades, the turbine has undergone strong improvements in efficiency thanks to the higher turbine inlet temperatures, which easily exceed the material's allowable temperature. Consequently, cooling systems play a key role in protecting the turbine components from high thermal stresses. Rib roughened internal cooling channels are one of the most consolidated cooling technologies in gas turbine blades. In this context, this activity aims to study the cooling performance of a traditional rib roughened internal cooling channel using numerical CFD simulations. Subsequently, the original channel shape was optimized to improve the wall heat transfer coefficient. During the first step of the activity, numerical results were validated with experiments available in the literature under two different flow conditions (Re=21500, 42000). The effect of the turbulence models and the grid size was estimated by comparing the numerical results of the heat transfer coefficient and friction factor with the experimental correlations. Then, an Adjoint optimization method was adopted to obtain the rib and channel shape that maximizes the convective heat transfer coefficient. Adjoint-based optimization is a class of gradient-based optimization algorithms that uses the so-called adjoint equations to calculate the gradients of the objective functions. This information is later used to deform the mesh in the direction of the optimal solution. In this activity, the optimization was performed on a simplified domain, extracted from the entire channel length, to obtain accurate results with lower computational resources. At the end of the optimization, each rib assumed a unique shape that adapts to the coolant flow to achieve the best thermal performance. The optimized ribs are characterized by a w-shape that promotes the generation of counter-rotating vortices and enhances the fluid interaction with the walls. Moreover, the channel cross-section was also modified, resulting in a further contribution to the optimal cooling performance. The impact of the main geometrical features that emerged during optimization was extensively discussed during a CFD postprocessing step. Finally, the error induced by the mesh deformation was quantified and corrected through a remeshing operation on the optimized geometry.