< Terug naar vorige pagina


Adjoint Shape Optimization based on Large Eddy Simulations through an Adaptation of the Reynolds-averaged Navier-Stokes Equations

Boek - Dissertatie

Industrial design developments are more and more assisted by computer optimization algorithms to speed up the design process. As nowadays the real designs are close to their optimal configurations, the challenge lies in the extraction of the last percentages of improvement, thus accurate evaluations of the performance are required while maintaining the fast industrial time scales. The present thesis proposes a methodology to be used in this context. Initially, a classical adjoint shape optimization based on the Reynolds-averaged Navier-Stokes solution for the evaluation of the flow field has been implemented, in order to obtain a first optimal design at a low computational cost. This approach increases the design performance but could lead to a flawed evaluation of the flow characteristics. The correct prediction of the flow field is of crucial importance to drive numerical optimization to a real optimum design, thus a more accurate simulation is needed. An accurate evaluation of the flow field is ensured by the integration of Large Eddy Simulations in the optimization process, hereby avoiding the known convergence issues related to the chaotic flow motion. The computational cost associated with a more accurate prediction is kept to a minimum by reducing the number of expensive evaluations thanks to the use of a gradient based method in which the adjoint approach is used to evaluate the gradient of the objective function. The calculated gradients are linked to a node-based constrained morphing routine, allowing a large design freedom. The integration with a robust mesh morpher solver leads to successive automatic steps towards the design improvement. Design modifications take into account constraints and limitations. Finally, the feasibility of the design is guaranteed by the application of a smoother with the aim to avoid rough external surfaces. The routines developed have been applied to the optimization of the U-bend test case with the aim to reduce its pressure losses and consequently increase the performance of cooling systems of gas turbines. A validation against experiments has been performed to determine the accuracy of the numerical simulations.
Jaar van publicatie:2020