CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Study of mesh-moving techniques applied to high-fidelity simulations of compressible flows

Estudiants que han llegit aquest projecte:

Director/a: VENTOSA MOLINA, JORDI

Departament: MMT

Títol: Study of mesh-moving techniques applied to high-fidelity simulations of compressible flows

Data inici oferta: 28-01-2026     Data finalització oferta: 28-09-2026


Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
Adaptive Mesh Refinement, Compressible flows, high-fidelity simulations
 
Descripció del contingut i pla d'activitats:
Sonic and transonic flows are characterized by the presence of shock waves and very strong gradients of pressure, density, and velocity. These flow regimes are common in many engineering applications, such as nozzles, turbomachinery blade passages, and external aerodynamics. Accurately resolving shock waves is essential, as small numerical errors in their prediction can significantly affect the quality of the results.
High-fidelity CFD methods are capable of capturing the complex physics of compressible flows with shocks, but they usually require very fine meshes in shock regions. Classical approaches rely on mesh refinement, which increases the number of grid points and leads to a significant rise in computational cost. An alternative strategy is mesh moving, where the total number of nodes is kept constant and their positions are adjusted to concentrate resolution in regions where shocks occur.
This thesis focuses on the application of mesh-moving techniques for high-fidelity CFD simulations of sonic and transonic flows. The goal is to improve shock resolution by relocating mesh nodes according to flow features, instead of increasing mesh density. The student will simulate different flow configurations with shocks and assess the benefits and limitations of mesh moving compared to fixed meshes.
 
Overview (resum en anglès):
High-fidelity Computational Fluid Dynamics (CFD) simulations require substantial computational resources to accurately capture complex compressible phenomena such as shock waves. Adaptive Mesh Refinement (AMR) techniques address this challenge by dynamically concentrating computational resolution where it is most needed. This thesis evaluates the computational efficiency and numerical accuracy of an r-refinement AMR strategy.

The research was conducted in two phases. First, a dedicated one-dimensional Euler solver was developed to solve the Sod shock tube problem, providing a controlled environment for implementing and investigating the mechanics of dynamic mesh deformation and performing an initial evaluation of its performance. Second, the methodology was extended to SOD2D, a high-fidelity CFD solver. A comprehensive parametric study was then conducted to assess the influence of the compression factor, mesh velocity factor, and shock-sensor threshold on both solution accuracy and computational cost.

The results demonstrate that the performance of AMR is strongly dependent on both the parameters defining the strategy and the characteristics of the underlying flow. In highly transient configurations involving multiple moving shocks, such as a two-dimensional Riemann problem, frequent mesh adaptations introduce a computational overhead that offsets the accuracy gains when compared with uniform static mesh refinement. In these configurations, the most efficient parameter combinations are those that minimize the number of required mesh adaptations. Conversely, for localized and stationary shocks, such as those generated by a supersonic compression ramp, the AMR strategy provides a more favorable accuracy-to-cost trade-off and substantially improves shock resolution. In this case, configurations that induce stronger mesh compression around the shock achieve the best overall performance. Overall, while r-refinement AMR can mitigate artificial dissipation and improve the resolution of shock waves, its computational viability depends on balancing the benefits of localized resolution enhancement against the costs associated with continuous mesh adaptation.

© CBLTIC Campus del Baix Llobregat - UPC