The Fluid Mechanics Group, University of Zaragoza (UZ), together with the associated Laboratory for Combustion Research (LITEC) is probably the largest interdisciplinary team in Spain devoted to combustion research, both theoretical and experimental.

UZ used APUS-CFD and the Grid-computing infrastructure (FLOWGRID) to simulate a swirl-type burner. This is a piece of equipment encountered (in different variants) in many Engineering devices in which energy needs to be generated, including domestic boilers, power-station furnaces, industrial ovens and mobile and stationary gas turbines.

The aim of this evaluation aspect is whether APUS-CFD is capable of reproducing the main flow features, as observed and measured from experiments. For the burner, it is expected to find similar outlet profiles. For the combustion chamber, it is expected to find internal and external recirculation regions of similar location and strength to those found in the experimental data. It is also expected to find similar mixing fields, showing the mixing of the two gas streams.

Three different simulations have been carried out with APUS-CFD: the fluid flow in the burner, the inert fluid flow in the combustor and finally the combusting flow in the combustor. Some of the results from these three cases are presented below.

The modelling interest of this case lies in the rather complicated internal geometry. This can be appreciated in the solid model of the figure below. The mesh generated for the simulation of the burner is an unstructured mesh with 2.3 Million tetrahedral cells. Part of the mesh is shown in the figure below.
The Standard k-e turbulence model was used for this simulation; upstream of the burner a flat velocity profile was assumed.

Burner geometry	Burner mesh

This burner geometry serves partly the purpose of generating swirl in the flow, the calculation of which is critical for defining boundary conditions for the combustion chamber. Hence, the evaluation of the accuracy of the results focuses in the contours of the velocity components at the outlet of the burner.

Burner mesh	Velocity contours

The Figure above shows the contours of the Velocity magnitude on the axial plane of the burner.
The velocity profiles right at the burner outlet are shown in the figures below for the y-velocity and z-velocity components. Comparisons of APUS-CFD results with another commercial solver show very good agreement.

Comparison of Velocity components at the outlet of the burner

The mesh used to simulate the combustion chamber has 1 Million tetrahedral cells. Details of the mesh can be seen in the figure below. A close-up of the mesh shows details of the inlets to the combustor, where air inlets are shown in 'blue' and gas inlets in 'red'.

Combustor Mesh	Close-up of the Mesh

The main flow features in this case are the swirl zone near the burner, and the development of the velocity profile along the combustor. Contours of the velociy magnitude are illustrated in the Figure below, together with the region of recirculation. Comparison of the results with APUS-CFD, in the form of velocity profiles at different z positions, are also shown. The values of y velocities are higher in the zone near the burner, due to the strong swirl of the flow in that zone.

Velocity contours in the axial plane	Location of the recirculation region

Velocity profiles at z=8.9 cm	Velocity profiles at z=16 cm

The parallel version of the APUS-CFD solver was benchmarked and tested during the validation exercise. The effect of number of processors on performance was assessed and the numerical efficiency of the solver was observed. For all the tests performed the solver converged with almost the same number of iterations, regardless the number of processors used (from 1 CPU to 64 CPUs).
Execution times for a mesh of 1.8 Million cells are shown in the figures below. The 2-CPU system is a Windows-based AMD-Opteron 3.0GHz server, with 4-GBytes of memory. The 32/64-CPU system is the Altix 3700 Bx2 (1600 GHz). The graph below shows the time reduction for 1000 iterations. Although the 2-CPU AMD-Opteron is a very fast server, it still requires almost 16 hours to reach a converged solution. The same problem however runs in less than 20 mins on a 64-CPU SGI.

Performance of APUS-CFD

Motorsport
Automotive
Power Generation
Land Transportation