Using MeshTools for Improving Productivity of the CFD Process in Turbomachinery Design: Case Study - Turbine Leading Edge Filleting

28th April 2006, Carlos Felipe Favaretto, Cambridge Flow Solutions Ltd

Introduction

Computational Fluid Dynamics (CFD) has long become a daily tool for gas turbine designers. The possibility of analyzing a large number of different design configurations unsupervised and overnight has enabled designers to reduce the design time and costs.

One of the potential bottlenecks in this process, however, can be regarded as the number of man-hours required to set-up the optimization task from the initial CAD data. Another bottleneck could exist in the mesh generation loop if geometrical changes are made at CAD level and manually. On top of that, often the critical issue in practice, unless the morphed mesh is checked by a robust and reliable quality control routine there is no guarantee that the configurations stored in the database have an acceptable degree of accuracy or are even solvable. With all these issues in mind, the usage of a novel mesh manipulation application (MeshTools) was tested for a typical turbine design task, via the customised application of our standardised MeshTools routines into a self-contained tool to suit the specific requirements of deformation and quality control for turbomachinery CFD analysis. (Alternative applications of the same core technology can be found on our website.)

Description Of Case Study

The first step in the process was to obtain a fluid flow solution for the baseline geometry. Calculations were conducted to verify whether the mesh near the wall was suitable enough for resolving the laminar sublayer. The normalized wall distance (y +) distribution showed that the original grid needed re-meshing near the blade and hub surfaces.

MeshTools was used as a means to control the distance between the viscous layer nodes and the wall, as shown in Fig.1. The original FLUENT case file was converted to a universal format for CFS applications. To preserve the critical mesh quality needed for viscous effects, a viscous layer pointer file was created by MeshTools after reading basic input data from two text files, such as target fraction of element edge length, grid surface number and number of layers for re-meshing. The pointer file stores all viscous layer tree information required for later calculations with MeshTools.

Figure 1 - Flowchart illustrating normalized wall distance modification by use of MeshTools.

After creating the pointer file, new FLUENT mesh files are generated in seconds. This saves a considerable time for the user by eliminating the necessity of revisiting the mesh generator and minimizing the number of clicks required for performing a simple task.

Figure 2 - Flowchart illustrating mesh morphing by use of MeshTools.

The next step after obtaining a satisfactory baseline mesh was to use the core libraries within MeshTools for morphing the mesh according to the desired design modification, as shown in the flowchart (Fig.2). A free-form deformation (FFD) box was specified by vectors stored in the ffddata.dat file. Biasing factors for the FFD control point distribution along the edges of the box were used to control the region of influence of the FFD. Displacement values for the FFD control points were specified in the same text file. The original FFD box and the one with displaced control points are shown in Fig.3.

Figure 3 - FFD boxes with original and displaced control points. a) Original FFD box b) FFD box with displaced nodes

After reading from file all necessary data MeshTools calculates the parametric location of the surface nodes to be morphed in the FFD box. After displacing the control points, the location of the morphed surface nodes is updated from its parametric location in the FFD box. The coordinates of the viscous layer tree nodes are also updated by applying the same displacement as their root surface nodes. In the case where blade viscous layers overlap hub viscous layers a blending function was used for smoothing the nodes at that location. The remaining nodes outside the viscous layers are displaced by using a spring model analogy. This approach usually produces a smooth morphed mesh with minimal cell quality issues, as shown in Fig.4. There are cases, however, where the FFD is relatively complex and it is impossible to avoid the generation of negative volume cells or highly distorted cells. For those cases, MeshTools has special built-in routines that optimize the cell quality. These routines are particularly important in the case of unsupervised design optimization where low quality meshes may produce erroneous results and compromise the final answer.

At this stage, the MeshTools application for this test case is complete, and will produce any number of high-quality meshes around the default design in seconds. Clearly there are many alternative tool configurations – alternate FFD configuration, alternate user-input configuration, prism-layer-based mesh quality control, externally specified kinematics – and other applications can be seen on our website.

Figure 4 - Original and morphed meshes in the viscous layer overlapping region a) Original FFD box b) Morphed Mesh

CFD Results

Figure 5 shows limiting streamline plots for the original and morphed geometries. It is clearly observed that the location of the saddle point was shifted towards the leading edge due to the convex shape of the fillet. The endwall separation line was moved towards the leading edge of the blade, indicating that the penetration of the inlet boundary layer for the filleted case is smaller than the original one. As a consequence, the formation of the new boundary layer beneath the passage vortex occurs at an earlier stage for the filleted leading edge configuration. The surface restricted lines on the suction surface of the filleted blade show a change in the passage vortex lift-off line. The local stagnation pressure loss coefficient contours are shown in Fig.6 at 25%, 48%, 68% and 84% of axial chord. From a qualitative perspective, the size of the high loss region is clearly modified by the morphed geometry. This suggests that the current numerical model was capable of predicting the physical effects caused by the shape of the leading edge fillet.

Figure 5 - Limiting streamline plots for the original and morphed geometries. a) Original FFD box b) Morphed Mesh

Figure 6 - Total pressure loss contours at 25%, 48%, 68% and 84% of axial chord (range: blue = low, red = high).   a) Original FFD box   b) Morphed Mesh

Conclusions

MeshTools proved to be a valuable tool for turbomachinery designers. A simple design change was handled in a matter of minutes without the necessity of revisiting the CAD system or rerunning the mesh generation routine. Customised configuration of standard functionality enabled MeshTools flexible control of the viscous layer nodes, which can be of great help and is indeed critical when tuning the mesh distribution near the walls. Normalized wall distances (y +) were easily managed in a matter of seconds. The final morphed mesh produced by MeshTools was of a high quality and was successfully imported and solved with the commercial code FLUENT.

The CFD results showed some physical insight on the effects of leading edge filleting. The objective here was not to present an optimized leading edge fillet but to demonstrate the capabilities that a turbine designer would have at hand with MeshTools.

This test case demonstrated that MeshTools is a very powerful tool to be integrated with fully automated design optimization systems. Thanks to its built-in and robust quality control routines as well as the automatic y+ control grid requirements from a particular type of CFD solver and turbulence model can be guaranteed.