Figures & data
Table 1. Weighting function for SG using second-order polynomial and 5 data points.
Figure 8. Grid-independency for wall pressure distribution along the upper and lower lines of the symmetric plane.
![Figure 8. Grid-independency for wall pressure distribution along the upper and lower lines of the symmetric plane.](/cms/asset/f053c195-d956-4677-bcc1-9d86b6692cf3/gipe_a_1894143_f0008_oc.jpg)
Figure 10. Three geometries with the same centre-line curvature and different cross-sections, and (a) their corresponding pressure distributions along the upper and lower lines of the symmetry plane, and (b) average pressure of cross-section.
![Figure 10. Three geometries with the same centre-line curvature and different cross-sections, and (a) their corresponding pressure distributions along the upper and lower lines of the symmetry plane, and (b) average pressure of cross-section.](/cms/asset/0877ca5a-7aa0-4e69-8a8d-08a918ce53f6/gipe_a_1894143_f0010_oc.jpg)
Figure 11. Evolution of geometries and their corresponding pressure distributions from the initial guess to the target geometry.
![Figure 11. Evolution of geometries and their corresponding pressure distributions from the initial guess to the target geometry.](/cms/asset/4e029d7e-36db-444c-8381-550363dd346a/gipe_a_1894143_f0011_oc.jpg)
Figure 13. Results of inverse design with similar cross-sectional profiles (n = 2) for the initial and target geometry using an equally angled grid and radial spines after 150 geometry corrections.
![Figure 13. Results of inverse design with similar cross-sectional profiles (n = 2) for the initial and target geometry using an equally angled grid and radial spines after 150 geometry corrections.](/cms/asset/d5cca9d6-13a7-42d9-a28f-433f0bc445fc/gipe_a_1894143_f0013_oc.jpg)
Figure 14. Results of inverse design with similar cross-sectional profiles (n = 4) for the initial and target geometries using an equally angled grid and radial spines after 150 geometry corrections.
![Figure 14. Results of inverse design with similar cross-sectional profiles (n = 4) for the initial and target geometries using an equally angled grid and radial spines after 150 geometry corrections.](/cms/asset/2bc9cc6a-99e6-4dfc-ab9e-0dfe1a4dace7/gipe_a_1894143_f0014_oc.jpg)
Figure 15. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometries (n = 2) using an equally angled grid and radial spines after 150 geometry corrections.
![Figure 15. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometries (n = 2) using an equally angled grid and radial spines after 150 geometry corrections.](/cms/asset/b5e55ae2-0c9a-4a55-b3c1-43d80d60ff09/gipe_a_1894143_f0015_oc.jpg)
Figure 16. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometries (n = 2) using radial spines after 200 geometry corrections.
![Figure 16. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometries (n = 2) using radial spines after 200 geometry corrections.](/cms/asset/87e078fb-0f59-4c53-82d7-814a1790bba3/gipe_a_1894143_f0016_oc.jpg)
Figure 17. Results of inverse design with similar cross-sectional profiles (n = 2) for the initial and target geometries using an equally angled grid and vertical spines after 200 geometry corrections.
![Figure 17. Results of inverse design with similar cross-sectional profiles (n = 2) for the initial and target geometries using an equally angled grid and vertical spines after 200 geometry corrections.](/cms/asset/a60ab566-6dbb-4dcb-8acf-b63293735ca8/gipe_a_1894143_f0017_oc.jpg)
Figure 18. Results of inverse design with similar cross-sectional profiles (n = 4) for the initial and target geometries using an equally angled grid and vertical spines after 200 geometry corrections.
![Figure 18. Results of inverse design with similar cross-sectional profiles (n = 4) for the initial and target geometries using an equally angled grid and vertical spines after 200 geometry corrections.](/cms/asset/89962336-20ca-409c-8a42-23e9dadbd230/gipe_a_1894143_f0018_oc.jpg)
Figure 19. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometry (n = 2) using an equally angled grid and vertical spines after 200 geometry corrections.
![Figure 19. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometry (n = 2) using an equally angled grid and vertical spines after 200 geometry corrections.](/cms/asset/59f9e2a8-a8f8-46d0-80c8-8c6b63e918a6/gipe_a_1894143_f0019_oc.jpg)
Figure 20. Results of inverse design with similar cross-sectional profiles (n = 2) for the initial and target geometries using an equally angled grid and horizontal spines after 200 geometry corrections.
![Figure 20. Results of inverse design with similar cross-sectional profiles (n = 2) for the initial and target geometries using an equally angled grid and horizontal spines after 200 geometry corrections.](/cms/asset/bcf71ba9-3393-4a0d-8dd8-a6ef826ae9f6/gipe_a_1894143_f0020_oc.jpg)
Figure 21. Results of inverse design with similar cross-sectional profiles (n = 4) for the initial and target geometries using an equally angled grid and horizontal spines after 200 geometry corrections.
![Figure 21. Results of inverse design with similar cross-sectional profiles (n = 4) for the initial and target geometries using an equally angled grid and horizontal spines after 200 geometry corrections.](/cms/asset/acd69697-4d5f-4aa5-832b-dc3b14a07299/gipe_a_1894143_f0021_oc.jpg)
Figure 22. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometry (n = 2) using an equally angled grid and horizontal spines.
![Figure 22. Results of inverse design with different cross-sectional profiles for the initial guess (n = 6) and target geometry (n = 2) using an equally angled grid and horizontal spines.](/cms/asset/52f6595e-2543-4a33-b221-f8f42228b192/gipe_a_1894143_f0022_oc.jpg)
Figure 23. Comparison of wall pressure distributions for the target and final geometry with an equally angled grid along the lateral lines of the wall at various sectors.
![Figure 23. Comparison of wall pressure distributions for the target and final geometry with an equally angled grid along the lateral lines of the wall at various sectors.](/cms/asset/2d677b07-27d8-46f0-8880-1856786870c4/gipe_a_1894143_f0023_oc.jpg)
Figure 24. Comparison of streamline-based and equally angled grid generation for the initial guess and target geometry with the same cross-sectional profile.
![Figure 24. Comparison of streamline-based and equally angled grid generation for the initial guess and target geometry with the same cross-sectional profile.](/cms/asset/42dcc939-8e96-47f7-9226-cf45b6aca091/gipe_a_1894143_f0024_oc.jpg)
Figure 25. Comparison of equally angled (a) and streamline-based (b) grid generation for the initial guess and target geometry with different cross-sectional profiles.
![Figure 25. Comparison of equally angled (a) and streamline-based (b) grid generation for the initial guess and target geometry with different cross-sectional profiles.](/cms/asset/69a169b8-4068-4403-9db4-eeb326dadda7/gipe_a_1894143_f0025_oc.jpg)
Figure 26. Comparison of wall pressure distributions for the target and final geometry with streamline-based grid along the wall streamlines
![Figure 26. Comparison of wall pressure distributions for the target and final geometry with streamline-based grid along the wall streamlines](/cms/asset/62459ac7-c394-4e81-80f2-7c0d0d69cc5d/gipe_a_1894143_f0026_oc.jpg)
Figure 27. Results of inverse design with different cross-sectional profiles for the initial guess (n = 4) and target geometry (n = 2) using the streamline-based grid and horizontal spines after 200 corrections.
![Figure 27. Results of inverse design with different cross-sectional profiles for the initial guess (n = 4) and target geometry (n = 2) using the streamline-based grid and horizontal spines after 200 corrections.](/cms/asset/6e61ca57-a625-4c8b-845f-b611e95c3c7a/gipe_a_1894143_f0027_oc.jpg)
Table 2. Conditions for convergence and unique solution for the inverse design of the s-shape diffuser.
Figure 28. Optimum pressure distribution along the upper and lower lines of the symmetry plane of an S-duct as a TPD for quasi-3D inverse design.
![Figure 28. Optimum pressure distribution along the upper and lower lines of the symmetry plane of an S-duct as a TPD for quasi-3D inverse design.](/cms/asset/9d968090-7b04-42a8-a323-9488e17f089b/gipe_a_1894143_f0028_ob.jpg)
Figure 29. Symmetric plane of the quasi-3D designed S-duct corresponding to the optimum pressure distribution.
![Figure 29. Symmetric plane of the quasi-3D designed S-duct corresponding to the optimum pressure distribution.](/cms/asset/4db27d6b-a15a-4f0c-8c84-379671210c93/gipe_a_1894143_f0029_ob.jpg)
Figure 32. Streamlines and flow separation corresponding to the plane adjacent to the symmetry plane (section D) for the quasi-3D designed S-duct.
![Figure 32. Streamlines and flow separation corresponding to the plane adjacent to the symmetry plane (section D) for the quasi-3D designed S-duct.](/cms/asset/dcbcf912-5bc5-4b26-86ce-4f19be0029cb/gipe_a_1894143_f0032_oc.jpg)
Figure 38. Contours of stagnation pressure in various axial positions for the quasi-3D and 3D designed S-ducts.
![Figure 38. Contours of stagnation pressure in various axial positions for the quasi-3D and 3D designed S-ducts.](/cms/asset/567f28f4-1e18-4576-8582-a3f8f8ac08e0/gipe_a_1894143_f0038_oc.jpg)
Table 3. Comparison between quasi-3D designed S-duct performance and 3D designed.