To ensure that the optimised blade for forward flight does not compromise the performance in hover, the rotor with 17 degrees sweep and 11 degrees anhedral was analysed in hover. In addition, its twist was optimised for hover performance. The original blade had a twist of 16 degrees. In addition, a twist of 21 and 11 degrees was analysed using the HMB time marching method. As in the hover case in chapter 7, the optimisation was for maximising the FM over a range of thrust values. A database of 9 values were used i.e. three collective settings for each twist distribution. Table 8.5 shows the results.
It can be seen that the optimised blade performs better than the original blade even with the
Sweep Anhedral Twist CT CQ FM
17.1 11 11 2.20E-002 2.46E-003 0.646576 17.1 11 11 1.50E-002 1.11E-003 0.760250 17.1 11 11 9.78E-003 5.89E-004 0.707279 17.1 11 21 2.20E-002 2.05E-003 0.767487 17.1 11 21 1.49E-002 1.06E-003 0.785730 17.1 11 21 1.02E-002 6.20E-004 0.714826 17.1 11 16 2.20E-002 2.11E-003 0.744721 17.1 11 16 1.49E-002 1.07E-003 0.786982 17.1 11 16 9.94E-003 5.90E-004 0.721870 20 0 16 2.20E-002 2.25E-003 0.700278 20 0 16 1.50E-002 1.10E-003 0.767443 20 0 16 9.85E-003 5.96E-004 0.706070
Table 8.5: Initial CFD database for Hover Optimisation.
same twist distribution in terms of both FM and CQ. An ANN was trained to predict the FM based on the thrust and these predicted values were used to create a database to train the ANN for the optimisation function parameter: FMmaxand∇FM. These predictions are shown in Figure 8.15. The ANN surface created for FM against collective and twist can be seen in Figure 8.16(a). The optimisation function was as follows:
OF V = 0.48F Mmax−0.52∆F M−0.04 (8.20)
Once the GA was run, the optimum was found to be approximately 14.5 degrees of twist, a difference of -1.5 degrees from the original twist distribution as shown in Figure 8.16(b). As can be seen, the change in the OFV around that value of twist changes very little with twist. Therefore, it can be assumed that the hover performance is improved with the new planform design. This is expected, as it is well-accepted that in hover, less sweep and more anhedral benefits the rotor13.
FMmax ∇FM
(a)
(b)
Figure 8.16: (a) ANN prediction of the FM, (b) Optimum twist distribution for hover of the blade optimised for forward flight.
Chapter 9
Fuselage Parameterisation and
Optimisation
Fuselage drag is a major contributor to the overall drag of a helicopter because of its bluff, not streamlined, shape and its additional components such as non-retractable landing gear, weapons, rear-facing surfaces etc. Also, helicopters tend to yaw and fly at some pitch which makes it difficult to obtain a streamlined fuselage at all conditions146.
At low speeds, the effect of the rotor wake on the fuselage also becomes significant104 and further interactional effects contribute to the drag.
9.1
Grid Generation
For the grid generation, standard multi-block topologies (as described in Section 2.3) were gen- erated using ICEM-Hexa and these were projected on the fuselage shapes. The system for HMB makes use of the ICEM-Hexa scripting language to generate geometries in an easy-to-use fashion. These are combined with pre-existing multi-block topologies to produce the meshes.
The outline of the mesh generation process is as follows:
• Step 1: Components of a generic fuselage are generated using ICEM replay files (extension .rpl) from parameterisation coefficients. Example of such a script can be found in Appendix B.8.1.
• Step 2: An ICEMCFD replay file loads the components in ICEM and produces points, curves and surfaces as shown in Figure 9.1.
• Step 3: A pre-defined topology is loaded in a far-field domain as shown in Figure 9.1. • Step 4: The blocks are then re-associated with the geometry.
• Step 5: The mesh is exported to HMB format.
The replay files were generated using a fortran code129. This code (Appendix B.8.3 - B.8.2) reads in a set of parameterisation coefficients and creates a set of geometry points and an ICEMCFD replay file. The replay file can be run in ICEMCFD where it opens the point geometry created by the fortran code and creates the surfaces of the fuselage body. Depending on the geometry, this program has the ability to create matching patch surfaces or full lofted surfaces. Also, if point data is directly available, the program can read this data directly and create the replay files. It also has the ability to close the ends of the fuselage with a surface or at a point. Then another replay file builds the far-field geometry around the fuselage after which the blocking can be associated and the mesh generated.
This parameterisation method was applied to the ROBIN104 and ROBIN-mod710 bodies. Grids were generated and some initial results were obtained which can be found in Appendix C. However,
to demonstrate the optimisation process on a fuselage body, they were not selected in preference to the recent results from JAXA on their JMRTS fuselage7. This body was parameterised using the same technique for the ROBIN fuselage and the optimisation was performed using some of these parameters.
Number of blocks 160 Number of cells 3.3 million Wall spacing 1×10−5
Table 9.1: Table summarising mesh properties for the fuselage.
(a)
(b)