In this section, the CFD analysis process described in previous sections is used to provide aerodynamics performance within an overall aircraft design task. A short to medium-range transportation aircraft is selected as a test case to demonstrate the impact of the implemented CFD automated chain over the aircraft synthesis process. A grid refinement study is made to determine the resolution accuracy of the mesh used in CFD simulations. The overall fuel burn obtained by the mission analysis making use of the CFD computed polars is compared with the results obtained by using the VLM solution, and against a pure empirical-based synthesis.
The overall aircraft synthesis process setup in this work is based on a multi-fidelity architecture, to account for the CFD-based analysis process described in the previous sections. The implemented design workflow architecture is shown in Fig. 7.
In the design workflow, the Top Level Aircraft Requirements (TLAR), are specified for the synthesis. The first module is the conceptual aircraft design tool, VAMPzero , which is used as aircraft initializer to provide the initial overall synthesis of the aircraft performance, such as the fuel consumption and operating empty mass (OEM). Based on a multi-fidelity architecture, the program allows making use of the aircraft performance values evaluated by external tools. If any of the aircraft characteristics are already defined in the input dataset, they will be directly inherited instead of being recalculated by VAMPzero analysis modules. This feature allows integrating the presented CFD-based analysis process into the overall synthesis other than using the aerodynamics characteristics estimation available internally to the conceptual tool.
To provide affordable solution, the aerodynamic performance used in this design study is obtained by solving the Euler equations. As a result, the skin friction drag is not accounted into the CFD-based results. To obtain realistic mission fuel values, an estimation of the friction drag is obtained by a method based on the flat plate equivalency for the aircraft wetted area. An available representative engine is chosen to provide performance maps of fuel flow and thrust for pre-defined engines depending on the flight conditions, i.e., Mach number, altitude and thrust setting. Thereafter, mission analysis is performed to simulate an aircraft’s flight on a given flight mission profile, and to determine the mission block fuel of the design mission depending on the given aerodynamic polars, the engine performance and the aircraft geometry.
After the configuration is initialized by conceptual design module, the resulting model is progressed to the other analysis components in the workflow. The aerodynamic performances are evaluated with the described CFD analysis chain. The design workflow architecture allows using tools with different levels of fidelity, such as a conceptual tool, VLM method and CFD method, to evaluate the aerodynamic performance for the synthesis process. Afterwards, the aerodynamics performances are modified by considering friction drag estimation. Hence, mission analysis is performed to update the mission fuel mass based on the conceptual results (e.g., for the design masses), and on the CFD analysis (for the aerodynamics). With the updated mission fuel mass, the design is forwarded once more to the synthesis process, to account for the updates provided by the aerodynamics and mission modules, and to perform an updated synthesis of the aircraft. Currently, in the design process, the geometry of the aircraft at cruise condition is fixed and the aerodynamic performance of the aircraft is assumed to be unchanged. The low speed performance and the control surfaces size of the aircraft are accounted by the conceptual synthesis tool. Thereafter, with the updated synthesized values of OEM and MTOW from the conceptual design tool (VAMPzero in this study), a new mission analysis is performed. The design loop is executed till the convergence of the design masses (OEM, MTOW, and Fuel Mass). In this way, the convergent solution accounts for all the snowball effects in the aircraft synthesis process.
The configuration used in this design case is the D150, which is an A320 like aircraft and has been used as baseline aircraft in previous studies [23, 24]. The main top level aircraft requirements (TLAR) are reported in Table 5. Figure 8 shows the initialized configuration, which is in CPACS format, and visualized by the CPACS geometry interpreter TiGL Viewer.
Before the design case is carried out, a grid refinement study is made. A sequence of three refined grids with grid sizes ranging from 0.9 million cells to 2 million cells, named coarse, medium, and fine, respectively, is generated by varying the global factor defined in Ggrid. A series of Angle of Attack (AOA) from −4° to 4° is run for each grid to generate drag polars at the cruise Mach number of Ma = 0.78. The polars are shown in Fig. 9. It can be observed that the coarse grid is not sufficiently resolved to match the other two polars. However, the medium and fine grids are nearly indistinguishable from each other.
A description of all grids used in this work, as well as the C
D for each grid at C
L = 0.5 in Ma = 0.78 are given in Table 6. The medium grid offers 40% computational savings compared to the fine grid and with acceptable accuracy. To provide an efficient evaluation of the aerodynamic performance, the medium grid is used in the later aircraft synthesis study.
The overall aircraft synthesis results are compared for three cases:
Pure conceptual-based synthesis.
Multi-fidelity synthesis with VLM-based aerodynamics.
Multi-fidelity synthesis with CFD Euler-based aerodynamics.
Figure 10 shows the comparison of drag polars between VLM method and Euler simulation for Mach number equal to 0.2, 0.5 and 0.78. An angle of attack sweep from −4° to 4° is run for each grid to generate drag polars at each Mach number. Each polar include 10 points. In total, 30 CFD simulations are performed to estimate aerodynamic behavior of the configuration and each simulation cost 5 min with 4 cores. As expected, the differences between VLM and Euler polars in subsonic regime are relatively small. However, in the cruise condition, due to the high wave drag, significant differences are shown. For example, at C
L = 0.5, the difference of the C
D can up to 100 drag counts. It worth to be noticed that the wing shape used in CFD analysis is initialized by the conceptual design tool and has not been designed in purpose, which result to unpractical large wave drag compare to that of a satisfactory design where the wave drag takes 4–5% of the total drag.
The result of the synthesis process, such as the take-off mass (MTOM), fuel mass (MFM) and operating empty mass (OEM) for three synthesis cases are shown in Table 7. With extensive available database for conventional configurations, the conceptual synthesis process is calibrated on real aircraft data. As a result, the conceptual design results provide a good reference for comparison. Both multi-fidelity synthesis results, which make use of VLM and Euler CFD simulations, show a difference with the conceptual design case, as shown in Table 7. The main difference for the three cases is due to the fuel consumption at cruise phase, which results from the difference of drag predicted with aerodynamic tools of varying fidelity levels.
Further, due to the snowball effects accounted into the iterative synthesis process, the drag difference results into a different trimming condition at the cruise phase, which is reported in Table 8.
It is obviously that the under estimation of fuel for VLM case is due to the absence of the wave drag. On the other hand, the Euler case tends to give a higher fuel consumption value. The overestimated wave drag at cruise condition results into an increased fuel consumption, and thrust requirements, leading to a higher C
L values to trim the aircraft at the different mission points. Further, as the wave drag predicted by the Euler simulation is highly sensitive to the wing’s geometrical representation, it is crucial to provide suitable input file to the CFD-based synthesis process. Hence, the transition from the conceptual to the CFD-based analysis, needs to account for an enhancement of the geometry quality as well. The automated process presented here may be further extended to generate proper wing shape for CFD simulation within the multi-fidelity synthesis at the early design stages, and this work will also help to decrease redesign effort at the later stages.