Axisymmetric Capsule Shape Variations
One way to achieve the required L/D is to use a nonaxisymmetric shape similar to the AFE shape mentioned previously. A computer-generated shape optimization approach was pursued to attempt to optimize an OML that exhibited some of the desirable characteristics without necessarily being axissymmetric.
The investigation of various "optimized" shapes used the optimization capabilities of the CBAERO computer code. These optimized shapes held the aft-body shape fixed, while the heat shield shape was optimized to meet the trim and L/D constraints. CBAERO permits the very general optimization of the configuration shape, where the actual nodes of the unstructured mesh are used as the design variables. For instance, a typical capsule mesh contained approximately 20,000 triangles and 10,000 nodes. Full shape optimizations were performed where the Cartesian coordinates (x,y,z) of each node were used as design variables. In the example discussed below, there would be 3 x 10,000 = 30,000 design variables.
Often, only the heat shield was optimized, thus reducing the total number of design variables. Figure 5-32 shows the axisymmetric baseline CBAERO grid. The orange region contains those triangles that lie within the optimization region (2,774 nodes, or 8,322 design variables). Figure 5-33 shows one optimization result in which L/D was optimized with the moment constrained to zero and the volume held constant. The resultant geometry exhibits a "trim tab" on the upper windward surface, which the optimizer has produced in an attempt to trim the
Figure 5-32. Baseline Axisymmetric Shape CBAERO Grid
Figure 5-33. An Optimization Result from CBAERO Where the Moment was Constrained to Zero and the Volume Held Constant vehicle while maintaining both the required L/D of 0.4 and the vehicle volume. The surface also exhibits some concavities, which may lead to increased heating or other complex effects. More recent optimization studies have imposed constraints on concavities, and it may be desirable to revisit these optimized shapes or start with the AFE baseline.
The engineering level analysis of CBAERO, as well as efficient coding of the gradient process, enables these optimized solutions to be performed with tens of thousands of design variables and multiple constraints in a matter of minutes-to-hours on a typical desktop Personal Computer (PC). The results shown here typically took 100 to 200 design iterations and less than 60 minutes on a PC laptop.
Various candidate designs were shown to meet both the trim and L/D requirements; however, the complexity of the shapes led to the desire to investigate simpler (but nonaxisymmetric) shapes that might obtain similar results.
Various rotated heat shield concepts were also investigated to examine their ability to reduce the required "z" offset in the CG to trim the vehicle at the desired L/D of 0.4. The various configurations analyzed were capable of reducing the "z" offset; however, the shapes all failed to meet the required L/D of 0.4.
Post a comment