Adaptive Store Separation

Investigator:  A. Wissink and R. Meakin, Army AFDD (AMCOM)

Application Overview

    Cases which involve relative motion between vehicle components represent an important class of problems to which the present methodology is directly applicable. The method is demonstrated here for a generic store separating from a delta wing.  The case involves a wing that is a NACA 64A010 airfoil in section and a clipped delta wing in plan form with a 45 degree leading edge sweep angle.  The store has a tangent-ogive forebody, a cylindrical centerbody, and a tangent-ogive afterbody. The store diameter (cylindrical section) is 0.067L, where L is the wing root-chord length. The simulation corresponds to a freestream Mach number of 1.2, 0 degree angle-of-attack, and a Reynolds number of 3 million (based on L=15 feet). The problem geometry is indicated in Picture 1, with selected surfaces from the near-body volume grid system shown in Pictures 2 and 3.

Methodology    

    The near-body portion of the domain is discretized using six grid components and 0.47 million grid-points. The default off-body system, adapting only to the proximity of the wing and store, results in 72 grid components and 1.15 million grid-points.  The simulation is initiated for the store in carriage position and carried out to convergence, as measured by force and moment histograms. The store is then released from the wing according to a prescribed motion. Alternatively, the store could be allowed to respond to applied separation forces and aerodynamic loads with virtually no increase in computational expense. However, given the hypothetical nature of this case, a prescribed path is adopted instead.

    Prior to separation, solution adaption is activated, with an adapt frequency of 100 time-steps. Solution and domain connectivity accuracy is maintained in this way. Of course, solution accuracy is maintained as a result of error estimates and adaptive spatial partitioning and refinement. Domain connectivity accuracy, on the other hand, is maintained automatically by virtue of the off-body grid management system. Near-body grids are always surrounded by level-1 (fine) off-body grids with comparable resolution capacity, as indicated in Pictures 2 and 3.  Other overset grid techniques capable of simulating moving body applications do not share this favorable characteristic. Unless special ``drop-zone'' grids are created a-priori, the compatibility characteristics between interpolation donor and intergrid boundary point elements generally degrade with movement of the respective grid systems.

Results

    Picture 4 shows contours from the instantaneous Mach fields with the store in carriage position on the default grid system.   Picture 5 shows contours from the instantaneous Mach fields immediately after the third adapt cycle. Outlines of the finest level (i.e. level-1) Cartesian grids are shown with the solutions to give an indication of the refinement used in the respective background grid systems. The fidelity of the shock and wake system captured in the adapted grid system is superior to that resolvable by the default grid system.

    The simulation is carried out on an IBM-SP. In order to fit the problem into the core memory of the machine, the off-body domain is decomposed after each adaption step to maintain a maximum of 500K grid points per processor. The initial configuration with the default system of grids is small enough to fit on four processors. Each subsequent adaption step added points and, consequently, more groups (i.e. processors) are required to maintain the limit.

Significance

1. Meakin, R., ``Unsteady Aerodynamic Simulation of Static and Moving Bodies Using Scalable Computers,'' AIAA-99-3302-CP, pp. 469-483, 14th AIAA Computational Fluid Dynamics Conference, Norfolk, VA, June 1999.