Scalable Domain Decomposition for Parallel Solution of 3D Finite Element Multibody Rotorcraft Aeromechanics

Abstract

A specialized mesh partitioner is developed for large-scale multibody three-dimensional finite element models. This partitioner enables modern domain decomposition algorithms to be leveraged for the parallel solution of complex, multibody, three-dimensional finite element-based rotor structural dynamics problems. The partitioner works with any domain decomposition algorithm, but contains special features for FETI-DP, a state-of-the-art iterative substructuring algorithm. The algorithm was implemented into an aeroelastic rotor solver X3D, with several modifications to improve performance. The parallel solver was applied to two practical test cases: the NASA Tiltrotor Aeroacoustic Model (TRAM) and the NASA Rotor Optimization for the Advancement of Mars eXploration (ROAMX) rotor blade. The mesh partitioner was developed from two sets of requirements: one standard to any domain decomposition algorithm and one specific to the FETI-DP method. The main feature of the partitioner is the ability to robustly partition any multibody structure, but with several special features for rotary-wing structures. The NASA TRAM, a 1/4 scale V-22 model, was specially released by NASA as a challenge test case. This model contained four flexible parts, six joints, nearly twenty composite material decks, a fluid-structure interface, and trim control inputs. The solver performance was studied for three test problems of increasing complexity: 1) an elementary beam, 2) the isolated TRAM blade, and 3) the TRAM blade and hub assembly. A key conclusion is that the use of a skyline solver for the coarse problem eliminates the coarse problem scalability barrier. Overall, the principle barrier of computational time that prevented the use of high-fidelity three-dimensional structures in rotorcraft is thus resolved. The two selected cases provided a template for how 3D structures should be used in the future. A detailed aeromechanical analysis of the NASA TRAM rotor was conducted. The solver was validated against experimental results in hover. The stresses in the blade and hub components were examined, illustrating the unique benefit of 3D structures. The NASA ROAMX blade was the first rotor blade to our knowledge designed exclusively with 3D structures. The torsional stability, blade loads, blade deformations, and 3D stresses/strains were evaluated for multiple blade designs before the final selection. The aeroelastic behavior of this blade was studied in steady and unsteady hover. Inertial effects were found to dominate over aerodynamics on Mars. The rotor blade was found to have sufficient factor of safety and damping for all test conditions. Over 20 thousand cases were executed with detailed stresses/strains as means of downselection, demonstrating the efficiency and utility of the parallel solver, and providing a roadmap for its use in future designs.