Toward Improved Aeromechanics Simulations Using Recent Advancements in Scientific Computing

Many types of aeromechanics simulations are very expensive to perform because the computational cost per time step is quadratic or cubic in the discretization. To improve this situation, it is shown how such simulations can be accelerated by using two approaches: 1. Utilizing powerful parallel hardware for high computational throughput; 2. Developing effective fast algorithms to reduce the computational complexity. The hardware method has been implemented by using parallel programming on graphic processors with hundreds cores, and the algorithmic method was realized by an efficient fast multipole method. An application is shown where the use of both methods is combined to give nearly two orders of magnitude reduction in computational cost.