A Parallel Direct Cut Algorithm for High-Order Overset Methods with Application to a Spinning Golf Ball

TL;DR

This paper introduces a GPU-accelerated parallel direct cut algorithm for overset methods, improving efficiency and scalability in high-order curved grid simulations, demonstrated through complex flow problems including a spinning golf ball.

Contribution

It presents a novel overset domain connectivity algorithm optimized for GPU acceleration and high-order curved grids, enhancing computational efficiency and applicability.

Findings

Algorithm is validated on benchmark problems.

Efficient handling of high-order curved grids.

Successful simulation of flow over a spinning golf ball.

Abstract

Overset methods are commonly employed to enable the effective simulation of problems involving complex geometries and moving objects such as rotorcraft. This paper presents a novel overset domain connectivity algorithm based upon the direct cut approach suitable for use with GPU-accelerated solvers on high-order curved grids. In contrast to previous methods it is capable of exploiting the highly data-parallel nature of modern accelerators. Further, the approach is also substantially more efficient at handling the curved grids which arise within the context of high-order methods. An implementation of this new algorithm is presented and combined with a high-order fluid dynamics code. The algorithm is validated against several benchmark problems, including flow over a spinning golf ball at a Reynolds number of 150,000.