Ray tracing core accelerated high performance implementation for infrared radiation signature simulation using reverse Monte Carlo method

Abstract

Infrared radiation signature is a key characteristic of target and useful for many inverse analyses. Specially, infrared radiation signature simulation plays an increasingly important role in the field of stealth design of air-vehicle and target identification. To improve the simulation efficiency while ensure accuracy, this study presents two novel implementations based on reverse Monte Carlo method. For the first time, hardware ray tracing cores (RT cores) are leveraged to accelerate infrared calculation considering heterogeneous participating media and walls, which is the key innovation compared to the conventional method that only using CUDA cores for acceleration. The first novel method, termed RT method, determines the intersections using bounding volume hierarchy tree (BVH) that implemented on RT cores throughout computational process. Whereas the other, termed CUDA-RT method, determines the intersections between rays and boundaries based on BVH tree that implemented on RT cores, other intersection calculations are done with the help of the adjacency relationships between the grids and are implemented on CUDA cores. Additionally, an improved ray-intersection method is proposed to address complex meshes. A comparison between the numerical and experimental infrared radiation signatures of a rocket-plume case showed the high accuracy of the novel methods. Moreover, a series of numerical examples indicate that both novel methods significantly improved the computational efficiency. Specifically, the CUDA-RT method has the minimum calculation time, approximately 10 % of the reference method with a small additional GPU memory overhead, facilitating improved stealth design and target identification capabilities.