Speaker
Description
This work presents the development of a software tool for view factor (VF) and radiative exchange factor (REF) calculations on ESATAN-TMS geometry files using Monte Carlo ray tracing (MCRT). VFs and REFs describe the radiation heat transfer between two surfaces. The software can use both the central processor unit (CPU) and the graphics processing unit (GPU) to perform the computations. It is verified against ESATAN-TMS, which shows that the same accuracy levels can be achieved. Since a large number of rays need to be traced to achieve accurate results, calculating VFs and REFs through MCRT is computationally expensive. Several methods to reduce the computation time are presented, together with the capabilities and limitations of the software.
The CPU implementation uses vectorized MATLAB code to partially parallelize the most demanding calculations on a single CPU core. Furthermore, bounding volumes are used to accelerate the ray traversal. Since the rays do not affect each other, the tracing process can easily be parallelized. This is accomplished by using multiple CPU cores to trace multiple rays simultaneously. While these methods decrease the computation time on the CPU by a factor of up to 200, the CPU implementation is still around 100 times slower than ESATAN-TMS.
To further reduce computation times, the parallel architecture of GPUs is utilized. The GPU code is written using the NVIDIA OptiX API to access dedicated ray tracing cores on GPUs. These cores are specifically designed to handle ray tracing tasks such as ray-triangle intersection testing. By distributing the workload across thousands of parallel threads, up to 400 times faster calculation times than ESATAN-TMS are achieved.
