hide
Free keywords:
obstructed view factors, radiative heat transfer, building simulations, task parallelism, hemicube, OpenGL
Abstract:
We present a fully parallelized, high-performance, OpenGL-based approach to compute obstructed area-to-area (A2A) view factors for radiative heat transfer. The A2A view factors are computed from the defining surface integral by Gaussian quadrature. The values of the integrand, i.e. the point-to-area view factors, are computed using an OpenGL-based hemicube (HC) method to efficiently solve the obstruction problem by exploiting the hardware-accelerated visibility testing on modern graphics cards. The final steps to maximize hardware usage are rendering the HC and performing the numerical quadrature in parallel such that data transfer times are completely shadowed by computations. To demonstrate the power of our approach we compute the A2A view factor matrices for a warehouse equipped with ceramic infrared heaters and a test cabin conforming to EN 442 containing a section of a panel radiator. To judge the quality of the results we measure the deviation from unity of the area-weighted column sums of the view factor matrix and the error in radiant flux balance. Compared to a previous, ray-tracing-based implementation, we gain three orders of magnitude in speed in the view factor computation. Conservation of radiant flux is also substantially improved.
