Performance and accuracy enhancements of radiative heat transfer modeling via Monte Carlo.

摘要

Two ways to reduce the computational requirements of radiative heat transfer Monte Carlo simulation are explored. First, an efficient algorithm for tracing particles in large, arbitrarily complex, planar geometries containing nonparticipating media is presented. For arbitrary triangles and/or convex planar quadrilaterals, an efficient intersection algorithm is discussed in detail. After surveying several techniques used in ray tracing to limit the number of surfaces tested, the method of Uniform Spatial Division (USD) is implemented. The efficiency of the intersection algorithm and USD are demonstrated by timing results.; Second, improving the accuracy of the Monte Carlo results by applying reciprocity and closure is explored. Statistical theory is applied to the reciprocity estimation smoothing (RES) technique which combines reciprocity enforcement through estimation and closure enforcement through the technique of least-squares smoothing. By examining a large number of runs of two large geometries, several RES methods are compared to find the best method. The effects of the RES method on surfaces and individual results between surfaces are also explored. Estimates of the improvements caused by the RES method that can be calculated from the results of a single run are also derived.