Fast Layered Manufacturing Support Volume Computation on GPUs

We present a GPU-accelerated algorithm for computing a fast approximation of the volume of supports required for layered manufacturing in a given build direction, one criterion often used to choose a direction that requires less time and material. In a sequence of rendering passes that project the part in the given build direction, we use depth peeling to identify faces bounding supports. We exploit programmable graphics hardware to compute the total height of all supports at each projected pixel location, scale the values by pixel area, and finally sum over all pixels to find the total volume of supports. For sample parts tested, our algorithm achieves over 99% accuracy and running times ranging from .2 seconds, for a part with 1,252 facets and depth complexity 2, to 1.86 seconds, for a part with 419,798 facets and depth complexity 9.