Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Automated optimization of non-imaging optics for luminaires

: Kudaev, S.; Schreiber, P.


Mazuray, L. ; Society of Photo-Optical Instrumentation Engineers -SPIE-, Bellingham/Wash.; European Optical Society -EOS-:
Optical design and engineering II : 13 - 16 September 2005, Jena, Germany
Bellingham/Wash.: SPIE, 2005 (SPIE Proceedings Series 5962)
ISBN: 0-8194-5980-1
Conference "Optical Design and Engineering" <2005, Jena>
Conference Paper
Fraunhofer IOF ()
Edge-ray principle; LED light source; non-imaging optic; non-sequential ray tracing; optimization algorithm

Specifics of non-imaging optical systems require special algorithms for automated optimization. We have implemented two methods into commercially available optical design software, which are robust and numerically effective. The first one is a modification of the edge-ray principle. In this case the optimization criterion should be expressed in geometrical terms (like, for example, collimation of an extended light source). This gives us the possibility to design not only CPC-like collimators, but also rather complex refractive-reflective (RXI-like) devices. For the second (more general) case the optimization criterion is expressed in energetic terms. In this case stochastic behavior of the merit function due to Monte-Carlo ray-tracing procedure limits the applicability of standard optimization routines available in optical design software. We have realized a direct optimization algorithm, which does not calculate the derivatives of the merit function leading to reduced sensitivity with respect to local statistical deviations. The proposed algorithm is deterministic and does not suffer from redundant trials of random search. As a parametric description for the objects to be optimized we propose the use of piecewise Bezier splines. This allows relative strong shape bending but requires control for intersections. A "red-blue intersection reporting" algorithm is realized as a constraint for optimization.