The use of graphics card and nVidia CUDA architecture in the optimization of the heat radiation intensity.

*(English)*Zbl 1374.65114
Chleboun, J. (ed.) et al., Programs and algorithms of numerical mathematics 17. Proceedings of the 17th seminar (PANM), Dolní Maxov, Czech Republic, June 8–13, 2014. Prague: Academy of Sciences of the Czech Republic, Institute of Mathematics (ISBN 978-80-85823-64-6). 150-155 (2015).

The paper focuses on the heating model of the mould surface. This model can be used in the automotive industry for optimal properties of the final product (artificial leather): material structure and colour tone. The main goal of the optimization is to determine the position of heaters, but this goal is extremely difficult to compute using classical gradient methods due to a large number of local extremes. Thus, authors used a differential evolution algorithm, but the solution process is very long for complex surfaces (more than 1000 hours). So, authors modified the whole process to be parallel and they also accelerated it using computations on a graphic card. By this approach, very significant speedups (more than 50) were achieved.

For the entire collection see [Zbl 1329.65003].

For the entire collection see [Zbl 1329.65003].

Reviewer: Šimeček, Ivan (Praha)

##### MSC:

65K10 | Numerical optimization and variational techniques |

80A20 | Heat and mass transfer, heat flow (MSC2010) |

65Y05 | Parallel numerical computation |

49J20 | Existence theories for optimal control problems involving partial differential equations |

##### Keywords:

heat radiation intensity; differential evolution algorithm; parallel programming; GPGPU; gradient methods
PDF
BibTeX
XML
Cite

\textit{J. Mlýnek} et al., in: Programs and algorithms of numerical mathematics 17. Proceedings of the 17th seminar (PANM), Dolní Maxov, Czech Republic, June 8--13, 2014. Prague: Academy of Sciences of the Czech Republic, Institute of Mathematics. 150--155 (2015; Zbl 1374.65114)

Full Text:
Link