Diz 61 en

Z DCEwiki
Skočit na navigaci Skočit na vyhledávání
Verze k tisku již není podporovaná a může obsahovat chyby s vykreslováním. Aktualizujte si prosím záložky ve svém prohlížeči a použijte prosím zabudovanou funkci prohlížeče pro tisknutí.

Parallel Algorithms for Optimization of Production Systems

Author: Libor Bukata

Disertační práce 2018

The industrial production involves complex processes that directly determine the throughput and manufacturing cost, therefore, it is not surprising that there is a great demand for computer-aided optimization to improve the profitability. Such optimization is, however, typically computationally expensive, and therefore, it is very beneficial to use modern multi-core processors or graphics cards, which can accelerate the optimization about one to two orders of magnitude, in order to find better-optimized processes in a limited time. The transition to the parallel optimization, however, often requires the redesign of the algorithms and good knowledge of architecture. For that reason, it cannot be taken as granted in Operations Research. In this thesis, we propose novel parallel algorithms to solve two problems that are important to optimize production processes. The first problem is the energy optimization of robotic cells which goal is to minimize the total energy consumption without any deterioration in the throughput. The second problem is the Resource Constrained Project Scheduling Problem that is a universal problem applicable in, e.g., the metallurgical industry and assembly shop scheduling. The performance of our algorithms was verified on benchmark datasets. The experiments revealed that the Hybrid Heuristic and Branch & Bound algorithm can optimize industrial-sized robotic cells with up to 12 robots, compared to the existing works where 4 robots were considered at maximum. The Tabu Search algorithm, on the other hand, is designed for graphics cards and its performance is superior to other existing Tabu Search implementations. Besides the benchmarks, the outcomes were also used to optimize an existing robotic cell from Škoda Auto with the result of 20 % energy saving, which indicates that if the optimization is widely used in industry, it will improve the environmental and financial sustainability. The cooperation with industrial partners (Blumenbecker, Škoda Auto) continues within the eRobot project, which main goal is to integrate the proposed algorithms into the digital factory software in order to make the optimization accessible to designers of robotic cells.