A fully automated evolutionary optimization of a non-standard pump from galvanic industry is presented applying the island model for a high degree of parallelization on large supercomputers. The aim of this optimization is to improve the efficiency and the head of the pump at a specified operating point. The investigated pump is parameterized by nine degrees of freedom. The results of three optimization runs are compared with the baseline. In three optimization runs, the initialization, the migration strategy as well as the number of fully connected islands and the population size per island are varied. In addition, different definitions of fitness are used. The results of the optimization runs yield increased efficiencies and wider operating ranges than the baseline. The definition used for calculating fitness depends on the focus of the desired performance improvement.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Optimization of a Non-standard Pump from Galvanic Industry with the Island Model

  • Ines Waldsteiner,
  • Alexander Tismer,
  • Stefan Riedelbauch

摘要

A fully automated evolutionary optimization of a non-standard pump from galvanic industry is presented applying the island model for a high degree of parallelization on large supercomputers. The aim of this optimization is to improve the efficiency and the head of the pump at a specified operating point. The investigated pump is parameterized by nine degrees of freedom. The results of three optimization runs are compared with the baseline. In three optimization runs, the initialization, the migration strategy as well as the number of fully connected islands and the population size per island are varied. In addition, different definitions of fitness are used. The results of the optimization runs yield increased efficiencies and wider operating ranges than the baseline. The definition used for calculating fitness depends on the focus of the desired performance improvement.