<p>A common issue of free-form optimized shell structures is the organic appearance of the shape, which makes it hard for engineers to identify distinguishable features. This is a limiting factor for the acceptance of the method in industrial design, such as the design of metal sheets. This work introduces the generation of engineering shapes such as beading patterns in free-form shape optimization. To this end, the open-close operation from grayscale morphology is combined with the node-based shape regularization technique called Vertex Morphing. Gradient-based optimization of shell structures suffers from a highly non-convex design space, leading to many local minima. To avoid early convergence in a local minimum, the Adam momentum technique is applied to the gradients. With this technique, we may traverse different local minima and effectively explore the design space. We demonstrate the capabilities of the presented workflow using both academic examples and an industrial lightweight application.</p>

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

Engineering features in free-form shape optimization using morphological operators

  • Bastian Devresse,
  • David Schmölz,
  • Armin Geiser,
  • Kai-Uwe Bletzinger

摘要

A common issue of free-form optimized shell structures is the organic appearance of the shape, which makes it hard for engineers to identify distinguishable features. This is a limiting factor for the acceptance of the method in industrial design, such as the design of metal sheets. This work introduces the generation of engineering shapes such as beading patterns in free-form shape optimization. To this end, the open-close operation from grayscale morphology is combined with the node-based shape regularization technique called Vertex Morphing. Gradient-based optimization of shell structures suffers from a highly non-convex design space, leading to many local minima. To avoid early convergence in a local minimum, the Adam momentum technique is applied to the gradients. With this technique, we may traverse different local minima and effectively explore the design space. We demonstrate the capabilities of the presented workflow using both academic examples and an industrial lightweight application.