Showing 1–2 of 2 results for author: Geiß, J

Hole Seeding in Level Set Topology Optimization via Density Fields

Authors: Jorge L. Barrera, Markus J. Geiss, Kurt Maute

Abstract: Two approaches that use a density field for seeding holes in level set topology optimization are proposed. In these approaches, the level set field describes the material-void interface while the density field describes the material distribution within the material phase. Both fields are optimized simultaneously by coupling them through either a single abstract design variable field or a penalty t… ▽ More Two approaches that use a density field for seeding holes in level set topology optimization are proposed. In these approaches, the level set field describes the material-void interface while the density field describes the material distribution within the material phase. Both fields are optimized simultaneously by coupling them through either a single abstract design variable field or a penalty term introduced into the objective function. These approaches eliminate drawbacks of level set topology optimization methods that rely on seeding the initial design domain with a large number of holes. Instead, the proposed approaches insert holes during the optimization process where beneficial. The dependency of the optimization results on the initial hole pattern is reduced, and the computational costs are lowered by kee** the number of elements intersected by the material interface at a minimum. In comparison to level set methods that use topological derivatives to seed small holes at distinct steps in the optimization process, the proposed approaches introduce holes continuously during the optimization process, with the hole size and shape being optimized for the particular design problem. The proposed approaches are studied using the extended finite element method for spatial discretization, and the solid isotropic material with penalization for material interpolation using fictitious densities. Their robustness with respect to algorithmic parameters, dependency on the density penalization, and performance are examined through 2D and 3D benchmark linear elastic numerical examples, and a geometrically complex mass minimization with stress constraint design problem. △ Less

Submitted 24 September, 2019; originally announced September 2019.

MSC Class: 74P05; 74P20; 49Q10; 49Q12

Semantic Word Clouds with Background Corpus Normalization and t-distributed Stochastic Neighbor Embedding

Authors: Erich Schubert, Andreas Spitz, Michael Weiler, Johanna Geiß, Michael Gertz

Abstract: Many word clouds provide no semantics to the word placement, but use a random layout optimized solely for aesthetic purposes. We propose a novel approach to model word significance and word affinity within a document, and in comparison to a large background corpus. We demonstrate its usefulness for generating more meaningful word clouds as a visual summary of a given document. We then select keywo… ▽ More Many word clouds provide no semantics to the word placement, but use a random layout optimized solely for aesthetic purposes. We propose a novel approach to model word significance and word affinity within a document, and in comparison to a large background corpus. We demonstrate its usefulness for generating more meaningful word clouds as a visual summary of a given document. We then select keywords based on their significance and construct the word cloud based on the derived affinity. Based on a modified t-distributed stochastic neighbor embedding (t-SNE), we generate a semantic word placement. For words that cooccur significantly, we include edges, and cluster the words according to their cooccurrence. For this we designed a scalable and memory-efficient sketch-based approach usable on commodity hardware to aggregate the required corpus statistics needed for normalization, and for identifying keywords as well as significant cooccurences. We empirically validate our approch using a large Wikipedia corpus. △ Less

Submitted 11 August, 2017; originally announced August 2017.