Refine
Year of publication
Document Type
- conference proceeding (presentation, abstract) (14) (remove)
Has Fulltext
- yes (14) (remove)
Is part of the Bibliography
- no (14)
Institute
Begutachtungsstatus
- begutachtet (1)
- peer-reviewed (1)
Developing new deep-drawn sheet metal parts is a challenging task due to conflicting demands for low costs, durability, and crash properties. Ensuring manufacturability throughout geometrical changes adds to the complexity, leading engineers to rely on experience-driven iterative design changes that compromise requirements and lack reproducibility. Finite Element (FE) simulation models are employed to ensure manufacturability, albeit at the expense of high computational costs and delays in part development.
To improve efficiency, a Machine learning (ML)-centered approach was proposed to ensure manufacturability. However, the limited availability of data raises uncertainty about whether a low- or high-dimensional ML approach is most suitable for drawability assessment.
This work compares the accuracy of a low-dimensional, feature-based Linear Support Vector surrogate and an adapted high-dimensional PointNet model under different dataset sizes. The dataset is composed of parametrically generated, U-shaped structural sheet metal parts.
We use a one-step simulation scheme and evaluate results with a Forming Limit Diagram (FLD) to label drawability. Results show the point of transition to be at about 500 training samples, from which onwards Deep learning is advantageous.
Moreover, the generalizability of these models is tested on a second dataset with topologically similar components. This is to assess the potential for a geometrically more comprehensive evaluation. We discuss several influences on model performances and outline future potentials.
The primary consumers of plant exudates – in exact fungi and bacteria, are representative for the soil succession level from bare soil, which is bacterial dominant to old growth forest constitute by fungal dominance. In a specified level of soil succession, a special kind of plant family benefits on the Fungal to Bacteria Ratio. The ability to determine this ratio in situ without complex chemical applications is part of the project Electronical Laboratory for Intelligent Soil Examination (ELISE). Several mechanical and optical tests on soil samples are covered within this Project. To analyze the fungal to bacteria ratio, samples are prepared automatically – in a defined and reproductive procedure – to generate slides for shadowing microscopy. The samples are observed by a camera, which is attached to a transmitted light microscope. The automatic analysis, done with computer vision algorithms, aims to quantify bacterial and fungal biomass in the actual sample view. Moreover, the algorithm can classify organisms according to their color and shape.
To get a processable picture, several images from different focal levels must be taken through the sample thickness. Parts of each image, are in focus at the actual layer, are merged to a whole depth of field picture, by focus stacking.
This produced picture is used to classify, locate and quantify – in first step filamentous organisms e.g. fungal by image sematic segmentation. The result represents an image sized mask, which indicates the class of fungi with class equivalate values at the pixel positions – covered by the organism. This information is used to calculate the fungal mass per gram soil.
To quantify the bacterial biomass two approaches are implemented. For low density of bacterial existence, the individual bacteria is counted for a part of the field of view by an image detection algorithm to be extrapolate afterwards to the mass per gram soil. For high density of bacterial occurrence, specified regions of interest with only bacteria present are chosen. An image classification which has been pretrained by pictures of bacterial density patterns – previously determent by making the sample countable due to preforming sample dilutions, is done. The second option for high density bacterial count is, to automatically preform dilutions until the image detection is confidently countable.
To ensure a usable confidence score a statistical approach of many fields of view is taken.
The sample preparation for shadowing microscopy, to examinate biological soil properties, is time consuming, manual work. The outcome depends on subjective skills of the operator, furthermore the results are mostly not quantitative. The database on biological soil properties is mostly not sufficient for an integrated modelling on an multidisciplinary scale. This project combines three progressive approaches to develop a tool that is easy to use and gives in situ results that can be used for many purposes.
This paper summarizes six presentations in a session of the track “Use of Simulation for Manufacturing Applications”. The research work deals with the following key issues of this track: Modelling of process problems in manufacturing; Solutions of planning problems in manufacturing; Simulation of processes in manufacturing. This publication shows that the contributions in this track address research questions that are of high importance for industrial practice as well as current research directions such as stochastic optimization or the efficient search of large solution spaces.
With the progress in modern medicine, it was possible to significantly reduce the risks of birth for mother and child. One aspect that has received less attention so far is the risk of injury to the accoucheurs (obstetricians and midwives) during the birth process. Indeed, studies indicate that 92% of midwives suffer from musculoskeletal disorders, with the lower back being the main cause of complaints (72%). The aim of this study was to investigate two commonly used postural techniques used by accoucheurs during childbirth and to analyze the resulting load on the lower back using the AnyBodyTM musculoskeletal simulation software.