04 Fakultät Energie-, Verfahrens- und Biotechnik

Permanent URI for this collectionhttps://elib.uni-stuttgart.de/handle/11682/5

Browse

Filters

2020 - 20223

Settings

Institute: search.filters.UBSInstitut.Materialprüfungsanstalt Universität Stuttgart (MPA Stuttgart, Otto-Graf-Institut (FMPA))Author: search.filters.author.Schmauder, Siegfried

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    ItemOpen Access
    Experimental investigations of micro-meso damage evolution for a Co/WC-type tool material with application of digital image correlation and machine learning
    (2021) Schneider, Yanling; Zielke, Reiner; Xu, Chensheng; Tayyab, Muhammad; Weber, Ulrich; Schmauder, Siegfried; Tillmann, Wolfgang
    Commercial Co/WC/diamond composites are hard metals and very useful as a kind of tool material, for which both ductile and quasi-brittle behaviors are possible. This work experimentally investigates their damage evolution dependence on microstructural features. The current study investigates a different type of Co/WC-type tool material which contains 90 vol.% Co instead of the usual <50 vol.%. The studied composites showed quasi-brittle behavior. An in-house-designed testing machine realizes the in-situ micro-computed tomography (µCT) under loading. This advanced equipment can record local damage in 3D during the loading. The digital image correlation technique delivers local displacement/strain maps in 2D and 3D based on tomographic images. As shown by nanoindentation tests, matrix regions near diamond particles do not possess higher hardness values than other regions. Since local positions with high stress are often coincident with those with high strain, diamonds, which aim to achieve composites with high hardnesses, contribute to the strength less than the WC phase. Samples that illustrated quasi-brittle behavior possess about 100-130 MPa higher tensile strengths than those with ductile behavior. Voids and their connections (forming mini/small cracks) dominant the detected damages, which means void initiation, growth, and coalescence should be the damage mechanisms. The void appears in the form of debonding. Still, it is uncovered that debonding between Co-diamonds plays a major role in provoking fatal fractures for composites with quasi-brittle behavior. An optimized microstructure should avoid diamond clusters and their local volume concentrations. To improve the time efficiency and the object-identification accuracy in µCT image segmentation, machine learning (ML), U-Net in the convolutional neural network (deep learning), is applied. This method takes only about 40 min. to segment more than 700 images, i.e., a great improvement of the time efficiency compared to the manual work and the accuracy maintained. The results mentioned above demonstrate knowledge about the strengthening and damage mechanisms for Co/WC/diamond composites with >50 vol.% Co. The material properties for such tool materials (>50 vol.% Co) is rarely published until now. Efforts made in the ML part contribute to the realization of autonomous processing procedures in big-data-driven science applied in materials science.
  • Thumbnail Image
    ItemOpen Access
    A physically based material model for the simulation of friction stir welding
    (2020) Panzer, Florian; Shishova, Elizaveta; Werz, Martin; Weihe, Stefan; Eberhard, Peter; Schmauder, Siegfried
    A physically based material model, taking into account the interdependence of material microstructure and yield strength, is presented for an Al 5182 series aluminum alloy for the simulation of friction stir welding using continuum mechanics approaches. A microstructure evolution equation considering dislocation density and grain size is used in conjunction with a description of yield stress. In order to fit experimental stress-strain curves, obtained from compression tests at various strain rates and temperatures, phenomenological relationships are developed for some of the model parameters. The material model is implemented in smoothed particle hydrodynamic research code as well as in the commercial finite element code Abaqus. Simulations for various strain rates and temperatures were performed and compared with experimental results as well as between the two discretization methods in order to verify the material model and the implementation. Simulations provide not only an accurate approximation of stress based on temperature, strain rate, and strain but also an improved insight into the microstructural evolution of the material.
  • Thumbnail Image
    ItemOpen Access
    Many-scale investigations of the deformation behavior of polycrystalline composites: I - machine learning applied for image segmentation
    (2022) Schneider, Yanling; Prabhu, Vighnesh; Höss, Kai; Wasserbäch, Werner; Zhou, Zhangjian; Schmauder, Siegfried
    Our work investigates the polycrystalline composite deformation behavior through multiscale simulations with experimental data at hand. Since deformation mechanisms on the micro-level link the ones on the macro-level and the nanoscale, it is preferable to perform micromechanical finite element simulations based on real microstructures. The image segmentation is a necessary step for the meshing. Our 2D EBSD images contain at least a few hundred grains. Machine learning (ML) was adopted to automatically identify subregions, i.e., individual grains, to improve local feature extraction efficiency and accuracy. Denoising in preprocessing and postprocessing before and after ML, respectively, is beneficial in high quality feature identification. The ML algorithms used were self-developed with the usage of inherent code packages (Python). The performances of the three supervised ML models - decision tree, random forest, and support vector machine - are compared herein; the latter two achieved accuracies of up to 99.8%. Calculations took about 0.5 h from the original input dataset (EBSD image) to the final output (segmented image) running on a personal computer (CPU: 3.6 GHz). For a realizable manual pixel sortation, the original image was firstly scaled from the initial resolution 1080x1080 pixels down to 300x300. After ML, some manual work was necessary due to the remaining noises to achieve the final image status ready for meshing. The ML process, including this manual work time, improved efficiency by a factor of about 24 compared to a purely manual process. Simultaneously, ML minimized the geometrical deviation between the identified and original features, since it used the original resolution. For serial work, the time efficiency would be enhanced multiplicatively.