DaRUS (University of Stuttgart)
Not a member yet
2037 research outputs found
Sort by
Replication Data for: A Deep-Learning Based Incidence Operator for Adjustable and Solver-Agnostic Modelling of Adaptive Façades
No full textThis dataset contains the files:
- code: All Python files to replicate result data, figure and table generation.
- data/raw: The training data including all investigated feature sets as well as the target. The training data for the incidence operator (y) were generated using the Grasshopper file and the plugins presented in the study (https://doi.org/10.18419/opus-17170). For the feature matrix X1, a uniform grid of 11 tint states between [0% and 100%] was generated. The 10,000 zenith and azimuth angles per discrete tint state are determined using the Fibonacci distribution. The feature sets X2,X3,X4 are derived from X1 as described in the paper. Additionally, this folder contains the Grasshopper file with all information about the investigated parametric design.
- data/results: The Optuna study results for all feature sets.
- figures: The result figures within the paper.<br
Replication Data for: "BSP k-Means"
No full textThis dataset contains reference implementations in C++ and CUDA for the algorithm presented in the paper "BSP k-Means". The purpose of the data is to allow replication of the reported results
Direct and Indirect Measurement of Complex Poisson's Ratio - Direct Measurement in Tension
No full textThis data set contains directly determined complex Poisson's ratio from axial and transversal strain measurements. Here, the axial and transverse strains were measured locally with strain gauges (K-CXY3-0015-3-350-O, HBK, Darmstadt, Germany) on cylindric polymethyl methacrylate (PMMA, EH-Design, Wörrstadt, Germany) samples with a diameter of d = 5 mm.
Frequency measurements were performed with a rheometer (MCR 702, linear motor, Anton-Paar, Graz, Austria) in the range of 1 Hz to 100 Hz with an axial strain of 0.01 % at constant temperatures in the range of 15 °C to 105 °C. 500 periods were measured per frequency and recorded using a measuring amplifier (Universal Amplifier MX1615B, HBK, Darmstadt, Germany).
Transversal and axial strain is then measured on the PMMA sample with strain gauges in tension mode.
The material response in the time domain is transformed to the frequency domain using the Fast Fourier Transform. This gives the axial and transverse amplitude as well as the axial and transverse phase shift. With the variable from the frequency domain, the complex Poisson's ratio is calculated in post-processing.
The data set contains the calculated complex Poisson's ratio of three measured PMMA samples
Direct and Indirect Measurement of Complex Poisson's Ratio - Direct Measurement in Compression
No full textThis data set contains directly determined complex Poisson's ratio from axial and transversal strain measurements. Here, the axial and transverse strains were measured locally with strain gauges (K-CXY3-0060-3-350-O, HBK, Darmstadt, Germany) on cylindric polymethyl methacrylate (PMMA, EH-Design, Wörrstadt, Germany) samples with a diameter of d = 30 mm.
Dynamic mechanical analysis (DMA) was performed with the piezoelectric actuator (8P-035.20P, Physik Instrumente, Karlsruhe, Germany) driven by the high power amplifier (E-482, Physik Instrumente, Karlsruhe, Germany). Small strain amplitudes
excitation in the frequency range from 0.1 Hz to 1000 Hz are performed. To ensure the oscillation around a strain amplitude, a static preload is applied by a universal testing machine (RM50, Schenk, Germany).
Transversal and axial strain is then measured on the PMMA sample with strain gauges in compression mode. The material response in the time domain is transformed to the frequency domain using the Fast Fourier Transform. This gives the axial and transverse amplitude as well as the axial and transverse phase shift. With the variable from the frequency domain, the complex Poisson's ratio is calculated in post-processing.
The data set contains the calculated complex Poisson's ratio of three measured PMMA samples
exaFOAM Microbenchmark MB11 - Pitz&Daily Combustor
No full textThis work is part of the exaFOAM project that aims to enable the open-source CFD software OpenFOAM to exploit massively parallel HPC architectures and overcome performance scaling bottlenecks. The case is based on the experiment carried out by Pitz and Daily 1983, who measured a combustion flow formed at a backward facing step. The goal of the work was to study the turbulent shear layer during a combustion process in conditions similar to those of the industrial and aircraft gas turbine combustors. The premixed combustion is stabilized by recirculation of hot products which are mixed with cold reactants in a turbulent shear layer. The setup with a backward facing step is one of the simplest configurations reproducing these conditions. Detailed information and case setup can be found in the README.md file contained in the case setup file
Replication Data for: Learning Compensation of the State-Dependent Transmission Errors in Rack-and-Pinion Drives
No full textThis dataset contains all experimental data that is shown within the paper "Learning Compensation of the State-Dependent Transmission Errors in Rack-and-Pinion Drives".
Rack-and-pinion drives are commonly used in large machine tools to provide linear motion of heavy loads over long travel distances. A key concern in this context is the achievable path accuracy, which is limited by assembly and manufacturing tolerances of the gearing components in conjunction with load-dependent deformation and the inherent backlash of the system. To address this issue, this paper presents a method for robust modeling of the individual and state-dependent transmission errors of a drive utilizing a two-stage machine learning approach. Based on this, the position control is extended to include an error compensation, which suppresses the modeled deviations in the mechanical system including the position-dependent backlash. The achievable increase in path accuracy as well as the robustness of the approach are evaluated and quantified by an experimental validation on a system with industry standard components.
The data are structured to correspond to the figures in the publication and are available in TAB or Excel format:
Fig. 2 TE measurements:
Measured transmission errors of the examined rack-and-pinion drive in both directions of motion under varying external load.
Fig. 4 Path errors:
Comparison of calculated and measured path errors for different velocities with no external load.
Fig. 6 Model training:
Training data for the deformation regression models and the predictions of the trained neural network and the regression tree ensemble.
Fig. 8 Compensation validation sine:
Evaluation of the compensation of the transmission errors and backlash for a sinusoidal trajectory.
Fig. 9 Compensation validation overall:
Evaluation of the improvement of the path accuracy by the compensation for varying loads and velocities.
</ul
Data for: Transforming Laser-Scanned 750 kW Turbine Surface Geometry Data into Smooth CAD for CFD Simulations
No full textNote for access: The data is available to anyone interested, but in order to monitor access, we ask that interested users request access by logging in by using the account of their academic institution, selecting the files they want, and clicking "Request Access" If you do not have access through your institution, please contact us by clicking "Contact Owner", enter your email address, and mention the list of files you need. In both cases, please include a note stating your institution and your purpose for using the data.
This dataset contains the point cloud files, the resulting airfoil B-splines, connectors, surface meshes, and surface CAD files for the three blades presented in the paper "Transforming Laser-Scanned 750 kW Turbine Surface Geometry Data into Smooth CAD for CFD Simulations," which correspond to the three blades of the WINSENT test site.
In this dataset, the blades are named according to the numbers identified on the tip of each physical blade: 008, 022, and 025, corresponding to blades A, B, and C, respectively, in the paper.
Note: each blade was processed independently by the script, and as such, there will be small differences at the root. This means that the match between the blade root and the hub attachment point should be adjusted individually for each blade prior to simulation. Also, note that the blades are aligned along the z-axis, and the tower was thus tilted equivalently to the shaft axis; it is likely necessary to untilt the whole turbine prior to simulation. Additionally, note that the 0-pitch angle does not match the angle in the x-y plane in the given blade geometries. To have the blade in the almost (should be better than +/- 0.5°) proper 0° pitch position, you should rotate the blade by -7.5° around the z-axis.
The explanations for the individual files are provided in the accompanying file descriptions. Supplementary information is also given here for convenience.
The most accurate reconstructions of the blades are those with the closed trailing edge. These reconstructions closely adhere to the flat and slightly rounded trailing edge observed on the physical blades and from the scan data.
The FreeCAD files include surfaces generated from 193 original 3D-smoothed B-splines created by the splprep package in Python, as part of the automated reconstruction, interpolation, and smoothing program described in the paper. Both the FreeCAD and .iges CAD files are recognized as the official blade geometries for the WINSENT wind turbines. The differences between the closed and open trailing edge versions of the provided blade sufaces are localized to the actual trailing edge and do not perceptibly influence the trailing edge thickness. To facilitate the use of the open trailing edge CAD data, a surface being only the trailing edge itself is also provided. For those interested in CFD or generating meshes of the turbine blades, the Plot3D structured and STL surface meshes of the blade (with the last 0.2% of the blade radius cut off and left as a hole) generated directly from the splprep B-splines from the developed software are available. Alternatively, the same data is also provided as a series of Pointwise connectors. Both the Plot3D and connectors are in ASCII format, whereas STL is in binary format.
For those wanting to vary the level of smoothing used in the 3D-smoothing step, the B-splines prior to 3D-smoothing are also provided in Python's Pickle format.
There are two source cloud datasets given: A) The original clouds from the laser scanning campaign, with most surrounding artifacts removed but without any modifications to the point data; these clouds are in ASCII format (.txt), compressed as bz2 tar archives. The first three columns provide x, y, z coordinates, and the fourth column the intensity of the laser reflection, which is necessary for realigning the scans using the targets that were placed on the blade surfaces. See the
origData_ScanCampaignBladeArrangement.pdf file to see which scan file corresponds to which blade and side. The cloud file names beginning with Friday are for all pressure and suction side scans, and those beginning with Monday are for all leading and trailing edge scans. Note that for the LE and TE scans, the blades were slightly bent under their own weight and in different directions for each edge. No bending was observed in the PS and SS scans. B) The manually preprocessed clouds for each of the three blades. Preprocessing includes unbending of the leading and trailing edge clouds, removal of any non-blade artifacts, aligning/merging of the clouds for the suction and pressure sides to yield only one such cloud per blade, the division of the leading and trailing edge clouds at approximately the leading and trailing edges themselves, and finally, the separate treatment of the tip portion to eliminate the need for automatic alignment at the tip. These clouds are in CloudCompare format (.bin) because the automatic treatment by the developed software was partially conducted using the CloudComPy API.
Notes about the WINSENT test site and its data
This repository is meant to contain geometric data of the WINSENT turbines: the tower and hub data was closely approximated using a partial laser scan and photos of the standing Northern wind turbine while the blades come from a very careful reconstruction of the Southern turbine's blades. At the time of writing this document, in January 2024, both Northern and Southern turbines are identical by design, and the differences between their blades are expected to lie within the tolerances seen between the three blades of the Southern turbine.
For all data coming from the sensors installed on-site, please register and log in here: https://winsent-gui.zsw-bw.de/
For the positions of the two turbines and four met masts, either use the provided 'WINSENT_Test_Site.kml' file (e.g. by loading it in Google Earth) or the following coordinates (precise to the meter):
GK3 (Gauß-Krüger, Bessel, Zone 3)
Northern Turbine: (3561699, 5392296)
Southern Turbine: (3561656, 5392158)
NW Metmast: (3561565, 5392292)
SW Metmast: (3561520, 5392153)
NE Metmast: (3561823, 5392305)
SE Metmast: (3561791, 5392165)
WGS84
Northern Turbine: Latitude: 48.6652184, Longitude: 9.8365878
Southern Turbine: Latitude: 48.6639818, Longitude: 9.8359836
NW Metmast: Latitude: 48.6651956, Longitude: 9.8347682
SW Metmast: Latitude: 48.6639502, Longitude: 9.8341368
NE Metmast: Latitude: 48.6652871, Longitude: 9.8382723
SE Metmast: Latitude: 48.6640314, Longitude: 9.8378171
The terrain data is also available in the WINSENT_Elevation.tec and WINSENT_Trees_Buildings.tec files, but an official request must be made to receive them. Don't forget to also request the WINSENT_Terrain_Infos.pdf file to have some context on the data.
The authors gratefully acknowledge the funding of the project WINSENTvalid (grant no. 03EE2048C) by the German Federal Ministry for Economic Affairs and Climate Action (BMWK).
This work has been partially supported by the MERIDIONAL project, which receives funding from the European Union’s Horizon Europe Programme under the grant agreement No. 101084216.
The support of Prof. Norbert Haala and Dr. Michael Kölle from the Institute for Photogrammetry and Geoinformatics of the University of Stuttgart for the preparation, execution and postprocessing of the scanning process is also recognized
Data for: Electronic Moment Tensor Potentials include both electronic and vibrational degrees of freedom
No full textData for "Srinivasan, P., Demuriya, D., Grabowski, B. et al. Electronic Moment Tensor Potentials include both electronic and vibrational degrees of freedom. npj Comput Mater 10, 41 (2024). doi:10.1038/s41524-024-01222-9
The dataset contains three folders:
Data for the four figures in the manuscript. This also includes the thermodynamic properties with different contributions.
Nb and 3. TaVCrW: Both these contain the respective MTPs that make up the eMTP, along with the training set for each of the MTPs, the INCAR and KPOINTS used for generating the training data; the Hesse matrices used as reference in the thermodynamic integration; and the free energy contributions in parametrised form.
further details about the dataset can be found in the Methods section of the manuscript.</p
VQA-MHUG
No full textWe present VQA-MHUG - a novel 49-participant dataset of multimodal human gaze on both images and questions during visual question answering (VQA), collected using a high-speed eye tracker. To the best of our knowledge, this is the first resource containing multimodal human gaze data over a textual question and the corresponding image. Our corpus encompasses task-specific gaze on a subset of the benchmark dataset VQAv2 val2. Our dataset is unique in that it is the first to provide real human gaze data on both images and corresponding questions and, as such, allows researchers to jointly study human and machine attention.
We use our dataset to analyse the similarity between human and neural attentive strategies learned by five state-of-the-art VQA models: Modulated Co-Attention Network (MCAN) with either grid or region features, Pythia, Bilinear Attention Network (BAN), and the Multimodal Factorised Bilinear Pooling Network (MFB). While prior work has focused on studying the image modality, our analyses show - for the first time - that for all models, higher correlation with human attention on text is a significant predictor of VQA performance. This finding points at a potential for improving VQA performance and, at the same time, calls for further research on neural text attention mechanisms and their integration into architectures for vision and language tasks, including but potentially also beyond VQA.</p
ASTE: An artificial solver testing environment for partitioned coupling with preCICE
No full textThis dataset contains a software archive of the artificial solver testing environment ASTE (version v3.3.0), its corresponding documentation (as a PDF) and the compatible ASTE turbine tutorial. The documentation of ASTE is its README.md (in 'docs/README.md') converted to a PDF. The turbine tutorial is part of the preCICE tutorials in the preCICE distribution v2404.0. The dataset is connected to a publication in the Journal of Open Source Software (see related software).
The artificial solver testing environment (ASTE) allows to replace and imitate models coupled via the coupling library preCICE by artificial ones, potentially in parallel distributed across multiple ranks on distributed memory.
This helps in the development of preCICE, adapters, or simulation setups by reducing the necessary software components, simplifying execution workflows, and reducing runtime of the case.
This version of ASTE was tested with preCICE v3.1.2. Other preCICE v3 releases should be compatible as well.</p