DaRUS (University of Stuttgart)
Not a member yet
2037 research outputs found
Sort by
micro-XRCT data set of Carrara marble with artificially created crack network: fast cooling down from 500°C
No full textmicro-XRCT data set (projection and reconstructed images) of a Carrara marble core with a diameter of 5 mm. The sample was extracted with a water-cooled diamond drill from a bigger sample with a diameter of 30 mm and a length of 80 mm, which was previously subjected to the following thermal treatment to initiate cracks:
Starting at room temperature (20°C), the sample was heated up at a rate of 3 K/min to 500°C. After a holding time of 2 hours at this temperature, the sample was subsequently quenched in water at room temperature (20°C).
Additionally, a segmentation result, into solid and crack regions, is provided using the U-Net model based on the code provided by Lee et al. (2021, 2022). Note that the segmentation result significantly overestimates the crack volume as described and discussed in detail in Lee et al. (2022) and Ruf et al. (2025). A way to tackle this problem is shown in the related publication Ruf et al. (2025)
Thin Maple Bilayer Dataset
No full textThis dataset contains measured and derived features of 64 small-scale mapple wood bilayer samples. A wood bilayer consists of two bonded layers with differing material or anatomical properties. When exposed to changes in moisture, the layers expand or contract at different rates, producing curvature. This makes bilayers a fundamental system for studying hygro-mechanical behavior, moisture-driven actuation, and material-informed design.
Each bilayer sample in this dataset was subdivided into 19 segments, enabling segment-level analysis that captures local variations in swelling behavior and curvature response.
The samples were fabricated with four passive–active thickness ratios: 2:8, 3:7, 4:6, and 5:5. These ratios highlight how geometric configuration influences moisture-induced deformation.
All data are provided in CSV format, where each row represents a single segment. Please check the engineering metadata for the definition of each column
Kinetic modeling of enzymatic cephalexin synthesis with neural ODEs and surrogate-accelerated Bayesian inference
No full textα-Amino ester hydrolases (AEHs) offer a promising route to the stereoselective synthesis of β-lactams such as cephalexin. However, published kinetic studies have encountered difficulty when extended beyond fitting of the data, indicating practical non-identifiability of the underlying kinetic models. Here, we address this issue using Bayesian inference combined with a reaction-consistent neural ODE surrogate that substantially accelerates parameter estimation. This framework enables efficient development of complex enzyme kinetic models even on limited hardware while providing rigorous uncertainty quantification of all parameters. To account for batch-dependent differences in active enzyme concentration, it was treated as a free parameter in each time series. Using this approach, the number of kinetic parameters was reduced from 12 to 9, and a useful kinetic model was obtained which is identifiable, mechanistically consistent, and predictive even under high substrate conditions.
Available Models
models/model_04.json: The most comprehensive 12-parameter model including all major reaction pathways, competitive inhibition, substrate inhibition, and detailed enzyme regulation mechanisms. This model provides the most biologically detailed description but requires the most parameters to be estimated.
models/model_06.json: A streamlined 9-parameter model that simplifies some regulatory interactions while maintaining core kinetic behavior. This represents a good compromise between detail and parameter identifiability.
models/model_07.json: An intermediate 10-parameter model that includes additional regulatory terms compared to Model 06, capturing more complex enzyme behavior under varying substrate conditions.
models/model_08.json: An optimized 9-parameter model that balances predictive accuracy with parameter parsimony. This model was developed through systematic model reduction to retain essential kinetic features while minimizing parameter uncertainty.
models/model_04_no_e0.json: Identical to Model 04 but with fixed enzyme concentration (E₀) rather than estimating it from data. Use this when enzyme concentration is known or measured separately.
models/model_08_no_e0.json: Identical to Model 08 but with fixed enzyme concentration. This provides a direct comparison of modeling approaches with and without enzyme concentration estimation.
Model File Structure and Components
Each model file (JSON format) contains a complete mathematical description of the kinetic system:
Species definitions: Lists all chemical species with their names and symbolic identifiers used in equations
Constants: Fixed parameters like enzyme concentration (p0) that may be estimated or held constant
ODEs: The system of ordinary differential equations describing how each species concentration changes over time. These equations encode the reaction kinetics and mass balances.
Parameters: Adjustable kinetic parameters (rate constants, binding affinities, inhibition constants) with their prior distributions for Bayesian inference
Algebraic assignments: Complex mathematical expressions that define reaction rates, enzyme-substrate complexes, and regulatory terms as functions of the parameters and species concentrations
The models use symbolic mathematics where enzyme-substrate complexes and reaction rates are expressed algebraically, making them both interpretable and computationally efficient.
System Requirements
Software Dependencies
The analysis pipeline requires several specialized Python packages for scientific computing, probabilistic programming, and machine learning:
<pre
class="sourceCode bash">pip install catalax
Hardware Requirements
The computational analysis is moderately demanding due to Bayesian MCMC sampling and neural network training:
CPU: Multi-core processor (recommended: 12+ cores) - MCMC chains run in parallel across available cores for efficient sampling
RAM: 16GB minimum, 32GB recommended - Memory requirements peak during MCMC sampling when storing large arrays of posterior samples
Operating System and Python Version
Supported OS: Linux or macOS (primary testing on macOS)
Python version: 3.10 or higher required for compatibility with JAX and NumPyro
Shell: Bash-compatible shell for running analysis scripts
How to Reproduce
Quick Start
Install dependencies:
<pre
class="sourceCode bash">pip install catalax
Train the neural ODE surrogate:
<pre
class="sourceCode bash">jupyter notebook TrainNeuralODE.ipynb
# Run all cells to create trained/rateflowode.eqx
Run the complete analysis:
<pre
class="sourceCode bash">export XLA_FLAGS="--xla_force_host_platform_device_count=12" # Adjust number for your CPU cores
chmod +x fit_all.sh
./fit_all.sh
What This Does
The analysis pipeline:
Uses Bayesian inference (MCMC) to estimate kinetic parameters with uncertainty quantification
Compares multiple model complexities (Models 04, 06, 07, 08)
Treats enzyme concentration as a free parameter in each experiment
Generates diagnostic plots and statistical summaries
Saves all results to the results/ directory
Individual Model Analysis
To analyze just one model:
<pre
class="sourceCode bash">python run_inference.py models/model_08.json
For analysis without enzyme concentration estimation:
<pre
class="sourceCode bash">python run_inference.py models/model_08_no_e0.json --no-e0
Outputs
Statistical Results Files
These files contain the quantitative outcomes of the parameter estimation and model evaluation:
{model_name}_summary.csv: Comprehensive MCMC parameter statistics including posterior means, standard deviations, 95% credible intervals, effective sample sizes (ESS), and R-hat convergence diagnostics. This file provides the key numerical results for parameter interpretation.
{model_name}_samples.nc: Complete posterior distribution samples stored in NetCDF format. Contains 10,000 samples × 12 chains for each parameter, enabling detailed uncertainty analysis, prediction intervals, and further statistical computations.
{model_name}_metrics.json: Model performance metrics including various error measures (L1, L2 losses), coefficient of determination (R²), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). These metrics allow comparison of model quality and complexity.
{model_name}_mean_e0.npy: Estimated enzyme concentrations for each experimental measurement (when E₀ estimation is enabled). This file contains the posterior mean enzyme concentrations that can be used for subsequent analyses or experimental validation.
Visualization Outputs
(plots/ subdirectory)
Diagnostic and result plots for model assessment and interpretation:
Trace plots: Time series of MCMC samples for each parameter, allowing visual inspection of mixing and convergence
Corner plots: Two-dimensional projections of parameter correlations and marginal distributions
Posterior distributions: Histograms and density plots showing parameter uncertainty
Model fit plots: Comparison of model predictions vs. experimental data over time
MCMC diagnostics: Monte Carlo Standard Error (MCSE) and Effective Sample Size (ESS) plots to assess sampling quality
Fitted Model Files (models/ subdirectory)
Updated model definitions with estimated parameters:
{model_name}_bi.json: Model with parameters set to Bayesian posterior means. This represents the most probable parameter values given the data and priors, suitable for point predictions and further analysis.
{model_name}_fitted.json: Model with parameters optimized using deterministic methods. These parameters minimize prediction errors and are typically used for the best-fit model predictions.
</ul
Replication Data for: "Simulation of the magnetic Ginzburg-Landau equation via vortex tracking"
No full textData for reproducibility of the numerical simulations of the research paper "Simulation of the magnetic Ginzburg-Landau equation via vortex tracking". The .zip file contains everything needed to generate the plots shown in the paper, as well as the code to run your own simulations with the numerical method presented in the paper.
You just have to extract the .zip file from this repository to access the files and the associated README.md
Optimization of magnetoelectric composites using convolutional neural networks - scripts and dataset
No full textThis repository is intended to host the supplementary material (dataset and scripts) associated with the article “Optimization of magnetoelectric composites using convolutional neural networks”. The HDF5 file 'Microstructure_data.h5' contains the dataset of microstructural images and their effective properties of interest, computed using the finite element method. The individual archive files represent specific deep learning tasks discussed in the associated article. They contain the scripts for building, training, and evaluating the Deep-Inception-based surrogate models for each task.
The document 'README.pdf' summarizes all the necessary information pertaining to the effective usage of this repository such as the contents of the HDF5 file and the various archive files, the required Python libraries, and the procedure to execute the scripts. We recommend reading this document first before proceeding further
Identification and Validation Data for Robust Inverse Feedforward with Hybrid Modeling
No full textExperimental dataset for model identification and validation of feedforward control on a five-axis milling machine.
This dataset belongs to the Open Access publication "Robust Inversion-Based Feedforward Control With Hybrid Modeling for Feed Drives"
(doi: 10.1109/TCST.2024.3512862)
A detailed description of the setup can be found in the publication.
Feedback controllers from frequency inverter:
PI velocity controller with Kp = 0.035 [Nm/(rad/s)], Tn = 2.6 [ms].
P position controller with Kp = 35 [1/s].
The dataset contains the following folders (Please refer to corresponding setup.txt for more details):
const_velocities
measurements for identification with different constant velocities as reference.
velocity_sweeps
measurements for identification with velocity sweeps to estimate the analytical approximation of velocity control loop (with constant velocity offset + sweeps as reference)
execution_time_GP
execution time of feedforward compensation with GP model with unit [μs].
measured in TwinCAT 3 system with the function block "Profiler" (PLC Library-> Tc2_Utilities)
robustness_analytic_model
robustness test against errors in analytical model G0.
robustness_GP_model
robustness test against errors in data-driven GP model.
tracking results
tracking results compared to baseline controller with velocity and acceleration feedforward
</ul
Replication Data for: Experimental Investigation on Adaptive Preload Control for Cable-Driven Parallel Robots
No full textMeasurement data and analysis scripts to investigate a novel adaptive preload control (APC) for cable-driven parallel robots. The measurement data is provided as csv files and consists of data from two devices: the robot control system and the coordinate measuring system (laser tracker). The data from the robot control system is stored in the "experiments/robot" directory. The laser tracker data is stored in "experiments/tracker". Additionally, the motion profiles are stored as nc file in the directory "experiments". The robot-specific parameters are stored in COPacabana.xml. To analyze and plot the data, the "main.py" script must be executed using python and the installed packages in "requirements.txt". Additionally, the WiPy.pyd package must be imported
Replication Data for: INSTINCT: a flow‐based open‐source PNT framework for satellite navigation and sensor fusion
No full textIn order to process the data with INSTINCT a certain file structure is needed
The INSTINCT executable should be placed at the top level
Files with extension .flow need to be placed inside a folder called flow
All data files need to be placed inside a folder called data/GPS-Solutions
The final file structure should look similar to this:
.
├── data
│ └── GPS-Solutions
│ ├── Dataset-1
│ └── Dataset-2
├── flow
└── instinct / instinct.exe
To run a .flow file:
Open INSTINCT
On top in the menu bar press File > Open Flow and select the wanted flow
If all nodes are green, run the flow from the menu with Run > Run Flow
</ul
Trained Vanilla Models on Synthetic Permeability Fields, 101 Data Points
No full textModels are trained with [git: DDUNet] on 101 data points (dp).
Both, vanilla UNet and DDU-Net, can be applied directly end-to-end.
For inference follow the guidelines of Heat Plume Prediction to prepare raw data, then apply the models as described in [git: DDUNet].
Based on raw data from https://doi.org/10.18419/darus-5064
Slab Building Blocks
No full textOverview
This dataset defines building blocks for composing floor slabs with a vast range of triangular and rectangular configurations for spans mainly between four and twelve meters. The shapes are generated at regular intervals for different resolutions: 0,1m, 0,2m, and 1.0m. This dataset is meant as a Benchmark Set definition for the project "Universal Timber Slab (UTS)" under deliverable D1.2.
Approach
It is assumed that by combining the range of shapes described here, the basis for reproducing architectural floorplans is largely covered.
Data Structure
This data format uses the BHoM framework. See: https://bhom.xyz/.
It is possible to load the dataset with the core version of BHoM, even though a UTS BHoM Toolkit was used to organize the data. This approach ensures the data is recognized by our computational design method and aligns it with future developments.
All data is stored in JSON files that contain BHoM objects.
Dataset Descriptions
Triangular Slabs' datasets are generated by always defining the longest edge of a triangle, generating combinations of different edge lengths. Rectangular Slabs' datasets are generated by sampling the width and length of rectangular bays with the specified resolution in the defined range.
Triangular Slabs every 1.0m – Default
Resolution: 1.0m
Longest edge range: 4 to 12m
Samples: 166
Filename: UTS_Datasets_Tri_100cm.json
Triangular Slabs every 0.2m – Optional
Resolution: 0.2m
Longest edge range: 4 to 12m
Samples: 15,510
Filename: UTS_Datasets_Tri_20cm.json
Triangular Slabs every 0.1m – Optional
Resolution: 0.1m
Longest edge range: 4 to 12m
Samples: 119,579
Filename: UTS_Datasets_Tri_10cm.json
Rectangular Slabs every 1.0m – Optional
Resolution: 1.0m
Width range: 2 to 12m
Length range: 4 to 12m
Samples: 61
Filename: UTS_Datasets_Rec_100cm.json
Rectangular Slabs every 0.1m – Optional
Resolution: 0.1m
Width range: 2 to 12m
Length range: 4 to 12m
Samples: 4,941
Filename: UTS_Datasets_Rec_10cm.json
</ol