The research data management practices described in this README were developed in accordance with the *Research Data Management Plan Handbook: Guidelines for Developing a Research Data Management Plan* (Tas et al., 2026) for CRC 1615 SMART Reactors for Future Process Engineering. The handbook provides the underlying framework for documenting, organizing, preserving, and publishing research data in alignment with the FAIR principles and the research data management requirements of the German Research Foundation (DFG).[^1]

[^1]: Tas, E., Ruprecht, D., & Knopp, T. (2026). Research Data Management Plan Handbook CRC 1615 (Version V3). Zenodo. https://doi.org/10.5281/zenodo.21126620

## Project Title

CRC 1615 – SMART Reactors for Future Process Engineering, Sub-project C04: SMART continuously operated fluidised bed for spray granulation with self-regulating residence time distribution

## Related Publication

Shaik, A., Pietsch-Braune, S., Rautenbach, R., Weiland, C., Schlüter, M., Pirker, S., & Heinrich, S. (2026). Data-driven Transport Modeling for Tracer Injection and Mixing Time Screening in Stirred Tank Reactors. *Chemical Engineering Science* (submitted). DOI: [To be assigned upon publication]

## Principal Investigators / Authors

- **Asif Shaik** (ORCID: 0009-0007-2062-6317), Hamburg University of Technology, Institute of Solids Process Engineering and Particle Technology
- **Swantje Pietsch-Braune** (PI), Hamburg University of Technology, Institute of Solids Process Engineering and Particle Technology
- **Ryan Rautenbach**, Hamburg University of Technology, Institute of Multiphase Flows
- **Christian Weiland**, Hamburg University of Technology, Institute of Multiphase Flows
- **Michael Schlüter**, Hamburg University of Technology, Institute of Multiphase Flows
- **Stefan Pirker**, Johannes Kepler University Linz, Department of Particulate Flow Modelling
- **Stefan Heinrich**, Hamburg University of Technology, Institute of Solids Process Engineering and Particle Technology

## Funding Acknowledgement

This data set was generated as part of the DFG-funded project *CRC 1615: SMART Reactors for Future Process Engineering* (DFG Project Number: 503850735).

---

# 1. General Information

## Data set Title

Data-driven Transport Modeling for Tracer Injection and Mixing Time Screening in Stirred Tank Reactors

## Short Description

This dataset contains M-Star LBM simulations (v3.12.21, D3Q19 lattice with Smagorinsky LES, Cs=0.1) executed on NVIDIA H200 GPU with 768 GB RAM, producing velocity databases at 4.7×10⁶ lattice points. The M-Star database comprises 900 frames sampled at 200 Hz (Δt=0.005 s) over t=25-29.5 s, establishing a recurrence window of τ_rec=4.5 s. The rCFD implementation requires ANSYS Fluent 2023 R1 or later with Visual Studio 2017+ (Windows) or GCC 4.8+ (Linux) for UDF compilation, operating on a 1.8×10⁶ cell FVM mesh with time step Δt_rCFD=0.025 s and calibrated face-swap diffusion parameter f_swap=0.0825, generating a ~45 GB cell-to-cell velocity database. The reactor configuration consists of a 30 L stirred tank (96 mm ID × 600 mm height) with dual impellers (PBT + RT) operating at 200 rpm in deionized water (998.2 kg/m³) under isothermal single-phase conditions. Post-processing workflows utilize Python 3.8+ (numpy, pandas, matplotlib, scipy, pyvista, opencv-python) and ParaView 5.11+ for VTK visualization.

## Date of Data Collection

2023–2025

## Geographical Coverage

Hamburg, Germany

## Keywords

recurrence CFD, rCFD, stirred tank reactor, mixing time, lattice Boltzmann method, M-Star, computational speedup, experimental validation, parametric optimization, tracer injection

---

# 2. Methodological Information

## Data Collection and Processing

Velocity field data were generated using M-Star CFD v3.12.21 (commercial lattice Boltzmann solver) on NVIDIA H200 GPU. The D3Q19 lattice model with Smagorinsky LES turbulence model (Cs=0.1) was employed at resolutions of 4.7×10⁶ lattice points. Velocity fields were sampled at 200 Hz over t=25-29.5 s (900 frames total) to capture statistically stationary turbulent flow. Data conversion from node-centered LBM format to cell-centered FVM format was performed using custom PyVista scripts, achieving R²=0.972 correlation with NRMSE=16.7%. Experimental mixing time measurements were obtained using conductivity probes (4 M NaOH tracer) at two locations within the stirred tank, with 11 experimental repetitions yielding τ₉₅=25±3 s.

## Experimental Design / Study Context

The study validates recurrence CFD (rCFD) methodology for rapid mixing time prediction in a 30 L dual-impeller stirred tank reactor (pitched blade turbine + Rushton turbine, 200 rpm). rCFD exploits pseudo-periodic flow patterns by reusing pre-computed velocity fields from a short recurrence window (4.5 s) to simulate extended mixing processes (up to 60 s). The calibrated rCFD model (f_swap=0.0825) was validated against experimental conductivity probe measurements and compared with conventional CFD methods. A parametric study explored 90 tracer injection locations (3 radial × 3 height × 10 angular positions) to identify optimal injection strategies minimizing mixing time. The workflow consists of: (1) M-Star LBM velocity field generation, (2) LBM-to-FVM data conversion, (3) rCFD database creation, (4) rCFD mixing simulations with species transport, and (5) post-processing and validation.

## Data Validation and Quality Assurance

All velocity field conversions were validated through correlation analysis (R²=0.972) and normalized root-mean-square error (NRMSE=16.7%) comparing node-centered LBM data with cell-centered FVM interpolations. rCFD mixing time predictions (τ₉₅=24.5±1.0 s) were validated against experimental measurements (τ₉₅=25±3 s, n=11 repetitions), demonstrating closer agreement than M-Star (τ₉₅=30.1±3.0 s) or ANSYS Fluent (τ₉₅=28.1±2.7 s). Parametric study results were verified for physical consistency (e.g., injection near baffles yielding slower mixing, mid-radius injection providing optimal performance). All Python post-processing scripts were tested on Linux and Windows platforms with reproducible figure generation.

---

# 3. Data and File Overview

## List of Files and Structure

| File / Folder | Description | Format | Size |
|---------------|-------------|--------|------|
| `01_Reactor_Geometry/` | CAD models, STL files of 30 L stirred tank reactor (vessel, impellers, baffles, probes) | STL, MSB | ~10 MB |
| `02_M-Star_LBM_Simulations/` | M-Star LBM velocity field database, input.xml files | VTK, HDF5, XML | ~166 GB* |
| `03_rCFD_Implementation/` | rCFD framework: LBM-to-FVM conversion scripts, cell-to-cell velocity database, ANSYS Fluent case files, UDFs | Python, Binary, C, Fluent case | ~50 GB |
| `04_Experimental_Validation_Data/` | Conductivity probe time-series measurements (P1, P2), mixing time summary (11 repetitions) | CSV, TXT, PDF | ~5 MB |
| `05_Parametric_Injection_Study/` | 90-case parametric study results: mixing times for all injection locations, heatmaps, optimal configurations | CSV, PNG, SVG | ~20 MB |
| `06_Post_Processing_Scripts/` | Python scripts for flow field verification, validation comparisons, tracer visualization, computational speedup analysis | Python, CSV, PNG | ~50 MB |

*Note: M-Star velocity database may be provided as subset or external link due to size. Full database regenerable from input.xml files.

## File Naming Convention

- M-Star VTK output: `[geometry_name].[timestamp].vtp` (e.g., `StaticBody.2.46246e-04.vtp`)
- rCFD database files: `ip_[frameID].txt` (e.g., `ip_000001.txt`)
- Parametric study results: `mixing_time_r[radial]_theta[angle]_y[height].csv`
- Python scripts: `[analysis_purpose]_[dataset].py` (e.g., `velocity_validation_LBM_vs_FVM.py`)

## Number of Records / Observations

- M-Star velocity fields: 900 frames (200 Hz sampling, 4.5 s duration)
- rCFD parametric study: 90 injection configurations
- Experimental measurements: 11 repetitions of mixing time experiments
- Validation probes: 2 conductivity probe locations per experiment

---

# 4. Access and Licensing Information

## Repository and Persistent Identifier

Published via TORE (TU Hamburg Open Research Data Repository), DOI: https://doi.org/10.15480/882.16811

## License for Use

**Recommended**: CC BY 4.0 (Creative Commons Attribution 4.0 International)

All data (geometry, velocity fields, experimental measurements, parametric study results, post-processing scripts) are released under CC BY 4.0, requiring attribution via citation of the primary publication and dataset DOI.

**Restriction**: The rCFD core algorithm is proprietary to Johannes Kepler University Linz (Prof. Stefan Pirker). Access to the rCFD solver source code requires permission from stefan.pirker@jku.at. This dataset provides application-specific setup files and user-defined functions (with proper citations) but not the core rCFD solver.

## Access Restrictions

Open Access. No embargo period.

**Note**: Due to file size constraints (~166 GB M-Star velocity database), the full velocity field dataset may be provided as external download link or subset. Complete regeneration instructions using input.xml files are included in `02_M-Star_LBM_Simulations/README.md`.

## Text for Citation

Shaik, A., Pietsch-Braune, S., Rautenbach, R., Weiland, C., Schlüter, M., & Heinrich, S. (2026). Recurrence CFD for Stirred Tank Reactors: Experimental Validation and Parametric Optimization Dataset. TORE - TU Hamburg Open Research Data Repository. https://doi.org/10.15480/882.16811

---

# 5. Reproducibility and Software Dependencies

## Software Required

**M-Star LBM Simulations**:
- M-Star CFD v3.12.21 (commercial LBM solver, https://www.mstarcfd.com/)
- NVIDIA H200 GPU (or equivalent CUDA-capable GPU with 768 GB RAM)

**rCFD Simulations**:
- ANSYS Fluent 2021 R1 or later (commercial CFD solver)
- Visual Studio 2017+ (Windows) or GCC 4.8+ (Linux) for UDF compilation
- rCFD core algorithm (requires permission from Prof. Stefan Pirker, stefan.pirker@jku.at)

**Post-Processing and Data Conversion**:
- Python 3.8 or later
- Required Python packages: numpy, pandas, matplotlib, scipy, pyvista, opencv-python
- ParaView 5.11+ (for VTK visualization, https://www.paraview.org/)

**Hardware Recommendations**:
- **Minimum** (post-processing only): 16 GB RAM, 4 CPU cores, 50 GB storage
- **Recommended** (rCFD execution): 64 GB RAM, 32 CPU cores, 200 GB SSD
- **M-Star simulations**: NVIDIA H200 GPU equivalent, 768 GB RAM, 500 GB storage

## Scripts and Workflow

All post-processing scripts are located in `06_Post_Processing_Scripts/` with subdirectories for each analysis task:
1. `01_Flow_Field_Verification/`: LBM-to-FVM conversion validation (Figure 3 in publication)
2. `02_Methodology_Validation/`: rCFD vs M-Star vs Fluent vs Experimental comparison
3. `03_Dynamic_tracer_conc/`: Tracer concentration evolution visualization
4. `04_Data_Extraction_Rautenbach_data/`: Extract experimental validation data from Rautenbach et al. (2026)
5. `05_Computational_speedup/`: Performance analysis and speedup calculations

Each subdirectory contains a `README.md` with execution instructions. Python scripts can be executed sequentially following the numbered directory structure.

## Reproducibility Notes

To reproduce key results from the publication:

1. **Velocity Field Database**: Use M-Star input.xml files in `02_M-Star_LBM_Simulations/01_Coarse_Resolution_4p7M/` to regenerate velocity fields (or use provided VTK files if available).

2. **LBM-to-FVM Conversion**: Execute Python scripts in `03_rCFD_Implementation/01_LBM_to_FVM_Conversion/` to convert node-centered M-Star data to cell-centered FVM format.

3. **rCFD Mixing Simulations**: Load ANSYS Fluent case files from `03_rCFD_Implementation/03_rCFD_test_setup_30L_STR/`, compile UDFs, and execute with f_swap=0.0825.

4. **Validation Comparison**: Run Python scripts in `06_Post_Processing_Scripts/02_Methodology_Validation/` to compare rCFD (τ₉₅=24.5±1.0 s), M-Star (29.1±1.4 s), Fluent (29.2±2.9 s), and Experimental (24.9±2.7 s) results.

5. **Parametric Study**: Results for all 90 injection configurations are available in `05_Parametric_Injection_Study/` as CSV files with accompanying heatmaps and statistical analyses.

**Note**: rCFD core solver access requires permission from Prof. Stefan Pirker (JKU Linz). This dataset provides all setup files and pre-computed results for validation purposes.

---

# 6. Ethical and Legal Aspects

## Data Protection

Not applicable. This dataset contains no personal or sensitive data. All data are computational simulation results, experimental sensor measurements from industrial equipment, and derived analyses.

## Consent Statement

Not applicable. No human participants or personal data involved.

---

# 7. Versioning and Updates

## Version Number

v2.0

## Date of Release

v1: 2026-03-03
v2: 2026-07-31

## Change Log

**v1.0** (2026-03-03): Initial release
- Complete dataset with 6 main directories
- M-Star LBM simulations (4.7M coarse resolution, reference to 90M high-res)
- Full rCFD implementation (LBM→FVM conversion, IP files, Fluent UDFs, calibrated f_swap=0.0825)
- Experimental validation data from Rautenbach et al. (2026)
- 90-case parametric injection study (2.25 hours computational time)
- Complete post-processing pipeline (Python scripts, validation results, figures)

**v2.0** (2026-07-31): second release
- All most same folder structure and data expcept one change: M-Star LBM simulations of 90M lattice is no more considered in the comparison due to unfair comparison

---

# 8. Contact Information

## Corresponding Author

**Name**: Asif Shaik
**Institution**: Hamburg University of Technology, Institute of Solids Process Engineering and Particle Technology
**Email**: shaik.asif@tuhh.de
**ORCID**: https://orcid.org/0009-0007-2062-6317

## Project Website

CRC 1615 SMART Reactors: https://www.tuhh.de/crc1615