Replication Data for: Bypass-Driven Particle Size Classification in a Taylor-Couette Crystallizer

Bimodal particle suspensions were injected as the tracer, and the pulse response was continuously monitored using a QICPIC online imaging system for PSD analysis, equipped with a custom 3D-printed flow-through cuvette designed for volume flow rates below 100 mL/min. Within the framework of a design of experiments, we first quantified the impact of volume flow rate and rotation rate on the vortex drift velocity ratio, which serves as an indicator of bypass intensity, assumed to be the primary driver of size classification.

Our results show that decreasing the vortex drift velocity ratio decreases the mean residence time of particles, with the effect being more pronounced for larger particles, clearly demonstrating a particle size classifying effect in Taylor vortex flow. Dataset Description Overview This dataset contains raw and processed data as well as the MATLAB scripts used for the data evaluation and figure generation. It includes particle size distribution (PSD) measurements, vortex drift velocity measurements, size-dependent residence time distribution measurements, drawings and CAD files of the developed custom QICPIC flow cell, and further process metadata.

Data is provided as raw data – mostly in the form of video recordings and QICPIC AVI videos – and processed in Excel files and MATLAB data structures, which also contain the relevant metadata of the experiments that were recorded offline. Data containing figures generated via MATLAB code are also included as PNG files, as presented in the related publication, along with the respective MATLAB scripts and helper functions.

Furthermore, diagrams displaying operating points not shown in the related publication are supplied. The major results are summarized in a results summary PDF file. The main experiments were conducted in terms of a full factorial design of experiments (DoE) with two factors.

The experimental error and the confidence intervals were based on a center-point experiment that was conducted four times. The following factors were investigated: A: Rotation rate → 400 rpm to 500 rpm B: Volume flow rate → 23 mL/min to 92 mL/min Structure and Video Raw Data The dataset is organized by the main chapters of the related publication. The data belonging to each main chapter are provided as separate folders, compressed as ZIP archives.

These ZIP archives can be downloaded as a single complete dataset. For chapters that contain data-based figures generated in MATLAB, the corresponding chapter folder is structured to provide both the final figure files and the underlying data required for traceability and reproducibility. The top level of each chapter folder contains the respective figure files used in the related publication.

The chapter folder additionally contains a subfolder named: Data This folder generally contains a MATLAB structure file that includes the relevant raw data, metadata, processed data, and evaluation results used for generating the corresponding figures. In some cases, additional Excel tables are provided, for example, for supplementary calculations, intermediate evaluation steps, or tabulated results. For experiments involving QICPIC recordings, the Data folder also contains the subfolder: ImageDescriptors_PSD_Data This subfolder contains the image descriptor data and particle size distribution data obtained from the evaluation of the QICPIC videos.

The QICPIC video evaluation was performed using custom MATLAB code that has been described and published elsewhere [1-3]. In addition, each chapter folder contains a folder named: Figure Addons This folder provides the MATLAB scripts used to generate the respective figures. If not all operating points or experimental conditions are shown as figures in the related publication, additional figures for the remaining operating points can also be found in this folder.

Due to their large file size, the video files are not included inside the individual chapter ZIP archives. Instead, all raw videos used for particle size distribution measurements, drift velocity measurements, and residence time distribution measurements are provided separately in the top-level folder 0-0_Video_RawData. The QICPIC raw video files were grouped according to the corresponding main chapters and, where applicable, individual experimental runs.

These files are uploaded as ZIP archives, as the file size could be reduced significantly this way. In contrast, the MP4 videos recorded with the camera could not be compressed by ZIP archiving. Therefore, these videos are stored in their original format.

The video file names follow a standardized naming convention to allow direct assignment to the corresponding chapter, measurement method, operating conditions, acquisition settings, video index, and recording date. Video file naming convention The general naming structure is: ChXX-YY_<Method>_<SpecificationOrDoE>_<OperatingParameters>_<AcquisitionParameters>_VidXX_YYMMDD.ext where: ChXX-YY indicates the paper chapter or section in which the data are used, e.g. Ch2-1, Ch3-1, Ch3-2, or Ch3-3. <Method> indicates the recording or measurement method, e.g. QICPIC or CAM. <SpecificationOrDoE> describes the sample, flow cell, or DoE factor level. <OperatingParameters> includes relevant process parameters such as rotation rate and volume flow rate. <AcquisitionParameters> includes video acquisition settings, if applicable, such as frame rate and compression factor (export interval, every Nth frame).

VidXX is a two-digit running video index, where multiple videos were necessary per experiment. YYMMDD is the recording date. File Types .txt, .xlsx, .m (MATLAB script), .mlx (MATLAB live script), .mat (MATLAB data file), .png (graphs), .step (3D CAD file) Use and Purpose This dataset supports the reproducibility of results from a study on particle size classification in a Taylor-Couette Crystallizer.

References [1] Heisel, S., Kovačević, T., Briesen, H., Schembecker, G., Wohlgemuth, K., 2017. Variable selection and training set design for particle classification using a linear and a non-linear classifier. Chem.

Eng. Sci. 173, 131–144. 10.1016/j.ces.2017.07.030. [2] Heisel, S., Rolfes, M., Wohlgemuth, K., 2018.

Discrimination between Single Crystals and Agglomerates during the Crystallization Process. Chem. Eng.

Technol. 41 (6), 1218–1225. 10.1002/ceat.201700651. [3] Lins, J., Harweg, T., Weichert, F., Wohlgemuth, K., 2022. Potential of Deep Learning Methods for Deep Level Particle Characterization in Crystallization.

Applied Sciences 12 (5), 2465. 10.3390/app12052465.