Dynamic structure tracking in multiphase flows using ultrafast X-ray tomography and imaging-assisted scanner control (ROOF)

Motivation:

Today, fast imaging techniques become more and more essential tools for the investigation of highly dynamic processes. One subject of constant scientific interest for such techniques are multiphase flows, which can be found in many technical processes such as chemical multiphase reactors, rectification and extraction columns, crude oil processing, solar thermal direct evaporators or in refrigeration and bioprocess engineering. Especially the modelling of multiphase flows by numerical methods (e.g. CFD) is quite difficult because of the multiplicity of physical effects that must be considered, e.g. for the transport of momentum, mass and heat on different scales.

Objectives:

In the research project, the available ultrafast X-ray tomography system ROFEX at TU Dresden has to be combined with imaging-assisted and automated scanner position control to track moving structures, e.g. particles or gas bubbles, along the height of a process space. Therefore, GPU-based image reconstruction and image analysis algorithms have to be developed in combination with a fast data transfer solution and then applied to a given X-ray tomography scanner. At the end of the first project phase, a validated method will be made available for future experimental studies. In the second project phase, this will be used to investigate the coalescence and breakup behavior of individual bubbles in a bubble swarm. The results will be used to test and optimize integral models of bubble behavior with previously unattainable insights.

Methods and results:

The main objective during this project is the demonstration of feasibility to track a single gas bubble within a gas bubble swarm in a bubble column. Therefore, the following tasks are accomplished:

• Development of a fast data transfer solution for coupling ROFEX to GPU processing
• Optimization and implementation of image reconstruction algorithms for a GPU system
• Development and implementation of automatically operating cross-sectional image analysis algorithms for the selection and direction tracing of single bubbles
• Technical implementation of a tracing unit with an electronic control
• Test of algorithms and hardware on the basis of exemplary experiments

The challenge is to guarantee the real-time capability of data transfer and the subsequent data analysis to be able to achieve real-time control. Thus, an average gas bubble rise velocity of 20 cm/s and a bubble equivalent diameter of ≥ 5 mm requires 80 control decisions per second to be made at minimum. For scenarios with high turbulences even a higher decision rate may be required.

Until now, a positioning unit was built which can position the ROFEX scanner vertically with a veclocity of up to 50 cm/s. Further, the data acquisition, transfer and processing was optimized using massively parallel techniques (FPGAs, GPUs). The latency between image acquisition and image reconstruction could be reduced to less than 2 ms with a maximum throughput of more than 2000 images per second.

In the second project phase of 30 months, the ROFEX scanner will be used to address the following questions:

• Further development of control algorithms for simultaneous detection of possible coalescence partners,
• Preliminary studies of coalescence and breakup in initially monodisperse bubbly flows,
• Experimental investigation of coalescence and breakup due to acceleration in the bubble wake as well as due to differences in buoyancy, and
• Experimental investigation of bubble breakup.

The results of the experimental work will be used to validate and, if necessary, optimize existing models for coalescence frequency and daughter bubble size distribution.

Publications:

D. Windisch, M. Bieberle, A. Bieberle, U. Hampel
Control concepts for image-based structure tracking with ultrafast electron beam X-ray tomography
Transactions of the Institute of Measurement and Control, 42 (4) (2020): 691-703. doi:10.1177/0142331219858048

D. Windisch, O. Knodel, G. Juckeland, U. Hampel, A. Bieberle
FPGA-based Real-Time Data Acquisition for Ultrafast X-Ray Computed Tomography
IEEE Transactions on Nuclear Science, 68 (12) (2021): 2779-2786.
doi: 10.1109/TNS.2021.3123837

D. Windisch, J. Kelling, G. Juckeland, A. Bieberle
Real-time Data Processing for Ultrafast X-Ray Computed Tomography using Modular CUDA based Pipelines
Computer Physics Communications, 287 (2023): 108719.
doi: 10.1016/j.cpc.2023.108719

D. Windisch, C. Kaever, G. Juckeland, A. Bieberle
Parallel Algorithm for Connected-Component Analysis
Algorithms, 16 (2) (2023): 80.
doi: 10.3390/a16020080