Interface Tracking Algorithms for Multiphase Flow Analysis

Image-processing algorithms that locate Taylor bubble interfaces in PIV-LIF and shadowgraphy images

The Taylor bubble experiments at the Jožef Stefan Institute record a bubble held in place by a downward water flow with a high-speed camera. Before velocity fields can be averaged around the bubble, or its interface compared with simulations, every image has to be split into gas and liquid. In a vertical pipe this is hard: light reflected from the air–water interface is deformed by refraction through the pipe wall and the water, so the signal-to-noise ratio is low.

PIV-LIF Images: Dynamic Masking

Particle Image Velocimetry with Laser Induced Fluorescence (PIV-LIF) needs a dynamic mask that separates the gas phase from the seeded liquid. I developed two algorithms for this:

  • Summation algorithm: uses the high seeding density of the images and builds a mask around the brighter particles
  • Maximum algorithm: uses the higher seeding density at the bubble nose and trailing edge

Both were tested on datasets with different bubble lengths, image resolutions and seeding densities. Both locate the bubble nose and trailing edge reliably, and the summation algorithm is the more robust of the two because it depends less on its settings. Neither reconstructs the full bubble shape accurately enough. The Python code and a small test dataset are available in the project repository.

Shadowgraphy Images: Interface Reconstruction

For bubble shape and interface dynamics the experiments also use shadowgraphy, where dedicated reconstruction algorithms turn each high-speed frame into a bubble contour.

Interface recognition on a shadowgraphy image of a Taylor bubble: (a) raw image, (b) processed image, (c) extracted bubble contour, (d, e) close-ups of the detected interface points.

The reconstructed contours showed:

  • Bubble shape: an asymmetric, bullet-train shape in fully turbulent flow (Re = 5600), but an axisymmetric bubble in transitional flow (Re = 1400)
  • Interface waves: disturbance waves travelling along the interface, tracked with enough sensitivity to resolve their correlated motion across the bubble surface
  • Simulation benchmarks: measured shapes and bubble speeds for comparison with VOF simulations in OpenFOAM and front-tracking simulations in TrioCFD

Results

The algorithms and the results above are published in (Kren & Mikuž, 2021; Kren et al., 2023; Tiselj et al., 2025).

References

2025

  1. Experimental and numerical study of Taylor bubble in counter-current turbulent flow
    Iztok Tiselj, Jan Kren, Blaž Mikuž, and 3 more authors
    Arabian Journal for Science and Engineering, 2025

2023

  1. TaylorExp.jpg
    Dynamics of Taylor bubble interface in vertical turbulent counter-current flow
    Jan Kren, Boštjan Zajec, Iztok Tiselj, and 4 more authors
    International Journal of Multiphase Flow, Aug 2023

2021

  1. Large interface tracking algorithm for air-water slug in turbulent flow
    Jan Kren and Blaž Mikuž
    In Proceedings of the 30th International Conference Nuclear Energy for New Europe, 2021
    NENE 2021