Citation: Kadish, S.; Schmid, D.; Son,
J.; Boje, E. Computer Vision-Based
Classification of Flow Regime and
Vapor Quality in Vertical Two-Phase
Flow. Sensors 2022, 22, 996. https://
doi.org/10.3390/s22030996
Academic Editors: Yangquan Chen,
Nunzio Cennamo, M. Jamal Deen,
Subhas Mukhopadhyay,
Simone Morais and Junseop Lee
Received: 24 December 2021
Accepted: 23 January 2022
Published: 27 January 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Computer Vision-Based Classification of Flow Regime and
Vapor Quality in Vertical Two-Phase Flow
Shai Kadish
1,
*, David Schmid
2
, Jarryd Son
1
and Edward Boje
1
1
Department of Electrical Engineering, University of Cape Town, Cape Town 8001, South Africa;
jarryd.son@uct.ac.za (J.S.); edward.boje@uct.ac.za (E.B.)
2
The European Organization for Nuclear Research (CERN), 1211 Meyrin, Switzerland; david.schmid@cern.ch
* Correspondence: kdssha001@myuct.ac.za; Tel.: +27-798-908-343
Abstract:
This paper presents a method to classify flow regime and vapor quality in vertical two-phase
(vapor-liquid) flow, using a video of the flow as the input; this represents the first high-performing
and entirely camera image-based method for the classification of a vertical flow regime (which is
effective across a wide range of regimes) and the first image-based tool for estimating vapor quality.
The approach makes use of computer vision techniques and deep learning to train a convolutional
neural network (CNN), which is used for individual frame classification and image feature extraction,
and a deep long short-term memory (LSTM) network, used to capture temporal information present
in a sequence of image feature sets and to make a final vapor quality or flow regime classification.
This novel architecture for two-phase flow studies achieves accurate flow regime and vapor quality
classifications in a practical application to two-phase CO
2
flow in vertical tubes, based on offline
data and an online prototype implementation, developed as a proof of concept for the use of these
models within a feedback control loop. The use of automatically selected image features, produced
by a CNN architecture in three distinct tasks comprising flow-image classification, flow-regime
classification, and vapor quality prediction, confirms that these features are robust and useful, and
offer a viable alternative to manually extracting image features for image-based flow studies. The
successful application of the LSTM network reveals the significance of temporal information for
image-based studies of two-phase flow.
Keywords: flow regime; vapor quality; computer vision; machine learning
1. Introduction
A flow regime describes the spatial distribution between the vapor and liquid phases
in a two-phase flow, with the different regimes being identified by the gas bubble character-
istics inherent to each regime [
1
]. The classification of flow regimes for multiphase flow
is essential in many industrial sectors, such as the energy, metallurgical, and processing
industries. The study of flow regime is important because it reveals essential information
about flow behavior, as well as the physical flow parameters of the two-phase flow under
investigation [
2
]. Different flow regimes can be observed across different flow channel
shapes, orientations, and operating conditions, with different flow regimes also arising
because of properties of the flow itself, including phase velocity and vapor quality [
2
].
Much of the literature aims to relate flow regime classes to the measured physical character-
istics of the flow and to define the regimes and their transitions in terms of these physical
characteristics, for a given set of experimental conditions [2–5].
The physical parameters or features used in the classification of flow regimes are classi-
cally extracted from the flow by direct measurement, using a variety of instruments
[6–15]
.
Some of these instruments require direct contact with the flow [
6
–
10
], while other methods
read flow data in a non-intrusive manner [
11
–
15
]. Image-processing techniques represent a
non-intrusive method by which to extract flow features for the purpose of flow regime study.
Sensors 2022, 22, 996. https://doi.org/10.3390/s22030996 https://www.mdpi.com/journal/sensors
