Computer Vision-Based Classification of Flow Regime and Vapor Quality in Vertical Two-Phase Flow

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 CO2 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.

Determination of Void Fraction in Microchannel Flow Boiling Using Computer Vision

“…impedance, temperature and differential pressure measurements). This allows not only the formulation of accurate models for the prediction of heat transfer, differential pressure and flow patterns in microchannels but also the use of deep learning algorithms for the tailored description of two-phase flows in different microchannel systems [3].…”

Gas–Liquid Two-Phase Flow Measurement Based on Optical Flow Method with Machine Learning Optimization Model

“…Their results showed a classification accuracy of 95.3% and an average root mean square error (RMSE) of 0.0038, with an average absolute percentage error of 6.3% for gas void fraction measurement. Shai Kadish et al [14] utilized computer vision techniques and deep learning to train CNNs and long short-term memory (LSTM) networks for classifying fluid flow states using video frames as features. They also measured steam mass flow rates within the range of (0.005 to 0.023) kg/s, with an average RMSE of 5% of full scale, achieving a classification accuracy of 92%.…”

Computer-Vision- and Deep-Learning-Based Determination of Flow Regimes, Void Fraction, and Resistance Sensor Data in Microchannel Flow Boiling

“…However, most of the studies focus on the prediction of heat transfer coefficients and pressure drops based on universal consolidated data under the use of artificial neural networks [ 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 ]. So far, only a handful of studies have used machine learning based on convolutional neural networks (CNNs) to automatically detect bubbles, classify flow regimes, and calculate void fractions from HSV images taken during two-phase flow processes [ 23 , 24 , 25 , 26 ].…”

“…In another study, Kadish et al [ 24 ] classified vapor quality and flow regimes of vertical two-phase (vapor-liquid) CO 2 flow images captured at frame rates of 10 fps and 30 fps, respectively, in an 8 mm diameter transparent circular channel using a CNN with ResNet101 for image feature extraction and a deep long short-term memory (LSTM) network to incorporate temporal information of image sequences. The model was trained on a data set of 39,261 manually labeled image frames using cross-entropy loss and the Adam optimization function at a learning rate of 10 −4 , a batch size of 256 for 60 epochs on an NVIDIA® Kepler™ K40 M GPU with 12 GB of GPU accelerator memory.…”