Non-intrusive classification of gas-liquid flow regimes in an S-shaped pipeline riser using a Doppler ultrasonic sensor and deep neural networks

Abstract

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Development of Gas-Liquid Slug Flow Measurement Using Continuous-Wave Doppler Ultrasound and Bandpass Power Spectral Density

“…The shift in Doppler is the fluctuating frequency of an acoustic wave when there is a movement or shift between the source and the acoustic receiver, and the frequency shift is in proportion to the acoustic source velocity [25,31]. The acoustic source velocity can be obtained by calculating the change in frequency between the acoustic source and the receiver.…”

“…The received output signals are then oozed and amplified by the flowmeter electronics. The Doppler frequency shift signals are the processed output signal, and this was obtained using a data acquisition card (NI-PCI-6040E) and a LabVIEW program which controlled a 10 kHz sampling frequency for 120 s for each dataset [25].…”

“…The airflow rate was adjusted by controlling the valves to achieve the desired flow regime. More details of the experimental facility can be found in [24,25,35]. The experimental setup is illustrated in Figure 3.…”

Development of Gas–Liquid Flow Regimes Identification Using a Noninvasive Ultrasonic Sensor, Belt-Shape Features, and Convolutional Neural Network in an S-Shaped Riser

“…Eyo et al [23] developed an online gasliquid objective flow regime identifier using conductance signals and kernel methods and achieved a performance accuracy of 90% against the subjective visual method. Furthermore, another method based on the twin-window feature extraction (TFE) algorithm and deep neural networks (DNN) achieved an outstanding accuracy of 96.28%; the twin-window strategy accomplished significant performance in the onedimensional signal [24]. Kuang et al further propose a flow regime identification benchmark, which mainly covers fully convolutional networks [25].…”

“…The main contributions of this research are as follows: (i) this research appears to be the first implementation of CWDU signals and a CNN-based classifier to identify the flow regime in an S-shaped riser; (ii) this research proposes a novel BSF feature extraction algorithm from the one-dimensional ultrasonic signals, which codes the information to a belt-shaped featurecompared to Fig. 1 Schematic diagram of the multiphase flow test facility [22], [24].…”