Combination of X-ray tube and GMDH neural network as a nondestructive and potential technique for measuring characteristics of gas-oil–water three phase flows. (15th January 2021)

• Record Type: Journal Article
• Authors: Roshani, Mohammadmehdi; Phan, Giang; Hossein Roshani, Gholam; Hanus, Robert; Nazemi, Behrooz; Corniani, Enrico; Nazemi, Ehsan
• Is Part Of: Measurement. Volume 168(2021)
• Publication Date: 2021-01-15
• Subjects: GMDH neural networks -- X-ray tube -- Flow pattern -- Volume fraction -- Gas-oil–water -- Three phase flow; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2020.108427
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A system including 1 X-ray tube and 2 NaI detectors combined with GMDH is proposed. 1 GMDH network considered for recognizing flow patterns of a 3 phase flow. 2 GMDH networks were implemented for predicting the volume fractions of components. The recorded spectra in the 2 detectors were assigned as the inputs of the networks. Applying the proposed method all the flow patterns were almost recognized correctly. Abstract: In this investigation, a fan-beam photon attenuation based system, including one X-ray tube and two sodium iodide crystal detectors, combined with group method of data handling (GMDH) neural network is proposed to recognize type of flow regime and predict gas-oil–water volume fractions of a three phase flow. One GMDH neural network was considered for recognizing flow patterns and two GMDH networks were implemented to predict the volume fractions. The recorded photon energy spectra from the two sodium iodide detectors were defined as the inputs of the three GMDH neural networks. The type of flow pattern and volume fractions were the output obtained from the first and the other two GMDH neural networks, respectively. Through the application of the proposed methodology, all of the flow patterns were recognized correctly except one single case. The volume fraction was also predicted with RMS error of less than 3.1.