Elsevier

Measurement

Volume 51, May 2014, Pages 34-41
Measurement

Precise volume fraction prediction in oil–water–gas multiphase flows by means of gamma-ray attenuation and artificial neural networks using one detector

https://doi.org/10.1016/j.measurement.2014.01.030Get rights and content

Highlights

  • Volume fraction was predicted in multiphase flow using one detector.

  • ANN model was proposed to predict oil, water and air percentages.

  • MCNP-4C code was used for simulation.

  • Trained ANN model predicted the percentages with MAE about 1%.

Abstract

Artificial neural network (ANN) is an appropriate method used to handle the modeling, prediction and classification problems. In this study, based on nuclear technique in annular multiphase regime using only one detector and a dual energy gamma-ray source, a proposed ANN architecture is used to predict the oil, water and air percentage, precisely. A multi-layer perceptron (MLP) neural network is used to develop the ANN model in MATLAB 7.0.4 software. In this work, number of detectors and ANN input features were reduced to one and two, respectively. The input parameters of ANN are first and second full energy peaks of the detector output signal, and the outputs are oil and water percentage. The obtained results show that the proposed ANN model has achieved good agreement with the simulation data with a negligible error between the estimated and simulated values. Defined MAE% error was obtained less than 1%.

Introduction

Multiphase flow measurement is a matter of high importance in oil and gas industry, especially in applications such as reservoir management, field development, operational control, flow assurance and production allocation [1]. By determining volume fractions of oil–water–gas in oil productions in combination with velocity measurement of each phase, the mass flow rate can be determined, which is a useful parameter for monitoring the productions. Utilizing nuclear techniques such as neutrons [2], [3] and gamma ray [4], [5], [6], [7], [8], [9], because of their ability to measure volume fractions without modifying the operational conditions and being non-invasive, is so useful. For the first time, Abouelwafa and Kendall [4] proposed a multi-energy gamma attenuation technique to resolve three-phase mixture component ratios. They examined various static mixture of oil–water–gas in a 0.1 m diameter pipe section using cobalt-57 (122 keV) and barium-133 (365 keV) radioisotopes and a lithium-drifted germanium based detector. Li et al. [10], also analyzed static mixtures in a cubic conduit using americium-241 (59.5 keV) and cesium-137 (662 keV) radioisotopes and a sodium iodide detector crystal. Tjugum et al. [11], produced a multi-beam gamma-ray instrument with an americium-241 source and three detectors. Two detectors, measured gamma ray attenuation across the pipe flow, while the third one measured the scattered radiation. The obtained results were more accurate than single beam. Generally, one of the difficulties in this field is the mathematical modeling to correlate the measurements with volume fractions. In such situations, artificial intelligence techniques, especially artificial neural networks (ANN) can be so helpful. Many researchers have used these techniques in gamma densitometry [7], [12], [13], [14], [16] in order to overcome such difficulties. Cong et al. [17] reviewed applications of ANNs in flow and heat transfer problems in nuclear engineering. In 2009, Salgado et al. [13] proposed a methodology based on neural network to predict volume fractions. They simulated a system comprised of three detectors (one of them for transmitted gamma rays and two of them for scattered) and dual energy gamma-ray source (Eu-152 with energy 121 keV and Ba-133 with energy 356 keV) using N-particle (MCNP) code. In this work, following their investigation, a system with MCNP code has been simulated, which number of detectors was reduced from three to just one detector and only transmitted gamma-rays were considered. Advantage of using less detectors, is the costs decrement as an important criterion in industry. Also it can make the operation of the system easier. When the input of ANN is a signal, one of the most important points is how to choose the features as the ANN inputs. In previous works [13], [14], 58 and 106 features using two and three detectors were considered as ANN inputs respectively. But in this study, only two features using one detector were selected. Our attention was concentrated on improving the precision of volume fractions prediction in annular regimes with only one detector using artificial neural network.

Section snippets

Simulation results

The first step in this investigation was the Monte Carlo simulations in order to generate the training and testing set for the ANN using MCNP. MCNP is a Monte Carlo N-particle code that can be used for neutron, photon, electron, or coupled neutron/photon/electron transport [18]. The Monte Carlo technique is a widely used simulation tool for radiation transport. MCNP-4C code was used to simulate gamma-ray absorption from a radiation source in annular regime in an oil–water–gas pipeline.

Results and discussion

Regression diagrams of Monte Carlo simulation results and predicted ones for both outputs have been shown in Fig. 4, Fig. 5. The comparison between simulated and predicted results (using ANN) has been tabulated for training and testing data in Table 3, Table 4, respectively.

According to Table 3, Table 4, Fig. 4, Fig. 5, it is clear that the predicted oil and water percentage using ANN model is close to the simulation results. Table 5 shows the obtained errors for the proposed ANN model, where

Conclusion

In this study, new simple system with one detector and a dual energy gamma-ray source was presented. Contrary to previous works, only two features from detector output signal was chosen as ANN inputs. Multi-layer perceptron neural network is used for developing the ANN model. The obtained results show that the proposed ANN model has achieved good agreement with the simulation data with minimum error. With above discussion, the proposed model is a sufficient tool to predict the volume fraction

Acknowledgment

The authors would like to thank Majid Khorsandi and Farzin Shama for their valuable comments that help to improve the manuscript.

References (22)

  • J. Wang et al.

    On the use of prompt gamma-ray neutron activation analysis for determining phase amounts in multiphase flow

    Meas. Sci. Technol.

    (2008)
  • Cited by (122)

    • Deep learning models for improved accuracy of a multiphase flowmeter

      2023, Measurement: Journal of the International Measurement Confederation
    View all citing articles on Scopus
    View full text