A Water Fraction Measurement Method Using Heuristic-Algorithm-Based Electrical Capacitance Tomography Images Post-Processing Technology

Bo Liu, Chenhui Tang, Kaihao Tang, Hongli HuIEEE Access 2020 · DOI

Abstract

This paper presents a sectional water fraction measurement method using electrical capacitance tomography (ECT) image. To achieve a desired measurement accuracy, an optimal threshold-search method based on fast simulated annealing (FSA) algorithm is applied for image binarization; and to guarantee the imaging time consumption, fast fixed point continuation (FFPC) iterative algorithm cooperating with the compressed sensing (CS) theory was applied to the image reconstruction. In this study, the imaging time consumption and the measurement accuracy of water fraction are used to compare the proposed method and other methods. A numerical model is established with COMSOL Multiphysics to conduct the simulative validation of the proposed method firstly and physical experiments are conducted then. The experiment results show the average relative error of water fraction is lower than 15%, where the even as low as 3.68%.

Introduction

In industrial processes such as natural gas extraction and chemical engineering, the precise measurement of the water fraction in gas-water flows is vital for operational efficiency and safety. Electrical Capacitance Tomography (ECT) has emerged as a prominent non-intrusive technique for visualizing internal pipe flows by measuring the capacitance changes between peripheral electrodes. However, ECT is inherently limited by the "inverse problem"—the number of measurement points is significantly smaller than the number of pixels in the reconstructed image, leading to ill-posedness and blurred results.

Figure 1: Schematic diagram of the ECT measurement system including the sensor, measurement circuit, and data processing unit.

Proposed Methodology

To overcome the limitations of traditional ECT, this study proposes a two-stage pipeline focusing on both reconstruction speed and post-processing accuracy.

Image Reconstruction via CS-FFPC

The first stage utilizes Compressed Sensing (CS) theory combined with a Fast Fixed Point Continuation (FFPC) iterative algorithm. By exploiting the spatial sparsity of fluid distributions (e.g., distinct gas and water phases), CS-FFPC achieves higher image quality and faster convergence than standard iterative methods like the Landweber algorithm.

Post-Processing via FSA Binarization

The core innovation lies in the image binarization step. Standard fixed-threshold methods often fail to accurately represent the physical boundaries of the water phase in blurry ECT images. The authors implement a Fast Simulated Annealing (FSA) heuristic to search for an optimal threshold. This search minimizes the "capacitance residual"—the difference between the actual measured capacitance and the theoretical capacitance derived from the binarized image. This physics-aware approach ensures that the calculated water fraction is consistent with the raw sensor data.

Figure 2: Cross-sectional view of the 12-electrode ECT sensor showing the imaging area and pipe wall.

Experimental Validation

The method was validated through both numerical simulations and physical experiments. A 12-electrode sensor model was developed in COMSOL Multiphysics to simulate various flow regimes, including bubble and annular flows. Subsequently, a physical test rig using acrylic tubes and water was employed to verify the system's performance under static conditions.

Results and Discussion

The proposed CS-FFPC and FSA-based method demonstrated significant improvements over traditional techniques:

  • Accuracy: Physical experiments yielded an average relative error in water fraction measurement of less than 15%, with a minimum error of 3.68%.
  • Image Quality: The Image Correlation Coefficient (ICC) was consistently higher than those produced by Linear Back Projection (LBP) and Landweber Iterative Algorithms (LIA).
  • Efficiency: The reconstruction process took approximately 44.37 ms, with the total processing time (including binarization) reaching roughly 110 ms. This performance is nearly four times faster than the Landweber method, making it suitable for near-real-time industrial monitoring.

Conclusion

This research successfully integrates heuristic optimization with compressed sensing to enhance ECT-based water fraction measurement. By tying the binarization threshold to physical capacitance residuals, the method achieves high accuracy without sacrificing the speed required for industrial applications. Future work will focus on optimizing software overhead and extending the method to dynamic, high-speed turbulent flows.

Summary updated

Abstract

This paper presents a sectional water fraction measurement method using electrical capacitance tomography (ECT) image. To achieve a desired measurement accuracy, an optimal threshold-search method based on fast simulated annealing (FSA) algorithm is applied for image binarization; and to guarantee the imaging time consumption, fast fixed point continuation (FFPC) iterative algorithm cooperating with the compressed sensing (CS) theory was applied to the image reconstruction. In this study, the imaging time consumption and the measurement accuracy of water fraction are used to compare the proposed method and other methods. A numerical model is established with COMSOL Multiphysics to conduct the simulative validation of the proposed method firstly and physical experiments are conducted then. The experiment results show the average relative error of water fraction is lower than 15%, where the even as low as 3.68%.
BibTeX
@inproceedings{liu2020water,
  author = {Bo Liu and Chenhui Tang and Kaihao Tang and Hongli Hu},
  title = {{A Water Fraction Measurement Method Using Heuristic-Algorithm-Based Electrical Capacitance Tomography Images Post-Processing Technology}},
  booktitle = {IEEE Access},
  year = {2020},
  doi = {10.1109/access.2020.3037721},
}
Figure