Electrical Impedance Tomography of Industrial Two-Phase Flow Based on Radial Basis Function Neural Network Optimized...

Area of Science:

• Engineering
• Computational Science
• Physics

Background:

• Industrial two-phase flow detection using Electrical Impedance Tomography (EIT) often relies on the Gauss-Newton algorithm.
• The Gauss-Newton algorithm exhibits limitations in imaging accuracy, particularly in complex scenarios involving multiple bubbles.
• Accurate imaging is crucial for process monitoring and control in industrial applications.

Purpose of the Study:

• To introduce and evaluate the Artificial Bee Colony-optimized Radial Basis Function Neural Network (ABC-RBFNN) for enhanced EIT image reconstruction.
• To improve the accuracy and robustness of EIT imaging for industrial two-phase flows, especially with complex bubble distributions.
• To assess the performance of ABC-RBFNN against existing methods like Gauss-Newton and standard RBFNN.

Main Methods:

• The ABC-RBFNN algorithm was applied to EIT for industrial two-phase flow imaging.
• Simulated electrode data from a 16-electrode EIT system using EIDORS-v3.10 software served as training data.
• Performance was evaluated using Image Correlation Coefficient (ICC) and Root Mean Square Error (RMSE) on noisy and noiseless datasets.
• Algorithm generalization was tested using various bubble models (size, quantity, shape).

Main Results:

• The ABC-RBFNN algorithm demonstrated superior performance over Gauss-Newton and RBFNN, achieving higher ICC and lower RMSE.
• The algorithm exhibited significant noise immunity, maintaining accuracy with noisy test data.
• ABC-RBFNN accurately determined bubble size and shape, confirming its generalization capability across different models.
• Validation with experimental data from a 16-electrode EIT device confirmed accurate identification of target size and position.

Conclusions:

• The ABC-RBFNN algorithm offers a significant advancement in EIT image reconstruction for industrial two-phase flows.
• Its enhanced accuracy, noise immunity, and generalization ability make it suitable for complex flow imaging.
• The findings provide a strong foundation for the practical implementation of ABC-RBFNN in industrial EIT applications.