This study presents a novel method for accurately measuring the void fraction in annular two-phase air-water flow using a concave 8-blade capacitive sensor, validated through both simulations and experiments. The capacitance values obtained from different electrode configurations were used to generate sinograms. Firstly, tomographic images of the flow were constructed from these matrixes using the back-projection algorithm. Then, the energy of the sinograms was used as the primary input for an Artificial Neural Network (ANN). An optimized Multilayer Perceptron (MLP) network was designed to predict void fractions with high accuracy. Replacing multiple inputs with the energy characteristic greatly enhanced computational efficiency. The method achieved a Mean Absolute Error (MAE) of 0.003 (training) and 0.002 (testing), with R2 scores of 0.9992 and 0.9997, and Root Mean Square Error (RMSE) of 0.005 (training) and 0.008 (testing), confirming the model’s robustness. These results highlight the enhanced sensitivity and precision of the proposed method in void fraction measurement. Moreover, tomographic reconstructions of flow patterns provided valuable insights into the material distribution within the system, contributing to improved process optimization and safety in high-speed fluid environments.
