Deep learning-based dual-modality information fusion of EMT and ERT for accurate solid phase fraction measurement in gas-liquid-solid three-phase flow

Abstract Accurate measurement of phase fraction distribution in gas-liquid-solid three-phase flow is critical for optimizing industrial processes such as petrochemical and bioengineering reactors. However, single-modality electrical tomography techniques face significant challenges due to complex flow dynamics, limited sensitivity, and the inability to simultaneously distinguish multiple phases with high precision. To address these limitations, this paper proposes a novel two-stage deep learning-based information fusion network that integrates electromagnetic tomography and electrical resistance tomography (ERT) data. The first-stage model (front-model) extracts the background liquid-solid phase fraction and corresponding empty-field ERT measurements, while the second-stage model (post-model) combines dual-modality data to reconstruct the solid phase fraction distribution. This approach effectively separates solid phase information from raw measurements and enhances image fusion quality. Simulation and experimental results demonstrate that the proposed model achieves high accuracy in predicting solid phase fractions, with a mean absolute percentage error of 1.53% in simulations and 5.2% in experiments. The model also exhibits strong noise robustness, maintaining performance at signal-to-noise ratios as low as 30 dB. This work advances the field by providing a reliable and non-invasive solution for phase fraction measurement in complex three-phase flow, with potential applications in industrial monitoring and control.