A GA-RELM-based image reconstruction method for accurate multiphase flow monitoring in annular electromagnetic tomography

Electromagnetic tomography (EMT) is a non-invasive imaging technique with significant potential for multiphase flow monitoring in industrial systems. However, image reconstruction in annular geometries remains challenging due to strong nonlinearity, high-dimensional sensitivity matrices, and complex phase distributions. In this study, a Genetic Algorithm–optimised Regularized Extreme Learning Machine (GA-RELM) framework is proposed for accurate reconstruction in annular EMT. A dedicated annular EMT model and a hybrid MATLAB–COMSOL simulation framework are developed to generate datasets containing 12 representative flow patterns with randomised spatial distributions. The proposed method is validated using a static experimental platform with multiple phase configurations. Simulation results demonstrate that GA-RELM achieves higher correlation coefficients and lower relative errors than ELM, GA-ELM, and RELM across diverse phase distributions. Noise analysis shows that the proposed method maintains stable performance under moderate noise conditions (SNR ≥ 30 dB). Spatial error analysis confirms improved boundary localisation and edge preservation. Experimental results verify that GA-RELM achieves the best overall performance, with the highest average correlation coefficient (0.755 ± 0.207) and the lowest relative error (0.249 ± 0.221) among compared methods. These results demonstrate that GA-RELM is an effective and robust reconstruction approach for high-resolution multiphase flow monitoring using annular EMT.