Numerical investigation and acoustic detection of flow boiling regimes in a helical tube

Boiling acoustics (BA) offers a noninvasive approach for two-phase flow regime identification, yet its correlation with specific boiling regimes under flow conditions remains poorly characterized. This study simulates flow boiling and corresponding acoustic emissions in a helical tube to establish the relationship between flow regimes and their acoustic signatures. Machine learning (ML) models, particularly a convolutional neural network (CNN), are trained to automate regime classification. The results reveal distinct acoustic features: plug flow exhibits high-frequency dominance, wavy flow is characterized by low-frequency components, and slug flow combines both. BA-based classification is numerically validated and further enhanced by the CNN model, which achieves 97.5% accuracy even under strong white noise (SD = 0.7). These findings demonstrate BA’s potential as a robust, real-time tool for monitoring and controlling flow boiling processes in industrial systems.