Prediction of Taylor flow in microchannels based on generative artificial intelligence

Flow pattern prediction in multiphase systems is essential for characterizing hydrodynamic properties and optimizing mass/heat transfer efficiency. Herein, we propose a generative artificial intelligence (GenAI) flow pattern prediction framework for rapidly processing and analyzing large–scale flow pattern image data, with the first application to predictive modeling of gas–liquid Taylor flow in microchannels. The forecasting results of this GenAI–based prediction framework do not consist of discrete flow pattern classification labels but rather intuitive, spatially resolved high-fidelity visualization results comparable to experimental observations under steady–state operating conditions (e.g., high–resolution flow pattern images captured by high–speed cameras). Notably, the proposed prediction framework overcomes the limitations of conventional methods that only provide category information of flow patterns. More importantly, the model evaluation results demonstrate that this framework can effectively model the correlation between operating conditions and corresponding flow characteristics within microchannels, thereby validating the great potential of this GenAI technology for multiphase flow research.