Nonlinear phenomena of gas–liquid microflow pressure drop and its machine learning model

Abstract The flow pressure drop of the gas–liquid system in the microscale is important for the design and optimization of microreaction processes; however, the understanding of this parameter is insufficient or even biased because the achievable gas–liquid microflow patterns are limited. Accordingly, this work designs different microdevices to generate all the surface tension‐dominated flow patterns (Taylor bubble, bubbly, zigzag bubble, bubble swarm, and bubble crystal), and the universal laws of the gas–liquid microflow pressure drop are revealed. The results show that the ultra‐wide bubble diameter range of 40–600 μm is realized. Especially, under the same working conditions, the change rule of microflow pressure drop will be entirely different based on the bubble size and flow pattern, and nonlinear microflow pressure drop curves (N‐shaped, increased, inverted V‐shaped, V‐shaped, and decreased) are observed. Finally, a data and mechanism co‐driven machine learning model is developed for the nonlinear gas–liquid microflow pressure drop.