Multiphysics modeling of thermochemical process in cement rotary kiln via computational fluid dynamics-discrete phase model and physics-informed neural network

Cement production is a cornerstone of modern infrastructure, but it is highly energy-intensive and carbon-intensive, with the clinker formation in cement kilns accounting for the majority of energy consumption and CO2 emissions. To investigate the complex multiphysics process of clinker formation, this work employs computational fluid dynamics-discrete phase model (CFD-DPM) and physics-informed neural networks (PINN). The CFD-DPM simulations not only accurately capture the velocity and temperature contours inside the kiln but also reproduce particle motion and the stepwise chemical reactions involved in clinker formation in different zones. In addition, the PINN algorithm is applied to solve the governing ordinary differential equations (ODEs) for CaCO3 decomposition, using the temperature of particles as input constraints. This approach accurately predicts the limestone mass fraction profile during calcination. According to the results from the parametric study, the location of reaction zones is determined by the temperature profile. The raw meal composition governs clinker formation by affecting the DPM mass concentration of individual phases. Air velocity controls the onset of CaCO3 decomposition and affects the DPM concentration of clinker phases. The integration of CFD-DPM and PINN provides a new paradigm for investigating and modeling complex two-phase systems with high accuracy and efficiency. The findings also provide important insights for reducing energy consumption and CO2 emissions in the cement industry and support operation optimization toward sustainable high-temperature manufacturing.