Predictive modelling of drag, lift and torque coefficients of cylindrical biomass particles under artificial intelligence

The escalating threats of climate change and fossil fuel depletion have greatly spurred the widespread utilization of renewable energy sources. Biomass energy stands out as the most abundant renewable energy resource. Among all biomass conversion technologies, direct combustion of biomass for heating and power generation represents the most common and cost-effective approach. Suspended combustion technology is widely employed in existing biomass combustion power plants worldwide. However, conventional modeling procedures often fail to adequately address the differences between coal and biomass combustion, particularly the significant impact of large-sized and highly non-spherical cylindrical straw biomass particles on their trajectories. This study utilizes OpenFOAM-body-fitted mesh direct numerical simulation (DNS) to generate a Computational Fluid Dynamics (CFD) dataset consisting of 300 data points for cylindrical particles, aimed at constructing a prediction model based on machine learning (N+1) for key aerodynamic parameters (drag coefficient, lift coefficient, and torque coefficient). The proposed model enables precise prediction of aerodynamic parameters for a wider range of different cylindrical particles and flow conditions, thus extending the existing Euler-Lagrangian model for non-spherical biomass multiphase flow.