Implementing Artificial Neural Networks to Estimate Coefficients of Drag, Lift, and Torques in Spherocylinder Particles




Molina, Sergio A.

Journal Title

Journal ISSN

Volume Title



The fluid mechanic modeling of Spherocylinder particles is a research topic of great importance, specifically, within the modelling of dispersion of pollutants field, coal combustion, and separation processes. To add, accurately estimating the coefficients of drag, lift, and torque of Spherocylinder particles enable a richer description of the fluid mechanical behavior that otherwise is underrepresented. An abundance of research, in literature, exhibits a mixture of experimental and numerical simulation techniques for estimating the hydrodynamic loads on non-spherical particles, each with varying levels of accuracy. A correlation limitation for the lift and torque coefficients estimates can be explored via artificial neural networks. A dataset consisting of over 1200 data points for the coefficients of drag, lift, and torque was generated via Direct Numerical Simulations (DNS) for Spherocylinders with aspect ratios ranging from 1 through 6, incident angles 0° through 90°, and Reynolds numbers from 0.1 through 300. To enable time efficient fluid modeling, with minimal personnel expertise on software, an accurate Muli-Layered Neural Network (ANN) model was proposed. The ANN achieved a percent relative error of less than or equal to 15% for the coefficient of drag, lift, and torque estimates when tested with a validation dataset of 45 random, unobserved datapoints. Lastly, the model estimates for the drag, lift, and torque relative to the total dataset, including trained and validation datasets) achieved correlation coefficients greater than 99.99%.


This item is available only to currently enrolled UTSA students, faculty or staff. To download, navigate to Log In in the top right-hand corner of this screen, then select Log in with my UTSA ID.


Artificial Intelligence, Coefficients of Drag lift and torque, Fluid Mechanics, Neural Networks, non-linear regression, Torque, Spherocylinder particles



Mechanical Engineering