Machine Learning in CFD

Fluid dynamics research has continuously benefitted from advances in applied mathematics, computer science, and computer engineering; the development of CFD being a prime example of this multidisciplinary contribution. Improved numerical analysis techniques led to stable and efficient algorithms, while vectorization and parallelization fueled the expansion of CFD from research codes to validated design software. As computing power has grown, so too has the generation of data and the complexity of flows simulated. Gaining insight into fundamental physics is now growing, but with that comes enormous amounts of data that represent opportunities for insight and exploitation. Machine learning (ML) is the next iteration in the aforementioned multidisciplinary collaboration. This field encompasses many techniques and approaches, including the classification of flow fields, prediction of nonlinear processes like aerodynamic stall, modeling of fine-scale structures, and unsupervised exploration of complex, voluminous data for hidden physics. This field should be rightly thought of as a new toolbox that will allow us to further understand and utilize fluid dynamics through increasingly complex simulations. In short, it is ready for application to distinctly nonlinear, unsteady, multiscale problems in fluid dynamics. In this special issue, readers will find applications of machine learning to modeling complex phenomena, classifying flows, and increasing the fidelity of existing methods. The readers will also find that much of the new jargon used in the ML community is remarkably similar to ideas used for decades in the CFD community. The guest editors have chosen papers that are representative of leading-edge CFD applications. In each case, ML is a means to obtain more than could normally be accomplished with traditional methods. We hope that this special issue will encourage readers to begin new studies in this growing field. Finally, we would like to thank Prof. Habashi, Editor-in-Chief of the IJCFD for providing us with supportive guidelines and comments throughout the process of this special issue. We would also like to express our gratitude to the editorial staff at Taylor & Francis and the reviewers for their valuable effort. Respectfully,