Data-Driven Enhancements to Transition and Turbulence Modeling Under Varying Pressure Gradients and Unsteadiness Effects

TitleData-Driven Enhancements to Transition and Turbulence Modeling Under Varying Pressure Gradients and Unsteadiness Effects
Publication TypeConference Proceedings
Year of PublicationSubmitted
AuthorsFang Y, Rosenzweig M, Reissmann M, Pacciani R, Marconcini M, Bertini F, Sandberg RD
Conference Name15th International ERCOFTAC Symposium on Engineering Turbulence Modelling and Measurements (ETMM15)
Conference Location22nd - 24th September 2025; Dubrovnik, Croatia
Abstract

Previous studies have shown that data-driven physical modeling can enhance numerical predictions for steady flows with pressure gradients (Fang et al., 2024). However, research on data-driven physical modeling for inherently unsteady flows remains limited. Investigating this area is crucial for two key reasons. First, improvements in mean quantities do not necessarily indicate an accurate representation of the underlying physics. When steady-state computations are applied to unsteady flows, inaccurate predictions may arise, with enhancements potentially stemming from data fitting rather than a genuinely refined physical model. Second, an improved mean-field prediction does not ensure the accurate representation of instantaneous flow dynamics. Our study proposes two strategies to extend the existing training framework for steady flows, facilitating improved physical modeling of unsteady flows with strong pressure gradients. First, unsteadiness information is incorporated into the cost function by selecting and computing phase-averaged data that exhibit significant discrepancies between LES and URANS calculations. This integration provides phase-averaged information as feedback to guide model training. Second, we refine the recently reconstructed laminar kinetic energy (LKE) transition model by Pacciani et al. (2025), specifically adapting it for wake-induced transition.

Fang, Y., Reissmann, M., Pacciani, R., Zhao, Y., Ooi, A. S., Marconcini, M., Akolekar, H. D., and Sandberg, R. D. (2024). Exploiting a transformer architecture for simultaneous development of transition and turbulence models for turbine flow predictions. In Turbo Expo: Power for Land, Sea, and Air (Vol. 88070, p. V12CT32A023). American Society of Mechanical Engineers.
Pacciani, R., Fang, Y., Metti, L., Marconcini, M., and Sandberg, R. (2025). A Reformulation of the Laminar Kinetic Energy Model to Enable Multi-mode Transition Predictions. Flow, Turbulence and Combustion, 114(1), 81-116.

Refereed DesignationRefereed