A Data-Driven Approach for Generalizing the Laminar Kinetic Energy Model for Separation and Bypass Transition in Low- and High-Pressure Turbines

The laminar kinetic energy (LKE) transition model has been proposed to predict the separation-induced transition, which is frequently experienced by low pressure turbines (LPT). In contrast to LPTs, high-pressure turbines (HPT) often are subject to bypass transition which currently is not captured well when using the LKE model. Because these two types are common transition modes for turbines, the current effort is focused on generalizing and improving the LKE model to be able to predict the different types of transition. In addition to modifying the transition model for better on-blade predictions, the turbulence model also needs improvements to more accurately capture the wake mixing. Hence, the main purpose of this study is to use a data-driven approach to simultaneously develop transition and turbulence closures suitable for a range of different turbine configurations. To achieve this, two strategies are adopted. The first is to employ a multi-case multi-objective CFD-driven model training framework. Previous research was limited to single-case multi-objective CFD-driven training, using only a T106A data set. Enhanced predictions were obtained both on the blade surface and wake mixing regions (Akolekar et al., GT2022-81091) for the training case, but the models' performance was not guaranteed for other cases. In the present multi-case training framework, the coupled training of the transition and turbulence models is performed on several LPT and HPT configurations to ensure better generalizability of the models. The second strategy employed is to derive an updated list of local non-dimensionalized variables, which serves as the inputs for the LKE model corrections. The training turbine cases include PakB, T106A, LS89, and T108. Among these turbines, LS89 is an HPT characterized by unsteady vortex shedding near the trailing edge. In order to capture this, for the first time an unsteady solver is utilized during the CFD-driven training, and the time-averaged results are used to calculate the cost function as part of the model development process. The trained models are then tested on turbine cases different from the training ones, and their performance are assessed in terms of pressure coefficient, wall shear stress and wake losses.