Upgrade the Simulation Accuracy of Your Aero-Engine with Fidelity Flow Solver

The current level of computing power is unprecedented. It has led to the development of advanced computational techniques that enable us to simulate larger systems and predict complex phenomena with greater accuracy. However, the simulation of turbomachinery systems still poses a challenge as the current practice is to simulate each component separately, thereby failing to account for the interactions between them. A new methodology has been developed based on the Reynolds-Averaged Navier-Stokes equations in Fidelity Flow solver to address this issue and enhance the aero-engine design in terms of efficiency, reliability, and low emission. This approach allows a fully coupled simulation of an entire engine using a single code. This blog post is about the new methodology, the flow solver technology, and the results achieved through its implementation.

Layout of the KJ66 MGT: impeller (1), diffuser (2), combustion chamber (3), HPT nozzle (4), HPT rotor (5), LPT nozzle (6), LPT rotor (7), deswirl vane (8), exhaust hood (9).

Methodology

A novel approach has been developed for conducting steady, time-accurate, and fully coupled simulations of complete aero-engine and gas turbine systems. This method employs the Nonlinear Harmonic (NLH) technique to capture unsteady effects, leading to computationally affordable times. The combustion processes in this approach rely on the Flamelet Generated Manifold (FGM), which is both efficient and reliable. Compared to an inert simulation, the computational overhead incurred in modeling the combustion process is only about 50%. Additionally, the Smart Interface approach is implemented to avoid the transportation of scalars throughout the entire system, ensuring the computational overhead is minimized by solving for combustion transport variables only where the flow is reactive.

Flow Solver

This study is carried out using the Cadence Fidelity Flow solver, which comprises both pressure-based and density-based solution schemes. The flow solver package is equipped with a wide range of physical models, including modules for turbomachinery modeling, Large Eddy Simulation (LES), Conjugate Heat Transfer (CHT), Fluid-Structure-Interaction (FSI), a Lagrangian module for sprays, as well as models for cavitation, radiation, multiphase flows, and combustion. Standard approaches such as the mixing plane, frozen rotor, and sliding mesh are implemented in the turbomachinery module. Furthermore, the Nonlinear Harmonic method is available for highly efficient calculation of unsteady flow in turbomachinery applications, which is discussed in the next section.

Nonlinear Harmonic Method (NLH)

The NLH method is a nonlinear approach that considers the impact of unsteadiness on the time-mean flow. These effects are deterministic stresses, manifesting as time-averaged products of periodic fluctuations. The beauty of the NLH method lies in its computational efficiency. It only necessitates determining a steady-state solution for the average flow field and steady-state solutions for the real and imaginary parts of each of the harmonics that the user chooses to resolve. The solution accuracy depends on the number of harmonics, but only a few harmonics are typically necessary to capture the unsteady effects.

Comparison between rotor-stator interaction using mixing plane and NLH method.

Fidelity Flow's NLH module accounts for rank-2 effects, which allow the modeling of interactions between adjacent rows and adjacent to adjacent rows with or without relative rotation speed. In other words, the NLH module can account for more complex, unsteady interactions. By using a rank-2 solution, the effect of clocking can be studied in post-processing mode. Additionally, the NLH module in Fidelity Flow offers the flexibility to define how many harmonics to use in each blade row, making the simulation process more customizable and efficient.

Smart Interfaces

Fidelity Flow offers a Smart Interface methodology that allows connecting different domains with varying physics and computational models. Smart Interfaces provide users with the flexibility to transfer information between blocks with different computational models while using physical models that are relevant only to their respective domains. With Smart Interfaces, users can define which state variables are exchanged between blocks and how they are coupled. Furthermore, physical models in one domain can be customized to account for states in another block as per the user's requirements. Smart Interfaces are defined using the OpenLabs module and loaded into the flow solver using dynamic libraries.

Sketch of Smart Interface coupling in unsteady full-engine simulation using NLH method.

Result and Analysis

The proposed method took 52 hours of wall-clock time using 288 processors on a cluster of Intel E5-2697 CPUs (2.6 GHz) to achieve convergence to a steady-state solution. The compressor and turbine operated at their design points, and no harmful phenomena like complete flow reversal due to surging, choking, or flash-back of the flame were encountered during the simulation. Additionally, the NLH method used in the simulation performed with rank-2 and 3 harmonics per perturbation reached a comparable level of convergence in 146 hours (6 days) in about 10k iterations on the finest multigrid level.

Reconstruction in time of the static temperature perturbation from the combustor at the HPT 90% span for the rank-2 NLH computation with 3 harmonics per perturbation.

Compared to the steady simulation with the classical mixing plane approach, the NLH method provides better insight into the inter-component interactions, allowing for an in-depth analysis of the effects of individual disturbances on each machine element. Additionally, the NLH method was used to study the effect of hot streaks on the first rows of the turbine. The impact of the potential field of the combustion chamber on the upstream components was also evaluated. According to the numerical results, the pressure perturbations in the diffuser row of the compressor due to the influence of the combustor are in the order of 20 Pascal, while pressure perturbations in the impeller due to the presence of the diffuser are higher by two orders of magnitude.

Reconstruction in time of the static pressure perturbation at the compressor/diffuser midspan for the rank-2 NLH computation with 3 harmonics per perturbation.

Conclusion

This study reveals that the NLH technique in Fidelity Flow solver is capable of handling unsteady rotor-stator interaction while being computationally efficient. Moreover, the Smart Interface methodology and Manifolds method enable linking blocks with different physical models while limiting combustion modeling in the combustion chamber. The full-engine simulation approach produces reliable results without significant discontinuities at the interfaces between the various blocks. Upon examining the interaction between components, it was discovered that temperature field inhomogeneities at the combustor outlet are conveyed downstream to the rotor of the first turbine stage, resulting in significant fluctuations in the thermal field. This study provides valuable insights into the full-engine simulation approach and its possible applications in the field of turbomachinery.

Read the whitepaper Methodology for Steady and Unsteady Full-Engine Simulations to learn more about this new turbomachinery simulation approach for complete aero-engine.