SBIR-STTR Award

Cascade Technologies and its collaborators from the Florida State University and Stanford University propose to develop a robust computational framework to enable geometrical optimization of complex non-axisymmetric exhaust systems with controlled accuracy and low computational cost. In this framework, Reynolds-averaged Navier-Stokes (RANS) surrogates will be constructed based on extensive databases from high-fidelity large eddy simulations (LES) of the exhaust system and jet plume. The RANS surrogates will then be used to efficiently explore the design space considered. Ultimately, the final optimized configuration will be once again analyzed using the LES tools to verify the performance gains predicted by the surrogate. In addition, a companion experimental investigation will provide the basis for a detailed validation and further confidence in the computational predictions. A coordinated Phase I plan is presented, which includes experimental measurements, LES predictions and RANS optimization for a rectangular nozzle. In phase I Option, the procedure will be extended to a SERN-inspired nozzle where a ramp downstream of the rectangular nozzle exit is explicitly included in the computations and experiments. Research directions and plans are outlined for Phase II work, with focus on fluid-structure coupling, introduction of robustness principles within the optimization cycle and innovative non-intrusive techniques for exhaust internal flow measurements.

Benefit:
Provide general framework for cost-effective LES/RANS optimization Improve design of exhaust ducts and nozzles

Keywords:
non-axisymmetric exhaust system, non-axisymmetric exhaust system, Large Eddy Simulation, RANS design optimization, rectangular nozzle, jet aeroacoustic measurements

The objectives of the proposed work are twofold. The first goal is to develop and validate GPU-based static and moving versions of Cascade's large eddy simulation (LES) software CharLES that would fully leverage existing (and future) GPU-accelerated systems accessible by NAVAIR and other DoD agencies. These software developments will be performed by Cascade. For the current project, the targeted capabilities are high-speed flows with shocks and wall modeling on both static and moving solvers. The application of interest is open rotors and realistic rotorcraft configurations. These choices were made to address key applications of interest to NAVAIR. The second goal is to develop modeling capabilities in the GPU-based solvers targeted towards efficient predictions of aerodynamics in realistic open rotors. In particular, the focus will be on actuator disk modeling to speed up simulation (when possible) and wall model closure with rotational effects to better capture the flow physics near rotor blades. The models development and validation for canonical flows and open rotor configurations will be done in collaboration with the Massachusetts Institute of Technology (MIT), the proposed Academic Subcontractors for this STTR project. In addition of the software and model developments, large eddy simulations will be conducted by Cascade and MIT for a range of cases to validate the solvers and models implementation, and assess the scalability and computational performance of the GPU-accelerated compressible flow solvers.

Benefit:
The main anticipated benefit for NAVAIR is to address the pressing need to accelerate simulation throughput and cut down on time-to-solution while keeping the accuracy and predictive capabilities of the computations. The proposed developments of scalable GPU-accelerated CFD solvers will enable NAVAIR to use existing untapped/under-utilized GPU resources to expand their computational allocations, improve simulation turn-around time and demonstrate the importance of investments in GPU-accelerated systems to the DoD HCPMP. The chosen applications of high-speed flows and rotorcrafts are highly relevant to NAVAIR, in particular in support of the V22 program. In terms of potential commercial applications, Cascade Technologies has entered in a number of joint development and software licensing agreements with major industrial OEMs (including General Electric, Solar Turbines, Boeing, United Technologies, and Honda) over the past decade. Most of these licensing and collaborations with our current clients precipitated from the need to more accurately predict flow phenomena in multi-physics ow environments (turbulent reacting flows, transonic aerodynamics, radiated noise, flow encountering boundary layer separation). In conjunction with advances in the simulation technologies (numerical methods, physical models), recent simulations have been demonstrating engineering utility of these calculations with Cascade's Software. However, the computational cost on CPU-based systems (in either a dollar sense or wall-clock sense) has now become the single, largest barrier for scaling usage inside these companies or expanding the use of the Cascade's software to other organizations that do not presently have a HPC infrastructure. Improving the performance of our Software product is therefore critical and even more so for the proposed application of open rotors, a class of problems involving moving boundaries that are particularly computationally expensive. Cascade's industrial clients have orally communicated that a cost reduction of a factor of 3 would be sufficient for commercial viability. By reducing the unit cost of computing by over an order of magnitude, the GPU-accelerated Software solution proposed herein could significantly accelerate the adoption of Cascade's software within our existing clients and expand licensing opportunities with other industrial clients across a range of industries, including aviation, aerospace, defense, power, and automotive.The validation simulations from this project will also provide persuasive evidence of improved performance and technology readiness in these areas.

Keywords:
wall modeling, rotorcraft predictions, GPU-accelerated systems, large-eddy simulation, CFD, moving-mesh compressible flow solver