Aeroacoustic Simulation Delivers Breakthroughs in Aircraft Noise Reduction

Aircraft manufacturers face increasingly stringent standards for reducing community noise. Conventional aircraft development methods based on engineering experience, past designs and flight testing will not suffice to meet future noise reduction targets. Computational Fluid Dynamics (CFD) software based on so-called Reynolds-averaged Navier-Stokes (RANS) methods has revolutionized aerodynamics engineering, but is insufficient for high-fidelity aeroacoustic simulation. However, the Lattice-Boltzmann-based technology of Exa Corporation’s PowerFLOW software provides aeroacoustic simulation accuracy comparable to wind tunnels and flight testing.

Community Noise

Flow structure visualization around the wing of an Embraer regional jet in highlift configuration

Aircraft noise, already a problem for many communities located near major airports, will only get worse with the continuing growth in air travel. Thus, on top of air carriers’ demands for more fuel-efficient, cheaper-to-operate airplanes, commercial aircraft manufacturers face increasingly stringent requirements to reduce the “community noise” produced by their aircraft. Already, communities, as well as individual airports, impose noise regulations that may affect which aircraft are allowed to land at given times, and charge airlines a fee if any of their aircraft are too noisy. All this has made noise reduction a critical competitive factor for aircraft manufacturers today.

Community noise has two main sources: engine noise and airframe noise. Airframe noise is caused principally by airflow around the aircraft’s landing gear, and around high-lift devices such as wing flaps and slats. The main sources of engine noise are the engine fan and the jet downstream of the engine. Figure 1 visualizes these noise sources based on predictions from Exa’s PowerFLOW code, using some of the unsteady flow visualization techniques available in PowerFLOW. In addition, a factor known as “installation noise” — the noise created by interactions of the jet from the engine as well as the main landing gear with components of the high-lift wing (e.g. the flaps), — can be a significant contributor to overall noise.

Over the past 20 years, aircraft engine manufacturers have made significant progress in reducing engine noise through advances such as high-bypass turbofans and innovative nozzle geometries. This has promoted airframe noise to the most urgent challenge now — especially during landing approach, when noise produced by the airframe is as loud as, or even louder than, propulsion noise.

At the same time, mitigating engine noise remains important in reducing takeoff noise. According to a study by Swen Noelting, Vice President, Aerospace, and Ehab Fares, Senior Technical Director of Aerospace Applications, both with CFD software developer Exa Corporation, “Despite huge reductions over decades, engine noise is still the biggest contributor to community noise during takeoff. The reductions have mostly been achieved by increasing the bypass ratio of jet engines, which cannot realistically be further increased due to geometrical constraints. Therefore, significant research effort has been put into better understanding and directly reducing fan and jet noise, the main contributors to engine noise.” [1]

Adding to the challenge of reaching aggressive noise reduction goals is the fact that these are often in conflict with other performance objectives. For example, the goal of increased fuel efficiency can be achieved by making the engines larger, but the resulting increase in fan radius will make the engine noisier. Similarly, reducing noise from highlift devices may impact the lift performance of an aircraft.

Traditionally, aircraft engineering organizations have relied on experience from past aircraft designs, and empirical methods to meet noise targets. There is little, if any, use of wind tunnels to reduce aircraft noise. Most often the procedure has been simply to engineer the aircraft with a high safety margin with respect to noise requirements, and then hope it will meet certification requirements in flight testing.

But increasingly stringent noise regulations are making these traditional processes inadequate — engineers will no longer be able to design in enough safety margin to be confident their new aircraft will meet certification targets. Wind tunnel testing is seldom (if ever) a viable solution for noise optimization in commercial aircraft development programs, due to the fact that a complete aircraft with all relevant noise sources cannot be tested in a wind tunnel at a scale that would adequately capture the noise sources. Testing of individual components — such as landing gear and high-lift devices — is possible in wind tunnels but also of limited use due to time and cost constraints and the inability to capture installation effects.

Digital Simulation

Figure 1. Visualization of the main noise sources of a commercial aircraft.

The alternative to wind tunnels — using digital simulation to test realistic concepts for airframe noise reduction — has been a challenging task. To accurately predict and mitigate airframe noise, for example, the landing gear’s complex geometry has to be represented at high fidelity in the simulation model, since even small geometry details can be significant noise contributors. In addition, the entire aircraft has to be simulated, in order to identify the complex interactions among landing gear, wings and other major aircraft components. This has proved to be an extremely challenging task for traditional RANS-based CFD tools.

Noelting and Fares note that significant progress has been made over the past several decades in simulating aerodynamic flows using CFD, and that this has fundamentally changed the design process in the aerospace industry. But the use of CFD has been focused mostly on modeling the aircraft in cruise configuration—not with flaps extended and landing gears deployed.

Major technological limitations remain to expand the use of CFD to more complex configurations. “The state of the art in CFD is based on Reynolds-averaged Navier-Stokes (RANS) methods which have proven accuracy and reliability for flow conditions at or near the design point in cruise configuration,” Noelting and Fares point out. “The usability of these methods is generally limited today, however, to relatively well-behaved flows without significant flow separation or unsteadiness. While it is expected that Large Eddy Simulations (LES) and hybrid RANS-LES methods will eventually be able to address these types of inherently unsteady flows, the computational cost associated with running Large Eddy Simulations for an industrial application is expected to be prohibitively large even in the decades to come.”

Figure 2. Flow structures around the nose landing gear of a business jet, highlighting the main noise sources.

Another challenge traditional RANS-based CFD methods are facing is the handling of complex geometries and grid generation. Creating the computational grids for configurations that are typically associated with separated flows, such as fully detailed high-lift wings—possibly including fairings and brackets, nacelles and landing gears— remains one of the greatest bottlenecks for current CFD tools. In addition, strong dependence of results on mesh quality, non-convergence of simulations and the requirement for deep CFD expertise are all critical factors slowing the expansion of the application of CFD in the aerospace development process.

One of the applications that have shown virtually no productive use of CFD in the development process is aeroacoustics and community noise. This includes airframe noise from landing gears and high-lift components, and jet and installation noise—all of which are generally addressed today only in flight testing. The limitations of traditional methods have made productive CFD use in this area impossible until now. This is now changing with the availability of Lattice-Boltzmann based codes such as Exa’s PowerFLOW.

Exa’s PowerFLOW software is gaining increasing recognition as the first CFD solution capable of effectively simulating the full complexities of the design factors that determine aircraft noise, and providing guidance for noise mitigation. A methodology developed in partnership with NASA has demonstrated that Exa’s software technology can be used in a way that delivers accuracy comparable to wind tunnel testing and flight testing. Experiments show that with PowerFLOW, users are able to predict airframe noise propagated to ground observers within an error range of 1 to 2 decibels from physical experiments—results that have been extended to engine noise prediction as well.

Noelting and Fares explain that Exa’s technological differentiation is grounded in “non-traditional methods such as the Lattice-Boltzmann method (LBM), combined with the best features of hybrid turbulence modeling.” LBM not only has the potential to tackle the advanced flow physics required, but also offers new possibilities for fully automatic volume mesh generation and parallelization.

Figure 3. Pressure fluctuations on the deployed flaps of a business jet.

The Lattice-Boltzmann method, including its implementation in Exa’s PowerFLOW, is based on kinetic theory. A relatively new CFD technology, LBM has only been developed over the last 25 to 30 years. In contrast to methods based on the Navier-Stokes (N-S) equations, LBM is based on a simpler and more general physics formulation. Its motivation is to simulate a fluid at a microscopic level where the physics is simpler and more general than the macroscopic, continuum approach taken by the Navier-Stokes equations.

Further, LBM can generally be combined with various turbulence modeling approaches including a standard turbulence model, an LES subgrid scale model and also hybrid approaches. The clear advantage of LBM methods is the very high temporal resolution, inherently efficient unsteady simulation algorithm, and low dissipation and dispersion of the numerical scheme. These advantages result in about one order of magnitude greater (in some cases even much more) computational efficiency than CFD solutions based on classical Navier-Stokes solutions of comparable quality.

Another key advantage of the Lattice-Boltzmann method is the high efficiency of computations on modern compute clusters with hundreds or thousands of interconnected nodes and tens of computational cores per node, due to the predominantly local nature of computational operations which dramatically reduces the core-to-core communications requirements compared with traditional RANS methods.

In summary, the main differences between N-S-based methods and LBM are:

  • LBM is based on a simpler representation of the flow physics; thus, the algorithmic implementation is less complicated.
  • LBM is most efficiently implemented on Cartesian grids. Cartesian grids allow for very robust automated grid generation of complex geometries.
  • LBM is inherently unsteady. For steady-state solutions, it is about one order more expensive than steady-state RANS methods, but for unsteady solutions, it is about one order less expensive than comparable N-S.
  • LBM offers low numerical dissipation, which makes it well suited for simulations of wakes and detached flows, and for aeroacoustics.
  • LBM methods are easily parallelized and can run efficiently on modern HPC architectures with thousands of computational cores.

Aeroacoustic simulations, such as the prediction of noise generated by landing gears and high-lift devices during approach and take-off are an almost ideal application for LBM. Acoustic simulations by definition have to be unsteady, and the extremely low numerical dissipation of LBM enables the capture of both the hydrodynamic pressure fluctuations on the component surfaces responsible for the noise generation, and the resulting coherent acoustical pressure fluctuations emanating from the structure which are responsible for the far-field community noise. In addition, the geometries that must be simulated for airframe noise cases tend to be highly complex and cannot be simplified for simulation, since even small geometry details may contribute significantly to the far-field noise.

Likewise, LBM offers a number of beneficial properties for mitigating jet, fan and installation noise. First, the low dissipation characteristic of the method is a fundamental enabler to capture the wake of a jet or propeller with sufficient accuracy. Second, the simplicity with regard to formulation of boundary conditions in the LBM approach allows efficient implementation of techniques to simulate rotating or moving geometries. The LBM implementation in PowerFLOW offers a sliding-mesh technique that enables very accurate simulations of rotating geometries, ideal for simulations of propellers or fans.

Exa has developed the methodology to simulate airframe and engine noise in a multi-year co-operation with NASA. Initially, isolated components such as nose and main landing gears were simulated and then compared to wind tunnel tests to ensure that all relevant noise sources were correctly captured. Figure 2 shows the flow structures and acoustic waves around a business jet main landing gear that was evaluated in the partnership with NASA.[2]

The next step in the NASA-Exa partnership included the evaluation of noise generated high-lift devices such as flap edges, and the development of concepts to reduce these noise sources. Figure 3 shows pressure fluctuations on the side edge of the flap which are one of the main overall contributors to airframe noise. In addition, installation effects were identified by investigating the interactions of a deployed landing gear with the high-lift components of the wing.

Using this methodology, PowerFLOW can be used to analyze the landing gear’s contribution to overall aircraft noise. This allows an aircraft manufacturer to give its landing-gear supplier detailed engineering data on how to configure and size the landing gear to achieve noise targets.

A number of studies have also been performed with PowerFLOW to validate the simulation of jet noise and fan noise. For example, simulation and comparison to physical experiment for one of the set points for the well-known SMC000 test case showed good agreement of both mean flow and fluctuating components of the flow.

In addition to predicting the overall noise levels of an aircraft and its components, Exa has developed methodologies to identify the sources of this noise—for example, to identify exactly which parts of the landing gear are causing noise—that give users the insight needed to remedy the problem. Its software can then be used to test various noise reduction concepts on a landing gear design.

For the future, aircraft manufacturers will have to accomplish major changes in the architecture of their aircraft in order to meet long-term noise reduction goals. Simulation technologies with the capabilities of PowerFLOW will be essential for that—the changes needed will simply be outside the scope and reach of conventional aircraft engineering practices.

Conclusion

Aircraft noise, already a problem for many communities located near major airports, will only get worse with the continuing growth in air travel. Thus, on top of air carriers’ demands for more fuel-efficient, cheaper-to-operate airplanes, commercial aircraft manufacturers face increasingly stringent requirements to reduce the “community noise” produced by their aircraft.

Community noise has two main sources: engine noise and airframe noise. Significant progress in reducing engine noise over the past two decades has promoted airframe noise to the most urgent challenge now. Airframe noise is caused principally by airflow around the aircraft’s landing gear, and around high-lift devices such as wing flaps and slats. Wind tunnel testing is seldom, if ever, a viable solution in commercial aircraft development programs due to cost and time constraints. But the alternative—using digital simulation to test and refine concepts for airframe noise reduction—is a challenging task. To be effective, the entire aircraft has to be simulated in order to identify the interactions among landing gear, wings and other components that are a major cause of noise. In addition, the landing gear’s complex geometry has to be represented at high fidelity in the simulation model, since even small geometry details can be significant noise contributors.

At the same time, mitigating engine noise remains important in reducing takeoff noise. In addition, “installation noise”—the noise created by interactions of the engine with components of the high-lift wing components, in particular the flaps— can be a significant contributor to overall noise. Finally, noise reduction goals often conflict with other performance objectives, so that tradeoffs must be quantified and evaluated.

Traditionally, aircraft engineering has relied on past experience, past aircraft designs, and empirical methods to meet noise targets. But increasingly stringent requirements are making these processes inadequate—conventional methods will not suffice to meet future noise reduction targets.

The longstanding alternative—CFD software based on Reynolds-averaged Navier-Stokes (RANS) methods—has revolutionized aircraft aerodynamics engineering, but its use has been restricted mostly to modeling the comparatively simple geometry of the aircraft in cruise configuration. For simulations of the aircraft with flaps extended and landing gear out, CFD has been much less used because of inherent limitations that make it impractical and insufficient for high-fidelity aeroacoustic simulation.

Today, however, software is capable of effectively simulating the full complexities of the design factors that determine aircraft noise. Enabling this capability is technology grounded in non-traditional methods such the Lattice-Boltzmann method, combined with the best features of hybrid turbulence modeling. A methodology developed in partnership with NASA recently demonstrated that Exa’s software technology delivers accuracy comparable to wind tunnel testing and flight testing, for both airframe noise and engine noise.

This article was written by Bruce Jenkins, Principal Analyst, Ora Research (Raleigh, NC). For more information, Click Here .