Final Report Summary - IDEALVENT (Integrated Design of Optimal Ventilation Systemsfor Low Cabin and Ramp Noise)
Executive Summary:
Reaching low levels of cabin and ramp noise is crucial to ensure the satisfaction of aircraft passengers and a safe working environment for the crew and personnel servicing the grounded aircraft. From the passenger and personnel standpoint, good thermal and acoustic comfort is important to minimize both stress symptoms and tiredness. Required to provide satisfactory air quality and temperature, the Environmental Control Systems (ECS) currently used are key contributors to the acoustic nuisance within the cabin and around the grounded aircraft. Reducing the amount of noise produced by the ECS has therefore a direct impact on the passenger satisfaction and personnel health and safety.
Unlike aircraft exterior noise, which has received considerable attention in past and currently running research projects, the noise emitted by confined flows in ECS assemblies involves complex mechanisms that haven’t been sufficiently investigated to permit the noise reduction wished by passengers and regulators. Acoustic and hydrodynamic interactions between subcomponents have so far been largely neglected despite being crucial, and are the focus of IDEALVENT.
In this project, thorough experimental studies as well as advanced analytical and numerical developments have been conducted in order provide a deeper understanding of these interaction mechanisms. Aerodynamic and acoustic installation effects were investigated by means of novel experimental techniques, highlighting the crucial importance of discriminating the noise emitted by a given ECS component (e.g. the blower unit) from the noise reflected by the remaining duct system (including its terminations in particular). Robust multi-modal decomposition techniques based on microphone and loudspeaker arrays were developed to this end. As a matter of fact, these modal decomposition techniques, applied as well to the numerical databases, proved essential to permit a meaningful cross-validation between measurements and simulations.
Integrated passive flow and noise control strategies have also been explored in this project, on laboratory test benches in a first stage, then on a full-scale industrial equipment in a second stage. The knowledge gained in the experimental and numerical investigations of the installation effects has permitted devising optimized strategies having the best potential for reducing ECS noise.
Based on the above considerations, it can be stated with confidence that the IDEALVENT project has been very successful, delivering nearly all the planned scientific and technological results, and opening unexpected new paths for better testing and modelling of complex ducted systems. The numerous results of the project are expected to find many fields of application beyond the initially intended one, e.g. for aircraft propulsion, or heating, ventilation and cooling systems used in surface transports and buildings.
Project Context and Objectives:
While a considerable number of European research projects conducted in the past and current Framework Programmes have addressed the exterior noise issues posed by aircrafts during take-off and landing, insufficient effort has been applied to understand and reduce the aerodynamic noise generated by the aircraft internal cooling and ventilation systems (grouped below under the acronym ECS for Environmental Control System). It has been acknowledged that the physical mechanisms involved in confined flow systems can be much more intricate than for exterior noise problems, e.g. because the separation between flow and acoustic fields becomes more arbitrary and less amenable to classical modelling approaches. Also, the intrinsic complexity of confined turbulent flows raises very specific challenges related to near-wall modelling and the prescription of adequate boundary conditions. Moreover, many interactions take place between sub-components of a given ECS: the blower unit, temperature and pressure regulating devices or simply duct connections. For these reasons, it proves difficult to devise the most promising noise reduction strategies and achieve the objectives of the Work Programme.
However, ECS components are strong contributors to the noise perceived within the cockpit and cabin, and to the acoustic environment of the personnel servicing the aircraft when grounded. Clearly, some research is needed to understand in detail the noise generating mechanisms within interacting ECS components, propose better design and sound mitigation guidelines, and eventually fulfil the competitive and societal objectives of aircraft manufacturers and European Communities altogether. The present project will address these important issues, in order to lead to an effective reduction of the cabin and ramp noise, and improve substantially both passenger comfort and airport personnel health and occupational safety.
Parts of the ventilation and cooling systems are installed within the aircraft belly-fairing casing, in order to ensure aircraft pressurization, provide temperature controlled air in the cabin and in the cockpit, and maintain a temperature compatible with the equipment present in the belly-fairing. In these systems, ram air is taken through ventilation inlets, used in heat exchangers and ejected below the aircraft belly through outlet vents. Together with the APU, these systems contribute significantly to the ramp noise (also called apron noise) perceived by personnel servicing the grounded aircraft. In the aeronautical field as in any other field of activity, occupational exposure to noise is one of the most significant health risks for workers, leading in some cases to irreversible hearing damage. Other adverse effects include hypertension, ischemic heart disease, annoyance, sleep disturbance and loss of concentration.
From the legislative viewpoint, ramp noise is dealt with by ICAO standards and regulations that have to be satisfied by aircrafts for certification. The European directive 2003/10 sets limits to the workers daily exposure which have a significant impact on the required workers protection equipment and on the servicing of a grounded aircraft if the noise levels exceed certain thresholds. Averaged permissible levels are established on an 8-hours daily working basis (Lex,8h), but at any instant a peak value Lpeak of 140 dB(C) cannot be exceeded. In no case the daily averaged exposure of a worker may exceed the limit of Lex,8h = 87 dB(A). These limits include the attenuation of any hearing protection equipment given to workers. When the daily equivalent level is over 80 dB(A) and/or the peak level exceeds 135 dB(C), the following measures have to be implemented: each worker must be provided with a proper information, and when applicable, a proper training regarding noise exposure, associated health effects and the use of protective equipment. Also, medical hearing checkups must be provided to workers asking for them. When the daily equivalent level is equal to or exceeds 85 dB(A) and/or the peak level is above 137 dB(C), technical and organizational measures must be implemented by the employer to reduce noise exposure. Areas where workers may be exposed to important noise levels must be indicated by appropriate signs and access to these areas has to be restricted whenever technically possible. This requirement can itself be contradictory to safety interests, as escape routes from an accident have to be as direct as possible.
Recent studies demonstrated that only a small minority of workers were exposed to daily-averaged levels below the limit of 85 dB(A). For some workers the level even exceeded the absolute permissible value of 87 dB(A). Apron noise is thus a very significant current problem that has to be tackled very seriously by ECS and aircraft manufacturers.
One of the issues with ECS design is that the various requirements that apply thereon including weight, space and noise, yield contradictory engineering solutions. The mass flow rate of ventilation air is usually provided by axial machines, which are intrinsically noisy as a result of the fluctuating forces exerted by the blades, yielding both tonal and broadband contributions to the sound spectrum. Because of space requirements, these machines must be as compact as possible, with the result that the rotation speed cannot be too low to provide the adequate flow rate. But noise power is typically proportional to a large power (5-6) of the blade Mach number. Moreover, the large fan speed combined with the large number of blades yields quite high frequencies for the tonal sound spectrum emitted by the blower, which will induce cut-on transversal modes contributing to the sound transmitted by the duct. These acoustic considerations must be included in the early stages of the design process to yield low-noise blowing units.
A quite generic configuration can be defined, surprisingly, for both cabin noise and ramp noise. A good example is the blower, which constitutes the main noise source for the most common ECS ramp noise applications, when using a pneumatic air conditioning pack, or when using a vapour cycle system for cooling purpose. The vapour cycle cooling system uses physical properties of a specific refrigerant fluid to cool fresh air (or re-circulated air), which is then distributed inside the cockpit or the cabin through complex duct systems. The blower will be predominant for recirculation noise, both in cockpit and cabin. While in the first example of ramp noise, the fan blade speed is almost supersonic at fan tip, resulting in quite high frequency noise, the two other application share a common frequency range and mass flow rate.
The blowing unit is a major contributor to ECS noise, but is not the only one. In addition, because of the large extent of an aircraft ventilation system and of tight space constraints, the layout of an aircraft ventilation network is quite complex, and includes many singularities such as valves, side-branches, manifolds and obstructions. Valves are used to control the flow rate at various locations of the piping system, which induce locally large velocities (in some cases choked flow) and hence strong aerodynamic noise. Diaphragms need to be placed in some ducts to provide an adequate pressure drop and a balanced flow rate across the network. Such singularities are also prone to strong noise generation. Flow noise is also produced by bends, which are inevitable in guiding the flow from the main blowing unit towards the cabin ventilation outlets and under-belly vent outlets. These components have to be included in the picture if one aims at reaching an acceptable sound pressure level perceived by the passenger/pilot or ground operator.
In order to meet certification requirements, it is thus quite important for an ECS manufacturer or integrator to be able to predict with a sufficient accuracy, and reduce where needed, the noise levels emitted by a given system. The prediction of aerodynamic noise, emitted by an individual valve for example, is not an easy task in itself, but is nowadays at reach with modern approaches and computational capabilities.
Since the advent of the acoustic analogy, significant progress has been achieved towards the understanding of the mechanisms through which sound is produced by an unsteady flow. The maturation of Computational Fluid Dynamics (CFD), permitting the modelling of the unsteady features of the flow field in particular, has paved the way to truly quantitative aeroacoustic simulations of individual components.
However, two key ingredients are still missing to provide a reliable ECS acoustic prediction model able to drive the design of quieter systems. Firstly, the interactions between subcomponents are crucial. Many of the studies carried out on individual components of an ECS system have indeed demonstrated a quite large sensitivity of the unsteady flow field, and thereby sound generation, with respect to their environment, expressed in terms of boundary conditions when considering modelling approaches. For example, the noise produced by the blowing unit is strongly dependant, both in spectral shape (tonal/broadband) and amplitude, on the inflow mean distortion and turbulence level. The same holds for singularities such as vanes, diaphragms and bends. For system designers, it is quite important to develop a system model assembly, composed of various duct component description and acoustic prediction laws. To enable this model development, sub-components need to be thoroughly tested and modelled, alone and coupled with their most recurring parts, defined as generic for a range of system distributions. To the present consortium’s knowledge, the multi-components interactions of the ECS have not been addressed previously, and constituted an important objective of this project.
Secondly, the numerical implementation of the acoustic analogy, based on unsteady CFD modelling (Large Eddy Simulation, Detached Eddy Simulation), still presents a prohibitive cost if design or optimization are targeted. Low-order modelling approaches, such as stochastic methods, or statistical approaches present much lower computational costs and give quite satisfactory results as long as the statistical models that are required as input are properly tuned. The combination of scale-resolved methods, based on time-resolved CFD, with statistical/stochastic methods is another important objective of this project, in order to permit studying and optimizing interactions effects with the current computational resources.
Simulating the ECS aero-acoustic performance is key to successful design, but the ultimate purpose is the reduction of the acoustic levels within the weight, space and efficiency constraints imposed by the aircraft manufacturer. In the aircraft industry, passive flow control strategies, which do not introduce additional energy in the controlled system, offer an interesting path that will be pursued in this project. The knowledge gained in the experimental and numerical investigation of the multi-components interactions is very instrumental in this respect. Indeed, having a better understanding of the mutual interactions causing additional tonal or broadband noise generation provides guidance about the flow manipulations that are the most appropriate to reduce the noise production. In this project, the costs and benefits of integrated passive flow and acoustic control approaches must be investigated by experimental as well as numerical means.
These passive approaches have been investigated for some time and with success in order to reduce the noise emitted by aeroengines, but require further optimization for the particular flow and frequency regimes encountered in ECS ducts. These innovative noise control approaches are to be first assessed on simplified (but representative) configurations, and on a production system in a second stage.
Project Results:
As a reminder, the research was articulated along a total of six work packages:
• WP1 coordinates the different strands of work in IDEALVENT.
• WP2 is focused on the experimental characterization of flow-acoustic and multi-component interactions. The purpose was to identify configurations relevant to the industrial application, and to provide exhaustive experimental databases permitting to i) understand the interactions mechanisms, ii) guide the methods developments in WP3, and iii) validate their application in WP4. In Task 2.1 the consortium has detailed the geometrical dimensions and range of flow regimes to be investigated in Task 2.2. In Task 2.2 a preliminary fast acoustic survey has been conducted on the configurations defined in Task 2.1 in order to identify relevant flow conditions that were investigated experimentally in greater detail in subsequent Tasks of WP2 and simulated in WP4. The relevant configurations were decided on the basis of the measured acoustic spectra compared with existing datasets made available by the final users. In Task 2.3 detailed flow and acoustic measurements have been performed for the valve-bend configuration. In Task 2.4 detailed flow and acoustic measurements have been performed for the fan with inflow distortion case. In Task 2.5 detailed flow and acoustic measurements have been performed for a tandem diaphragms case. In Task 2.6 the final user Embraer did perform acoustic measurements of one of its aircrafts grounded in its apron environment, in order to quantify the respective contributions of the ECS and of the APU at various locations around the aircraft.
• WP3 deals with the development of innovative simulation methodologies accounting for the interaction effects. The objective was to define the best modelling strategies for scale-resolved and statistical/stochastic approaches. In Task 3.1 the global specifications and ranges of applications of the different simulation approaches were detailed. In Task 3.2 the boundary conditions and interfaces required by the simulation tools were established. In Task 3.3 innovative scale-resolved and statistical approaches have been developed and improved for fan noise prediction. In Task 3.4 innovative scale-resolved and statistical approaches have been developed and improved for ducted turbulence noise prediction. In Task 3.5 multi-modal network models have been derived for the extraction of the active sound component from measurement or simulation datasets.
• WP4 applied the tools developed in WP3. The butterfly valve with upstream bend, fan with inlet distortion, and tandem diaphragms were investigated in Tasks 4.1 4.2 and 4.3 respectively.
• WP5 introduced noise mitigation strategies through passive flow and acoustic control. Porous liners and micro-perforates were investigated in Task 5.1. Turbulence screens were tested in Task 5.2.
Finally, WP6 integrated the outcomes of the previous WPs by providing guidelines on the integration of the sub-components of the ventilation system, and demonstrated their effectiveness in an industrial environment. The most promising control approaches tested in WP5 on the laboratory configurations were then tested on full-scale industrial equipment provided by Liebherr.
A summary of the most prominent results obtained in each of the WPs 2 to 6 is detailed below.
The WP2 final restricted and public reports (deliverables D2.7 and D2.8) discuss in length all the developments carried out in this WP, which can be summarized as follows:
• development of the fan test rig at VKI and of the valve test rig at KUL;
• preliminary flow and acoustic measurements on each test rig to verify that the correct operating conditions can be achieved and for the identification of relevant flow-acoustic installation effects;
• for the fan configuration, a theoretical analysis of the dominant sound generation mechanisms by application of analytical tools;
• development of an active multi-port sound extraction technique, based on the measurement of the modal reflection coefficients of the duct terminations and of the modal scattering matrices of the fan / valve components;
• quantification of the installation effects by application of the above-mentioned technique.
In order to investigate in WP2 the aerodynamic sound generated by an ECS fan (provided by Liebherr Aerospace) subjected to installation effects, a modular test-rig has been built at VKI. The modular structure allows investigating the noise generated by the axial fan, including both upstream and downstream installation effects. The loudspeaker modules are manufactured from aluminum by KTH in order to carry the loudspeakers which are used in the modal decomposition analysis. Upstream and downstream ducted microphone modules have been designed and manufactured by KTH as well, located upstream and downstream of the fan and diaphragm assembly. The in-duct microphone sections are used to measure the reflection coefficient downstream of the duct and for the modal decomposition measurements on the same rig. The modular structure of the test rig allows investigating installation effects through different configurations including empty duct, fan only, diaphragm only and fan-diaphragm configurations. The test rig also contains an auxiliary fan located outside of the anechoic room in order to generate flow for configurations in absence of the ECS fan. Besides the ECS fan mock-up of the fan with inflow and outlet distortion, investigated experimentally at VKI, also a mock-up of an ECS valve, provided by LTS, with inflow distortion is analyzed in the aeroacoustic test facility of KUL. Loudspeaker and microphone modules similar to those used in the VKI fan rig were designed by KTH for the multi-port eduction of the valve noise subjected to various inflow conditions.
Three main axes were followed in WP3 for the development of prediction methods adapted to the IDEALVENT configurations, which can be ranked by increasing computational cost (and presumably accuracy) as follows: statistical, stochastic and scale-resolved methods. The developments carried out thereto by the different partners are summarized in what follows.
To produce tractable solutions with the analytical formulations, the ECS fan stator vanes need to be simplified to rigid flat plates of zero thickness embedded in a uniform mean flow. The simplest strategy is to determine the fluctuating lift forces on the vanes with linearized isolated-airfoil theories and then to make the forces radiate. This two-step approach neglects the so-called cascade effect, which is defined as the effect of adjacent vanes on the aerodynamic and/or acoustic response of one of them. Because of the large overlap and the large solidity in the stator, the cascade effect is very important. The need for cascade response functions to replace simple Amiet’s theory has been recognized and motivated the development of ad hoc mathematical formulations in the literature. However, the main drawback of their formulation is that it cannot be extended for a three-dimensional annular cascade. In contrast, the mode-matching technique can overcome this limitation.
For the considered fan case, ECL has established that wake turbulence impinging on the stator vanes appears as the dominant source of broadband noise to be modelled in priority using a dedicated analytical model. Furthermore, the periodic (averaged) part of the wake velocity deficit is also responsible for the generation of tonal noise by impingement on the vanes. This makes wake interaction the dominant mechanism for both broadband and tonal contributions, for which the equivalent sources are distributed on the stator vanes. As a result, the rotor is just considered as a generator of vortical disturbances. The stator is the only part to be modeled geometrically. Independently of wake interactions the rotor also contributes to the tonal noise by two mechanisms, namely its interaction with the stationary inflow distortions (strongly depending on the inlet geometry) and its operation through the upstream potential field of the stator. Such mechanisms have been addressed independently, by means of a mode-matching technique developed by ECL in this project.
An alternative approach, still based on linearized airfoil theories but relying more directly on numerical approaches for the determination of the turbulence statistics and for the prediction of scattering effects, was researched by VKI for the prediction of the IDEALVENT fan noise. An automatic data extraction procedure from RANS CFD computations was developed. This has been implemented in a general framework and integrated in BATMAN (Broadband And Tonal Models for Airfoil Noise), the VKI in-house code implementing noise prediction methods based on Amiet’s theory for leading and trailing edge broadband noise and tonal interaction noise. A special reader module based on mexcgns matlab libraries has been implemented to interface CGNS TRACE CFD results from DLR with VKI post-processing routines. The VKI routines being based on strip theory, all the post-processing is performed at iso-radii, cuts being automatically performed on the blade surface, and the rotor and stator are post-processed separately. The extraction of data is used as inputs for the SISW methods using a combination of Amiet’s theory with the SISW Acoustic Transfer Vectors technique and the ECL method based on mode-matching technique.
With regard to stochastic approaches, DLR has further improved its CAA methodologies as needed to deal with the different duct and fan-noise problems. With respect to fan noise, DLR uses a hybrid strategy based on CFD tools optimized for aero-engine applications.
For fan noise, the hybrid approach involves three numerical tools:
• the URANS solver TRACE,
• the CAA solver PIANO,
• the turbulence synthesizer FRPM.
A URANS or Harmonic Balance (HB) computation of the fan-stage gives the fluctuation fields that are necessary to perform tonal noise prediction. A CAA computation is then carried out with PIANO, using the mean-flow solution given by TRACE as background flow for the perturbed equations. The turbulence-induced noise sources are generated in time domain by a turbulence generator - the so-called fast Random Particle Mesh (FRPM) method - which uses the turbulence statistics from URANS as input. This gives the broadband fan noise component. Three main developments were conducted in IDEALVENT, for the application of the method to rotor-stator interaction noise:
1. accounting for the mean flow periodic variation;
2. accounting for the so-called cyclo-stationarity of the turbulence properties;
3. accounting for the fact that the integral length scale and turbulent kinetic energy in the wake of the rotor differ from the region outside.
The cyclo-stationary (or periodic components) of the flow are extracted from a URANS or HB computation and prescribed as complex Fourier coefficients to the PIANO and FRPM computations.
For the prediction of the noise produced by ducted singularities such as valves, a hybrid prediction chain similar to that used for fan-noise predictions is applied by DLR. For the prediction of the stationary turbulent flow problem inside the duct RANS predictions are based on the unstructured DLR flow solver TAU. To summarize, the prediction chain involves the tools:
• the RANS solver TAU,
• the CAA solver PIANO (structured) / DISCO++ (unstructured),
• the turbulence synthesizer FRPM.
The developments carried in IDEALVENT in relation with these tools include:
1. the efficient extension of the Gaussian based FRPM method to more complex turbulence spectra;
2. the introduction of anisotropy into the modeling of stochastic fluctuations by means of an anisotropic scaling tensor that replaces the scalar scaling function;
3. the introduction of a vorticity mode selective relaxation source term (Eddy relaxation source term) in PIANO;
4. the coupling of FRPM with the Discontinuous Galerkin solver of DLR (DISCO++) to enable broadband noise prediction for complex geometries otherwise not resolvable by structured meshes. (applied in WP 4 to the butterfly-valve test-case).
In contrast with the approaches described above, the deterministic methods developed in this project are solving time- or frequency-dependent equations in a deterministic way, to compute either the noise produced by a given ducted element (fan or valve), or its sound scattering properties.
The major objectives of NTS consisted in the development of improved scale-resolved techniques for prediction of the ducted fan noise and of the noise produced by a stationary obstacle (e.g. diaphragm) inside the ECS duct.
The approaches developed by NTS for solving these two problems are essentially the same. They consist in Direct Noise Computation (DNC) based on compressible scale-resolving simulation of turbulence with the use of a hybrid RANS-LES method of the DES-type (Improved Detached Eddy Simulation – IDDES) combined with synthetic turbulence generator for “creation” of realistic turbulent content upstream of the noise-generating region (fan or diaphragm and the area with high level of turbulent fluctuations downstream of them). In principle, such an approach is capable of predicting both the tonal and the broadband components of the noise produced within the source area and its propagation over the ventilation duct and in the surrounding ambient air (of course, within a range of frequencies accurately resolved by the numerical method on the grid used in the computations). However, such computations require huge (currently non-affordable) computational resources. For this reason, the computational domain in which the turbulence-resolving simulations of the flow are performed has to be limited to a relatively small part of the duct neighboring the source region. This has two major consequences.
Firstly, in addition to IDDES computation of the flow in the limited computational domain, the procedure must involve auxiliary stages necessary for setting inflow/outflow boundary conditions. Secondly, a problem of extraction of acoustic signal from the simulations carried out with the use of the developed approach is far from trivial because in the vicinity of the source region (at least downstream of it) the acoustic signal turns out to be almost indistinguishable against much stronger pressure fluctuations caused by turbulence and, in particular, by impinging of turbulent structures on the duct walls. Therefore, in order to extract the acoustic signal, some supplementary acoustic techniques should be used, for which the detailed unsteady information obtained in the course of the scale-resolving IDDES of the flow serves as an input. Such techniques were developed by IDEALVENT partners (SISW, KTH, VKI, DLR).
All the elements of the outlined computational approach were implemented in the NTS in-house general purpose code for IDDES based on the two-equation k-ω SST background RANS turbulence model and (for the fan flow) on the same model with streamline curvature/rotation correction (SST-CC model). This is a structured finite-volume high-order CFD/CAA code accepting multi-block overlapping grids of Chimera type. It has been intensively used for a wide range of aerodynamic and aeroacoustic problems and shown to be rather reliable and accurate. The code was adapted to the computation of the flows with rotating elements with the use of the approach based on employing different reference frames in different grid blocks combined with the sliding interfaces technique for imposing the boundary conditions on the interfaces between rotating and stationary grid blocks.
Major new developments made in the course of this work, which ensured its overall success, are as follows.
1. Volumetric Synthetic Turbulence Generator (VSTG hereafter) is developed, which is based on introducing distributed volume sources (“body force”) into the momentum and turbulent kinetic energy transport equations.
2. A modification of the original IDDES method is proposed, which ensures significant weakening of the well-known flaw of DES, namely, a severe delay of the Kelvin-Helmholtz instability and transition from fully modeled to mostly resolved turbulence in the free and separated shear layers (RANS-to-LES transition).
Still in the framework of scale-resolved simulations, important developments were carried out by SISW for the coupling of time-resolved CFD solutions with their acoustic solver Virtual.Lab Acoustics. Three main axes of research were explored:
1. aeroacoustic source modelling for internal flows, by extending a method to express dipolar sources as Neumann-type boundary conditions in an acoustic boundary value problem;
2. statistical approaches for computing the scattering of broadband fan noise, combining an analytical statistical description of the incident field with a technology based on a boundary element method to compute scattering problems;
3. advanced high-order finite element solvers for duct acoustics including flow effects.
The first approach consists in transforming Curle’s dipoles into Neumann boundary conditions of the Helmholtz equation. This is done by evaluating numerically a (hypersingular) integral involving the flow pressure and an appropriate Kernel function. This new approach was originally mainly applied to exterior problems, such as airframe noise problems. An initial investigation on a basic interior flow (a simplified HVAC duct) prior to the start of this project, proved that the application to internal flows with complex geometries presents additional challenges which require a dedicated study, such as the treatment of free edges and junctions. Specific aspects related to the application of this approach to ducted flows have been investigated, such as strategies to minimize numerical leakage in interior/exterior problems, as well as the impact of the length of the source region. Furthermore, a new formulation for the Neumann boundary condition in the presence of acoustic impedance has been derived. Finally, the algorithm to convert dipoles into boundary conditions has been improved for cases dealing with incompressible CFD input with multiple frequencies running in parallel in several processors, as well as for cases with multiple CFD time windows for averaging. The latter is necessary for averaging the results of turbulent flow computations, which are chaotic by nature and typically require several realizations to converge to the experimental solution.
The second aspect developed by SISW related to the computation of broadband fan noise scattering for industry-standard applications. The approach combines an analytical formulation based on Amiet’s theory with the so-called Acoustic Transfer Vectors (ATV) approach. The acoustic transfer functions are computed through boundary element methods, based on the acoustic pressure and acoustic velocity Power Spectrum Density (PSD) of the broadband fan source. The approach is general and valid for any stochastic source characterized by their radiation PSD. This approach allows taking into account impedance values (to characterize for instance absorbing surfaces or micro-perforated materials) in a straightforward manner. As a validation case, the sound radiated by a fan inside a cooling unit was computed. The initial prototype was previously used to compute this case. The optimized implementation achieves a total speed-up factor of around 450 (run sequentially). The parallel implementation offers potential further gains in computation speed. This can reduce the typical computation cost for large case from several days to less than an hour, thus making it possible to tackle cases of industrial size in general, and, in particular, the fan problems to be computed in the WP4 of IDEALVENT.
The last subject of developments by SISW addresses its finite element solvers for duct acoustics, in order to achieve more accurate and efficient computations of the IDEALVENT configurations. The use of an axisymmetric implementation of the problem as a means to reduce the cost of three-dimensional computations was investigated. In particular, an axisymmetric approach to solving the linearized Euler equations (LEE) was evaluated using a frequency domain high-order FEM. The application of this model to the Munt problem, as a test case, showed results in good agreement with the reference solution. The acoustic duct modes are injected through a Perfectly Matched Layer (PML) and radiated outside the duct, where a PML is also applied in order to absorb the outgoing waves. In addition, the three-dimensional finite element solvers were optimized by means of a high-order adaptive finite element implementation, which had already shown an improved computational performance for the Helmholtz equation, including potential mean flow effects.
The determination of the scattering properties of ducted components is important for the characterization – through experiments or compressible flow simulations, of acoustic installation effects and of the active noise produced by the element under consideration. In that line, KTH extended their existing two-dimensional frequency-domain Linearized Navier-Stokes (LNS) solver, including the various boundary condition formulations and the interfacing schemes to interpolate the mean flow variables from the fine RANS (Reynolds Averaged Navier-Stokes) grid to the coarse acoustic mesh, to a full HPC-running three-dimensional frequency-domain LNS solver. This solver includes an additional source term contribution from the mean flow field solution to better model the viscous dissipation caused by the mean flow field variables. The updated formulation can be directly linked to multi-port network models, being solved in the frequency domain. This means one first computes the passive scattering matrix of a component by simulating the reflection and transmission using the LNS approach for both the up- and downstream side. This must be done for each propagating mode in the up- and downstream ducts to build the complete scattering matrix. Once the scattering matrix is known it can be combined with the up- and downstream boundary conditions also expressed as multi-ports.
Linearized Navier-Stokes and Linearized Euler Equation solvers were also topics of research by the KUL team. They developed in the past an in-house time-domain discontinuous Galerkin (DG) solver, which is parallelized using MPI. The solver uses a 3D quadrature-free implementation in the bulk of the domain while a local quadratic implementation in the vicinity of highly curved walls, significantly reduces the number of elements while maintaining the same accuracy. An optimized RK-time marching scheme further reduces the dispersion and dissipation errors and increases the maximum allowable time-step. As governing equations the Linearized Euler Equations (LEE), the Irrotational Linearized Euler Equations (ILEE) as well as the Linearized Navier-Stokes (LNS) equations can be used, dependent on the governing physics of the influence of the mean flow field onto the propagation of the acoustic waves. For the coupling between CFD output and their DG solver, KUL implemented a flexible interfacing scheme for the mapping between the CFD and the acoustic propagation meshes is developed. The adopted approach uses a global least square procedure for continuous mapping in combination with an additional anti-aliasing filtering.
KUL developed as well an efficient time domain impedance and transfer-admittance formulation. By writing the transfer-admittance matrix coefficients as a sum of first- and second-order systems, a recursive formulation can be obtained for explicit non-uniform multi-stage time integration schemes with constant time step. This technique was implemented into the in-house RKDG solver for the time-domain LEE and LNSE. The formulation was successfully verified for simple benchmark cases (both in two and three dimensions), without mean flow and with simplified grazing flow profiles for which the Ingard-Myers condition (unstable in time-domain) was not assumed and the boundary layer profile was fully described.
The WP4 was dedicated to the application of the simulation technologies that were upgraded in WP3 to the three main flow categories: the industrial valve, the tandem diaphragm and the industrial fan. The simulation approaches span across a range of CPU cost vs. accuracy spanning from stochastic to scale-resolved approaches.
Simulations of the broadband sound generation by the butterfly valve have been conducted by DLR Braunschweig using the unstructured CAA code DISCO++ together with the FRPM module to reconstruct turbulent sound sources from an inexpensive precursor CFD simulation solving RANS equations. The results obtained permitted to demonstrate the general applicability of stochastic sound sources to realistic three-dimensional system components, even though their superior efficiency, compared with scale-resolved approaches, is less obvious than for cases that exhibit some homogeneity in one direction, permitting a locally two-dimensional treatment.
The active and passive properties of the ducted industrial valve were simulated using the panel of approaches developed in WP3. The passive properties of the valve were computed by KUL using their in-house, time-domain Runge-Kutta Discontinuous Galerkin (RKDG) discretization of the Linearized Euler Equations (LEE). Linear elements, allowing the use of an efficient quadrature-free DG formulation, are used throughout the domain except in the vicinity of the curved walls of the circular duct where curved second-order elements are locally used locally around complex geometries. This mixed approach results in an optimal balance between the computational efficiency of the quadrature-free method and the geometrical flexibility of the second-order elements. Due to the small mesh size in the vicinity of the valve the use of quadratic elements in this region is not needed. For the inter-element communication, a Lax-Friedrich formulation is used for the convective fluxes. Boundary conditions are imposed by specifying the fluxes at the boundary faces. The time integration is done using a Runge-Kutta scheme, optimized for the spatial discontinuous Galerkin operator. An excellent agreement was found between the experimental and numerical results for all modes of the valve scattering matrix, except at frequencies slightly above the cut-on frequency where the numerical results show, similar as for the circular duct results, an increased transmission and a decreased reflection.
For the deterministic calculation of the active noise produced by the valve, multiple efforts were combined by several consortium partners. The CAD of the valve is made available by LTS. RANS computations have been performed by VKI using rhoSimpleFoam, the compressible steady solver of OpenFoam 2.4.0. The k-ω SST turbulence model is used with second order accurate schemes for all variables. Air is used as fluid with perfect gas properties and the Sutherland model is used for the variation of viscosity with compressible effects. Other properties are kept constant. Mass flow rate is used as inlet condition. As the corresponding value is not directly available from the experiments performed by KUL, but only the pressure drop and the corresponding volume flow rate, computations are performed in a range of inlet mass flow rate. The operating conditions corresponding to this specific pressure drop are then determined through the operating curve results.
VKI ran the unsteady computations using the SAS turbulence model of the CFD solver Fluent 16.2. It allows the resolution of the turbulence spectrum in unstable regions, resulting in a LES-like behavior in unsteady regions of the flow field. At the same time, it provides standard RANS capabilities in stable flow regions. As for the RANS computations, second order schemes are used for all variables and central bounded second order schemes are used in time. Same fluid properties are used as in the RANS computations. Furthermore, general Non-Reflecting Boundary Conditions are used to limit possible reflections at the inlet boundary conditions. At the outlet, a pressure far-field condition is applied that is a non-reflecting boundary condition based on the introduction of Riemann invariants for a one-dimensional flow normal to the boundary. Wall-pressure fluctuations (dipoles) and volumetric velocity data (quadrupoles) are extracted in the complete CFD domain, to be used as input to acoustic computations performed by SISW. These simulations were conducted solving a linearized acoustic propagation equation in potential flow, with volume-distributed sources based on vortex sound theory. The computational results show a difference between upstream and downstream acoustic levels that is similar to that of experiments, and the overall behavior seems to follow the same trends.
For the diaphragm simulations, two tandem configurations were considered with the separation distances between the diaphragms equal to 2 and 4 duct diameters. In addition, a baseline case of a single ducted diaphragm was analyzed. Overall, the NTS predictions of the wall-pressure spectra are in good qualitative and acceptable quantitative agreement with experiment. The main discrepancy resides in the low frequency end of the spectra of the upstream propagating noise and is somewhat more pronounced for the single diaphragm case. A cause of this discrepancy is not quite clear so far, but most likely it is associated with noticeable reflections of the low frequency sound waves by the duct inlet in the experiment. Particularly, because of these reflections, the presented comparison of the “raw” spectra from the simulations and experiment cannot be considered as fully consistent. A more consistent comparison was provided by KTH who has extracted from both the experimental and computational pressure signals the individual acoustic modes free of the impact of the turbulence impinging on the duct walls and inlet/outlet reflections of sound waves.
Using the compressible CFD data obtained by NTS, KTH applied its modal decomposition technique to extract the active noise produced by the tandem configuration, for several axial separations. The decomposition of numerical data brings a number of advantages compared to the empirical approach. Admittedly, the computed time samples are still much shorter than data gained from measurements; however, a plethora of sensoring-points is accessible, even inside the channel cross-sections. The decomposition matrix is hence highly overdetermined and guarantees a stable matrix inversion. Furthermore, in computations we can disable all the disturbing effects which usually complicate measurements.
Hybrid aeroacoustic computations were run by SISW for the single and tandem diaphragm configurations. The sources are based on the compressible CFD data provided by NTS. Since the CFD is also able to provide the acoustic levels, these are used together with the experimental data as a reference solution to compare the results of the hybrid approach. In the cases with two diaphragms, the sound production is concentrated around the second diaphragm, due to the effect of the turbulent flow generated behind the first diaphragm impinging on the second diaphragm. As a consequence, the dipole sources on the downstream diaphragm are the dominant source of sound, and the sound production is less sensitive to the type of flow (laminar or turbulent) that is injected upstream of the first diaphragm. Therefore, dipole sources are defined only on the downstream diaphragm in the hybrid computations. The results were compared to experimental measurements obtained in WP2. The hybrid computational approach seems to provide an overall good match with the experimental data and with the direct CFD results, thus confirming that the dipole sources on the downstream diaphragm are the dominant source of sound. For each case, two different hybrid computations are carried out: one based on the CFD with clean inflow, and another based on the CFD with upstream turbulence injection. It can be observed that this yields some differences at low frequencies, where the flow with the turbulent inflow tends to produce more noise. By contrast, the flow with the clean inflow seems to produce somewhat more high-frequency noise.
The modelling of the noise emitted by the industrial fan was tackled from different approaches: (U)RANS, DES, hybrid methods, using deterministic, stochastic and statistical methods.
Unsteady RANS simulations were performed by DLR using the Harmonic Balance method. The tonal noise is extracted from the URANS simulation by using the DLR in-house code Connect3D, which performs an extended Triple-Plane-Pressure (XTPP) mode matching to extract the amplitude and phase of the acoustic propagating modes.
For the DLR broadband simulations, the cascade is realized with all 10 stator vanes and a periodic boundary condition in the azimuthal direction. In axial direction radiation and outflow condition are combined with a sponge region to minimize reflections. The turbulence is generated by the FRPM method and imposed to the CAA domain using the LEE-relaxation method at the inflow patch (blue). Only a single vane is excited assuming uncorrelated noise for adjacent vanes. The turbulence is eliminated by a second patch to reduce spurious noise in the downstream microphones.
The approaches developed by NTS have been applied in WP4 for the prediction of the effect of inlet flow distortions on the noise emitted by the industrial fan. In the first (baseline) case the duct has a smooth bell-mouth inlet resulting in a "clean" flow upstream of the fan with turbulence confined within a thin boundary layer on the duct wall. In the two other cases, a significant non-uniformity of the flow and elevated levels of turbulent fluctuations in the upstream part of the duct away from its walls are created by replacing the bell-mouth inlet by inlets with a rectangular-to-circular transition of the duct cross-section and with a T-junction formed by two circular inlet pipes.
Fan computations were as well carried out by SISW, this time to establish the passive behavior of the fan as a purely scattering element, without mean flow (Helmholtz equation). The results indicated that the first drop in transmission of the plane wave mode and in the first azimuthal mode is due to the cavity in the stator area, while the second has to do with the cavity in the front of the fan. If we observe the pressure field in the near-field of the fan, this becomes apparent. With respect to the experimental data, the computed results show similar qualitative trends. For instance, both experiments and computations predict the drops in transmission due to the cavities at rotor and stator sides at similar frequencies. Nevertheless, the experimental results display in general lower levels of transmission, and less pronounced peaks, thus suggesting the presence of acoustic damping mechanisms that are not modeled in the acoustic computations.
The acoustic mesh used by SISW for the computations of the active contribution (with aeroacoustic sources) of the fan cases is similar to the one used for the passive contribution, but the rotor surface is excluded from the mesh, leaving only the stator, where the sources are defined, in the duct mesh. The mean flow is assumed to be negligible in the acoustic computations.
It was observed that the predicted broadband levels are in very good agreement with the CFD input. This suggests that indeed the dipole sources on the stator vanes are sufficient to characterize the broadband noise produced by the fan. By contrast, the levels at the BPF are significantly higher in the current computations. Since the acoustic mode at the BPF should be cut off, this indicates that there are some inaccuracies in the acoustic results preventing the cancellation of the acoustic peak. It must be pointed out that the acoustic mesh and the CFD mesh were not exactly coincident and may have generated some inaccuracies in the mapping of the flow pressure from the CFD mesh to the acoustic mesh. This numerical effect may be worsened by the close proximity of the BPF to the cut-on frequency of the first radial mode. In any case, the agreement of the broadband levels is quite satisfactory and confirms the validity of the approach to predict broadband fan noise. It is worth noting that the approach is also applicable if the CFD is incompressible and thus unable to provide acoustic predictions.
On the analytical side, the implementation by ECL of the MMBW (Mode-Matching in Bifurcated Waveguides) approach has first been assessed in a two-dimensional context and validated by comparison with alternative mathematical formulations, typically the Wiener-Hopf technique and numerical simulations. This was achieved both in the framework of the IDEALVENT project and during the French program SEMAFOR, by addressing the two following generic problems:
• transmission of an oblique acoustic wave (equivalent to a duct mode in an unwrapped configuration);
• generation of acoustic waves due to a single impinging oblique vortical gust.
When addressing pure sound transmission, the predicted acoustic fields are in complete agreement with those of alternative methods provided that the Kutta condition is implemented, but de-activating this condition leads to significant errors (increasing with increasing Mach number). In absence of Kutta condition, the energy balance for acoustic waves is very accurate. With the Kutta condition activated, part of the energy of the acoustic motion is lost because of conversion into vortical motion. When addressing sound generation by impinging vorticity, again a full agreement is found between all approaches/methods if the Kutta condition is implemented.
Finally, in WP4 LTS tested the applicability of an empirical prediction approach, based on similarity laws and existing data for other similar fans.
The WP5 dealt with the exploration of mitigation approaches to reduce the noise produced and propagated in the Environmental Control System (ECS). A strong constraint that was taken into account concerns the overall efficiency and reliability of the control. Passive strategies, which do not inject additional energy into the system, were adopted for the sake of safety and energy consumption. The mitigation operates on two fronts: flow and acoustic control, using turbulence grids to dampen out the inflow inhomogeneity and turbulence, and porous liners/microperforates for acoustic absorption. Both experimental and numerical research have been carried out in WP5, though not systematically for each control technology. A specific effort was brought on developing the knowledge around the use of so-called micro-perforated plates (MPP’s) for noise control in ducts. A MPP is essentially a perforated plate, with holes that are so small so that the viscous losses dominate over the inertia of the air moving at each hole. The classical application for MPP’s have been in room acoustics as so called panel absorbers. More recently it has been suggested that MPP’s also have a large potential for other applications for instance involving noise in ducts. This was the motivation behind including MPP’s in the study of new and innovative passive noise control measures within the IDEALVENT project.
KUL investigated experimentally the properties of a set of modal filters with one MPP panel mounted in the mid-plane over the full length of the filter. Applying the multi-port eduction method developed in WP2, they measured the power loss resulting from the insertion of the modal filters subjected to prescribed incident modes.
When a flow is present through the modal filter, the micro-perforated panel is subject to a two-sided grazing flow. For micro-perforated panels with circular and slit-shaped perforations, the presence of a grazing flow on one side of the panel increases the resistance and slightly decreases the reactance. Similar effects can be expected for two-sided flow over a MPP with slit-shaped perforations.
KTH had already investigated the principle of using an expansion chamber muffler with a micro-perforated sheet tube as a dissipative silencer prior to this project. The damping behaviour of such a filter is constituted by both the impedance of the cavity and the impedance of the micro-perforated plate which covers the cavity. The optimum wall impedance for plane wave damping is chosen as the so-called Cremer impedance adapted to circular ducts and flow. The computed impedance could easily be realized by designing cavity structures fitting to an existing micro-perforated layer. A problem in practical applications is, however, that the optimal impedance changes with frequency and modes. A silencer is therefore only maximal effective for a small frequency band and a certain acoustic mode. For this reason, it has been attempted in this project to combine the performance of a MPP modal filter with that of a Cremer filter, in a so-called Cremer-Modal Filter (CM-Filter). Two different arrangements were first investigated numerically: a serial one, and a cladded one. The serial CM-filter had a higher total loss for the (0,0)-mode, which is reasonable due to a longer effective Cremer-impedance. The serial-connected CM-filter became more effective to the (1,0)-mode above 2700 Hz and for the (2,0) mode for all frequencies.
Modelling work was also conducted by KUL, yielding promising results considering the approximations made in the choice of governing equations and the rational impedance fit.
Likewise, SISW performed some research on computational modeling aspects of perforate plates. The task consisted in extending the potential flow acoustics propagation solver to account for transfer impedance boundary conditions in the presence of grazing flow. A transfer admittance model was implemented in the Linearized Potential Equation (LPE) code of Sysnoise. It represents an extension of the transfer admittance functionality for cases with a non-zero mean flow. The implementation considered the possibilities of matching and non-matching meshes on either sides of the MPP, and was validated against standard impedance models for conditions where both modelling strategies apply.
The two novel noise reducing modules, the MPP (Micro-Perforated-Plate) modal and the Cremer filters, have been tested on the VKI test-rig in similar conditions than those used in WP2. Both modules have been designed by KTH and built by SNT. The modal filter is designed to damp the two lowest circumferential modes. The Cremer filter is optimized to damp the plane wave mode at the fan BPF.
Three different configurations were tested. Measurements are first performed on an empty configuration, with an empty duct upstream of the fan and the bellmouth used at the inlet. Using the same bellmouth inlet, the filters are then inserted upstream of the fan and measurements are performed for two configurations, with the Cremer-modal filters or modal-Cremer filters, in order to verify their relative efficiency depending on their relative position to the fan. The results have demonstrated the high efficiency of the serial Cremer-Modal filter unit. It both reaches the target of more than 10 dB damping at the fan BPF but also delivers a broad band damping. This is due both to the action of the modal filter and the Cremer silencer. Because the Cremer silencer can although it is tuned to a certain frequency be quite effective in a band around this frequency. It can also be noted although the Cremer silencer is optimized for the plane wave, it is still very or often more efficient for the higher order mode components at the BPF. However, the drawback is that is quite reflective for the non-optimum modes. This is compensated for by combining the Cremer silencer with the upstream modal filter which gives a broad-band damping to circumferential modes up to order 2.
Similar experimental investigations were performed by KUL on their test rig for the valve configuration. The results weren’t as promising as for the fan configuration, which can be explained by the fact that most of the aerodynamic sources excites the plane wave mode (0,0), while the modal filter is only damping noise of the energy which is contained in the azimuthal modes. As such, for the butterfly valve application the presence of a modal filter does not yield any significant noise mitigation.
On the basis of all the results obtained during the project, LTS drew a synthesis report summarizing the main project outcomes, with a focus on the quantification of the installation effects and potential of the mitigation strategies. In a nutshell the main conclusions can be summarized as follows:
• For the fan component:
o The diaphragm at the exhaust of a fan can have an important impact on noise radiated by the system:
▪ If the diaphragm doesn’t have a high pressure drop, the fan works on a similar operating point, in this case, the impact of the diaphragm on the fan noise is 2-3 dB.
▪ If the diaphragm induces a high pressure drop, the noise produce by the fan-diaphragm association can be very different: the noise is increased by about 15 dB. This change is not due to the diaphragm but to the fan which is near a rotating stall operating point and generates high noise levels.
o Inlet fan modules create inlet distortions that can add high noise level, especially at low frequencies.
These interactions between fan, diaphragms, and upstream distortions must be analyzed during the design of air distribution. The acoustic studies of isolated components are not sufficient to have a good noise prediction of the global system.
• For the valve component:
o The aerodynamic noise production of the butterfly valve is dominated by the static pressure drop over the valve.
o In the experimental set-up the valve is responsible for >90% of the total pressure drop.
o The far-field radiation becomes efficient above the cut-off frequency.
o No installation effects (presence of the bend, distance bend- valve, turbulence screens) are noticeable.
o It is possible to simplify the butterfly geometry as a flat plate to have similar acoustic behavior as the complex valve geometry.
The only mitigation guideline on this particular noise source amounts therefore to reducing the pressure drop over the valve.
The final activity of IDEALVENT, conducted by the final user LTS, consisted in integrating the outcomes of the project by testing the most promising noise mitigation strategies in one of its industrial systems. The main components of the system include:
• an inlet part manufactured by 3D Additive Layer Manufacturing,
• a condenser exchanger,
• a butterfly valve,
• a fan, identical to the one investigates in the VKI test rig.
To assess the effect of the different noise reduction units, the following configurations have been tested at inlet and outlet of the system:
• baseline (i.e. without any reduction noise device);
• splitter upstream the fan;
• Cremer upstream the fan;
• splitter and Cremer in serial (splitter upstream);
• Cremer and splitter in serial (Cremer upstream).
While a full accounting of the results of this test campaign is provided in the WP6 final deliverable, the main results can be summarized as follows:
• For the sound propagated towards the system inlet:
o The three configurations including the Cremer unit showed a good attenuation, in particular around the first BPF of the fan, where the noise reduction reaches 16dB (for the Cremer and splitter configuration).
o The Cremer unit (and Cremer + Splitter) shows an interesting broadband attenuation. The splitter and the Cremer units display 4dB noise reduction in high frequency above 4000Hz.
o The Cremer and Splitter configuration gives the best results with the combination of both effect noticed, with 9dB noise reduction on global noise level.
• For the sound propagated towards the system outlet:
o The configurations with the Splitter unit do not show a significant reduction on the measured noise at outlet of the system. However, the Cremer and Cremer + Splitter configurations give a slight attenuation on a wide frequency band.
o The best noise reduction result is obtained with the combination of Cremer and splitter units leading to 2.5 dB of noise reduction on global level.
To conclude, the IDEALVENT project has laid down an essential milestone in the investigation of the aerodynamic noise generated by Environmental Control Systems. Unprecedented efforts have been devoted to its experimental investigation, its prediction through a wide range of modelling approaches, and its mitigation by innovative passive control strategies. The project achievements are meant to bring rapid and long-term benefits for the design of better systems and to substantially improve passenger comfort and occupational health on the ramp.
The execution of the project didn’t require any significant deviation from the planned work program, as far as the global objectives and ambition of the project are concerned. Relatively minor alterations of the technical program were implemented where necessary, e.g. due to scientific outcomes making initially scheduled activities irrelevant or unnecessarily challenging. Such re-organization, which was agreed upon by the concerned partners during WP meetings or the 6-monthly regular consortium meetings, is inherent to any research project where by definition, the scientific outcomes can hardly be anticipated. Overall, the partners all agreed at the final meeting that the management of the project ran very smoothly and with a certain pragmatism focused on the final and most important objectives of the project.
The results produced by the project have been exploited by the various partners in different ways, depending on the type of results and partner activities. Generally, the results have been exploited by all partners to generate publications at international conferences and in peer-reviewed journals. This activity is important for maintaining the international reputation of the academia and research groups, to establish the young researchers in the scientific community, and to evidence the innovation potential of the industrial partners. To date, the project has resulted in a number of conference and journal papers, that have been made publicly available on the project website www.idealvent.eu wherever permitted by copyright ownership regulations.
A considerable amount of numerical and experimental data has been developed by the partners. In order to secure the future usability of these data, by the consortium and beyond, the consortium members endeavor to maintain and document a significant subset of key results to permit the validation of future numerical or modeling approaches. The datasets will be managed by the partners having generated them.
From a more academic viewpoint, the exploitation of the knowledge developed in the project is twofold. Firstly, the novel theoretical developments have been assimilated into the ex cathedra course material, and thereby disseminated to the future generations of engineers trained by the academic partners. Secondly, the application cases described above, inspired by industries and including experimental and numerical databases, constitute excellent training material for the final year Masters and PhD projects hosted by the universities.
Method-wise, the academia and research centres of the project have consolidated and upgraded their theoretical, experimental and numerical methods, which will be further used and developed in future projects. The software vendors have improved their solutions, integrated in the final product as novel libraries. The new tools being validated on industrial configurations can now be marketed towards the fields of activity of the project and beyond. Indeed, the scientific and technological solutions developed in this project are applicable to other sectors of activity than aeronautics including Heating, Ventilation and Air-Conditioning (HVAC) systems in surface transport, buildings or industrial cooling systems. The outcomes will accordingly be exploited in future research and commercial projects.
Last but not least, the project will have offered an unprecedented opportunity to the final users LTS and EMB to develop a quantitative assessment of the importance of installation effects in Environmental Control Systems, and of the impact towards the passengers and personnel servicing the aircraft on the apron. This knowledge, and the exposure of the final users to state-of-the-art modeling and mitigation strategies, will serve them to design better products and improve their competitiveness.
Potential Impact:
The IDEALVENT project contributes directly to several of the objectives of the ACARE Strategic Research Agenda for air transport, and Flightpath 2050 vision for aviation. The project outcomes will lead to novel aircraft integrated ventilation concepts for the reduction of cabin and ramp noise, which is being recognized as a growing issue for passenger comfort and safe aircraft operation and maintenance.
Regarding ramp noise first, our consortium estimates that for the positions farther from the APU (entrance door, some servicing positions), a reduction of the ECS noise sources will reduce the overall ramp noise by the same level. In other terms, designing a system 10 dB(A) quieter, the overall noise level at the regions specified above will also be reduced by 10 dB(A) as well. At the tail of the aircraft the noise levels will be still dominated by the APU, but EMB estimates that for this case a reduction of 3 dB(A) can still be expected. This clearly demonstrates that the IDEALVENT project can potentially have a very significant positive impact on the occupational health and safety of the personnel servicing the aircraft.
Then, cabin comfort is more than a question of convenience. The adverse effects that noise can have on health have been documented for quite some time, and include hearing impairment, hypertension, ischemic heart disease, annoyance, sleep disturbance and loss of concentration. Furthermore, the quality of the cabin acoustic environment is becoming a more pressing constraint since the spread in economy class of personalized flat screens for the in-flight entertainment, games, and services of all sorts that are nowadays making the concept of office-in-the-sky come true. Beyond the comfort issue, the economic success of a complete newly envisioned range of comfort items and customization concepts is at stake with cabin noise. Passenger comfort has become a challenging competitive selling point. In quantitative terms, EMB estimates that a reduction of 6 dB(A) in the ECS noise sources inside the cabin will lead to a decrease of 3 dB(A) in the overall cabin noise. The results obtained by LTS in the project do confirm these orders of magnitude, with Overall Sound Pressure Level reductions of about 9 dB measured at the inlet side of the ECS, and 2.5 dB reduction on the outlet side.
Beyond the aeronautical field, the tools and methods developed in the project will have very tangible impact on other fields such as ground transportation, where the noise emitted by train and automotive air-conditioning and cooling systems is a concern. Building ventilation systems are concerning a very large community as well, having a significant impact on the quality of the workspace, health and performance of the employees. In these various fields, the level of technological readiness of some modelling and mitigation strategies is somewhat lagging behind the level of expertise reached by the aeronautical industry. The presence of these non-aeronautical important societal actors within the Users Committee will greatly enhance the dissemination of mature tools and best practices developed within the project towards a wide industrial community affecting the life of many European citizens. One of the important outcomes of the IDEALVENT project will be providing the aeroacoustic simulation community with tools and recommendations on the best modelling practices for ventilation systems. The extensive and detailed experimental campaign that will be performed in WP2 on the one hand, and the advanced simulation tools that will be developed in WP3 and applied in WP4 on the other hand, will be employed to better understand the type and degree of interaction existing between the different components ventilation systems, at both aerodynamic and acoustic levels. Ventilation systems designers are eagerly expecting numerically efficient simulation strategies to reach optimized designs in a cost-effective and reliable way. This accumulated knowledge will be also very instrumental to devise control strategies in WP5, in order to manipulate the transient flow features and reduce noise production and transmission.
List of Websites:
https://www.idealvent.eu/
Reaching low levels of cabin and ramp noise is crucial to ensure the satisfaction of aircraft passengers and a safe working environment for the crew and personnel servicing the grounded aircraft. From the passenger and personnel standpoint, good thermal and acoustic comfort is important to minimize both stress symptoms and tiredness. Required to provide satisfactory air quality and temperature, the Environmental Control Systems (ECS) currently used are key contributors to the acoustic nuisance within the cabin and around the grounded aircraft. Reducing the amount of noise produced by the ECS has therefore a direct impact on the passenger satisfaction and personnel health and safety.
Unlike aircraft exterior noise, which has received considerable attention in past and currently running research projects, the noise emitted by confined flows in ECS assemblies involves complex mechanisms that haven’t been sufficiently investigated to permit the noise reduction wished by passengers and regulators. Acoustic and hydrodynamic interactions between subcomponents have so far been largely neglected despite being crucial, and are the focus of IDEALVENT.
In this project, thorough experimental studies as well as advanced analytical and numerical developments have been conducted in order provide a deeper understanding of these interaction mechanisms. Aerodynamic and acoustic installation effects were investigated by means of novel experimental techniques, highlighting the crucial importance of discriminating the noise emitted by a given ECS component (e.g. the blower unit) from the noise reflected by the remaining duct system (including its terminations in particular). Robust multi-modal decomposition techniques based on microphone and loudspeaker arrays were developed to this end. As a matter of fact, these modal decomposition techniques, applied as well to the numerical databases, proved essential to permit a meaningful cross-validation between measurements and simulations.
Integrated passive flow and noise control strategies have also been explored in this project, on laboratory test benches in a first stage, then on a full-scale industrial equipment in a second stage. The knowledge gained in the experimental and numerical investigations of the installation effects has permitted devising optimized strategies having the best potential for reducing ECS noise.
Based on the above considerations, it can be stated with confidence that the IDEALVENT project has been very successful, delivering nearly all the planned scientific and technological results, and opening unexpected new paths for better testing and modelling of complex ducted systems. The numerous results of the project are expected to find many fields of application beyond the initially intended one, e.g. for aircraft propulsion, or heating, ventilation and cooling systems used in surface transports and buildings.
Project Context and Objectives:
While a considerable number of European research projects conducted in the past and current Framework Programmes have addressed the exterior noise issues posed by aircrafts during take-off and landing, insufficient effort has been applied to understand and reduce the aerodynamic noise generated by the aircraft internal cooling and ventilation systems (grouped below under the acronym ECS for Environmental Control System). It has been acknowledged that the physical mechanisms involved in confined flow systems can be much more intricate than for exterior noise problems, e.g. because the separation between flow and acoustic fields becomes more arbitrary and less amenable to classical modelling approaches. Also, the intrinsic complexity of confined turbulent flows raises very specific challenges related to near-wall modelling and the prescription of adequate boundary conditions. Moreover, many interactions take place between sub-components of a given ECS: the blower unit, temperature and pressure regulating devices or simply duct connections. For these reasons, it proves difficult to devise the most promising noise reduction strategies and achieve the objectives of the Work Programme.
However, ECS components are strong contributors to the noise perceived within the cockpit and cabin, and to the acoustic environment of the personnel servicing the aircraft when grounded. Clearly, some research is needed to understand in detail the noise generating mechanisms within interacting ECS components, propose better design and sound mitigation guidelines, and eventually fulfil the competitive and societal objectives of aircraft manufacturers and European Communities altogether. The present project will address these important issues, in order to lead to an effective reduction of the cabin and ramp noise, and improve substantially both passenger comfort and airport personnel health and occupational safety.
Parts of the ventilation and cooling systems are installed within the aircraft belly-fairing casing, in order to ensure aircraft pressurization, provide temperature controlled air in the cabin and in the cockpit, and maintain a temperature compatible with the equipment present in the belly-fairing. In these systems, ram air is taken through ventilation inlets, used in heat exchangers and ejected below the aircraft belly through outlet vents. Together with the APU, these systems contribute significantly to the ramp noise (also called apron noise) perceived by personnel servicing the grounded aircraft. In the aeronautical field as in any other field of activity, occupational exposure to noise is one of the most significant health risks for workers, leading in some cases to irreversible hearing damage. Other adverse effects include hypertension, ischemic heart disease, annoyance, sleep disturbance and loss of concentration.
From the legislative viewpoint, ramp noise is dealt with by ICAO standards and regulations that have to be satisfied by aircrafts for certification. The European directive 2003/10 sets limits to the workers daily exposure which have a significant impact on the required workers protection equipment and on the servicing of a grounded aircraft if the noise levels exceed certain thresholds. Averaged permissible levels are established on an 8-hours daily working basis (Lex,8h), but at any instant a peak value Lpeak of 140 dB(C) cannot be exceeded. In no case the daily averaged exposure of a worker may exceed the limit of Lex,8h = 87 dB(A). These limits include the attenuation of any hearing protection equipment given to workers. When the daily equivalent level is over 80 dB(A) and/or the peak level exceeds 135 dB(C), the following measures have to be implemented: each worker must be provided with a proper information, and when applicable, a proper training regarding noise exposure, associated health effects and the use of protective equipment. Also, medical hearing checkups must be provided to workers asking for them. When the daily equivalent level is equal to or exceeds 85 dB(A) and/or the peak level is above 137 dB(C), technical and organizational measures must be implemented by the employer to reduce noise exposure. Areas where workers may be exposed to important noise levels must be indicated by appropriate signs and access to these areas has to be restricted whenever technically possible. This requirement can itself be contradictory to safety interests, as escape routes from an accident have to be as direct as possible.
Recent studies demonstrated that only a small minority of workers were exposed to daily-averaged levels below the limit of 85 dB(A). For some workers the level even exceeded the absolute permissible value of 87 dB(A). Apron noise is thus a very significant current problem that has to be tackled very seriously by ECS and aircraft manufacturers.
One of the issues with ECS design is that the various requirements that apply thereon including weight, space and noise, yield contradictory engineering solutions. The mass flow rate of ventilation air is usually provided by axial machines, which are intrinsically noisy as a result of the fluctuating forces exerted by the blades, yielding both tonal and broadband contributions to the sound spectrum. Because of space requirements, these machines must be as compact as possible, with the result that the rotation speed cannot be too low to provide the adequate flow rate. But noise power is typically proportional to a large power (5-6) of the blade Mach number. Moreover, the large fan speed combined with the large number of blades yields quite high frequencies for the tonal sound spectrum emitted by the blower, which will induce cut-on transversal modes contributing to the sound transmitted by the duct. These acoustic considerations must be included in the early stages of the design process to yield low-noise blowing units.
A quite generic configuration can be defined, surprisingly, for both cabin noise and ramp noise. A good example is the blower, which constitutes the main noise source for the most common ECS ramp noise applications, when using a pneumatic air conditioning pack, or when using a vapour cycle system for cooling purpose. The vapour cycle cooling system uses physical properties of a specific refrigerant fluid to cool fresh air (or re-circulated air), which is then distributed inside the cockpit or the cabin through complex duct systems. The blower will be predominant for recirculation noise, both in cockpit and cabin. While in the first example of ramp noise, the fan blade speed is almost supersonic at fan tip, resulting in quite high frequency noise, the two other application share a common frequency range and mass flow rate.
The blowing unit is a major contributor to ECS noise, but is not the only one. In addition, because of the large extent of an aircraft ventilation system and of tight space constraints, the layout of an aircraft ventilation network is quite complex, and includes many singularities such as valves, side-branches, manifolds and obstructions. Valves are used to control the flow rate at various locations of the piping system, which induce locally large velocities (in some cases choked flow) and hence strong aerodynamic noise. Diaphragms need to be placed in some ducts to provide an adequate pressure drop and a balanced flow rate across the network. Such singularities are also prone to strong noise generation. Flow noise is also produced by bends, which are inevitable in guiding the flow from the main blowing unit towards the cabin ventilation outlets and under-belly vent outlets. These components have to be included in the picture if one aims at reaching an acceptable sound pressure level perceived by the passenger/pilot or ground operator.
In order to meet certification requirements, it is thus quite important for an ECS manufacturer or integrator to be able to predict with a sufficient accuracy, and reduce where needed, the noise levels emitted by a given system. The prediction of aerodynamic noise, emitted by an individual valve for example, is not an easy task in itself, but is nowadays at reach with modern approaches and computational capabilities.
Since the advent of the acoustic analogy, significant progress has been achieved towards the understanding of the mechanisms through which sound is produced by an unsteady flow. The maturation of Computational Fluid Dynamics (CFD), permitting the modelling of the unsteady features of the flow field in particular, has paved the way to truly quantitative aeroacoustic simulations of individual components.
However, two key ingredients are still missing to provide a reliable ECS acoustic prediction model able to drive the design of quieter systems. Firstly, the interactions between subcomponents are crucial. Many of the studies carried out on individual components of an ECS system have indeed demonstrated a quite large sensitivity of the unsteady flow field, and thereby sound generation, with respect to their environment, expressed in terms of boundary conditions when considering modelling approaches. For example, the noise produced by the blowing unit is strongly dependant, both in spectral shape (tonal/broadband) and amplitude, on the inflow mean distortion and turbulence level. The same holds for singularities such as vanes, diaphragms and bends. For system designers, it is quite important to develop a system model assembly, composed of various duct component description and acoustic prediction laws. To enable this model development, sub-components need to be thoroughly tested and modelled, alone and coupled with their most recurring parts, defined as generic for a range of system distributions. To the present consortium’s knowledge, the multi-components interactions of the ECS have not been addressed previously, and constituted an important objective of this project.
Secondly, the numerical implementation of the acoustic analogy, based on unsteady CFD modelling (Large Eddy Simulation, Detached Eddy Simulation), still presents a prohibitive cost if design or optimization are targeted. Low-order modelling approaches, such as stochastic methods, or statistical approaches present much lower computational costs and give quite satisfactory results as long as the statistical models that are required as input are properly tuned. The combination of scale-resolved methods, based on time-resolved CFD, with statistical/stochastic methods is another important objective of this project, in order to permit studying and optimizing interactions effects with the current computational resources.
Simulating the ECS aero-acoustic performance is key to successful design, but the ultimate purpose is the reduction of the acoustic levels within the weight, space and efficiency constraints imposed by the aircraft manufacturer. In the aircraft industry, passive flow control strategies, which do not introduce additional energy in the controlled system, offer an interesting path that will be pursued in this project. The knowledge gained in the experimental and numerical investigation of the multi-components interactions is very instrumental in this respect. Indeed, having a better understanding of the mutual interactions causing additional tonal or broadband noise generation provides guidance about the flow manipulations that are the most appropriate to reduce the noise production. In this project, the costs and benefits of integrated passive flow and acoustic control approaches must be investigated by experimental as well as numerical means.
These passive approaches have been investigated for some time and with success in order to reduce the noise emitted by aeroengines, but require further optimization for the particular flow and frequency regimes encountered in ECS ducts. These innovative noise control approaches are to be first assessed on simplified (but representative) configurations, and on a production system in a second stage.
Project Results:
As a reminder, the research was articulated along a total of six work packages:
• WP1 coordinates the different strands of work in IDEALVENT.
• WP2 is focused on the experimental characterization of flow-acoustic and multi-component interactions. The purpose was to identify configurations relevant to the industrial application, and to provide exhaustive experimental databases permitting to i) understand the interactions mechanisms, ii) guide the methods developments in WP3, and iii) validate their application in WP4. In Task 2.1 the consortium has detailed the geometrical dimensions and range of flow regimes to be investigated in Task 2.2. In Task 2.2 a preliminary fast acoustic survey has been conducted on the configurations defined in Task 2.1 in order to identify relevant flow conditions that were investigated experimentally in greater detail in subsequent Tasks of WP2 and simulated in WP4. The relevant configurations were decided on the basis of the measured acoustic spectra compared with existing datasets made available by the final users. In Task 2.3 detailed flow and acoustic measurements have been performed for the valve-bend configuration. In Task 2.4 detailed flow and acoustic measurements have been performed for the fan with inflow distortion case. In Task 2.5 detailed flow and acoustic measurements have been performed for a tandem diaphragms case. In Task 2.6 the final user Embraer did perform acoustic measurements of one of its aircrafts grounded in its apron environment, in order to quantify the respective contributions of the ECS and of the APU at various locations around the aircraft.
• WP3 deals with the development of innovative simulation methodologies accounting for the interaction effects. The objective was to define the best modelling strategies for scale-resolved and statistical/stochastic approaches. In Task 3.1 the global specifications and ranges of applications of the different simulation approaches were detailed. In Task 3.2 the boundary conditions and interfaces required by the simulation tools were established. In Task 3.3 innovative scale-resolved and statistical approaches have been developed and improved for fan noise prediction. In Task 3.4 innovative scale-resolved and statistical approaches have been developed and improved for ducted turbulence noise prediction. In Task 3.5 multi-modal network models have been derived for the extraction of the active sound component from measurement or simulation datasets.
• WP4 applied the tools developed in WP3. The butterfly valve with upstream bend, fan with inlet distortion, and tandem diaphragms were investigated in Tasks 4.1 4.2 and 4.3 respectively.
• WP5 introduced noise mitigation strategies through passive flow and acoustic control. Porous liners and micro-perforates were investigated in Task 5.1. Turbulence screens were tested in Task 5.2.
Finally, WP6 integrated the outcomes of the previous WPs by providing guidelines on the integration of the sub-components of the ventilation system, and demonstrated their effectiveness in an industrial environment. The most promising control approaches tested in WP5 on the laboratory configurations were then tested on full-scale industrial equipment provided by Liebherr.
A summary of the most prominent results obtained in each of the WPs 2 to 6 is detailed below.
The WP2 final restricted and public reports (deliverables D2.7 and D2.8) discuss in length all the developments carried out in this WP, which can be summarized as follows:
• development of the fan test rig at VKI and of the valve test rig at KUL;
• preliminary flow and acoustic measurements on each test rig to verify that the correct operating conditions can be achieved and for the identification of relevant flow-acoustic installation effects;
• for the fan configuration, a theoretical analysis of the dominant sound generation mechanisms by application of analytical tools;
• development of an active multi-port sound extraction technique, based on the measurement of the modal reflection coefficients of the duct terminations and of the modal scattering matrices of the fan / valve components;
• quantification of the installation effects by application of the above-mentioned technique.
In order to investigate in WP2 the aerodynamic sound generated by an ECS fan (provided by Liebherr Aerospace) subjected to installation effects, a modular test-rig has been built at VKI. The modular structure allows investigating the noise generated by the axial fan, including both upstream and downstream installation effects. The loudspeaker modules are manufactured from aluminum by KTH in order to carry the loudspeakers which are used in the modal decomposition analysis. Upstream and downstream ducted microphone modules have been designed and manufactured by KTH as well, located upstream and downstream of the fan and diaphragm assembly. The in-duct microphone sections are used to measure the reflection coefficient downstream of the duct and for the modal decomposition measurements on the same rig. The modular structure of the test rig allows investigating installation effects through different configurations including empty duct, fan only, diaphragm only and fan-diaphragm configurations. The test rig also contains an auxiliary fan located outside of the anechoic room in order to generate flow for configurations in absence of the ECS fan. Besides the ECS fan mock-up of the fan with inflow and outlet distortion, investigated experimentally at VKI, also a mock-up of an ECS valve, provided by LTS, with inflow distortion is analyzed in the aeroacoustic test facility of KUL. Loudspeaker and microphone modules similar to those used in the VKI fan rig were designed by KTH for the multi-port eduction of the valve noise subjected to various inflow conditions.
Three main axes were followed in WP3 for the development of prediction methods adapted to the IDEALVENT configurations, which can be ranked by increasing computational cost (and presumably accuracy) as follows: statistical, stochastic and scale-resolved methods. The developments carried out thereto by the different partners are summarized in what follows.
To produce tractable solutions with the analytical formulations, the ECS fan stator vanes need to be simplified to rigid flat plates of zero thickness embedded in a uniform mean flow. The simplest strategy is to determine the fluctuating lift forces on the vanes with linearized isolated-airfoil theories and then to make the forces radiate. This two-step approach neglects the so-called cascade effect, which is defined as the effect of adjacent vanes on the aerodynamic and/or acoustic response of one of them. Because of the large overlap and the large solidity in the stator, the cascade effect is very important. The need for cascade response functions to replace simple Amiet’s theory has been recognized and motivated the development of ad hoc mathematical formulations in the literature. However, the main drawback of their formulation is that it cannot be extended for a three-dimensional annular cascade. In contrast, the mode-matching technique can overcome this limitation.
For the considered fan case, ECL has established that wake turbulence impinging on the stator vanes appears as the dominant source of broadband noise to be modelled in priority using a dedicated analytical model. Furthermore, the periodic (averaged) part of the wake velocity deficit is also responsible for the generation of tonal noise by impingement on the vanes. This makes wake interaction the dominant mechanism for both broadband and tonal contributions, for which the equivalent sources are distributed on the stator vanes. As a result, the rotor is just considered as a generator of vortical disturbances. The stator is the only part to be modeled geometrically. Independently of wake interactions the rotor also contributes to the tonal noise by two mechanisms, namely its interaction with the stationary inflow distortions (strongly depending on the inlet geometry) and its operation through the upstream potential field of the stator. Such mechanisms have been addressed independently, by means of a mode-matching technique developed by ECL in this project.
An alternative approach, still based on linearized airfoil theories but relying more directly on numerical approaches for the determination of the turbulence statistics and for the prediction of scattering effects, was researched by VKI for the prediction of the IDEALVENT fan noise. An automatic data extraction procedure from RANS CFD computations was developed. This has been implemented in a general framework and integrated in BATMAN (Broadband And Tonal Models for Airfoil Noise), the VKI in-house code implementing noise prediction methods based on Amiet’s theory for leading and trailing edge broadband noise and tonal interaction noise. A special reader module based on mexcgns matlab libraries has been implemented to interface CGNS TRACE CFD results from DLR with VKI post-processing routines. The VKI routines being based on strip theory, all the post-processing is performed at iso-radii, cuts being automatically performed on the blade surface, and the rotor and stator are post-processed separately. The extraction of data is used as inputs for the SISW methods using a combination of Amiet’s theory with the SISW Acoustic Transfer Vectors technique and the ECL method based on mode-matching technique.
With regard to stochastic approaches, DLR has further improved its CAA methodologies as needed to deal with the different duct and fan-noise problems. With respect to fan noise, DLR uses a hybrid strategy based on CFD tools optimized for aero-engine applications.
For fan noise, the hybrid approach involves three numerical tools:
• the URANS solver TRACE,
• the CAA solver PIANO,
• the turbulence synthesizer FRPM.
A URANS or Harmonic Balance (HB) computation of the fan-stage gives the fluctuation fields that are necessary to perform tonal noise prediction. A CAA computation is then carried out with PIANO, using the mean-flow solution given by TRACE as background flow for the perturbed equations. The turbulence-induced noise sources are generated in time domain by a turbulence generator - the so-called fast Random Particle Mesh (FRPM) method - which uses the turbulence statistics from URANS as input. This gives the broadband fan noise component. Three main developments were conducted in IDEALVENT, for the application of the method to rotor-stator interaction noise:
1. accounting for the mean flow periodic variation;
2. accounting for the so-called cyclo-stationarity of the turbulence properties;
3. accounting for the fact that the integral length scale and turbulent kinetic energy in the wake of the rotor differ from the region outside.
The cyclo-stationary (or periodic components) of the flow are extracted from a URANS or HB computation and prescribed as complex Fourier coefficients to the PIANO and FRPM computations.
For the prediction of the noise produced by ducted singularities such as valves, a hybrid prediction chain similar to that used for fan-noise predictions is applied by DLR. For the prediction of the stationary turbulent flow problem inside the duct RANS predictions are based on the unstructured DLR flow solver TAU. To summarize, the prediction chain involves the tools:
• the RANS solver TAU,
• the CAA solver PIANO (structured) / DISCO++ (unstructured),
• the turbulence synthesizer FRPM.
The developments carried in IDEALVENT in relation with these tools include:
1. the efficient extension of the Gaussian based FRPM method to more complex turbulence spectra;
2. the introduction of anisotropy into the modeling of stochastic fluctuations by means of an anisotropic scaling tensor that replaces the scalar scaling function;
3. the introduction of a vorticity mode selective relaxation source term (Eddy relaxation source term) in PIANO;
4. the coupling of FRPM with the Discontinuous Galerkin solver of DLR (DISCO++) to enable broadband noise prediction for complex geometries otherwise not resolvable by structured meshes. (applied in WP 4 to the butterfly-valve test-case).
In contrast with the approaches described above, the deterministic methods developed in this project are solving time- or frequency-dependent equations in a deterministic way, to compute either the noise produced by a given ducted element (fan or valve), or its sound scattering properties.
The major objectives of NTS consisted in the development of improved scale-resolved techniques for prediction of the ducted fan noise and of the noise produced by a stationary obstacle (e.g. diaphragm) inside the ECS duct.
The approaches developed by NTS for solving these two problems are essentially the same. They consist in Direct Noise Computation (DNC) based on compressible scale-resolving simulation of turbulence with the use of a hybrid RANS-LES method of the DES-type (Improved Detached Eddy Simulation – IDDES) combined with synthetic turbulence generator for “creation” of realistic turbulent content upstream of the noise-generating region (fan or diaphragm and the area with high level of turbulent fluctuations downstream of them). In principle, such an approach is capable of predicting both the tonal and the broadband components of the noise produced within the source area and its propagation over the ventilation duct and in the surrounding ambient air (of course, within a range of frequencies accurately resolved by the numerical method on the grid used in the computations). However, such computations require huge (currently non-affordable) computational resources. For this reason, the computational domain in which the turbulence-resolving simulations of the flow are performed has to be limited to a relatively small part of the duct neighboring the source region. This has two major consequences.
Firstly, in addition to IDDES computation of the flow in the limited computational domain, the procedure must involve auxiliary stages necessary for setting inflow/outflow boundary conditions. Secondly, a problem of extraction of acoustic signal from the simulations carried out with the use of the developed approach is far from trivial because in the vicinity of the source region (at least downstream of it) the acoustic signal turns out to be almost indistinguishable against much stronger pressure fluctuations caused by turbulence and, in particular, by impinging of turbulent structures on the duct walls. Therefore, in order to extract the acoustic signal, some supplementary acoustic techniques should be used, for which the detailed unsteady information obtained in the course of the scale-resolving IDDES of the flow serves as an input. Such techniques were developed by IDEALVENT partners (SISW, KTH, VKI, DLR).
All the elements of the outlined computational approach were implemented in the NTS in-house general purpose code for IDDES based on the two-equation k-ω SST background RANS turbulence model and (for the fan flow) on the same model with streamline curvature/rotation correction (SST-CC model). This is a structured finite-volume high-order CFD/CAA code accepting multi-block overlapping grids of Chimera type. It has been intensively used for a wide range of aerodynamic and aeroacoustic problems and shown to be rather reliable and accurate. The code was adapted to the computation of the flows with rotating elements with the use of the approach based on employing different reference frames in different grid blocks combined with the sliding interfaces technique for imposing the boundary conditions on the interfaces between rotating and stationary grid blocks.
Major new developments made in the course of this work, which ensured its overall success, are as follows.
1. Volumetric Synthetic Turbulence Generator (VSTG hereafter) is developed, which is based on introducing distributed volume sources (“body force”) into the momentum and turbulent kinetic energy transport equations.
2. A modification of the original IDDES method is proposed, which ensures significant weakening of the well-known flaw of DES, namely, a severe delay of the Kelvin-Helmholtz instability and transition from fully modeled to mostly resolved turbulence in the free and separated shear layers (RANS-to-LES transition).
Still in the framework of scale-resolved simulations, important developments were carried out by SISW for the coupling of time-resolved CFD solutions with their acoustic solver Virtual.Lab Acoustics. Three main axes of research were explored:
1. aeroacoustic source modelling for internal flows, by extending a method to express dipolar sources as Neumann-type boundary conditions in an acoustic boundary value problem;
2. statistical approaches for computing the scattering of broadband fan noise, combining an analytical statistical description of the incident field with a technology based on a boundary element method to compute scattering problems;
3. advanced high-order finite element solvers for duct acoustics including flow effects.
The first approach consists in transforming Curle’s dipoles into Neumann boundary conditions of the Helmholtz equation. This is done by evaluating numerically a (hypersingular) integral involving the flow pressure and an appropriate Kernel function. This new approach was originally mainly applied to exterior problems, such as airframe noise problems. An initial investigation on a basic interior flow (a simplified HVAC duct) prior to the start of this project, proved that the application to internal flows with complex geometries presents additional challenges which require a dedicated study, such as the treatment of free edges and junctions. Specific aspects related to the application of this approach to ducted flows have been investigated, such as strategies to minimize numerical leakage in interior/exterior problems, as well as the impact of the length of the source region. Furthermore, a new formulation for the Neumann boundary condition in the presence of acoustic impedance has been derived. Finally, the algorithm to convert dipoles into boundary conditions has been improved for cases dealing with incompressible CFD input with multiple frequencies running in parallel in several processors, as well as for cases with multiple CFD time windows for averaging. The latter is necessary for averaging the results of turbulent flow computations, which are chaotic by nature and typically require several realizations to converge to the experimental solution.
The second aspect developed by SISW related to the computation of broadband fan noise scattering for industry-standard applications. The approach combines an analytical formulation based on Amiet’s theory with the so-called Acoustic Transfer Vectors (ATV) approach. The acoustic transfer functions are computed through boundary element methods, based on the acoustic pressure and acoustic velocity Power Spectrum Density (PSD) of the broadband fan source. The approach is general and valid for any stochastic source characterized by their radiation PSD. This approach allows taking into account impedance values (to characterize for instance absorbing surfaces or micro-perforated materials) in a straightforward manner. As a validation case, the sound radiated by a fan inside a cooling unit was computed. The initial prototype was previously used to compute this case. The optimized implementation achieves a total speed-up factor of around 450 (run sequentially). The parallel implementation offers potential further gains in computation speed. This can reduce the typical computation cost for large case from several days to less than an hour, thus making it possible to tackle cases of industrial size in general, and, in particular, the fan problems to be computed in the WP4 of IDEALVENT.
The last subject of developments by SISW addresses its finite element solvers for duct acoustics, in order to achieve more accurate and efficient computations of the IDEALVENT configurations. The use of an axisymmetric implementation of the problem as a means to reduce the cost of three-dimensional computations was investigated. In particular, an axisymmetric approach to solving the linearized Euler equations (LEE) was evaluated using a frequency domain high-order FEM. The application of this model to the Munt problem, as a test case, showed results in good agreement with the reference solution. The acoustic duct modes are injected through a Perfectly Matched Layer (PML) and radiated outside the duct, where a PML is also applied in order to absorb the outgoing waves. In addition, the three-dimensional finite element solvers were optimized by means of a high-order adaptive finite element implementation, which had already shown an improved computational performance for the Helmholtz equation, including potential mean flow effects.
The determination of the scattering properties of ducted components is important for the characterization – through experiments or compressible flow simulations, of acoustic installation effects and of the active noise produced by the element under consideration. In that line, KTH extended their existing two-dimensional frequency-domain Linearized Navier-Stokes (LNS) solver, including the various boundary condition formulations and the interfacing schemes to interpolate the mean flow variables from the fine RANS (Reynolds Averaged Navier-Stokes) grid to the coarse acoustic mesh, to a full HPC-running three-dimensional frequency-domain LNS solver. This solver includes an additional source term contribution from the mean flow field solution to better model the viscous dissipation caused by the mean flow field variables. The updated formulation can be directly linked to multi-port network models, being solved in the frequency domain. This means one first computes the passive scattering matrix of a component by simulating the reflection and transmission using the LNS approach for both the up- and downstream side. This must be done for each propagating mode in the up- and downstream ducts to build the complete scattering matrix. Once the scattering matrix is known it can be combined with the up- and downstream boundary conditions also expressed as multi-ports.
Linearized Navier-Stokes and Linearized Euler Equation solvers were also topics of research by the KUL team. They developed in the past an in-house time-domain discontinuous Galerkin (DG) solver, which is parallelized using MPI. The solver uses a 3D quadrature-free implementation in the bulk of the domain while a local quadratic implementation in the vicinity of highly curved walls, significantly reduces the number of elements while maintaining the same accuracy. An optimized RK-time marching scheme further reduces the dispersion and dissipation errors and increases the maximum allowable time-step. As governing equations the Linearized Euler Equations (LEE), the Irrotational Linearized Euler Equations (ILEE) as well as the Linearized Navier-Stokes (LNS) equations can be used, dependent on the governing physics of the influence of the mean flow field onto the propagation of the acoustic waves. For the coupling between CFD output and their DG solver, KUL implemented a flexible interfacing scheme for the mapping between the CFD and the acoustic propagation meshes is developed. The adopted approach uses a global least square procedure for continuous mapping in combination with an additional anti-aliasing filtering.
KUL developed as well an efficient time domain impedance and transfer-admittance formulation. By writing the transfer-admittance matrix coefficients as a sum of first- and second-order systems, a recursive formulation can be obtained for explicit non-uniform multi-stage time integration schemes with constant time step. This technique was implemented into the in-house RKDG solver for the time-domain LEE and LNSE. The formulation was successfully verified for simple benchmark cases (both in two and three dimensions), without mean flow and with simplified grazing flow profiles for which the Ingard-Myers condition (unstable in time-domain) was not assumed and the boundary layer profile was fully described.
The WP4 was dedicated to the application of the simulation technologies that were upgraded in WP3 to the three main flow categories: the industrial valve, the tandem diaphragm and the industrial fan. The simulation approaches span across a range of CPU cost vs. accuracy spanning from stochastic to scale-resolved approaches.
Simulations of the broadband sound generation by the butterfly valve have been conducted by DLR Braunschweig using the unstructured CAA code DISCO++ together with the FRPM module to reconstruct turbulent sound sources from an inexpensive precursor CFD simulation solving RANS equations. The results obtained permitted to demonstrate the general applicability of stochastic sound sources to realistic three-dimensional system components, even though their superior efficiency, compared with scale-resolved approaches, is less obvious than for cases that exhibit some homogeneity in one direction, permitting a locally two-dimensional treatment.
The active and passive properties of the ducted industrial valve were simulated using the panel of approaches developed in WP3. The passive properties of the valve were computed by KUL using their in-house, time-domain Runge-Kutta Discontinuous Galerkin (RKDG) discretization of the Linearized Euler Equations (LEE). Linear elements, allowing the use of an efficient quadrature-free DG formulation, are used throughout the domain except in the vicinity of the curved walls of the circular duct where curved second-order elements are locally used locally around complex geometries. This mixed approach results in an optimal balance between the computational efficiency of the quadrature-free method and the geometrical flexibility of the second-order elements. Due to the small mesh size in the vicinity of the valve the use of quadratic elements in this region is not needed. For the inter-element communication, a Lax-Friedrich formulation is used for the convective fluxes. Boundary conditions are imposed by specifying the fluxes at the boundary faces. The time integration is done using a Runge-Kutta scheme, optimized for the spatial discontinuous Galerkin operator. An excellent agreement was found between the experimental and numerical results for all modes of the valve scattering matrix, except at frequencies slightly above the cut-on frequency where the numerical results show, similar as for the circular duct results, an increased transmission and a decreased reflection.
For the deterministic calculation of the active noise produced by the valve, multiple efforts were combined by several consortium partners. The CAD of the valve is made available by LTS. RANS computations have been performed by VKI using rhoSimpleFoam, the compressible steady solver of OpenFoam 2.4.0. The k-ω SST turbulence model is used with second order accurate schemes for all variables. Air is used as fluid with perfect gas properties and the Sutherland model is used for the variation of viscosity with compressible effects. Other properties are kept constant. Mass flow rate is used as inlet condition. As the corresponding value is not directly available from the experiments performed by KUL, but only the pressure drop and the corresponding volume flow rate, computations are performed in a range of inlet mass flow rate. The operating conditions corresponding to this specific pressure drop are then determined through the operating curve results.
VKI ran the unsteady computations using the SAS turbulence model of the CFD solver Fluent 16.2. It allows the resolution of the turbulence spectrum in unstable regions, resulting in a LES-like behavior in unsteady regions of the flow field. At the same time, it provides standard RANS capabilities in stable flow regions. As for the RANS computations, second order schemes are used for all variables and central bounded second order schemes are used in time. Same fluid properties are used as in the RANS computations. Furthermore, general Non-Reflecting Boundary Conditions are used to limit possible reflections at the inlet boundary conditions. At the outlet, a pressure far-field condition is applied that is a non-reflecting boundary condition based on the introduction of Riemann invariants for a one-dimensional flow normal to the boundary. Wall-pressure fluctuations (dipoles) and volumetric velocity data (quadrupoles) are extracted in the complete CFD domain, to be used as input to acoustic computations performed by SISW. These simulations were conducted solving a linearized acoustic propagation equation in potential flow, with volume-distributed sources based on vortex sound theory. The computational results show a difference between upstream and downstream acoustic levels that is similar to that of experiments, and the overall behavior seems to follow the same trends.
For the diaphragm simulations, two tandem configurations were considered with the separation distances between the diaphragms equal to 2 and 4 duct diameters. In addition, a baseline case of a single ducted diaphragm was analyzed. Overall, the NTS predictions of the wall-pressure spectra are in good qualitative and acceptable quantitative agreement with experiment. The main discrepancy resides in the low frequency end of the spectra of the upstream propagating noise and is somewhat more pronounced for the single diaphragm case. A cause of this discrepancy is not quite clear so far, but most likely it is associated with noticeable reflections of the low frequency sound waves by the duct inlet in the experiment. Particularly, because of these reflections, the presented comparison of the “raw” spectra from the simulations and experiment cannot be considered as fully consistent. A more consistent comparison was provided by KTH who has extracted from both the experimental and computational pressure signals the individual acoustic modes free of the impact of the turbulence impinging on the duct walls and inlet/outlet reflections of sound waves.
Using the compressible CFD data obtained by NTS, KTH applied its modal decomposition technique to extract the active noise produced by the tandem configuration, for several axial separations. The decomposition of numerical data brings a number of advantages compared to the empirical approach. Admittedly, the computed time samples are still much shorter than data gained from measurements; however, a plethora of sensoring-points is accessible, even inside the channel cross-sections. The decomposition matrix is hence highly overdetermined and guarantees a stable matrix inversion. Furthermore, in computations we can disable all the disturbing effects which usually complicate measurements.
Hybrid aeroacoustic computations were run by SISW for the single and tandem diaphragm configurations. The sources are based on the compressible CFD data provided by NTS. Since the CFD is also able to provide the acoustic levels, these are used together with the experimental data as a reference solution to compare the results of the hybrid approach. In the cases with two diaphragms, the sound production is concentrated around the second diaphragm, due to the effect of the turbulent flow generated behind the first diaphragm impinging on the second diaphragm. As a consequence, the dipole sources on the downstream diaphragm are the dominant source of sound, and the sound production is less sensitive to the type of flow (laminar or turbulent) that is injected upstream of the first diaphragm. Therefore, dipole sources are defined only on the downstream diaphragm in the hybrid computations. The results were compared to experimental measurements obtained in WP2. The hybrid computational approach seems to provide an overall good match with the experimental data and with the direct CFD results, thus confirming that the dipole sources on the downstream diaphragm are the dominant source of sound. For each case, two different hybrid computations are carried out: one based on the CFD with clean inflow, and another based on the CFD with upstream turbulence injection. It can be observed that this yields some differences at low frequencies, where the flow with the turbulent inflow tends to produce more noise. By contrast, the flow with the clean inflow seems to produce somewhat more high-frequency noise.
The modelling of the noise emitted by the industrial fan was tackled from different approaches: (U)RANS, DES, hybrid methods, using deterministic, stochastic and statistical methods.
Unsteady RANS simulations were performed by DLR using the Harmonic Balance method. The tonal noise is extracted from the URANS simulation by using the DLR in-house code Connect3D, which performs an extended Triple-Plane-Pressure (XTPP) mode matching to extract the amplitude and phase of the acoustic propagating modes.
For the DLR broadband simulations, the cascade is realized with all 10 stator vanes and a periodic boundary condition in the azimuthal direction. In axial direction radiation and outflow condition are combined with a sponge region to minimize reflections. The turbulence is generated by the FRPM method and imposed to the CAA domain using the LEE-relaxation method at the inflow patch (blue). Only a single vane is excited assuming uncorrelated noise for adjacent vanes. The turbulence is eliminated by a second patch to reduce spurious noise in the downstream microphones.
The approaches developed by NTS have been applied in WP4 for the prediction of the effect of inlet flow distortions on the noise emitted by the industrial fan. In the first (baseline) case the duct has a smooth bell-mouth inlet resulting in a "clean" flow upstream of the fan with turbulence confined within a thin boundary layer on the duct wall. In the two other cases, a significant non-uniformity of the flow and elevated levels of turbulent fluctuations in the upstream part of the duct away from its walls are created by replacing the bell-mouth inlet by inlets with a rectangular-to-circular transition of the duct cross-section and with a T-junction formed by two circular inlet pipes.
Fan computations were as well carried out by SISW, this time to establish the passive behavior of the fan as a purely scattering element, without mean flow (Helmholtz equation). The results indicated that the first drop in transmission of the plane wave mode and in the first azimuthal mode is due to the cavity in the stator area, while the second has to do with the cavity in the front of the fan. If we observe the pressure field in the near-field of the fan, this becomes apparent. With respect to the experimental data, the computed results show similar qualitative trends. For instance, both experiments and computations predict the drops in transmission due to the cavities at rotor and stator sides at similar frequencies. Nevertheless, the experimental results display in general lower levels of transmission, and less pronounced peaks, thus suggesting the presence of acoustic damping mechanisms that are not modeled in the acoustic computations.
The acoustic mesh used by SISW for the computations of the active contribution (with aeroacoustic sources) of the fan cases is similar to the one used for the passive contribution, but the rotor surface is excluded from the mesh, leaving only the stator, where the sources are defined, in the duct mesh. The mean flow is assumed to be negligible in the acoustic computations.
It was observed that the predicted broadband levels are in very good agreement with the CFD input. This suggests that indeed the dipole sources on the stator vanes are sufficient to characterize the broadband noise produced by the fan. By contrast, the levels at the BPF are significantly higher in the current computations. Since the acoustic mode at the BPF should be cut off, this indicates that there are some inaccuracies in the acoustic results preventing the cancellation of the acoustic peak. It must be pointed out that the acoustic mesh and the CFD mesh were not exactly coincident and may have generated some inaccuracies in the mapping of the flow pressure from the CFD mesh to the acoustic mesh. This numerical effect may be worsened by the close proximity of the BPF to the cut-on frequency of the first radial mode. In any case, the agreement of the broadband levels is quite satisfactory and confirms the validity of the approach to predict broadband fan noise. It is worth noting that the approach is also applicable if the CFD is incompressible and thus unable to provide acoustic predictions.
On the analytical side, the implementation by ECL of the MMBW (Mode-Matching in Bifurcated Waveguides) approach has first been assessed in a two-dimensional context and validated by comparison with alternative mathematical formulations, typically the Wiener-Hopf technique and numerical simulations. This was achieved both in the framework of the IDEALVENT project and during the French program SEMAFOR, by addressing the two following generic problems:
• transmission of an oblique acoustic wave (equivalent to a duct mode in an unwrapped configuration);
• generation of acoustic waves due to a single impinging oblique vortical gust.
When addressing pure sound transmission, the predicted acoustic fields are in complete agreement with those of alternative methods provided that the Kutta condition is implemented, but de-activating this condition leads to significant errors (increasing with increasing Mach number). In absence of Kutta condition, the energy balance for acoustic waves is very accurate. With the Kutta condition activated, part of the energy of the acoustic motion is lost because of conversion into vortical motion. When addressing sound generation by impinging vorticity, again a full agreement is found between all approaches/methods if the Kutta condition is implemented.
Finally, in WP4 LTS tested the applicability of an empirical prediction approach, based on similarity laws and existing data for other similar fans.
The WP5 dealt with the exploration of mitigation approaches to reduce the noise produced and propagated in the Environmental Control System (ECS). A strong constraint that was taken into account concerns the overall efficiency and reliability of the control. Passive strategies, which do not inject additional energy into the system, were adopted for the sake of safety and energy consumption. The mitigation operates on two fronts: flow and acoustic control, using turbulence grids to dampen out the inflow inhomogeneity and turbulence, and porous liners/microperforates for acoustic absorption. Both experimental and numerical research have been carried out in WP5, though not systematically for each control technology. A specific effort was brought on developing the knowledge around the use of so-called micro-perforated plates (MPP’s) for noise control in ducts. A MPP is essentially a perforated plate, with holes that are so small so that the viscous losses dominate over the inertia of the air moving at each hole. The classical application for MPP’s have been in room acoustics as so called panel absorbers. More recently it has been suggested that MPP’s also have a large potential for other applications for instance involving noise in ducts. This was the motivation behind including MPP’s in the study of new and innovative passive noise control measures within the IDEALVENT project.
KUL investigated experimentally the properties of a set of modal filters with one MPP panel mounted in the mid-plane over the full length of the filter. Applying the multi-port eduction method developed in WP2, they measured the power loss resulting from the insertion of the modal filters subjected to prescribed incident modes.
When a flow is present through the modal filter, the micro-perforated panel is subject to a two-sided grazing flow. For micro-perforated panels with circular and slit-shaped perforations, the presence of a grazing flow on one side of the panel increases the resistance and slightly decreases the reactance. Similar effects can be expected for two-sided flow over a MPP with slit-shaped perforations.
KTH had already investigated the principle of using an expansion chamber muffler with a micro-perforated sheet tube as a dissipative silencer prior to this project. The damping behaviour of such a filter is constituted by both the impedance of the cavity and the impedance of the micro-perforated plate which covers the cavity. The optimum wall impedance for plane wave damping is chosen as the so-called Cremer impedance adapted to circular ducts and flow. The computed impedance could easily be realized by designing cavity structures fitting to an existing micro-perforated layer. A problem in practical applications is, however, that the optimal impedance changes with frequency and modes. A silencer is therefore only maximal effective for a small frequency band and a certain acoustic mode. For this reason, it has been attempted in this project to combine the performance of a MPP modal filter with that of a Cremer filter, in a so-called Cremer-Modal Filter (CM-Filter). Two different arrangements were first investigated numerically: a serial one, and a cladded one. The serial CM-filter had a higher total loss for the (0,0)-mode, which is reasonable due to a longer effective Cremer-impedance. The serial-connected CM-filter became more effective to the (1,0)-mode above 2700 Hz and for the (2,0) mode for all frequencies.
Modelling work was also conducted by KUL, yielding promising results considering the approximations made in the choice of governing equations and the rational impedance fit.
Likewise, SISW performed some research on computational modeling aspects of perforate plates. The task consisted in extending the potential flow acoustics propagation solver to account for transfer impedance boundary conditions in the presence of grazing flow. A transfer admittance model was implemented in the Linearized Potential Equation (LPE) code of Sysnoise. It represents an extension of the transfer admittance functionality for cases with a non-zero mean flow. The implementation considered the possibilities of matching and non-matching meshes on either sides of the MPP, and was validated against standard impedance models for conditions where both modelling strategies apply.
The two novel noise reducing modules, the MPP (Micro-Perforated-Plate) modal and the Cremer filters, have been tested on the VKI test-rig in similar conditions than those used in WP2. Both modules have been designed by KTH and built by SNT. The modal filter is designed to damp the two lowest circumferential modes. The Cremer filter is optimized to damp the plane wave mode at the fan BPF.
Three different configurations were tested. Measurements are first performed on an empty configuration, with an empty duct upstream of the fan and the bellmouth used at the inlet. Using the same bellmouth inlet, the filters are then inserted upstream of the fan and measurements are performed for two configurations, with the Cremer-modal filters or modal-Cremer filters, in order to verify their relative efficiency depending on their relative position to the fan. The results have demonstrated the high efficiency of the serial Cremer-Modal filter unit. It both reaches the target of more than 10 dB damping at the fan BPF but also delivers a broad band damping. This is due both to the action of the modal filter and the Cremer silencer. Because the Cremer silencer can although it is tuned to a certain frequency be quite effective in a band around this frequency. It can also be noted although the Cremer silencer is optimized for the plane wave, it is still very or often more efficient for the higher order mode components at the BPF. However, the drawback is that is quite reflective for the non-optimum modes. This is compensated for by combining the Cremer silencer with the upstream modal filter which gives a broad-band damping to circumferential modes up to order 2.
Similar experimental investigations were performed by KUL on their test rig for the valve configuration. The results weren’t as promising as for the fan configuration, which can be explained by the fact that most of the aerodynamic sources excites the plane wave mode (0,0), while the modal filter is only damping noise of the energy which is contained in the azimuthal modes. As such, for the butterfly valve application the presence of a modal filter does not yield any significant noise mitigation.
On the basis of all the results obtained during the project, LTS drew a synthesis report summarizing the main project outcomes, with a focus on the quantification of the installation effects and potential of the mitigation strategies. In a nutshell the main conclusions can be summarized as follows:
• For the fan component:
o The diaphragm at the exhaust of a fan can have an important impact on noise radiated by the system:
▪ If the diaphragm doesn’t have a high pressure drop, the fan works on a similar operating point, in this case, the impact of the diaphragm on the fan noise is 2-3 dB.
▪ If the diaphragm induces a high pressure drop, the noise produce by the fan-diaphragm association can be very different: the noise is increased by about 15 dB. This change is not due to the diaphragm but to the fan which is near a rotating stall operating point and generates high noise levels.
o Inlet fan modules create inlet distortions that can add high noise level, especially at low frequencies.
These interactions between fan, diaphragms, and upstream distortions must be analyzed during the design of air distribution. The acoustic studies of isolated components are not sufficient to have a good noise prediction of the global system.
• For the valve component:
o The aerodynamic noise production of the butterfly valve is dominated by the static pressure drop over the valve.
o In the experimental set-up the valve is responsible for >90% of the total pressure drop.
o The far-field radiation becomes efficient above the cut-off frequency.
o No installation effects (presence of the bend, distance bend- valve, turbulence screens) are noticeable.
o It is possible to simplify the butterfly geometry as a flat plate to have similar acoustic behavior as the complex valve geometry.
The only mitigation guideline on this particular noise source amounts therefore to reducing the pressure drop over the valve.
The final activity of IDEALVENT, conducted by the final user LTS, consisted in integrating the outcomes of the project by testing the most promising noise mitigation strategies in one of its industrial systems. The main components of the system include:
• an inlet part manufactured by 3D Additive Layer Manufacturing,
• a condenser exchanger,
• a butterfly valve,
• a fan, identical to the one investigates in the VKI test rig.
To assess the effect of the different noise reduction units, the following configurations have been tested at inlet and outlet of the system:
• baseline (i.e. without any reduction noise device);
• splitter upstream the fan;
• Cremer upstream the fan;
• splitter and Cremer in serial (splitter upstream);
• Cremer and splitter in serial (Cremer upstream).
While a full accounting of the results of this test campaign is provided in the WP6 final deliverable, the main results can be summarized as follows:
• For the sound propagated towards the system inlet:
o The three configurations including the Cremer unit showed a good attenuation, in particular around the first BPF of the fan, where the noise reduction reaches 16dB (for the Cremer and splitter configuration).
o The Cremer unit (and Cremer + Splitter) shows an interesting broadband attenuation. The splitter and the Cremer units display 4dB noise reduction in high frequency above 4000Hz.
o The Cremer and Splitter configuration gives the best results with the combination of both effect noticed, with 9dB noise reduction on global noise level.
• For the sound propagated towards the system outlet:
o The configurations with the Splitter unit do not show a significant reduction on the measured noise at outlet of the system. However, the Cremer and Cremer + Splitter configurations give a slight attenuation on a wide frequency band.
o The best noise reduction result is obtained with the combination of Cremer and splitter units leading to 2.5 dB of noise reduction on global level.
To conclude, the IDEALVENT project has laid down an essential milestone in the investigation of the aerodynamic noise generated by Environmental Control Systems. Unprecedented efforts have been devoted to its experimental investigation, its prediction through a wide range of modelling approaches, and its mitigation by innovative passive control strategies. The project achievements are meant to bring rapid and long-term benefits for the design of better systems and to substantially improve passenger comfort and occupational health on the ramp.
The execution of the project didn’t require any significant deviation from the planned work program, as far as the global objectives and ambition of the project are concerned. Relatively minor alterations of the technical program were implemented where necessary, e.g. due to scientific outcomes making initially scheduled activities irrelevant or unnecessarily challenging. Such re-organization, which was agreed upon by the concerned partners during WP meetings or the 6-monthly regular consortium meetings, is inherent to any research project where by definition, the scientific outcomes can hardly be anticipated. Overall, the partners all agreed at the final meeting that the management of the project ran very smoothly and with a certain pragmatism focused on the final and most important objectives of the project.
The results produced by the project have been exploited by the various partners in different ways, depending on the type of results and partner activities. Generally, the results have been exploited by all partners to generate publications at international conferences and in peer-reviewed journals. This activity is important for maintaining the international reputation of the academia and research groups, to establish the young researchers in the scientific community, and to evidence the innovation potential of the industrial partners. To date, the project has resulted in a number of conference and journal papers, that have been made publicly available on the project website www.idealvent.eu wherever permitted by copyright ownership regulations.
A considerable amount of numerical and experimental data has been developed by the partners. In order to secure the future usability of these data, by the consortium and beyond, the consortium members endeavor to maintain and document a significant subset of key results to permit the validation of future numerical or modeling approaches. The datasets will be managed by the partners having generated them.
From a more academic viewpoint, the exploitation of the knowledge developed in the project is twofold. Firstly, the novel theoretical developments have been assimilated into the ex cathedra course material, and thereby disseminated to the future generations of engineers trained by the academic partners. Secondly, the application cases described above, inspired by industries and including experimental and numerical databases, constitute excellent training material for the final year Masters and PhD projects hosted by the universities.
Method-wise, the academia and research centres of the project have consolidated and upgraded their theoretical, experimental and numerical methods, which will be further used and developed in future projects. The software vendors have improved their solutions, integrated in the final product as novel libraries. The new tools being validated on industrial configurations can now be marketed towards the fields of activity of the project and beyond. Indeed, the scientific and technological solutions developed in this project are applicable to other sectors of activity than aeronautics including Heating, Ventilation and Air-Conditioning (HVAC) systems in surface transport, buildings or industrial cooling systems. The outcomes will accordingly be exploited in future research and commercial projects.
Last but not least, the project will have offered an unprecedented opportunity to the final users LTS and EMB to develop a quantitative assessment of the importance of installation effects in Environmental Control Systems, and of the impact towards the passengers and personnel servicing the aircraft on the apron. This knowledge, and the exposure of the final users to state-of-the-art modeling and mitigation strategies, will serve them to design better products and improve their competitiveness.
Potential Impact:
The IDEALVENT project contributes directly to several of the objectives of the ACARE Strategic Research Agenda for air transport, and Flightpath 2050 vision for aviation. The project outcomes will lead to novel aircraft integrated ventilation concepts for the reduction of cabin and ramp noise, which is being recognized as a growing issue for passenger comfort and safe aircraft operation and maintenance.
Regarding ramp noise first, our consortium estimates that for the positions farther from the APU (entrance door, some servicing positions), a reduction of the ECS noise sources will reduce the overall ramp noise by the same level. In other terms, designing a system 10 dB(A) quieter, the overall noise level at the regions specified above will also be reduced by 10 dB(A) as well. At the tail of the aircraft the noise levels will be still dominated by the APU, but EMB estimates that for this case a reduction of 3 dB(A) can still be expected. This clearly demonstrates that the IDEALVENT project can potentially have a very significant positive impact on the occupational health and safety of the personnel servicing the aircraft.
Then, cabin comfort is more than a question of convenience. The adverse effects that noise can have on health have been documented for quite some time, and include hearing impairment, hypertension, ischemic heart disease, annoyance, sleep disturbance and loss of concentration. Furthermore, the quality of the cabin acoustic environment is becoming a more pressing constraint since the spread in economy class of personalized flat screens for the in-flight entertainment, games, and services of all sorts that are nowadays making the concept of office-in-the-sky come true. Beyond the comfort issue, the economic success of a complete newly envisioned range of comfort items and customization concepts is at stake with cabin noise. Passenger comfort has become a challenging competitive selling point. In quantitative terms, EMB estimates that a reduction of 6 dB(A) in the ECS noise sources inside the cabin will lead to a decrease of 3 dB(A) in the overall cabin noise. The results obtained by LTS in the project do confirm these orders of magnitude, with Overall Sound Pressure Level reductions of about 9 dB measured at the inlet side of the ECS, and 2.5 dB reduction on the outlet side.
Beyond the aeronautical field, the tools and methods developed in the project will have very tangible impact on other fields such as ground transportation, where the noise emitted by train and automotive air-conditioning and cooling systems is a concern. Building ventilation systems are concerning a very large community as well, having a significant impact on the quality of the workspace, health and performance of the employees. In these various fields, the level of technological readiness of some modelling and mitigation strategies is somewhat lagging behind the level of expertise reached by the aeronautical industry. The presence of these non-aeronautical important societal actors within the Users Committee will greatly enhance the dissemination of mature tools and best practices developed within the project towards a wide industrial community affecting the life of many European citizens. One of the important outcomes of the IDEALVENT project will be providing the aeroacoustic simulation community with tools and recommendations on the best modelling practices for ventilation systems. The extensive and detailed experimental campaign that will be performed in WP2 on the one hand, and the advanced simulation tools that will be developed in WP3 and applied in WP4 on the other hand, will be employed to better understand the type and degree of interaction existing between the different components ventilation systems, at both aerodynamic and acoustic levels. Ventilation systems designers are eagerly expecting numerically efficient simulation strategies to reach optimized designs in a cost-effective and reliable way. This accumulated knowledge will be also very instrumental to devise control strategies in WP5, in order to manipulate the transient flow features and reduce noise production and transmission.
List of Websites:
https://www.idealvent.eu/