Vehicle Wind Noise Analysis Using a SEA Model with Measured Source Levels

<div class="htmlview paragraph">Automotive interior noise at mid and high frequencies is typically dominated by the airborne noise from acoustic sources that are spatially distributed around a vehicle. Each source is typically spatially compact (for example, a tire contact patch) but the source radiates sound that then propagates across the entire exterior surface of the vehicle. To characterize a source it is therefore necessary to know both the sound pressure level in the vicinity of the source and also the way in which sound from the source diffracts around the vehicle. The former depends on the details of the source, the latter typically depends on the overall vehicle geometry. When creating Statistical Energy Analysis (SEA) models of interior noise, the diffraction of airborne loads around a vehicle is often measured experimentally. Since SEA deals with averaged quantities, it is typically only necessary to obtain an approximate estimate of the space and frequency averaged sound pressure level across various surface regions of the vehicle (rather than a very detailed narrowband prediction of the diffraction).</div> <div class="htmlview paragraph">This paper provides a numerical investigation of the diffraction of typical acoustic sources around a vehicle using the Fast Multipole Boundary Element Method (FMM BEM). In particular, the paper investigates the sensitivity of the space and frequency averaged diffracted field to various modeling details (for example, geometric detail, mesh density/quality, surface impedance, frequency resolution etc.). The objective of the study is to investigate how much “detail” is needed in an FMM BEM model in order to get accurate predictions of the exterior acoustic loads for use in Statistical Energy Analysis (SEA) models. Significant reductions in computational expense can be obtained by matching the level of detail in the FMM BEM model to the level of detail required in the input results for an SEA model (and therefore avoiding “over computing” results that are subsequently frequency and space averaged).</div>

<div class="htmlview paragraph">Each time a new vehicle is developed, engineers face the challenge to develop the ideal sound insulation package. The goal is to attenuate powertrain, wind and road/tire noise from entering the vehicle while complying with cost, weight and packaging constraints. The design process is greatly facilitated if the engineer has effective tools to rapidly quantify how various sound insulation components contribute to the overall NVH performance of the vehicle.</div> <div class="htmlview paragraph">This paper discusses how an interactive vehicle acoustical design tool can be developed that assists the designer in making rapid decisions as to how to balance the performance of the various sound package components. The acoustical design tool is unique for each vehicle, and must take into account design decisions such as type of powertrain, body style, and numerous other factors in order to correctly predict the performance of the total package. Any good modeling tool must also take into account inputs from a reasonable range of operating conditions.</div> <div class="htmlview paragraph">A very good model for the acoustical behavior of the vehicle can be developed using data from measurements made while operating a sample vehicle at the target operating conditions. The authors will present a method for utilizing measurements together with some simple equations to predict the result when components of different designs are exchanged. The process of generating the needed data for this method from experiments on a vehicle is admittedly time consuming, if all of the possible design variations are to be considered. An alternative that the authors present in this paper is for the case in which a correlated Statistical Energy Analysis (SEA) model exists for the vehicle being considered [<span class="xref">1</span>]<sup><span class="xref">1</span></sup>. A great time savings is realized when the same acoustical behavior information about the vehicle can be extracted by running analytical experiments in the SEA model. The authors assert that even if an SEA model must be developed from scratch and correlated, this approach is preferable over the experimental approach.</div> <div class="htmlview paragraph">Once developed, an interactive design tool of the type being discussed has ongoing benefit to product development engineers by giving them a handy way to asses future design changes. If sound insulation components are later sacrificed in the interest of reducing vehicle costs or weight, the model can accurately portray the implication of the changes on sound levels in the vehicle. A case study is presented that demonstrates the use of this tool and its accuracy.</div>

Improved model predictive control for the flow field in a wind tunnel with model uncertainty

<div class="section abstract"><div class="htmlview paragraph">Prediction of flow induced noise in the interior of a passenger car requires accurate representations of both fluctuating surface pressures across the exterior of the vehicle and efficient models of the vibro-acoustic transmission of these surface pressures to the driver’s ear. In this paper, aeroacoustic and vibro-acoustic methods are combined in order to perform an aero-vibro-acoustic analysis of a Mercedes-Benz A-class. The exterior aero-acoustic method consists of a time domain incompressible Detached Eddy Simulation (DES) and an acoustic wave equation. The method is extended in this paper to account for convection effects when modelling the exterior sound propagation. The interior vibro-acoustic model consists of a frequency domain Finite Element (FE) model of the side glass combined with a generalized Statistical Energy Analysis (SEA) model of the interior cabin. The method is extended in this paper to account for additional geometric details that affect the direct and reverberant fields within the cabin. The overall approach is presented and validated against vibro-acoustics tests and wind tunnel measurements for two separate mirror designs. The aero-vibro-acoustic method achieves very good results and accurately predicts the change in interior noise resulting from an exterior body design change.</div></div>

<div class="section abstract"><div class="htmlview paragraph">Statistical Energy Analysis (SEA) is the standard analytical tool for predicting vehicle acoustic and vibration responses at high frequencies. SEA is commonly used to obtain the interior Sound Pressure Level (SPL) due to each individual noise or vibration source and to determine the contribution to the interior noise through each dominant transfer path. This supports cascading vehicle noise and vibration targets and early evaluation of the vehicle design to effectively meet NVH targets with optimized cost and weight.</div><div class="htmlview paragraph">A common misconception is that SEA is only capable of predicting a general average interior SPL for the entire vehicle cabin and that the differences between different locations such as driver's ear, rear passenger's ear, lower interior points, etc., in the vehicle cannot be analytically determined by an SEA model. However, because the interior acoustic energy distribution varies due to absorption and distance effects that can be modeled, an SEA model is capable of predicting the SPL at different interior locations with good accuracy at high frequencies.</div><div class="htmlview paragraph">This paper discusses the SEA modeling assumptions used to generate a typical model of a vehicle cabin interior and surrounding structure. The distribution of acoustic absorption and its effect on the local interior SPL responses are addressed. Measurements of transfer functions to various points of the vehicle interior from exterior and interior acoustic sources and structureborne sources for a typical vehicle are presented and compared to SEA model predictions. Observations and recommendations about typical interior transfer function correlation, modeling limitations, and use of the SEA model as a design tool are given.</div></div>

<div class="htmlview paragraph">A process is implemented to propagate uncertainties inherent to the Statistical Energy Analysis (SEA) modeling practice to variance in predictions. A Monte Carlo based approach is scripted for the VA-One environment to account for uncertainties in gross parameters of SEA model subsystems. The variance module of the commercial software is used to estimate possible variations in local modal properties. A first-order expansion solution is applied to integrate uncertainties in the power inputs of the system. The impact of each type of source is assessed in computing overall variance in predictions. The process is applied to analysis of in-flight interior cabin noise predictions using a simplified aft fuselage section SEA model.</div>

<div class="htmlview paragraph">Statistical Energy Analysis (SEA) models were created and analyzed of the Gulfstream Aerospace Corporation (GAC) G100 mid-cabin and G150 wide-cabin, high speed business jets to assist in the design efforts of the G150 cabin interior development program. The G100 was used as a model validation platform due to the existence of extensive test data. Once satisfied that the modeling technique and material properties were adequately characterizing the G100 aircraft, a G150 model was created and predictions were made. Additionally, optimization was performed successfully on the G150 SEA model to reduce the weight of the thermal / acoustic package without adversely affecting the acoustic performance. This analytical modeling effort was the first of its kind to be used at Gulfstream during an early development phase in helping to design and optimize a GAC lightweight interior thermal / acoustic package. This paper describes the specifics of analytical modeling as part of the G150 interior development effort by rendering an overview of the SEA models, constructs used to develop the models, applied loads, response of the models, and correlation of results to test data. Cambridge Collaborative Inc.'s SEA software SEAM3D and Engineous Software's optimization code iSIGHT were used in this analysis effort.</div>

<div class="section abstract"><div class="htmlview paragraph">Many sources and paths cause interior cabin noise. Some noise from an electric vehicle is unique and different from a vehicle with an internal combustion engine. Especially, whine noise occurs due to the particular orders of the electromagnetic force of an electric motor and transmission gears, which is tonal and usually reaches high frequencies. This paper covers structure-borne (SB) and airborne (AB) aspects to estimate whine, and the difference between the two characteristics is distinguished. The focus lies mainly on the process of virtual vehicle development and application for performance improvement. First, to predict SB whine, an e-powertrain is modeled as a finite element model (FEM), and electromagnetic (EM) forces are calculated. A vehicle model is also modeled as an FEM, in which interior sound packages are carefully modeled as they play an important role in the medium-frequency region. The e-powertrain and vehicle models (being simulated separately) are combined to obtain cabin noise up to 1.5 kHz. Design studies show that the stiffness of mount insulators and the panel stiffness of the vehicle can be substantial design variables to reduce the SB whine. Second, the study highlights a simulation method to predict interior airborne whine up to 8 kHz by combining the FEMs of the e-powertrain and the vehicle’s exterior cavity with a statistical energy analysis (SEA) model of a vehicle. Path contribution can be identified by defining source strength and acoustic transfer function of airborne paths. Design modifications, including encapsulation of the e-powertrain, show this simulation process could be practically useful to reduce the airborne whine at high frequencies.</div></div>

<div class="section abstract"> <div class="htmlview paragraph">This Recommended Practice provides a procedure for measuring and documenting the aerodynamic performance in a full-scale wind tunnel of passenger vehicles, i.e., mass-produced cars and light-duty trucks intended primarily for individual consumers. Testing or numerical modeling of pre-production and/or reduced-scale models is outside the scope of this document. Aerodynamic <i>development</i> procedures, i.e., methods to improve or optimize aerodynamic performance, are also excluded.</div> <div class="htmlview paragraph">It is well-known that aerodynamic performance results depend significantly on vehicle content and loading, as well as the wind tunnel itself (type, scale, and simulation qualities of the wind tunnel). Publication of non-standard test results causes unnecessary additional development work and incorrect perception of a vehicle’s anticipated aerodynamic performance by government, academia, and the general public.</div> <div class="htmlview paragraph">The intent of this document is to promote uniformity and traceability of published aerodynamic performance data acquired in a wind tunnel, thereby enhancing the public reputation and credibility of the aerodynamics discipline in the total vehicle development process.</div> <div class="htmlview paragraph">Measurement of the aerodynamic performance according to this Recommended Practice requires a full-scale wind tunnel test on a production-ready, full-scale vehicle according to specifications for vehicle loading, content, and wind tunnel conditions.</div> <div class="htmlview paragraph">The measurement and resulting documentation for a particular vehicle shall be reported only by the company that intends to produce the vehicle. Whenever there is an opportunity for interpretation of the document, a good faith effort shall be made to obtain the vehicle’s typical in-service performance and avoid finding the best possible performance under the best possible circumstances.</div> <div class="htmlview paragraph">Use of and compliance with this Recommended Practice is voluntary; however, it is mandatory for those advertising or otherwise publishing aerodynamic performance as “measured according to SAE J2881.” Conversely, any advertised or otherwise published result carrying this or a substantially similar designation must have the necessary documentation accepted by and on file with the SAE in accordance with the procedures defined in this Recommended Practice.</div> <div class="htmlview paragraph"><span class="xref">Appendices B</span> and <span class="xref">C</span> contain the necessary submittal documents for compliance with this Recommended Practice. When a submittal is made, the chairman of the Road Vehicle Aerodynamics Forum Committee (RVAC) or his/her designee shall review the submittal for proper completion according to the requirements herein, but without consideration of the technical merit of the claimed performance. After the chairman’s review, SAE will maintain an electronic database of all submittals.</div> <div class="htmlview paragraph">The database shall be available only to members and liasion members of the RVAC. Any RVAC member who retrieves information from the database shall do so with the understanding that the information is to be considered confidential within his or her company.</div></div>

<div class="htmlview paragraph">The high-frequency performance of automotive sound packages may be predicted by combining transmission loss and absorption calculations with statistical energy analysis (SEA) models. Vehicle SEA modeling therefore requires analytical tools that accurately describe the acoustic properties of layered structures such as trim panels. While widely accepted prediction models for fibrous materials exist, less work has been done on foam layers. This paper presents a numerical calculation of sound transmission through layered structures using a direct global matrix (DGM) solution. This numerical technique is used to compare several models for acoustic propagation in foams. Predictions are compared to transmission loss measurements for several panel configurations. A rigid porous model compares well with data up to 1 kHz for the foams studied, while foam elasticity appears to be important at higher frequencies.</div>

<div class="htmlview paragraph">There is increasing concern about potential differences in aerodynamic behavior measured in steady flow wind tunnel conditions and that which occurs for vehicles on the road. As tools become available for better simulation of on road conditions there is a growing practical value in understanding what range of conditions are important to simulate.</div> <div class="htmlview paragraph">Surface pressures measured on the sideglass of a European hatchback vehicle in the MIRA full scale wind tunnel are compared with those measured on-road. The on-road data corresponds to relatively calm, low yaw conditions and the time averaged pressure distributions on-road and in the wind tunnel at zero yaw were very similar.</div> <div class="htmlview paragraph">Variations in instantaneous aerodynamic yaw angle produces fluctuations in surface pressures but the sensitivity of instantaneous pressures to yaw angle was lower for the on-road measurements compared with steady state wind tunnel tests. Also, surface pressure unsteadiness was larger than could be attributed to yaw angle fluctuations alone.</div> <div class="htmlview paragraph">The response of instantaneous surface pressures to yaw angle fluctuations is reduced for higher frequency fluctuations as small scale turbulence gusts are less able to modify the flow over the entire vehicle. Examination of the spectral relationship between surface pressure and yaw angle indicates that the impact of yaw angle fluctuations at is reduced by a factor of 2 for reduced frequencies above unity.</div>

<div class="htmlview paragraph">High frequency responses of structural-acoustic systems may be predicted by statistical energy analysis (SEA) or energy finite element method (EFEM). To combine the good features of these two techniques, a simplified energy finite element method, referred to as EFEM<sup>0</sup>, has been developed recently. The EFEM<sup>0</sup> technique, which is compatible with SEA, integrates the joint coupling procedures for discontinuous systems and the finite volume formulation for continuous system. The EFEM<sup>0</sup> models have been verified either analytically or experimentally for one- and multi-dimensional systems.</div> <div class="htmlview paragraph">In this study, the EFEM<sup>0</sup> technique is applied to a passenger van for a noise control investigation. The general scheme is to incorporate the EFEM<sup>0</sup> coupling factors into a SEA model in order to release some SEA assumptions and improve the SEA model, especially for relatively high damping, strong coupling and direct field cases. The EFEM<sup>0</sup>-SEA integrated model for the trimmed van is validated by experimental data. Several noise control strategies will be discussed for the integrated model to optimize the sound package.</div>

This paper introduces the interval perturbation method into the sensitivity analysis of statistical energy analysis (SEA) models. The damping loss factors and coupling loss factors are treated as the design parameters, while the modal energy of each subsystem is considered as the decision-making target. The variation of the design parameter is characterized by an interval form, and the ratio between different interval widths is used to represent the sensitivity of the decision-making target to the design parameter. Then the sensitivity factor matrix that can help a designer find the design parameters giving the greatest benefit in noise and vibration reduction is provided. The presented interval approach is verified through simulated studies for a two-plate coupling system and a three-plate coupling system. The work of this paper can enrich and develop the theory of the sensitivity analysis of SEA models.

<div class="htmlview paragraph">The unsteady environment road vehicles are exposed to is subject of many investigations that are currently made. Yet, the approaching flow is only one aspect of unsteady forces acting on the vehicle. Unsteady wake structures also lead to time-varying surface pressures and consequently fluctuating forces even in steady and low turbulent flows. However, little is known about the influence of realistic flow conditions, i.e. as found on road, on the unsteady surface pressures and wake structures of a vehicle.</div> <div class="htmlview paragraph">Therefore, to derive a deeper understanding of the unsteady aerodynamic properties of a vehicle this paper presents results of measurements conducted on a vehicle body both in smooth and turbulent flow conditions in the IVK model scale wind tunnel. Unsteady surface pressure measurements in the area where separation occurs and the base of the vehicle were made together with time accurate total pressure measurements in the wake. The analysis of periodic surface pressure fluctuations and flow structures reveals remarkable changes in time averaged and instantaneous properties of the flow field when the turbulence intensity is brought to levels typical for on road conditions.</div> <div class="htmlview paragraph">These findings give insight into the interaction between realistic turbulence levels and changes in the flow around the vehicle body. From this the impact on unsteady forces is derived.</div>

In order to improve the quality of flow field in wind tunnel in China, The 2.4 m × 2.4 m intermittent wind tunnel control Mach number is required to achieve the accuracy of 0.001. Dynamic matric control is a great way to handle multi-variable, time delayed nonlinear control system, and there are nonlinear, time-delayed problems of Mach number control in intermittent wind tunnel. The prediction models use step response coefficients to be the parameters of the wind tunnel system, however, the time of wind tunnel test is about 90s, if the quantity of step response coefficients N is too large, it spends a long time collecting step response coefficients and we have limited time for aircraft blowing test. We need to get four different kinds of step response coefficients. So, we have to use incomplete step response coefficients as prediction model, If the number N is less than the number of the stable value of the step response, the truncation error is generated in the prediction process. In order to get a less number N and at the same time eliminate the truncation error, this paper improves the dynamic matrix control algorithm to get better results.