Sie sind auf Seite 1von 16

Efcient Calculation of SEA Input Parameters Using a Wave Based Substructuring Technique

P. Ragnarsson1 , B. Pluymers1 , S. Donders2 , W. Desmet1


1 K.U.Leuven,

Department of Mechanical Engineering, Celestijnenlaan 300 B, B-3001, Heverlee, Belgium e-mail: patrik.ragnarsson@mech.kuleuven.be
2 LMS

International, Interleuvenlaan 68, B-3001, Leuven, Belgium

Abstract
This paper presents a new and efcient method to calculate point mobilities from subcomponents of a full structure. When performing dynamic analyses of subcomponents the result normally tends to strongly depend on the boundary condition selected for the component. The goal is to simulate the results one would have received if the full model would be analysed. Normally free or clamped boundary conditions are used. In this paper waves extracted in a Wave based substructuring analysis will be used to dene a more realistic boundary condition. The results show great promise and for the three investigated cases the results are a better match to the reference calculation than both free and clamped boundary conditions. It is especially clear that the results are accurate for a wider frequency range. While free boundary condition can be suitable for high frequencies and less suitable for low frequencies and clamped boundary condition can be suitable for low frequencies but less suitable for high frequencies, the WBS boundary condition yields good results over the entire frequency range.

1 Introduction
In modern vehicle design processes the use of Computer Aided Engineering (CAE) has increased dramatically in recent years. In order to reduce time to market and to minimize the number of prototypes for a vehicle manufacturer it is crucial for a design engineer to get early and accurate predictions of new designs. Low-frequency methods such as the Finite Element Method (FEM) and the Boundary Element Method (BEM) have, due to increasing computer speed, been able to run deterministic models to higher and higher frequencies, while Statistical Energy Analysis (SEA) has become the standard method for both acoustic and vibration analysis in the high-frequency range. However, there is still a frequency gap where no mature prediction method exists today [1]. In recent years much research have been focusing on lling this gap by extending the applicable range of the deterministic methods towards higher frequencies and by lowering the valid frequency range for SEA. Langley and Bremer developed a method in which deterministic and statistical techniques are combined [2]. The stiff components (long wave length) are modeled with Finite Element (FE) and the soft components (short wave length) are modeled with SEA. In classical SEA, the input parameters are normally evaluated using different analytical formulas e.g. for innite plate theory. For high frequencies this theory works well, but when trying to expand the use of SEA towards lower frequencies, other methods are needed. Lyon,
1579

1580

P ROCEEDINGS OF ISMA2008

DeJong [3] and later Manning [4] have shown that the point impedances at input and output locations can be used to better estimate the input power and the output response. This is a well known method and for many years measurements have been used to get a better estimation of these parameters [5]. Lately the use of deterministic methods such as FEM to better estimate these parameters has shown promising results [6] [7]. However, to run a complete FE calculation into the mid-frequency region is highly time consuming for industrial sized models. To really capitalize on the strength of the deterministic methods, the calculation times must be reduced. This paper focuses on a method to perform subcomponent modeling in order to retrieve the important SEA parameters (input and response point mobility, coupling loss factors (CLF) etc.) without having to run the full model. A subcomponent can be any part of the complete model of interest, in vehicle analysis this can be e.g. a b-pillar or a section of the oor panel. The goal is to run a fast calculation of a subcomponent and to retrieve a result that is as close as possible to the result one should expect from a calculation of the full model. Since the subcomponent is a part of a larger system the a dominant error source will be the boundary condition at the subcomponent interface. A classical way to treat this interface is to simplify the boundary conditions to a free or a xed boundary condition. This works well for high frequencies and for low frequencies, but for the mid-frequency range neither of these boundary condition types yields accurate results. Therefore, improved methods are needed. Wave-Based Substructuring (WBS) [8], [9] is a powerful substructuring method which uses a set of basis functions (waves) to describe the behavior of the interfaces. In this work it will be investigated if the WBS formulation can be used as an efcient way to dene more realistic boundary conditions. A new wave extraction criterion is introduced to create a boundary condition which is as close as possible to the as in assembly condition. The results will then be compared to results obtained in an analysis of a complete FE model of the same structure. In this paper the wave extraction is performed on the same assembly as the subcomponent originates from, but a future step in this research project is to expand the method so an FE calculation of another similar system can be used. In that way an already existing model (i.e. a predecessor of the new model) can be used for the wave extraction and the waves can then be applied to a subcomponent of the new car model already before a full FE model of the new car has been created. This will then be a powerful method for the design engineers to try out new subcomponents early in the design phase.

2 Theory
2.1 Classical SEA

The analysis procedure using SEA is to divide the studied structure into structural and acoustic subsystems with a dened level of subsystem damping and coupling between the subsystems and to then solve the power balance equation for these subsystems for one or more applied external power input excitations. The power balance equation for a subsystem s can be written as [3]: Ws,in = Ws,diss +
s=r

Ws,r

(1)

where Ws,in is the input power, Ws,diss is the power dissipated through damping mechanisms in subsystem s, and Ws,r is the net power transmitted from subsystem s to subsystem r , (Ws,r = Wr,s ). The transmitted power can be written as: Ws,r = s,r Es r,s Er (2)

where s,r and r,s are the coupling loss factors between the different subsystems, is the radian frequency, and E is the time-averaged energy stored in the subsystem. The coupling loss factors are related to each other

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1581

through the modal densities n(), as is shown in Equation (3). This relation is commonly known as the SEA reciprocity relation. For all cases in which the modal densities between subsystem s and subsystem r are not the same, the coupling loss factors s,r and r,s are not reciprocal. By instead introducing the concept of a reciprocal coupling factor r,s as the product of radial frequency, modal density, and coupling loss factor: s,r = s,r n()s = r,s n()r = r,s (3)

the denition of modal power s as the ratio of total energy to the modal density of the subsystem can be used. The transmitted power between subsystems can now be written as the product of the coupling factor s,r and the difference between the modal powers of the subsystems, s and r . This means that the net energy ow between subsystems is simply proportional to the difference in subsystem energy per modal density in connected subsystems. Ws,r = s,r Es Er n()s n()r = s,r (s r ) (4)

In a similar way the loss factor s,diss , can be introduced as a product of the radial frequency , the damping loss factor and the modal density n(): s,diss = s n()s (5)

With this denition of loss factor, the dissipated power in a subsystem Ws,diss may also be written in terms of the damping factor s,diss and the modal power s : Ws,diss = s Es = s Es = s,diss s n()s (6)

As in Equations (1) - (4), the dissipated power is obtained in terms of the of modal power and a coefcient dependent on the frequency, modal density and damping loss factor. The principal advantage of expressing the coupling and damping in terms of reciprocal coupling factors and loss factors rather than in terms of coupling loss factors and damping loss factors is that the linear power balance Equation (1) can now be written in matrix form as a symmetric, positive denite matrix using the notations from Equations (1) - (6):

W1,in 1,diss + s 1,s 1,2 1,3 . . . W2,in 1,2 2,diss + s 2,s 2,3 . . . = . . . .. . . . . . . .

1 2 . . .

(7)

Besides the obvious computational advantages (stability and solution speed) of solving a symmetric, positive denite matrix, there is an additional advantage that the coupling factors and the modal powers are parameters related to quantities that are easier to measure or calculate by FEA or other hybrid analysis methods. The substitution of reciprocal coupling factors is also a key feature for the practical implementation of the SEAFEA point mobility method that is described in further detail in the following section.

2.2 SEA-FEA point mobility approach


One frequently studied parameter that is important to understand when predicting vehicle noise is the transfer function (TF) from an excitation in one part of the vehicle to the response in another part of the vehicle. An example of such a transfer function that can be of importance is the structural transfer function from an applied force at an engine mount to the vibration of a oor panel or the structural-acoustic transfer function to the sound pressure level inside the car cabin near the drivers ear. The following equations are based on

1582

P ROCEEDINGS OF ISMA2008

the formulations presented by Manning in [4]. First, a dimensionless power transfer function (W T F ) can be obtained by dening a new term based on Equation (7): W T Fs,r = r Ws,in (8)

where W T Fs,r is the power transfer function relating the modal power of the response subsystem r to the input power of the source subsystem s. In a typical vehicle development situation the transfer function of interest is often the transfer function between the input force and the response velocity (also called the transfer mobility function): T Fs,r = vr Fs (9)

This transfer mobility function T Fs,r can be written as a function of the SEA power transfer function W T Fs,r by expressing the input power as a function of the drive point conductance Gs (real part of the 2 mobility) and the mean square of the applied force Fs . In classical SEA, the average subsystem conductance is used. The average is then taken over the spatial extent of the subsystem and over a band of frequencies. However, the relation shown in Equation (10) can also be applied to single frequencies and to individual points as long as the proper conductance is used.
2 Ws,in = Fs Gs

(10)

The modal power r can also be dened as a function of the drive point conductance Gr and the average 2 mean square velocity vr of the subsystem: r =
2 vr 2 Gr

(11)

Then the square of the transfer function between an input force in subsystem s and the response in subsystem r can be written as: |T Fs,r |2 = vr Fs
2 2

2 r Gr Gs 2 = Gs Gr W T Fs,r Ws,in

(12)

This very useful result gives an expression for the transfer mobility function based only on the power transfer function (SEA results) and the point conductance for both the excitation and the response point (which can be retrieved from measurement or FEA). This result is the basis for the point mobility approach presented in [4] and used in this paper. Calculation of the point conductance for important input and output points will be used to enhance the results from the SEA calculations. FE calculations will also be used to enhance the coupling factors. Manning also shows that the coupling factor for a point connection between two subsystems can be expressed in terms of the point mobilities for these subsystems: s,r = 1 4 Gs Gr 2 |Ys + Yr |2 (13)

where Y is the complex valued point mobility . In classical SEA theory the point mobility functions are calculated from expressions for the innite (or semi-innite) system. The accuracy of this approach depends on the averaging bandwidth and the modal overlap factor The modal overlap is a function of modal density and damping and the highest value is obtained when both of these parameters are large for a certain bandwidth. A wide bandwidth and/or high modal overlap result in a lower expected variance of the response from the mean prediction, so that the accuracy of the result compared to a measurement is expected to be good. This approach has shown to be a very accurate approximation for the high-frequency range where the modal

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1583

overlap generally is high. This is the underlying reason that SEA has come to be accepted as the standard tool for high-frequency CAE analysis in vehicle design engineering. For the mid- and low-frequency ranges, where the modal overlap values tend to be smaller, the denition of the coupling factor in Equation (13) using expressions for innite systems may lead to large errors compared to measured response data. The reason for this is that the smaller modal overlap is, the more the measured response (averaged over third octave bands) tends to depend on single deterministic phenomena. By instead using FEA to calculate these point mobility functions, the transfer function accuracy can be improved for the mid- and low-frequency range. This can of course be performed using a full FE model of the investigated system, but if the system is large e.g a car body in white (BIW) or a full vehicle model, the computational cost will prohibit to stretch the calculations towards higher frequencies. The point mobilities can also be calculated on only a part of the system. The boundary conditions for the interface between the remainder and the subcomponent must then be added to the model. Normally free or clamped boundary conditions are applied. This simplication can often lead to results where some of the dynamic behavior of the complete system is missed. The method which will be investigated in this paper is to apply the waves which connect the parts in Wave-Based Substructuring [8], [9] as boundary conditions. In section 2.3 the Wave-Based Substructuring theory is explained.

2.3

Wave-Based Substructuring

Wave-Based Substructuring (WBS) is a powerful substructuring method in which the dynamic behavior of the interfaces are described by a set of basis functions (waves). This reduces the model size compared to conventional substructuring methods as Component Mode Synthesis (CMS) as the number of waves needed to describe the dynamics of the interfaces is normally much smaller than the number of physical degrees of freedom (DOF) of the interfaces [8], [9]. Instead of connecting the nodes from two neighboring subsystems directly via continuity equations, WBS connects two subsystems via a set of virtual nodes (the wave participation factors). This makes the method very suitable for subcomponent modeling. An undamped system with no external forces can be described by the FE matrix equation: M + Kx = 0 x (14)

If this system is divided up into non overlapping substructures the model can be divided into internal DOFs xi and junction DOFs xj . The equation of motion can then be divided into sub-matrices and written as: Mii Mij Mji Mjj i x j x + Kii Kij Kji Kjj i x j x = 0 fj (15)

In WBS the junction DOFs are expressed as a linear combination of a set of waves (V ). These waves are weighted with a participation factor p: xj = V p The equation of motion, Equation (15), can then be written as: Mii Mij V T V Mji VT Mjj V i x p + Kii Kij V T V Kji VT Kjj V i x p = 0 (17) (16)

V fj

The waves can be seen as the modal displacement shapes dened only at the interface of the substructure and are obtained from a modal analysis from the full FE model. Since the waves are a projection of the orthonormal system modes on a subset of the system DOFs, orthonormality of the waves can not be guaranteed. Therefore some post processing of the waves must be performed aiming at orthonormalization of the waves and at selecting the minimum number of waves needed to represent the dynamics of the interfaces.

1584

P ROCEEDINGS OF ISMA2008

To avoid numerical issues in the solution of ill conditioned problems, it is therefore crucial to perform an orthogonalization on the interface modal displacement shapes, and to apply a selection threshold to keep only the relevant waves. For this purpose, one can use Singular Value Decomposition (SVD) to decompose the interface modal displacement shapes into a set of orthonormal waves and to rank these waves from the wave with the highest singular value to the wave with the lowest singular value. The selection step then consists of applying a threshold on the singular values, and keeping only the waves that correspond with a singular value higher than the dened threshold. Physically, this corresponds to discarding those waves that have a below-threshold contribution to the mathematical description of the complete vector space that describes the interface dynamics. In conventional WBS, the interface modal displacement shapes are obtained from a full FE analysis of the complete system; work is ongoing to lower this requirement, for instance by obtaining the interface modal displacement shapes only from a smaller subassembly [10]. As in all substructuring methods, continuity and equilibrium must be imposed at the boundaries. For a rigid connection between two subsystems (a) and (b), continuity of the interface displacement and equilibrium of the reaction forces are applied: xj
(a)

= xj

(b)

and

fj

(a)

= fj

(b)

(18)

Transformed into a WBS framework the continuity conditions will be: p(a) = p(b) and V T fj
(a)

= VT fj

(b)

(19)

Only rigid connections are discussed in this paper, however, the theory for elastic connections can be found in [8] and [9].

2.4 Using WBS waves as boundary conditions for a subcomponent modeling approach
Note that in conventional WBS, the orthogonalized waves are used to accurately describe the complete interface dynamics in the assembled system. When one uses a limited number of wave degrees of freedom instead of the (typically much larger number of) physical degrees of freedom at the interface, one obtains a smaller description of the interface dynamics. This results in a more efcient reduction procedure of large components with large connection interfaces and more efcient system-level calculations. In this paper however, the aim is to develop a subcomponent modeling approach, where the WBS technique is used to create a new type of boundary condition. This boundary condition shall describe the interface dynamics as if the subcomponent were connected to the rest of the assembly. For this purpose, one cannot use exactly the same selection procedure as in conventional WBS. The aim is to keep a subset of the waves, such that in a subcomponent analysis, you describe the interface dynamics to represent an as in assembly boundary condition. This boundary condition must be more accurate than a rigid or free-free boundary conditions. If one uses all the waves (as would be done by keeping the wave selection procedure of conventional WBS), in a component-level analysis, one has merely obtained a slightly more efcient model to predict the free-free behavior of that component. 2.4.1 Deriving a selection criteria for the waves As described in section 2.3, the waves connecting two components in WBS are calculated from a modal analysis of the assembled system. This leads to a wave set with a large number of waves. Some of them are important to the dynamic behavior of the subcomponent and some are not. The problem when using the waves as a boundary condition without connecting the subcomponent to the remainder is that all waves

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1585

contribute to the boundary condition with equal strength. Therefore a calculation step where the most important waves are selected has to be performed. In this paper the selection procedure is based on the Modal Effective Parameter (MEP). The MEP is a product of the modal displacement value at the input DOF and the modal displacement value at the output DOF, performed for each mode [11]. This product is scaled by the modal mass. The MEP between DOF m and DOF n for the r-th mode is dened as: Xmn,r = m,r n,r mr (20)

The MEP can be described as a measure of how much a specic global mode contributes to the transfer function between two points. When all the MEPs are calculated, the modes can be sorted in a descending order based on the MEP value. If the mode set then is truncated to contain only the modes contributing to a certain percentage of the total cumulated MEP relating to a transfer function of interest, the global modes most important to this transfer function are selected, see Figure 1.

Figure 1: Original mode set is rearranged so that the Mode with the highest MEP comes rst. This limited mode set is then used for the further analysis i.e for the wave extraction and wave orthonormalization.

Numerical results

As discussed in Section 1, a fast and accurate method to perform subcomponent modeling is crucial for using calculated point mobilities as a tool for evaluating the SEA parameters. In this paper three examples of subcomponents are presented and the point mobility for a number of locations on each subcomponent is calculated using different techniques. The results are then compared to a reference calculation performed on a full BIW. The investigated sub-components are the back part of the roof, the B-pillar and a part of the spare wheel panel, see Figure 2. A modal analysis is then performed on the subcomponents with three different boundary conditions, namely free, clamped and the WBS boundary condition. Free and clamped boundary conditions are straightforward to apply, and the procedure for the calculation and application of WBS boundary condition has been described in detail in section 2.4.1.

3.1 Calculation procedure


The model used in this paper is a Chrysler Neon BIW containing about 240 000 nodes and 230 000 elements. From this model different parts have been cut out to simulate the different subcomponents. A modal analysis has been performed up to 700 Hz and an FRF synthesis is performed based on the extracted modes. The FE calculation of the modes has been performed in MSC.Nastran 2004 [12] and the model setup, preprocessing, WBS denition, forced response analysis and postprocessing have been performed in LMS Virtual.Lab [11]. The results are then averaged both over frequency (third octave bands) and spatially (average of the responses for all the IO points on the subcomponent, see Figure 3) .

1586

P ROCEEDINGS OF ISMA2008

Figure 2: The three different subcomponents considered in the paper, extracted from the full model for a clearer view. 1. Back of the roof, 2. B-pillar and 3. Spare wheel panel.

Figure 3: The three different subcomponents considered in the paper, with the IO points marked with yellow dots. 1. Back of the roof (34 points), 2. B-pillar (3 points) and 3. Spare wheel panel (15 points).

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1587

3.2

Results

First the truncation limit for the wave set has to be determined. The modes are arranged according to descending MEP. For the back of the roof the truncation has been based on the total cumulated MEP of this rearranged mode set. Three different values ranging from 5% to 20% have been considered and the resulting FRFs for three different WBS boundary condition cases have been compared to the reference solution. The comparison has been performed by subtracting the reference FRF from the WBS FRF for each third octave band from 10 Hz to 500 Hz, see Figure 4.

Figure 4: The absolute value of the difference between the reference FRF and three different realizations of the truncation limit (5%, 10% and 20%) In the difference plot it can be seen that the 10% truncation limit seems to give the best performance. To further investigate this, the absolute values of the differences for all third octave bands are summed up to a total difference. In Table 1 these summed differences conrm the conclusion from the difference plot. Truncation limit Summed difference. [(m/s)/N] 5% 0.225 10% 0.172 20% 0.201

Table 1: The absolute values of the differences for all third octave bands summed up to a total difference. The 10% truncation limit is used for all three subcomponents in this study. For this application, this allows the wave set to be reduced from about 400 waves to approximately 5-20 waves depending on the subcomponent. Future research will focus on a more generic method to select the truncation criteria. In the next sections the results from the numerical study are shown. The three different subcomponents are presented with the different boundary conditions as well as the reference calculation. The results are presented rst in a narrow band analysis (f = 1 Hz) and then in a third octave band analysis. 3.2.1 Back of the roof For the back of the roof 20 randomly distributed points have been used for the spatial average. The results shown are the averaged point mobility for the back of the roof in the vertical direction (Z-direction). Global modes corresponding to a truncation limit of 10 % of the total cumulated MEP are used to calculate the waves. Figures (5) and (6) show that the WBS boundary condition gives better results than the clamped and the free boundary condition. For low frequencies (< 150 Hz), the free boundary condition greatly overestimates the mobility and for high frequencies (> 150 Hz) the clamped boundary condition generally gives to high mobility (can easiest be seen in the third octave plot). The WBS boundary condition however, manages to

1588

P ROCEEDINGS OF ISMA2008

Figure 5: Comparison of point mobilities for different boundary conditions for the back of the roof [(m/s)/N]. The reference result is taken from a full BIW calculation.

Figure 6: Comparison of point mobilities for different boundary conditions for the back of the roof in third octave band [(m/s)/N]. The reference result is taken from a full BIW calculation.

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1589

capture both the low and high frequency behavior of the subcomponent. This can be seen even clearer if the absolute value of the difference compared to the reference calculation is plotted, see Figure 7.

Figure 7: The absolute value of the difference between the reference FRF and the three different boundary conditions, namely Free, Clamped and WBS boundary condition for the back of the roof.

3.2.2 B-pillar For the B-pillar the results shown are the spatial average of three points on the B-pillar in the direction perpendicular to the side of the car (Y-direction). For the WBS boundary condition the modes contributing to 10 % of the total cumulated MEP are used to calculate the waves.

Figure 8: Comparison of point mobilities for different boundary conditions for the B-pillar [(m/s)/N]. The reference result is taken from a full BIW calculation. As can be seen in Figure 8 and Figure 9, the WBS boundary condition captures the real behavior of the assembled system much better than clamped or free boundary condition. This is true for the entire frequency range, as can bee seen in Figure 10.

1590

P ROCEEDINGS OF ISMA2008

Figure 9: Comparison of point mobilities for different boundary conditions for the B-pillar in third octave band [(m/s)/N]. The reference result is taken from a full BIW calculation.

Figure 10: The absolute value of the difference between the reference FRF and the three different boundary conditions, namely Free, Clamped and WBS boundary condition for the B-pillar.

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1591

3.2.3 Spare wheel panel For the spare wheel panel 15 randomly distributed points have been used for the spatial average. The results shown are averaged point mobility for the spare wheel panel in the vertical direction (Z-direction). Global modes contributing to 10 % of the total cumulated MEP are used to calculate the waves.

Figure 11: Comparison of point mobilities for different boundary conditions for the spare wheel panel [(m/s)/N]. The reference result is taken from a full BIW calculation.

Figure 12: Comparison of point mobilities for different boundary conditions for the spare wheel panel in third octave band [(m/s)/N]. The reference result is taken from a full BIW calculation. Over the full frequency range the WBS boundary condition is better or equal to both free and clamped boundary condition, see Figure 13. However, the 10% truncation limit used for all three cases might not be the optimal limit for this case. Future research will focus on nding a more generic method to select the truncation limit.

Conclusions

In this paper, a novel approach has been presented to accurately dene as in assembly boundary conditions on individual structural subcomponents. This is important to improve the quality of a point mobility calcu-

1592

P ROCEEDINGS OF ISMA2008

Figure 13: The absolute value of the difference between the reference FRF and the three different boundary conditions, namely Free, Clamped and WBS boundary condition for the B-pillar. lation on a structural component, which in turn is important for the accurate estimation of SEA parameters. Since running industrial sized models into the mid-frequency region is often very time consuming, there is also a need for a method were accurate results can be obtained from a smaller model. The novel approach is based on the Wave-Based Substructuring (WBS) technique, which allows writing the interface between substructures in terms of a limited number of basis functions (waves). In this paper, it is shown that for the denition of accurate boundary conditions, you only need a very limited number of waves. It is demonstrated in this paper that the most important waves can be chosen based on the Modal Effective Parameter (MEP). The MEP is a product of the modal displacement value at the input DOF and the modal displacement value at the output DOF, performed for each mode. This product is then scaled by the modal mass. The MEP can be described as a measure of how much a specic global mode contributes to the transfer function between two points. As can be seen in the results in Section 3.2, the method to use waves from a WBS analysis as boundary conditions shows great promise. It is clearly shown that the WBS boundary condition gives better results for the entire investigated frequency range. While free boundary conditions can be suitable for high frequencies and less suitable for low frequencies and clamped boundary conditions can be suitable for low frequencies but not for high frequencies, the WBS boundary condition gives accurate results over the entire frequency range. There is still work to be done on nding a robust method to select the truncation limit. Another drawback of the method as presented in this paper is that an FE calculation of the full model has to be performed in order to extract the waves. The next step in this research project is to expand the method such that an FE calculation of another similar system can be used. In that way an already existing model (i.e. a predecessor of the new model) can be used for the wave extraction and the waves can then be applied to a subcomponent of the new car model already before a full FE model of the new car has been created. This will then be a powerful method for the design engineers to try out new subcomponents early in the design phase.

Acknowledgements

The European Commission is gratefully acknowledged for their support of the Marie Curie EST project SIMVIA2 (http://www.simvia2.eu, contract nr. MEST-CT-2005-020263), from which the rst author received a Research Training grant. Furthermore, we kindly acknowledge IWT Vlaanderen for their support of the ongoing research IWT-070337 MIDAS - next generation numerical tools for mid-frequency acoustics.

M EDIUM AND HIGH FREQUENCY TECHNIQUES

1593

References
[1] W. Desmet. Mid-frequency vibro-acoustic modelling: challenges and potential solutions. In Proceedings of ISMA 2002, pages 835862, Leuven, Belgium, 2002. [2] R.S. Langley and P. Bremner. A hybrid method for the vibration analysis of complex structural-acoustic systems. Journal of the Acoustical Society of America, 105(3):16571671, 1999. [3] R.H. Lyon and R.G. DeJong. Theory and Application of Statistical Energy Analysis. ButterworthHeinemann, Boston, second edition edition, 1995. [4] J.E. Manning. Use of Measured Mobility to Improve SEA Predictions in the Mid-Frequency Range. In Proceedings of DETC99, Las Vegas, Nevada, USA, 1999. Sept. 12-15. [5] J.E. Manning. SEA Models To Predict Structureborne Noise In Vehicles. In SAE Noise and Vibration Conference Proceedings, 2003-01-1542, 2003. [6] P. Ragnarsson, K. De Langhe, J. Betts, and C.T. Musser. Using FE Analysis to improve SEA at MidFrequencies. In Proceedings of Internoise 2007, Istanbul, Turkey, 2007. August 28-31. [7] G. Borello and L. Gagliardini. Virtual SEA: Towards an Industrial Process. In SAE Noise and Vibration Conference Proceedings, number 2007-01-2302, St. Charles, Illinois, 2007. [8] S. Donders, R. Hadjit, K. Cuppens, M. Brughmans, and W. Desmet. A Wave-Based Substructuring Approach for Faster Vehicle Assembly Predictions. In Proceedings of NOVEM 2005, Saint-Rapha l, e France, 2005. April 18-21. [9] S. Donders. Computer-aided engineering methodologies for robust automotive NVH design. PhD thesis, K.U.Leuven, Department of Mechanical Engineering, Division PMA, Leuven, Belgium, February 2008. available online: http://hdl.handle.net/1979/1698. [10] P. Cermelj, W. Desmet, S. Donders, B. Pluymers, and M. Boltezar. Basis Functions and Their Sensitivity in the Wave-Based Substructuring Approach. In Proceedings of ISMA 2008, Leuven, Belgium, 2008. Sept. 15-17. [11] LMS International. LMS Virtual.Lab Rev. 7B SL1, 2008. [12] MSC. MSC Nastran 2004, 2004.

1594

P ROCEEDINGS OF ISMA2008

Das könnte Ihnen auch gefallen