A Comprehensive Vehicle NVH Performance Correlation Study

Farshid Haste, Ravi Kumar, Ramandeep Gill, Sukhpal PannuFVPT NVH CAE, DaimlerChrysler Corporation, Auburn Hills MI 48326

Kuang-Jen Liu, Robert ShaverNVH Development & Engineering, DaimlerChrysler Corporation, Auburn Hills MI 48326

ABSTRACTOne of main NVH (Noise, Vibration and Harshness) deliverables for Chrysler Development System (CDS) is a fullycorrelated finite element model to be used for the referencing and the target setting process for the futureprogram. To satisfy CDS requirements, a comprehensive feasibility study using the finite element modelingtechniques along with the experimental methods to validate the CAE (Computer Aided Engineering) model hasbeen launched by NVH CAE and NVH Development and Engineering in DaimlerChrysler. The objective of thisstudy was to evaluate the fidelity of current CAE processes based on the best practice approach and to simulate awide range of vehicle NVH performances with minimal model compensation. Parallel to this effort, a full battery ofexperimental tests including modal testing, bushing rate and engine mount measurement, and standard load-casetesting were performed on a vehicle in the component, sub-system, and full vehicle level. A correlation wasestablished and proposed, and the correlation results based upon one production vehicle were analyzed. A set ofrecommendations to improve testing and simulation process was developed and presented in this paper.

INTRODUCTION

Reliable NVH Simulation: An Important Tool in Gaining Competitive Advantage

In the last decade, automotive industry has experienced a tremendous increase in the level and intensity ofcompetition. OEMs faced with global over-capacity, reduced margins and increased customer demands, areconstantly looking for innovative ways of differentiating their products in an increasingly competitive market.Automakers believe that “overall quality perception” is a major decision factor when it comes to customers’perception of a vehicle and corresponding buying decision. Vehicle noise and vibration is considered to be animportant part of quality perception. However, as a result of increased competition, automakers want to developvehicles with competitive NVH performance in shorter design cycles and with reduced development and vehiclecosts. Number of hardware prototypes has reduced dramatically and today it is more important than ever before toprovide reliable NVH simulation for all NVH vehicle performance attributes. However, due to the complexity of theNVH simulations, subjective nature of the NVH operating conditions and inherent randomness especially in highfrequencies, large-scale system-level full-vehicle NVH modeling suffers from a quality perception problem. Withoutaddressing this “perception” problem, CAE engineer’s contribution will be limited.

Chrysler Design System and NVH Simulation

To address the correlation issue with NVH models, NVH team has imbedded a special provision in the CDS [1].As part of the CDS process, before the start of any new program, NVH community should deliver a fully correlatedbaseline model to the product team simulation group. This certified model should be used in all programsimulations as a reference point. All the design directions and trade-offs made based on NVH simulation shouldbe referenced back to this model. The reference vehicle should be selected from a production vehicle with stablemanufacturing process. All the subjective ratings for NVH attributes should be performed and recorded for thatvehicle. In doing these a full set of subjective and objective baseline information will be available. Since the modelis correlated or at least the relationship between test and simulation data is known, the results from the simulationof the next level vehicle can be easily interpreted and credible recommendations to guide the design and projectthe performance of the next level vehicle can be made. To make sure the correlated model is ready for the nextgeneration vehicle program, work on the new model starts when the last program ends. Lead time is very crucial.

Minivan Correlation Project

A 2001 model year vehicle was the Chrysler current production minivan and this vehicle was used as thereference vehicle for the future minivan program. Since the NVH model of the reference vehicle is required for theprediction of the new vehicle model in the CDS process, a comprehensive vehicle NVH performance correlationstudy between the finite element model and the experimental validation testing was launched to develop arepresentative minivan vehicle model. A current production minivan was secured and dedicated to this correlationproject, and a full battery of experimental tests including all the NVH operating and non-operating tests wereperformed on the full vehicle, sub-system, and component levels.

Test Plan and Corresponding CAE Loadcases

The operating tests include the objective measurements for each of the NVH load cases in a controlled NVHlaboratory environment. The coarse road noise, rough road shake, impact harshness, and powertrain noise andvibration are the standard NVH load cases used for the benchmark of vehicle level NVH. Additional transfer pathanalysis (TPA), panel mobility, and running mode analysis were also considered for this correlation study. Thefollowing are brief descriptions of each standard NVH loadcase:

The coarse road noise test is a NVH load case to simulate the noise in the vehicle on coarse road surfaces.The noise often has qualities of boom, roar, and sometimes ringing. This test was conducted in a semi-anechoic chamber equipped with a 10-foot dynamometer, where a pre-determined coarse road surface wasbuilt-in to the chassis roll. A frequency spectrum, an equivalent to the coarse road surface, was establishedas an excitation for the NVH CAE simulation.

The rough road shake test is a NVH load case to simulate the vibration in the vehicle in low frequencies. Theshake is felt mostly in the steering wheel, seat, and floor of the vehicle. This test was conducted in a semi-anechoic chamber with a 6-foot dynamometer, where a pre-determined cleat was used on the smooth chassisroll surface. No simulation was performed for this test.

The impact harshness test is a NVH load case to simulate the ride harshness. Not only the harshness noisebut also the harshness feel are captured in this test. Again, this test was conducted in a semi-anechoicchamber with a 6-foot dynamometer, where a pre-determined cleat was used on the smooth chassis rollsurface. A frequency spectrum, an equivalent to the cleat profile, was established as an excitation for the NVHCAE simulation

The powertrain noise and vibration is a NVH load case to quantify the overall engine noise, intake noise,exhaust noise, and powertrain vibration, which comes through not only the steering wheel, seat, and floor, butalso the throttle acceleration pedal, clutch pedal, shifter, and door. This test can be conducted on the road orin an engine dynamometer, and the typical engine operating conditions are the wide-open throttle (WOT), partthrottle (PT), and idle. The engine mount excursions and the engine gas pressure data were also acquired tocorrelate with engine excitation loads for the powertrain CAE simulation. Gas pressure loads combined withthe inertial loads were used as a source of excitation for a detailed CAE model of the P/T and driveline. CAEmodel simulated the structure-borne portion of P/T noise. Air-borne portion was not accounted for.

The non-operating tests include the vehicle level modal tests for the full vehicle, trimmed body, and body-in-white(BIW), the component level modal tests for the power train, exhaust system, steering column system, seat, andsuspension and chassis components, and the body and acoustic sensitivity (A/F & P/F) measurements. All theseloads were closely replicated by CAE loads. Furthermore, all the engine mounts, transmission mounts andbushing rates were measured under different engine operating conditions in order to provide right properties fordifferent CAE simulations.

CAE Modeling Plan

The complete 2001 model year minivan CAE model was built to represent the vehicle that was tested. All the bodyand chassis components were modeled in detail. The trim components were modeled using the best practiceprinciples. Detail models of closures and seats were used in the model. There was a very good weight correlationbetween the full system CAE model and the test vehicle, with the mass of the CAE model being 1910 kgcompared to the test vehicle mass at 1950 kg. All the bushings and mounts were tested. These tested rates wereused in the CAE model. Table 1 summarizes the components that make up the full system CAE model. Figure 1shows the picture of the full system CAE model. The standard NVH loadcases of impact harshness andpowertrain noise were updated in collaboration with the NVH lab so that the loading of the CAE model for theseloadcases simulated the testing conditions in the NVH lab as closely as possible.

2001 Model year Minivan CAE Model Comments

Full System model 324392 nodes, 388121 elementsBody-in-white (BIW) 169465 nodes, 167503 elementsWelds Area welds using rigid elementsTrimmed Body-in-White 218800 nodes, 212356 elementsPower train Modal model provided by the Power train CAE groupChassis All the components modeled in sufficient detailEngine mounts and bushings 6 DOF*, Translation DOF measured, Rotational DOF calculatedSeats DMIG based on the detailed model provided by the supplierDoors DMIG based on the detailed modelTrim, Non-structural masses Modeled as concentrated masses located at CG’s**Tires Modal model provided by the supplierBolts and fasteners Modeled using rigid components

*DOF – Degrees of freedom**CG – Center of gravity

Table 1

Figure 1

Correlation Results

The correlation project generated a large amount of data. A summary of this data is presented below.

Body-in-white (BIW) normal modes correlation

Table 2 gives the summary of the normal modes for the minivan BIW with the windshield in place. Theexperimental results are given in column 2 while the CAE results are given in the column 3 of the table. The MAC(Modal Assurance Criteria – a measure of mode shape correlation) value is 70 % and above for the major modes.The CAE predicted modes are within a range of -7% to +6% of the experimental modes.

Table 2

Trimmed body-in-white normal modes correlation

Table 3 shows the summary of the normal modes correlation between the CAE (analytical) and test data for thetrimmed body-in-white. The MAC values for trimmed body-in-white (full vehicle with the exception of chassis,power train and driveline system) are smaller than those for body-in-white (refer Table 2). The MAC valuesimprove, as shown in column 5 of the table, when certain localized sub-components like the seats, fuel tank,instrument panel and closure panels are removed from the test model. The CAE predicted modes are within arange of -8% to +15% of the experimental modes.

Table 3

Body-in-white Driving Point FRF (Frequency Response Function) Correlation

In driving point FRF, the point of excitation and response are the same. As an example driving point FRF on frontleft rail is shown in Figure 2. Less than 150 Hz the correlation was very high irrespective of the mesh size and typeof load application. With local mesh refinement and distributed load, the CAE data starts correlating very well withthe test values in the higher frequency range.

Figure 2

Body-in-white transfer path correlationIn the transfer path analysis carried out for this correlation project, the body-in-white is excited at front and the rearrails and response is measured at locations on the upper radiator cross member, lower radiator cross member,front rail, mid-B-pillar, mid-floor, lower D-pillar, mid-roof, mid-of-rear header. Figure 3 shows the correlationbetween the test data and the CAE data for the mid-roof, the point furthest from the source of excitation. The CAEdata and test data follow the same trends.

Figure 3

Impact harshness correlation

Figure 4 shows the correlation for the middle seat front outboard anchor on the driver side of the vehicle for impactharshness. The CAE loads simulate the front wheels out-of-phase load input to the vehicle on the 6-footdynamometer with the cleat as explained above in the test plan. The CAE response correlates very well with thetest data in the range if interest, especially at 14 Hz peak in the spectrum.

Figure 4

Coarse road noise correlationFigure 5 shows the correlation between the test and CAE data, for sound pressure measured at driver’s left earwith coarse road input. The improvement in correlation with the current CAE model, which has the updated trimand bushings, over the legacy (old) CAE model in the whole frequency range, is noticeable.

Figure 5

Powertrain noise correlationThe sound pressure correlation measured at driver’s right ear for 2nd order part throttle powertrain loads is shownin Figure 6. Overall for the powertrain noise loadcase, the degree of correlation was not as good as the otherloadcases.

Figure 6

CONCLUSIONS ANDS LESSONS LEARNED

MODELINGIn our overall evaluations of correlation results, it was agreed that the best practice NVH FEA modeling standardsas established in DaimlerChrysler Chrysler group are reliable. Under majority of loads, the model showed goodcorrelation within the expected test to test variation range. Some more specific conclusions are as follows:

Low Damped Sub-systems: Overall level of correlation for sub-systems like BIW, were from good toexcellent, considering the test-to-test variation range. The typical 25 mm element size with area weldrepresentation and a 1-2% damping provided excellent results for both modal and FRF analysis up to 100 Hz.However, to assure accuracy of point mobility analysis in higher frequencies, 100 to 400 Hz, a detailedrepresentation of force application point is paramount. Transfer mobility values were less sensitive to localmeshing modifications in either load application or measurement points.

High Damped Sub-Systems: Gaining overall mode shape correlation for a sub-system like Trim-BIW provedto be difficult. However, further analysis showed that the low MAC values for some modes were affected bysome heavily damped components like seats. More attention should be spent on modeling high-dampedcomponents with built-in mechanisms like seats.

Low Frequency Shake Analysis: Full vehicle system model results correlated well with most of test results inlow frequency road shake events like Road Harshness. Most important factor in deciding on correlation was tomake sure the test procedure and post-processing are in synch with the corresponding CAE procedure andvice versa.

High Frequency Road Noise: The updated model showed remarkable improvement in comparison to thelegacy model. We used both random and deterministic load applications and we found that correlation did notimprove enough to justify the more complex random analysis. The road noise results also showed goodreliability of supplier provided tire modal models even in higher frequencies.

High Frequency P/T Structure-borne Noise: In comparing the test and CAE results for the P/T noise, wehad to deal with a testing limitation: the test procedure to measure the gas pressure data was different fromthe one used to measure the P/T noise inside the cabin. When the CAE results based on a flexible P/T model,detailed vehicle system model was compared to the measured noise, the results were not as satisfactory asroad induced loadcases. Even when P/T model was by-passed and direct engine-side of the engine mountdata was used, the correlation to the available test data did not improve much. It is recommended that furthertesting should be conducted to resolve this issue.

TESTING:

Bushing and Engine Mount Measurements: Measuring all the bushings and engine mount rates underdifferent pre-loads proved to be a complicated task. Defining multi-axial preloads by using the CAE model andcorresponding bookkeeping took a lot of time and resources and can not be trivialized or left unsupervised to athird party.

In-Vehicle P/T Gas Pressure Measurement: It is recommended to include the gas pressure measurementand engine mount excursion in the standard in-vehicle powertrain NVH testing.

RECOMMENDATIONS:

Importance of Detail CAE Modeling: With improved pre and post processing tools and availability of robusttools to reduce detailed components to dense representation, like DMIGs, it is important that all the significantstructural and non-structural components of the vehicle be represented to avoid the guess work by the CAEengineer. Obviously, it is the duty of CAE manager to balance this requirement with the analytical capacity oftheir overall system. A good first test for any vehicle model is to see how close it represents the actual vehicleweight. Our work proves that a very good weight correlation is attainable with a reasonable effort.

Importance of Standard and Representative Test Procedures: The procedure for testing operations,placement of transducers and post-processing analysis should be standardized and coordinated with all thestakeholders including the simulation team. It is essential that the standard test procedure be representative ofwhat customer will experience on the road.

CAE and Testing Community Coordination: Without close coordination between testing and simulationcommunity, major progress in the quality of NVH simulation is not possible. The interdependency is too high tobe left to casual person-to-person interactions. As our project results showed, this coordination is achievablewith an organizational commitment.

ACKNOWLEDGMENT:We like to acknowledge the work and support provided by our current and previous colleagues in the course ofthis project.

Mr. Tamim Arif and Mr. Sohail Rana for their outstanding CAE work. Mr. William Danos II for his great testingsupport. Mr. Greg Goetchius for his help in starting the project and guidance. Mr. Kenneth Buczek, manager ofNVH laboratory, for his continuing support and Mr. Joe Vitous, FVPT/CAE Manager, for his allocation of resourcesand support.

REFERENCES1. Dong, B., et. Al., “Process to Achieve NVH Goals: Subsystem Targets via “Digital Prototype” Simulations,” SAENoise and Vibration Conference, SAE 1999-01-1692, 1999.