SAE 2015 NVH Conference Structure Borne NVH WorkshopSource 2015 SAE NVC Structure Borne NVH Workshop Slide 8 Frequency Range:up to 1000 Hz System Characterization • Source of Excitation

SAE 2015 NVH ConferenceStructure Borne NVH Workshop

Alan DuncanAlan Duncan Siemens PLM @ HondaNVH Simulation Specialist

2015 SAE NVC Structure Borne NVH Workshopwww.AutoAnalytics.com

Slide 1Slide 1

NVH Simulation Specialist

Greg GoetchiusGreg Goetchius AtievaSr Technical Specialist, NVH

Jianmin GuanJianmin Guan Altair Engineering NVH Manager

Contact Email: [email protected]



Structure Borne NVH WorkshopWorkshop Objectives -1. Review Basic Concepts of Automotive Structure Bor ne Noise.

2. Propose Generic Targets.

3. Present New Technology Example.


Slide 2Slide 2

Intended Audience –• New NVH Engineers.

• “Acoustics” Engineers seeking new perspective.

• “Seasoned Veterans” seeking to brush up skills.

Structure Borne NVH Workshop• Introduction

• Low Frequency Basics

•LIVE Noise Attenuation Demo


Slide 3Slide 3

•LIVE

• Mid Frequency Basics

• New Technology

An Overview of Stochastic Simulation

• Closing Remarks



• LIVE Noise Attenuation Demo


Slide 4Slide 4


• New Technology


• Closing Remarks

Rideand

Handling

ImpactCrashWorthiness

Competing Vehicle Design Disciplines


Slide 5Slide 5

NVH Durability

Automotive Engineering Objectives are Timeless


Slide 6Slide 6

Structure Borne Noise and Vibration

Vibrating

Source


Slide 8Slide 8

Frequency Range: up to 1000 Hz

System Characterization• Source of Excitation

• Transmission through Structural Paths

• “Felt” as Vibration

• “Heard” as Noise

Structure Borne NoiseAirborne Noise

Res

pons

eAutomotive NVH Frequency Range


Slide 9Slide 9

Res

pons

e

Log Frequency

“Low”

Global Stiffness

“Mid”

Local Stiffness+

Damping

“High”

Absorption+

Mass+

Sealing+

Damping

~ 150 Hz ~ 1000 Hz ~ 10,000 Hz

Low Frequency Basics• Source -Path-Receiver Concept

• Single DOF System Vibration

• NVH Source Considerations

• Receiver Considerations


Slide 10Slide 10


• Vibration Attenuation StrategiesProvide Improved IsolationMode ManagementNodal Point MountingDynamic Absorbers






Slide 11Slide 11



RECEIVER

Structure Borne NVH Basics


Slide 12Slide 12

PATH

SOURCE






Slide 13Slide 13



APPLIED FORCE

F = FO sin 2 π f t

TR = F / F

Single Degree of Freedom Vibration

2n)ff(21 ζ+=


Slide 14Slide 14

m

k c FT

TR = FT / F

TransmittedForce

= fraction of critical damping

fn = natural frequency

f = operating frequency

( )2n

22n

2n

ff2)ff(1)ff(21ζ

ζ+−

+=

ζmk

3

4

Tran

smis

sibi

lity

Rat

ioVibration Isolation Principle

m

APPLIED FORCE

F = FO sin 2 π f t

k c F

TR = FT / F0.25

0.15

0.1


Slide 15Slide 15

0 1 2 3 4 50

1

2

Tran

smis

sibi

lity

Rat

io

1.414Frequency Ratio (f / fn)

k c FTTransmitted

Force

Isolation Region

1.0

0.5

0.375






Slide 16Slide 16



NVH Source Considerations


Slide 17Slide 17

Two Main Sources

Suspension Powertrain

Typical NVH Pathways to the Passenger


Slide 18Slide 18

PATHS FOR

STRUCTURE BORNE

NVH

Structure Borne NVH Sources


Slide 19Slide 19

Structure Borne NVH Sources

Primary Consideration:

Reduce the Source first as much as possible because whatever enters the


Slide 20Slide 20

possible because whatever enters the structure is transmitted through multiple paths to the receiver.

Transmission through multiple paths is more subject to variability.






Slide 21Slide 21



Receiver ConsiderationsSubjective to Objective Conversions

Subjective NVH Ratings are typically based on a 10 Point Scale resulting from Ride Testing

A 2 ≈≈≈≈ 1/2 A 1

Receiver Sensitivity is a Key Consideration


Slide 22Slide 22

A 2 ≈≈≈≈ 1/2 A 1

Represents 1.0 Rating Change

TACTILE: 50% reduction in motion

SOUND : 6.dB reduction in sound pressure level ( long standing rule of thumb )

� � � � � � � � � �

6.0

1.0

5.0

Human Sensitivity to Tactile Vibration


6.0

10 Pt Tactile Rating

Equation:

Standard Rating =

8.19 - 4.34 * LOG10(V)

V; Velocity in mm/sec

5.0

6.0

Sou

nd P

ress

ure

Leve

l -

dB

Loudness Level - Phons

Human Sensitivity to Sound Level

1.0


Slide 24

6.0

Frequency - Hz

Sou

nd P

ress

ure

Leve

l

Range ofInterest

Minimum AudibleField Limit

10 Pt AcousticRating

Equation:

Standard Rating =

13.6 - (0.175) * SPL

SPL; Sound Pressure Level, dBA

� � � � � � � � � � � � � � � � � � � � � � � � � � �






Slide 25Slide 25



Symbolic Model of Unibody Passenger Car8 Degrees of Freedom

318

2 4

5


Slide 26Slide 26

Total 2178.2 Kg (4800LBS)Mass Sprung 1996.7 Kg

Unsprun g 181.5 Kg (8.33% of Total)Powertrain 181.5 Kg

Tires 350.3 N/mmKF 43.8 N/mmKR 63.1 N /mmBeam mass lumped on grids like a beam M2,3,4 =2 * M1,5

From Reference 6

316 7

5

1 2 4 5

8

3

Suspension

Engine Mass

EngineIsolator

Flexible Beam for Body

8 Degree of Freedom Vehicle NVH Model


Slide 27Slide 27

6 7

Tires

Wheels

SuspensionSprings

8 Degree of Freedom Vehicle NVH ModelForce Applied to Powertrain Assembly

Feng

1 2 4 5

8

3


Slide 28Slide 28

Forces at Powertrain could represent a First OrderRotating Imbalance

6 7

Engine Isolation Example

Response at M id Car

0.0100

0.1000

1.0000

Vel

ocity

(m

m/s

ec)

Constant Force Load; F ~ A 15.9 Hz

8.5 Hz

7.0 Hz


Slide 29Slide 29

0.0001

0.0010

0.0100

5.0 10.0 15.0 20.0

Frequency Hz

Vel

ocity

(m

m/s

ec)

700 Min. RPM First Order UnbalanceOperation Range of Interest

318

2

6

4

753311

8822

66

44

7755

Engine Isolation Example

Response at M id Car

0.0100

0.1000

1.0000

Vel

ocity

(m

m/s

ec)

Constant Force Load; F ~ A

Engine Idle Speed Operating Shapes

15.9 Hz

8.5 Hz


Slide 30Slide 30

0.0001

0.0010

0.0100

5.0 10.0 15.0 20.0

Frequency Hz

Vel

ocity

(m

m/s

ec)

700 Min. RPM First Order UnbalanceOperation Range of Interest

318

2

6

4

753311

8822

66

44

7755

Operating Shapes at 700 RPM 7.0 Hz

Concepts for Increased Isolation“Double” isolation is the typical strategy for further improving isolation of a given vehicle design.

Second Level of Isolation is at Subframe


Slide 31Slide 31

Suspension Bushing is first level

Subframe is Intermediate Structure

Isolation is at Subframe to Body Mount

8 Degree of Freedom Vehicle NVH ModelRemoved Double Isolation Effect

1 2 4 5

8

3


Slide 32Slide 32

WheelMass

Removed6 7

Double Isolation ExampleVertical Response at DOF3

3.0E+00

4.0E+00

5.0E+00

6.0E+00

Vel

ocity

(m

m/s

ec)

Base Model

Without Double_ISO

318

2

6

4

753311

8822

66

44

7755


Slide 33Slide 33

0.0E+00

1.0E+00

2.0E+00

3.0E+00

5.0 10.0 15.0 20.0

Frequency Hz

Vel

ocity

1.414*fn






Slide 34Slide 34



Mode Management Chart

0 5 10 15 20 25 30 35 40 45 50HzFirst Order Wheel/Tire Unbalance V8 Idle

Hot - Cold

EXCITATION SOURCESInherent Excitations (General Road Spectrum, Recipr ocating Unbalance, Gas Torque, etc.)Process Variation Excitations (Engine, Driveline, A ccessory, Wheel/Tire Unbalances)

CHASSIS/POWERTRAIN MODESSuspension Hop and Tramp Modes


Slide 35Slide 35

Hz

Hz0 5 10 15 20 25 30 35 40 45 50

0 5 10 15 20 25 30 35 40 45 50

Ride ModesPowertrain Modes

Suspension Longitudinal ModesExhaust Modes

BODY/ACOUSTIC MODES

Body First Bending First Acoustic Mode

Steering Column First Vertical BendingBody First Torsion

(See Ref. 1)

8 Degree of Freedom Vehicle NVH ModelBending Mode Frequency Separation

Beam Stiffness was

1 2 4 5

8

3


Slide 36Slide 36

Beam Stiffness was adjusted to align Bending

Frequency with Suspension Modes and then

progressively separated back to Baseline.

6 7

Response at Mid Car

10.00

100.00

Vel

ocity

(mm

/sec

)

18.2 Hz Bending

13.Hz Bending

10.6 Bending

8 DOF Mode Separation Example

18.2 Hz

13.0 Hz

10.6 Hz318

2

6

4

753311

8822

66

44

7755


Slide 37Slide 37

0.10

1.00

5 10 15 20Frequency Hz

Vel

ocity

(mm

/sec

)

Response at Mid Car

10.00

100.00

Vel

ocity

(mm

/sec

)

18.2 Hz Bending

13.Hz Bending

10.6 Bending

8 DOF Mode Separation Example

318

2

6

4

753311

8822

66

44

7755

Bending @ 10.6 Hz

Bending @ 13.0 Hz


Slide 38Slide 38

0.10

1.00

5 10 15 20Frequency Hz

Vel

ocity

(mm

/sec

)

All Operating Shapes at 10.6 Hz Bending

@ 18.2 Hz






Slide 39Slide 39



First Bending : Nodal Point Mounting ExampleMount at Nodal Point


Slide 40Slide 40

Front input forces Rear input forces

Locate wheel centers at node points of the first be nding modeshape to prevent excitation coming from suspension input motion.

Side View

First Torsion: Nodal Point Mounting ExamplesMount at Nodal Point


Slide 41Slide 41

Passenger sits at node point for First Torsion.

Side View

Transmission Mount of a3 Mount N-S P/T is nearthe Torsion Node.

Rear View

Engine

Powertrain Bending Mode Nodal Mounting

1 2 4 53


Slide 42Slide 42

Mount system is placed to support Powertrain at the Nodal Locations of the First order Bending Mode. Best compromise with Plan View nodes should also be considered.

6 7

1 2 4 5

8

3

8 Degree of Freedom Vehicle NVH ModelBending Node Alignment with Wheel Centers

Redistribute Beam Masses to move Node Points to

Align with points 2 and 4


Slide 43Slide 43

6 7

Response at Mid-Car

2.0E+00

3.0E+00

4.0E+00

Vel

ocity

(m

m/s

ec)

Node ShiftedBase Model

First Bending Nodal Point Alignment

318

2

6

4

753311

8822

66

44

7755


Slide 44Slide 44

0.0E+00

1.0E+00

5.0 10.0 15.0 20.0Frequency Hz

Vel

ocity

Response at Mid-Car

2.0E+00

3.0E+00

4.0E+00

Vel

ocity

(m

m/s

ec)


First Bending Nodal Point Alignment

318

2

6

4

753311

8822

66

44

7755

Operating Shapes at

18.2 Hz


Slide 45Slide 45

0.0E+00

1.0E+00

5.0 10.0 15.0 20.0Frequency Hz

Vel

ocity

Node Shifted Model

Diagnosis: Increase at 10.2 Hz


Slide 46

Response at Mid-Car

0.0E+00

1.0E+00

2.0E+00

3.0E+00

4.0E+00

5.0 10.0 15.0 20.0Frequency Hz

Vel

ocity

(m

m/s

ec)


Motion Experienced when Bending is Present

Motion Experienced when Bending is Removed

Diagnosis: Increase at 10.2 HzUp Position


Slide 47

Center - Undeformed Position

Down Position






Slide 48Slide 48



YO

xSDOF

Dynamic Absorber Concept

M

YO

x

Auxiliary Spring-Mass-Damperm = M / 10

2DOFM


Slide 49Slide 49

Powertrain Example of Dynamic Absorber

Anti-Node Identifiedat end of Powerplant


Slide 50Slide 50

k c

Absorber attached at anti-node acting in the Vertical and Lateral plane.

Tuning Frequency = √√√√ k/m

m

[Figure Courtesy of DaimlerChrysler Corporation][Figure Courtesy of DaimlerChrysler Corporation][Figure Courtesy of DaimlerChrysler Corporation][Figure Courtesy of DaimlerChrysler Corporation]

Baseline Sound Level63 Hz Dynamic Absorber63 + 110 Hz Absorbers

10 dB


Slide 51Slide 51[Figure Courtesy of DaimlerChrysler Corporation][Figure Courtesy of DaimlerChrysler Corporation][Figure Courtesy of DaimlerChrysler Corporation][Figure Courtesy of DaimlerChrysler Corporation]

Low Frequency Basics - Review• Source -Path-Receiver Concept





Slide 52Slide 52



Presentation and References at www.AutoAnalytics.com DOWNLOAD Link



•Live Noise Attenuation Demo


Slide 53Slide 53

•Live


• New Technology


• Closing Remarks

GregGoetchius

Toolbox Demo Noise Test Results

85

61

68

63

70

80

90

SP

L (d

BA

)Tool Box Demo Test Results

85

6168

63


61 63

55

47

39

30

40

50

60

1) Baseline:Imbalance, No

Isolation

2) Imbalance +Isolation

3) No Imbalance,No Isolation

4) No Imbalance,No Isolation +

Damping

5) No Imbalance +Isolation +Damping

6) #5 + Absorption 7) #6 + InsulatorMat

SP

L (d

BA

) 6155

47

39



•LIVE Noise Attenuation Demo


Slide 58

•LIVE


• New Technology

Uncertainty and NVH Scatter

• Closing Remarks

Alan Duncan

Mid Frequency NVH Fundamentals


Slide 59

This looks familiar!Frequency Range of Interest has changed to

150 Hz to 1000 Hz

Typical NVH Pathways to the Passenger


Slide 60

PATHS FOR

STRUCTURE BORNE

NVHNoise Paths are thesame as Low

Frequency Region

Mid-Frequency Analysis CharacterStructure Borne Noise

Airborne Noise

High modal densityand coupling in

source, path andreceiver

� Body Mode separation is less practical in the Mid-Frequency range

� Mode separation in the Chassis is still possiblefor 1 st Order modes.

� NEW STRATEGY: Control Interaction betweenSource and Receiver ( Chassis and Body )


Slide 61

Res

pons

e

Log Frequency

“Low”

Global Stiffness

“Mid”

Local Stiffness+

Damping

“High”

Absorption+

Mass+

Sealing

~ 150 Hz ~ 1000 Hz ~ 10,000 Hz

Mid-Frequency Analysis Character

Control Measures for Mid Frequency Concerns

Effective Isolation


Slide 62

Attenuation along Key Noise Paths



Effective Isolation


Slide 63


Classical SDOF: Rigid Source and Receiver

“Real Structure”Flexible (Mobile)Source and Receiver

Isolation Effectiveness

1.0


Slide 64

1.0 1.414 10.0

f / f n

Effectiveness deviates from the classical developme nt as resonances occur in the receiver structure and in the foundati on of the source.

Isolation Region

Mobility• Mobility is the ratio of velocity response at the excitation point on structure

to the force applied in the direction of the veloci ty

Mobility =Velocity

Force

• Mobility, related to Admittance , characterizes Dynamic Stiffness of the structure at load application point


Slide 65

the structure at load application point

Mobility =Frequency * Displacement

Force

=Frequency

Dynamic Stiffness

• The isolation effectiveness can be quantified by a theoretical model based on analysis of mobilities of receiver, isolator and so urce

• Transmissibility ratio is used to objectively defin e measure of isolation

TR =Force from source without isolator

Force from source with isolator

Isolation

V rReceiverV rReceiverV

V


Slide 66

V ir

V is

F r

F ir

Source

F is

Fs V s

Isolator

VF s =

Y i + Y r + Y s

V r

F r

Receiver

Source

F s V s

VF s =

Y r + Y s

(See Ref. 8)

TR = ( Y r + Y s ) //// ( Y i + Y r + Y s )

Y r : Receiver mobility

Y :

V m

V im

V if

F m

F im

Receiver

F if

Isolator

TR =Force from source without an isolator

Force from source with an isolator

Isolation


Slide 67

• For Effective Isolation (Low TR) the Isolator Mobility must exceed the sum of the Source and Receiver Mobilities.

Y s : Source mobility

Y i : Isolator mobility

Source

F if

F f V f

Recall that K 1Y ∝

TR = ( ) //// ( ) K body

1K source

1+

K body

1+

K iso

1K source

1+

K iso

K sourceK body

K iso1.0 5.0 20.0

1.0 0.67 0.55 0.51

Designing Noise Paths


Slide 68

5.0

20.0

0.55 0.29 0.20

0.51 0.20 0.09

Generic targets:body to bushing stiffness ratio of at least 5.0source to bushing stiffness ratio of at least 20.0

Body -to-Bushing Stiffness RatioRelationship to Transmissibility

Tran

smis

sibi

lity

Rat

io T

R

0.3

0.4

0.5

0.6

Target Min. = 5 gives TR = .20

For a source ratio of 20


Slide 69

Stiffness Ratio; K body / K iso

Tran

smis

sibi

lity

Rat

io T

R

0

0.1

0.2

0.3

1 2 3 4 5 6 7 8 9 10

gives TR = .20



Effective Isolation


Slide 70


Identifying Key NVH PathsKey NVH paths are identified by Transfer Path Analy sis (TPA)

Fi

TactileTransfer

AcousticTransfer

Break the system at the points where the forces


Slide 71

Operating loads Operating loads

points where the forces enter the body (Receiver)

Total Acoustic Response is summation of partial res ponses over all noise paths

Pt = ΣΣΣΣ paths [P i ]

Identifying Key NVH Paths

Fi Operating loads createForces (F i) into body atAll noise paths

F F

F FF

F F

P = ΣΣΣΣ [P ]

P/V

(Kbody )V/F P/F

= ΣΣΣΣ [ F * (P/F) ]


Slide 72

Pt = ΣΣΣΣ paths [P i ]

Key NVH Measures of Interest

PiPt = f ( (V/F)iFi (P/F)i )

= ΣΣΣΣ paths [ F i * (P/V) i * (V/F)i]

= ΣΣΣΣ paths [ F i * (P/F) i ]

Identifying Key NVH PathsTPA Example: Dominant Paths over Frequencies

Pi


Slide 73

P/T Load

Crank torque

Identifying Key NVH PathsTPA Example: Contribution at One Transfer Path

Partial response from a particular path: P i = (P/F)i * Fi

P/T Load

Crank torque

Pt

Pi Peak at 91 Hz


Slide 74

(P/F)i

�

Fi

�

Pi

*

Peak at 91 Hz

High Forces

Ranking Key NVH PathsTPA Example: Sum of Key Transfer Paths at One Peak

Total Response: P t = ΣΣΣΣ paths [P i ]

TO

TA

L SO

UN

D


Slide 75

P/T Load

Crank torqueP/T Load

Crank torque

TO

TA

L SO

UN

D

Path N

o. 3

Path N

o. 1

Path N

o. 2

REMAINING PATHS

Recall for Acoustic Response Pt

Pt = ΣΣΣΣ paths [P i ] = ΣΣΣΣ paths [ F i * (P/V) i * (V/F)i]

“ Downstream ” Effects: Body Panels

(P/V)i �� “Downstream” (Body Panel) System Dynamics: Three Co ncepts:

1. Panel Acoustic Contribution

Defines “the where”(End of the Structural Path)

Increased Stiffness

2. Panel StiffnessIncreasing Stiffness


Slide 76

(End of the Structural Path)

Increased Damping

3. Panel DampingIncreasing Damping

Generic Noise Path Targets

K iso

K body

> 5.0K iso

K source

> 20.0

StructuralMobility < 0.2 to 0.3 mm/sec/N


Slide 77

AcousticSensitivity

< 50 - 60dBL/N

Panel Damping Loss Factor

> .10 - .20

Primary : Minimize the Source Force

< 1.0 N

Final Remarks on Mid Frequency Analysis

• Effective isolation at dominant noise paths is crit ical• Effective isolation at dominant noise paths is crit ical

• Reduced mobilities at body & source and softenedbushing are key for effective isolation

• Reduced mobilities at body & source and softenedbushing are key for effective isolation

• Mode Separation remains a valid strategy as modes• Mode Separation remains a valid strategy as modes


Slide 78

• Other means of dealing with high levels of response[@ Source: Tuned dampers & Isolator

placement at nodal locations ] and[@ Receiver: damping treatments] are effective

• Other means of dealing with high levels of response[@ Source: Tuned dampers & Isolator

placement at nodal locations ] and[@ Receiver: damping treatments] are effective

• Mode Separation remains a valid strategy as modesin the source structure start to participate

• Mode Separation remains a valid strategy as modesin the source structure start to participate

Structure Borne NVH: Concepts Summary

• Source -Path-Receiver as a system1. Reduce Source

2. Rank and Manage Paths

3. Subjective Response: Human Sensitivity

• Mode Management

• Source -Path-Receiver as a system1. Reduce Source

2. Rank and Manage Paths

3. Subjective Response: Human Sensitivity

• Mode Management


Slide 79

• Mode Management

• Nodal Point Placement

• Effective Isolation (Source and Receiver)

• “Downstream” (Body Panel) Considerations

• Mode Management

• Nodal Point Placement

• Effective Isolation (Source and Receiver)

• “Downstream” (Body Panel) Considerations

This presentation is an extension to the ‘NVH Scatter and Uncertainty’ topic from the 2013 workshop.

• Conceptual review of select Stochastic

Simulation studies



Slide 81

• Implications of Stochastic Simulation

• Approaches to Stochastic Simulation


Simulation studies




Slide 82


Two categories of Stochastic Simulation:• A typical stochastic simulation process

Review of Stochastic Simulation


Slide 83

• Quantify noise factors in loads

• Quantify noise factors in design

• Identify control factors

• Perform Monte-Carlo simulation

• Evaluate vehicle population based

metrics

A Typical Stochastic Simulation Process

Noise Factors


Slide 84

metrics

• Optimize vehicle tuning

Applying Six Sigma Tools to the Rear Driveline Syst em for Improved Vehicle Level NVH Performance, Amit Zutshi, etc.

Response SpreadParameter Effects

Two categories of Stochastic Simulation:• A typical stochastic simulation process• Capturing worst case combinations of ‘error

states’ – loads variations� Driveline imbalance� Wheel/Tire force variations



Slide 85

� Wheel/Tire force variations

Capturing the ‘Worst Case’ in DrivelineImbalance

An Application of Variation Simulation – Predicting Interior Driveline Vibration Based on Production Va riation of Imbalance and Runout, Glenn A. Meinhardt, etc.


Slide 86

Planes of Imbalance Phasing of Imbalance

Imbalance Distribution ‘Worst Case’ Response

Capturing the ‘Worst Case’ in Wheel/Tire Force Variations

Determination of Interior NVH Levels from Tire Whee l Variations using a Monte Carlo Process, M. Qatu, etc


Slide 87

Types of Forces Application of Forces

Distribution of Variation‘Worst Case’ Response

Comments:• The ‘Worst Case’ response cannot be found from te sting prototypes • It can be found only thru stochastic simulation b y analyzing sufficient no. of samples





Slide 88


• Capturing experimentally observed response spread – design variations

� Parametric modeling� Non-parametric modeling

Magnitude of 99 Structure Borne Noise Transfer Func tions for Rodeo’s at the Driver MicrophoneMeasurements from Kompella and Bernhard ( Ref. 8 ) 1993 Society of Automotive Engineers, Inc.

Review of Slides from the 2013 WS


Slide 89

Frequency ( Hz )

12 dB Variation

Freymann, BMW NVH Scatter Results



Slide 90

Frequency ( Hz )

Acoustic Scatter from Simulation of the vibroacoustic behavior of a vehicle due to possible tolerances in the component area and in the production process.

© 2000 Society of Automotive Engineers, Inc.Reproduced with permission from paperby Freymann, et. Al. (Ref. 9)

Experimentally detected Scatter in low frequency vibroacoustic behavior of production vehicles.• Scatter is the “Physics”

• NVH is unique in this hyper-sensitive behavior to small variations

Capturing the Response Spread Parametric Study


Slide 91

Comments: • Limited to Few Discrete Variations around the Loc al Design Change• Brute Force – Full Vehicle Runs to Produce Bands

P/T Load2nd Order

95% UprNominal 50% Mean95% Lwr

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

Capturing the Response Spread Non-Parametric Study


Slide 92

� � � � � � � � � ! " # $ % � & % � ' ( ) * % ( � + ) " % ( $ + + � ! , ( " ( $ � & " - � . * - # * % , - ( % / $ ( ) ( ) * * 0 ! * # $ * & ( % ' � # ( ) * 1 � � $ & 2

& � $ % * � � # " ! ) % � ' ( ) * � . , - $ � ' ( ) * � 3 � % $ & . 4 5 6 7 % + " - * 8 * 0 ! * # $ * & ( % ' � # ( ) * � � + " # % 5 ( ) $ & 2 # " 9 - $ & * % 7 : * " &

; " - , * � ' ( ) * * 0 ! * # $ * & ( % 5 ( ) $ + < 2 # " 9 - $ & * 7 : = * " & + � ! , ( " ( $ � & " - � . * - 5 . " % ) * . - $ & * 7 : = * " & ; " - , * � ' ( ) *

# " & . � # * % ! � & % * 5 $ . ( ) $ & % � - $ . - $ & * 7 : > ? @ + � & ' $ . * & + * # * 2 $ � & 8 ( ) * , ! ! * # " & . - � / * # * & ; * - � ! * % " # * ( ) * , ! ! * #

" & . - � / * # ( ) $ + < % � - $ . - $ & * % �

95% Lwr

NOTE: The Method has Quantified the Model and Manufacturi ng Combined Uncertainty.

Comments:• The 95% Confidence Bands encapsulate the Measured Scatter of 20 Vehicles.• Half-Bandwidth Scatter (between Mean and Upr 95%) is similar to K-B results.





Slide 93


• Capturing experimentally observed response spread – design variations

� Parametric modeling� Non-parametric modeling

Many OEMS are actively pursuing some form of stochastic simulation


Simulation studies




Slide 94



• Implications to CAE model validation

• Implications to CAE driven design

decision making

Implications of Stochastic Simulation


Slide 95

• Implications to CAE based problem

diagnostics



decision making



Slide 96


diagnostics

Test Variation Band10. dB; 50-150 Hz20. dB; 150-500 Hz

Reference Baseline Confidence CriterionFor Operating Response Simulations

REF. 8FUDGE

FACTORS

Sou

nd F

RF



Slide 98

Test Upper BoundTest Band AverageTest Lower Bound

Confidence Criterion:Simulation result mustfall within the band oftest variation.

Simulation Prediction

Sou

nd F

RF

Frequency Hz

However, obtaining a band of test variation is beco ming less practical by day, thus the only hope is to cre ate a band from CAE thru stochastic simulation

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

CAE Model Validation

P/T Load2nd Order



Slide 99

� � � � � � � � � ! " # $ % � & % � ' ( ) * % ( � + ) " % ( $ + + � ! , ( " ( $ � & " - � . * - # * % , - ( % / $ ( ) ( ) * * 0 ! * # $ * & ( % ' � # ( ) * 1 � � $ & 2

& � $ % * � � # " ! ) % � ' ( ) * � . , - $ � ' ( ) * � 3 � % $ & . 4 5 6 7 % + " - * 8 * 0 ! * # $ * & ( % ' � # ( ) * � � + " # % 5 ( ) $ & 2 # " 9 - $ & * % 7 : * " &

; " - , * � ' ( ) * * 0 ! * # $ * & ( % 5 ( ) $ + < 2 # " 9 - $ & * 7 : = * " & + � ! , ( " ( $ � & " - � . * - 5 . " % ) * . - $ & * 7 : = * " & ; " - , * � ' ( ) *

# " & . � # * % ! � & % * 5 $ . ( ) $ & % � - $ . - $ & * 7 : > ? @ + � & ' $ . * & + * # * 2 $ � & 8 ( ) * , ! ! * # " & . - � / * # * & ; * - � ! * % " # * ( ) * , ! ! * #

" & . - � / * # ( ) $ + < % � - $ . - $ & * % �

Comments:• There is no physical sample corresponding to the no minal design• One should generally not expect the nominal respons e to match any

single test sample

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �


P/T Load2nd Order


TestMeasurement


Slide 100

� � � � � � � � � ! " # $ % � & % � ' ( ) * % ( � + ) " % ( $ + + � ! , ( " ( $ � & " - � . * - # * % , - ( % / $ ( ) ( ) * * 0 ! * # $ * & ( % ' � # ( ) * 1 � � $ & 2

& � $ % * � � # " ! ) % � ' ( ) * � . , - $ � ' ( ) * � 3 � % $ & . 4 5 6 7 % + " - * 8 * 0 ! * # $ * & ( % ' � # ( ) * � � + " # % 5 ( ) $ & 2 # " 9 - $ & * % 7 : * " &

; " - , * � ' ( ) * * 0 ! * # $ * & ( % 5 ( ) $ + < 2 # " 9 - $ & * 7 : = * " & + � ! , ( " ( $ � & " - � . * - 5 . " % ) * . - $ & * 7 : = * " & ; " - , * � ' ( ) *

# " & . � # * % ! � & % * 5 $ . ( ) $ & % � - $ . - $ & * 7 : > ? @ + � & ' $ . * & + * # * 2 $ � & 8 ( ) * , ! ! * # " & . - � / * # * & ; * - � ! * % " # * ( ) * , ! ! * #

" & . - � / * # ( ) $ + < % � - $ . - $ & * % �

Measurement

Author:Focus of Correlation Effort

All Test Peaks in Band: Results imply test and CAE results are consistent

Comments:• Test CAE comparison can be properly understood only in the context of

a population based analysis• The population context is key to avoid falling into the trap of thinking in

overly simplistic terms about CAE model confidence

• Sample to sample validation• At the vehicle level, one CAE sample cannot be expe cted to

match one random test sample

• Model updating to test - leads to a model of the tes t sample



Slide 101

• Sample to sample validation• At the vehicle level, one CAE sample cannot be expe cted to

match one random test sample

• Model updating to test - leads to a model of the tes t sample

• Population to population validation



Slide 102

• Population to population validation• Population based validation – CAE population to a

reasonable number of test samples

• What if the populations don’t match? – need to start from the

ground up (See Dascotte Ref. #20)• Components – less affected by variation

• Subsystem and vehicle – more and more affected by va riation



decision making



Slide 103


diagnostics

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

CAE Driven Design Decision Making

P/T Load2nd Order


TestMeasurement


Slide 104

� � � � � � � � � ! " # $ % � & % � ' ( ) * % ( � + ) " % ( $ + + � ! , ( " ( $ � & " - � . * - # * % , - ( % / $ ( ) ( ) * * 0 ! * # $ * & ( % ' � # ( ) * 1 � � $ & 2

& � $ % * � � # " ! ) % � ' ( ) * � . , - $ � ' ( ) * � 3 � % $ & . 4 5 6 7 % + " - * 8 * 0 ! * # $ * & ( % ' � # ( ) * � � + " # % 5 ( ) $ & 2 # " 9 - $ & * % 7 : * " &

; " - , * � ' ( ) * * 0 ! * # $ * & ( % 5 ( ) $ + < 2 # " 9 - $ & * 7 : = * " & + � ! , ( " ( $ � & " - � . * - 5 . " % ) * . - $ & * 7 : = * " & ; " - , * � ' ( ) *

# " & . � # * % ! � & % * 5 $ . ( ) $ & % � - $ . - $ & * 7 : > ? @ + � & ' $ . * & + * # * 2 $ � & 8 ( ) * , ! ! * # " & . - � / * # * & ; * - � ! * % " # * ( ) * , ! ! * #

" & . - � / * # ( ) $ + < % � - $ . - $ & * % �

Measurement

Comments:• Performance of the nominal design often does not fo llow the mean• Nominal performance is a poor representation of pop ulation characteristics• Nominal response curve misses peaks present in the population

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

DETERMINISTIC MODEL –Nominal Response

STOCHASTIC MODEL –Mean Response

. Amplitude Range = 15 dB

. Decision is Clearer that both Designs are Equal Performers

Blue: New Design

Green: Base

Population Based Metrics


Slide 105

� � � � � 8 4 � � $ & 2 & � $ % * $ & % $ . * " + " # ; % # ! �

� ( � + ) " % ( $ + � . * - - $ & 2 � � � ! " # $ % � & � ' ( ) * * . $ " &

; " - , * � ' " 1 " % * - $ & * + � & ' $ 2 , # " ( $ � & 5 2 # * * & 7 / $ ( ) "

� . $ ' $ + " ( $ � & % * ( 5 1 - , * 1 � - . 7

P/T Load2nd Order

� � � � � 8 4 � � $ & 2 & � $ % * $ & % $ . * " + " # ; % # ! �

' * ( * # $ & $ % ( $ + � . * - % � � � ! " # $ % � & � ' " 1 " % * - $ & *

+ � & ' $ 2 , # " ( $ � & 5 2 # * * & 7 / $ ( ) " � . $ ' $ + " ( $ � & % * ( 5 1 - , *

1 � - . 7 �

. Amplitude Range = 30 dB

. Need Tradeoff Judgment

Performers

Comments:• Population based metrics, such as Mean or 95% Upper Bound are more stable• They can be directly related to the percentage of c ustomers experiencing

unacceptable NVH response



decision making



Slide 106


diagnostics

P/T Load2nd Order


� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

Multiple Sample Analysis


Slide 107

� � � � � � � � � ! " # $ % � & % � ' ( ) * % ( � + ) " % ( $ + + � ! , ( " ( $ � & " - � . * - # * % , - ( % / $ ( ) ( ) * * 0 ! * # $ * & ( % ' � # ( ) * 1 � � $ & 2

& � $ % * � � # " ! ) % � ' ( ) * � . , - $ � ' ( ) * � 3 � % $ & . 4 5 6 7 % + " - * 8 * 0 ! * # $ * & ( % ' � # ( ) * � � + " # % 5 ( ) $ & 2 # " 9 - $ & * % 7 : * " &

; " - , * � ' ( ) * * 0 ! * # $ * & ( % 5 ( ) $ + < 2 # " 9 - $ & * 7 : = * " & + � ! , ( " ( $ � & " - � . * - 5 . " % ) * . - $ & * 7 : = * " & ; " - , * � ' ( ) *

# " & . � # * % ! � & % * 5 $ . ( ) $ & % � - $ . - $ & * 7 : > ? @ + � & ' $ . * & + * # * 2 $ � & 8 ( ) * , ! ! * # " & . - � / * # * & ; * - � ! * % " # * ( ) * , ! ! * #

" & . - � / * # ( ) $ + < % � - $ . - $ & * % �

Comments:• Better ability to capture performance problems with multiple simulation samples• NVH diagnostics based on the worse case sample is s imilar to diagnostics

using hardware


Simulation studies




Slide 108



• Parametric approach

• Non-parametric approach

• Hybrid approach

Approaches to Stochastic Simulation


Slide 109

• Hybrid approach

Parametric Approach


Slide 110

Comments: • Limited to Few Discrete Variations around the Loc al Design Change• Brute Force – Full Vehicle Runs to Produce Bands • Provides better ability to control variations to reduce spread• The complexity of this approach often lead to ina ction

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

� � � � � � � � � �

Standard Eqn. forStructural-Acoustic Coupling Modal Model

Typical Derivation for creating the randomized Dynamic Matrix with Gaussian Distribution.Non-Parametric : Direct Changes in Modal

Non-Parametric Approach


Slide 111

Changes in Modal Matrices.

The scatter created is a function of:Dispersion Parameter: �

Once determined, � is a constant controlling the amplitude level of scatter.

7 Dispersion Constants are needed:

3 for Structure: M, D, K

3 for Acoustics: M, D, K

1 for Coupling: C n,m

Comments: • This example is inspirational – it shows there is a way forward even when lacking sufficient knowledge on variations• The challenge is to find a way to control the spr ead

• Non-parametric – to decompose vehicle level spread t o subsystem level variations

• Decomposition methods, such as modal participation and TPA,

have been widely used in NVH

Hybrid Approach


Slide 112

• Non-parametric - decomposition to subsystem level• Decomposition methods, such as modal participation and TPA,

have been widely used in NVH

• It would be very beneficial to be able to decompose the vehicle

response spread to contributions from the subsystem level

Hybrid Approach


Slide 113

• Parametric – subsystem and component level

• Much reduced no. of variations

• Possibility to reduce vehicle level spread by contr olling variation in critical subsystems and components

How is stochastic simulation going to change the perception?

Conclusions

� It fails to correlate with test data

� It fails to capture problems at the early upfront stage

Today, many people consider relying on simulation in early upfront design is ‘very risky’ due to real or perceived problems:

� Ensure proper CAE population to samples of test comparison

� Use of multiple simulation sample analysis can capture more peaks


Slide 115

the early upfront stage

� CAE derived design fixes do not always fix the vehicle and can sometimes make the problem worse

� CAE predicted performance can mysteriously change significantly after a model update that cannot be explained by the physical design differences

analysis can capture more peaks

� Understanding that simulation sample to test sample difference could be the cause – multiple sample test data will be helpful

� Use population based metrics, which are much more stable than nominal results, can help us better relate to the impact to the customer

Primary References (Workshop Basis: 4 Papers)

1. A. E. Duncan, et. al., “Understanding NVH Basics”, IBEC, 1996

2. A. E. Duncan, et. al., “MSC/NVH_Manager Helps Chrysler Make

Quieter Vibration-free Vehicles”, Chrysler PR Article,

March 1998.

3. B. Dong, et. al., “Process to Achieve NVH Goals: Subsystem

Structure Borne NVH References


116

3. B. Dong, et. al., “Process to Achieve NVH Goals: Subsystem

Targets via ‘Digital Prototype’ Simulations”, SAE 1999-01-1692,

NVH Conference Proceedings, May 1999.

4. S. D. Gogate, et. al., “’Digital Prototype’ Simulations to Achieve

Vehicle Level NVH Targets in the Presence of

Uncertainties’”,

SAE 2001-01-1529, NVH Conference Proceedings, May

2001

Supplemental Reference Recommendations5. T.D. Gillespie, Fundamentals of Vehicle Dynamics, SAE 1992

(Also see SAE Video Lectures Series, same topic and author)

6. D. E. Cole, Elementary Vehicle Dynamics, Dept. of Mechanical Engineering, University of Michigan, Ann Arbor, Michigan, Sept. 1972



117

7. J. Y. Wong, Theory of Ground Vehicles, John Wiley & Sons, New York, 1978

8. M. Norton, Fundamentals of Noise and Vibration Analysis for Engineers, Cambridge Press, 1989

New Technology References (2013 NVH Workshop)

9. Sol, A.; Van Herpe, F.; “Numerical Prediction of a Whole Car Vibro-Acoustic

Behavior at Low Frequencies”; SAE # 2001-01-1521

10. Durand, J.; Gagliardini, L.; Soize, C.; “Nonparametric Modeling of the

Variability of Vehicle Vibroacoustic Behavior”; SAE # 2005-01-2385; SAE 2005

Noise and Vibration Conference Proceedings.

11. Durand, J.; Gagliardini, L.; Soize, C.; “Structural-acoustic modeling of



Slide 118

automotive vehicles in presence of uncertainties and experimental identification

and validation”; Journal of the Acoustical Society of America 24, p 1513-1525

12. Jund, A.; Gagliardini, L.; et.al.; “AN INDUSTRIAL IMPLEMENTATION OF

NON-PARAMETRIC STOCHASTIC MODELLING OF VEHICLE

VIBROACOUSTIC RESPONSE” CONVEBONOV Workshop, University of

Sussex, Brighton, UK, March 27-28 2012

New Technology References (2015 NVH Workshop)13. Amit Zutshi; Bhaskar Avutapalli; et.al.; “Applying Six Sigma Tools to the Rear Driveline

System for Improved Vehicle Level NVH Performance”; SAE # 2007-01-2286

14. Glenn A. Meinhardt; Zhaohui Sun; et.al.; “An Application of Variation Simulation –

Predicting Interior Driveline Vibration Based on Production Variation of Imbalance and

Runout”; SAE # 2011-01-1543

15. Mohamad S. Qatu; Roger King; et.al.; “Determination of Interior NVH Levels from Tire

Wheel Variations using a Monte Carlo Process”; SAE # 2011-01-1580



Slide 119

Wheel Variations using a Monte Carlo Process”; SAE # 2011-01-1580

16. Perry Gu; Michael McKee; “An Innovative Method of Simulating Tire Non-Uniformity

Forces for Vehicle Vibration Sensitivity Measurements” SAE # 2009-01-2086

17. Mark J. Moeller; Robert S. Thomas; “An Assessment of an FEA Body Model for Design

Capability” SAE # 2001-01-1401

18. Eddy Dascotte; “Linking FEA with Test” SOUND AND VIBRATION/APRIL 2004, p. 12-

16

19. L. Gagliardini; “Dispersed vibroacoustic responses of industrial products: what are we

able to predict?” ISMA 2014, Keynote Speech

Documents

SAE 2015 NVH Conference Structure Borne NVH WorkshopSource 2015 SAE NVC Structure Borne NVH Workshop Slide 8 Frequency Range:up to 1000 Hz System Characterization • Source of Excitation