FRF-Substructures in NVH Analysis of Powertrains

Use of FRF-Substructures in NVH Analysis of

Powertrains in Full-Vehicle Analysis

5th International Conference: Dynamic Simulation in Vehicle Engineering

Uwe Fiedler, CDH AG

Mladen Chargin, CDH AG

Volker Kreissig, Daimler AG

Use of FRF-Substructures in NVH Analysis

Contents

2

1. Introduction

2. FE-model size and calculation time

3. Reduction ProcedureSuperelement method

FRF-Substructure method

4. Example

5. Conclusions


Introduction

3

• Finite element simulations play a

central role in full-vehicle analysis.

• Complex load cases such as:- Road excitation

- Engine excitation

- Gearbox whine

need detailed models.

• The challenge is to simulate

a wide range of excitations

over a large frequency range

in an acceptable time.

• Detailed, complex FE models

require long calculation times.


Introduction

4

Many structural derivatives and variants

must be evaluated:• Body variants: sedan, station, coupe,

cabriolet, wheelbases

influence on interior acoustics

• Chassis variants: 4x2 or 4x4,

active suspension strategies,

subframe, brakes, wheel sizes

• Powertrain variants: petrol, diesel,

parallel hybrid, plug-in hybrid,

4x2 or 4x4, 4/6/8/12 cylinder,

inline or V-engine, cylinder deactivation

Axles and powertrain are not independentMany models to evaluate

large variant matrix


Contents

5

1. Introduction




4. Example

5. Conclusions


FE-Model Size and Calculation Time

6

Analysis of a single subcase as an example: Gear box whine.

• Excitation between two gears of the rear axle differential.

• Frequency response up to 1000Hz.

• Evaluation at comfort points response(sound pressure),

acceleration and forces at mounts.


FE-Model Size and Calculation Time

7

Model size of Mercedes S-class:Structure:

40M DOFs

~48,000 modes < 1,500Hz

Fluid:

1M DOFs

~5,000 modes < 2,000Hz

Calculation Time:

generating matrices: 85 min

AMLS for structural modes: 225 min

AMLSF for fluid modes: 5 min

forming modal matrices: 70 min

solving modal system equation: 55 min

data-recovery: 10 min

Total calculation time 7.5 h

Current approach: full vehicle analysis is time consuming

focusing on few variants, optimisation not feasible!


Contents

8

1. Introduction




4. Example

5. Conclusions


Reduction Methods

9

Some parts are identical for different structural

variants. The powertrain department, for example, can

use a single body for many functional assessments.

This presents an opportunity for model reduction. 1) Standard superelement method

2) New FRF-Substructure method


Contents

10

1. Introduction




4. Example

5. Conclusions


Reduction 1: Superelement

11

An external superelement of body with:- 27M DOFs : 45,000 modes < 1,500 Hz

- 270 boundary-DOFs: coupling and response points

is impossible to generate with current HPC hardware

(1TB RAM, 500TB of disk space) within a reasonable time.

Up to 600Hz with 12,000 modes: - elapsed time 32h

- file size 4.4GB

Using a body-SE is much faster for a calculation of

full vehicle modes, but- many structure modes (huge SE-size)

- frequency response remains costly because of the same number of modes

- disadvantage: acoustic analysis is not possible


Contents

12

1. Introduction




4. Example

5. Conclusions


Reduction 2: FRF-Substructure

13

Can we use body transfer functions only ?


Reduction 2: FRF-Substructure Theory

14

• FRF Equation of motion

• Partition Eq. (1) into 2 parts

Partition 2 contains body, output, and loaded DOFs

• is the “dynamic flexibility”

(1)

-w 2 M[ ]+ iw B[ ]+ K[ ]éë ùû u[ ] = F[ ]

A[ ] = -w 2 M[ ]+ iw B[ ]+ K[ ] A[ ] u[ ] = F[ ]

A11 A12

A21 A22

é

ëê

ù

ûú

u1

u2

é

ëê

ù

ûú =

0

F2

é

ëê

ù

ûú

A11 A12

A21 A22

é

ëê

ù

ûú

u1

H

é

ëê

ù

ûú =

0

I

é

ëê

ù

ûú

H[ ]


Reduction 2: FRF-Substructure Theory

15

• Invert Possibly ill-conditioned, especially

when it contains both structure and fluid!

• is the “dynamic stiffness”

• Any number of additional FE elements can be added or

subtracted from any DOF in

• In modal FRF generation, add residual vectors to DOFs in

to improve the accuracy of the FRF Substructure

Added Elements FD Substructure

H[ ]

H[ ] H i[ ]-1

H i[ ]-1

-w 2 M[ ]+ iw B[ ]+ K[ ] + H[ ]-1é

ëùû

u[ ] = F[ ]


Contents

16

1. Introduction




4. Example

5. Conclusions


Example of Use

17

Mercedes S-Class FE-Model:

Generation of an FRF Substructure of a trimmed body

to couple to a chassis FE-Model:

- 42 mounting points -> 252 coupling DOFs

- 22 response DOFs

- 0 excitation DOFs

Transfer function matrix generated in 7h, file size 1.4 GB.


Example of Use

18

Use of FRF Substructure: Chassis FE-model combined

with trimmed body FRF-Substructure

Final model consists of:13M DOFs for the chassis FE model

2,100 structural modes < 1,500Hz

+ FRF-Substructure matrix of trimmed body

Total calculation time: 51 min

+


Result Quality

19


Result Quality

20


Result Quality

21


Calculation Time Comparison

22

Full FE-modell Superelement FRF-Substructure

Body Chassis Body Chassis

Model DOFs 40M 27M 13M 27M 13M

Modes < 1500Hz 48000 12000<600Hz 2100 45000 2100

One time matrix

genereation32h 7h

Combined job 7.5h 2h 51min

use of FRF-Substructures in NVH Analysis

Contents

23

1. Introduction




4. Example

5. Conclusions


Conclusions

24

FRF-Substructure approach:

• much more efficient than use of superelements

• improved quality of results compared to superelement method

• acoustic response is possible

• FRF Substructure can represent any part of the model

• very useful for optimisation, especially mount optimisation

Thank you for kind attention!www.cdh-ag.com

FRF-Substructures in NVH Analysis of Powertrains

Documents

Transcript of FRF-Substructures in NVH Analysis of Powertrains