correlation of simulation models using concept modeling

Correlation of Simulation Models using Concept Modeling Dr. Jörgen Hilmann, Joe Abramczyk Ford Motor Company

Dr. Salvatore Scalera RLE International

Andreas Arlt SFE GmbH Berlin

Correlation of Simulation Models using Concept Modeling

Dr. Jörgen Hilmann, Joe Abramczyk, Ford Motor CompanyDr. Salvatore Scalera, RLE International; Andreas Arlt, SFE GmbH BerlinKeywords: RADIOSS, HyperMesh, MotionView, SFE Concept, Correlation, Front Impact, Side Impact

ABSTRACTPre-requisite for efficient vehicle programs is the CAE driven development process as described in [1], [2] "Up Front CAE". Both CAD and CAE models based on special Concept Software as SFE CONCEPT are used to analyze the attribute performance as safety, NVH or durability and commodity studies concerning the weight, manufacturing, or package. A database of concept models is used to minimize the modeling effort and to maximize the re-usability of components [3]. The accuracy of the models increases over time and leads to an increasing amount of available Concept models in databases. Due to the increased acceleration of the development process, these models are critical in providing direction on vehicle architecture in the early stages of a program. Due to the high importance of these decisions it is mandatory to trust the results of this early CAE models. Correlation of this Concept models to test or reference mainstream CAE models creates the confidence in this approach. The correlation focuses on two aspects: 1. the level of detail required to capture the detailed folding characteristic of the structure (e.g. siderail or B-Pillar) and 2. the process chain used to process the raw output from SFE Concept into RADIOSS Include files (e.g. gap, contacts, spot-welding, adhesives, bolts). This process chain is implemented using HyperMesh in batch mode, details may be found in [4]. This process chain is tuned to latest program modeling approaches and to meet the desired correlation status. In this presentation RADIOSS safety concept models are correlated to different impact modes. Tools and methods are explained focusing on both the automated evaluation of simulation output and the judgment of the correlation quality. The main criteria defining the correlation fitness are video overlay of vehicle and dummies kinematics, curve comparison of vehicles acceleration, velocity, and intrusion. Furthermore the dummy sensors have been evaluated. The combination of the above mentioned steps enable an accelerated and more confident concept phase allowing for more alternatives being analyzed in a more holistic and detailed manner as described in [5], [3]. This is a pre-requisite for the creation of efficient designs under the constraints of an increasingly accelerated development process.

BIBLIOGRAPHY[1] E. Schelkle and H. Elsenhans,

"Virtual Vehicle Development in the Concept Stage – Current Status of CAE and Outlook on the Future", 3rd MSC Worldwide Aerospace Conference & Technology Showcase in Toulouse 2001

[2] Jörgen Hilmann (Ford) und Uwe Wagner (Ford), „CAE driven development process for the early vehicle development phase.“ IABC Confernce 2007 in Berlin

[3] Michael Keimes, Dr. Jörgen Hilmann, Martin Lichter, Dr. Uwe Wagner, "Optimierungsstrategien für Leichtbauprojekte" VDI Leichtbaukonfernz Ludwigsburg, 2011.

[4] Jörgen Hilmann (Ford) und Hans Zimmer (SFE GmbH)„Development and application of an automated model built process chain for the Preprogram and Concept phase using SFE CONCEPT and the Altair Hyperworks package.“ EHTC Konferenz in Strassbourg 2008

[5] K.H. Volz and H.Zimmer, "Optimizing Topology and Shape for Crashworthiness in Vehicle Product Development", IABC Confernce 2007 in Berlin

2011 European HyperWorks Technology Conference

Jörgen Hilmann, Joe Abramczyk, Salvatore Scalera Andreas Arlt

Contents

• Introduction

• Motivation for Concept Modeling

• Concept Models & Libraries

• Model built and Include Files

• Correlation quality: Procedure and Criteria

• Vehicle Samples

• Q&A

Correlation of Simulation Models using Concept Modeling



Electrical

Chassis

Linkto “LHS“

Veh. Dyn.

…

Body

Basic Design

Mechanical Package

Powertrain Integration

Vehicle Integration

Electrical

Concepts

SQ & V (NVH)

Chassis & Veh.Dyn.

Project LeadB-Car

Project LeadC-Car

Project LeadCD-Car

Project LeadCommercial

Body Exterior

Body Interior & Attributes

- Technical Specialists

- Supervisors

- Base heads

- Contractors

… is a Mini Product Development Department

Introduction

Total Basic Design Team with Matrix Organization



Update2

Update 3

Updateb

UpdateI

Update1

Safety

Pre-decessor

NVH

Pre-decessor

Durability

Update n

Updatec

Updatex

UpdateII

Separate CAE Attribute Workstreams – OLD StateUpdate Effort for each attribute & Risk for misaligned program assumptions

Dev. Time

Mo

de

l Q

ua

lity



Predecessor

• Concept libraries are populated during the concept modeling work

• Re-use of concept modules increases over time driven through best practice initiatives e.g. for joint execution or Pillar designs

• A process chain is implemented using Altair Hypermesh processing re-occurring pre-processing tasks in automated batch mode

• The full vehicle models are assembled from different sources using the RADIOSS Include files:

- Correlated CAE models / modules from the mainstream development teams

- Parts / Modules developed in the concept modeling department using e.g. SFE Concept

Concept Modeling in CAE driven development processUsage of Concept Module Libraries & Standardized Pre-Processing Procedure & Include Assembly Process

Concept module library

Process Chain

meshing, using

HyperWorks package

in batch mode

RADIOSS

Include

Assembly

Process



model updateVCS 3 Level

Safety NVH Durability

Baseline study

Component studiesConcept development& Iterations

Virtual design verification

Safety NVH

New modelVCS 1 Level

Safety NVH Durability

model updateVCS 2 level

Dev. Time

Integrated CAE Attribute Workstreams – Current StateUpdate Effort shared among the team & No risk for misaligned program assumptions



Predecessor

Mo

de

l Q

ua

lity

• In order to built CAE concept model confidence it is mandatory to verify the predictive power of CAE Concept models using correlation studies.

• Herein the full vehicle models (including the concept modules) reflect the <Job #1> design intent of the pre-decessor model.

• The Correlation-Phase is used to familiarize the team with the new program and to ensure the processes used are capable.

• Two Objectives for Correlation: 1. Geometry and Mesh level of detail2. Process to built the full vehicle model

Price of the new process: Demand for Correlation Starting the work from correlated reference models does not require up front preparation

Correlated reference CAE Model

Substitution of concept modules

Correlation of the new model

Analysis of concept alternativesusing the same processes

Concept module

Modules from the reference model



Difficult to standardize the correlation approach,

due to differentiation through:

• Application to geometry or the model built process chain; or both.

• Applied Car Line: B-CAR / C-CAR / CD-CAR / Light Commercial Vehicles

• Availability of reference data: Mainstream CAE and/or real test results

• Different Impact Modes: Front, Side, Rear, …

• Different Requirements / Markets

As a consequence the correlation task is challenging and

the required timing difficult to predict.

Correlation typesStarting the work from correlated reference models does not require up front preparation



Mainstream CAE

SFE Concept FE Mesh

FORD FOCUS

SFE Concept Geometry

Required accuracy of the modelThe Level of detail is a function of the impact mode and



Within Ford concept models

using the same numerical code,

e.g. RADIOSS for Safety analysis.

The level of detail ranges between both:

- A rough box section to represent an inertia effect of structure, which

is not considered to deform

- A detailed geometry with all holes and depressions if considered

being important e.g. under axial compression in a frontal impact.

ODB – EuroNCAP Frontal Offset / IIHSModel correlation status is satisfactory for:

• Siderail behavior

• Overall intrusion values

(although slight under-prediction of Toeboard intrusion)

• Crash pulse

This CAE model is w.r.t. the offset impactwell suited to support A to B comparisons of architecture studies

Correlation Sample: Frontal Offset Key take aways of this presentation



ConventionalCAE model

Concept model

Offset

Deformable

Barrier

64kph To

eB

oa

rd

Da

sh

bo

ard

Co

wl

Up

pe

r_C

ow

l

A_

Po

int

CC

_B

ea

m

rid

ed

ow

n a

rea

Reference model -- -- -- -- -- -- --

Concept model

delta to reference-14mm -1mm 2mm 3mm -6mm 2mm -3mm

Displacement delta values

Straight Front – US-NCAP, SDGModel correlation status is overall satisfactorybut there is a minor difference in the siderail bending.

As a consequence the pulse shape showa slight deviation.

For A to B Comparisons may be used.

Correlation Sample: Full Frontal Rigid Barrier Key take aways of this presentation



Full Frontal

Rigid Barrier

56kph To

eB

oa

rd

Da

sh

bo

ard

Co

wl

Up

pe

r_C

ow

l

A_

Po

int

CC

_B

ea

m

rid

ed

ow

n a

rea

Reference model -- -- -- -- -- -- --

Concept model -12mm -15mm 2mm 9mm -3mm 2mm -5mm

ConventionalCAE model

Concept model

Blue: Mondeo reference CAE modelGold: Mondeo SFE CONCEPT Model

(with carry over parts e.g. Platform, PT,…)

Correlation Sample: Ford Mondeo Frontal Offset Concept Upper Structure combined with mainstream residual modules



Dual Color Overlay: A to B comparison

@

Striker

@ Thorax@ Pelvis

@ B-Pillar-Mid

B-Pillar velocities

Doorvelocities

Correlation Sample: Ford Mondeo Side Impact Concept Upper Structure combined with mainstream residual modules

Blue: Mondeo reference CAE modelGold: Mondeo SFE CONCEPT Model

(with carry over parts e.g. Platform, PT,…)

B-Pillar velocities



0 ms

50 ms

100 ms

AVI

Correlation Method: Film / CAE Overlay Overlay of a physical Test Film with a RADIOSS Crash Simulation: High Speed barrier side impact



Accelerometer readings

A B

C

Method:

- Create the average curve of all test curves

- Add ±15% of the average peak of the curves to define a channel

Observation:

• The accelerations evaluated using CAE model are almost always

contained in the band fluctuation of real acceleration

• The peak of the B-pillar_@_rocker acceleration is slightly

underestimated

B-pillar @

striker

B-pillar @

beltline

A

B

CB-

pillar

@

rocker

Correlation Method: Curve comparison using channels Overlay of a physical curve measurements with RADIOSS time history readings

Black: CAE modelGray: Test curvesBlue channel borders





0,00

0,05

0,10

0,15

0,20

0,25

0,30

0,35

0,40

0,45

0,50

0,55

0,60

0,65

0,70

0,75

0,80

0,85

0,90

0,95

1,00

1,05

1,10

1,15

1,20

1,25

1,30

1,35

1,40

1,45

1,50

1,55

HIC36Shoulder deflection [mm]

Ave5RibDis [mm]

T1 acceleration [g´s]

T12 acceleration [g´s]

Pelvis acceleration [g´s]

Iliac Fy

Acetabulum Fy

2nd row

Normalization: The bar charts represent the max. injury criteria

divided by the arithmetic mean of available test max. injury criteria.

“1” being the average of the available test curve maxima for this criteria, e.g. HIC36

Test Variation: The CAE results (orange bar) are in the range of test values,

except the Pubic Load for the 1st row and Spine Acceleration in the 2nd row.

Correlation Method: Peak curve comparison Overlay of a physical curve measurements with RADIOSS time history readings



0,00

0,05

0,10

0,15

0,20

0,25

0,30

0,35

0,40

0,45

0,50

0,55

0,60

0,65

0,70

0,75

0,80

0,85

0,90

0,95

1,00

1,05

1,10

1,15

1,20

1,25

1,30

1,35

1,40

1,45

HIC36 MaxThoRib Abdomen Force Spine lowe

acceleration

Pelvis acceleration Pubic Fy

1st row

• Concept modeling significantly changed the vehicle development

towards a CAE driven Development Process,

however requires to built confidence in the used simulation models.

� Demand for correlation!

• Methods are presented using visual and statistical methods to determine

the correlation quality

• These methods are objective and measurable; they help to be as

accurate as necessary to reflect the system behavior without capturing all

details of one single test result,

� prevents the risk of overfitting / overpredicting to one test.

• Having solutions for concept modeling, process chain and correlation

under control you can use this approach as a generic approach for

structure development.

• Very helpful is the coupling with Topology Optimization e.g. Optistruct.

and the consideration of system noises � Robustness.

ConclusionsKey take aways of this presentation


Jörgen Hilmann, Joe Abramczyk, Salvatore Scalera, Andreas Arlt

“white Paper”Part F

Competitor

Part BPredecessor

Part C

Referencemodel

Part A

OtherSegment

Part E

Evolution Part D

Topology Opt.Altair Optistruct

Beam representation of the major Load Pathes

incl. Sensitivities

SWOTDemands / Wishes

• System xyz is not weight

efficient enough

• Demand to improve vehicle

performance abc

• Part t.b.d. is not package

efficient enough

• …

Concept development considering all available findings

Materialgaugeoptimization and Materialgrade optimization

Safety ODB/FF

NVH Static / Dyn.

Weight

Architecture Study

?

Selecting the best Subsystems A holistic Approach of Evolution & Revolution



Thanks for your attention!

Correlation of Simulation Models using

Concept Modeling

Team of Authors:

Dr. Jörgen Hilmann, Joe Abramczyk Ford Motor Company

Dr. Salvatore Scalera RLE International

Andreas Arlt SFE GmbH Berlin

correlation of simulation models using concept modeling

Technology

available concept models

database of concept

early cae models

cae driven development

sfe concept intoradioss

process chain

confident concept phase

special concept software