Vehicle NVH Europe
Acoustic
Centre
CologneVehicle NVH Europe
Acoustic
Centre
Cologne
24�November�2011 FordProprietary
Automated Multi�Disciplinary Optimization
(MDO) Process Development and Application
on Vehicle Program
Norbert Hampl, Vehicle NVH CAE, Ford Werke GmbH
Giri Nammalwar, Global CAE, Ford Motor Company
Presentation to WOST 8.0
based on a Presentation by
Giri Nammalwar, Behrooz Shahidi, Bruno Barthelemy, Nand Kochhar (all Ford Motor Company), Nitin Sharma, RadhaKrishnan (both DEP Inc., Troy, MI USA) to
2011 SIMULIA Customer Conference
Barcelona, Spain, May 18, 2011
Page 2 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�Develop tools and technologies to significantly reduce the
total execution time of full vehicle Multi-Disciplinary
Optimization (MDO).
� Implement MDO process on a vehicle program to achieve
weight reduction while maintaining performance for
Crash, Body NVH & Full vehicle NVH
�Deploy the automated MDO process and associated tools
and techniques on all future Ford vehicle programs
Objectives
Page 3 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�Long time to parameterize detailed full vehicle models
�Lack of robust tool to compare design content
�Extremely long design generation time
�Limited design space exploration
�Lack of tools to automatically post-process large number
of runs
Current Challenges
Page 4 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�Single parametric model for all attributes
�Modular and scalable morphing methodology
�Model synchronization tool
�Designer Environment to submit and post-process
multiple jobs
�CAD morphing to generate CAD data for the optimized
design
Required Technologies and Processes
Page 5 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Automated (MDO) Process Flow
MESHWORKS
(Common Control-
Blocks to parameterize
models from different
disciplines)
Crash
Design1
Design2
Design3
:
:
Design ‘n’
NVH (BODY)
Design1
Design2
Design3
:
Design ‘n’
Durability
Design1
Design2
Design3
:
:
Design ‘n’Isight
(Optimizer)
DOE
Design Variables
& Limits
Input-Output
matrix
Response
Surface
Optimized design
Objectives
Constraints
Ford in-house process
NVH Solver
Crash Solver
Ford & Meshworks scripts
UNIFY
Baseline - Crash
Base – BODY NVH
Base - Durability
Base – VEH. NVH
Ford in-house Vehicle NVH Solver
• First step is to unify Crash, NVH & Durability models
• MDO Process integrates the Parametrized full vehicle CAE
models with Ford’s solvers & output processing scripts
• Tools developed to fully automate the design generation,
submission and results extraction process
Page 6 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�Attributes chosen for MDO were Safety, Body NVH, Vehicle NVH and Durability.
�Model synchronization to commonize design content across different attributes.
�Safety model was parameterized to have shape, weld and gauge variables.
�Critical shapes and sections @ A, B, C, D pillars, roof header and rocker, weld pitch at key locations and 60 BIW component gages were included as parameters
�Parametric models of Body NVH, Vehicle NVH and Durability were generated from safety model.
Process Application on Vehicle Program
Page 7 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
MDO –Attributes and Load Cases
Offset Frontal
Body NVH Side Impact
Roof Crush
Vehicle NVH
Page 8 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Attribute Model Synchronization
Crash Model NVH Model
COMP10078
DG93-F02908
Comp-name Image
COMP150007S71-A02509
Image Comp-name
• MESHWORKS ‘Syncing’ geometrically compares every part on
Crash modes with every part on NVH model
• Provides % geometric matching of parts (even though part names
and mesh sizes are different)
• Applies parameters from one model to the other
Page 9 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Comprehensive Parametrization
Motor Compartment Weld
Variable
• Complex sections with
multiple parts can be
parametrized rapidly and
easily
• Automatic conversion of spot
welds to weld lines
• Control of weld pitch or no. of
welds on each weld line
C-Pillar Section Variable
D-Pillar Section Variable
Gauge of
60 BIW
parts
Page 10 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�DOE matrix generated through Isight using Optimal
Latin Hypercube Method
�The DOE designs were generated in batch mode
using DEP Meshworks
�Design generation through parallel processing using
multiple Vista machines to achieve fast turnaround
time
Automated Design Generation
Page 11 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
G a u g e P a r a m e t e r S h a p e P a r a m e t e r W e l d P a r a m e t e r Macro
Para
meter
Crash NVHNCAP Roof LINCAP IIHS Trim Model BIW Durability Model
• Isight was used for DOE matrix generation
• All models for 440 designs generated as a batch process
driven through DEP Meshworks
• Completely automated and robust process
DOE-Matrix & Design Generation
Page 12 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Safety
NVH
Durability
Job Completion StatisticsJob Completion Statistics
Safety Body CAE DurabilityVehicle NVH(2)
# of Runs
Time (days)(1)
1,776 2,6642,664 444 444
16 5 1 5
Meshworks Design Environment
HPC
Scripts
Python
Scripts
Automated Job Submission
Page 13 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�Tool developed within Meshworks Designer
Environment to automate multi-attribute response
extraction for crash, body NVH and vehicle NVH
results.
�Ford’s in-house processes of results extraction were
seamlessly integrated with DEP Meshworks for
complete automation.
�Significant time reduction achieved by automatically
extracting (using batch process) over 100 responses
from about 3000 LS-Dyna, Nastran runs.
Multi-Attribute Response Extraction
Page 14 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
• NCAP
• IIHS
• LINCAP
• Roof
SafetySafety Body CAEBody CAE
• Global Torsion Mode
• Global Bending Mode
• Noise and Vibration
Transfer Functions
• Equivalent Stiffness
• Dynamic Stiffness
• Bending Stiffness
• Torsion Stiffness
DurabilityDurability
• Cumulative Damage
Vehicle NVHVehicle NVH
• P/T (@idle rpm)
• Rough Road
Output Responses Tracked
WEIGHTWEIGHT
Page 15 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
� Isight was used to evolve optimized designs using
Response Surface (RSM) Based Optimization
�Combination of Numerical Sequential Programming
and Genetic Algorithms were used to obtain optimal
design
�Models corresponding to optimal designs were
generated and validated for different attributes
RSM Based Optimization
Page 16 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Response Surface Model (RSM)
MESHWORKS
(Common Control-Blocks to parameterize
models from different disciplines)
Crash
Design1
Design2
Design3
:
:
Design ‘n’
NVH (BODY)
Design1
Design2
Design3
:
Design ‘n’
Durability
Design1
Design2
Design3
:
:
Design ‘n’Isight
(Optimizer)
DOE
Design Variables
& Limits
Input-Output
matrix
Response Surface
Optimized design
Objectives
Constraints
Ford in-house process
NVH Solver
Crash Solver
Ford & Meshworks scripts
UNIFY
Baseline - Crash
Base – BODY NVH
Base - Durability
Base – VEH. NVH
Veh. NVH Solver
RSM FOR VEH. NVH OUTPUTS
• Response Surface Models (RSM) were created for all the responses –
Crash, Body NVH and Vehicle NVH
• Accuracy analysis was carried out by comparing RSM predictions with
actual solver results
• This RSM forms the basis for subsequent optimization & sensitivity
studies
Page 17 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
RSM FOR DYN. STF
Optimization Scheme using RSM
• Several optimization runs were carried out using the RSM
• Different choices of constraints and their levels were tried out
• Different optimization algorithms were tried out
• This resulted in several optimal designs with different degrees
of mass saving and associated performance levels
• Constraints were set in general at baseline or target levels
Page 18 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Insight into design space using Pareto plots
• Pareto plots were created to study the ‘contribution’ of
each design variable to a chosen output response
Page 19 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
-1.50E-08
-1.00E-08
-5.00E-09
0.00E+00
5.00E-09
1.00E-08
1.50E-08
2.00E-08
0.0 19.8 39.7 59.5 79.4 99.2
Baseline
Design 1
Design 2
Attribute Confirmation Runs - Results
Body NVH: Noise Transfer Function Vehicle NVH: Steering Wheel Response
Baseline
Optimized Design
Safety: Acceleration Response
Baseline
Design 1
Optimized Design
Baseline
Design 1
Optimized Design
Weight Savings
of 5 lbs.
Page 20 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
Metric for StartMetric for StartMetric for StartMetric for Start����Up Vibrations:Up Vibrations:Up Vibrations:Up Vibrations:
( )4
0
4
,, ∫=
=
T
t
wzyx dttaVDV 4444
Zyx VDVVDVVDVRSQ ++=
• aw (t) is the frequency�weighted acceleration in time�domain.
• T is the signal duration.
C344_gf_base: 250.32C344_gf_base: 250.32C344_gf_base: 250.32C344_gf_base: 250.32
C344_gf_3421: 263.46C344_gf_3421: 263.46C344_gf_3421: 263.46C344_gf_3421: 263.46
Single-Number Metrics
Page 21 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
�High Degree of Automation
�Wide Range of Design Parameters
�Automated Synchronization Across Attributes
�Rapid Throughput during design exploration
�Practical Application at any stage of the Vehicle
Development
Key Benefits of Automated MDO Process
Page 22 of 22 Automated MDO Process on Vehicle ProgramFordProprietary
� Developed an automated MDO Process and successfully applied the process on a vehicle program to realize 5 lbs of weight savings.
� Addressed many of the challenges that exists today in a traditional MDO process by leveraging the current technology and by developing new tools to seamlessly integrate Ford Processes.
� Tools and techniques developed enable Ford to execute Full vehicle MDO several times during the product development phases
� Early stages of product development with a focus to optimize and develop vehicle architecture
� During the platform development phase to optimize the underbody components
� Finally during the upperbody development phase to optimize the tophatcomponents.
� New tools developed as part of this study are currently used by Ford engineers to automate MDO tasks
� The process is highly scalable to include additional attributes, load cases or design variables
Conclusions