nasa perspective on certification by analysis...7 august 2012 martin s. annett structural dynamics...
TRANSCRIPT
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
NASA Perspective on Certification by Analysis
Use of Dynamic Analysis Methods for Aircraft Seat Certification
Wichita, KS
7 August 2012
Martin S. Annett Structural Dynamics Branch
NASA Langley Research Center Hampton, VA
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Outline
• Introduction
• Vertical Impulsive Testing and Analysis of ATD’s
• Multi-Dimensional Calibration of Full-Scale Crash Test Article
• Remarks
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Vertical Impulsive Testing of Hybrid II and III ATDs
– Three standard configurations
• Hybrid III Straight Spine (Aero)
• Hybrid III Curved Spine (Auto)
• Hybrid II Straight Spine
– Two loading conditions
• High Magnitude, short duration
• Low Magnitude, High duration
– 2 Finite Element Models
• LSTC
• FTSS
• Examined
– FTSS vs. LSTC Model
– Test vs. Model
– Test Configurations
FTSS LSTC Test
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
• Lumbar loads show results for both pass and fail on FAR 27.562
• Pelvis accelerations show considerable scatter
• Calibration of Hybrid III ATD FEM performed to improve response – Mesh refinement
– Components switched form rigid to deformable
– Foam constitutive model
– Load transfer from pelvis to torso
– Preloading of ATD on seat
Vertical Impulsive Testing of Hybrid II and III ATDs
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
• THOR-NT developed by NHTSA
• All existing THOR’s currently scheduled to be upgraded to the THOR-K version
• (16) drop tests performed with the 50th percentile THOR-NT ATD in the 14-ft Vertical Drop Tower in Building 1262 at LaRC
• Purpose of the tests – Validate THOR finite element models currently in
development
– Compare results to similarly conducted tests previously completed with Hybrid II and III ATDs
– Add additional data to the growing test database for THOR development.
• JSC and LaRC are interested in evaluating many of these recently developed ATDs under spacecraft and aircraft landing loads.
Vertical Impulsive Testing of THOR/NT ATD
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Vertical Impulsive Testing of THOR/NT ATD
Chair Pulses representative of FAR, DOD, and NASA Orion Requirements
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Vertical Impulsive Testing of THOR/NT ATD
Initial Correlation Efforts:
- Analysis performed by
Virginia Tech (THOR-NT FEM
v. 2011)
- Evaluate sensitivity of
kinematics, accelerations, and
force results in the FEM prior
to correlation to test results
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Vertical Impulsive Testing, Comparision of Hybrid II, Hybrid III, and THOR/NT
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Full-Scale Crash Test Objectives
• To evaluate the performance of an external Deployable Energy Absorber (DEA) under realistic crash test conditions
• To generate test data to validate a system-integrated LS-DYNA FEM that includes accurate physical representations of all critical components
• To establish methodologies for FEM development, calibration, and validation that will ultimately lead to crash certification by analysis
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s MD-500 Crash Test with DEA
DEA
Test with DEA
Test Parameters Nominal
Conditions MD-500
with DEA Vehicle Weight (lb) 2,900 2,940
Linear Velocity (ft/sec)
Forward 40. 38.8 Vertical 26. 25.6 Lateral 0 0.5
Attitude (deg)
Pitch 0 -5.7 Roll 0 7.0 Yaw 0 9.3
Angular Velocity
(deg/sec)
Pitch Rate 0 0.4 Roll Rate 0 1.1 Yaw Rate 0 4.8
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Baseline LS-DYNA FEM
LS-DYNA Model 400,000 Elements
266,000 representing DEA
LSTC Hybrid III
FEM
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Comparison-Crash Test with DEA
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Pilot Seat Box Vertical Acceleration Passenger Floor Vertical Acceleration
Acceptable Test/Analysis Correlation for Airframe Response
Comparison-Crash Test with DEA
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s MD-500 Crash Test without DEA
Test Parameters Nominal
Conditions MD-500
Without DEA
Vehicle Weight (lb) 2,900 2,906
Linear Velocity (ft/sec)
Forward 40. 39.1
Vertical 26. 24.1
Lateral 0 0.6
Attitude (deg)
Pitch 0 -6.2
Roll 0 1.9
Yaw 0 2.1
Angular Velocity
(deg/sec)
Pitch Rate 0 0.5
Roll Rate 0 0.7
Yaw Rate 0 1.7
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Comparison-Crash Test without DEA
Increased number of airframe elements
from 134,000 to 250,000
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Pilot Region Vertical Acceleration Passenger Floor Vertical Acceleration
Comparison-Crash Test without DEA
Inconsistencies in Acceleration Time Histories
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Inconsistencies in Deformation Patterns
Comparison-Crash Test without DEA
Pilot Subfloor Deformation
Keel Beam Shear panel
Forward Forward
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Rationale -Multi-Dimensional Calibration
• FEM deficiencies became apparent when severe loads and highly nonlinear responses were introduced for the test without the DEA • A comprehensive and systematic calibration was conducted
• Is modification of existing parameters sufficient? • Is more physical detail required in the model?
• Process implemented is divided into several steps: 1. Metric definition 2. Selection of candidate parameter set 3. Definition of parameter uncertainty model 4. Estimation of variance-based sensitivity 5. Model parameter calibration
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Sensors (19 locations,
23 accelerometers, primarily oriented vertical)
Multi-dimensional Calibration FEM
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Calibration Metrics
Metric 1: Let be the 2-norm of a response vector v and
let and . The probability to reconcile
test with analysis given N model realizations is bounded by:
),(min ptQp
),(max ptQp
NtQtQProbM ee /1))()((1
2),( vptQ
No
rm o
f R
esp
on
se
Time (sec)
N LS-DYNA Solutions
Uncertainty Bounds
Probability Bands
range
Parameter 1
Par
amet
er 2
y1y2
range
Parameter Space(all parameters in the range are equally likely)
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Calibration Metric 1: Uncertainty Bounds
Interim Calibration Cycle
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.161000
1100
1200
1300
1400
1500
1600
1700
1800
Time (sec)
PV
Bo
un
ds
Test
Upper Bound
Lower Bound
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.160
50
100
150
200
250
300
Time (sec)
PV
Bo
un
ds
Test
Upper Bound
Lower Bound
No. Parameter Description Nominal
Upper
Bound
Lower
Bound
1 belly thickness (in) 0.09 0.12 0.08
2 keel beam thickness (in) 0.04 0.07 0.03
3 seat box thickness (in) 0.1 0.12 0.08
4 seat box bulkhead thickness (in) 0.05 0.07 0.03
5 yield stress (psi) 40,000 45,000 35,000
Accele
rati
on
2-N
orm
Velo
cit
y 2
-No
rm
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Variance Based Global Sensitivity Analysis
• Analysis of variance (ANOVA) is used for parameter sensitivity
• LS-DYNA simulations limited compared to number of solutions required
• Response surface surrogates (ERBF) are used to estimate additional solutions
Vel
oci
ty
Time (sec)
Vari
an
ce C
on
trib
uti
on
Yield Stress
Seat Box Bulkhead Thickness
Seat Box Thickness
Keel Beam Thickness
Belly Thickness
Belly
Impact
Interim Calibration Cycle
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Calibration Metric 1: Uncertainty Bounds
Final Calibration Cycle: Keel Beam & Seat Pan Thicknesses
Increased
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.160
50
100
150
200
250
300
350
Time (sec)
PV
Bo
un
ds
Test
Upper Bound
Lower Bound
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.161000
1100
1200
1300
1400
1500
1600
1700
1800
Time (sec)
PV
Bo
un
ds
Test
Upper Bound
Lower Bound
No. Parameter Description Nominal
Upper
Bound
Lower
Bound
1 belly thickness (in) 0.09 0.12 0.08
2 keel beam thickness (in) 0.12 0.15 0.10
3 seat box thickness (in) 0.1 0.12 0.08
4 seat box bulkhead thickness (in) 0.05 0.07 0.03
5 yield stress (psi) 40,000 45,000 35,000
Accele
rati
on
2-N
orm
Velo
cit
y 2
-Norm
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Calibration Metric 2: Impact Shapes
Test
LS-DYNA
Impact Shape =1 16.8 % Contribution
Impact Shape =1 36.6 % Contribution
1 2 3 4 5 6 7 8 9 10
35.57
12.90
10.32
6.98
5.00
4.05
3.50
3.31
2.57
2.25
Test
LS
-DY
NA
MAC
Frequency (Hz)
16
.77
14
.75
9.4
6
6.7
8
5.6
4
5.3
9
4.9
3
4.2
3
3.5
1
3.3
1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1 2 3 4 5 6 7 8 9 10
36.60
13.82
8.92
6.18
4.74
3.92
3.49
3.05
2.84
2.37
Test
LS
-DY
NA
MAC
Frequency (Hz)
16
.77
14
.75
9.4
6
6.7
8
5.6
4
5.3
9
4.9
3
4.2
3
3.5
1
3.3
1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Test
LS
-DY
NA
Test
LS
-DY
NA
Baseline
Run # 44Final Calibration Cycle
Baseline
Acceleration
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Final Calibrated FEM
LS-DYNA Model 590,000 Elements
266,000 representing DEA
Modified LSTC
Hybrid III FEM
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Pilot Region Vertical Acceleration Passenger Floor Vertical Acceleration
Calibrated FEM-Crash Test without DEA
-60
-40
-20
0
20
40
60
80
0 0.05 0.1 0.15
Acc
ele
rati
on
(g)
Time (sec)
Test
Analysis-Original Model
Analysis-Calibrated Model
Filter: 60 Hz Butterworth
-30
-20
-10
0
10
20
30
40
50
60
70
80
0 0.05 0.1 0.15
Acc
ele
rati
on
(g)
Time (sec)
Test
Analysis- Original Model
Analysis- Calibrated Model
Filter: 60 Hz Butterworth
No.
Parameter
Description Nominal
Upper
Bound
Lower
Bound Calibrated
1 belly thickness (in) 0.09 0.12 0.08 0.089
2
keel beam thickness
(in) 0.12 0.15 0.10 0.12
3
seat box thickness
(in) 0.1 0.12 0.08 0.11
4
seat box bulkhead
thickness (in) 0.05 0.07 0.03 0.065
5 Yield Stress (psi) 40,000 45,000 35,000 35,210
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Pilot Region Vertical Acceleration Passenger Floor Vertical Acceleration
Calibrated FEM-Crash Test with DEA
-5
0
5
10
15
20
0 0.05 0.1 0.15 0.2
Acc
ele
rati
on
(g)
Time (sec)
Test
Analysis-Original Model
Analysis-Calibrated Model
Filter: 60 Hz ButterworthFilter: 60 Hz Butterworth
-5
0
5
10
15
20
0 0.05 0.1 0.15 0.2
Acc
ele
rati
on
(g)
Time (sec)
Test
Analysis-Original Model
Analysis-Calibrated Model
Filter: 60 Hz Butterworth
No.
Parameter
Description Nominal
Upper
Bound
Lower
Bound Calibrated
1 belly thickness (in) 0.09 0.12 0.08 0.089
2
keel beam thickness
(in) 0.12 0.15 0.10 0.12
3
seat box thickness
(in) 0.1 0.12 0.08 0.11
4
seat box bulkhead
thickness (in) 0.05 0.07 0.03 0.065
5 Yield Stress (psi) 40,000 45,000 35,000 35,210
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s Conclusions
• Multiple parameters required to calibrate ATD response, mostly relegated to pelvic/lumbar region
• For crash test of full-scale airframe, multi-dimensional model calibration was performed based on two new calibration metrics:
(1) 2-norm acceleration and velocity bound metric (Time variation)
(2) Orthogonality of test and analysis impact shapes (Spatial variation)
• Response surfaces generated from a subset of DYNA simulations for variance analyses
• Calibration cycles identified critical parameters and provided effective guidance to improve model responses for both tests, with and without DEA
NA
SA F
un
dam
enta
l Aer
on
auti
cs-
Ro
tary
Win
g C
rash
wo
rth
ines
s
Remarks
• Important considerations when conducting severe crash tests for the purpose of validating analytical models
• Sensor suite must cover all critical components, and should be located to avoid high-frequency saturation of the acceleration output.
• Accelerometers should be chosen to ensure their velocity integration is accurate (limit drift)
• Multiple validation metrics should be applied between test and analysis which comprehensively identify modeling deficiencies, evaluate parameter importance, and propose required model changes.
• Building block approach to model calibration breaks up the problem into more manageable subsystems
The objective of crash certification by analysis achievable using
similar methodologies