static and dynamic model update of an inflatable ......lucas g. horta and mercedes c. reaves...
TRANSCRIPT
Langley Research Center
Static and Dynamic Model Update of an Inflatable/Rigidizable Torus Structure
(Work In Progress)
Lucas G. Hortaand Mercedes C. Reaves
Structural Dynamics BranchNASA Langley Research Center
FEMCI Workshop 2005May 5, 2005
Langley Research Center
Background• The TSU* Hexapod is a generic test bed that incorporates design features of structures that can be packed for launch and deployed in space
•Inflatable/rigidizable structures can be inflated to more than 12 times their packaged size enabling new space mission scenarios
•However, new fabrication techniques also bring new modeling challenges
*TSU= Tennessee State University
Langley Research Center
Objective and MotivationObjective-To develop a mathematical and computational procedure to reconcile differences between finite element models and data from static and dynamic tests
Motivation- All missions envisioned for inflatable and rigidizable systems will require state of the art controls and accurate analytical models. In the Hexapod case, the initial model missed many important modes
10 20 30 40 50 60 70 80 90 100 11010
-5
10-4
10-3
10-2
10-1
Freq. (Hz)M
ag
Scale Factor = 0.11241 Sensor No. 12 Shaker 2
TestNASTRAN
10 20 30 40 50 60 70 80 90 100 110-150
-100
-50
0
50
100
Freq. (Hz)
Pha
se (
deg)
Initial ModelShaker No. 2
Langley Research Center
Approach•FEM developed using commercial tools
•Static test results used to update stiffness values
•Dynamic test results used to update mass distribution
•Probabilistic assessment used to compute most likely set of parameters
•Update parameters computed via nonlinear optimization (genetic and gradient based)
•Computational tools based on MATLAB
•System ID using the Eigensystem Realization Algorithm
•Incremental update of components starting with torus
Langley Research Center
Model Update Procedure
Develop baseline
finite element model Select
parameters for update
Define parameter bounds & probability
distributionCompute output
probability & sensitivity values
Reduce parameter set using sensitivity and
probability results Solve optimization for
updated parameters &
probability
Langley Research Center
Computational Framework
NASTRANStatic
Dynamic
Bulk DataFile
Output DataFile
MATLABOptimizationProbabilitySensitivitySystem ID
External Inputs
1- Parameter selection and prob. distribution
2- Experimental data
Langley Research Center
Torus Description and NomenclatureUrethane Joint (typ.)
Tube (typ.)O.D. 0.0181 m
1.858 m Joint flanget = 0.00508 m Inner joint
t = 0.0063 m
Outer jointt = 0.0032 m
Tube (typ.)t = 0.00042 m
Langley Research Center
Main Sources of Modeling Uncertainties
•Fabrication irregularities due to materials and fabrication methods (irregular cross-section, irregular geometry, etc)
•Unknown stiffness of adhesive bonded joints
•Composite material property uncertainties, “effective single ply” and rule of mixture approximations
Langley Research Center
STATIC TEST RESULTS
Langley Research Center
Static Test Configuration
Support 3
Point load
AluminumBracket
Support 1
Support 2
SteelBlock
Torus static test set-up Instrumentation set-up
Retro-reflectiveTarget
Laser displacement sensor
Langley Research Center
Input Load and Sensor Location for Static Tests
Support 1
Support 2Support 3
Loading Point
1 2 34
56
7
8
910
1112131415
1617
18
19
20
2122
2324
Langley Research Center
Static Test Parameters Selected for Update
1. Torus tube thickness
2. Urethane joint:- Inner joint thickness- Outer joint thickness- Joint flange thickness
3. Torus tube orthotropic material properties:E1, E2
4. Urethane joint modulus of elasticity E
Langley Research Center
Output Probability StatementHow probable is it to predict the measured
output?
Linear Torus Solution
Probability Bands
w range
w
v v range
Output No. 1Output No. 2
Updated parameters
Sample Space 300 points
Displacement Results: Test, Baseline, and Updated Model
Measurement locations
Target 1Z = -4.54 mm
Load =42.1 N
Support 1
Support 2
Support 3
Torus structure displacement under load
-6
-5
-4
-3
-2
-1
0
1
2
3
1 2 3 4 6 7 8 9 10 11 12 14 15 16 17 18 19 20 22 23 24
Measurement location
BaselineUpdateUpdate GenAverage Test
Ver
tical
Dis
plac
emen
t (m
m)
Torus static test results
σ/µ=1.1 σ/µ=0.01
Support 1
Support 2Support 3
Loading Point
1 2 34
56
78
910
111213141516
1718
1920
2122
23 24
Langley Research Center
Static Update Summary
14.1-13.90.008730.011570.011680.00860.01016Flange thickness
16.920.00.005270.005070.007610.005070.00634 Inner jointthickness
-11.019.60.003520.003790.00380.002530.00317Outer jointthickness
9.4130.00039120.0003760.0004960.00030240.000432Tubethickness
-6.103.21E+093.03E+093.33E+092.73E+093.03E+09Joint E
5.405.83E+106.16E+106.77E+105.5E+106.16E+10Tube E1,E2
% ChangeGenetic
% ChangeFMIN
OptimumGenetic
OptimumFMIN
UpperBound
LowerBoundBaselineParameters
Langley Research Center
Dynamic Test Results
Langley Research Center
Dynamic Test Configuration
Force gageSpring/cablesuspension
Suspension 1
Suspension 2
Laser Vibrometer
Shaker
Suspension 3
Retro-reflective
Target (typ.)
Langley Research Center
Shaker and Sensor Locations for Dynamic Tests
Suspension 1Shaker 1Z
Suspension 2Shaker 2Z
C
CC
C
C
C
C
C
C
C
C
C
1416
17
1
54
2
7
8
11
10
13 15
18
3
6
9
12
Shaker4 XY
Suspension 3
Langley Research Center
Sample Set of Identification Results (Drive Point 1 & 2)
0 20 40 60 80 100 12010
-5
100
105 Predicted (- -) and Real FRFs for input No. 1 and output No. 11
Frequency (Hz)
Am
plitu
de
0 20 40 60 80 100 120-200
-100
0
100
200
Frequency (Hz)
Pha
se (
deg)
0 20 40 60 80 100 12010
-4
10-2
100
102 Predicted (- -) and Real FRFs for input No. 2 and output No. 17
Frequency (Hz)A
mpl
itude
0 20 40 60 80 100 120-200
-100
0
100
200
Frequency (Hz)
Pha
se (
deg)
Shaker No. 1
Shaker No. 2
Langley Research Center
Principal Values for Experimental Frequency Response Functions
20 40 60 80 10010
-2
10-1
100
Frequency (Hz)
Pri
nci
pal
Va
lue
s
1279.81433.81194.8955.8Joint 7 density
990.761433.81194.8955.8Joint 8 density
1063.51433.81194.8955.8Joint 10 density
1150.51433.81194.8955.8Joint 11 density
1239.21433.81194.8955.8Joint 13 density
1433.81433.81194.8955.8Joint 14 density
1080.21433.81194.8955.8Joint 16 density
1231.41433.81194.8955.8Joint 17 density
10931433.81194.8955.8Joint 1 density
980.591433.81194.8955.8Joint 2 density
996.081433.81194.8955.5Joint 4 density
1204.61433.81194.8955.5Joint 5 density
1770.42200.81834.01467.2Torus tube density
2400.43502.22918.52334.8Spring 3
3169.83502.22918.52334.8Spring 2
2752.93502.22918.52334.8Spring 1
Optimum ValueUpper BoundNominal ValueLower BoundParameter^
Dynamic Update Summary*
* list of parameters not based on statistical analysis
Langley Research Center
Principal Component Comparison Test/Nominal/Updated
100
101
102
10-3
10-2
10-1
100
Principal Value := 1
Mag
Frequency (Hz)
TestUpdatedNominal
100
101
102
10-3
10-2
10-1
Principal Value := 3
Mag
Frequency (Hz)
TestUpdatedNominal
Maximum Principal ValueMaximum Principal Value Minimum Principal Value
Langley Research Center
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1.123
1.123
1.228
4.138
4.138
4.338
13.734
13.734
20.798
20.803
38.463
38.969
52.754
52.785
68.307
68.310
86.639
86.673
92.301
92.302
101.684
113.262
113.295
123.661 4.3
6
4.4
3
4.5
5
4.7
5
14.1
6
14.
25
14.
33
14.
51
14.
54
14.
58
18.
23
36.
94
37.
17
49.
75
85.3
4
91.1
4
ERA Identified Modes
Ana
lysi
s M
odes
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Orthogonality of Test/Analysis Modes
Animation ofAnalysis Mode
Animation of Test Mode
Langley Research Center
Work To Be Done• Establish repeatability of test data• Complete probabilistic assessment of dynamic FRF to verify
parameter selection• Compute updated parameters for dynamic case using non-
gradient based optimizer • Verify computational accuracy with refined FEM• Implement error localization algorithm to examine potential FEM
problem areas• Compute Modal Assurance Criterion using reduced mass matrix• Re-test with increased sensor count to improve spatial
resolution for test and analysis modes• Refine system ID results to improve areas near FRF zeros
Langley Research Center
Concluding Remarks•Computational procedure using MATLAB in place•Initial updates completed using static and dynamic tests results
•Updated solution for static case is in good agreement with test
•Selected parameters for static analysis showed low output probability values when used to predict observed solution
•Two step approach using output probabilities followed by optimization provides good physical insight into the problem
•Dynamic test and analysis results need to be re-visited
•Procedure computationally intensive but well suited for multi-processor systems