Beruflich Dokumente
Kultur Dokumente
Assemblies
Michael Tonks
Outline
300
250
200
Background 150
100
bending-only
Method
50
plate elements
0
5 10 15 20 25 30 35 40
node
A
Example vA
A
equilibrium position
Geometric
B
nominal position
vB
B
Covariance
Current Research
Introduction
Statistical Tolerance
Analysis (STA)…
statistically predicts critical
assembly variations
Why STA?
Part variations propagate
through assemblies
Predicting effects of
tolerance stack-up allows
designer to avoid assembly
problems
Flexible Parts in Assemblies
Flexible parts deform when assembled due to warping
and dimensional variations
Produces internal stresses and deflections
Rigid body STA cannot predict deformations and
stresses in flexible parts
A method for analyzing flexible parts is needed
Where is Flexible STA Needed?
Aircraft
Skin Panels
Automotive
Body Panels
Plastic
Assemblies
What is Flexible STA?
Analysis method combines
STA with finite element
analysis (FEA)
Rigid body STA gives part
misalignment
FEA used to model parts as flexible
bodies, rather than rigid
Statistically predicts forces, stresses,
and deformations from assembling
flexible parts
The Flexible STA Method
Flexible STA Method Steps
1. Determine misalignment
2. Model the compliant parts using FEA
3. Calculate covariances
A. Material
B. Geometric
4. Statistical FEA Solution for assembly
forces, stresses, and deformations
1. Determine Misalignment
Rigid body STA is used to find the misalignment. A mean and
standard deviation of the gap is found.
A
equilibrium position
vA
nominal position
vB
B
Gap Vectors:
VA, VB = Variation of parts A and B from nominal
A , B = Equilibrium deflection after assembly
2. Model Parts using FEA
Parts are meshed in an FEA Program:
Create geometry
Mesh each compliant part, place
nodes at fastener locations
Apply displacement boundary
conditions
Output the global stiffness matrices
Create Super Elements MSC.Nastran
Create Equivalent Stiffness Matrix MSC.Marc
Create Super Element Matrices
The equivalent stiffness matrix is condensed by eliminating
interior nodes that have no displacement
Matrix reduced to create equivalent super-element matrix
Small and easy to analyze
Contains all important information
boundary nodes
boundary DOF’s coupled DOF’s
Kbb Kbi
b Fb
interior nodes
Kib Kii
i Fi Reduced stiffness matrix:
coupled DOF’s interior DOF’s
Kr = K bb - Kbi Kii-1 Kib
Calculate equivalent stiffness
The Super-element matrices from the mating parts are
combined to form an equivalent stiffness matrix
Two linear
springs in series:
KA KB
FA FB
A
B
Stiffness ratios
Equilibrium: FA = -FB
A = [(KA + KB) KB]
-1
0 = KrA
Gap:
Stiffness:
0 = A -
FA = KA
FB = KB
A
B
B
}
B = -[(KA + KB) KA]
-1
0
0 = KrB0
x x x
200 200
150 150
100 100
50 50
0 0
5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
node node
• 5,000 FE solutions •2 FE solutions
•Large sample size •Very similar results
required for accuracy
•Very Fast
• Slow
Method Comments
Covariance matrix critical
for accurate results 300
Material covariance
250
200
method well established 150
0
Best way to characterize 5 10 15 20 25
node
30 35 40
y
x
Advantages
Simple to apply
Can handle long wavelengths
Disadvantages
Does not consider frequency spectrum
Must be matched to wavelength of surfaces
Spectral Analysis Method
Models surface variation as a finite sum of discrete sinusoids
FFT Frequency 2
spectrum
1.5
Auto-
0.5
spectrum
1.5
0
0 20 40 60 80
1
frequency
IFFT Auto-
0.5 correlation
2
0
0 20 40 60 80 1
frequency
0
-1
-2
-1 -0.5 0 0.5 1
separation distance
Create Covariance Matrix
Because the covariance is assumed to be independent of
location, every row of the covariance matrix is similar
Variance, the diagonal terms of the covariance matrix, is
the autocorrelation function at zero separation
The autocorrelation function can be shifted to form each
row of the geometric covariance matrix:
Autocorrelation Geometric Covariance Matrix
0
Wavelength Effect
Wavelengths of surface variation effects stress in part
Standard deviation of assembly stress x
5000
4000
3000
\L=1 \L=1/2
2000
1000
\L=1/3 \L=1/4
Spectral Analysis Method
Advantages
Models surface variations
Takes in account wavelengths of variations
Disadvantages
Cannot model wavelengths longer then part lengths
Current Limitations