Mucool meeting April 26, 2002 presentation Donna Kubik.

Mucool meeting April 26, 2002 presentationDonna Kubik

The FEA sensitivity analyses emphasizethe need to accurately determine the thickness of the window:

Deflection comparison between original & modified Window geometry

0.000

0.500

1.000

1.500

2.000

2.500

0 20 40 60 80 100 120 140

Applied pressure (psi)

De

fle

cti

on

(m

m)

original

modified

--------330um--------345um

Comparison of Stress built up rate between original & modified geometry

0.0

50.0

100.0

150.0

200.0

250.0

300.0

350.0

0 20 40 60 80 100 120 140


Def

lect

ion

(m

m)

Original

Modified

Original

Modified

Burst at 117psi for 330um

Burst at 123psi for 345um

Radius = 0.5mm

Radius = 2.0mm

Str

ess

(MP

a)Room temp

If the window is machined to 345um

instead of the design thickness of 330um,

the burst pressure is increased by 5%.

How to determine the window thickness via photogrammetry

Need for a surface model

What we have done:

VANGO’s TIN

Small patch sphere fit (SPSF)

What we might do next:

Extended application of SPSF

Modified TIN



What we have done:

VANGO’s TIN



Extended application of SPS

Modified TIN


Do not need to model the surface to compare the window shape with design.

Calculate design z for the x,y of each target & compare to measured z.

Setup to measure window shape

Measure the concave and convex sides in same coordinate system, moving projector and camera from one side to the other

Difference between measured shape and design

Concave Convex

Whisker = z(measured)-z(design)*

*Given the design radius of curvature of the concave and convex surfaces, z(design) was calculated for the (x,y) position of each target


A surface model is needed to determine the thickness

It is necessary to determine z for matching pairs of x,y on the concave and convex surfaces or the window.

The required positions may be different from the target locations.

Problem of convex and concave alignment

convex concave

Problem of alignment

convex concave

This makes it difficult to determine the window thickness



What we have done:

VANGO’s TIN

Spherical patch surface fit (SPSF)



Modified TIN

VANGO VANGO uses a TIN (Triangular Irregular Network) to create a surface

model.

VANGOPoint on triangle is always < true point.

This is important to interpret any periodicities seen in the data;

period of an oscillation caused by the TIN would be ~6mm

~6mm

VANGO error

d

Rr

2cos

1sin

2

RRd

R

r

d = f(radius of curvature, size of triangle)

D’D D’

Errors in thickness due to VANGO TIN worst case geometry*

D’=D-14.6um D’=D+15umIDEAL*Assumes 6mm chord and design radii 30.000cm and 30.844cm Note: Nominal thickness (D) = 330um

D’D D’

Errors in thickness due to VANGO TIN:intermediate and best case*

D’=D-0.4um D’=DIDEAL*Assumes 6mm chord and design radii 30.000cm and 30.844cm Note: Nominal thickness (D) = 330um

D’

D’=D+(~10um)

D

IDEAL

TIN geometry for center of Window 4

VANGO TIN

Result for the center:

Thickness = 341.0um ( 5.5um) + (- ~10um)

Convex RMS = 3.6um Concave RMS = 4.1um

The difference (thickness) should have an RMS of about 5.5um

VANGO error bars

Each point requires a unique error bar.

Error = f(phase between concave & convex triangles,

size of triangles,

radius of curvature)

Difference between measured and design thickness

-100

-80

-60

-40

-20

0

20

40

601 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101

105

109

113

117

121

Radial Distance (mm)

Va

ria

tio

n (

um

)

North

East

West

South

N

W E

S

Error bars

Error for North and South is greater for large radii due to location of alignment slots.

Here a typical triangle has sides of 6x20x20 mm



What we have done:

VANGO’s TIN




Modified TIN


Fit a sphere to the points near the point in question.

The sphere fit gives the equation of the sphere:


thickness theis and between difference The

surfacesconvex and concaveboth for thisDo

y)(x, desired choosing zfor Solve

unknowns are ,, where

''' 2222

convexconcave zz

zyx

rzzyyxx


To find the thickness at the center of the window,

solve for z choosing x=0 and y=0.

Small patch sphere fit (SPSF) errors

Intrinsic precision of the points used in the fit.

The basic assumption is that the intended shape was a sphere


Result for the center:

Thickness = 331.6um 5.5um

Error of spherical fit: Convex RMS = 3.6um Concave RMS = 4.1um

The difference (thickness) should have an RMS of about 5.5um

Other info from SPSF

Design radius of curvature = 308.44mm

x (mm) y (mm)

z (mm) radius(mm) rms (mm)

All points except 0_000, 1_xxx, 2_xxx, R_xxx

0.1393 0.0880 -254.9681 305.8315 0.0099

Only 0_000, 1_xxx, 2_xxx

0.4536 -0.2756 -216.2946 267.2156 0.0036

Only 0_000, 1_xxx, 2_xxx, 3_xxx

0.3569 -0.1239 -232.2820 283.1955 0.0068

0_000

1_xxx

2_xxx

3_xxx

R_xxx

Other info from SPSF

x (mm) y (mm)

z (mm) radius(mm) rms (mm)

All points except 0_000, 1_xxx, 2_xxx, R_xxx

0.1393 0.0880 -254.9681 305.8315 0.0099

Only 0_000, 1_xxx, 2_xxx

0.4536 -0.2756 -216.2946 267.2156 0.0036

Only 0_000, 1_xxx, 2_xxx, 3_xxx

0.3569 -0.1239 -232.2820 283.1955 0.0068

Design radius of curvature = 308.44mm

Summary of what we’ve done so far to determine thickness

Obtained two window thickness measurements at the center:

VANGO TIN 341.0um ( 5.5um) + (- ~10um)

SPSF 331.6um 5.5um

Additional support for the thickness value

derived from SPSF:

LN burst Window 4ZOOM in on pressurization

-2.00E+01

0.00E+00

2.00E+01

4.00E+01

6.00E+01

8.00E+01

1.00E+02

1.20E+02

1.40E+02

1.60E+02

6000 6200 6400 6600 6800 7000 7200 7400 7600

time (s)

Pre

ssu

re (

psi

)

0

20

40

60

80

100

120

Tem

per

atu

re (

oh

ms)

pressure (psi)

temp 1

temp 2

152psi

Comparison of Stress built up rate between original & modified geometry

0

50

100

150

200

250

300

350

0 20 40 60 80 100 120 140


Defl

ecti

on

(m

m)

Original

Modif ied

Original

Modif ied

burst press @123psi on as built dimension

burst press @117psi on design dimension

a reduction of 5% in rupture stress if Window is machined to as built dimension

Stress built up Vs test pressure

0

50

100

150

200

250

300

350

400

450

0 20 40 60 80 100 120 140 160 180

Test pressure (psi)

Vo

n-M

ises (

MP

a)

r=.5mm

r=5mm

r=10mm

First yield at 75 psiburst press.@162psi on as built thk. Of 0.345mm

rupture burst pressure expected on design thickness

Based on material properties at 80K

Based on material properties at room temp.

The predicted burst pressure (162psi) was greater than the observedburst pressure (152psi).

This implies that thickness assumed in the FEA (345um) was greater than the true thickness.

This would be consistent with a measured thickness of 331um.

What next?

We haven’t found the thinnest part of window yet,

only the thickness at a chosen point…which we chose to be the center.



What we have done:

VANGO’s TIN




Modified TIN


We haven’t found the thinnest part of window yet, just the thickness at a chosen point.

BUT,

we could do SPSF for many points and search for minimum.

Pseudo code to determine thinnest part of window using SPSF

Chose number of points in “Small Patch”

Choose r=design or r=f(measurement)

Choose (x,y)’s

Do sphere fit

Now do the same for the other surface

Calculate differences

Search for minimum to find thinnest part of window



What we have done:

VANGO’s TIN




Modified TIN

Modified TIN Use a sphere to approximate the triangular surface between measured points.

What to use for radius of curvature?

~6mm

Is the measurement repeatable?Yes.

Measurements were repeated 3 times over ~1 week

Two of those times the MET-L-CHEK was very thin and some points were rejected.

HOWEVER, those points that were measured agreed well

For only one of the three was the TIN created

(the one without rejected points - Sasha’s MET-L-CHEK)

So check repeatability by z-plane fit to flange

Targets on flange

Many projected targets fall on the flange

Is the measurement repeatable?Yes.

Flange thickness (mm)

RMS (mm)

Diff from design thickness (mm)

Thick MET_L_CHEK 54.0896 0.018 0.1096

Thin MET-L-CHEK 54.0924 0.030 0.1124

Thin MET-L-CHEK 54.0624 0.019 0.0824

How else can improvements be made?

Upgrade the camera:

Refined calibration

(one GSI’s most dearly guarded secret,

second to how they achieve 1/50 pixel resolution!)

New lenses

Faster processor

Network capability

How much would we gain?

Result in 30% (maybe 50%) improvement in RMS

So for determining the window thickness, this would

result in change RMS=5.5um to RMS=3.85um

Mucool meeting April 26, 2002 presentation Donna Kubik.

Documents

um slide

tin slide

window thickness slide

d d errors

target slide

nominal thickness d

um d ideal tin geometry

um radius