Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Statistical Tolerancing
Fritz Scholz
Dimensions and properties of parts are not exactly what they should be.
That means that a lot of dimensions have to fit just about right.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0
The Root Sum Square (RSS) paradigm does not work here!
2
IBM Collaboration: Disk Drive Tolerances
D
H
B
A C
3
Coordination Holes for Aligning Fuselage Panels
ideal
perturbed holes
4
Main Ingredients: Mean, Variance & Standard Deviation
Z x
X f (x) (density), CDF F(x) = P(X x) = f (t) dt .
Z x
Mean: = X = E(X) = t f (t) dt
Z x
Variance: 2 = 2X = var(X) = E((X )2) = E(X 2)2 = (t )2 f (t) dt
p
Standard Deviation: = var(X)
5
Rules for E(X) and var(X)
For constants a1, . . . , ak and random variables X1, . . . , Xk
we have for Y = a1X1 + . . . + ak Xk
It is this latter property that justifies the existence of the variance concept.
q
Y = a21var(X1) + . . . + a2k var(Xk )
6
Central Limit Theorem (CLT) I
Y = 1 + . . . + n and Y2 = 21 + . . . + 2n .
7
CLT: Example 1
standard normal population uniform population on (0,1)
1.2
0.4
0.8
Density
Density
0.2
0.4
0.0
0.0
2 0 2 4 0.0 0.2 0.4 0.6 0.8 1.0
x1 x2
Density
3
2
1
0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
x3 x4
8
CLT: Example 2
Central Limit Theorem at Work
0.30
0.25
0.20
Density
0.15
0.10
0.05
0.00
2 0 2 4 6
x1 + x2 + x3 + x4
9
CLT: Example 3
standard normal population uniform population on (0,1)
1.2
0.4
Density
Density
0.6
0.2
0.0
0.0
4 2 0 2 4 0.0 0.2 0.4 0.6 0.8 1.0
x1 x2
0.8
4
Density
Density
0.4
2
0.0
0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
x3 x4
Weibull population
0.8
Density
0.4
0.0
x5
10
CLT: Example 4
Central Limit Theorem at Work
0.00 0.15 0.30
Density
2 0 2 4 6
x1 + x2 + x3 + x4
1 2 3 4 5 6 7
x2 + x3 + x4 + x5
11
CLT: Example 5
standard normal population uniform population on (0,1)
1.2
0.4
0.8
Density
Density
0.2
0.4
0.0
0.0
4 2 0 2 4 0.0 0.2 0.4 0.6 0.8 1.0
x1 x2
0.8
0.4
Density
Density
0.4
0.2
0.0
0.0
x3 x4
12
CLT: Example 6
Central Limit Theorem at Work (not so good)
0.20
0.15
Density
0.10
0.05
0.00
0 10 20 30 40
x1 + x2 + x3 + x4
13
CLT: Example 7
standard normal population uniform population on (0,1)
0.06
0.4
0.04
Density
Density
0.2
0.02
0.00
0.0
4 2 0 2 4 0 5 10 15 20
x1 x2
0.8
4
Density
Density
3
0.4
2
1
0.0
0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
x3 x4
14
CLT: Example 8
Central Limit Theorem at Work (not so good)
0.08
0.06
Density
0.04
0.02
0.00
20 10 0 10 20 30 40
x1 + x2 + x3 + x4
15
What is a Tolerance?
Tolerances recognize that part dimensions are not what they should be.
should be = nominal or exact according to engineering design
Exact dimensions allow mass production assembly using interchangeable parts
16
Simple Examples
Example 1: A disk should have thickness 1/800 with .00100 tolerance, i.e.,
the disk thickness should be in the range
Example 2: A stack of ten disks should be 1.2500 high with .0100 tolerance,
i.e., the stack height should be in the range
17
Disk Stack
= 0.125 0.001
1 8
0.01
1.25
18
Worst Case or Arithmetic Tolerancing
This gives the total possible stack height range as [1.2400, 1.2600] .
19
disk stack/tolerance stack
worst case
low stack
.02''
worst case
high stack
.125''
1.25''
20
Worst Case or Arithmetic Tolerancing in Reverse
and if the disk tolerances are to be the same for all disks (exchangeable),
.0100/10 = .00100
tolerances to the individual disks (item tolerances).
End tolerances can create very tight and unrealistic item tolerances. Costly!
21
Worst Case Analysis or Goal Post Mentality
nominaltol nominal+tol
nominal
tol -
= Statistical Tolerancing
22
Statistical Tolerancing Assumption
23
Normal Histogram/Distribution of Disk Thicknesses
1200
Histogram of 10,000 Thicknesses
600
400
200
18 16
0
disk thickness
24
Why Does Statistical Tolerancing Work
Under the normal population model = we will see about 13.5 out of 10, 000
disks with thickness .12600 .
The chance of randomly selecting such a fat or fatter disk is .00135 = 13.5/10, 000
The chance of having such bad (thick) luck ten times in a row is
.00135 . . . .00135 = (.00135)10 = 2.01 1029 (!!!)
25
The Insurance Principle of Averaging
We look forward to the day when everyone will receive
more than the average wage.
Australian Minister of Labour, 1973
26
Distribution of Stack Heights
.0031600 = 10 .00100 10 .00100 = .0100
27
Normal Histogram/Distribution of Stacks
400
300
Histogram of 10,000 Stack Heights
Density
200
100
0
28
Root Sum Square (RSS) Method
For S = D1 + . . . + D10, with independent disk thicknesses Di, we have
q
var(D1 + . . . + D10) = 2D1 + . . . + 2D10
p
S =
This is referred to as the Root Sum Square (RSS) Method of tolerance stacking.
Numerically TOLi = TOL?i are the same, they are just different in what they
represent: statistical variation range versus worst case variation range.
Again, TOLS and TOL?S represent statistical and worst case variation ranges,
but they are not the same since
q q
TOL21 + . . . + TOL210 = (TOL?1)2 + . . . + (TOL?10)2 TOL?1 + . . . + TOL?10
30
Statistical Tolerancing Benefits
31
RSS for General n
More generally when the TOLDi are not all the same
q
TOLS = TOL21 + . . . + TOL2n or TOL?S = TOL?D1 + . . . + TOL?Dn
Reduce the largest TOLDi to get greatest impact on TOLS . TOL?D ???
i
32
RSS = Pythagorean Shortcut
..........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
.......... ... ............ ... .
......... ........................ . ....
.......... ............................
.. . . .......... .
.... ............................ ....
. . .. . . . . . ..... . .. .
....
...
.
..
.
...................................................
. ...
.. . .. .
.......... ... ..................................... ...
. . . . . . .. .......... . .. . . ..
. .
. .. .
. .........................................................
. ...
. . . . ..... ... . .........................................
. . . ...... . . . . .
.. ... . . .....
.. .
......................................... ...
. . . .. . ...
T1
q
T1 + T2 + T3 > T12 + T22 + T32
33
Benderizing
As much as RSS gives advantages over worst case or arithmetic tolerancing
it was found that the RSS tolerance buildup was often optimistic in practice.
A simple remedy was proposed by Bender (1962) and it was called Benderizing.
The motivation? When shop mechanics were asked about the dimension accuracy
they could maintain, they would respond based on experience memory.
E(D) = and
Z +TOL
D 1
2D = (t )2 dt substituting (t )/TOLD = x
TOLD 2TOLD
Z 1
1 2
= TOL2D x dx with dt/TOLD = dx
1 2
" #1 !
x 3 1 3 (1)3 TOL 2
= TOL2D = TOL2D = D
6 6 6 3
1
= D = TOLD/ 3 or 3D = 3 TOLD = c TOLD, c = 3 = 1.732.
35
Uniform Part Variation Impact on TOLS
For n 3 the distribution of S is approximately normal, i.e., S N (S , 2S )
see next slide.
q
TOLS = 3S = (3D1 )2 + . . . + (3Dn )2
S = n D = TOLS = 3S = n 3 D = n 3 TOLD = n c TOLD,
i.e., we have a uniform distribution penalty factor c = 3 = 1.732.
Recall that under normal part variation we had: TOLS = n TOLD.
0.7
0.6
0.6
0.5
0.4
Density
Density
0.4
0.3
0.2
0.2
0.1
0.0
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
U1 + U2 + U3 U1 + + U4
37
Uniform Part Variation: Comparison with Worst Case
? TOL?S
TOLD = or TOL?S = n TOL?D,
n
TOLS = 3 n TOLD < TOL?S = n TOL?D when 3 < n.
38
Motivating the 3 cT Link
Increasing that scale by a factor should increase and T by that same factor .
c = 1 c = 1.732
c = 1.225 c = 1.369
c = 1.5 c = 1.306
40
Distribution Inflation Factors 2
c = 1
c = 1
= = 3 c = 2.023
c = 1.134
= = .6
p = .7 , g = .4
= = 2
c = 1.342 c = 1.512
41
Details on Distribution Inflation Factors 1
The factors c are chosen such that for finite range densities we have
3 D = c TOLD
cnormal = 1
q
cuniform = 3, ctriangular = 1.5, celliptical = 1.5, ccos = 3 1 8/2
q
ctrapezoidal = 3(1 + k2)/2 where 2k is the range of the middle flat part.
42
Details on Distribution Inflation Factors 2
For a = b the beta density is symmetric around .5 cbeta = 3/ 2a + 1.
q
cDIN = 3[(1 p)(1 + g) + g2]
43
RSS with Mixed Distribution Inflation Factors
q
S = 1 + . . . + n and S = 21 + . . . + 2n
q q
TOLS = 3S = (31)2 + . . . + (3n)2 = (c1TOL1)2 + . . . + (cnTOLn)2
44
Statistical Tolerancing by Simulation
Calculate the resulting critical assembly dimension, i.e., draw ten thicknesses
from a distribution of thicknesses and compute the stack height (sum).
Repeat the above many times, Nsim = 1000 (or Nsim 1000) times.
45
Statistical Tolerancing by Simulation & Iteration
46
Is Linear Tolerance Stack Special?
47
Crankcase Tolerance Chain
L1
L3 L5
L2 L4 L6
G = L1 L2 L3 L4 L5 L6 = L1 (L2 + . . . + L6)
48
Input/Output Black Box
............................................................................................................
........................................................................
X1 ........................................................................
........................................................................
........................................................................
........................................................................
...
........................................................................
Y = f (X1, . . . , Xn)
........................................................................
........................................................................
........................................................................
........................................................................
........................................................................
........................................................................
Xn ........................................................................
........................................................................
........................................................................
........................................................................
....................................
49
Smooth Functions f
When the output Y varies smoothly with small changes in the Xi, then
Y a0 + a1 X1 + . . . + an Xn
for all small perturbations in X1, . . . , Xn around 1, . . . , n.
f (1, . . . , n) n f ( , . . . , )
1 n
ai = and a0 = f (1, . . . , n) i
i i=1 i
50
Good Linearization Example
7
6
5
y
3
2
51
1.0 1.2 1.4 1.6 1.8 2.0
x
Medium Linearization Example
6
4
y
2
0
52
1.0 1.5 2.0 2.5
x
Poor Linearization Example
6
4
2
y
2
4
53
1.0 1.5 2.0 2.5 3.0 3.5 4.0
x
The Sensitivity Coefficients or Derivatives
Y2 = 2a0+a1X1+...+anXn
2 2 2 2 2
= a0 + a1X1 + . . . + anXn = a1X1 + . . . + anXn
54
The General Tolerance Stack Formula
and by 3 Xi = ci TXi
2 2
(3Y )2 = 3 a1 X1 + . . . + 3 an Xn
2 2
= a1c1TX1 + . . . + ancnTXn
q
2 2
TOLY = 3Y = a1c1TOLX1 + . . . + ancnTOLXn
56
Sensitivities and CLT
max 21, . . . , 2n
0 , as n
21 + . . . + 2n
max a2121, . . . , a2n2n
0 , as n
a2121 + . . . + a2n2n
It makes sense to allow for some kind of mean shift or targeting error
while still insisting on having all or most part dimensions within
specified tolerance ranges.
58
Two Strategies of Dealing with Mean Shifts
In either case combine this in worst case fashion or arithmetically with the
RSS part variation stack.
The reason for the last worst case stacking step is that the mean shifts
represent persistent effects that do not get played out independently and
repeatedly for each produced part dimension.
59
Mean Shifts Stacked Arithmetically
part dimension X1
part dimension X2
probability density
part dimension X3
60
Mean Shifts Stacked via RSS
part dimension X1
part dimension X2
probability density
part dimension X3
61
Mean Shifts within Tolerance Interval
For the part variation to stay within tolerance there has to be a tradeoff
between variability and mean shift.
62
Mean Shifts, Variability & C pk
However, this does not control the mean shift. We could have U and
C pk = 1. Then all part dimensions would be near U = worst case stacking.
63
Bounded Mean Shifts
i = iTi 0 i 1 .
But maintain C pk 1
iTi + 3i Ti = 3i (1 i)Ti
64
Arithmetically Stacking Mean Shifts
This grows on the order of n and not n, but with a reduction factor.
1 = . . . = n = 0 = RSS stacking.
65
RSS Stacking of Mean Shifts
The ci are the penalty factors for the distributions governing the mean shifts.
The ci are the penalty factors for the distributions governing part variation.
66
Distributions with Mean Shift I
c = 1 c = 1.732
c = 1.225 c = 1.369
c = 1.5 c = 1.306
67
Distributions with Mean Shift II
c = 1
c = 1
= = 3 = = .6
c = 2.023
c = 1.134
= = 2 p = .7 , g = .4
c = 1.342 c = 1.512
68
Other Variants
Rather than dividing up TOL into mean shift and a 3 range (by squeezing
down 3 to maintain C pk 1) we can increase TOL to the sum of the original
TOL0 = 3 plus the mean shift represented as a fraction of the increased
TOL, i.e.,
3i TOL0i
TOLi = 3i + iTOLi or TOLi = = .
1 i 1 i
For details on how the stacking formulas change see the provided reports.
69
Actuator
o
Ao TA Ro TR
B+
B-
max,o
min,o
B
70
Actuator Case Study
The leg B, representing the actuator, can be adjusted such that the angle agrees
exactly with a specified value 0.
From there B can extend or contract by an amount thus changing the angle
to a maximum and minimum value max and min, respectively.
71
The Question of Interest
How much variation of max and min around max,0 and min,0 can we expect
due to the variations in A and R over their respective tolerance ranges
A0 TA and R0 TR?
72
Geometric Considerations
73
Statistical Tolerancing via Simulation
The simplest way of dealing with the variation behavior of = max and
= min due to variation in A and R is through simulation = R.
Get N -vectors of A and R values from N (A, (TA/3)2) and N (R, (TR/3)2).
The results using N = 1, 000, 000 simulations is shown on the next slide.
It used theta.simNN and took just a few seconds to run.
T1 = 0.359o
0 = 55o
2.0
Density
1.0
0.0
max 0
min,, 0 0 = 15.999o
2.5
T2 = 0.467o
0 = 55o
2.0
Density
1.5
1.0
0.5
0.0
min 0 75
Statistical Tolerancing via RSS
q q
T1 = a2max,A TA2 + a2max,R TR2 and T2 = a2min,A TA2 + a2min,R TR2 ,
where
max max min min
amax,A = , amax,R = , amin,A = , and amin,R =
A R A R
All derivatives are evaluated at the nominal values (A0, R0) of (A, R).
These RSS formulae come from the linearization of x (A, R) near (A0, R0), i.e.,
x
x
x(A, R) = x(A0, R0) + (A A0) + (R R0) ,
A A=A0,R=R0
R A=A0,R=R0
which is then taken as an approximation for x (A, R) near (A, R) = (A0, R0).
76
Approximation Quality
When TA or TR get too large, quadratic terms may come into play normality???
All this assumes of course that x is differentiable near (A, R) = (A0, R0).
There are tolerance situation where differentiability is an issue and in that case the
RSS paradigm does not work.
77
The Derivatives
s
x 1 (sx A)(sx R)
=
A 1 + (sxA)(sxR) A sx(sx Bx)
s (s B )
x x x
and
s
x 1 (sx A)(sx R)
= .
R 1 + (sxA)(sxR) R sx(sx Bx)
s (s B )
x x x
Next we have
s s 1
(sx A)(sx R) (sx A)(sx R) (sx A)(sx R)
= 2
A sx(sx Bx) sx(sx Bx) A sx(sx Bx)
and
s s 1
(sx A)(sx R) (sx A)(sx R) (sx A)(sx R)
= 2 .
R sx(sx Bx) sx(sx Bx) R sx(sx Bx)
78
More Derivatives
Bx A R cos(0) Bx R A cos(0)
=q and =q
A R
A2 + R2 2AR cos(0) A2 + R2 2AR cos(0)
(sx A) 1 A R cos(0) (sx R) 1 A R cos(0)
= 1 and = +1
A 2 B A 2 B
(sx A) 1 R A cos(0) (sx R) 1 R A cos(0)
= +1 and = 1
R 2 B R 2 B
sx 1 A R cos(0) sx 1 R A cos(0)
= +1 and = +1
A 2 B R 2 B
(sx Bx) 1 A R cos(0) (sx Bx) 1 R A cos(0)
= 1 and = 1 .
A 2 B R 2 B
79
And More Derivatives
(sx A)(sx R)
A sx(sx Bx)
1
= 2 2
(sx R) (sx A) + (sx A) (sx R) sx(sx Bx)
sx (sx Bx) A A
(sx A)(sx R) (sx Bx) sx + sx (sx Bx)
A A
(sx A)(sx R)
R sx(sx Bx)
1
= 2 2
(sx R) (sx A) + (sx A) (sx R) sx(sx Bx)
sx (sx Bx) R R
(sx A)(sx R) (sx Bx) sx + sx (sx Bx) .
R R
80
And More Derivatives
Rather than just using these expressions as they are it is advisable to simplify them
somewhat to avoid significance loss in the calculations.
R x
(sx R) (sx A) + (sx A) (sx R) = [1 cos(0)] + [A R cos(0)]
A A 2 2B
R x
(sx Bx) sx + sx (sx Bx) = [1 + cos(0)] [A R cos(0)]
A A 2 2B
A x
(sx R) (sx A) + (sx A) (sx R) = [1 cos(0)] + [R A cos(0)]
R R 2 2B
A x
(sx Bx) sx + sx (sx Bx) = [1 + cos(0)] [R A cos(0)] .
R R 2 2B
81
RSS Calculations
The R function deriv.theta produced the following derivatives for A0 = 12.8,
R0 = 6, 0 = 55, and = 1.6
max min
= .00006636499 and = .004038650
A A
and
max min
= 0.04473785 and = 0.05810921 .
R R
The RSS calculation using normal variation for A and R then gives the following
values for T1 and T2 based on TA = .12 and TR = .14
The derivatives of max and min with respect to A are smaller than
the derivatives with respect to R by at least an order of magnitude.
Important when considering other distributions governing the variation of A and R.
82
Numerical Differentiation
The derivation of the derivatives was quite laborious, but R code is compact.
83
Numerical Differentiation Example
max min
.00006636269 and .004038651
A A=A0,R=R0 A A=A0,R=R0
and
max min
0.04473777 and 0.05810908 .
R A=A0,R=R0
R A=A0,R=R0
These agree very well with the derivatives obtained previously via calculus.
84
Revisit RSS for Linear Combinations
85
CLT and Adjustment Factors
a212X a2n2Xn
( )
1
max ,..., is small,
a212X + . . . + a2n2Xn a212X + . . . + a2n2Xn
1 1
q q
TY = 3Y = a21(3X1 )2 + . . . + a2n(3Xn )2 = c21a21T12 + . . . + c2na2nTn2 .
The next few slides show simulations with 0 = 55 and = 1.6 and
87
(A, R) (U , N ) Simulation Output, Nsim = 106
4
0 = 55o
Density
2
1
0
max 0
3.0
0 = 55o
Density
2.0
1.0
0.0
min 0
88
(A, R) (N , U ) Simulation Output, Nsim = 106
2.0
1.5 max,, 0 0 = 15.325o T1 = 0.622o
0 = 55o
Density
1.0
0.5
0.0
max 0
2.0
1.5
1.0
0 = 55o
0.5
0.0
min 0
(A, R) (U , U ) Simulation Output, Nsim = 106
2.0
1.5 max,, 0 0 = 15.325o T1 = 0.622o
0 = 55o
Density
1.0
0.5
0.0
max 0
2.0
1.5
1.0
0 = 55o
0.5
0.0
min 0
(A, R) (U , U ) Simulation Output, Nsim = 106
0 = 55o
Density
10
5
0
max 0
20
15
10
0 = 55o
5
0
min 0
RSS Calculation with Inflation Factors
Applying the RSS formula assuming a uniform distribution for both A and R we get
q
360
T1 = (.00006636269)2 3 .122 + (.04473777)2 3 .142 = 0.6215642
2
and
q
360
2 2 2
T2 = (.004038651) 3 .12 + (.05810908) 3 .14 2 = 0.8087691
2
using the inflation factor c = 3 and the numerical derivatives in both cases.
Reasonable agreement with the values .622 and .81 from simulation.
Not surprising when linearization is good. We are simply using the variance rules.
However, T1 and T2 do not capture the variation range of x , since the CLT fails.
R U (6 .15, 6 + .15)
93
Varying A, R, ?0 and Uniformly
1.0
max,, 0 0 = 15.325o T1 = 1.38o
0.8
0 = 55o
0.6
Density
0.4
0.2
0.0
14 15 16 17
max 0
1.0
0.8
0 = 55o
Density
0.4
0.2
0.0
18 17 16 15 14 94
min 0
Final Comments
the effect of the CLT when sufficiently many contributing inputs are involved
95
Voltage Amplifier
96
Output Voltage V0
R2
E1 1 + R E2 R2
1
V0 = f (E1, E2, R1, R2, R3, R4) = R3
1+R R1
4
Nominal values:
E1 = 1V , E2 = 1V , R1 = 10, R2 = 100, R3 = 10, and R4 = 100.
= V0 = 20V .
97
The Derivatives
R2
V0 1+R V0 R2
1
= R
, =
E1 1 + 3 E2 R1
R4
E1
V0 E1 R2 E2 R2 V0 R1 E2
= R
2+ , = R
R1 1 + R3 R1 R21 R2 1 + 3 R1
R4
4
R2 R2
V0 E 1 1 + R1 1 V0 E 1 1 + R1 R3
= 2 , = 2 2
R3 R3 R4 R4 R3 R4
1+R 1+R
4 4
98
V.amp.simN2U4(del=.1)
> V.amp.simN2U4(del=.1)
$V0
[1] 20
$delta
[1] 0.1
$derivatives
[1] 10.000000000 -10.000000000 -1.909090909 0.190909091
+ -0.090909091 0.009090909
$sigmas
[1] 0.33314890 0.33326565 1.10195837 1.10243208
+ 0.05248074 0.05246410
$nominals
[1] 1 -1 10 100 10 100
99
V.amp.simN2U4(del=.1)
Ei ~ N(i, (i)2) , Ri ~ U(i i , i + i )
0.25
= 0.1
0.20
0.15
Density
0.10
0.05
0.00
14 16 18 20 22 24 26
V0
100
V.amp.simN2U4(del=.05)
> V.amp.simN2U4(del=.05)
$V0
[1] 20
$delta
[1] 0.05
$derivatives
[1] 10.000000000 -10.000000000 -1.909090909 0.190909091
+ -0.090909091 0.009090909
$sigmas
[1] 0.16657056 0.16676230 0.55079156 0.55108759
+ 0.02627634 0.02624854
$nominals
[1] 1 -1 10 100 10 100
101
V.amp.simN2U4(del=.05)
Ei ~ N(i, (i)2) , Ri ~ U(i i , i + i )
0.5
= 0.05
0.4
0.3
Density
0.2
0.1
0.0
17 18 19 20 21 22 23
V0
102
V.amp.simU6(del=.1)
> V.amp.simU6(del=.1)
$V0
[1] 20
$delta
[1] 0.1
$derivatives
[1] 10.000000000 -10.000000000 -1.909090909 0.190909091
+ -0.090909091 0.009090909
$sigmas
[1] 0.57739282 0.57698360 1.10221137 1.10199967
+ 0.05251682 0.05253420
$nominals
[1] 1 -1 10 100 10 100
103
V.amp.simU6(del=.1)
Ei ~ U(i i , i + i ) , Ri ~ U(i i , i + i )
= 0.1
0.20
0.15
Density
0.10
0.05
0.00
14 16 18 20 22 24 26
V0
104
V.amp.simN6(del=.1)
> V.amp.simN6(del=.1)
$V0
[1] 20
$delta
[1] 0.1
$derivatives
[1] 10.000000000 -10.000000000 -1.909090909 0.190909091
+ -0.090909091 0.009090909
$sigmas
[1] 0.33348276 0.33332256 0.63653780 0.63714909
+ 0.03031808 0.03029352
$nominals
[1] 1 -1 10 100 10 100
105
V.amp.simN6(del=.1)
Ei ~ N(i, (i)2) , Ri ~ N(i, (i)2)
0.4
= 0.1
0.3
Density
0.2
0.1
0.0
16 18 20 22
V0
106
V.amp.simN6(del=.05)
> V.amp.simN6(del=.05)
$V0
[1] 20
$delta
[1] 0.05
$derivatives
[1] 10.000000000 -10.000000000 -1.909090909 0.190909091
+ -0.090909091 0.009090909
$sigmas
[1] 0.16656830 0.16669687 0.31840687 0.31774622
+ 0.01514106 0.01513453
$nominals
[1] 1 -1 10 100 10 100
107
V.amp.simN6(del=.05)
Ei ~ N(i, (i)2) , Ri ~ N(i, (i)2)
0.8
= 0.05
0.6
Density
0.4
0.2
0.0
18 19 20 21
V0
108
Some Final Comments
109
References on Statistical Tolerancing
Altschul, R.E. and Scholz, F.W. (1994). Case study in statistical tolerancing. Man-
ufacturing Review of the AMSE 7, 52-56.
Bates, E.L. (1947). How to increase tolerances and obtain closer fits. The Iron
Age, July 3rd.
Bowker, A.H. and Lieberman, G.J. (1959). Engineering Statistics, Prentice Hall,
Englewood Cliffs, N.J., 51-64.
Cox, N.D. (1986). Volume 11: How to Perform Statistical Tolerance Analysis. Amer-
ican Society for Quality Control, 230 West Wells Street, Milwaukee, Wisconsin
53203.
Ettinger & Bartky (1936). Basis for determining manufacturing tolerances. The
Machinist, October 3rd.
111
References on Statistical Tolerancing
Evans, D.H. (1974). Statistical tolerancing: The state of the art. Part I. Back-
ground Journal of Quality Technology 6, 188-195.
Evans, D.H. (1975). Statistical tolerancing: The state of the art. Part II. Method for
estimating moments. Journal of Quality Technology 7, 1-12.
Evans, D.H. (1975). Statistical tolerancing: The state of the art. Part III. Shifts and
drifts. Journal of Quality Technology 7, 72-76.
Evans, D.H. (198?/9?). Probability and its Application for Engineers, Chapter 9:
Tolerancing , Error Analysis, and Parameter Uncertainty.
112
References on Statistical Tolerancing
Greenwood, W.H. and Chase, K.W. (1987). A new tolerance analysis method for
designers and manufacturers. Trans. ASME, J. of Engineering for Industry 109,
112-116.
Harry, M.J. and Stewart, R. (1988). Six Sigma Mechanical Design Tolerancing.
Motorola Government Electronics Group, 8201 E. McDowell Rd., Scottdale, AZ
85257, Ph. (602) 990-5716.
113
References on Statistical Tolerancing
Nigam, S.D. and Turner, J.U. (1995). Review of statistical approaches to tolerance
analysis. Computer-Aided Design 27, 6-15.
114
References on Statistical Tolerancing
Scholz, F.W. (1995). Tolerance stack analysis methods, a critical review. ISSTECH-
95-021 Boeing Information & Support Services.
115
References on Statistical Tolerancing
Shapiro, S.S. and Gross, A.J. (1981). Statistical Modeling Techniques, Marcel
Dekker, Chapter 7, Analysis of Systems, 268-326.
Srinivasan, V., OConnor, M.A., and Scholz, F.W. (1995). Techniques for compos-
ing a class of statistical tolerance zones. ISSTECH-95-022 Boeing Information &
Support Services.
Wadsworth, H.M., Stephens, K.S., and Godfrey, A.B. (1986). Modern Methods for
Quality Control and Improvements, Chapter 11, 408-433. John Wiley & and Sons,
New York.
116