Professional Documents
Culture Documents
Tolerance Stack Up Analysis Complete Paper PDF
Tolerance Stack Up Analysis Complete Paper PDF
Gary Whitmire
Page 1
December, 2004
Many engineering design offices observe the principle of 100% interchangeability when considering the effect of
component tolerances upon the assembly and the functioning of the finished product. In such cases decisions on
the component tolerance are based upon the assumption that the extreme assembly conditions can and will be met
in practice. Where the tolerances for the assembled product are small, this policy leads to an impractical tolerance
for the individual parts, despite the fact that in many cases the probability of encountering parts at their extreme
conditions is very improbable. Where safety is of paramount importance, it is understandable that designers
should insist upon infallible interchangeability, no matter the cost. However, there are many cases where
insistence upon 100% interchangeability is not warranted and an occasional failure to assemble or to function
would not be serious, particularly when judged in the light of the overal1 economic gain in production. It is often
comparatively simple to estimate the maximum cumulative effect upon the assembly and thus ascertain if the
functioning of the assembly is maintained.
One way to ensure that all assemblies fit the designated size range is to determine the assembly tolerance by
simply adding the tolerances of the individual parts. This method is safe and is usually practical for assemblies
consisting of only two or three parts. However, for assemblies made up of several parts, the totals are usually too
large to be acceptable. One method for overcoming this problem is to provide the widest practical component
tolerances based upon the statistical fact that it is unlikely that all maximum-tolerance parts or all minimumtolerance parts would ever be brought together in the same assembly.
This paper presents a methodology for calculating a tolerance stack-up analysis to predict worst case assembly
variation for 100% and 99.73% interchangeability based on preferred component tolerances, or to develop
component tolerances that will support assembly design goals.
Master basic tolerance concepts including the difference between tolerance and variation, how bilateral,
unilateral, and limit dimension tolerances differ.
Get interchangeability when the tolerances are too tight for manufacturing.
It is understood that changes in temperature affect the size of parts. Engineering standards all over the
world state that all dimensions apply at 68 F (20 C).
When doing a tolerance analysis, it is assumed that all manufactured parts will meet the dimensional
requirements stated on the drawing.
Page 2
December, 2004
When doing a tolerance analysis, it is assumed that all parts are rigid in the free state and that they retain
that shape when assembled. However, it is understood that a part may not be rigid in its free state due to
length, size, shape or type of material.
It is assumed that all parts are machined to the middle of their dimensions and controlled by a quality
control program.
A tolerance analysis is typically done to determine if the parts will assemble 100% of the time, or will
assemble 99.73% ( 3 sigma) of the time statistically.
A tolerance analysis is typically done to determine if the parts will function properly at worst condition.
A tolerance analysis will determine if statistical tolerancing will give a lower cost product.
A fully documented tolerance analysis provides a record of the design dimensional requirements that can
be reviewed at a later date in case of a product failure.
It impresses the boss, because not everyone knows how to fully document a dimensional design
condition.
Page 3
December, 2004
Gary Whitmire
30
0
4X .875
4X R
A
12
RADIAL
ASSEMBLY
LINEAR
.75
.62
.375
.12
SECTION A-A
Figure 1
The Tolerance Analysis Form is shown in Figures 2 and 3. This form provides a way to standardize and increase
the accuracy of the tolerance analysis.
Page 4
December, 2004
Gary Whitmire
Assembly #
Rev.
Direction
Direction
Dimensions Dimensions
Today's Date
Rev.
Page
of
RMS
+
Tolerance
Square the
Tolerance in "D"
1
2
3
4
5
6
7
8
9
TOTAL
TOTAL
+
Algebraic
Sum
+
+ (For 3
100% interchangeability
Figure 2
sigma)
Page 5
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
A through F
4
+
Direction
Direction
Dimensions Dimensions
RMS
+
Tolerance
Square the
Tolerance in "D"
1
2
3
4
5
6
7
8
9
TOTAL
TOTAL
10
+
Algebraic
Sum
11
+
12
+
+ (For 3
100% interchangeability
Figure 3
sigma)
Page 6
December, 2004
Page 7
December, 2004
b. Starting at the START of the loop, whichever (left or right / up or down) direction will result in a
dimension going through the START/END gap must be the positive direction for all dimensions
going in the same direction. In Figure 4, the 2.62 .01 dimension is the positive dimension. The other
three dimensions are negative dimensions.
c. Mark these dimensions with positive and negative signs.
6. Tabulate the POSITIVE, NEGATIVE dimensions and the PLUS/MINUS tolerance in columns B, C, and
D. The order is not important.
7. Total the B, C, and D columns.
8. Establish the ALGEBRAIC SUM of columns B and C.
9. Bring down the total from column D.
The ALGEBRAIC SUM and the PLUS/MINUS tolerance is the 100% condition of the snap ring gap.
Page 8
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
START
END
.084 .025
#3
+ 2.62 .01
#1 _
#2
_ .43
.41
+.000
1.901 - .005
#4
.220
.215
Direction
Direction
Dimensions Dimensions
Tolerance
Square the
Tolerance in D
1.8985
.0025
.42
RMS
.01
.01
2.62
.2175
.0025
5
6
7
8
9
+ 2.62
2.536
.025
.025
TOTAL
TOTAL
2.62
2.536
Algebraic
Sum
.084
+
+
100% interchangeability
Figure 4
Figure 5 is a workshop problem for the reader and Figure 6 is the solution.
(For 3 sigma)
Page 9
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
unknown
length
+.008
1.000 .000
.190
.186
Direction
Direction
Dimensions Dimensions
RMS
+
Tolerance
Square the
Tolerance in D
1
2
3
4
5
6
7
8
9
TOTAL
TOTAL
+
Algebraic
Sum
+
+
100% interchangeability
Figure 5
(For 3 sigma)
Page 10
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
END
+.008
1.000 .000
START
.190
.186
Direction
Direction
Dimensions Dimensions
1
2
RMS
+
Tolerance
Square the
Tolerance in D
.188
.002
.004
1.004
3
4
5
6
7
8
9
+ 1.004
.188
+ .006
TOTAL
TOTAL
+ 1.004
.188
Algebraic
Sum
+ .816
+ .006
+
+
100% interchangeability
Figure 6
(For 3 sigma)
Page 11
December, 2004
This next subject explains the theory of statistical dimensioning, and outlines the procedure for using this system.
Most designs use arithmetic dimensioning which means dimensioning such that the tolerances of component parts
of an assembly are combined to give the high and low limits. This procedure assumes that all component parts
will be produced within the drawing specifications, and that the resulting assembly fit is equally likely to be
anywhere between the cumulative tolerance limits. This results in the following:
(a) The component parts have very close tolerances in order to provide the desired fit.
(b) Secondary machining operations are needed.
(c) Adjustments must be designed into the assembly in order to assure proper fit and function.
An alternative to arithmetic dimensioning is statistical dimensioning which is the assigning of dimensions and
tolerances to assembly components on the basis that the assembly tolerance is equated to the square root of the
sum of the squares of the individual tolerances. This is not as technical as it sounds. This approach assumes that
the resultant assembly fit is concentrated in the middle of the cumulative tolerance limits rather than equally likely
anywhere between the cumulative tolerance limits.
Page 12
December, 2004
Responsibility
It is the responsibility of everyone connected with statistical dimensioning to recognize the following:
(a) Engineers must understand not only the basic application techniques, but also the necessity of producing
parts to the mean dimension with a normal distribution.
(b) Manufacturing Engineering must design tools to produce parts around the mean dimension as required by
the drawing specifications.
(c) Quality Control must consult with Manufacturing Engineering to provide the necessary inspection
equipment, assist manufacturing in setting controls, audit the lot as a final inspection on the controlled
dimensions, and, where necessary, determine process capability information using the controlled
dimensions and forward the results to Manufacturing Engineering.
Normal Distribution
A normal distribution is formed when manufacturing processes produce dimensions in a random manner about the
mean, with a majority of the dimensions close to the mean and a decreasing number occurring away from the
mean. If the dimensions, from a reasonably stable process, are measured and recorded according to size, a plot of
the resulting frequency distribution will approximate the shape shown in Figure 7.
Page 13
December, 2004
mean
X
68.27%
1
2
3
94.45%
99.73%
Figure 7
Page 14
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
#3
+ 2.6200
#1 _
#4
.2175
Direction
Direction
Dimensions Dimensions
of
.0100
#2_
1.8985 .0025
RMS
+
Tolerance
Square the
Tolerance in D
.0025
.00000625
.01
.01
.0001
.0001
.0025
.00000625
.42
2.62
.2175
.4200
.0100
.0025
1.8985
2
3
Page
99.73%
.084 .0146
Rev.
START
END
5
6
7
8
9
+ 2.62
2.536
.025
TOTAL
.0002125
TOTAL
2.62
.0146
2.536
Algebraic
Sum
.084
.0002125
.025
+ .084
100% interchangeability
Figure 8
+ .0146
(For 3 sigma)
Page 15
December, 2004
Page 16
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
This page illustrates the four tolerance zones can be increase and still meet
the original design requirements shown on page 14 of .084 .025. This is
done by using the 99.73% probability. The formula for the percent of the part
tolerance is 1 = 1
= .5 ;
.5 X .025 = .0125
N
END
START
.084 .025
#3
+ 2.6200
#1 _
#4
2
3
C
1.8985
.42
2.62
.2175
.0125
#2_
1.8985 .0125
.2175
Direction
Direction
Dimensions Dimensions
99.73%
.4200
.0125
.0125
RMS
+
Tolerance
Square the
Tolerance in D
D
.0125
E
.00015625
.0125
.0125
.0125
.00015625
.00015625
.00015625
5
6
7
8
9
+ 2.62
2.536
+ .05
TOTAL
.000625
TOTAL
+ 2.62
.025
2.536
Algebraic
Sum
+ .084
.000625
+ .05
+ .084
+ .025
+ (For 3 sigma)
100% interchangeability
Figure 9
Page 17
December, 2004
Gary Whitmire
Figure 10 is a Workshop Problem for the Reader and Figure 11 is the solution.
Assembly #
Rev.
Today's Date
Rev.
Page
of
.043.004
.043.004
.113.004
.375.005
.062 .002
.080.003
Direction
Direction
Dimensions Dimensions
RMS
+
Tolerance
Square the
Tolerance in D
1
2
3
4
5
6
7
8
9
TOTAL
TOTAL
+
Algebraic
Sum
+
+ (For 3 sigma)
100% interchangeability
Figure 10
Page 18
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
99.73% .134.009
.043.004
.043.004
START
END
.113.004
+ .375.005
.062 .002
.080.003
Direction
Direction
Dimensions Dimensions
C
.043
.375
2
3
4
5
6
RMS
+
Tolerance
Square the
Tolerance in D
E
.000016
D
.004
.005
.000025
.003
.080
.062
.002
.000009
.000004
.113
.043
.004
.004
.000016
.000016
.241
+ .022
7
8
9
+ .375
TOTAL
.000086
TOTAL
+ .375
.0093
.241
Algebraic
Sum
+ .134
.000086
+ .022
+ .134
+ .009
+ (For 3 sigma)
100% interchangeability
Figure 11
Page 19
December, 2004
Gary Whitmire
Workshop Problem Using Figure 11, repeat the process shown in Figure 9. The solution is given in Figure 12.
Assembly #
Rev.
Today's Date
Rev.
Page
of
This page illustrates the six tolerance zones that can be increase and still meet
the original design requirements shown on page 17 of .134 .022. This is
done by using the 99.73% probability. The formula for the percent of the part
tolerance is 1 = 1
= .408 ;
.408 X .022 = .009.
N
.043.009
.134.022
.043.009
.113.009
.375.009
.062 .009
.080.009
Direction
Direction
Dimensions Dimensions
C
.043
.375
.080
.062
3
4
5
6
RMS
+
Tolerance
Square the
Tolerance in D
E
.00008
D
.009
.009
.00008
.009
.009
.00008
.00008
.113
.043
.009
.00008
.009
.00008
.241
+ .054
7
8
9
+ .375
TOTAL
.00048
TOTAL
+ .375
.022
.241
Algebraic
Sum
+ .134
.00048
+ .054
+ .134
+ .022
+ (For 3 sigma)
100% interchangeability
Figure 12
Page 20
December, 2004
I have never found the information on the next few pages in a technical book. I learned these principles from IBM
35 years ago when I was working for them.
In the next example, shown in Figure 13, the top surface of one part must be .0005 to .018 below the top surface
of another part.
.0005 - .018
.125
Unknown Length
.250
Figure 13
Analysis
(a) The average clearance is (.0005 + .018) / 2 = .009 (.00925). Therefore, the required mean length (labeled
Unknown Length in Figure 13) is .009 + .125 + .250 = .384. The total tolerance to meet the design
requirements of .018 - .0005 is: .018 - .0005 = .0175 or .00875. The three dimensions (.125, .250, and
.384) will have an individual tolerance of .00875/3 = .0029. The better number, shown in Figure 14, is
.0028.
(b) When considering statistical tolerancing, the assembly tolerance as stated above is .00875. Figure 15
illustrates the ratio of statistical addition to arithmetic addition using the formula one divided by the
square root of 3 = 58%. Thus 58% of the assembly tolerance that can be assigned to the three dimensions,
is an approximation. Therefor .58 X .0175 = .010 or .005 for each dimension. Recheck this using the
square root of the sum of the squares to determine that the design requirements have been met.
2
Square root of .005 + .005 + .005 =.00866 X 2 = .0173 (an approximation of .0175), so, all is good
because .0173 is less than .0175.
(c) All tolerance values have been adjusted using the most economical distribution based on available process
information. Note that the values used to determine the statistical sums are squared. Thus changes in the
largest tolerances will have the greatest effect on the final distribution of tolerances.
Page 21
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
START
END
.0028
.1250.0029
.2500
.0029
.0028
Direction
Direction
Dimensions Dimensions
2
3
C
.1250
.2500
.3840
.0028
.0029
.3840
RMS
+
Tolerance
D
.0029
.0029
.0029
Square the
Tolerance in D
.0028
.0028
.0028
4
5
6
7
8
9
.0084
+ .384
.375
+ .0087
TOTAL
TOTAL
+ .384
.375
Algebraic
Sum
.0084
+
.0087
.009
+
100% interchangeability
+
+
Figure 14
(For 3 sigma)
STATISTICAL SUM
=
ARITHMETIC SUM
100
N(tol.)2
=
N(tol.)
N
N
N
1
X
=
=
N
N
N
N N
To illustrate the ratio of statistical addition to arithmetic
addition of the formula above:
80
PERCENT
Page 22
December, 2004
INVERSE OF ROOT X
60
40
20
4 5 6 7 8 9 10 11 12 13 14
NUMBER OF DIMENSIONS
Figure 15
Figure 16 changes the tolerance from .0028 to .0049. Using statistical tolerancing, assume that meeting the
design requirements 99.73% of the time is acceptable.
Page 23
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page
of
START
END
.1250 .0049
.3842
.250
Direction
Direction
Dimensions Dimensions
2
3
C
.1250
.2500
.3840
.0049
.0049
RMS
+
Tolerance
Square the
Tolerance in D
E
.000024
.000024
.000024
D
.0049
.0049
.0049
4
5
6
7
8
9
+ .384
.375
+ .0147
TOTAL
.000072
TOTAL
+ .384
.0085
.375
Algebraic
Sum
+ .009
.000027
+ .0147
+ .009
100% interchangeability
Figure 16
+ .0085
(For 3 sigma)
Page 24
December, 2004
Figure 17 shows a functional clearance gap of .005 to .019. The average gap is (.005 + .019) / 2 = .012, the mean
dimension for the internal opening is .125 + .500 + .375 + .250 + .250, and the average gap of .012 = 1.512. The
gap tolerance is .019 - .005 = .014 or .007.
.500 .005
.375 .003
.250 .002
.250 .002
Figure 17
In the previous example, all parts had the same tolerance. In this example, the parts have different tolerances, so
the formula is different. The tolerance for the internal opening 1.512 dimension will be X as established by the
following calculation:
2
.007 = .000043 + X
.000049 = .000043 + X
2
X = .000049 - .000043
2
X = .000006
X = .000006
X = .002449 or .0024
The plus/minus is an approximation. A better tolerance is .0025. Therefore, the dimension for the opening is
1.512 .0025. The problem is that the five inside parts add up to be 1.500 .020. There is no way the required
gap of .005 to .019 can be met. By changing all of the tolerances to .0012 as shown in Figure 18, the design
requirements can be met, but this is more expensive.
Page 25
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page 1 of 1
The gap tolerance is .019 - .005 = .014 or .007. This .007 / 6 = .0012.
The other tolerance combinations can be used as long as the total .007 is
maintained.
For 100% interchangeability:
The maximum gap is .012 + .007 = .019
The minimum gap is .012 - .007 = .005
END
START
.125 .0012
.500 .0012
+1.512 .0012
.375 .0012
.250 .0012
.250 .0012
Direction
Dimensions
Direction
Dimensions
C
.125
.500
.375
A
1
2
3
4
.250
.250
5
6
1.512
RMS
+
Tolerance
Square the
Tolerance in "D"
D
.0012
..0012
.0012
.0012
.0012
.0012
7
8
9
+ 1.512
1.500
+ .007
TOTAL
TOTAL
+ 1.512
1.500
Algebraic
Sum
.012
+ .007
100% interchangeability
+
for
+ 3 sigma
Figure 18
This is arrived at by going back to the original tolerance and converting to statistical tolerancing (see Figure 19).
Page 26
December, 2004
Gary Whitmire
Assembly #
Rev.
Today's Date
Rev.
Page 1 of 1
END
START
.125 .001
.500 .005
+1.512 .0025
.375 .003
.250 .002
.250 .002
Direction
Dimensions
Direction
Dimensions
C
.125
.500
.375
A
1
2
3
4
.250
.250
5
6
1.512
RMS
+
Tolerance
Square the
Tolerance in "D"
D
.001
E
.000001
.005
.003
.002
.000025
.000009
.000004
.000004
.002
.0025
.000006
7
8
9
+ 1.512
1.500
+ .0155
TOTAL
.000049
TOTAL
+ 1.512
Algebraic
Sum
.007
1.500
.012
+ .0155
.000049
100% interchangeability
+ .007
+ .012
Figure 19
This completes the paper on How to Do a Tolerance Analysis. I hope it will be useful.
for
+ 3 sigma