Journal of ASTM International, March 2005, Vol. 2, No.

3
Paper ID JAI12878
Available online at www.astm.org
John D. Reiff, M.Sc.
1


A Method for Calculation of Fastener Torque
Specifications Which Includes Statistical Tolerancing
ABSTRACT: The Monte Carlo simulation method is applied to quantify the probability of exceeding the
maximum desired bolt exploitation for fasteners tightened to a torque specification. The simulation
results are utilized to develop a method for the calculation of torque specifications such that the
probability that a bolt will exceed the maximum desired bolt exploitation is thirty-two parts per million
(ppm).
KEYWORDS: torque, friction, fastener, calculate

Nomenclature
The following is a list of the nomenclature used. Most terms are consistent with the
nomenclature used in the VDI procedure for Systematic Calculation of High Duty Bolted Joints
(VDI 2230).

A
0
Smallest cross-section area of bolt
A
S
Effective tensile stress cross-section of the bolt thread per ISO 898-1
D
0
Outside diameter of bolt at the smallest cross-section, A
0
(smaller of D
S
or D
T
)
D
2
Pitch diameter of bolt thread
D
3
Minor diameter of bolt thread
D
km
Effective diameter for friction at the contact of the head of the driven fastener
D
S
Diameter at stress cross-section A
S

D
T
Shank diameter of bolt neck
D
W
Outside diameter of the contact area under the head of the driven fastener
F
M
Assembly preload, bolt tensile load at tightening
F
M,

Allowable bolt preload, bolt tensile load at which the equivalent stress considering
tension and torsion is R
p,0.2

F
M,MIN
Minimum assembly preload expected from tightening to the specified torque
F
M,MAX
Maximum assembly preload expected from tightening to the specified torque
M
A
Assembly input torque
M
A,MIN
Maximum assembly input torque
M
A,MAX
Minimum assembly input torque
M
A,PRE
Assembly prevailing torque
M
G
Assembly thread torque, moment in the bolt neck
P Pitch of the bolt thread
R
y
Actual proof stress for a bolt
R
p,0.2
Minimum 0.2 % proof stress of bolt material per ISO 898-1

Manuscript received 16 May 2004; accepted for publication 29 November 2004; published March 2005.
1
General Motors Corporation, 30001 Van Dyke Ave., Warren, MI.
Copyright 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.
JOURNAL OF ASTM INTERNATIONAL 2
d
i
Inside diameter of hollow bolt
d
h
Inside diameter of the contact area under the head of the driven fastener

th
Half flank angle of the bolt thread (/6 for ISO thread)

G

Coefficient of friction between bolt and nut thread

G,MIN

Minimum coefficient of friction between bolt and nut thread

G,MAX

Maximum coefficient of friction between bolt and nut thread

K

Coefficient of friction at the contact of the driven fastener head

K,MIN

Minimum coefficient of friction at the contact of the driven fastener head

K,MAX

Maximum coefficient of friction at the contact of the driven fastener head

Degree (%) of exploitation of bolt yield stress desired at maximum assembly
condition

Introduction
In order to minimize product cost and mass, bolted attachments should be designed with the
smallest fasteners possible. This is achieved in part by designing a torque specification that
results in the maximum possible preload, but with acceptable margin against failure due to over-
stressing during tightening or in service. To accomplish this, the probability density function for
the ratio of bolt preload to the targeted allowable bolt preload must be determined. This is not a
simple task because the ratio is a non-linear function of random variables. The bolt preload
variation is related to the variation in input torque, driven fastener head friction, and fastener
thread friction. At the same time the variation in the targeted allowable preload is related to the
variation in bolt yield strength and fastener thread friction.
A method for the calculation of assembly torque specifications is developed for the general
case, which includes bolts with reduced shank diameter, bolts that are hollow, and bolts or nuts
with prevailing torque. The method includes the calculation of torque specification limits based
on the simultaneous occurrence of extreme values of process variables, and it is referred to as the
Extreme Value Method. The upper torque specification limit is calculated so that the bolt
preload is equal to the desired maximum value when the extreme values of process variables
occur. The lower torque specification limit is calculated so that the assembly process will be
statistically capable for the selected assembly tool.
The Monte Carlo computer simulation method is then applied to simulate the tightening
process with normally distributed process variables and to calculate the non-normal distribution
that describes the probability of exceeding the maximum desired bolt exploitation. The
simulation results are used to modify the Extreme Value Method so that the probability that a
bolt will exceed the maximum desired bolt exploitation is 32 ppm.

The Extreme Value Method
For known values of bolt geometry, bolt material properties, prevailing torque, and fastener
head and thread friction, the Extreme Value Method is used to calculate the torque specification
limits (M
A,MIN
and M
A,MAX
) and the bolt preload limits (F
M,MIN
and F
M,MAX
). The bolt preload
limits are used to calculate the capability of the fastened joint to resist the loads expected in
service.
REIFF ON FASTENER TORQUE SPECIFICATIONS 3
First, the equation for maximum allowable initial bolt preload (F
M,MAX
) is developed so that
the equivalent stress in the outer fiber of the bolt due to tensile and torsional loads at the
minimum section is at the desired maximum allowable stress (R
P,0.2
). The exploitation factor
() is a factor to provide a margin against yielding the bolt during tightening to account for an
expected increase in bolt load during service or to account for possible error in the estimation of
the limiting values of friction. The maximum allowable bolt preload is at the most conservative
value when the bolt has minimum yield strength (R
P,0.2
) and minimum thread friction (
G,MIN
).

( ) ( )( )
( )
2
2 1
3 1
2
,
2
2 1
2
2 1 ,
2 1 , ,
3 1 2
3 3 1 4 6
6
K K
K K M K K K K M
K K M F
PRE A PRE A
PRE A MAX M
+
+
+ =
(1)

Where:
( )
2
2 2
0
0
1
4
(
(

+
=
i
d D
D
K ,
(

+ =
th
MIN G
D
P
K

cos 2 2
2 ,
2
, and
( )
2
2 2
0 2 . 0 ,
3
4
(
(

=
i P
d D R
K

For the case where the minimum bolt cross section is the threaded section:

2
3 2
0
D D
D
+
= (2)

And for the case where the minimum bolt cross section is the shank:

T
D D =
0
(3)

The mathematical model of the tightening process developed by Motosh [2] is modified here
to include the prevailing torque term:

( )
|
|
.
|

\
|
+ +

=
2 cos 2 2
2
,
km K
th
G
PRE A A
M
D D P
M M
F

(4)

Where:
2
h w
km
d D
D
+
= (5)

This equation is rearranged to solve for input torque:

|
|
.
|

\
|
+ + + =
2 cos 2 2
2
,
km K
th
G
M PRE A A
D D P
F M M

(6)

The maximum torque limit is calculated by substituting the maximum allowable initial
preload (F
M,MAX
) from Eq 1, the minimum value of head friction (
K,MIN
), and the minimum value
JOURNAL OF ASTM INTERNATIONAL 4
of thread friction (
G,MIN
) into Eq 6. This maximum torque results in the maximum allowable
initial preload in the worst-case conditions for friction with respect to increasing preload.


(

+ + + =
2 cos 2 2
, 2 ,
, ,
km MIN K
th
MIN G
MAX M PRE A MAX A,
D D
P
F M M

(7)

The minimum torque limit is calculated from the maximum torque based on the desired torque
tolerance:
( )
( )
% ,
% ,
, ,
1
1
TOL A
TOL A
MAX A MIN A
M
M
M M
+

= (8)

Where:
NOM A
TOL A
TOL A
M
M
M
,
,
% ,
= (9)

( )
2
, ,
,
MIN A MAX A
TOL A
M M
M

= (10)

Finally, the bolt minimum initial preload when assembled to the minimum torque is
calculated by substituting M
A,MIN
into Eq 4 with maximum values of head and thread friction:

( )
2 cos 2 2
, 2 ,
, ,
,
km MAX K
th
MAX G
PRE A MIN A
MIN M
D D
P
M M
F

+ +

= (11)

The application of these equations requires that the coefficients of head and thread friction be
calculated with equations that also include the consideration of prevailing torque. The equations
are:
|
|
.
|

\
|

2
cos 2
,
2
P
F
M M
D
M
PRE A G
th
G
(12)
And,
( )
M km
G A
K
F D
M M
=
2
(13)

In summary, the Extreme Value Method is applied as follows:

1. Laboratory tests are performed. Multiple samples of the attachment are tightened while
the input torque, bolt tension, and thread torque are measured. The coefficients of head
friction and coefficient of thread friction are calculated for each sample with Eqs 12 and
13. The statistics, mean and standard deviation, for the head and thread friction are
calculated from the individual samples.
REIFF ON FASTENER TORQUE SPECIFICATIONS 5
2. The values
K,MIN
,
K,MAX
,
G,MIN
, and
G,MAX
, are calculated as the three standard
deviation limits of the individual friction values calculated above.
3. The required torque specification tolerance M
A,TOL %
is determined so that the desired
assembly tool will be process capable.
4. The maximum desired degree of exploitation of the bolt during the tightening process
() is selected.
5. Equations 1, 7, 8, and 11 are applied to calculate the torque specification limits (M
A,MIN
and M
A,MAX
) and the bolt initial preload limits (F
M,MIN
and F
M,MAX
).

The Extreme Value Method (Simple Case)
For the case where prevailing torque is zero, the bolt is solid, and the shank is not reduced.
The equations above reduce to the following set of equations. Equation 14 is the same as derived
in the procedure VDI 2230.

2
2 2 3
,
cos 1
4
3 1
(
(

|
|
.
|

\
|
+
+
+
=
th
G
y S
M
D
P
D D
R A
F

(14)
2
,
2 2 3
2 . 0 ,
,
cos 1
4
3 1
(
(

|
|
.
|

\
|
+
+
+
=
th
MIN G
P S
MAX M
D
P
D D
R A
F

(15)

|
|
.
|

\
|
+ +
=
2 cos 2 2
2 km K
th
G
A
M
D D P
M
F

(16)

Where:
2
h w
km
d D
D
+
= (17)

(

+ + =
2 cos 2 2
, 2 ,
, ,
km MIN K
th
MIN G
MAX M MAX A
D D
P
F M

(18)

( )
( )
% ,
% ,
, ,
1
1
TOL A
TOL A
MAX A MIN A
M
M
M M
+

= (19)
Where:
NOM A
TOL A
TOL A
M
M
M
,
,
% ,
= (20)
( )
2
, ,
,
MIN A MAX A
TOL A
M M
M

= (21)
JOURNAL OF ASTM INTERNATIONAL 6

2 cos 2 2
, 2 ,
,
,
km MAX K
th
MAX G
MIN A
MIN M
D D
P
M
F

+ +
= (22)

The equations for the coefficients of head and thread friction reduce to the more standard
equations as defined in the ISO standard for Fasteners Torque/Clamp Force Testing (16047)
and in the German national standard, Determination of Coefficient of Friction of Bolt/Nut
Assemblies Under Specified Conditions (DIN 946).
|
|
.
|

\
|
=

2
cos 2
2
P
F
M
D
M
G th
G
(23)
And:
( )
M km
G A
K
F D
M M
=
2
(24)

Statistical Analysis
The probability that the extreme values of the process variables (M
A,MAX
,
K,MIN
,
G,MIN
, and
R
P,0.2
) will occur simultaneously is very small, and therefore the Extreme Value Method is very
conservative. The application of statistical tolerancing methods results in an increase of the
torque specification limits and therefore bolt preload, with an acceptable probability that over-
loading of the bolt will occur. Equations for calculation of the torque specification limits (M
A,MIN
and M
A,MAX
) are developed so that the probability is approximately 32 ppm that the ratio of the
actual bolt preload to the allowable bolt preload (F
M
/ F
M,
.) will exceed the exploitation factor
(). This is equal to a four-standard-deviation (4) probability for a normal distribution. Also,
equations for the calculation of the bolt preload limits (F
M,MIN
and F
M,MAX
) are developed so that
the probability is 0.0014 (on each side) that the bolt preload will be outside of those limits. This
is equal to a three-standard-deviation (3) probability for a normal distribution.
A common analytical method used to determine the density function of a complex transfer
function is the error transmission formula [2], which is a Taylors series expansion technique.
The application of this method results in a formula for the standard deviation of the bolt preload
that is a function of the standard deviations of M
A
,
K
, and
G
.
2
2
2
2
2
2
2
G K A M
G
M
K
M
M
A
M
F
F F
M
F

|
|
.
|

\
|

+
|
|
.
|

\
|

+
|
|
.
|

\
|

= (25)
However, two assumptions made in the derivation of the error transmission formula do not
apply for this problem:
1. The random variables are assumed to be independent of each other. This does not apply
in this case since the head friction and thread friction are sometimes dependent on each
other, in the case where the bolt is the driven fastener.
2. The higher order terms of the Taylors series expansion are assumed insignificant and are
excluded. In this case, the bolt preload is sufficiently non-linear within the normal
REIFF ON FASTENER TORQUE SPECIFICATIONS 7
tolerances of the process variables, and the higher-order terms of the Taylors series
expansion are significant.

Because of the complexity of the transfer functions and the dependencies between variables,
a Monte Carlo computer simulation is chosen to model the tightening process and calculate the
probability functions needed. The computer uses a random number generator to draw values
from the distributions of the process variables (R
y
, M
A
,
K
, and
G
) and computes the values for
F
M
and F
M,
. The computed values are then summarized in histograms for analysis. Various
distributions of the process variables are simulated to allow for a generalization of results. The
simulation procedure for each test case is as follows:

1. The bolt geometry properties (P, D
T
, D
2
, D
3
, d
i
, d
h,
A
s
, D
km
, and
th
) are modeled as
constants. Values are selected from the appropriate ISO specifications.
2. Values for head friction (
K
) and for thread friction (
G
)

are randomly generated to fit
normal distributions. The means and standard deviations are selected for each test case,
and for some cases the thread friction is generated to be directly dependent on head
friction.
3. Values of torque (M
A
) are randomly generated to fit a normal distribution with the three
standard deviation limits equal to M
A,MIN
and M
A,MAX
. The values for M
A,MIN
and M
A,MAX

are calculated from the appropriate equations.
4. Values of bolt yield strength (R
y
) are randomly generated to fit a normal distribution so
that R
p,0.2
is the lower three-standard deviation limit and the upper three-standard
deviation limit is R
p,0.2
times the ratio of maximum to minimum specified hardness for the
bolt material.
5. All variation is statistically independent except for the variation in
K
and
G
as stated
above.
6. The maximum allowable bolt preload (F
M,
) is calculated for each simulated sample with
values of R
y
and
G
randomly generated.
7. The bolt initial preload (F
M
) is calculated for each simulated sample with values of M
A
,

K
, and
G
randomly generated.
8. The ratio F
M
/F
M,
is calculated for each simulated sample.

Example Simulation Result for the Extreme Value Method
The example model is a standard M10 bolt per ISO 4162 with grade 10.9 material per ISO
898-1. The statistics for the head and thread friction are values for a commonly used organic
finish, and the torque specification limits are calculated using the Extreme Value Method. The
results are shown in the first column of Table 1, labeled Extreme Value Method (EVM), and in
Figs. 1 and 2. Results are shown for cumulative probability values that are equivalent to three
standard deviations (3) for a normal distribution (0.0014 and 0.9986) and for cumulative
probability values that are equivalent to four standard deviations (4) for a normal distribution
(0.000032 and 0.999968). Here are some summary comments from the example simulation.

1. The probability that the bolt preload will be outside of the preload limits (F
M,MIN
and
F
M,MAX
) is significantly lower than the desired probability.
JOURNAL OF ASTM INTERNATIONAL 8
2. The probability is about 32 ppm that the ratio F
M
/F
M,
exceeds 92.5 %. The desired result
is that if the probability is about 32 ppm, the ratio F
M
/F
M,
exceeds 100 %, since the
exploitation factor for this example is 1.0.
3. The distributions of F
M
and the ratio F
M
/F
M,
are not normal since the distribution is not
symmetrical about the mean value. This is proof that the error transmission formula
should not be applied.
4. As expected, the results show that the Extreme Value Method is over conservative.

Summary comments from other simulations of the Extreme Value Method are as follows:

1. For the example shown above, the values of head friction and thread friction are
dependent. In the case where the values are independent, the probability that the bolt
preload will be greater than F
M,MAX
or less than F
M,MIN
is decreased. However, the
probability that the bolt will yield is approximately the same as above.
2. In the case where the variation for head and thread friction are increased by 50 %, the
probability that the bolt preload will be greater than F
M,MAX
or less than F
M,MIN
is also
decreased. However, the probability that the bolt will yield is again approximately the
same.

TABLE 1Simulation results for an M10 bolt.
Extreme Value Method Statistical Method
Inputs
100 % 100 %

G,nom
0.13 0.13

G,std dev
0.01 0.01

K,nom
0.13 0.13

K,std dev
0.01 0.01
Calculations
M
A,MIN
(Nm) 51.4 55.6
M
A,MAX
(Nm) 69.6 75.2
F
M,MIN
(KN) 22.8 27.2
F
M,NOM
(KN) 32.3 34.9
F
M,MAX
(KN) 46.5 45.2
Simulation Results
F
M
(KN) (min sample) 23.1 25.3
F
M
(KN) (Prob. = 0.000032) 22.9 25.0
F
M
(KN) (Prob. = 0.0014) 25.3 27.5
F
M
(KN) (Prob. = 0.5000) 32.4 35.0
F
M
(KN) (Prob. = 0.9986) 41.9 45.6
F
M
(KN) (Prob. = 0.999968) 45.1 49.1
F
M
(KN) (max sample) 46.1 50.7
F
M
/F
M,
(Prob. = 0.9986) 85.0 % 92.5 %
F
M
/F
M,
(Prob. = 0.999968) 92.5 % 100.2 %
F
M
/F
M,
(max sample) 92.4 % 101.7 %
F
M
/F
M,MIN
(min sample) 101.3 % 92.3 %
F
M
/F
M,MIN
(Prob. = 0.000032) 100.4 % 91.2 %
F
M
/F
M,MIN
(Prob. = 0.0014) 111.0 % 100.4 %
F
M
/F
M,MAX
(Prob. = 0.9986) 90.1 % 100.1 %
F
M
/F
M,MAX
(Prob. = 0.999968) 97.0 % 108.6 %
F
M
/F
M,MAX
(max sample) 99.1 % 112.2 %
REIFF ON FASTENER TORQUE SPECIFICATIONS 9

Extreme Value Method
M10 Flange bolt
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
2
1
2
6
3
1
3
6
4
1
4
6
5
1
5
6
Fm Fm,
F
r
e
q
u
e
n
c
y
FM,MIN F
M,MAX
Extreme Value Method
M10 Flange bolt
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
5
0
%
5
5
%
6
0
%
6
5
%
7
0
%
7
5
%
8
0
%
8
5
%
9
0
%
9
5
%
1
0
0
%
1
0
5
%
Fm/Fm,
F
r
e
q
u
e
n
c
y
4 (equiv.)

FIG. 1Distribution of F
M
and F
M,
FIG. 2Distribution of F
M
/F
M,
for the EVM.
for the EVM.

Statistical Targeting
Statistical targeting of the torque specification is accomplished by modification of the
equations based on the results from the Monte Carlo simulations. The goals as stated above are:

1. To develop equations for calculation of the torque specification limits (M
A,MIN
and M
A,MAX
)
such that the probability is approximately 32 ppm that the ratio of the bolt preload to the
allowable bolt preload (F
M
/ F
M,
.) will exceed the exploitation factor ().
2. To develop equations for the calculation of the bolt preload limits (F
M,MIN
and F
M,MAX
) such
that the probability is 0.0014 (each side) that the bolt preload will be outside of those limits.

The revised equations are shown here for the simple case where there is no prevailing torque
and the bolt is solid. Modifications are placed in brackets {} to be seen easily.

2
2 2 3
,
cos 1
4
3 1
(
(

|
|
.
|

\
|
+
+
+
=
th
G
y S
M
D
P
D D
R A
F

(26)
JOURNAL OF ASTM INTERNATIONAL 10
{ }
2
,
2 2 3
2 . 0 ,
4 ,
cos 1
4
3 1
925 . 0
(
(

|
|
.
|

\
|
+
+
+
=
th
MIN G
P S
M
D
P
D D
R A
F

(27)

|
|
.
|

\
|
+ +
=
2 cos 2 2
2 km K
th
G
A
M
D D P
M
F

(28)

Where:
2
h w
km
d D
D
+
= (29)

{ }
(

+ + =
2 cos 2 2
, 2 ,
4 , ,
km MIN K
th
MIN G
M MAX A
D D
P
F M

(30)

( )
( )
% ,
% ,
, ,
1
1
TOL A
TOL A
MAX A MIN A
M
M
M M
+

= (31)

Where:
NOM A
TOL A
TOL A
M
M
M
,
,
% ,
= (32)

( )
2
, ,
,
MIN A MAX A
TOL A
M M
M

= (33)

{ }
2 cos 2 2
9 . 0
, 2 ,
,
,
km MIN K
th
MIN G
MAX A
MAX M
D D
P
M
F

+ +
= (34)

{ }
2 cos 2 2
0 . 111
, 2 ,
,
,
km MAX K
th
MAX G
MIN A
MIN M
D D
P
M
F

+ +
= (35)
|
|
.
|

\
|
=

2
cos 2
2
P
F
M
D
M
G th
G
(36)

( )
M km
G A
K
F D
M M
=
2
(37)

REIFF ON FASTENER TORQUE SPECIFICATIONS 11
The example above is simulated again, but this time with the torque specification limits
(M
A,MIN
and M
A,MAX
) and the bolt preload limits (F
M,MIN
and F
M,MAX
) calculated using Eqs 30, 31,
34, and 35. The results are shown in the second column of Table 1 and in Figs. 3 and 4:

1. The probability that the bolt preload will be outside of the preload limits (F
M,MIN
and F
M,MAX
)
is approximately equivalent to 3 limits for a normal distribution. This is the desired result.
2. The probability is equivalent to 4 limits (32 PPM) that the ratio F
M
/F
M,
exceeds 100 %.
This is the desired result.

Considerations for Setting Exploitation Factor
Normally, the exploitation factor () is a factor to provide for a margin against yielding the
bolt during tightening to account for an expected increase in bolt load during service. A
reduction in the exploitation factor should also be considered in order to compensate for
inaccurate estimates of the limiting values of head and thread friction, for example, if a small
sample size is used in the laboratory test.

Statistical Tolerance Method
M10 Flange bolt
0
50
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
2
1
2
6
3
1
3
6
4
1
4
6
5
1
5
6
Fm Fm,
F
r
e
q
u
e
n
c
y
F
M,MIN
F
M,MAX
Statistical Tolerance Method
M10 Flange bolt
0
50
100
150
200
250
300
350
400
450
500
550
600
650
5
0
%
5
5
%
6
0
%
6
5
%
7
0
%
7
5
%
8
0
%
8
5
%
9
0
%
9
5
%
1
0
0
%
1
0
5
%
Fm/Fm,
F
r
e
q
u
e
n
c
y
4 (equiv.)

FIG. 3Distribution of F
M
and F
M,
FIG. 4Distribution of F
M
/F
M,
for the EVM.
for the EVM.

Conclusion
For known values of bolt geometry, bolt material properties, prevailing torque, and fastener
head and thread friction, the Extreme Value Method calculates the torque specification limits
(M
A,MIN
and M
A,MAX
) and the bolt preload limits (F
M,MIN
and F
M,MAX
) so that the probability is
JOURNAL OF ASTM INTERNATIONAL 12
approximately 32 ppm (equivalent to 4 for a normal distribution) that the bolt equivalent stress
will exceed the desired maximum allowable stress (R
P,0.2
). In other words, if is set to 100 %,
then the probability is approximately 32 ppm that the equivalent stress in the bolt will exceed the
minimum yield stress R
P,0.2
. This is quite conservative since few bolts will actually have yield
strength at the minimum specified value.
Application of statistical tolerancing methods results in a modification to the Extreme Value
Method. Equations 2637 describe the method for the simple case where there is no prevailing
torque and the bolt is solid. The result is an increase to the calculated torque specification and an
increase in overall bolt preload, so that the probability is approximately 32 ppm (equivalent to
4 for a normal distribution) that the bolt equivalent stress will exceed the desired maximum
allowable stress (R
y
). In the case where is set to 100 %, the probability is approximately 32
ppm that the bolt will yield, with the actual distribution of bolt yield strength considered. The
result is an 8 % increase in average bolt tension with a very small risk of yielding a bolt during
assembly.

References
[1] Motosh, N., Development of Design Charts for Bolts Preloaded up to the Plastic Range, J.
Eng. Ind., August 1976.
[2] Cramer, H., Mathematical Methods for Statistics, Princeton University Press, 1942.