Professional Documents
Culture Documents
P16608coll23 21134
P16608coll23 21134
by
D. ANDY HISSAM
A DISSERTATION
HUNTSVILLE, ALABAMA
2013
ii
iii
ABSTRACT
The School of Graduate Studies
The University of Alabama in Huntsville
Title Detailed Methodology for Determining Torque Limits to Maximize Preload for High-
Strength Threaded Fasteners
High-strength threaded fasteners are an amazing product of engineering, and one of the
most common methods of making structural attachments. They are frequently used in structural
joints because of two important features: such fasteners allow disassembly and reassembly of the
joint, and they can generate high clamping loads called preload.
To achieve the full benefit of threaded fasteners, the initial preload must be maximized.
The benefits of high preload are well documented and include improved fatigue resistance,
increased joint stiffness, and better resistance to vibration loosening. For joints in shear, preload
resists relative motion, or slip, of the clamped members. Many factors, like elastic interactions
and embedment, tend to lower the initial preload placed on the fastener. These factors provide
additional motivation to maximize the initial preload. Also, in aerospace applications, maximized
preloads help to achieve the full structural capacity from the fasteners, thereby minimizing
weight. Of the available methods for controlling fastener preload, "torque control" is the most
common. Unfortunately, determining the torque to maximize preload is problematic and greatly
complicated by the large preload scatter generally seen with torque control.
maximize preload for high-strength threaded fasteners. The methodology accounts for the large
scatter in preload found with torque control, and therefore, addresses the statistical challenges of
dealing with preload uncertainty. In developing this methodology, the probabilities of yielding
the bolt during installation were calculated. To calculate these probabilities, the expected scatter
iv
v
ACKNOWLEDGMENTS
In all of life's great endeavors, we require the support of others. I could easily list one
hundred people that have helped me during the course of my continued education and this
research. Without their help, this advanced degree and research would not have been possible. I
want to thank everyone that has made a contribution, regardless of how big or small. However,
there are a few people that I would like to give special thanks. These people played a major role
First, I want to thank Dr. Mark Bower for encouraging me to continue my academic
studies. Without his support and encouragement, I would have never pursued this degree. He is
truly missed. I also want to thank Dr. John Gilbert for accepting me as a graduate student and
serving as my doctoral advisor. My first graduate class was with Dr. Gilbert back in 1998, and I
still remember how much I enjoyed his class. Therefore, how appropriate that I complete my
graduate studies working with him. I want to thank Neill Myers for acting as a sounding board
for many of my ideas. Over the past five years, we have had many technical and philosophical
discussions on fastener design and bolt preload. He always provided a fresh perspective when I
was struggling with a problem. I also want to thank John Forbes for his many years of support,
and most recently, for proofreading my dissertation. He performed a very thorough technical
review and provided many excellent editorial comments. I want to thank James Hodo and Bruce
Bice in the MSFC Materials Lab for helping me conduct fastener tests. Their background in
I want to thank my parents, Frank and Rosemary, who from my early days as an
undergraduate, encouraged my education and supported me financially. Their moral support and
encouragement have always been fundamental to my success. I want to thank my wife and
daughter, Deborah and Melaina, for their faithful support. For the past several years, they have
had to tolerate stacks of books in the bedroom, spreadsheets taped to the walls, and a dominated
vi
computer. I could not have done it without them. Finally, I need to recognize my employer,
NASA, for providing me full financial support and the resources to do much of this research.
Ronald McDonald in the training office has been especially patient throughout my many years of
study. I also want to thank David Whitten and Richard Stroud for their support and for
recognizing the importance of this research -- which should ultimately benefit many projects and
programs. Thanks again to all. Your support and friendship are truly appreciated.
vii
TABLE OF CONTENTS
Page
Chapter
I. INTRODUCTION ...................................................................................................... 1
D. Torque Control............................................................................................... 4
F. Objectives/Motivation ................................................................................... 7
G. Scope.............................................................................................................. 8
1. Friction.............................................................................................. 20
viii
D. Literature Review .......................................................................................... 23
D. Testing ........................................................................................................... 39
V. CONCLUSIONS ........................................................................................................ 89
ix
APPENDIX B: Randomized Order for Prevailing Torque Tests ....................................... 101
x
LIST OF FIGURES
Figure Page
2.4 Depiction of Friction in Bolt Thread and Under the Nut During Tightening ............. 21
2.5 The Central Limit Theorem Predicts the Preload PDF Will Be Normal .................... 23
3.2 Statistical Interference Between Installation Stress and Bolt Strength ....................... 35
4.1 Axial and Torsional Stress with Constant Input Torque (1/2"-20 Fastener
Torqued to 925 in-lbs) ............................................................................................ 48
4.2 Effective and Axial Stress with Constant Input Torque (1/2"-20 Fastener
Torqued to 925 in-lbs) ............................................................................................ 49
4.3 Axial and Torsional Stress When Effective Stress Reaches 112,500 psi
(1/2"-20 Fastener) ................................................................................................... 50
4.4 Input Torque to Generate 112,500 psi Effective Stress, Including Effects
of Prevailing Torque (1/2"-20 Fastener) ................................................................. 51
4.5 Input Torque to Generate 112,500 psi Effective Stress for Different
Bolt Sizes ................................................................................................................ 52
4.6 Achievable Preload When Bolt is Torqued to 112,500 psi Effective Stress
for Different Bolt Sizes ........................................................................................... 53
4.7 Percent Preload Lost to Torsion for Different Bolt Sizes (each torqued to
112,500 psi effective stress) .................................................................................... 54
xi
4.8 Percent Preload Lost to Average Prevailing Torque for Different Bolt Sizes
(each torqued to 112,500 psi effective stress) ......................................................... 55
4.9 Magnitude of Effective Stress in Bolt Cross Section for Various Conditions
(1/2"-20 Bolt) .......................................................................................................... 56
4.10 Overlap of Distribution Curves for Different Population Coverages ......................... 67
4.11 Plot of k-values for 90%, 95%, and 99% Coverage with 95% Confidence,
One-sided ................................................................................................................ 70
4.12 Torque-Tension Plot for 3/8" Bolts ............................................................................ 71
4.16 Normal Probability Plot for Preload Sample of Unmodified Lock Nuts .................... 80
A.1 Preloaded Bolt with Strain Gauge Wired into Wheatstone Bridge ............................. 97
C.1 Axial and Torsional Stress with Constant Input Torque (1/4"-28 Fastener
Torqued to 75 in-lbs) .............................................................................................. 103
C.2 Effective and Axial Stress with Constant Input Torque (1/4"-28 Fastener
Torqued to 75 in-lbs) .............................................................................................. 104
C.3 Axial and Torsional Stress When Effective Stress Reaches 112,500 psi
(1/4"-28 Fastener) ................................................................................................... 104
C.4 Input Torque to Generate 112,500 psi Effective Stress, Including Effects
of Prevailing Torque (1/4"-28 Fastener) ................................................................. 105
xii
LIST OF TABLES
Table Page
4.5 Test Results for 1/2" Bolts When Input Torque Equals 800 in-lbs ............................. 73
xiii
LIST OF SYMBOLS
Fn normal force
J polar second moment of area based on the bolt's tensile stress area, in4
K nut factor
P thread pitch, in
Tτ thread reaction torque (torque that generates torsional stress in the bolt), in-lbs
xiv
β thread half-angle, degrees
γ utilization factor
μ coefficient of friction
σn bolt axial preload stress (at the radius of the tensile stress area), lbs/in2
bolt torsional shear stress (at the radius of the tensile stress area), lbs/in2
CV coefficient of variation
Fo test statistic to determine the equality of variances from two normal distributions
n number of samples
xv
sFT covariance factor for Fi and Tτ due to common elemental random error sources, in-lb2
to test statistic to determine the equality of the means of two normal distributions with
unknown variances
̅ sample mean
z z-value
α significance level, %
μD difference between the means of bolt yield strength and effective stress , lbs/in2
σD standard deviation of the difference between means of bolt yield strength and effective
stress, lbs/in2
xvi
LIST OF ACRONYMS
CV Coefficient of Variation
UN Unified
xvii
Chapter I
INTRODUCTION
This chapter includes the purpose of the study, a statement of the problem, the motivation
for the research, and an outline of the dissertation. It also provides a brief description of three key
elements of this research: bolt preload, torque control, and prevailing torque.
Threaded fasteners have long been one of the most common methods for making
structural attachments. Such fasteners are used in structural joints because of two important
features: they allow disassembly and reassembly of the joint, and they can generate high
clamping loads known as preload. The latter is important since the full benefit of a threaded
Of the available techniques for controlling fastener preload, “torque control” is the most
common. After installation, a small percent of the initial preload is lost to phenomena such as
embedment. Embedment occurs when high spots on the threads and joint plastically yield due to
the initial high contact pressure. The remaining preload, or residual preload, is the preload that
The purpose of this study is to develop a detailed methodology for determining torque
1
B. Statement of the Problem
In order to maximize the bolt's initial preload using torque control, the installation torque
must be determined. The torque that will maximize the initial preload will stress the bolt to very
near its yield point. The bolt will yield when the effective stress due to tension and torsion
A maximized installation torque, or torque limit, is dictated by the stresses placed on the
bolt during the assembly process. These initial stresses are a combination of tension and torsion.
Figure 1.1 shows the normal (σ) and shear (τ) stresses on the bolt shank created by load, , and
torque, . After the assembly torque is removed, a significant portion of the torsion stress in the
bolt will dissipate. For this research, testing has been conducted to investigate the magnitude of
this drop in stress. Other factors, like embedment, will tend to further decrease the preload stress
in the bolt. These reductions in stress will provide the additional load capacity in the bolt needed
to accommodate the external loads seen in service. From fundamental joint theory, it is known
that a preloaded bolt will only see a small percent of the external load. This percent is largely
dictated by the joint factor and loading plane. Without preload, or if preload is lost, the bolt sees
Tτ
σ
τ
Fi
Tτ
Fi
2
Since torque limits are based on a bolt's initial stress, they can be determined by testing
individual fastener combinations, i.e., a single bolt and nut. Therefore, there is no need for more
Unfortunately, there is a large inherent variability in bolt preload when torque control is
used. This large variability, or scatter, in preload is due primarily to the variability in friction.
Since many factors influence friction, its variability can be high. In order to maximize bolt
preload, preload scatter must be thoroughly addressed. However, existing methods for
determining torque limits fail to fully consider this scatter, serving as a primary motivation for the
study at hand.
The present research is significant because it can benefit all mechanical systems that
contain fasteners. When preload is maximized, not only is the full fastener capacity utilized, but
system safety and reliability are improved. It is well known, for example, that fasteners in highly
preloaded joints are better isolated from insidious vibration loads which are often responsible for
premature joint failure. In a paper by V. E. Kahle, insufficient preload was identified as a root
cause of failure in a helicopter accessory drive. The report states, "failure occurred via a fatigue
mechanism, with the physical root cause for failure attributable to insufficient clamping load in
service, as well as other mechanical factors with respect to the connection [1]."
Three key elements of the present research are: bolt preload, torque control, and
C. Bolt Preload
Bolt preload is the tension placed on a bolt at assembly. This tension can be very high
and often approaches the yield strength of the bolt. High bolt preload has many advantages and is
typically desired in structural joints. Its advantages include improved fatigue resistance,
increased joint stiffness, and better resistance to vibration loosening. High preload will also
3
increase the margin on joint separation. For joints in shear, it resists relative motion, or slip, of
the clamped members. Many factors, like elastic interactions and embedment, tend to lower the
initial preload placed on the fastener. These factors provide additional motivation to maximize
the initial preload. There are many methods of controlling bolt preload. The most common and
economical is torque control and the focus of this research. Torque control is discussed in more
majority of applications. The benefits of high preload have long been recognized. In a 1963
paper by Bernie Cobb, it states, "the rule is, always tighten a bolted assembly to the maximum
permissible preload [2]." More recently, VDI 2230 states, "the aim is to utilize the bolt strength
Finally, in developing this methodology, the bolt is assumed to be the weak link in the
fastener assembly. In other words, it is assumed that the bolt will fail before the nut or clamped
members. The nut should always be able to generate the full strength of the bolt. This is
generally the best accepted design practice. However, if the bolt is not the weak link, adjustments
D. Torque Control
Torque control is one method of applying preload to a bolt. Preload is placed on a bolt
by applying a predetermined torque on the nut (or bolt head), usually applied with a calibrated
torque wrench. Since only the installation torque is measured, torque control is an indirect
method of applying preload. Good design practice requires that fasteners in a structural
attachment be tightened incrementally using a defined pattern. Therefore, the final installation
torque is applied on the last tightening pass for a given group of fasteners.
Torque control is the most common method of applying preload to a bolt [4, 5]. As
Bickford states, "torque is king [5]." It is the most common because it is easy to implement and
4
relatively inexpensive. Unfortunately, it has one major drawback. The preload scatter associated
with torque control is very high, higher than almost any other method of applying preload.
Therefore, torque control is not appropriate when precise preload control is needed and should
not be used for preload critical joints. A preload critical joint is any joint that must develop a
specific preload with minimal scatter in order to operate properly and safely. In general a preload
critical joint cannot tolerate the preload scatter typically seen with torque control. Nevertheless,
torque control is sufficient for a majority of applications. Most joints can tolerate the wide range
There are two general approaches encountered for determining torque limits. The first
approach directly accounts for torsional stress. It considers both the axial preload and the
torsional stresses generated in the bolt during assembly. The second approach only accounts for
axial preload stress. It may account for torsional stress, but only indirectly. It does so by limiting
the axial stress at assembly. For example, the preload stress may be limited to 75% of yield
strength. By taking a percentage of yield strength, the intent is to cover both torsional stresses
and preload scatter. However, to ensure that bolt preload is in fact maximized, the first approach
must be taken.
Determining the torque that will maximize preload is problematic and greatly
complicated by the large preload scatter generally seen with torque control. When a known
torque is applied to a bolt, a preload results. Unfortunately, the resulting preload is a random
variable with large variability. Accounting for preload scatter was one of the main objectives of
this research.
Although more precise (and expensive) methods of applying preload exist, many
aerospace flight programs still use torque control. Two recent examples are the Environmental
Control and Life Support System (ECLSS) and the J2-X engine, both managed at Marshall Space
Flight Center (MSFC). ECLSS, currently flying on the International Space Station, provides
oxygen and drinking water for the flight crew and removes carbon dioxide from the cabin air.
5
Torque values for the ECLSS hardware were taken from MSFC-STD-486, Torque Limits for
Threaded Fasteners [6]. Similarly, the J2-X engine, currently under development, also uses
torque control for most of its threaded fasteners. One exception is the injector-main combustion
chamber interface which uses ultrasonic measurement for preload control. This critical joint
requires greater preload accuracy than can be obtained with torque control.
E. Prevailing Torque
Prevailing torque is the torque required to overcome a locking feature added to fasteners
to help maintain preload and provide additional resistance to vibration loosening. Although high
preload by itself will provide good resistance to vibration loosening, in severe vibration
environments, an additional locking feature is often desired. The prevailing torque locking
feature is generally added to the nut. Nuts with this feature are called prevailing torque nuts or
locking nuts.
When a locking nut is first installed, a torque must be applied to the nut to overcome the
prevailing torque. This torque is applied until the nut comes into contact with the clamped
members. Additional torque is then applied to preload the bolt. It is important to note that
prevailing torque creates no preload in the bolt. This research investigated the impact that
prevailing torque has on achievable preload and how it should be handled at assembly.
A wide variety of locking nuts are available. A common type is the all-metal, deformed
thread, locking nut. It is used in many aerospace applications. In this type, a portion of the
thread is purposely deformed, often on a cylindrical section at the top of the nut. During
installation, the deformed thread then causes interference with the bolt thread. This interference
creates prevailing torque. Another common type is the plastic-insert locking nut. An insert of
material, often nylon, is placed in the thread to create the prevailing torque. A common form uses
a ring of nylon retained at the top of the nut. The inner diameter of the nylon ring is smaller than
the outer diameter of the bolt thread. Therefore, when the nut is installed on the bolt, interference
6
results, creating prevailing torque. Due to the plastic material, these locking nuts cannot be used
in high-temperature applications. This type of locking nut has been evaluated in this research.
F. Objectives/Motivation
The primary objective of this research was to develop a methodology for determining
torque limits to maximize bolt preload for high-strength fasteners. This methodology had to
account for the large preload scatter typically seen with torque control, and therefore, address the
statistical challenges of dealing with this scatter. To meet these challenges, the probability of
yielding the bolt during installation had to be determined. In addition, the preferred sample size
for torque limit testing had to be defined. A secondary objective of this research was to
investigate the impact that torsion and prevailing torque have on achievable preload. Torsion and
prevailing torque will reduce the amount of preload that can be placed on a bolt, but to what
degree?
combination is defined as a specific bolt type, nut, washer(s), and lubricant. By defining specific
fastener combinations, fastener strength and geometry are fixed. Since friction is known to be the
main influence on preload and preload scatter, testing specific fastener combinations limits the
variables that influence friction. Once a torque limit is determined for a specific fastener
combination, it can be applied to any joint using that combination. Doing so eliminates the need
and the cost of lot-specific testing or joint-specific testing. Lot-specific testing requires testing a
sample of the fasteners that will be used in service. Joint-specific testing also requires testing the
exact joint configuration. While these approaches for determining torque limits may be better,
they are also much more costly and not an option for most designs or projects.
The torque limits created by this methodology can be placed in torque tables that are tied
directly to each specific fastener combination. Any joint design using that fastener combination
can use the torque limits shown in its torque table. Each torque table should clearly list the
7
associated fastener combination. The torque table should also list the maximum and minimum
preloads determined during testing. This information would greatly benefit the design engineer
determining torque limits. All existing methods fail to fully address the statistical challenges of
dealing with preload uncertainty (or scatter). Some of the better methods may tell a user to apply
a "minimum" coefficient of friction or a "minimum" nut factor when calculating torque limits, but
they fail to define "minimum." Therefore, a methodology was needed that provides specific
G. Scope
preload is beneficial in most structural applications. However, there are a few exceptions. In
these applications, a maximized preload may not be appropriate. Examples include: joints with
gaskets, joints where the bolt is threaded into a weak material, or joints where thermal loading
selected because they can be preloaded to a high value. High-strength fasteners generally have
ultimate strengths equal to or greater than 120 ksi. Nevertheless, this methodology could be
3) applications where the bolt is tightened slowly. When fasteners are tightened with a
torque wrench, this is considered a slow operation. When fasteners are preloaded quickly, for
Finally, several areas were not investigated by this research. Although they are important
topics in fastener design and analysis, the following were not objectives of this research:
8
1) To develop a new model for describing torque-tension relationships. Mathematical
models already exist that describe how input torque translates into preload. Although one of
these models was applied in this research, there was no intention of developing a new model for
2) To investigate cases where bolts are preloaded to yield. In some applications, the bolt
by Bickford [5] and VDI 2230 [3]. A drawback of preloading bolts to yield is that they can only
3) To develop new methods for selecting bolt sizes and for analyzing bolted joints.
These are vast topics within themselves. To select a bolt size, the minimum required clamping
load (preload) must be determined for the application. A comprehensive joint analysis requires
evaluation of all service loads and environments. In addition, the geometry of the joint must be
evaluated. These subjects are covered in detail by many sources [3, 5, 7, 8, 9].
maximize preload for high-strength threaded fasteners. It addresses the statistical challenges of
dealing with preload uncertainty when attempting to maximize preload. This methodology,
developed for specific fastener combinations, accounts for both torsion and prevailing torque
development of this research. First, the screw thread and the geometry responsible for generating
preload are discussed. Next, two fundamental relationships for describing the torque-tension
interaction are addressed. This is followed by a discussion on preload scatter and preload
distribution and the many variables affecting them. The chapter concludes with a review of the
9
most relevant literature on torque control, bolt preload, and methods for determining torque
limits.
Chapter III outlines the overall approach that was taken to meet the objectives of this
research. The approach includes a combination of analytical studies and testing. Two analytical
studies were conducted which investigated: 1) the relationship between torque, preload, and
stress during bolt installation, and 2) the probability that the tightening process will yield a bolt.
In addition to the analytical studies, two series of tests were performed. The first investigated the
torsional stresses generated in a bolt during tightening. The second investigated the effect of
Chapter IV summarizes the results from this research. It presents the most significant
findings from the two analytical studies and the two series of tests outlined in Chapter III. These
results are used to make key decisions about the final methodology. The chapter concludes by
summarizing the detailed steps of the methodology. Specific recommendations are given for key
statistical calculations.
Chapter V presents the major conclusions from this research. It also provides guidelines
10
Chapter II
BACKGROUND
This chapter provides background material that was important to the development of this
research. First, the screw thread and the geometry responsible for generating preload are
discussed. Next, two fundamental relationships for describing the torque-tension interaction are
addressed. This is followed by a discussion on preload scatter and preload distribution and the
many variables affecting them. The chapter concludes with a review of the most relevant
literature on torque control, bolt preload, and methods for determining torque limits.
The screw is one of the classical simple machines identified centuries ago. However, the
screw can also be considered a special form of another simple machine, the inclined plane.
Fundamentally, the screw thread is an inclined plane wrapped on a helix. Most screws also have
a characteristic of another classical machine, the wedge. Fastener threads, which have an angle
between the threads, wedge against one another as they are tightened. This wedging action
creates additional friction and tends to keep the fastener threads locked together.
The ramp of the thread is a result of the pitch angle (λ). During tightening, one thread is
forced up the ramp of the other thread. As more preload is generated, the normal force between
the two threads increases. As a result, the torque to turn the nut (or bolt) must increase. While
the UN (Unified) thread is the subject of this research, the ramp also appears in all types of
threads including square and Acme. During tightening, sliding occurs in the direction of this
ramp.
11
The wedge effect of the thread is due to the thread angle (generally 60°). This wedge
creates a clamping force on the threads which increases the normal force and friction, an increase
that is relatively large due to the large thread angle. This wedge effect does not occur in square
threads.
B. Torque-Tension Relationships
Several mathematical models exist that relate input torque and preload. These models are
often used to calculate an installation torque. Some are very basic and others are more detailed.
Two of the most common models, and a third variation, are discussed below.
1. Short-Form Equation
The most common and simplest equation relates the input torque to preload by a constant,
called the nut factor, and the nominal bolt diameter. The nut factor attempts to incorporate all the
factors that influence the torque-tension relationship. This equation, often called the short-form
equation, is:
where
When using the short-form equation, a nut factor (K) is either estimated, taken from a
table, or determined by testing. The preload is taken as a percent of either proof strength or
yield strength. This percentage is generally assumed to be constant (e.g., 75% of yield).
Therefore, the short-form equation does not address the role torsion plays in achieving maximum
12
preload. To optimize bolt preload using this approach, the percent of yield strength would need
Using a nominal value of K in the short-form equation will provide a nominal value for
the installation torque. Therefore, a minimum K value (Kmin) is often recommended for
calculating installation torque, and a maximum K value (Kmax) for calculating minimum clamping
loads. Unfortunately, statistical definitions of Kmin and Kmax are generally not provided.
Some might argue that the "nut factor" approach for determining installation torque has
been used successfully for decades. This may be true, but the success was likely due to the
2. Long-Form Equation
Many additional equations are found in the engineering literature that relate input torque
to preload, thread geometry, and the coefficient of friction (COF). One of the first papers to
define this relationship was presented by N. Motosh in the Journal for Engineering for Industry in
1976 [10]. His paper shows how input torque results in bolt stretch, thread friction, and under-nut
friction. Other sources show similar equations [3, 5, 11, 12, 13]. On first observation, these
equations may appear to be different. But in each case, with a little algebraic manipulation,
substitutions, and/or simplification, these equations can be shown to be the same. The most
common form of this equation, discussed by Bickford in Introduction to the Design and Behavior
of Bolted Joints [5], is shown below. Often referred to as the "long-form equation," it is derived
from a free-body diagram of a fastener and the summation of forces experienced during the
tightening process.
( )
13
where
The thread geometry and terminology used in Equation (2.2) are shown in Figure 2.1.
force normal
to thread
λ = pitch
angle
2rb
β = thread
half-angle
2rt
P = thread pitch
2β
14
When expanded, this equation becomes:
Therefore, the total input torque can be viewed as the summation of three torque terms:
. (2.4)
The first term is the torque required to stretch the bolt. The mechanical work done by this
torque is converted totally into strain energy. This strain energy is stored as stretch and twist in
the bolt. The second term ( ) is the torque needed to overcome friction between the bolt and
nut threads. The mechanical work done by this torque is converted into both strain energy and
heat due to friction. The strain energy is stored as twist in the bolt. The heat is generated as the
threads slide over one another. The third term is the torque needed to overcome friction
between the rotating nut and washer. The mechanical work done by this torque is converted
totally into heat due to friction. This heat is generated as the nut slides over the washer. By
observing the long-form equation, the relative magnitudes of these torque terms are shown to be
approximately 10%, 40%, and 50%, respectively [5]. Approximately 10% of the input torque
will stretch the bolt, approximately 40% will be lost to thread friction, and the remainder is
needed to overcome bearing surface friction. Therefore, the majority of input torque goes into
overcoming friction. The relative proportions of these terms are shown in Figure 2.2.
The long-form equation will be used to investigate the impact that torsion and prevailing
torque have on achievable preload and how torque is reacted in a structural joint. Although the
long-form equation itself is not used in the final methodology for determining torque limits, it
15
10% Bolt Stretch
Ti = Tb+ Tth + Ts
Input Torque (Ti)
Tb+ Tth
When this equation is used to calculate an installation torque for a particular fastener, the
nominal geometry values for P, rt , β, and rb are typically used. Although a tolerance is
associated with each of them, these tolerances are small and contribute very little to the preload
scatter. These tolerances are given in fastener and thread specifications. Unlike the nominal
values used for geometry, the minimum coefficient of friction (μmin) should be used in these
calculating an installation torque. Similarly, the maximum coefficient of friction (μmax) should be
Finally, certain variables are not captured by the long-form equation. For example, the
equation does not reflect the variation between the bolt thread and nut thread due to their
tolerances. This variation can significantly affect the torque-tension relationship [14].
16
3. Long-Form Equation with Prevailing Torque
The long-form equation with prevailing torque is identical to Equation (2.2) except for an
additional prevailing torque term (Tpr). Note that prevailing torque is a separate term that is
independent of preload. However, some in the engineering community argue that prevailing
( )
The magnitude of prevailing torque varies with the size of nut. Locking nut
specifications define the acceptable range of prevailing torque. Table 2.1 provides a list of nut
sizes and the associated maximum prevailing torque. It also lists the minimum suggested
installation torques for grade 8, lubricated bolts according to MSFC-STD-486 [6]. The ratio of
maximum prevailing torque to minimum applied torque is also given. As Table 2.1 shows, the
prevailing torque can be a substantial percent of the installation torque, especially for smaller
thread sizes.
17
Table 2.1 Comparison of Prevailing Torque to Installation Torque
C. Preload Scatter
Preload scatter is the natural variability that occurs in preload when bolts are torqued.
Even when apparently identical fasteners are installed to an identical torque value, there can be
considerable variation in the preload. Preload scatter is often called preload uncertainty. The two
terms are considered synonymous in this paper. With torque control, preload scatter is often
reported at ±25% when lubrication is applied to the fasteners. This percentage jumps to ±35%
when lubrication is not used. Although preload scatter occurs in all preload methods, it is
It is important to realize that the large variability seen in preload during testing is not due
to the errors associated with the measurement system. Instead, it is due to the large number of
factors that affect the torque-tension relationship. During testing, preload and input torque can be
measured with a high degree of accuracy. Although preload is not measured in the field, the
18
Test results presented in [15] show just how dramatic preload variation can be.
Figure 2.3 shows plots corresponding to 50 tests conducted on a series of 1/4 inch bolts over a
two-day period in December, 2008. This scatter in preload occurred even though the bolts were
±35%. However, what do these percentages represent? Generally, this question goes
unanswered. By itself, a percentage does not give an indication of the population coverage. For
example, if the distribution is thin and tall, ±25% will cover much more of the population than
when the distribution is short and flat. The population coverage will be an important
consideration when it comes to the interference analysis discussed in Chapter III. Specifying the
preload scatter in terms of a standard deviation is preferred, for example, ±2 standard deviations
19
or ±3 standard deviations. A standard deviation provides a better indication of the population
coverage, especially if the distribution is normal. Bickford states that, "scatter can usually be
Many experimental studies have been performed to identify the factors that impact the
[5]. The natural variation in these factors also leads to the large preload scatter. Some of these
sources of error are major contributors, while others play a lesser role. The use of a prevailing
torque device in the fastener system introduces yet another source of error and preload
uncertainty. Ultimately, friction has been found to be the main influence on preload uncertainty.
Since many factors impact friction, their collective effect can be considered the main contributor
to preload scatter.
1. Friction
Friction is a force that resists the relative motion between parts. Friction is an empirical
property and can only be determined through testing. The friction force can be increased by
either: 1) increasing the normal force, or 2) increasing the COF. As previously stated, during the
where
20
This equation is valid at the condition of impending slippage, or while the body is slipping. If the
body is about to slip, μ is static friction. If the body is slipping, μ is dynamic friction. Dynamic
During the installation of a bolt, friction forces occur in two locations: 1) in the threads,
and 2) under the nut. A simplified representation of the friction force in fasteners is shown in
Figure 2.4. In this representation, the normal force is created by the bolt preload, and the applied
force results from the applied torque. Although preload is the source of the normal force on the
threads, the geometry of the thread form ultimately determines its magnitude.
Normal Force
(Due to Preload)
Figure 2.4 Depiction of Friction in Bolt Thread and Under the Nut During Tightening
Determining the normal force between the nut and washer is very straightforward.
However, determining the normal force for a fastener thread is challenging due to the complex
geometry of the thread form. First, thread geometry is defined by the pitch angle (λ), the angle of
the ramp that the opposing thread must slide up. The pitch angle is typically less than 10° for fine
threads. Second, thread geometry is defined by the thread angle, which is 60° for UN series
threads. This is the angle that causes a wedging action when the fastener is tightened. These two
angles are then wrapped in a helix to make up the thread. Every surface on the thread is a unique
combination of these two angles and its location on the helix. Friction acts on all thread surfaces
that are in contact. The resulting friction force helps to prevent the fastener from backing off
during use.
21
When applying the long-form torque-tension equation, the COF used should correspond
to the preload value to be placed on the bolt. Often the COF is assumed to be constant and not to
vary with increases in preload. However, testing in this research has shown that this is not always
the case. By using a purely empirical approach to determine torque limits, this non-linearity
becomes a non-issue. In addition, the two COF terms in the long-form equation are often
assumed to be equivalent. However, again, testing conducted for this research has shown that this
Deformed-thread, all-metal locking nuts increase friction by compressing the bolt thread
which increases the thread normal force. This additional friction creates the locking feature and
also causes the prevailing torque. Nylon insert locking nuts also increase friction by compressing
the bolt thread. As the nylon insert locking nut is threaded onto the bolt, a thread is formed (not
cut) in the nylon insert. This creates a compressive force on the bolt thread. The nylon material
also tends to completely fill the area between the bolt threads, increasing the resistance to
vibration loosening.
2. Preload Distribution
Another important consideration when dealing with preload is the shape of its
reports on the shape of preload distribution. For example, a recently released NASA standard for
the analysis of threaded fastener systems, NASA-STD-5020, states that "multiple tests have
shown the distribution to usually be normal" [9]. But, Bickford reports that preload distribution is
The central limit theorem can provide insight on the expected distribution for preload.
Since so many factors (or error sources) influence the torque-tension relationship, the central limit
theorem tells us to expect a normal distribution. Coleman summarizes this fact by stating, "the
central limit theorem states that if X is not dominated by a single error source but instead is
22
affected by multiple, independent error sources, then the resulting distribution for X will be
approximately normal [16]." So unless one factor (or error source) dominates all others, a normal
distribution can be expected. Even though the error sources themselves may not follow a normal
distribution, the resulting preload will be normal due to the large number of error sources.
Preload PDF
Figure 2.5 The Central Limit Theorem Predicts the Preload PDF Will Be Normal
D. Literature Review
A thorough literature review was conducted as part of this research. The reviewed
material came from many sources including books, journal papers, fastener specifications and
standards, Internet sites, and publications from technical societies like the Industrial Fasteners
Institute (IFI).
Many books have been written on the subject of bolted joint design and analysis. Two of
the most prominent are Bickford's Introduction to the Design and Behavior of Bolted Joints [5]
and the Handbook of Bolts and Bolted Joints [7]. They provide a thorough overview on all
23
subjects related to the bolted joint. Although they discuss methods for determining installation
torque, they do not completely address all of the issues with preload uncertainty (scatter). Books
on mechanical design, including Shigley's Mechanical Engineering Design [12] and Juvinall's
Fundamentals of Machine Component Design [13], also provide details on bolted joint design and
determining installation torque. But, again, they fail to fully address the statistical nature of
dealing with preload uncertainty. Similarly, a 2008 report from Sandia National Labs [17]
documents the current state-of-the art in bolted joint design and analysis. Although the preload
uncertainty associated with torque control is discussed, it does not completely address the impact
Two important papers on preload and torque control were written by Reiff [18, 19]. In
the first paper, he develops a torque-tension relationship that includes prevailing torque. He
argues that prevailing torque must be considered when defining installation torque, otherwise
errors will result. He defines the torque-tension relationship in terms of coefficient of friction and
prevailing torque (basically a variation of the long-form equation with prevailing torque). In the
second paper, he begins to address statistical tolerancing. Again, he defines the torque-tension
relationship in terms of coefficient of friction. The minimum and maximum values of preload are
taken as three standard deviations (3σ) from the mean. Unfortunately, he does not provide any
rationale or justification for using 3σ values. He also does not address the statistical penalty that
joints is developed. He then uses these friction coefficients to relate installation torque to
preload. But, he doesn't fully consider the variability that can exist in preload for a given set of
parameters. He calculates a nominal preload value (again using the long-form equation) based on
installation could benefit all areas of bolted joint analysis, thereby reinforcing the need for the
24
development of the proposed methodology. One area of active research in joint analysis is self-
loosening. In a study by Nassar and Housari [21], it is shown that "for a known amplitude of the
cyclic excitation, bolt tension will have to be below a threshold value in order for the self-
loosening to start." Therefore, if the preload is above this threshold value, self-loosening can be
prevented. In another study by Dinger and Friedrich [22], preload generation and self-loosening
The self-loosening process is characterized by either complete or localized slip at the screw head
and thread contact surfaces. Finally, in paper by Kim, et al [23], different methods for modeling
the bolted joint were investigated. Four types of finite element models were studied -- all models
Many specifications, standards, and handbooks on bolted joint design and analysis are
also available. The most comprehensive specification on this subject is Systematic Calculation of
High Duty Bolted Joints, VDI 2230 [3]. It provides a very detailed methodology for analyzing
high strength bolts and for determining installation torque. Tables of installation torque with
respect to coefficient of friction are included. The specification requires that minimum
coefficient of friction values be used for installation torques, but it fails to define minimum. It
also provides tables of expected coefficients of friction, but the range of values is large, making
coefficient of friction for each specific application, the problem of defining "minimum" still
remains. VDI 2230 recognizes that torsion will limit the amount of preload that can be placed on
the bolt, and therefore accounts for torsion stress when determining torque limits.
Another notable standard, Torque Limits for Threaded Fasteners, MSFC-STD-486 [6],
also provides a method for determining installation torque. It requires that the preload stress not
exceed 65% of tensile yield stress of the bolt. But, using a constant percentage does not address
variability in preload and torsion as friction varies. Therefore, the method used by
MSFC-STD-486 will not maximize bolt capacity for each unique fastener combination.
25
Finally, Threaded Fasteners - Tightening to Proper Tension, MIL-HDBK-60 [11],
provides a method for determining torque values and has similarities to VDI 2230. It defines the
equation. It also accounts for torsion stress when determining torque limits. But, it does not
specifications for conducting torque-tension testing that should be mentioned. These include
ISO 16047, SAE J147, and NASM 1312-15 [24, 25, 26]. Although these specifications provide
detailed steps for conducting the tests, no detail is provided on how to reduce the test data or
calculate maximum or minimum values. ISO 16047 specifically states that it does not apply to
Lastly, dozens of torque tables can be found in fastener catalogs and on the Internet,
mostly for commercial or industrial grade fasteners. Their format and content vary greatly.
Some provide the clamping loads associated with the torque values, but they generally fail to
address the variability that can exist in these clamping loads. Some of these torque tables specify
bolt and lubricant, but none define the complete fastener combination as was done in this
methodology. The better ones warn that they provide suggested torque values and are for
guidance only.
26
Chapter III
APPROACH
methodology for determining torque limits to maximize preload for high-strength threaded
fasteners. A secondary objective, closely related to the first, was to investigate the impact that
torsion and prevailing torque have on achievable preload. This chapter outlines the overall
approach that was taken to meet these objectives. The approach includes a combination of
All elements of a fastener system have an impact the torque-tension relationship. This
fact is recognized by MSFC-STD-486 which states, "Among the numerous variables which can
significantly influence the torque induced tensile load in a fastener system are the number and
type of washers used and the type of nut used [6]." Therefore, torque limits should be developed
for specific fastener combinations when feasible, not individual bolts. For this study, a fastener
combination is defined as a specific bolt type, nut, washer(s), and lubricant. Defining fastener
combinations in this way limits the factors that influence friction, the major contributor to preload
scatter. Parameters not directly associated with friction, such as bolt length and clamped
material, are not specified as part of this definition. Two examples of a fastener combination are
shown in Figure 3.1. Note the differences in the geometry of the bolt head and nut.
27
1/4" Fastener Combination #1 1/4" Fastener Combination #2
Although lot-specific and joint-specific testing may provide better results, the cost
associated with this testing is often prohibitive. A much lower-cost alternative is to develop
torque limits for fastener combinations. These torque limits can be used with any joint using
these combinations. In addition, these torque limits can be placed in torque tables which list the
specific fastener combination, therefore, making fastener selection easier. The maximum and
To best explain this methodology, several fundamental concepts related to fastener design
and stress analysis must be defined. These include the distortion energy criterion, the tensile
This methodology was based on a fundamental argument: Install bolts with as much
initial preload as possible without yielding them. To help achieve this goal, the distortion energy
28
For ductile isotropic materials, the distortion energy criterion predicts yielding when the
effective stress (or von Mises stress) reaches the material's yield strength [27]. In other words,
where
Since fasteners, in general, are made of ductile materials, this criterion is appropriate. For
example, the A286 fasteners tested in this research have a percent elongation of 10-15%. In
aerospace design, yield strength is frequently called the "allowable" yield strength.
In this research, the tensile and torsional stresses are based on the bolt's tensile stress
area. The tensile stress area is a theoretical area used in stress calculations for threaded fasteners.
Originally determined through experimentation, the diameter of the tensile stress area was found
to fall between the thread's pitch and minor diameters. Several equations have been developed
The most common equation, and the one used in this research, is:
[ ] [ ]
where
29
is the radius of the tensile stress area.
This area is used to calculate tensile stress in the bolt, and its equivalent radius is used to
calculate the torsional stress. In reality, the state of stress in a bolt is very complex due to the
bolt's geometry. Some localized yielding may occur in the bolt thread and in the fillet of the bolt
head. However, this occurs well before gross yielding and does not compromise the integrity or
function of the fastener. For this application, this localized yielding and associated stress
concentrations have been ignored because preloading a bolt is a static operation. Preload is
typically applied a limited number of times, and often only once. In addition, fasteners have the
ductility to accommodate small amounts of localized yielding that might occur. Nevertheless,
when cyclic loading (dynamic loading) occurs in the service life of the bolt, a fatigue analysis
should be performed.
To implement this methodology, the yield strength of the bolt must be known. The yield
strength of commercial-grade bolts are generally provided by the bolt specification (e.g., SAE
J429). These specifications typically provide ultimate, yield, and proof loads. Unfortunately, the
same is not true for aerospace-grade bolts (e.g., NAS bolts). In general, aerospace bolt
specifications only provide the bolt's ultimate tensile strength. The designer and analyst must
either estimate the yield strength or determine it by testing. The methodology developed in this
dissertation specifies testing as the method for determining yield strength. However, when a
yield strength is provided by a bolt specification, its value can be used instead.
Testing to determine the yield strength of bolts is slightly different from testing standard
tensile specimens. Standard tensile specimens have a uniform cross-section in the gauge section.
In contrast, the cross-section of the bolt changes along its length, which includes the bolt head,
30
Fastener test specifications identify two primary methods for determining the yield
strength of bolts: 1) the Johnson's 2/3 method, and 2) the parallel offset method. The Johnson's
2/3 method is specified by NASM 1312-8 (Fastener Test Methods, Method 8, Tensile Strength)
[28] and MSFC-STD-486 (Torque Limits for Threaded Fasteners) [6]. The parallel offset method
is specified by ASTM F606 (Standard Test Methods for Determining the Mechanical Properties
of Externally and Internally Threaded Fasteners) [29] and ISO 898-1 (Mechanical Properties of
Fasteners Made of Carbon Steel and Alloy Steel) [30]. Either method is acceptable for
Finally, as the bolt's material strength increases, the yield and ultimate strengths typically
converge. As a result, a very high strength bolt preloaded close to its yield point will be closer to
its ultimate strength (Su) than a lower strength bolt. Also, material ductility generally decreases
as its strength increases. Therefore, for very high strength bolts (> 180 ksi), changes in the
methodology as presented may be desired. For example, the installation limit stress may need to
C. Analytical Studies
Two analytical studies were conducted to help make final decisions about the subject
methodology. The first study investigated the relationships between torque, preload, and stress in
a preloaded bolt by utilizing the long-form torque-tension equation and the distortion energy
criterion. The second study employed a process known as statistical interference to investigate
1. Torque-Preload-Stress Relationships
What are the relationships between input torque, preload, and effective stress in a
preloaded bolt, and how do they vary with changes in the coefficient of friction? In addition,
31
what impact do torsion and prevailing torque have on achievable preload? To address these
questions, the effective stress and long-form torque-tension equations can be used.
where
is the bolt's torsional shear stress (at the radius of the tensile stress area).
where
where
J is the polar second moment of area based on the bolt's tensile stress area.
32
[( ) ( ) ] (3.6)
Input torque is related to preload by the long-form equation presented in Chapter II and is
( )
However, only a portion of the input torque creates torsion in the bolt. The torque associated
with under-nut friction causes no torsional stress in the bolt. The portion of input torque that does
( )
or
. (3.8)
( )
or
. (3.10)
When Equations (3.6) and (3.7) are combined, the resulting equation can be solved for bolt
preload (Fi):
33
= . (3.11)
√ [ ( )]
Now, bolt preload is expressed in terms of COF, effective stress, and geometry. Variations of this
equation are found in [3] and [11]. When the effective stress is set equal to a permissible stress,
the preload will be maximized. Once the maximized preload is determined, Equation (2.2) can be
used to calculate the installation torque (or torque limit) that corresponds to this preload. Many
Equations (3.6) and (3.9). Solving the resulting equation for preload is greatly complicated by the
Both the maximum bolt stress experienced during installation and the bolt yield strength
are random variables. Therefore, when measured, their values will naturally vary from fastener to
fastener. Both variables can be described by probability density functions (PDFs). A PDF is a
mathematical model used to represent the distribution of a continuous random variable. When
plotted together, the PDF of bolt stress and the PDF of bolt strength will overlap. The amount of
overlap will depend on the distance between their means and the flatness of their distributions.
Figure 3.2 shows an example of probability distributions for bolt stress and bolt yield
strength. Since the variability in stress is greater than the variability in strength, its curve is
shorter and wider. The shaded area provides a relative indication of the probability of yielding
the bolt. However, integrating the area does not give the actual probability of yielding. The
method for determining this probability, called statistical interference analysis, is described
34
0.16
Bolt Yield Strength
0.14
Probability Density 0.12
0.1
0.08
0.06
Bolt Stress During
Installation
0.04
0.02
0
40 60 80 100 120 140 160
Bolt Stress and Yield Strength (ksi) Area Represents
Probability of
Yielding the Bolt
Figure 3.2 Statistical Interference Between Installation Stress and Bolt Strength
The probability that a load will exceed strength is given by [31, 32]:
where
This equation integrates (sums) the probability that a load, x, occurs at the same time that the
strength ≤ x. Therefore, the probability that a load equals x, is multiplied by the probability that
the strength ≤ x, and then summed over the full range of x. Remember, in probability, if events
35
are independent, then the probability of them both occurring together is the product of the
When the distributions of strength and load are normal, the mathematics for calculating
this probability is relatively straightforward. From this point forward, the equations for
determining probability are given in terms of bolt yield strength and effective stress. First, a new
variable (D) is defined as the difference between bolt strength and effective stress [31, 32]. The
bolt will yield when D is less than zero. Since both strength and stress are normally distributed,
the new variable (D) will also be normally distributed. The mean of this new variable (μD) is:
where
where
Therefore, the probability of exceeding the yield strength becomes [31, 32]:
( )
where
36
Taking this approach allows the use of the standard normal probability table to determine
probabilities. According to Sarafin [31], even when the distributions are not normal, this
approach can be taken to roughly estimate reliability. However, the assumed distributions are
more important the farther out in the tails the overlap occurs.
Equation (3.12) can also be used to find the probability of yielding for other distributions
mathematics becomes much more complicated. Computerized numerical integration will likely
be required to solve the equation. Weibull distributions are often used to model strength
The coefficient of variation (CV ), defined as the ratio of the standard deviation and the
mean, provides a good indication of the variability (or scatter) in a random variable. That is:
where
The standard deviation by itself does not indicate the degree of variability. In order to fully
describe variability, the standard deviation must be considered in terms of the magnitude of the
mean. A distribution with a large CV will be short and wide while a distribution with a small CV
will be tall and skinny. In Chapter IV, coefficients of variation are used to determine the mean
and standard deviation of yield strength and effective stress. The coefficient of variation is also
Ideally, the methodology for determining torque limits would rely totally on statistics and
probability. Unfortunately, factors exist that cannot be captured by testing. For example, testing
37
environment may not be as controlled, possibly leading to additional variability in the results.
The experience of the installer may also lead to additional variability. In the laboratory, torque
can be applied in a very consistent manner. But, in the field, the experience of the installer could
influence the results. To account for such factors, a knock-down factor may be applied to the bolt
yield strength. This factor accounts for any variable not captured by the torque limit testing.
VDI 2230 applies such a knock-down factor which it calls a "utilization factor" [3]. VDI 2230
recommends a value of 0.9 which will be adopted here. When the bolt's yield strength is
multiplied by this utilization factor (γ), the result becomes the "permissible" bolt stress which is
used to calculate torque limits. In aerospace design, this "permissible" stress is called the "limit"
stress.
3. Tolerance Intervals
The discussion in the previous section on statistical interference was based on the mean
(μ) and standard deviation (σ) of a population. In reality, however, the true population mean and
standard deviation are never known. Instead, their values are estimated from a set of samples.
The greater the number of samples, the better the estimate will be. Since the methodology is
based on a finite number of fastener tests, the effect of sample size must be addressed. It does so
A tolerance interval is a range about the sample mean that has a certain probability of
containing a specified percentage of the parent population. Tolerance intervals can be either
one-sided or two-sided. Since this methodology is interested in the interference of the upper tail
of the bolt stress distribution with the lower tail of the bolt strength distribution, one-sided
tolerance intervals are used. A one-sided tolerance interval is found by first multiplying the
sample standard deviation by a k-value. This product is then added to or subtracted from the
mean to give either the upper or the lower limit of the tolerance interval. The selected k-value
38
depends on the desired population coverage, the confidence level, and the sample size. A small
sample size results in a larger k-value, and therefore a greater "statistical penalty."
Table 3.1 provides a list of k-values for one-sided tolerance intervals for normal
distributions. Values are taken from DeVore [34]. This table provides three levels of population
coverage (90%, 95%, and 99%), each at two confidence levels (95% and 99%). Determining the
One-sided Intervals
Confidence Level 95% 99%
% of Population Captured 90% 95% 99% 90% 95% 99%
2 20.581 26.260 37.094 103.029 131.426 185.617
3 6.156 7.656 10.553 13.995 17.370 23.896
4 4.162 5.144 7.042 7.380 9.083 12.387
5 3.407 4.203 5.741 5.362 6.578 8.939
6 3.006 3.708 5.062 4.411 5.406 7.335
7 2.756 3.400 4.642 3.859 4.728 6.412
8 2.582 3.187 4.354 3.497 4.285 5.812
Sample Size n 9 2.454 3.031 4.143 3.241 3.972 5.389
10 2.355 2.911 3.981 3.048 3.738 5.074
15 2.068 2.566 3.520 2.522 3.102 4.222
20 1.926 2.396 3.295 2.276 2.808 3.832
30 1.777 2.220 3.064 2.030 2.516 3.447
100 1.527 1.927 2.684 1.639 2.056 2.850
200 1.450 1.837 2.570 1.524 1.923 2.679
300 1.417 1.800 2.522 1.476 1.868 2.608
∞ 1.282 1.645 2.326 1.282 1.645 2.326
D. Testing
In addition to the analytical studies above, two series of tests were performed as part of
this research. The first series of tests investigated the torsional stresses generated in a bolt during
tightening. The second series investigated the effect of prevailing torque on preload.
39
1. Preload-Torsion Tests
A series of tests was run to investigate the impact of torsion on achievable preload. The
tests were conducted on a machine specifically designed for torque-tension testing. This machine
and the associated bolt fixture are shown in Figure 3.3. A section view through the bolt and bolt
fixture is shown in Figure 3.4. Preload and torsional stresses were measured every 0.2 seconds as
input torque was gradually applied to the fastener. Preload and torsional stresses were also
measured after the input torque was removed. The decay in preload and torsional stress was then
Test Bolt
Load cell to measure preload (Fi)
Bolt fixture
Bolts were instrumented with strain gauges to measure the torsional shear stress in the
bolt. Figure 3.5 shows one of the instrumented bolts. Although load cells exist for measuring
thread reaction torque directly, none were available for this research. By using strain gauges, the
torsional stress was measured on the shank of the bolt. For this measurement, a two-element
40
strain gauge was bonded to each bolt shank. A schematic of the strain gauge and the
configuration of the Wheatstone bridge are shown in Appendix A. The equation for calculating
the shear stress due to torsion in a preloaded bolt is also derived in Appendix A. During testing,
Equation A.11 was used to convert output voltage (VG) into torsional stress (τ). In these tests,
Fixture for
Holding Bolt
During Test
torsional stress was measured at the major diameter of the bolt. To obtain the torsional stress at
the radius of the tensile stress area , this stress would be multiplied by the ratio of the tensile
stress area diameter and the major diameter, which assumes that torsional stress varies linearly
41
Figure 3.5 Strain Gauge Installed on a 1/2" A286 Bolt
When applying this methodology, the thread reaction torque will be measured, not the
torsional stress. Once the thread reaction torque is known, the torsional stress may be calculated
For these tests, A286 bolts and locking nuts were used. A286 is one of the most popular
materials for aerospace fasteners because of its excellent high and low temperature properties. In
addition A286 is not susceptible to stress corrosion cracking [35]. Both 3/8" and 1/2" bolts were
tested. Figure 3.6 shows pictures of an instrumented bolt installed on the torque-tension machine.
The upper-right picture shows the torque transducer that was used to measure input torque.
42
Strain
Gauge
Torque
Transducer
A series of tests was run to investigate the impact of prevailing torque on preload at high
values. Grade 8 bolts and locking nuts were used for these tests. Two sets of locking nuts with
nylon inserts were tested: one set as received from the manufacturer, and a second set with the
locking feature removed. The locking feature was removed by machining the inside diameter of
the nylon ring until no interference would exist with the bolt thread. This machining was
performed on a standard lathe. Figure 3.7 illustrates the two locking nut configurations.
All locking nuts used for these tests came from the same manufactured lot. If different
lots had been used, a true evaluation of the effect of prevailing torque on preload could not have
been made because other factors besides prevailing torque could have led to differences in the
measured preload.
43
Machined
Nylon Ring
Two simple comparative experiments were conducted. The first investigated whether
prevailing torque impacts the magnitude of preload, especially at higher preload values. This
experiment compared the mean preload with a locking feature to the mean preload without a
locking feature. The null hypothesis for the first experiment was:
The test statistic (to) for this hypothesis test was [36]:
̅ ̅
where
44
and are the sample sizes.
(3.20)
The second experiment investigated whether prevailing torque impacts preload variance.
Variance is simply the square of the standard deviation. This experiment compared the preload
variance with a locking feature to the preload variance without a locking feature. The null
The test statistic (Fo) for this hypothesis test was [36]:
where
(3.24)
All tests were run in a random order to help guard against nuisance factors. A random
number generator in Excel® was used to create the random order. The test order is shown in
45
Appendix B. Note: Non-locking nuts were also tested (as shown in Appendix B). However,
these nuts were tested for future evaluation, so their results are not presented here.
As a final evaluation of this test data, normal probability plots were generated. These
plots provide a graphical means of determining if the data approximates a normal distribution.
46
Chapter IV
RESULTS
This chapter summarizes the results from this research. It begins by presenting the
findings from two analytical studies. The first study investigated the relationships between
torque, preload, and stress. The second study investigated the statistical interference between bolt
stress and strength during the installation process. The chapter then presents the results from two
series of tests. These tests investigated the impact that torsion and prevailing torque have on
in preload and thread reaction torque propagate into the uncertainty in effective stress. The
A. Torque-Preload-Stress Relationships
As described in Chapter III, when the effective stress equation and the thread reaction
torque equation are combined, the resulting equation can be solved for bolt preload. This
combined equation will be expressed in terms of effective stress, geometry, and coefficient of
friction. When the effective stress is set equal to the limit stress, the preload will be maximized.
This maximized preload can be determined for any given coefficient of friction and bolt size.
When the prevailing torque is excluded, solving this equation is relatively straightforward.
However, the presence of prevailing torque greatly complicates the solution. With the help of a
spreadsheet such as Excel®, solving this equation becomes manageable. For this study, the
coefficients of friction in the threads (μt) and on the bearing surface (μb) are assumed equivalent.
47
Large spreadsheets were generated to both make these calculations and to plot the results.
For this study, bolt sizes ranging from 1/4" to 3/4" were selected. Fastener geometry was based
on the NAS 670x series of bolts and NAS 1805 nuts. Note: This section does not address
In Figures 4.1 and 4.2, the variations in stress are shown for a constant input torque. The
torque value was selected from MSFC-STD-486 for 160 ksi bolts lubricated with molybdenum-
disulfide (MoS2) dry film lubricant. In these plots, the maximum effective stress varies.
120000
100000
80000
Stress (psi)
Axial
60000 (Preload)
Stress
40000
Torsional
Stress
20000
0
0 0.1 0.2 0.3
Coefficient of Friction
Figure 4.1 Axial and Torsional Stress with Constant Input Torque (1/2"-20 Fastener Torqued
to 925 in-lbs)
Figure 4.1 shows that for constant input torque, both axial and torsional stresses decrease
with increasing coefficient of friction (COF). There is a dramatic drop in axial stress as much of
the input torque is robbed by friction. There is also a gradual drop in torsional stress. It might be
expected that the torsional stress would increase with increases in friction. However, this plot is
for a constant input torque, so more of the input torque is being absorbed by the increased bearing
48
surface friction. Note: Figure 4.3 illustrates that as input torque is allowed to increase (to
maintain a constant effective stress), the torsional stress does increase with coefficient of friction.
300000
250000
200000
Stress (psi)
Effective
150000 Stress
Axial
100000
(Preload)
Stress
50000
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.2 Effective and Axial Stress with Constant Input Torque (1/2"-20 Fastener
Torqued to 925 in-lbs)
Similarly, Figure 4.2 shows that both axial and effective stresses decrease with an
increase in the COF. This result is not unexpected. The figure also shows that the axial stress
due to bolt preload is the main contributor to effective stress. For lower COF values, torsional
stress increases the effective stress by only 5-15%. Therefore, for well lubricated bolts, the stress
due to torsion does not reduce the achievable preload as much as might be expected. However,
for higher COF values, the torsional stress has a more dramatic impact on effective stress. For
example, as the COF approaches 0.2, approximately 30% of the effective stress is due to torsion.
In Figures 4.3 through 4.8, variations in stress, input torque, and preload are given for a
constant permissible effective stress. In these plots, the input torque varies until the defined
effective stress is reached. For this study, the permissible effective stress was set to 112,500 psi.
49
This value is 90% of the assumed yield strength (125 ksi) for the 160 ksi A286 bolt used in this
analysis.
120000
100000
80000
Stress (psi)
Axial
(Preload)
60000 Stress
40000 Torsional
Lubricated Stress
Fasteners
20000
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.3 Axial and Torsional Stress When Effective Stress Reaches 112,500 psi
(1/2"-20 Fastener)
Figure 4.3 shows the axial and torsional stresses in a 1/2"-20 bolt when the effective
stress reaches 112,500 psi. The axial preload stress drops dramatically as the coefficient of
friction increases. In other words, achievable preload drops as friction increases. In contrast, the
torsional stress increases with increases in coefficient of friction. This increase in torsional stress
is effectively stealing from the achievable preload. This plot clearly shows that to maximize
preload, friction must be reduced in the fastening system. This can be easily accomplished by
applying lubrication to the bolt threads and to the bearing surface (under the nut). With proper
lubrication, the coefficient of friction should fall within the range shown in Figure 4.3.
Over a realistic range of COF values, this plot also shows that axial (preload) stress is
always greater than torsional stress. The same is true for all bolt sizes. At lower COF values, the
axial stress can be 3 to 5 times greater. At higher COF values, it can be 1.5 to 2 times greater.
50
Since torsional stress is always lower than axial stress, torsion contributes less to the effective
stress. So, even given the factor of '3' in the effective stress equation for plane stress (refer to
Equation 3.6), the axial stress will always be the biggest contributor to effective stress.
2000.0
1500.0 No
Input Torque (in-lbs)
Prevailing
Torque
1000.0 Avg
Prevailing
Torque
Max
500.0 Prevailing
Torque
0.0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.4 Input Torque to Generate 112,500 psi Effective Stress, Including Effects of Prevailing
Torque (1/2"-20 Fastener)
Figure 4.4 shows the input torque required to generate an effective stress of 112,500 psi
in a 1/2"-20 bolt. As expected, this input torque increases as friction increases. Still, even though
the input torque is increased, the achievable preload drops (as shown in Figure 4.3).
This figure also shows the effect prevailing torque has on input torque. At lower COF
values, more input torque is required when prevailing torque is present. However, at higher COF
values, where torsional stresses become a larger percent of the effective stress, the required input
torque actually drops when prevailing torque is present. The prevailing torque values used in this
analysis were taken from the nut specification and are assumed to be constant. In reality, the
51
7000.0
6000.0
Input Torque (in-lbs)
5000.0 3/4-16
Fastener
4000.0 1/2-20
Fastener
3000.0 3/8-24
Fastener
2000.0 1/4-28
Fastener
1000.0
0.0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.5 Input Torque to Generate 112,500 psi Effective Stress for Different Bolt Sizes
Figure 4.5 provides the input torque required to generate 112,500 psi effective stress for
multiple bolt sizes. Again, as the coefficient of friction increases, so does the required input
torque. The additional input torque is needed to overcome increased friction in the fastening
system. In addition, as bolt size increases, so does the percent increase in torque.
52
50000
40000
3/4-16
Fastener
Preload (lbs)
30000
1/2-20
Fastener
20000 3/8-24
Fastener
1/4-28
10000 Fastener
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.6 Achievable Preload When Bolt is Torqued to 112,500 psi Effective Stress for
Different Bolt Sizes
Figure 4.6 shows the achievable preload for different fastener sizes when the fastener is
torqued to 112,500 psi effective stress. Again, as shown in Figure 4.3, preload drops with
increases in coefficient of friction and the corresponding increases in torsional stress. This drop
is more dramatic as the bolt size increases. Therefore, higher preloads can be achieved when
Figures 4.3 through 4.6 clearly show that when determining torque limits, torsional stress
must be considered in order to maximize preloads. In more traditional methods, preload is taken
as a percentage of yield or proof strength, regardless of the torsional stress. Therefore, using
53
40%
0%
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.7 Percent Preload Lost to Torsion for Different Bolt Sizes
(each torqued to 112,500 psi effective stress)
Figure 4.7 shows the impact that torsion has on achievable preload. The percent of
preload lost to torsion is presented for various bolt sizes (each torqued to an effective stress of
112,500 psi). The percent preload lost is found by dividing the achievable preload using torque
control by the achievable preload using pure stretch. At lower COF values, the percent preload
lost ranges from 5 to 15%. At higher COF values, however, this percentage reaches 20% and
higher. Although greatest for the 1/4"-28 bolt, these percentages do not vary significantly with
bolt size.
54
20%
0%
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure 4.8 Percent Preload Lost to Average Prevailing Torque for Different Bolt Sizes
(each torqued to 112,500 psi effective stress)
Finally, Figure 4.8 illustrates the impact that prevailing torque has on achievable preload.
The percent preload lost to prevailing torque is presented for different bolt sizes (again, with each
torqued to an effective stress of 112,500 psi). The average prevailing torque was used for this
analysis, so greater percentages will occur when the prevailing torques are at their maximum
values. The losses are greatest for smaller fastener sizes. For the 1/4" fastener, preload loss is
5% to 7% for low coefficients of friction and increases to 10% for higher values. For fastener
sizes 1/2" and greater, the percent preload lost to prevailing torque is below 3% for lower
coefficients of friction and remains below 6% for higher values. This plot shows that there is a
55
σeff = 99,130 psi σeff = 90,270 psi
Figure 4.9 Magnitude of Effective Stress in Bolt Cross Section for Various Conditions
(1/2"-20 Bolt)
Up to this point, this paper has focused on the effective stress that occurs at the outer
surface of the bolt (at the tensile stress area diameter). But, how does the effective stress vary
across the section of the bolt? Figure 4.9 shows the magnitude of the effective stress in the bolt
cross-section for various loading conditions. The two upper views illustrate the effective stress in
a bolt during installation: the left view for a COF of 0.10 and the right view for a COF of 0.15.
For each, the effective stress at the center is equal to the preload stress. As shown earlier, the
preload is greater when the COF is lower. The two lower views show the magnitude of effective
transverse load is applied, the outer fibers of the beam experience the greatest bending stresses.
These stresses decrease linearly to zero at the neutral axis. So, even if the beam's outer fibers
reach the yield point, the majority of its cross section remains below the yield point, and the beam
56
is far from complete failure. As stress further increases and once a majority of the
cross-section has exceeded the yield point, gross failure will occur as a plastic hinge is formed.
Like a transversely-loaded beam, a bolt preloaded by torque experiences the greatest torsional
stress on the outer fibers of the fastener. These torsional stresses decrease linearly toward the
center of the bolt. However, in contrast to the beam, the fastener also experiences a large axial
stress when torqued. At high preloads, when the fastener’s outer fibers reach the yield point, the
combined torsion and tension place a much larger percentage of the bolt's cross section near the
When the distributions of bolt stress and strength are known, the probability of yielding
the bolt during installation can be determined. The method for calculating this probability, called
In order to perform a statistical interference analysis, the mean and standard deviation of
each distribution must be known. Unfortunately, this information is not always readily available.
For example, material and bolt specifications typically provide minimum values for strength, but
include nothing about its variability. Not even the Metallic Materials Properties Development
and Standardization (MMPDS) provides this information. One standard, ASTM E8/E8M-09
(Standard Test Methods for Tension Testing of Metallic Materials) [37], was found that contains
coefficient of variation (CV) data for materials. CV values for 0.2% offset yield strength are
provided and range from 0.01 to 0.02. According to Sarafin, the “coefficient of variation for
strength … is less than 0.05 for most commonly used, ductile metal alloys [31]." Therefore,
material strength coefficient of variation values from 0.01 to 0.05 were investigated.
The statistical data for bolt stress can be even more elusive. Chapter II stated that preload
uncertainty is often given in terms of a percentage (e.g. ±25%), but most sources fail to define the
statistical meaning of this percentage. One exception is Bickford, who states, "scatter can usually
57
be assumed to equal ±3 times the standard deviation [7]." Data encountered by the author further
supports this claim. However, for this study, the uncertainty in effective stress is needed, not the
uncertainty in preload. The absolute and relative uncertainties in effective stress were found to be
no greater than the uncertainty in preload (see section F in this chapter for a complete discussion
on this topic). Based on this finding, three levels of uncertainty in effective stress were
investigated: ±25%, ±35%, and ±45%. For each case, the percentage was assumed to equal ±3
standard deviations.
(4.1)
where
The z-value is a standardized variable that represents the number of standard deviations from the
mean.
Next, the limit stress is set equal to yield stress. That is:
(4.2)
or
. (4.3)
Preload is maximized when the bolt’s maximum effective stress reaches the limit
(4.4)
58
where
By applying Equations (4.1) through (4.4) and Equation (3.16), the probabilities of
exceeding the yield strength during bolt installation can be calculated. Tables 4.1, 4.2, and 4.3
summarize these probabilities which are given in both parts per million (PPM) and parts per
thousand (PPT). Each table represents a different level of effective stress uncertainty (±25%,
±35%, and ±45%) and lists probabilities for four different strength distributions, each at five
levels of population coverage. The magnitude of the population coverage is dictated by the
calculation is provided in Appendix D and corresponds to the case indicated by † in Table 4.1.
The selected population coverage will dictate the relative position of the two
distributions. Since the maximum effective stress is set equal to the yield strength, the curves
move farther apart as the coverage is increased. For example, in Figure 4.10, the plot on the right
represents 99% coverage for both strength and stress, which corresponds to the case indicated
by † in Table 4.1. The plot on the left represents 95% coverage and corresponds to the case
indicated by ‡ in Table 4.1. Since the overlap in the two curves on the left is much greater, the
chance of yielding a bolt is much greater. From Table 4.1, 21 bolts out of 1000 would be
expected to yield for 95% coverage. This number drops to 2 bolts out of 1000 for 99% coverage.
59
Table 4.1 Probabilities of Exceeding Yield Strength Based on Stress-Strength Interference
(Stress Uncertainty ±3σ = ±25%)
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 3σ and Yield Strength = μ - 3σ (99.87% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 128.87 1.29 125 0.083 100.00 8.33 -3.42 3.09E-04 309 0 1.00
125 0.02 132.98 2.66 125 0.083 100.00 8.33 -3.77 8.16E-05 82 0 1.00
125 0.03 137.36 4.12 125 0.083 100.00 8.33 -4.02 2.92E-05 29 0 1.00
125 0.04 142.05 5.68 125 0.083 100.00 8.33 -4.17 1.53E-05 15 0 1.00
125 0.05 147.06 7.35 125 0.083 100.00 8.33 -4.23 1.15E-05 11 0 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 2.33σ and Yield Strength = μ - 2.33σ (99.0% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
60
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 127.98 1.28 125 0.083 104.68 8.72 -2.64 4.10E-03 4102 4 1.00
125 0.02 131.11 2.62 125 0.083 104.68 8.72 -2.90 1.85E-03 1853 2 1.00 †
125 0.03 134.39 4.03 125 0.083 104.68 8.72 -3.09 9.92E-04 992 1 1.00
125 0.04 137.85 5.51 125 0.083 104.68 8.72 -3.21 6.53E-04 653 1 1.00
125 0.05 141.48 7.07 125 0.083 104.68 8.72 -3.28 5.24E-04 524 1 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 2σ and Yield Strength = μ - 2σ (97.72% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 127.55 1.28 125 0.083 107.14 8.93 -2.26 1.18E-02 11826 12 1.00
125 0.02 130.21 2.60 125 0.083 107.14 8.93 -2.48 6.57E-03 6569 7 1.00
125 0.03 132.98 3.99 125 0.083 107.14 8.93 -2.64 4.12E-03 4122 4 1.00
125 0.04 135.87 5.43 125 0.083 107.14 8.93 -2.75 3.00E-03 2995 3 1.00
125 0.05 138.89 6.94 125 0.083 107.14 8.93 -2.81 2.50E-03 2503 3 1.00
Table 4.1 (continued)
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 1.645σ and Yield Strength = μ - 1.645σ (95% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Safety Factor
125 0.01 127.09 1.27 125 0.083 109.93 9.16 -1.86 3.18E-02 31766 32 1.00
125 0.02 129.25 2.59 125 0.083 109.93 9.16 -2.03 2.12E-02 21182 21 1.00 ‡
125 0.03 131.49 3.94 125 0.083 109.93 9.16 -2.16 1.53E-02 15330 15 1.00
125 0.04 133.80 5.35 125 0.083 109.93 9.16 -2.25 1.22E-02 12219 12 1.00
125 0.05 136.20 6.81 125 0.083 109.93 9.16 -2.30 1.07E-02 10679 11 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 1.282σ and Yield Strength = μ - 1.282σ (90% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Safety Factor
125 0.01 126.62 1.27 125 0.083 112.93 9.41 -1.44 7.47E-02 74722 75 1.00
61
125 0.02 128.29 2.57 125 0.083 112.93 9.41 -1.57 5.77E-02 57737 58 1.00
125 0.03 130.00 3.90 125 0.083 112.93 9.41 -1.68 4.70E-02 46955 47 1.00
125 0.04 131.76 5.27 125 0.083 112.93 9.41 -1.74 4.05E-02 40498 40 1.00
125 0.05 133.56 6.68 125 0.083 112.93 9.41 -1.79 3.69E-02 36935 37 1.00
Table 4.2 Probabilities of Exceeding Yield Strength Based on Stress-Strength Interference
(Stress Uncertainty ±3σ = ±35%)
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 3σ and Yield Strength = μ - 3σ (99.87% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 128.87 1.29 125 0.117 92.59 10.80 -3.33 4.28E-04 428 0 1.00
125 0.02 132.98 2.66 125 0.117 92.59 10.80 -3.63 1.42E-04 142 0 1.00
125 0.03 137.36 4.12 125 0.117 92.59 10.80 -3.87 5.39E-05 54 0 1.00
125 0.04 142.05 5.68 125 0.117 92.59 10.80 -4.05 2.54E-05 25 0 1.00
125 0.05 147.06 7.35 125 0.117 92.59 10.80 -4.17 1.54E-05 15 0 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 2.33σ and Yield Strength = μ - 2.33σ (99.0% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
62
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 127.98 1.28 125 0.117 98.28 11.47 -2.57 5.03E-03 5025 5 1.00
125 0.02 131.11 2.62 125 0.117 98.28 11.47 -2.79 2.63E-03 2629 3 1.00
125 0.03 134.39 4.03 125 0.117 98.28 11.47 -2.97 1.48E-03 1484 1 1.00
125 0.04 137.85 5.51 125 0.117 98.28 11.47 -3.11 9.37E-04 937 1 1.00
125 0.05 141.48 7.07 125 0.117 98.28 11.47 -3.21 6.72E-04 672 1 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 2σ and Yield Strength = μ - 2σ (97.72% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 127.55 1.28 125 0.117 101.35 11.82 -2.20 1.38E-02 13799 14 1.00
125 0.02 130.21 2.60 125 0.117 101.35 11.82 -2.38 8.58E-03 8578 9 1.00
125 0.03 132.98 3.99 125 0.117 101.35 11.82 -2.53 5.63E-03 5632 6 1.00
125 0.04 135.87 5.43 125 0.117 101.35 11.82 -2.65 4.00E-03 3995 4 1.00
125 0.05 138.89 6.94 125 0.117 101.35 11.82 -2.74 3.10E-03 3096 3 1.00
Table 4.2 (continued)
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 1.645σ and Yield Strength = μ - 1.645σ (95% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Safety Factor
125 0.01 127.09 1.27 125 0.117 104.87 12.24 -1.81 3.54E-02 35447 35 1.00
125 0.02 129.25 2.59 125 0.117 104.87 12.24 -1.95 2.56E-02 25617 26 1.00
125 0.03 131.49 3.94 125 0.117 104.87 12.24 -2.07 1.92E-02 19207 19 1.00
125 0.04 133.80 5.35 125 0.117 104.87 12.24 -2.17 1.51E-02 15141 15 1.00
125 0.05 136.20 6.81 125 0.117 104.87 12.24 -2.24 1.26E-02 12631 13 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 1.282σ and Yield Strength = μ - 1.282σ (90% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Safety Factor
125 0.01 126.62 1.27 125 0.117 108.74 12.69 -1.40 8.03E-02 80310 80 1.00
63
125 0.02 128.29 2.57 125 0.117 108.74 12.69 -1.51 6.54E-02 65432 65 1.00
125 0.03 130.00 3.90 125 0.117 108.74 12.69 -1.60 5.46E-02 54564 55 1.00
125 0.04 131.76 5.27 125 0.117 108.74 12.69 -1.68 4.69E-02 46895 47 1.00
125 0.05 133.56 6.68 125 0.117 108.74 12.69 -1.73 4.17E-02 41673 42 1.00
Table 4.3 Probabilities of Exceeding Yield Strength Based On Stress-Strength Interference
(Stress Uncertainty ±3σ = ±45%)
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 3σ and Yield Strength = μ - 3σ (99.87% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 128.87 1.29 125 0.150 86.21 12.93 -3.28 5.14E-04 514 1 1.00
125 0.02 132.98 2.66 125 0.150 86.21 12.93 -3.54 1.98E-04 198 0 1.00
125 0.03 137.36 4.12 125 0.150 86.21 12.93 -3.77 8.19E-05 82 0 1.00
125 0.04 142.05 5.68 125 0.150 86.21 12.93 -3.95 3.85E-05 39 0 1.00
125 0.05 147.06 7.35 125 0.150 86.21 12.93 -4.09 2.15E-05 21 0 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 2.33σ and Yield Strength = μ - 2.33σ (99.0% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
64
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 127.98 1.28 125 0.150 92.63 13.89 -2.53 5.64E-03 5640 6 1.00
125 0.02 131.11 2.62 125 0.150 92.63 13.89 -2.72 3.25E-03 3247 3 1.00
125 0.03 134.39 4.03 125 0.150 92.63 13.89 -2.89 1.94E-03 1944 2 1.00
125 0.04 137.85 5.51 125 0.150 92.63 13.89 -3.03 1.24E-03 1243 1 1.00
125 0.05 141.48 7.07 125 0.150 92.63 13.89 -3.13 8.64E-04 864 1 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 2σ and Yield Strength = μ - 2σ (97.72% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Safety
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Factor
125 0.01 127.55 1.28 125 0.150 96.15 14.42 -2.17 1.51E-02 15064 15 1.00
125 0.02 130.21 2.60 125 0.150 96.15 14.42 -2.32 1.01E-02 10075 10 1.00
125 0.03 132.98 3.99 125 0.150 96.15 14.42 -2.46 6.93E-03 6931 7 1.00
125 0.04 135.87 5.43 125 0.150 96.15 14.42 -2.58 4.99E-03 4987 5 1.00
125 0.05 138.89 6.94 125 0.150 96.15 14.42 -2.67 3.80E-03 3797 4 1.00
Table 4.3 (continued)
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 1.645σ and Yield Strength = μ - 1.645σ (95% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Safety Factor
125 0.01 127.09 1.27 125 0.150 100.26 15.04 -1.78 3.77E-02 37728 38 1.00
125 0.02 129.25 2.59 125 0.150 100.26 15.04 -1.90 2.87E-02 28724 29 1.00
125 0.03 131.49 3.94 125 0.150 100.26 15.04 -2.01 2.23E-02 22293 22 1.00
125 0.04 133.80 5.35 125 0.150 100.26 15.04 -2.10 1.78E-02 17806 18 1.00
125 0.05 136.20 6.81 125 0.150 100.26 15.04 -2.18 1.47E-02 14737 15 1.00
Max Eff Stress = 100% Strength, Max Eff Stress = μ + 1.282σ and Yield Strength = μ - 1.282σ (90% COVERAGE)
Strength Effective Stress Stress-Strength Interference
Yield CV Mean SD Max CV Mean SD Φ Probability PPM PPT Safety Factor
125 0.01 126.62 1.27 125 0.150 104.84 15.73 -1.38 8.37E-02 83677 84 1.00
65
125 0.02 128.29 2.57 125 0.150 104.84 15.73 -1.47 7.06E-02 70550 71 1.00
125 0.03 130.00 3.90 125 0.150 104.84 15.73 -1.55 6.02E-02 60224 60 1.00
125 0.04 131.76 5.27 125 0.150 104.84 15.73 -1.62 5.23E-02 52303 52 1.00
125 0.05 133.56 6.68 125 0.150 104.84 15.73 -1.68 4.64E-02 46371 46 1.00
The final column in these tables lists the factor of safety, which is found by dividing the
yield strength by the maximum effective stress. In all cases the factor of safety is 1.0. This
column was added to make an important point about factor of safety. A factor of safety by itself
gives no indication of the reliability of a part or design. However, as observed in these tables, the
reliability varies greatly with the statistical definition of yield strength and maximum effective
stress. Yet, the factor of safety is the same for all cases.
These tables provide the probabilities of yielding a bolt when the strength and stress
distributions are known (i.e., the true population means and standard deviations are known).
Therefore, the utilization factor mentioned in Chapter III has not been included. The utilization
factor is only applied in the final methodology. When applied, it accounts for factors that may
impact the stress distribution, yet are not captured by the torque-tension testing.
These tables were created for bolts with 125 ksi yield strength. However, similar tables
can be made for bolts with other yield strengths. Although the means and standard deviations
will vary, the probabilities will remain the same. As yield strength increases, both the effective
stress and strength distributions will shift to the right and flatten out. The distance between their
means will also increase. Nevertheless, provided the same coefficients of variation are assumed,
Finally, these tables provide important information needed to define the methodology.
Specifically, they provide results for various percentages of population coverage and coefficients
of variation. These results are needed to define an acceptable probability of yielding the bolt
during installation.
66
0.16 0.16
0.14 0.14
125 ksi 125 ksi
0.12 Yield 0.12 Yield
0.1 Strength 0.1 Strength
95% coverage 99% coverage
0.08 0.08
0.06 0.06
67
0.04 Effective Stress 0.04 Effective Stress
Probability Density
Probability Density
0.02 0.02
0 0
40 60 80 100 120 140 160 40 60 80 100 120 140 160
Bolt Effective Stress and Yield Strength (ksi) Bolt Effective Stress and Yield Strength (ksi)
A tolerance interval provides a range that has a given probability of containing a specified
percentage of the population -- a probability based on a sample from the population. Since the
distributions of effective stress and bolt strength are based on samples (i.e., a defined number of
fastener tests), tolerance intervals are used by the methodology for determining torque limits.
The equations in this section are similar to the equations in Section B. The equations in
Section B are expressed in terms of assumed population parameters (μ and σ). In contrast, the
equations in this section are expressed in terms of estimates of population parameters ( ̅ and S)
The lower limit of the tolerance interval for the bolt’s yield strength is defined as:
̅ (4.5)
where
is the one-sided tolerance interval factor for the yield strength, and
When yield strength is multiplied by the utilization factor discussed above, the product becomes
(4.6)
or
̅ . (4.7)
To maximize bolt preload, input torque is increased until the bolt’s effective stress reaches the
limit stress. This effective stress, the bolt's maximum effective stress , is defined as:
68
̅ (4.8)
where
is the one-sided tolerance interval factor for the effective stress, and
or when:
̅ (̅ ) . (4.10)
The input torque that corresponds to this condition is the torque limit.
From the previous section, it was shown that for 99% coverage, the probability of
yielding the bolt during installation falls between 1 in a 1000 and 6 in a 1000. However, that
analysis was based on a known population mean and standard deviation. In practice, one must
rely on a sample to estimate the population mean and standard deviation. Therefore, a k-value
must be used. A k-value, like the z-value used in the previous section, covers a certain
percentage of the population. However, unlike the z-value, the k-value has an associated
confidence level. A confidence level of 95%, most commonly used in research, is adopted here.
So, while the analysis in the previous section can help us select the appropriate percent coverage,
more information is still required to select the k-value, since the k-value also depends on the
sample size.
Figure 4.11 shows one-sided k-values plotted against sample size for 90%, 95%, and 99%
coverage, all at a 95% confidence level. A distinct knee occurs for sample sizes between 10-20.
Therefore, if at least 20 data points can be afforded (both in time and cost), then the "statistical
69
penalty" associated with sample size is greatly reduced. For very large sample sizes, the k-values
converge to the z-values (1.282 for 90% coverage, 1.645 for 95%, and 2.326 for 99% coverage).
5 k-value,
k-value
99%/95%
4
k-value,
95%/95%
3
k-value,
2 90%/95%
0
0 20 40 60 80 100 120
Sample Size
Figure 4.11 Plot of k-values for 90%, 95%, and 99% Coverage with 95% Confidence, One-sided
D. Preload-Torsion Tests
Preload-torsion tests were conducted to investigate the impact of torsion on the preloaded
bolt. Both torsion and preload were measured during and after application of preload. During
these tests, an electric motor gradually applied input torque to the nut until a specified preload
was reached. The motor then maintained the torque for a short period (10 seconds). The motor
was then reversed until the preload was totally removed. Preload was measured with a load cell
that was integral to the torque-tension equipment. The bolts were instrumented with strain
gauges to measure the torsional stress. In total, eight bolts (four 3/8" bolts and four 1/2" bolts)
were instrumented. Each bolt was preloaded three times (three cycles). The test results from the
70
Figure 4.12 shows curves for the first torque-tension cycle for three of the 3/8" bolts. For
two bolts, the fasteners were lubricated with MoS2 paste (Dow Corning Molykote P-37). For the
third, no lubrication was applied (the nut was silver plated, but silver is a poor lubricant). In
terms of the short-form equation, the slope of each curve equals 1/KD. The steeper the slope, the
lower the fastening system friction will be. The intersection point with the x-axis indicates the
prevailing torque for each bolt's locking nut. For each bolt, the prevailing torque, as measured
with a torque wrench, closely matched the value indicated by the test system's torque transducer.
Lubricated with
10000
MoS2 Paste
8000
Tension (lbs)
6000 Prevailing
Torque
No Lubrication
4000
2000
0
0 100 200 300 400 500 600 700
Torque (in-lbs)
For a given input torque, the preload is noticeably higher when lubrication is used, thus
illustrating the impact that lubrication has on the resulting preload. The installation torque for the
non-lubricated bolt could be increased to achieve the same preload as the lubricated bolt.
71
However, since the torsional stress will be higher, the same maximum preload could not be
Table 4.4 provides the test results for the two 3/8" bolts lubricated with MoS2 paste.
Since each bolt was preloaded three times, there are a total of six tests. Each bolt was tested with
the same type of nut, washer, and lubricant, therefore, they are considered members of the same
"specific fastener combination" as defined by this study. For an input torque of 400 in-lbs, the
table includes the measured preload and torsional stress as well as the calculated effective stress
for each test. The mean ( ̅ ), standard deviation ( ), the relative uncertainty ( ̅ ), and three
times the relative uncertainty (3 ̅ ) are shown at the bottom of the table.
Table 4.4 Test Results for 3/8" Bolts When Input Torque Equals 400 in-lbs
3/8"-24 Bolts
Thread
Preload Torsional Reaction Effective
Preload Stress Stress Torque Stress
Test # (lbs) (lbs/in2) (lbs/in2) (in-lbs) (lbs/in2)
1 6444.7 73401.5 23484.9 172.8 83919.0
2 6937.8 79018.5 20998.2 154.5 86986.7
3 6921.2 78829.6 20829.0 153.2 86692.9
4 7327.2 83452.9 22794.9 167.7 92321.2
5 9558.3 108865.0 21664.9 159.4 115150.7
6 9471.2 107871.9 20648.2 151.9 113645.9
Four important observations can be made from these results. First, the magnitude of
preload stress is approximately four times the magnitude of torsional stress. This result aligns
with the finding presented in Section A of this chapter which predicted preload stresses 3 to 5
72
times greater than torsional stresses for lubricated bolts (see Figure 4.3). Second, the absolute
and relative uncertainties in torsional stress (highlighted in purple) are much lower than the
absolute and relative uncertainties in preload stress (highlighted in blue). This result was true for
all bolts tested. Third, the absolute and relative uncertainties in effective stress (highlighted in
red) are less than the absolute and relative uncertainty in preload stress (highlighted in blue). It
might be expected that the uncertainties in effective stress would be higher than the uncertainties
in preload. However, due to propagation of uncertainty and the form of the effective stress
equation, the uncertainties in effective stress will likely be less than the uncertainties in preload
Fourth, the preload scatter for these bolts (highlighted in green) is much higher than generally
experienced for lubricated bolts. Two test runs experienced especially high preloads which
Similar observations are made for the two lubricated 1/2" fasteners, whose results are
Table 4.5 Test Results for 1/2" Bolts When Input Torque Equals 800 in-lbs
1/2"-20 Bolts
Thread
Preload Torsional Reaction Effective
Preload Stress Stress Torque Stress
Test # (lbs) (lbs/in2) (lbs/in2) (in-lbs) (lbs/in2)
1 13786.3 86218.4 20301.5 366.2 93113.2
2 17446.4 109108.0 20121.5 363.0 114539.0
3 17648.8 110373.9 19703.9 355.4 115529.8
4 13644.1 85329.2 18065.1 325.9 90885.2
5 16340.5 102192.0 16612.1 299.7 106165.4
6 17055.1 106660.8 17189.0 310.1 110738.0
73
In a final series of tests, torsional stress was measured in the bolt shank both during
application of torque and after the installation torque was removed. For these tests, the input
torque was applied with a torque wrench because the torque-tension machine could not release
the input torque (to simulate removal of the torque wrench). The same 3/8" bolts shown in
Figure 4.12 were tested. Each bolt was torqued to 580 in-lbs. The resulting torsional stresses are
shown in Figure 4.13. Since the installation torque was applied with a torque wrench, a saw-
30000
Torsion (lbs/in2)
Lubricated with
MoS2 Paste No Lubrication
20000
10000
20 Minutes
0
0 200 400 600 800 1000 1200 1400
Time (sec)
Figure 4.13 Torsional Stress in 3/8" Bolts During Application of Torque and After Installation
Torque is Removed
Clearly, torsional stress reached its maximum value at the maximum input torque. This
maximum torsional stress is due to the sum of the torque to stretch the bolt (Ts), the torque to
overcome thread friction (Tth), and the prevailing torque (Tpr). Immediately after the torque
wrench was removed (within 1 sec), the torsional stress dropped steeply. This drop in torsion was
74
greater for the lubricated bolts. After this initial drop, the torsion continued to decay gradually
over the next 20-30 seconds. For the 3/8" fasteners, the torsion continued to drop slightly over
the next 10 minutes. After that, the torsion remained nearly constant, even after 24 hours. For
the 3/8" bolts, the total drop in torsion ranged from 50 to 75%. Similar results were observed for
the 1/2" bolts, whose total drop in torsion ranged from 40 to 60%. Again, the torsion remained
The next figure in this section, Figure 4.14, shows the axial load (preload) during these
tests. Like torsion, the preload reached its maximum value at the maximum input torque.
However, unlike torsion, preload did not drop after the installation torque was removed. Not
even a slight decrease in preload was noted, even after 24 hours. This result was surprising
because embedment is usually experienced after a bolt has been preloaded. The embedment can
often lead to a 5% decrease in preload. However, this did not occur in these tests, neither for the
75
1956C32-6-ST1-01.txt 1956C32-7-ST1-01.txt 1956C32-8-ST1-01.txt
14000
12000
8000
6000 No Lubrication
4000
2000
20 Minutes
0
0 200 400 600 800 1000 1200 1400
Time (sec)
Figure 4.14 Axial Load (Preload) in 3/8" Bolts During Application of Torque and After
Installation Torque is Removed
preload. Two simple comparative evaluations of the test data were performed. The first
compared the preload means for two nut types: one with a locking feature and another with the
locking feature removed. The second evaluation compared the preload variances for these same
nuts. Since all the locking nuts came from the same lot, a true comparison between the two types
of nuts could be made. Ten 1/2"-20 nuts of each type were tested. All fasteners were lubricated
with MoS2 paste (Dow Corning Molykote P-37). The preloads were measured at an installation
Table 4.6 gives the results of the first evaluation. Based on the test data, the mean
preload without prevailing torque is approximately 5% higher than the mean preload with
prevailing torque. However, is this result statistically significant? Two significance levels (2.5%
76
and 5%) were considered. At a significance level of 2.5%, the data does not provide sufficient
evidence to conclude that the mean preload without prevailing torque is greater. However, at a
significance level of 5%, there is sufficient evidence to show a difference. Therefore, one can be
up to 95% confident that the mean preload with prevailing torque removed is greater.
Table 4.6 Comparison of Preload Means When Prevailing Torque is Present and When
Prevailing Torque is Removed (1/2"-20 Bolt)
Test
Statistic for
Test Data Summary Means Hypothesis Test
Nut Type Mean SD n to γ α tα , γ Null Hypothesis
Prev. Torque
14557 747 10 1.98 19.1 2.5% ±2.093 DON'T REJECT
Removed
Prev. Torque
13813 923 10 5% ±1.729 REJECT
Present
Table 4.7 shows the results of the second evaluation. Based on the test data, the preload
variance with prevailing torque is approximately 24% higher than the variance without prevailing
torque. However, is this result statistically significant? Again, two significance levels (5% and
10%) were considered. For both significance levels, the data does not provide sufficient evidence
to show a difference. Therefore, even at a significance level as high as 10%, one cannot conclude
that the preload variance (or standard deviation) is greater when the prevailing torque is present, a
somewhat unexpected result. If a greater number of lock nuts had been tested, a difference may
77
Table 4.7 Comparison of Preload Variances When Prevailing Torque is Present and When
Prevailing Torque is Removed (1/2"-20 Bolt)
Test
Statistic for
Test Data Summary Variances Hypothesis Test
Nut Type Mean SD n Fo α F* Null Hypothesis
Prev. Torque 14557 747 10 1.528 5% 3.18 DON'T REJECT
Removed
Prev. Torque 13813 923 10 10% 2.44 DON'T REJECT
Present
F* = Fα, n1-1, n2-1
Using the data from these tests, normal probability plots were also generated. Normal
probability plots are used to determine if data comes from a normal distribution. Data points are
plotted against the z-values that would be expected if the data was indeed normal. When data
points fall close to a straight line, the distribution can be assumed to be approximately normal. In
other words, the more the points deviate from a line, the less likely that the distribution is normal.
Figure 4.15 presents the normal probability plot for the modified lock nuts (without prevailing
torque). Since the points fall reasonably close to a straight line, one can conclude that the preload
78
2
1.5
0.5
z-value
-0.5
-1
-1.5
-2
13000 13500 14000 14500 15000 15500 16000
Preload (lbs)
Figure 4.15 Normal Probability Plot for Preload Sample of Modified Lock Nuts
Figure 4.16 presents the normal probability plot for the unmodified lock nuts (with
prevailing torque). In this case, the points fall further from a straight line, with a distinct hump in
the middle of the plot. Therefore, one cannot easily conclude that this data is normally
distributed. Since all the locking nuts came from the same manufactured lot, one can conclude
that the presence of the locking feature caused the preload distribution to be less normal.
79
2
1.5
0.5
z-value
-0.5
-1
-1.5
-2
13000 13500 14000 14500 15000 15500 16000
Preload (lbs)
Figure 4.16 Normal Probability Plot for Preload Sample of Unmodified Lock Nuts
The data from this testing was also used to compare the coefficient of friction in the
threads ( ) and the coefficient of friction on the bearing surface ( ). The COF on the bearing
surface (under the nut) was found to be approximately 60% greater. At first, this might seem
surprising since the same lubricant was used on both surfaces. However, the COF is also dictated
by variables like surface finish. It is unlikely that the surface finish on the threads and on the
Finally, how should prevailing torque be handled at installation? Two approaches can be
taken. The first approach measures prevailing torque for each locking nut and adds it to the
installation torque (or torque limit). The installation torque is typically defined on an engineering
drawing or an installation specification. The rationale for this approach is to account for the
variability in prevailing torque that can occur between individual nuts. It is argued that since the
magnitude of prevailing torque has no impact on the preload, then adding its measured value to
the specified installation torque should result in a more consistent preload. However, for this
80
approach to be effective, the methodology used to determine the installation torque should not
include the effect of prevailing torque. Otherwise, prevailing torque would be effectively added
twice, once during the testing to determine the torque limit and once again at installation. In
other words, the installation torque should be determined without using the locking nut.
Consequently, testing might not reflect the locking nut's actual geometry or the true impact of
prevailing torque on preload. In addition, the torque-limit testing would not include the torsional
stress created by prevailing torque. These issues represent a significant drawback to this
approach.
The second approach still measures prevailing torque at installation, but only to ensure
that it is within specification. The prevailing torque is not added to the installation torque. The
effects of prevailing torque are captured by the test data that is used to define the installation
torque. Therefore, its effect already appears in the torque limit. As a result, it is not added to the
specified installation torque at assembly. The methodology presented in this paper takes the
second approach. Prevailing torque is treated as another variable that contributes to preload
scatter.
F. Propagation of Uncertainty
The methodology for determining torque limits developed by this research requires three
measurements: preload , thread reaction torque , and input torque . Two of these
measurements ( and ) are required to calculate the effective stress. The uncertainty in both of
these measurements will propagate into the uncertainty in the effective stress. The effective
stress equation containing these two variables, repeated here for convenience, is:
[( ) ( ) ] .
81
As stated in Chapter II, the tolerances associated with fastener geometry are small and contribute
very little to the preload uncertainty. Therefore, nominal values are typically used and are
assumed to be constant.
The propagation of uncertainty in both preload and thread reaction torque can be
determined by the Taylor Series Method [16]. Applying this method, a Taylor series expansion is
used to evaluate the uncertainty in effective stress based on the uncertainties in measured preload
and thread reaction torque. In general terms, the combined standard uncertainty in the calculated
( ) ( ) ( )
( ) ( ). (4.11)
When determining torque limits, the large random uncertainties will easily dwarf the systematic
uncertainties. Preload and thread reaction torque can be accurately measured within ±1%
(a combination of systematic and random uncertainty). This uncertainty is very small compared
to the ±25% to ±35% natural variation in preload (random uncertainty). When a propagation of
uncertainty calculation is made, the small systematic uncertainties will be negligible. Therefore,
Once the systematic uncertainty terms are eliminated, the propagation of uncertainty
( ) ( ) ( ) ( )( )
82
where
measurement,
is the random standard uncertainty (or standard deviation) of the thread reaction
is the covariance factor for Fi and Tτ due to common elemental random error
sources.
The final term in this equation, the correlated effects term, accounts for uncertainties due to
common error sources, that is, error sources that influence both measurements. In uncertainty
analysis, the correlated effects due to random errors are often neglected because measurement
uncertainties are assumed to be independent. As a result, the correlated effects term for random
( ) ( ) ( )
However, for this application, it was not known if the measurements of preload and thread
To check for the presence of a correlated effects term, the uncertainty in effective stress
shown in Table 4.4 can be compared to the uncertainty in effective stress calculated with
Equation (4.13). The uncertainty in Table 4.4 is calculated directly from the test results, that is, it
is the calculated standard deviation of the effective stress. In contrast, the uncertainty obtained
from Equation (4.13) is based on the standard deviations of preload and thread reaction torque
and propagation of uncertainty theory. If the two results are the same (or are nearly the same),
then a correlated effect is not present. If the results are different, then a correlated effect must
exist.
83
From Table 4.4, the uncertainty in effective stress for the 3/8" bolts is 14,172 lbs/in2.
Based on Equation (4.13), the uncertainty in effective stress is 14,437 lbs/in2 -- a 2% difference.
Therefore, this suggests little correlation between the measurements of preload and torsional
stress. Similar results were obtained for the 1/2" bolts. Originally, it was suspected that there
would be a correlated effect, so this result was unexpected. Since these results are based on
Without a correlated effects term, Equation (4.13) predicts that the uncertainty in
effective stress will be less than the uncertainty in preload. This is due to the form of the
effective stress equation which takes the square root of the sum of the squares of the preload
stress and torsional stress. The test results shown in Tables 4.4 and 4.5 confirm this prediction.
For both the 3/8" and 1/2" bolts, the uncertainty in effective stress is less than the uncertainty in
preload.
Once the preload and torsional stresses that cause the effective stress to reach the limit
stress are determined, how then are these stresses tied to an input torque, the torque limit? In a
data acquisition system where multiple time-varying measurements are being made,
measurements are effectively taken at the same instant. Since input torque can also be measured
very accurately, the uncertainty associated with its measurement will be negligible. In this
respect, at a given time, input torque can be treated as a known value. Therefore, once the desired
combination of preload and torsional stress is found, their values can be linked directly to a
G. Methodology Summary
This section describes in detail the methodology for determining torque limits to
maximize preload for high-strength fasteners. The methodology, based on the results and key
findings from this research, can be used to determine torque limits for any fastener combination.
84
Specific recommendations for tolerance interval factors and sample sizes are given. However,
these recommendations could be easily modified to meet the user's needs or preferences. For
example, Tables 4.1 through 4.3 could be used to select an alternate probability of yielding the
• Procure 12 bolts each (of a desired size) from 3 lots, for a total of 36 bolts. Obtaining
bolts from three different lots may not always be feasible. In these cases, bolts of three
different lengths should be procured. This introduces some of the variability that would be
• Perform axial tension (stress-strain) tests on 6 bolts from each lot (18 total tests).
Selecting this number of bolts for testing helps to lower the "statistical penalty" discussed in
Chapter III.
• Determine the yield strength for each bolt using the Johnson’s 2/3 method or the 0.2%
parallel offset method. The yield strength for the sample is then:
̅ (4.5)
A k-value based on 99% coverage and 95% confidence is recommended. This k-value should
come from a one-sided tolerance interval table. When yield strength is provided by a bolt
• Set the limit stress (Slimit) equal to 90% of the bolt yield strength (Sy).
(4.6)
• Procure 6 sets of nuts and washers (of the same size as the bolts) from 3 lots, for a total
of 18 sets. Theses nuts and washers will be used with the remaining 18 bolts for torque-
tension tests.
85
• Specify lubricant, when used, and exactly where it is applied. Also, specify whether the
Note: Since each fastener combination is preloaded multiple times, the effect of preload
"cycle" will be reflected in the torque limit. The variation in preload with "cycle" can be
significant [6]. This variation was observed in this research, especially when lubrication was
• Perform three torque-tension tests for each fastener set (54 total tests). A fastener set
includes a bolt, a nut, a washer for under the nut, a washer for under the bolt head, and a
lubricant. Measure input torque, preload, and thread reaction torque through a full range of
input torque values. During each test, the bolt’s effective stress should not exceed the limit
stress, i.e., the data acquisition system should stop the test when the limit stress (permissible
effective stress) is reached. For each test and for the full range of input torque values,
[( ) ( ) ] (3.6)
• Generate a table (or curve) of input torque vs upper bound effective stress based on the
test data. Determine the upper bound effective stress for the full range of input torque values
using:
̅ (4.14)
A k-value based on 99% coverage and 95% confidence is recommended. The k-value should
• Determine the torque limit (Tlimit) from the input torque vs upper bound effective stress
(4.9)
86
or when:
̅ (̅ ) . (4.10)
Graphically, the torque limit can be found by drawing a horizontal line from the limit stress to
the intersection point on the input torque-upper bound effective stress curve. It is important
to note that this curve may not always be perfectly linear, especially when locking nuts are
used. From this intersection point, a line drawn vertically to the input torque axis defines the
torque limit.
If locking nuts were a part of the fastener combination, then the effect of prevailing
torque will be captured in the test results. Therefore, the prevailing torque is not added to the
• In the torque table, clearly list the applicable fastener combination (bolt, nut, washer(s),
and lubricant), the torque limits with an allowable tolerance, the maximum and
minimum preloads (Fmax and Fmin), and the bolt yield strength (Sy) obtained from
87
Limit stress (S limit) = 90% of yield strength = 0.9(Xstrength - kstrength Sstrength )
Effective Stress (psi)
88
Chapter V
CONCLUSIONS
This research has developed a methodology for determining torque limits for high-
of Chapter IV, maximizes the preload that can be placed on the fasteners during installation.
Once a torque limit is determined for a specific fastener combination, it can be applied to any
joint using the same combination. This research was motivated by the major shortcoming found
in other methods for determining torque limits: failing to fully address the statistical challenges
of dealing with preload scatter. Since preload scatter can be so significant (sometimes exceeding
±35%), it must be carefully considered, especially when maximized preload is desired. To meet
these challenges, the probability of yielding the bolt during installation had to be determined. In
addition, the preferred sample size for torque limit testing had to be defined.
During installation, the scatter (or uncertainty) in the bolt's effective stress was shown to
be less than the scatter in bolt preload. This result was observed in the torque-tension test data
from two bolt sizes. It was also predicted by propagation of uncertainty theory. The expected
scatter in effective stress was needed to determine the probabilities of yielding the bolt during
tightening.
This methodology, recognizing that friction is the main factor affecting the torque-tension
relationship and preload scatter, creates torque limits for specific fastener combinations. A
fastener combination is defined as a specific bolt type, nut, washer(s), and lubricant. The
fundamental approach is to control only the variables that influence friction. Therefore, variables
such as bolt length and clamped member material are not explicitly controlled. As a result, torque
89
limit testing is independent of the bolt length -- eliminating the need to test multiple bolt lengths.
However, this approach does require that each applicable fastener size (1/4"-28, 3/8"-24, etc.) be
tested.
When applying this methodology, tolerance interval factors (k-values) based on 99%
coverage and 95% confidence are recommended for both bolt strength and permissible effective
stress. Using these factors, the probability of yielding the bolt falls approximately between 1 in a
1000 (0.1%) and 6 in a 1000 (0.6%), which in the author's opinion is acceptable for bolt
installation. For most applications, the immediate consequences of yielding a single bolt during
installation are not catastrophic. In addition, yielding only occurs at the bolt's outer diameter, not
across its full section (as shown in Figure 4.9). More importantly, the consequences of not taking
full advantage of maximized preload can be far more serious. In contrast, a k-value based on
minimize the statistical penalty associated with limited testing. This recommendation is based on
the sharp increase in k-values that occurs below 20 data points. Furthermore, in order to capture
the variability between lots, the fasteners for these tests should be selected from multiple lots. A
The specific fastener combinations should be clearly listed with the torque tables to
simplify fastener selection. Design engineers are often required to select an appropriate
combination of bolts, nuts, and washers. Therefore, having specific combinations already defined
simplifies this selection. In addition, the maximum and minimum preload values observed during
testing should always be listed with the torque tables. Sharing this information with the designer
Using this methodology, the prevailing torque of each locking nut should be measured at
installation to ensure that it is within specification. However, the measured prevailing torque is
90
not added to the installation torque because its effect is already included in the installation torque
value.
This research also investigated the impact of torsion and prevailing torque on achievable
preload. The results clearly show that torsion can significantly limit the achievable preload,
particularly when friction is high. At lower COF values, the percent preload lost ranges from 5 to
15%. At higher COF values, however, this percentage can reach 20% and higher. Therefore,
torsional stress must be considered when determining torque limit, otherwise, bolt preload cannot
be consistently maximized.
Prevailing torque is also shown to marginally reduce the achievable preload that can be
placed on the bolt. For the 1/4" fastener, preload loss is 5 to 7% for lower coefficients of friction
and increases to 10% for higher values. For fastener sizes 1/2" and greater, the percent preload
lost to prevailing torque is below 3% for lower coefficients of friction and remains below 6% for
higher values. Even though prevailing torque can be a large percentage of the installation torque
(as shown in Table 2.1), the prevailing torque itself does not add significantly to the effective
stress. Therefore, prevailing torque will reduce the magnitude of the achievable preload, just not
significantly.
Other significant observations and conclusions taken from this research include:
1) During fastener analysis, the coefficients of friction in the threads and under the nut
are sometimes assumed to be the same. However, in one series of tests, the coefficient of friction
under the nut was found to be significantly higher (60%) than the coefficient of friction in the
threads. Therefore, making such an assumption could lead to significant error in the calculated
bolt stress and preload when using relationships like the long-form equation.
2) Torsional stress in the bolt was shown to drop by 40 to 75% immediately after the
installation torque was removed. This drop was greatest when a lubricant was used. Most
91
3) The distribution of preload was found to be approximately normal when prevailing
torque was absent. However, when prevailing torque was present, the distribution of preload
appeared to be less normal. Since all of the locking nuts came from the same manufactured lot,
one can conclude that the presence of the locking feature was the reason for this difference.
lubrication increases the maximum preload that can be placed on the bolt. It also reduces the
5) In addition to a bolt, nut, and washers, this methodology can also be applied to other
easy-to-implement methodology is more likely to be utilized than one that is overly complicated
and hard to execute. The test equipment required to measure input torque, preload, and thread
reaction torque is readily available. Strain gauges were used in this research to measure torsional
stress; however, instrumentation exists for measuring thread reaction torque directly. For
example, load cells available from PCB Load and Torque will measure both bolt tension and
Fortunately, software programs like Excel® make this a relatively easy task. Once spreadsheets
Finally, specifications are available, as needed, for conducting bolt tension and torque-
tension testing. NASM 1312-8 [28] and ASTM F606 [29] provide detailed instructions for
conducting bolt tensile testing. ISO 16047 [24], SAE J174 [25], and NASM 1312-15 [26]
92
Even though this methodology is relatively easy to implement, significant up-front cost
and resources would be required to develop a database of torque limits. This includes the cost of
purchasing and testing the fasteners. However, this cost should be viewed as an investment since
the torque limits developed can be applied to any future joint design that uses the same fastener
combination. In the long run, money is saved by eliminating the need for lot-specific or joint-
specific testing. In a world of limited resources, developing torque limits for specific fastener
Now that a new methodology for determining torque limits has been developed, it should
will require tensile testing of bolts, something not done as part of this research. It will also
require procurement of fasteners from multiple lots. Torque limits for several fastener
Several areas of future work should also be considered. First, this research investigated
the probability of yielding the bolt when the load and strength distributions are normal. Even
when the distributions are not totally normal, the method of statistical interference used in this
research can be used to approximate this probability. However, what about cases where these
distributions deviate significantly from normal? Future work should investigate such cases.
Second, the methodology developed in this research was based on just reaching the yield
point, as shown in Figure 4.9. However, how much yielding can occur before a bolt becomes
compromised or non-functional? In its current revision, VDI 2230 actually makes an adjustment
to the polar moment of inertia that allows the bolt's cross section to go fully plastic. Future work
should investigate the amount of combined stresses that cause the bolt to become non-functional.
93
Third, more testing should be conducted to investigate the correlated effects between the
uncertainties in preload and thread reaction torque. This testing can also be used to directly
Finally, only one type of prevailing torque device was tested as part of this research, the
lock nut with a nylon ring insert. However, as discussed in Chapter I, several additional types of
prevailing torque devices are available. Future work should investigate the effect of other types
of prevailing torque devices on bolt preload, especially the all-metal locking nut which is
94
APPENDICES
95
APPENDIX A
When a bolt is preloaded by torque control, both tension and torsion are created in the
bolt. The current fastener test equipment at MSFC is only capable of measuring bolt tension
(preload) and input torque. For this research, a measurement of torsional shear stress was also
needed. To make this measurement, two-element strain gauges were bonded to the shanks of
multiple bolts as shown in Figure A.1. The two elements of the strain gauge were wired into the
In this appendix an equation is derived for torsional shear stress based on the output
voltage from the Wheatstone bridge. The torsional shear stress is expressed in terms of the
Wheatstone bridge output voltage (VG ), the excitation voltage (VEX ), the gauge factor (G ), and
During assembly, the general state of stress on the outer surface of a preloaded bolt is
shown in Figure A.2. The first stress element is oriented along the axis of the bolt. The second is
oriented at 45° to the bolt axis. The axes of the gauge elements themselves are also aligned at 45°
96
+ε
VG VEX
-ε
Simplified
R4 Schematic
R3
R1
VG VEX
R2
Figure A.1 Preloaded Bolt with Strain Gauge Wired into Wheatstone Bridge
Bolt tension will cause equal strains (both in magnitude and sign) in the two elements of
the strain gauge. Since the strains are equal and since the gauge elements are mounted in adjacent
arms of the Wheatstone bridge, the effect of bolt tension will cancel. Therefore, tension will not
affect the output voltage of the bridge. Similarly, the effects from temperature changes, cross
In contrast, bolt torsion will cause equal but opposite strains in the two elements of the
strain gauge, resulting in a voltage output in the Wheatstone bridge. Since there are two gauge
elements, the output voltage will be doubled. This fact is reflected in the derivation to follow.
97
45°
Bolt
Preload
Bolt
Torsion
BOLT BOLT
AXIS AXIS
Since the effects of tensile stress cancel, the Wheatstone bridge will only output a voltage
due to torsion (assumes no bending occurs in the bolt). Therefore, the strain gauges, wired in this
configuration, only indicate strain due to torsion. The stress elements for torsion only are shown
in Figure A.3.
45°
σ σ
τ
τ
τ
τ
σ σ
BOLT BOLT
AXIS AXIS
98
Derivation of Torsional Shear Stress Equation
First, the excitation voltage (input voltage) for the Wheatstone bridge is related to the
Since the effects of tension cancel, only the resistance changes due to torsion require
where
The gauge factor is provided by the gauge manufacturer and will have an approximate value
For shear due to torsion, σxx = - σyy so Equation (A.7) simplifies to:
99
Solving for σ provides:
Since the magnitude of the normal stress (σ ) and shear stress (τ ) are equal for pure torsion,
Therefore, the shear stress due to torsion is now expressed in terms of one measured quantity (VG )
and four known quantities (E, G, VEX, and ν ), which is the desired result. This equation was used
to convert the output voltage measured during testing into torsional stress values.
100
APPENDIX B
Random Sorted
Numbers Random Test
RAND() Nut Type Numbers Random Test Number
101
Table B.1 (continued)
102
APPENDIX C
This appendix includes the torque-preload-stress plots for 1/4"-28 bolts. These plots are
identical to several of the plots shown in Chapter IV for 1/2" bolts except for the bolt size. They
120000
100000
80000 Axial
Stress (psi)
(Preload)
60000 Stress
40000 Torsional
Stress
20000
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure C.1 Axial and Torsional Stress with Constant Input Torque (1/4"-28 Fastener Torqued
to 75 in-lbs)
103
300000
250000
200000
Stress (psi)
Effective
150000 Stress
Axial
100000 (Preload)
Stress
50000
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure C.2 Effective and Axial Stress with Constant Input Torque (1/4"-28 Fastener
Torqued to 75 in-lbs)
120000
100000
80000
Stress (psi)
Axial
(Preload)
60000 Stress
40000 Torsional
Stress
20000
0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure C.3 Axial and Torsional Stress When Effective Stress Reaches 112,500 psi
(1/4"-28 Fastener)
104
250.0
200.0 No
Input Torque (in-lbs) Prevailing
Torque
150.0
Ave
Prevailing
Torque
100.0
Max
Prevailing
50.0 Torque
0.0
0 0.05 0.1 0.15 0.2 0.25 0.3
Coefficient of Friction
Figure C.4 Input Torque to Generate 112,500 psi Effective Stress, Including Effects of Prevailing
Torque (1/4"-28 Fastener)
105
APPENDIX D
This appendix presents the method used to calculate the probabilities shown in Tables
4.1, 4.2, and 4.3 of this paper. The method for determining the mean and standard deviation of
yield strength and effective stress for a given CV and percent population coverage is also
demonstrated. This sample calculation corresponds to the case identified by † in Table 4.1.
STRENGTH
Given:
- yield strength (Sy) = limit stress (Slimit) = 125 ksi, the approximate yield strength of a
_______________________________________________________________________
(3.16)
(4.3)
_______________________________________________________________________
Therefore
106
and
EFFECTIVE STRESS
Given:
_______________________________________________________________________
(3.16)
(4.4)
_______________________________________________________________________
Therefore
and
107
PROBABILITY
_____________________________________________________________________
( )
√
______________________________________________________________________
Therefore
[ ]
( )
√
which means approximately 2 bolts in a 1000 will exceed the yield strength.
108
REFERENCES
[2] Cobb, Bernie J. "Torque and Strength Requirements for Preloading Bolts." Product
Engineering (1963): 62-66.
[3] Verein Deutscher Ingenieure. Systematic Calculation of High Duty Bolted Joints,
VDI 2230, 2003.
[4] Industrial Fasteners Institute. Fastener Standards. 6th ed. Cleveland: Industrial Fasteners
Institute, 1988.
[5] Bickford, John H. Introduction to the Design and Behavior of Bolted Joints. 4th ed. Boca
Raton: CRC Press, 2008.
[6] NASA Marshall Space Flight Center. Torque Limits for Threaded Fasteners, MSFC-STD-
486B, 1992.
[7] Bickford, John H., and Sayed Nassar. Handbook of Bolts and Bolted Joints. New York:
Marcel-Dekker, 1998.
[8] NASA Johnson Space Center. Criteria for Preloaded Bolts, NSTS 08307, 1998.
[9] National Aeronautics and Space Administration. Requirements for Threaded Fastening
Systems in Spaceflight Hardware, NASA-STD-5020, 2012.
[10] Motosh, N. "Development of Design Charts for Bolts Preloaded up to the Plastic Range."
Journal of Engineering for Industry (1976): 849-851.
[12] Shigley, Joseph E., and Charles R. Mischke. Mechanical Engineering Design. 5th ed. New
York: McGraw-Hill, 1989.
[13] Juvinall, Robert C., and Kurt M. Marshek. Fundamentals of Machine Component Design.
5th ed. Hoboken: John Wiley & Sons, 2012.
[16] Coleman, Hugh W., and W. Glenn Steele. Experimentation, Validation, and Uncertainty
Analysis for Engineers. 3rd ed. Hoboken: John Wiley & Sons, 2009.
109
[17] Brown, Kevin H., Charles Morrow, Samuel Curbin, and Allen Baca. "Guideline for Bolted
Joint Design and Analysis: Version 1.0." Sandia National Laboratories, SAND2008-0371,
2008.
[18] Reiff, John D. "A Procedure for Calculation of Torque Specifications for Bolted Joints
with Prevailing Torque." Journal of ASTM, Vol. 2, No. 3, JAI12879 (2005).
[19] Reiff, John D. "A Method for Calculation of Fastener Torque Specifications Which
Includes Statistical Tolerancing." Journal of ASTM, Vol. 2, No. 3, JAI12878 (2005).
[20] Croccolo, D., M. DeAgostinis, and N. Vincenzi. "Failure Analysis of Bolted Joints: Effect
of Friction Coefficients in Torque-Preloading Relationship." Engineering Failure Analysis
(2011): 364-373.
[21] Nassar, S., and B. Housari, "Effect of Thread Pitch and Initial Tension on the Self-
Loosening of Threaded Fasteners." Journal of Pressure Vessel Technology, Vol. 128,
(2006).
[22] Dinger, G., and C. Friedrich. "Avoiding Self-Loosening Failure of Bolted Joints with
Numerical Assessment of Local Contact State." Engineering Failure Analysis (2011):
2188-2200.
[23] Kim, Jeong, Joo-Cheol Yoon, and Beom-Soo Kang. "Finite Element Analysis and
Modeling of Structure with Bolted Joints." Applied Mathematical Modeling 31 (2007),
895-911.
[25] SAE International. Torque-Tension Test Procedure for Steel Threaded Fasteners - Inch
Series, SAE J174, 2013.
[26] Aerospace Industries Association. Fastener Test Methods, Method 15, Torque-Tension,
NASM 1312-15, 1997.
[27] Cook, Robert D., and Warren C. Young. Advanced Mechanics of Materials. 2nd ed.
Upper Saddle River, NJ: Prentice-Hall, 1999.
[28] Aerospace Industries Association. Fastener Test Methods, Method 8, Tensile Strength,
NASM 1312-8, 2001.
[29] ASTM International. Standard Test Methods for Determining the Mechanical Properties of
Externally and Internally Threaded Fasteners, Washers, Direct Tension Indicators, and
Rivets, ASTM F606, 2011.
[31] Sarafin, Thomas P. Spacecraft Structures and Mechanisms, From Concept to Launch.
Torrance, CA: Microcosm, Inc, 1995.
110
[32] Haldar, Achintya, and Sankaran Mahadevan. Probability, Reliability, and Statistical
Methods in Engineering Design. New York: McGraw-Hill, 2000.
[34] Devore, Jay L. Probability and Statistics for Engineering and the Sciences. 6th ed.
Belmont, CA: Brooks/Cole, 2004.
[35] Barrett, Richard T. Fastener Design Manual, NASA Reference Publication 1228, 1990.
[36] Hines, William W., Douglas C. Montgomery, David M. Goldsman, and Connie M. Borror.
Probability and Statistics in Engineering. 4th ed. Hoboken: John Wiley & Sons, 2003.
[37] ASTM International. Standard Test Methods for Tension Testing of Metallic Materials,
ASTM E8/E8M-09, 2011.
[38] PCB Load and Torque, Inc., accessed October 14, 2013, http://www.pcbloadtorque.com.
[39] Dally, James W., and William F. Riley. Experimental Stress Analysis. 3rd ed. Boston:
McGraw-Hill, 1991.
111