Engineering Failure Analysis 122 (2021) 105279

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

The effects of joint design, bolting procedure and load eccentricity

on fatigue failure characteristics of high-strength steel bolts
Abdul Kareem Abdul Jawwad a, *, Nabeel ALShabatat b, Mofid Mahdi c
a
Industrial Engineering Department, The University of Jordan, Amman 11942, Jordan
b
Mechanical Engineering Department, Tafila Technical University, Tafila 66110, Jordan
c
Mechanical Engineering Department, King Faisal University, Al-Ahsa 31982, Saudi Arabia

A R T I C L E I N F O A B S T R A C T

Keywords: A detailed failure analysis was applied to a set of broken high-strength steel bolts originally fitted
Fatigue failure on a heavy fuel oil meter. The investigation included both experimental and numerical analysis.
High-strength bolts Experimental analysis confirmed failure of the bolts by a premature fatigue fracture process with
Load eccentricity
failure being predominantly at thread runout. The present failure was prompted by the presence
FEM
Stress concentration factor
of relatively high stresses within the bolts – sometimes exceeding the material’s yield strength. A
two-stage finite element modelling (FEM) was used to model loading patterns within the overall
meter assembly and also on individual bolts. FEM results have shown that loading patterns and
the associated tensile stresses within bolts are the result of varying degrees of loading eccentricity,
which itself was found to be the result of a neglectful bolting procedure. Furthermore, failure
location is confirmed to be the result of a major design weakness where thread runout was found
to coincide with flange interface. Contrary to some published results, fatigue crack front curva­
ture was shown by experimental evidence to be related to the degree of axial misalignment be­
tween crack front ends rather than mean stress level. FEM results indicated that stress
concentration factor is inversely related to the degree of load eccentricity and to a much lesser
degree to the applied load. Finally, some operating and design recommendations are made in the
light of present results as preventive measure of such failures in the future.

1. Introduction

1.1. Literature review

Bolted joints are among the most widely used methods of joining in engineering applications. This is mainly due to the wide
availability of bolts and related accessories in different sizes and styles from various materials, with varying mechanical and chemical
properties, in both SI and imperial dimensions. Moreover, bolted joints enjoy the capability and convenience of being assembled,
disassembled and re-assembled. Despite this wide spread of bolted joints and the majority of these joints serving their functions
satisfactorily, many failures have been, and are still being, reported to take place in such joints either within the members being joined
or within the bolts themselves. Fatigue failure is believed to be the predominant failure mode in bolts [1–4]. Fatigue failure of bolts is
normally controlled by several factors including loading conditions, geometry, and fastener material among others. The following

* Corresponding author.
E-mail address: [email protected] (A.K. Abdul Jawwad).

https://doi.org/10.1016/j.engfailanal.2021.105279
Received 12 November 2020; Received in revised form 27 January 2021; Accepted 27 January 2021
Available online 5 February 2021
1350-6307/© 2021 Elsevier Ltd. All rights reserved.
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 1. Most frequent fatigue-failure locations in bolts;(a) under-head-to-shank transition, (b) thread runout and (c) first engaged thread
adapted from [16]

discussion sheds some light on major factors governing fatigue performance of bolts.
Presence of stress concentration sites. Fatigue crack initiation in bolts has been predominantly associated with the presence of stress
concentration sites within bolt geometry. In this regard, three distinctive sites for fatigue failure initiation have been identified in
relevant literature [1–3,5–15]; namely first engaged thread, thread runout and under-head-to-shank transition with relative failure
percentages of 65, 20 and 15%, respectively, Fig. 1 [15,16]. The special geometry of fastener threads results in these threads basically
being a series of notch type stress concentrations [1,12,15]. This is expected to raise the local stresses at thread root to levels much
higher than the average stress within the overall bolt cross section; thereby leading to these sites being preferential sites for fatigue
crack initiation. The level of increase in local stresses at thread roots is dependent on the stress concentration factor (Kt) of the
particular thread which was found to increase with smaller thread root radii [2,6,12,13] and larger bolt diameters [15]. Furthermore,
it has been shown that, upon loading, each thread would also change the stress concentration effect on adjacent threads [1,4,8,17,18].
Charlton and Vancouver [19] postulated that 30% of the load is carried by the first engaged thread while around 75% of the load is
carried by the first three threads and basically the first six threads take the complete load. This pattern has been further supported by
finite element modeling (FEM) of stress distribution in bolts [8]. As such, the first thread root normally sees the highest local stress
levels which is in agreement with this site being the most frequently reported for fatigue crack initiation in bolts [2,3,8,9,12,13,15].
Thread runout is the gradual transition from normal thread depth to full shank diameter (or zero thread depth) in order to avoid a
sudden change to full shank diameter, which for machined bolts is normally achieved over one and a half turns [1,20]. While this may
result in lower stress levels within the last full thread root compared to an abrupt change, it has been shown that higher stress levels are
still present within the thread runout [20]. In terms of fatigue failure initiation, thread runout is especially risky if it coincides with
joint intersection as, in this case, it would be prone to further stress amplitude magnifications imposed by the presence of any possible
eccentricity at the joint interface [1,2,5,7,21,22]. The effect of load eccentricity will be discussed later on in this introduction.
The level of stress concentration at the under-head transition was reported to depend on the under head fillet radius and fastener
size, where smaller fillet radii and larger fastener diameters were reported to result in higher stress concentration levels [23]. It has
been suggested that under head fillet radius should be at least 0.08 D for bolt size up to 19 mm and 0.10 D for bolt sizes larger than 19
mm [17].
Some practical measures have been suggested for lowering stress concentration levels, and thereby improving bolt fatigue strength
including new nut designs [21,24], modified runout form and dimensions [12,25] and using different thread types [12]. These topics
are, however, beyond the scope of the present paper.
Preload. Preload, also known as pretension or clamping force, is the amount of tensile stress imposed on the fastener as a result of
initial tightening of the bolted joint prior to the application of any service loads. Preload is normally determined by the amount of
torque applied to the fastener [26]. If the fastener is likely to experience cyclic loading, the role of preload becomes to limit the
amplitude of cyclic loading on the fastener to below the endurance limit [27]. The application of a proper amount of preload has been
repeatedly emphasized as a safety measure against bolts fatigue failure [1,2,3,7,8,9,13,15,21] with most of bolt fatigue failures being
associated with either inadequate or excessive preload [1,8,9,19,27,28]. As a rule of thumb, a preload value of around 70% of the
bolt’s material yield stress is recommended [29]. Available literature, however, contains some apparent controversy as whether higher
or lower preloads, than this non-specific value, produce longer fatigue lives. While some studies indicated any increase in preload is
likely to result in lowering the endurance limit in proportion to the increase in the bolt’s mean stress created by this preload [2,13],
other studies have indicated that fatigue life increased in proportion to the applied preload [9,27,30–32].
This controversy can better be explained by considering the role of the preload in controlling the cyclic loads imposed on bolts by
service conditions. If the bolted joint is subjected to swinging and/or vibrational motion, as in the case of rotating machinery, street
light posts and wind turbine towers, then increasing preload is expected to improve fatigue performance of bolts and fasteners through
the reduction of vibrational motion amplitude seen by the bolts. This has led some industries including automobile engine manu­
facturers to adopt yield-point tightening in an attempt to reduce the engine weight by reducing the bolt size [33]. On the other hand if
the preload inflicts a load addition on top of service loads, as in the case of pressure containing vessels, then a decrease in bolt’s
endurance limit proportional to the increase in the mean stress would be expected.

2
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 2. Set of eight broken bolts received for analysis.

It is worth mentioning that the amount of preload applied to the fastener was shown not to significantly affect the stress con­
centration factor [2].
Eccentric loading. The presence of loading eccentricity is believed to reduce fatigue life of bolts by an amount proportional to the
increase in local stresses produced by eccentricity [2,8,13–15]. Despite the many failures associated with this phenomenon, very few
experimental studies have been devoted to the evaluation of the effects of load eccentricity on fastener fatigue performance
[13–15,34–37]. Nakagome et al. [36] studied the effect of load eccentricity on fatigue limit by varying the distance between the bolt
axis and the loading position. This variation created different degrees of reactions which in turn changed the net load seen by the bolt.
Their results have shown that the greater distances between the two positions resulted in lower fatigue limits. This is explained by the
fact that longer distances result in higher eccentricity levels and thus higher bending stresses inflicted on the bolts. A series of studies
by Patterson and co-workers [14,34,35,37] have considered several aspects of load eccentricity. Their results have shown that the
nominal stress amplitude that causes failure at a certain number of cycles decreases with eccentricity. As fatigue in bolts is considered
to be a local phenomenon, it would be controlled by the stress state at the thread root rather than the nominal stress. The role of
eccentricity was explained through its effect in changing the ratio of the local stress amplitude to the nominal stress amplitude which in
a way or another reflects the influence of stress concentration [14]. In simple terms eccentricity would adversely affect fatigue per­
formance of bolts in case of “local” fatigue limit, at the thread root, is reached at lower nominal stress as a result of extra bending
stresses inflicted on the convex side of the bolt.
Other factors. In addition to the main three factors discussed above, there are several other factors which have been also reported to
affect fatigue performance of bolts including bolt size, relative dimensions, materials and method of manufacture. The effects of these
factors are not usually simple and significant levels of synergy are reported to exist between the different effects. Griza et al. [38]
indicated that increasing bolt length at the same tightening torque resulted in an increase in the endurance limit of the joint. This was
explained by the reduction in bolt’s stiffness relative to that of the joint (members to be joined). This effect has been substantiated by
the theoretical analysis of Reemsnyder [27] and Grover [39] which indicated that the net load in the fastener increases with increasing
preload, service load and to a lesser extent with the increase in the ratio of joint stiffness to bolt stiffness. The results of Dragoni [40]
have indicated that the endurance limit of bolts slightly increases as the thread pitch is decreased for small low-grade steel bolts, while
on the other hand, major increase of up to 40% in endurance limit was obtained through larger pitch sizes for large high-grade steel
bolts.
Contrary to the general expected trend of dependence of fatigue strength on general strength level, experimental results of Cho et al.
[8] indicated that bolts with higher tensile strength exhibited lower bolt fatigue limits based on nominal stress levels. The authors have
then investigated the effect of material’s strength level on the local stress at thread root using finite element analysis (FEA). Their FEA
results have shown that when the bolt fatigue limit and mean stress are estimated with the stress at thread root the bolt fatigue limit
increases with the tensile strength. Results obtained by O’Brien and Metcalfe [1] indicated, however, that beyond a tensile strength of
around 965 MPa no gain in fastener’s fatigue strength may be obtained through increased material’s strength.
This apparent contradiction between the dependence of fatigue strength on tensile strength between the general (smooth) members

3
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 3. Leaking fuel oil meter with broken bolts.

and bolts was further explained by Dragoni [40] on the basis of interdependencies between stress concentration, which is geometry
dependent, and notch sensitivity, which is material dependent.
The manufacturing route was found to significantly affect bolts fatigue performance through its effects on surface finish, material’s
strength and stress state at thread roots [16,41]. Experimental results [42] have shown that bolts with rolled threads generally have
higher fatigue strength than bolts with machined threads, which may exceed double the value in high strength bolts. The inferior
performance of machined threads was partially explained by the rough surface finish introduced by machining [16]. Marcelo et al.
[41] evaluated fatigue strength of bolts with rolled threads heat treated (quenched and tempered) with different tempering tem­
peratures and also of bolts with thread rolling after heat treatment. Their results indicated that tempering temperature after thread
rolling has no significant effect on fatigue strength of bolts. They found, however, that when thread rolling is carried out after the
quenching and tempering treatment fatigue strength significantly improved. The results were explained by the introduction of
compressive residual stresses at the thread surface including thread roots. In terms of cyclic loading, fatigue life of bolts was found to
decrease with increasing both the maximum load amplitude and stress range [42].

1.2. Brief case history

A set of eight 3/4 in. (~19 mm) broken bolts was received for analysis, Fig. 2. The broken bolts were a subset of a total of 24 bolts
originally fitted on a flanged positive displacement heavy-fuel–oil meter Fig. 3. Seven of these bolts had both the head side and the nut
side available and one bolt had only the nut side available for analysis, Fig. 2. The heads of all failed bolts had the characteristic “Grade-
5′′ symbol, shown in the lower right corner of Fig. 2. The operating company indicated that, before this failure, the meter had been in
continuous operation for four years after a rebuild service with fluid temperature and pressure ranges, of 56 – 62 ◦ C and 40 – 60 psi
(0.27 to 0.41 MPa), respectively. The company indicated, however, that pressure spikes up to 320 psi (2.41 MPa) were occasionally

4
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 4. Type ASTM A307-B bent bolts.

Table 1
Chemical analysis results for the broken and bent bolts (wt.%).
Sample Broken grade-5 bolts SAE J 429 (1999) – Grade-5 Bent bolt (at 8o’clock) ASTM A307 (type B)

Bolt No. 3 Bolt No. 8

C 0.34 0.38 0.28 – 0.55 0.21 0.33 max

Si 0.19 0.20 0.22
Mn 0.96 0.98 0.84 1.25 max
P 0.007 0.008 0.03 0.008 0.041 max
S 0.011 0.012 0.05 0.018 0.051 max
Cr 0.26 0.26 0.06
Mo 0.01 0.01 0.02
Ni 0.06 0.06 0.07

reported.
The client indicated the bolts to have been found out broken after a leak was noticed on the meter with no leaks observed before
that time. The eight broken bolts were fitted “in-a-row” on the bottom part of the meter flange. The leakage was associated with side
occurrences including the ejection of the rubber O-ring and the flange gasket from their original positions, Fig. 3. In addition to the
broken bolts, three non broken bolts, with the characteristic symbol “ A307-B ”, were found bent and provided for analysis, Fig. 4. The
positions of the three non-broken bolts are also shown in Fig. 3 (4, 8 and 12o’clock positions). The new shiny bolts and the eyebolts
seen in Fig. 3 were fitted for meter removal and lifting purposes. The operating company has indicated that upon meter inspection after
the failure “all” the16 remaining bolts were found “finger tight”. No information was available about any details of the final rebuild
service or the tightening torque applied to the bolts.

2. Experimental procedure

The experimental procedure followed in the present case study included most of the routine steps normally carried out during
metallurgical failure analysis such as visual inspection, optical and electron microscopy, fractography, chemical analysis and me­
chanical testing. This is normally done for the dual objective of determining the parts fitness-for-service and the root cause of the
failure. Finite element modelling (FEM) was also incorporated in this study in order to testify and substantiate the failure mechanism
hypothesised based on experimental results as well as obtaining useful and usable information for future design and operating con­
siderations. Before the start of examination each of the broken bolts was given, and marked with, a serial number between 1 and 8

5
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 5. Metallographic sections of broken bolts; macro section (a) and microstructure (b).

Fig. 6. Microstructure of type A307-B bent bolt located at the 8 o’clock position.

based on its respective position on the flange (as informed by the operating company) for ease of comparison and traceability. These
positions are also indicated on the meter flange in Fig. 3.

3. Results and discussion

3.1. Chemical analysis

Chemical analysis was performed for two of the broken bolts (bolts No. 3 and 8) and one of the bent bolts (bolt located at the 8
o’clock position) by means of optical emission spectrometry (OES) according to ASTM E1009 (2006). The results are presented against
corresponding values specified by the relevant standards in Table 1. The values in Table 1 indicate that the chemical composition of the
broken bolts conforms to compositional specifications set by SAE Standard J429 (1999) for Grade 5 steel fasteners and the chemical
composition of the bent bolt conforms to compositional specifications set by ASTM Standard A307 for type-B steel fasteners.

3.2. Metallographic examination

Metallographic specimens, containing longitudinal sections of broken bolts Nos. 2, 3 and 8 and the bent bolt at the 8 o’clock
position were chosen for metallographic examination. The specimens were mounted in Bakelite, ground and polished in accordance
with ASTM E3-01, and then etched in 2% nital solution to reveal the microstructure in accordance with ASTM E407-07.
A macro section of a broken bolt is shown in Fig. 5-a where fatigue crack initiation is observed to have initiated at thread root and
the fracture surface is seen relatively flat. This figure also reveals that the present bolts were manufactured by machining, and not

6
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Table 2
Hardness values for bolts.
Bolt Reading 1 Reading 2 Reading 3 Reading 4 Average SAE J429 (1999) ASTM A307 (type B)

No. 2 (HRC) 27.5 26.8 26.8 25.6 26.7 25–34 ——

No. 8 (HRC) 28.0 26.4 27.6 28.3 27.6 25–34 ——
Bent bolt (HRB) 87.3 85.2 83.1 88.3 86 —— 69 − 95

Fig 7. Association of thread runout with bolts fracture (a) and presence of tinting on fracture surfaces (b).

thread rolling, as the structure was homogeneous throughout the whole section with no evidence of thread rolling.
The microstructures of the three broken bolts were found to be similar and composed mainly of tempered martensite with some
white patches scattered within the structure; most likely being bainite, Fig. 5-b. This microstructure is expected from the chemical
composition and the heat treatment specified in the SAE standard J429 (1999) (Grade-5 fasteners shall be fully austenitized, quenched
and tempered with recommended minimum tempering temperature of 427 ◦ C).
The microstructure of the bent bolt is shown in Fig. 6 and shows a typical ferritic-pearlitic structure commensurate with the
chemical composition observed for this bolt. The directional characteristic of the structure indicates the forming direction during
production.

3.3. Hardness testing

Hardness testing was performed on cross (transverse) sections from broken bolts No. 2 and 8, using HRC hardness measurements,
and bent bolt at the 8 o’clock position, using HRB hardness measurements, according to ASTM standard E18-12. Four readings were
taken from each bolt and the average was calculated and reported. The results are given in Table 2 together with hardness limits
specified by relevant standards. Results in Table 2 indicate both the broken and bent bolts to conform to hardness limits specified by
their corresponding standards.
Results obtained from chemical analysis, metallography and hardness measurements indicate that Grade-5 bolts used on this meter
meet the specifications indicated by the relevant standards and are suitable for the present application. In fact Grade-5 bolts were
recommended by the manufacturer of this meter unit. The use of type ASTM A307-B bolts, however, was not as per manufacturer
recommendations.

3.4. Visual and low magnification examination

Visual examination revealed all the bolts to have failed at the thread runout Fig. 7-a, with varying degrees of bluish tinting observed
on fracture surfaces, Fig. 7-b, which may indicate some oxidation and/or exposure to heat.
The fracture surfaces of the broken bolts were examined at low magnification under a stereomicroscope. Fracture surfaces
belonging to bolt No. 5 are shown, as an example, in Fig. 8-a and b.
The two fracture surfaces exhibit common features normally associated with fatigue failure; namely the relatively flat surface, the
presence of beach marks and striation bands and the division of the fracture surface into two distinct regions of fatigue and final
fracture. Furthermore, higher magnification imaging of the crack initiation site, Fig. 8-c, revealed the presence of numerous ratchet
marks. Ratchet marks are normally the sites where individual cracks meet to form a larger crack before all the cracks meet up and
coalesce into one major fatigue front under the influence of the main acting stress. The presence of relatively large number of ratchet

7
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 8. Fracture surfaces of bolt No.5 (a) head side, (b) nut Side, (c) fatigue crack initiation site and (d) band of coarse striations.

marks is a positive indication that fatigue crack initiation had taken place under severe stress concentration conditions (in this case at
thread root) where, initially, many points along the runout thread root were “competing” for crack initiation. Moreover, bands of
coarse fatigue striations were observed on fractures surfaces. Some studies have mistakenly referred to these features as beach marks.
Beach marks are normally observed on fatigue crack surfaces as indications of major changes in load and are normally separated by
striation bands. Fatigue striations are normally considered as characteristic micro-features of fatigue failures. The fact that the stri­
ations seen in Fig. 8 were visible at relatively low magnification is an indication that fatigue crack has attained sufficient length to
cause large crack propagation “jumps” for each loading cycle, especially under the presence of high stress concentration in accordance
with Paris law. This has been confirmed by larger magnification imaging using scanning electron microscopy (SEM), Fig. 8-d. More
discussion on the morphology of fracture surfaces and influence of loading conditions will be given in the paragraphs below.
Fracture surfaces of the broken bolts are shown, in a collective manner, in Fig. 9. In this figure fatigue and final fracture areas are
separated by dashed yellow curves and crack propagation directions are indicated by red arrows for convenience. In regards to the
actual (in situ) crack propagation direction(s) it is naturally assumed that all cracks have propagated in the same general direction
dictated by the main overall load with some minor deviations due to the respective radial positions of individual bolts.
Two main observations can be seen on these fracture surfaces:

i- A pattern of increasing fatigue area, at the expense of final fracture area, is seen to take place as bolt location traverses along the
flange circumference in the counter clockwise direction, Fig. 3, starting at bolt No. 1 and reaching a maximum at bolt No. 5 (the 6
o’clock position). The pattern is then reversed where the fatigue area is seen to decrease as bolt location moves from bolt No. 5 to
bolt No. 8. This pattern of change in the final fracture area indicates that bolt No. 5, located at the 6 o’clock position, had expe­
rienced the lowest overall stress level and also that this overall stress level was increasing as bolt location moved away from the 6
o’clock position in either direction; reaching a maximum at bolt No. 1 close to the 8 o’clock position. This rather systematic pattern

8
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 9. Fracture surfaces of broken bolts No. 1 to 8, (a) to (h), respectively; crack propagation directions are indicated by the red arrows.

9
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 10. Example of contact patterns between bolt head and flange/washer surface; (a) focus at fracture surface showing fatigue crack initiation site
and (b) focus at bolt head showing preferential rubbing on the region aligned with fatigue crack initiation site.

of stress level variation would rule out this to be a result of “deliberate” variations in the preload applied to the bolts. A possible
cause, which is compatible with this load distribution pattern, could be the exposure of the bolts to varying degrees of bending
moment with the associated load eccentricity [13–15]. This possibility was initially checked and demonstrated by inspecting the
nature of contact pattern between the bolts and the flange/washer surface. Fig. 10 shows an example of a broken bolt with image
focus being alternated between the fracture surface and the bolt head from inside. It is clear, from Fig. 10, that the region of bolt
head aligned with crack initiation site had suffered increased amount of rubbing against the flange/washer surface compared to the
rest of bolt head, indicating the bolts to have been at an angle with the flange and not coaxially aligned with the bolt holes. At this
point it is logical to conclude that eccentricity was highest around the 8 o’clock position with decreasing level until reaching the 6
o’clock position and then assuming an increasing pattern again until reaching bolt No. 8. This point will be discussed further in
sections 4 and 5.
ii- Fatigue crack observed on fracture surface of bolt No. 1 is seen to exhibit a relatively flat front, while all other fatigue cracks
exhibited curved fronts with varying degrees of curvature towards final fracture. Moreover, the degree of curvature is seen to be
independent on the overall mean stress level; for example bolts No. 3 and 7 have comparable mean stress levels (indicated by their
relative final fracture areas) with fatigue crack on fracture surface No. 3 exhibiting significantly higher degree of front curvature.
This observation, at least initially, indicates that fatigue crack final front morphology has no relationship to the load level; or its
counterpart – crack length. This result is in disagreement with previously reported results, which indicated higher mean stresses,
and thus shorter fatigue cracks, to result in cracks with curved fronts having a crescent shape [14,43]. A clear example is the fatigue
crack seen on bolt No. 1 which is perfectly flat despite the fact that it had apparently experienced the highest mean stress level. A
closer look at fracture surfaces in Fig. 9, however, reveals that the morphology of fatigue crack front is a dynamic rather than a
static phenomenon; i.e., crack fronts are initially seen flat and gradually change to a curved shape upon traversing from crack
initiation site to the final fracture area. The present results suggest that fatigue crack front morphology is controlled by the
“instantaneous” stress field at the crack front, rather than the overall stress level alone, which in the specific case of bolts and
threaded fasteners, is also influenced by the helical geometry of the threads. In all the present cases crack initiation is seen to take
place at a region along thread root which is approximately normal to bolt axis, i.e. at a portion of thread root with small axial
misalignment. This is expected to result in a relatively flat initial crack front. Upon crack growth, however, and as crack-front end
points travel along thread helix, the degree of axial misalignment between crack ends increases. As the axial misalignment reaches a
level where a flat front cannot be maintained, then based on local stress levels, the crack may continue assuming one of two paths;
the crack may either maintain its flatness and cut through thread flank or continue travelling along the thread helix with some
curvature in the crack front. Present results indicate that the second scenario is more likely to take place as stress levels at thread
(helix) root are highest, thereby favouring crack propagation along thread helix with the introduction of crack front curvature. The
presence of the aforementioned axial misalignment is normally signified by the presence of a pair of ridges, which represent the
fractured ends of the thread, on fracture surfaces. Such ridges are indicated in Fig. 9-c. The present results indicate that the role of
increased axial misalignment (evidenced by the presence of crack ridges) is to slow crack propagation rate at the ends of crack front
compared to the middle region thereby resulting in the observed curvature. Accordingly, and supported by the patterns seen in
Fig. 9, the degree of curvature is seen to be dependent on “how early” the crack front encounters a ridge (large axial misalignment)
on its way towards final fracture. This pattern of curvature dependence on presence of crack ridge(s) is clearly seen for all fracture
surfaces in Fig. 9. For example fatigue crack on bolt No. 3 has encountered a ridge “much earlier” than all other cracks and so
appears to have the largest degree of curvature, while fatigue cracks on bolts No. 2 is seen to encounter a ridge relatively “late” and
is seen to exhibit much more flatter crack front. It is worth mentioning here that the crack front on bolt No. 1 (Fig. 9-a) remained flat
and did not change to a curved morphology as final fracture took place before the crack front became close to the ridge. The reason

10
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 11. Parts of a star pattern followed in bolting a 20-bolt flanged joint.

Fig. 12. Load distribution of broken bolts superimposed on load patterns of three bolting passes for 8 bolts in a row;
adapted from [44]

11
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 13. Bending observed on an A307-B bolt located at the 8 o’clock position.

why crack propagation rates slow down at crack front edges compared to the middle region, with increased axial misalignment,
may need further research efforts to be fully explained.

4. Proposed failure mechanism

As in any failure investigation, especially those involving the human factor, the exact case history may not be fully known and/or
easily concluded based on the available physical evidence alone. As such, one should utilize available physical evidence side by side
with accumulated experience, available literature; rationale and applicable analytical techniques available in order to conceive the
most probable sequence of events which had led to the observed failure. Assembly of such bolted flanges is normally carried out using
either a two-pass or a three-pass tightening procedure with a star (or a cross) pattern for transition between bolts as shown in Fig. 11.
For instance, if a three pass procedure is followed, bolts would be tightened to around 35% of their full clamping in each pass. As a
result of elastic interactions between individual stress fields of the bolts, each of these bolting passes would normally have a distinct
“signature” load pattern [44].
Fig. 12 shows the loading pattern expected for the eight broken bolts under consideration, based on the previous discussion,
superimposed on typical signature load patterns for eight bolts in-a-row during a three-pass tightening procedure [44]. From Fig. 12 it
is obvious that the load pattern for the broken bolts does not match any of the common signature load patterns. This implies that (i) this
load pattern was not a result of the common star (or cross) transition pattern and (ii) the load distribution among the broken bolts was
not the result of elastic interactions of individual stress fields of the bolts. Keeping in mind the load eccentricity demonstrated in
section 3.4, then this load pattern most probably would be the result of the flanges exerting unequal load shares on these eight bolt due
to the flanges being assembled in a non parallel configuration. During the assembly of heavy rigid flanged joints there are several
scenarios which may lead to unparallel flange surfaces including uneven gasket compression, flange/joint rocking and deformed flange
surfaces [45,46].
Based on the above discussion, the following steps summarize the most probable sequence of events that led to the present failure:

(i) In order to avoid the rocking motion of this rather heavy joint (individual flange diameter and thickness of 700 and 25.4 mm,
respectively), and especially as the interface between the two flanges is sloping rather than being perfectly horizontal or
vertical, the technicians who assembled this meter decided to first “pin” the two flanges together with the bolt at the 12 o’clock
position without any tightening.
(ii) The technicians carried on by fitting Grade-5 bolts at and around the 6 o’clock position on the flange with a “snugging”
tightening (probably with the good intention that proper tightening procedure would be resumed once the flanges are secured
in place). The decision to place the bolts at these locations first and not to follow the usual star pattern might have been
encouraged by the easy access to this area, Fig. 3. This procedure, however, is known to result in the flanges open up at the 12
o’clock position; a phenomenon known as “bird-mouthing” [45] where the flanges open at one side resemble the shape of an
open bird mouth.
(iii) Apparently as bolts’ positions moved away from the 6 o’clock position, the technicians were faced with increasing difficulty in
securing bolts in place as, with the present angular opening, hole positions on one flange are expected to shift from matching
hole positions on the other flange. A common procedure for mild shifting is the use of “hammer” tools to guide bolts into place.
Present results, however, revealed that in the present case hole-shifting was too large that the technicians resorted to the use of
“softer bolts”. The amount of both angular opening and hole-shifting may be estimated from the bending observed on the softer
A307-B bolts. Fig. 13 shows a bending of 5.5◦ (180◦ − 174.5◦ ) on the bolt located at the 8 o’clock position. This is about double
the maximum of possible bending within an individual flange with a hole diameter of 20.2 mm and a flange thickness of 25 mm
(bending angle = tan-1((20.2–19 mm)/25 mm) = tan-1 (0.048) = 2.75◦ ). This indicates that the bolt located at this position had
experienced maximum possible hole-shift of 2.4 mm (2 × 1.2 mm within each flange). Given the present flange dimensions,
simple rotation and trigonometric calculations can show that a 1.2 mm hole-shift requires a 27 mm vertical transition, i.e., 2 mm
higher than the flange thickness, which resembles half the vertical opening between the two flanges at this location with an
effective angle of about 0.7◦ . The use of softer bolts at the 8 and 4 o’clock bolt positions might have seemed to be a practical (but
not a smart) decision to help, at least partially, close the gap between the two flanges and reduce further hole-shifting.

12
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 14. Overall flow meter assembly; (a) discretized mesh, (b) solid model, (c) a cutaway section showing meter from inside with the gap between
flanges and (d) enlarged section at the 6 o’clock position.

(iv) Despite the apparent flange gap and the associated hole-shifting the technicians decided - again – to go ahead with the assembly
and “force” the flanges to close by tightening all the other remaining bolts between the 8 and 4 o’clock positions. Probably this
was done in order to avoid restarting the assembly over (probably and over again). In fact most of the force placed on the bolts
during this stage would have been used up in bringing the flanges to close, through elastic deformation of bolts and flanges, and
not as a tightening or clamping force [44,46]. The presence of an elastic O-ring would have compensated for this lack of enough
clamping force, especially at the relatively low working pressures involved.
(v) The bolting procedure explained above is believed to have resulted in the expected load pattern on the eight broken bolts but
this needs to be explored further.
(vi) Based on the present results it is unlikely that any extra tightening was applied to the broken bolts, as otherwise their individual
stress fields would have interacted and changed the observed load pattern.
(vii) It is believed that the stress levels inflicted on the eight bolts were high enough to cause premature fatigue failure under the
cyclic loading (pressure variations) reported for the meter. The details of this, however, need to be clarified and confirmed
which will be the subject of the next section.
(viii) Based on the sequence of events proposed above, it is believed that bolts at and around at the 6 o’clock position had failed in
succession with the highly loaded bolt(s) failing first and so on. The leak on the meter was not, however, observed before all of
the 8 bolts had fractured. This would be the result of the fact that as long as there was one intact bolt, that would have been
enough to hold the flanges down together. The failure of the last bolt (most probably near the 6 o’clock position) would, in this
case, be “the straw that broke the camel’s back” and resulted in flange opening up and the observed leakage on the meter.
Normally, under constant overall load and uniform load distribution between bolts, the first bolt(s) to fail would exhibit the
largest fatigue-to-final-fracture zone ratio. This is due to the fact that the load born by this (these) bolt(s) would be redistributed
on the remaining (smaller number of) bolts thereby raising the individual mean stress levels and causing failure at lower number
of cycles manifested through increased final fracture areas compared to the first bolt(s) to brake. This trend of increasing final
fracture area relative to the fatigue would continue until all bolts have fractured. In the present case, however, the two con­
ditions above are not present as, first, the overall load is not constant and is governed by the amount of elastic deformation of the

13
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 15. Meter assembly model after flange closing; (a) general view indicating displacement; (b) non-complete flange closure and (c) hole-shifting
between the upper and lower flanges.

flange which would be gradually relived upon successive fracture of individual bolts due to elastic recovery of the flange.
Second, the load distribution was found to be extremely uneven amongst individual bolts, and to be dependent on bolt position,
with bolts close to the 6 o’clock position seeing the lowest load levels. This would have caused the observed relational pattern
between fatigue-to-final-fracture area ratios and failure sequence.
(ix) At this point the flanges and all the other non-broken bolts would have gone through appreciable amount of elastic recovery
with almost complete loss of clamping force, which explains the fact that all the non-broken bolts were found “finger-tight”
upon inspection after the failure.

The proposed sequence of actions and the associated failure mechanism need, however, to be verified and confirmed. This is
discussed in the next section (FEM).

5. Finite element modelling (FEM) of the problem

In order to confirm/testify the above assumptions together with the proposed failure mechanism the current problem was modelled
in two stages; a first stage for the modelling of the overall assembly and a second stage for modelling loading patterns within individual
bolts.

14
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Table 3
Details of bolts loading and effective angle of eccentricity.
Bolt Number 1 2 3 4 5 6 7 8

Hole-shift (mm) 2 1.45 0.82 0.45 0.252 0.45 0.82 1.45

Bend angle (◦ ) 4.5 2.1 0 0 0 0 2.1 2.1
Main flange angle (◦ ) 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7
Total effective contact angle (◦ ) 5.2 2.8 0.7 0.7 0.7 0.7 0.7 2.8
Load (kN) 175 90.33 48.073 15.835 1.598 8.297 27.731 85.702

Fig. 16. Problem geometry; (a) full problem geometry and (b) final model geometry.

5.1. Overall meter assembly model

The overall assembly of the flow meter was carried out in order to estimate the effect of flange face tilting on load patterns within
the assembly. This was done using SOLIDWORKS® 2020 software with the automatic meshing option, Fig. 14. Plain carbon steel with
yield strength and elastic modulus of 220 MPa and 200 GPa, respectively, was assumed for the material. Meter dimensions were given
by the manufacturer. In this model, the lower half of the assembly was kept fixed while movement was confined to the upper half
thereby resembling the actual case. As the software requires non-zero bolt pretention, a minor load of 150 N was applied to each bolt
resulting in near zero (~0.5 MPa) pretention. This was done in order to test the effects of the bolting procedure under the safest
conditions of no pretention.
At the first sight, Fig. 14 shows that the bolt thread runout coincides with the interface between the two flanges. This is a major
design weakness and is known to promote failure at this location especially if load eccentricity is present [2,21,22]. This observation
explains, at least partially, the reason why all the broken bolts have failed at thread runout in the present case.
Fig. 15 shows different views of the meter assembly after the two flanges were forced to close using a distributed load. A maximum
displacement of about 10 mm is seen around the 12 o’clock position with decreasing magnitude away from this position until reaching
nearly zero mm towards the 6 o’clock position, Fig. 15-a. Fig. 15-b indicates, however, that this closure at the 12 o’clock position
resulted in the upper flange bowing rather than forming a flat interface with the lower flange. This bowing action is expected to result
in a twofold influence; (i) persistence of the hole-shifting between the two flanges and (ii) non-even distribution of load on the bolts
around the 6 o’clock position with higher loads being inflected on those bolts away from the 6 o’clock position. The former effect is
evident in Fig. 15-c where a shift in bolt holes between the two flanges is seen at the 12 o’clock position. This would have a role in
causing the observed bending on the bolt located at that location, Fig. 4. It is interesting though that the amount of bending observed
on this bolt (around 2.5◦ ) is less than that observed on bolts located at the 4 and 8 o’clock positions (around 5◦ ) thereby reflecting the
effect of flange closure in reducing the amount of hole-shifting.
Relative load distribution on the bolts was obtained directly from the software which, in SOLIDWORKS®, is normally given as
“reactions on the connectors”. The results are shown in Table 3. The values in Table 3 also contain the effective angle of contact
between the upper flange and the bolts. This angle is the summation of the flange opening angle (0.7◦ ) and bolt bending angle.
Respective bolt bending angles were determined by calculating the amount of hole-shifting at the particular bolt location which itself is
dependent on the flange opening angle. From Table 3 it can be seen that both bolt-shifting and bolt bending angle increase as bolt
location departs away from the 6 o’clock position. It is also interesting to note that bolts number 3 to 7 did not suffer any bending. This
comes from the fact that their bolt-shifting was less than 1.2 mm which is the clearance between the bolt and the bolt hole in the flange
and, as such, the overall effective angle on these bolts would be the same as that of the flange opening angle. Two observations
regarding the load patterns in Table 3 need to be clarified. First the uneven load distribution between similar bolt positions on either

15
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 17. Modeling of flange-bolt contact, (a) Top view and (b) oblique view.

Fig. 18. 3-dimentional bolt modeling, (a) SOLID187-type element and (b) complete body mesh.

side of the 6 o’clock position e.g. bolts No. 2 vs. No.8, No. 3 vs. No.7 and No. 4 vs. No.6. This is mainly due to the presence of unequal
number of bolts on either side which is likely to have considerable effects on the elastic interactions and the restraint on the flange.
Second, although the load variation pattern agrees with that predicted in Fig. 12, relative amounts do not seem to be in agreement with
those predicted from the final fracture area. For instance the load on bolt No. 5 was predicted to be around 20% of that on bolt No. 1
while the values in Table 3 indicate an insignificant load share of only around 0.9%. While the values in Table 3 indicate load dis­
tribution at the onset of flange closure, they do not necessarily reflect actual load levels around bolt fracture times. From the data in
Table 3 it is predicted that bolt No. 1 is likely to fail first at which instance a small share of its load would be relieved by flange elastic
recovery while the rest of this load share is expected to be carried over to the rest of the bolts with bolts No. 2 and No. 8 receiving the
highest load shares and becoming both the first line of defense and the next candidates for failure. This cycle of bolt failure, relaxation
and load shedding continues until the last bolt (most probably No. 5) fails thereby resulting in complete elastic recovery of the flange.
The implication of these loading patterns will be the subject of the next section.

5.2. Individual bolt loading

FEM of individual bolt loading was done using ANSYS® software. The model was based on the results obtained so far, i.e., load
shares and effective contact angles between bolts and the upper flange. It is worth mentioning that modeling considered only the upper
portion of the bolt covering the thread runout (represented by a round groove) and bolt shank. The decision to include this section only
was based on the fact that all bolts have failed within this region and on the bolt-flange setup seen in the overall assembly model,
Fig. 14-b. Furthermore, previous studies have indicated that thread modeling by a round groove provides similar results as in the case

16
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 19. Pairing between target and deformable surface elements.

Fig. 20. Loading conditions and the constrains applied in the present problem.

of using a helical thread [8].

5.2.1. FE model details

i- Problem geometry
The current problem geometry is illustrated in Fig. 16-a where the flange is replaced by an inclined plate. The loading conditions
and the associated boundary conditions are symmetrical with respect to the y-direction (horizontal), therefore, the problem may be
reduced to the lower half of the domain as shown in Fig. 16-b.
For contact modeling purposes, the flange has been replaced by a contact surface the slope of which is governed by the particular
given inclination angle, Fig. 17.
ii- Types of elements used
Four element types were utilized in the present problem as follows:

17
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 21. FEM results for bolt No. 1; (a) longitudinal section and (b) across the root of thread runout.

a- Low-stiffness spring element: One spring element with very low stiffness (10 N/m) is employed to prevent the rigid body motion of the
bolt by initial balancing of the applied tensile force.
b- Three-Dimensional Structure Element. SOLID187-type element is used to model the main bolt body. SOLID187 is a quadratic 3-dimen­
sional element which is well suited to model irregular mesh geometries. The element is defined by 10 nodes (I through R) having
three degrees of freedom at each node, i.e., translations in the nodal ×, y, and z directions, Fig. 18-a. A mesh of a 20,422 elements
was used to model the main bolt body with individual element size of 1.25 mm, Fig. 18.-b
c- Target and contact elements: TARGE170-type element was used to model the flange “target” surface. The target is treated as a rigid
surface; while a CONTA174-type element (8-Node Surface-to-Surface contact) is used to represent contact and sliding between 3-D
target surfaces (TARGE170) and a deformable surface. The contact conditions are met when the target surfaces and the contact
body surfaces are paired. Typical pairing used in this study is shown in Fig. 19.

iii- Applied loading and boundary conditions

The applied loading and the constraints are shown in Fig. 20. The target (flange) is a fully constrained surface. The lower surface of
the bolt is constrained in x-y plane and subjected to a given tensile stress to simulate the tensile force. In terms of the current problem,
these conditions represent the constraints on the bolts inflected by the lower flange and the tensile forces resulting from upper flange
closure, respectively.
iv- Material properties
The bolt is modeled as an elastic linearly work hardening material such that E = 200 GPa, υ = 0.3, work-hardening rate is 2.8 GPa
and a yield stress of 600 MPa. These properties are typical of Grade-5 bolts.

5.2.2. FEM results

Fig. 21 shows the resulting stress field for bolt No. 1. Fig. 21 shows that even under near-zero pretention, stress levels beyond the
yield stress have been generated within this bolt as a result of the followed bolting procedure (mean stress of about 620 MPa in the
main shank body and around 800 MPa in the reduced section). This is known to promote accelerated fatigue failure under fluctuating
load conditions [32]. Fig. 21-b shows stress distribution across a plane containing groove (runout) root indicating that load eccentricity
has resulted in an uneven stress distribution about the bolt axis. Fig. 21-b also shows a transition from a maximum tensile stresses of
around 1700 MPa at the root of runout region located at the bolt side in contact with the flange to a maximum compressive stress of
around 1500 MPa located at the opposite side of thread runout. Present results indicate that bolt No. 1 is likely to suffer accelerated
fatigue failure at thread runout region [2,21,32,43]. This comes from the fact that fatigue failure is dependent on both the maximum
stress amplitude and the stress range [43]. In the present case the stress range is limited by pressure fluctuations experienced by the
meter assembly. The maximum stress range for individual bolts was calculated under the normal working pressure variations (section
1.2) and found to be only about 7 MPa while that under the reported occasional pressure spikes was found to be about 106 MPa.
Present results show that the deciding factor in the present case was the maximum stress amplitude and not the stress range. Even
under the maximum expected stress range the overall stress would still be below the upper allowable limit of 550 MPa if the stress in
this bolt was maintained at the recommended pretention stress value set by the manufacturer and common practice of, i.e., 420 MPa.
Results for the eight bolts are presented collectively in Fig. 22. The same trend is seen for the rest of the bolts as that of bolt No. 1 in
terms of maximum stress location being at the root of thread runout except that lower stress levels are seen thereby reflecting the
effects of lower loads. It is noted also that stress transition across the sides of bolt section became flatter with lower overall loads, i.e.,
smaller differences between maximum (tensile) and minimum (compressive) stresses across the runout region. As discussed earlier
these stress fields are considered to be “a snapshot” at the moment of flange closure but would be subject to continuous changes

18
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 22. FEM results for individual bolts, (a) to (h) bolt No. 1 to 8, respectively.

19
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Table 4
Stress concentration factor for the different angles and mean stress levels.
Bolt number 1 2 3 4 5 6 7 8

Total effective angle (◦ ) 5.2 2.8 0.7 0.7 0.7 0.7 0.7 2.8
Mean stress (MPa) 816 422 225 74 7.5 38.7 129 400
Max. Stress (MPa) 1730 1210 980 640 73.4 365 820 1190
Stress Concentration Factor (Kt) 2.1 2.9 4.4 8.7 9.9 9.4 6.3 3

12
Total effective angle (°)

10
Stress Concentartion
Factor
8

0
1 2 3 4 5 6 7 8
Bolt Number
Fig. 23. Dependence of stress concentration factor on load eccentricity.

especially upon successive failure of the respective bolts. Furthermore, these stress levels are the resultant of near-zero initial pre­
tention. If higher pretention loads were applied to the bolts (which might well be the case) then much higher stresses are likely to be
generated in the bolts, thereby promoting much more accelerated failure of the bolts.
A final interesting observation is that stress concentration factor (Kt), defined as the ratio between maximum stress, at the root of
thread runout, and the mean stress within the reduced section, Table 4, is seen to have a strong inverse correlation with the contact
angle and, to a much lesser degree with the overall load level. These relationships are clearly seen in Fig. 23.

6. Conclusions and recommendations

6.1. Conclusions

A detailed failure analysis was carried out in the present study for a set of eight high-strength steel bolts originally fitted on a heavy
fuel oil meter. Experimental results indicated failure to have taken place by fatigue fracture process with all bolts failing at thread
runout. The failure at thread runout is believed to be the combined result of a major joint design weakness where this region was found
coinciding with the interface between the two flanges and the loading patterns on the bolts. Fracture surfaces of the bolts have
exhibited a systematic variation in final fracture area which was related to the loading shares to which the particular bolts had been
subjected. Experimental evidence has further shown that load eccentricity, evidenced by the contact pattern between bolt heads and
the upper flange, had played a major role in determining the aforementioned loading trend on bolts. The resulting eccentricity itself is
believed to be the result of neglectful bolting procedure which might have been coupled with other sources of eccentricity such as
deformed flange surface (through improper surface grinding) and/or uneven gasket compression. The effects of load eccentricity were
evaluated by a two-stage EEM procedure where minor flange tilting (as small as 0.7◦ ) at near zero bolt pretention were shown to result
in significantly high stress levels – approaching or even exceeding the material’s yield strength in some bolts. FEM results not only were
in agreement with experimental evidence in regards to failure location at thread runout, but also in regards to the expected sequence of
bolts’ failure evidenced by the respective load shares for individual bolts. FEM results have further indicated that stress concentration
factor at the root of thread runout has a strong inverse correlation with the amount of eccentricity (inclination angle) and a weaker
inverse correlation with the mean stress. From another angle, present experimental evidence indicated that, contrary to published
results, the degree of fatigue crack front curvature is related to the presence of axial displacement between crack-front ends and not to
the mean stress level.

20
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

Fig. 24. Modified joint design.

6.2. Recommendations

In order to avoid recurrence of such failures in the future, the following measures may be recommended:

i- Bolting sequence and procedure as specified by the manufacturer should be followed. This would include bolt type and size,
bolting sequence, number of tightening cycles and torquing level in each cycle as well as observing maximum allowable hole-
shift and flange opening. It has been repeatedly stated that meticulous design and manufacturing efforts can be easily spoiled if
“the man with the wrench” decides to do so [1,47].
ii- Modifying the present design either by slight reduction in the flange thickness and/or using a different bolt length so that thread
runout location does not coincide with flange interface, Fig. 24. This is expected to greatly reduce local stress levels at thread
runout and prevent failure at such location. By doing so this measure is also expected to result in much longer bolt fatigue lives.
iii- The use of spring washers and/or vibration resistant bolts [45] maybe considered as this may help filter out small degrees of
misalignment and/or vibration.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgment

The experimental part of this study was carried out at the laboratories of ACUREN Group INC., Mississauga, Ontario, Canada, where
the first author held the position of a senior metallurgical failure analyst. The help and support of ACUREN technical and managerial
staff is highly appreciated. The authors would like to express their deep appreciation to Mr. R. Abu-Rshaid for his distinct efforts in
preparing the SOLIDWORKS geometrical model for the meter assembly.

References

[1] M.J. O’Brien, R.G. Metcalfe, High strength engineering fasteners: design for fatigue resistance, J. Fail. Anal. andPreven. 9 (2009) 171–181, https://doi.org/
10.1007/s11668-009-9213-6.
[2] A.R. Shahani, I. Shakeri, Experimental evaluation of the effect of preload on the fatigue life of bolts, Int. J. Steel Struct. 15 (3) (2015) 693–701, https://doi.org/
10.1007/s13296-015-9015-5.
[3] D.J. Benace, Technical brief: Avoiding bolt failures, J. Fail. Anal. Preven. 7 (2007) 79–80, https://doi.org/10.1007/s11668-007-9015-7.
[4] F.R. Hutchings, P.M. Unterweiser, Failure analysis: The British engine technical reports, American Society for Metals, 1981.
[5] M. Zamanzadeh, E. Larkin, R. Mirshams, Fatigue failure analysis case studies, J. Fail. Anal. Preven. 15 (2015).
[6] Z.-W. Yu, X.-L. Xu, Fracture failure of bolts assembled in tuning- Wheel of a marine diesel engine, J. Fail. Anal. Preven. 16 (2016).
[7] P. Palai, P. Sinha, T. Venugopalan, Analysis of slide gate mounting bolt failure in steel ladle due to fatigue, J. Fail. Anal. Preven. 14 (2014).
[8] S.-S. Cho, H. Chang, K.W. Lee, Dependence of fatigue limit of high-tension bolts on mean stress and ultimate tensile strength, Int. J. Automot. Technol. 10 (4)
(2009) 475–479, https://doi.org/10.1007/s12239-009-0054-9.
[9] F. Casanova, C. Mantilla, Fatigue failure of the bolts connecting a francis turbine with the shaft, Eng. Fail. Anal. 90 (2018) 1–13, https://doi.org/10.1016/j.
engfailanal.2018.03.015.
[10] G.H. Majzoobi, G.H. Farrahi, N. Habibi, Experimental evaluation of the effect of thread pitch on fatigue life of bolts, Int. J. Fatigue 27 (2005) 189–196, https://
doi.org/10.1016/j.ijfatigue.2004.06.011.

21
A.K. Abdul Jawwad et al. Engineering Failure Analysis 122 (2021) 105279

[11] D. Gao, W. Yao, T. Wu, Failure analysis on the axial-connected bolts of the thin-walled cylinder under random vibration loading, Eng. Fail. Anal. 105 (2019)
756–765, https://doi.org/10.1016/j.engfailanal.2019.06.043.
[12] P. Honarmandi, J.W. Zu, Elasto-plastic fatigue life improvement of bolted joints and introducing FBI method, Mech. Based Des. Struc. 33 (3–4) (2005) 311–330,
https://doi.org/10.1080/15367730500374381.
[13] S. Novoselac, D. Kozan, T. Ergic, D. Damjanovic, Fatigue damage assessment of bolted joint under different preload forces and variable amplitude eccentric
forces for high reliability, In: G. Pluvinage, L. Milovic (Eds.), Fracture at all Scales. Lecture Notes in Mechanical Engineering. Springer, Cham, 2017, 239-268.
https://doi.org/10.1007/978-3-319-32634-4_13.
[14] J.W. Hobbs, R.L. Burguete, P.F. Heyes, E.A. Patterson, The effect of eccentric loading on the fatigue performance of high-tensile Bolts, Int. J. Fatigue 22 (6)
(2000) 531–538, https://doi.org/10.1016/S0142-1123(00)00004-9.
[15] S. Lochana, A. Mehmanparasta, J. Wintle, A review of fatigue performance of bolted connections in offshore wind turbines, Procedia Structural Integrity 17
(2019) 276–283, https://doi.org/10.1016/j.prostr.2019.08.037.
[16] B. Eccles, Fatigue failure of bolts. https://www.boltscience.com/pages/fatigue-failure-of-bolts.pdf, 2004.
[17] J.G. Ritchie, Fatigue of fasteners and studs, in: The Failure of Metals by Fatigue, Proceedings of a Symposium, pp. 260–289, University of Melbourne, December
2–6, 1946.
[18] S.D. Antolovich, A. Saxena, Fatigue failures, in: ASM Handbook, ninth ed., vol. 11, Failure Analysis and Prevention, ASM International, 1986, pp. 113.
[19] R.S. Chariton, Threaded fasteners: Part 1 – failure modes and design criteria of connections, in: Corrosion,pp. 13-17, March, Houston, Texas, 2011.
[20] W.P. Saunders, Turbine Steam Path Mechanical Design and Manufacture, vol, PennWell Books, IIIb, 2004.
[21] R.G. Baggery, Evaluating the effects of overtorque in bolts, J. Fail. Anal. andPreven. 1 (2001) 41–46, https://doi.org/10.1007/s11668-006-5012-5.
[22] J.D. Chalupnik, Stress concentrations in fastener-thread roots, Exp. Mech. 8 (1968) 398–404, https://doi.org/10.1007/BF02326000.
[23] T.F. Lehnhoff, B.A. Bunyard, Fastener thread and head fillet stress concentration factors, J. Press. Vessel Technol. 122 (2) (2000) 180–185, https://doi.org/
10.1115/1.556168.
[24] Climax Molybdenum Company, Three Keys to Satisfaction, New York, 1960.
[25] A.I. Yakushev, Effect of Manufacturing Technology and Basic Thread Parameters on the Strength of Threaded Connexions, Pergamon Press, Oxford, 1964.
[26] D.J. Wulpi, Understanding How Components Fail, American Society for Metals, Metals Park, 1985.
[27] H. Reemsnyder, Fatigue of mechanically fastened joints, in: ASM Handbook, tenth ed., vol. 19, Fatigue and Fracture , ASM International, 1996, pp. 287–294.
[28] J.H. Bickford, An Introduction to the Design andBehavior of Bolted Joints, third ed., Marcel Decker, New York, 1995.
[29] P. Schaumann, R. Eichstädt, Fatigue strength of preloaded hot-dip galvanized bolt assemblies with very large diameters, in: A. Zingoni (Ed.), Insights and
Innovations in Structural Engineering, CRC Press, Florida, Mechanics and Computation, 2016, pp. 578–583.
[30] A.R. Shahani, I. Shakeri, Experimental evaluation of the effect of preload on the fatigue life of bolts, Int. J. of Steel Struct. 15 (2015) 693–701, https://doi.org/
10.1007/s13296-015-9015-5.
[31] Y. Choi, B. Lim, Fatigue life prediction of high strength bolt subjecting to torque, Key Eng. Mater. 345 (346) (2007) 299–302, https://doi.org/10.4028/www.
scientific.net/KEM.345-346.299.
[32] J.M. Minguez, J. Vogwell, Theoretical analysis of preloaded bolted joints subjected to cyclic loading, Int. J. Mech. Eng. Educ. 33 (4) (2005) 349–357, https://
doi.org/10.7227/IJMEE.33.4.5.
[33] P.W. Wallace, Increasing joint performance by tightening bolts to yield, in:J. Bickford, S. Nassar (Eds.), Handbook of Bolts and Bolted Joints, New York, 1998,
pp. 591− 602.
[34] J. Hobbs, R.L. Burguete, P.F. Heyes, E.A. Patterson, The effect ofeccentric loading on the fatigue performance of high tensile bolts, Int. J. Fatigue 22 (6) (2000).
[35] J. Hobbs, R. Burguete, P. Heyes, E. Patterson, A photoelastic analysis of crescent-shaped cracks in bolts, J. Strain Anal. Eng. 36 (1) (2001) 93–100, https://doi.
org/10.1243/0309324011512649.
[36] M. Nakagome, K. Yasuda, M. Mizuno, Fatigue strength of boltedconnections under conditions of eccentric clamping and eccentricloading, Jpn. Soc. Precis. Eng.
52 (6) (1986) 1050–1054, https://doi.org/10.2493/jjspe.52.1050.
[37] R. Burguete, E. Patterson, Photoelastic stress freezing of bolts in bending using real parameters, in: Conference on Modern Practice in Stress Vibration Analysis,
pp 403-413, University of Sheffield, 1993.
[38] S. Grizaa, M.E. da Silva, S.V. Santos, E. Pizzio, T.R. Strohaecker, The effect of bolt length in the fatigue strength of M24x3 bolt studs, Eng. Fail. Anal. 34 (2013)
397–406, https://doi.org/10.1016/j.engfailanal.2013.09.010.
[39] H.H. Grover, Estimation of fatigue life of welded, riveted, and bolted structures, in: G. Sines, J.L. Waisman (Eds.), Metal Fatigue, University of California
Engineering Extension Series, McGraw-Hill, 1959, pp. 307–324.
[40] E. Dragoni, Effect of thread pitch on the fatigue strength of steel bolts, Proc. Inst. Mech. Eng., Part C 211 (8) (1997).
[41] A.L. Marcelo, A.Y. Uehara, R.M. Utiyama, I. Ferreira, Fatigue properties of high strength bolts, Procedia Eng. 10 (2011) 1297–1302, https://doi.org/10.1016/j.
proeng.2011.04.216.
[42] Fatigue and fracture. ASME handbook, vol. 19, Ohio, 1996.
[43] J. Toribio, B. González, J.-C. Matos, F.-J. Ayaso, Fatigue behavior of bolted joints, Met. Mater. Int. 18 (4) (2012) 553–558, https://doi.org/10.1007/s12540-
012-4001-3.
[44] G. Bibel, Tightening groups of fasteners in a structure and the resulting elastic interactions, in: J. Bickford, S. Nassar (Eds.), Handbook of Bolts and Bolted Joints,
CRC Press, New York, 1998, pp. 451–478.
[45] J.H. Bickford, M.E. Looram, Good bolting practices: A reference manual for Nuclear Power Plant Maintenance Personnel:, Volume 1, Large bolt manual,
California, 1987.
[46] J. Bickford, Behavior of a Bolted Joint during Assembly, in: J. Bickford, S. Nassar (Eds.), Handbook of Bolts and Bolted Joints, CRC Press, New York, 1998,
pp. 425–450.
[47] J.O. Almen, On the strength of highly stressed, dynamically loaded bolts and studs, SAE Transactions 52 (1944) 151–158.

22