Engineering Failure Analysis 46 (2014) 179–187

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

High cycle fatigue of welded structures: Design guidelines

validated by case studies
P.R. Fry ⇑
CMP Engineers Pty Ltd, P.O. Box 247, Spring Hill, Brisbane, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Experience with fatigue failures, of both welded and non-welded details, has demonstrated
Received 5 August 2014 that the design guidelines in many fatigue codes are inadequate for fabricated equipment
Accepted 27 August 2014 that requires a life exceeding 107 cycles. Stress ranges from strain gauge testing and finite
Available online 6 September 2014
element analysis of vibrating screens, combined with the actual life results, have been used
to establish design criteria applicable to machinery operating in the giga-cycle range. Case
Keywords: studies from recent investigations are used to illustrate the validity of the design criteria.
Giga-cycle
Ó 2014 Elsevier Ltd. All rights reserved.
Fatigue
Welded
Vibrating
Screen

1. Introduction

In the mining and minerals processing industries almost all the equipment is subjected to cyclic loading. This ranges from
vibrating screens, centrifuges, feeders and mills in processing plants to excavators, shovels, trucks, long wall chocks, con-
veyor pulleys and load-haul-dump vehicles in open pit and underground mining.
Vibrating screens, which are used a throughout the mineral processing industry for a variety of functions including sizing,
feeding and washing, operate at frequencies that result in a very high number of cycles during their required life. Typically
operating at about 1  106 cycles per day, a life of 109 cycles is reached in less than 3 years. As a result they provide an ideal
platform for evaluating fatigue design criteria.
Design guidelines in various international standards do not adequately address fatigue in the giga-cycle regime [1,2] and
in 2001 a fatigue design approach was proposed [3], this was refined in 2006 [4], on the basis of additional field data. Recent
investigations have provided further evidence that the proposed guidelines are valid, as illustrated by the two case studies
described.

2. Fatigue of vibrating screens

2.1. Loads, life and environmental requirements

Screen designs are very variable in terms of size, design details, drive mechanisms and functionality. However there are
elements that are common to all screen designs. Fig. 1 illustrates the typical construction of large screens. Two vertical side
plates are connected at the top by a drive beam or ‘‘bridge’’, on which are mounted exciters with rotating eccentric masses.

⇑ Tel.: +61 7 38392045.

E-mail address: ryan@cmpeng.com.au

http://dx.doi.org/10.1016/j.engfailanal.2014.08.010
1350-6307/Ó 2014 Elsevier Ltd. All rights reserved.
180 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187

At the bottom the side plates are connected by structural steel beams, which support the screening panels, in some cases two
decks of panels are used. Attached to the side plates are pedestals so that the whole screen can be mounted on springs. The
various connections and joints are made using bolts, welds or swage lock bolts.
Screens are designed to vibrate at approximately 45° to the horizontal so that material moves forward along the screening
deck. This motion is obtained by adjusting the eccentric masses on the exciters and results in constant amplitude oscillation
at the operating frequency.
The loads and resultant cyclic stresses are almost entirely due to dynamic acceleration of the machine mass, typically
around 5 g zero to peak. There is also some loading influence from the flow of material over the screen, particularly on
the longitudinal and cross members supporting the screen decks. However the stress range in normal operation is typically
very similar with and without material on the deck. The loading therefore approximates to a constant amplitude cyclic
stress. Although there are start-up and shutdown transients, the associated stresses are typically lower than during normal
operation, except at the spring support components and attachments.
Additional stresses can occur if there are natural frequencies of the structure close to the operating frequency. Individual
panels, beams or parts of the screen can resonate, or the entire screen can resonate in a particular mode. However the design
intent is obviously to avoid operating close to any natural frequencies of the screen. Overloading or impact with adjacent
structures can also increase the stresses.
Most plant operators expect to obtain several years life at least from the processing equipment, and consequently a life in
excess of 109 cycles is typically required. Screens are generally well protected from the corrosive mine water as well as the
abrasive materials used in the plants but, because mineral processing frequently requires wet screening and the protection
breaks down, data has been obtained and included for both dry and corrosive environments [4].

2.2. Vibrating screens failures

Mineral processing requirements dictate that these machines vibrate in the range of usually between 12.5 and 25 Hz, but
typically 14 to 17 Hz. Depending on the utilisation these frequencies translate into a very high number of cycles in a
relatively short time, Table 1. For example a vibrating screen operating at 16.5 Hz could exceed 1  107 cycles in only
200 h (less than 14 days) and 109 cycles in approximately 3 years. Consequently fatigue cracking is a common mode of
failure, Fig. 2.

2.3. Fatigue design standards

Fatigue failures on vibrating screens have for many years challenged the fatigue limit adopted in most international
standards for fatigue design [1–4]. For example, fatigue design codes such as BS 7608-1993 ‘‘Fatigue Design and Assessment
of Steel Structures’’ [5], AS 4100-1990 ‘‘Steel Structures’’ [6], ANSI/AWS D1.1-94 ‘‘Structural Welding Code – Steel’’ [7] and BS
EN 1993-1-9:2005 ‘‘Eurocode 3: Design of steel structures – Part 1–9: Fatigue’’ [8], in general do not provide adequate guid-
ance on the approach to adopt for very high cycle fatigue regimes. For example BS 7608 assumes that in clean air, or
with corrosion protection, there is a non-propagating stress range, So, corresponding to N = 1  107 cycles. Similarly BS EN
1993-1-9 and AS 4100 assume a constant stress range fatigue limit, DrD, at 5  106 cycles [6,8]. One standard that does
indicate a lower fatigue limit stress range is the American Bureau of Shipping ‘‘Fatigue Assessment of Offshore Structures’’
in which the S–N curves are extended to 1  108 cycles [9].
One of the possible reasons for this deficiency in the standards is that the time required to obtain sufficient statistically
relevant data would be prohibitive, since testing machines would be committed for many months on each test sample.
Consequently field experience provides practical and very useful results from which design criteria can be extracted.
There is significant evidence from failures that the fatigue limit, or non-propagating stress range, in vibrating screens is at
a much lower stress range than proposed in the various international standards [4]. An endurance limit, such as So at 1  107
cycles defined in BS 7608, is definitely not appropriate for the low stress, very high cycle applications encountered in mine

Fig. 1. Views of typical large screen structures.

 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187 181

Table 1
Typical cycles for a screen operating at 16.5 Hz and 80% utilisation.

Time of operation
hours Calendar equivalent at 24 h/day @ 80% utilisationa Cycles
24 1 day 1.1  106
168 7 days/week 8.0  106
7884 365 days/1 year 4.2  108
23,652 3 years 1.2  109
*
Note that some plants operate less than 24 h/day.

Fig. 2. Typical fatigue cracks in the side plate of a screen.

process plants, for both welded and non-welded details. Since screens are manufactured using standard fabricating
techniques it is also reasonable to assume the same failure criteria would apply to other equipment. This is confirmed by
failures on conveyor pulleys which experience approximately 2  107 cycles per year but are known to fail in fatigue after
several years in service [10].

3. Proposed fatigue design guideline

A guideline for the design of equipment subjected to stress cycles exceeding 1  107 cycles was presented by Fry, Klos and
Angus in 2006 [4]. This guideline was a refinement, based on additional field experience, of the results previously reported by
Fry and Klos in 2001 [3]. Recent investigations, discussed below, have further demonstrated that the proposed guideline is
valid for the design and analysis of vibrating equipment operating in the giga-cycle regime.
Data from numerous failures in a range of welded and non-welded details was normalised, arbitrarily, to a BS 7608 Class
D detail, and plotted for comparison with the Class D S–N curve [4]. The graph, Fig. 3, shows the recommendations of BS 7608
including: the fatigue limit at 1  107 cycles, the S–N lines for operating in sea water and the dual slope line for variable
amplitude loading. (The recommendation to reduce the life by a factor of two for operating in sea water is not included
in this diagram.)

Fig. 3. Proposed S–N design curves for high cycle fatigue of vibrating equipment. Note: Data plotted are normalised to Class D.
182 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187

Table 2
Endurance limit for alternative interpretation of S–N line.

BS 7608, 2.3% probability of failure

Weld So 107 Endurance limit for typical conditions on vibrating screens, Endurance limit for areas at bottom of screen that are exposed to
detail cycles Sr at 4  108 cycles, knee at 107, slope = m + 2 corrosion, Sr at 1  109 cycles, knee at 107, slope = m + 2
B 100 54 46
C 78 40 34
D 53 26 21
E 47 22 19
F 40 19 16
F2 35 17 14
G 29 14 12
W 25 12 10

The data confirms that the endurance limits at 1  107 cycles in BS 7608 and at 5  106 cycles in other international
standards are not valid for fabricated, corrosion protected, vibrating equipment operating in a mine process plant environment.
However continuing the S–N curves at the slope, m = 3, from the 1  107 point for ‘unprotected joints in sea water’ (and by infer-
ence other corrosive environments) is overly conservative for unprotected details in process plants. Based on the data the
proposed design curve is the dual slope line recommended by BS 7608 for variable amplitude loading in which the slope, m,
changes to m + 2 at 1  107 cycles. An endurance limit for corrosion protected details is strongly indicated to exist at the stress
range corresponding to a life of 4  108 cycles. In areas which are continuously wet and with poor corrosion protection an
endurance limit at 1  109 cycles is proposed, but without any reduction in life penalty at lower numbers of cycles. In cases
of severe corrosion this recommendation should be reviewed by the designer as a more conservative approach may be
appropriate.
Table 2, and Figs. 4 and 5, show the endurance limit based on the proposed design guidelines applicable to each weld class
for typical conditions on a screen (protected with paint, glued on rubber or epoxy wear resistance, operating in air and
intermittently wet) and for locations where corrosion protection is poor and the detail is continuously wet.

4. Screen failure case studies

4.1. Case study 1 – cross beam failure from cleat weld

The cross beams on two screens, 3.6  6.1 m and 3.6  7.3 m, operating at 990 rpm (16.5 Hz) at a 48° stroke angle, failed
after 4060 and 4338 h in operation, equivalent to 2.41  108 and 2.58  108 cycles, 211 and 226 days at 80% utilisation. The
failures initiated at angle cleats welded to the top flange of the SHS to which the longitudinal deck supports are attached,
Fig. 6. The top of the beams are protected with rubber and are therefore in a non-corrosive environment. The beams are
stress relieved after welding.

Fig. 4. Proposed S–N design curves for high cycle fatigue in non-corrosive or corrosion protected conditions.
 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187 183

Fig. 5. Proposed S–N design curves for high cycle fatigue in corrosive conditions.

Fig. 6. Fatigue failure of beam initiated at toe of weld at a cleat.

One of the cross beams was replaced and the other had been repaired by welding the crack and then adding doubler
plates over the cracked area, to extend the life until a replacement beam could be installed. The repair was done in-situ
and the beam was not stress relieved after the repair. When the screen was inspected after a further 11 days in service small
cracks had developed at a life equivalent to 1.25  107 cycles, Fig. 7.
The doubler plate weld detail was assessed using BS 7608. A crack at the toe of a lap joint is a Class F2 detail (Type 11.4)
but the notch effect where the weld runs off the plate would result in a lower fatigue life consequently the S–N curve for a
Class G (Type 8.7) has been used. The life between the repair and the development of the crack was used to derive the oper-
ating stress range, Fig. 8. The stress range in the centre of the beam was consequently estimated to be 35 MPa, and could be
less if a lower probability of failure was assumed.
Using the stroke angle, beam orientation and derived stress range in the middle of the beam, the stress range was calcu-
lated in the corner of the beam, where the original failure had initiated at the cleat. The value was modified for the stress
relief heat treatment according to BS 7608, resulting in an effective operating stress range of 34 MPa. The life that would
be expected for the new screen, with the original, profile blended and stress relieved weld detail can then be determined
using the S–N curve for a BS 7608 Class E detail. A life of 2.3  108 cycles was predicted which correlated well with the actual
life of the two beams that failed, assuming 80% utilisation, Fig. 9.
This investigation validated the use of the S–N curves previously proposed [4], and confirmed the fact that a fatigue limit
at 1  107 cycles cannot be applied to vibrating screens: the repair failed at over 1  107 cycles and the original detail failed
at over 1  108 cycles.

4.2. Case study 2 – cross beam failure from drilled hole

A SHS cross beam on a 4.2  8.5 m screen, which operates at 850 rpm (14.17 Hz), failed after approximately 150 days in
operation, equivalent to 1.5  108 cycles, assuming 80% utilisation. Cracking initiated at a 40 mm diameter drilled hole in the
lower face of the beam.
184 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187

Fig. 7. Fatigue failure at toe of repair doubler plate.

Fig. 8. Operating stress range derived from life of repair.

The crack was weld repaired and blended without stress relieving. The screen operated for a further 10 days at a reduced
stroke with the damaged beam, which was then removed and replaced. It was inspected and the weld repair had re-cracked,
Fig. 10, and an adjacent hole was also found to be cracked.
After the failure the stroke on the screen had been reduced from ±4.64 g to ±2.8 g to lower the stress range, and increase
the probability of the cross beam surviving until the next available opportunity to replace the beam.
During the 10 days after the repair, it operated at 100% utilisation equivalent to 1.22  107 cycles.
The particular weld used for the repair – a steel tube welded from one side and dressed by grinding is not explicitly clas-
sified in BS 7608. Assuming that the repair would be classified as a Class F2 weld detail, welded from one side with no back-
ing strip (Type 7.5) according to BS 7608 and then improved by one class by grinding, it would be appropriate to treat it as a
Class F weld. Then, using the number of cycles to failure the stress range at the hole would have been approximately 51 MPa,
depending on the probability of failure, Fig. 11. (In a failure analysis scenario the 50% curve provides an appropriate estimate
of the operating stress range.) Note that this stress range is for the screen operating at the reduced stroke of ±2.8 g.
For the condition where the screen is operating at its normal stroke the stress range would be increased by a factor of 1.66
(4.64/2.8 = 1.66) to 85 MPa. Since the hole is a non-welded detail the effective stress range is reduced by 0.8 (Clause 3.45), to
68 MPa. The life of the beam with a drilled hole, giving a class B detail (Type 1.3), would therefore be predicted to be between
about 1  108 cycles at 2.3% probability of failure (96 days) and 3  108 cycles for a 50% probability of failure (288 days). This
compares well with the actual time in service of approximately 150 days (1.5  108 cycles), Fig. 12.
 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187 185

Fig. 9. Life prediction estimated from operating stress range.

Fig. 10. Beam after removal from screen with repaired and re-cracked hole left and unrepaired hole at right also with a crack.
186 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187

Fig. 11. Operating stress range derived from life of repair.

Fig. 12. Life prediction estimated from operating stress range.

The manufacturer of the screen later confirmed that the stress levels derived from the failure of the repair were consistent
with the design analysis.
As for the first case study, this investigation validated the use of the S–N curves previously proposed [4], and confirmed
the fact that a fatigue limit at 1  107 cycles cannot be applied to vibrating screens: the repair failed at slightly over 1  107
cycles and the original hole detail failed at over 1  108 cycles.

5. Conclusions

Fatigue cracking failures frequently occur after months in operation at lives well in excess of 1  107 cycles. Consequently
it is concluded that there is no endurance limit, at 5  106 cycles [6,8] or at 1  107 cycles [5] in the typical vibrating screen
environment, even for details that are operating in air and are well protected against corrosion.
 P.R. Fry / Engineering Failure Analysis 46 (2014) 179–187 187

Recent case studies have provided further evidence that the fatigue limits recommended in some design standards are not
adequate at high cycles and that the design guidelines previously proposed [4] are indeed valid and should be used in design-
ing vibrating equipment.
The following approach for design is recommended: use the BS 7608 fatigue design method with the recommended
slopes, m, to 1  107 cycles and m + 2 from 1  107 to 4  108 cycles with an endurance limit at 4  108 cycles for all loca-
tions that are either dry or corrosion protected. For details that suffer from poor corrosion protection and are continually wet
the endurance limit should be reduced to the stress corresponding to 1  109 cycles.
For non-welded details, as recommended by BS 7608, the effective stress range is determined by taking 60% of that part of
the stress range that is compressive and 100% of that part of the stress range that is tensile.
These design guidelines are based on experience with vibrating screens and should be directly applicable to other fabri-
cated equipment, particularly in mineral processing plants. However the method has not been critically evaluated for other
types of equipment and machinery. Consequently engineers will need to use their own judgement in deciding whether the
recommendations described are appropriate for the application under consideration.

References

[1] Fry PR, Greenway ME. An approach to assessing structural integrity and fatigue failures in vibrating equipment. In: Proceedings of the second national
conference on fracture, University of the Witwatersrand, Johannesburg; 26–27 November 1984.
[2] Steyn J. Fatigue failure of deck support beams on a vibrating screen. Int J Pres Ves Piping 1995;61:315–27.
[3] Fry PR, Klos AJ. High cycle fatigue of welded structures: failures and design of vibrating equipment, engineering materials conference, Melbourne
Australia. The Institution of Materials, Australasia; 2001.
[4] Fry PR, Klos AJ, Angus RB. High cycle fatigue of welded structures: design guidelines for equipment operating in the giga-cycle regime. In: Engineering
Materials Conference, Springer, Berlin; 2006.
[5] BS 7608:1993 fatigue design and assessment of steel structures. British Standards Institution; 1993.
[6] AS 4100-1990: steel structures, standards Australia. Sydney Australia; 1990.
[7] ANSI/AWS D1.1:2010 structural welding code-steel. American Welding Society; 2010.
[8] BS EN 1993-1-9:2005: Eurocode 3: Design of steel structures – Part 1–9: Fatigue. British Standards Institution; 2005.
[9] American Bureau of shipping fatigue assessment of offshore structures; 2003.
[10] Wolf T. Application of weldment fatigue methods to conveyor pulley design bulk material handling by conveyor belt II conference proceedings; March
1998. ISBN-10: 0-87335-157-6.