Engineering Fracture Mechanics 180 (2017) 315–329

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

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

Failures of high strength stainless steel bridge roller bearings:

A review
P. Noury a,⇑, K. Eriksson b
a
Luleå University of Technology, Division of Structural and Fire Engineering, Luleå 97187, Sweden
b
Luleå University of Technology, Division of Solid Mechanics, Luleå 97187, Sweden

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

Article history: With the innovation of elastomeric bearings in the mid-1950s steel bearings lost their
Received 25 May 2017 interest and significance both in research and development and subsequently even in
Accepted 6 June 2017 application. Steel bearings were gradually abandoned in bridges, followed by the technical
Available online 11 June 2017
literature and design standards. However, a great number of steel bearings remain today in
service world-wide and will pose their particular challenges in the future. To the author’s
Keywords: knowledge, just in Sweden, high strength stainless steel bearings still exist in no less than
Bridge roller bearing
some 650 bridges. In recent years, a large number of such bearings have failed with an
Failure analysis
Finite element method
alarmingly high frequency in Sweden during a period of six to twenty years after installa-
Linear elastic fracture mechanics tion making them a serious maintenance cost issue.
After a brief summary of the history of high strength stainless steel bearings, the paper
reviews service experience of such bearings in Sweden and elsewhere. Accompanying
finite element analyses were performed in order to gain insight into the likely failure
mechanism. Finally, this comprehensive review leads to a conclusion that identifies the
causes of the failures occurred and makes some recommendations.
Although previous investigations of the stainless steel bearings have not been able to
clearly identify the cause(s) of the failures occurred, it is found that the failures primarily
occurred due to initiation of cracks through stress corrosion cracking followed by fatigue
crack growth requiring a certain stress range and a sufficiently large number of cycles until
final failure ensued through sudden and instable fracture after fatigue growth to a critical
crack size.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction

It is well-known that bridge bearings have occupied a major part of the maintenance budget of bridges. In theory, bear-
ings should last for the entire design life of a bridge. However in practice, replacement has often been found necessary after
around 30 years. Rollers made of high strength stainless steels may crack even after only a few years in service thus giving
rise to serious maintenance problems, often involving complicated jacking procedures in bearing replacement work and con-
comitant interruption of traffic. The cost for the replacement of bearings, which is in general higher than for the original
installation, may become excessive if the design does not envisage and facilitate possible replacement operations (through
accessibility, extra jacking areas, etc.) [1].

⇑ Corresponding author.
E-mail address: [email protected] (P. Noury).

http://dx.doi.org/10.1016/j.engfracmech.2017.06.004
0013-7944/Ó 2017 Elsevier Ltd. All rights reserved.
316 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Nomenclature

a crack depth
Cv charpy impact energy
D roller diameter
E modulus of elasticity
fu ultimate strength of material
Keff effective crack driving force
KI mode I linear elastic stress intensity factor
KII mode II linear elastic stress intensity factor
KIII mode III linear elastic stress intensity factor
KIc critical value of KI at onset of brittle fracture in Mode I
KIIc critical value of KII at onset of brittle fracture in Mode II
L effective length of roller
N’Rd design resistance per unit length
R radius of roller
RP0.2 0.2% proof strength
Rm tensile strength
a roller rotation in degrees
cm partial material safety factor
m Poisson’s ratio

Roller bearings are perhaps the simplest type of movable bearings. They have a long history and are a convenient option
for bridges or roof trusses [2]. Roller bearings in a bridge, besides transferring the superstructure loading to the supports,
permit longitudinal deformation of the superstructure due to temperature variations and traffic loading. Cast iron and struc-
tural steel were extensively used for manufacturing of roller bearings between 1850 and 1950. A comprehensive review of
the evolution of cast iron and steel bridge bearings in this period in the countries of Central Europe where German was spo-
ken is given by Wetzk [3].
Kreutz Company in Germany is well-known as the inventor of high strength stainless steel and surface hardened roller
bearings [4]. At the end of the 1950s, the Kreutz Company introduced the expensive chromium electroplating technique
to provide corrosion resistance and to increase surface hardness for bridge rollers. One of the first applications of such bear-
ings was the Norderelbe Bridge in Hamburg in 1958. The rollers were made of C45 (1.0721) or C15 (1.0401) chromium plated
steel. It was however found that corrosion and accumulation of corrosion products in the contact areas was often the cause of
failure of chromium plated rollers.
High local compressive stresses occur generally along the roller contact lines. To reduce roller diameter, the use of stain-
less steel roller bearings with a high surface hardness and low friction began in Germany and Switzerland in the 1970s.
X90CrMoV18 (1.4112), the earliest stainless steel grade, also introduced by the Kreutz Company, however often failed in
a brittle manner that was attributed to a high carbon content of 0.85–0.95%. This steel grade was therefore replaced by
X40Cr13 (1.4034), a quenched and tempered high strength stainless steel. The Kreutz stainless steel roller bearings made
of X40Cr13 quickly superseded the chrome plated bearings.
As business ran so well for Kreutz stainless steel bearings, the Maschinenfabrik Esslingen (ME) Company, considering the
high cost of stainless steel, tried to reduce the cost by using the mating surface method in which the rollers were made of
structural steel on which a stainless steel weld metal of X40Cr13 was deposited and then machined to the required dimen-
sions. The rollers were called Corroweld roller bearings. Thanks to the high hardness level of weld layers made from
X40Cr13, Corroweld rollers allowed a length of up to 12 times the roller diameter. These bearings behaved in more ductile
manner compared to the Kreutz stainless steel bearings due to the ductile bulk material.
Gutehoffnungshütte Company improved the Corroweld bearings. In their application rollers were made of steel X40CrI3
and stainless steel welds deposited only on the bearing plates of structural steel. However, both the original and the
improved Corroweld types of bearings were found to fail after only a few years in service. The failures were attributed to
incisions and bore holes at the end face of rollers, intended for guiding purposes Fig. 1(a,b).
In 1967, at the end of the roller bearing era, the Kreutz Company, realizing that an incision results in stress concentration,
modified the design of the stainless steel bearings. The incisions at the end face of the roller were abandoned and in later
design replaced by an appropriate rack and pinion gear system Fig. 1(c). The roller and the roller plates were made of steel
X40Cr13 and were double tempered in order to facilitate transformation of retained austenite to martensite.
In the late 1970 s and early 1980 s, failures of the modified Kreutz stainless steel roller bearings were first reported in
Germany.
 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 317

Fig. 1. Guiding mechanisms for roller bearings: (a) incision, (b) bolted plate and (c) rack and pinion gear system.

In 1976, failures of such bearings in the Öland Bridge in Sweden were reported [5]. The bearings were installed and loaded
just by the bridge superstructure in 1970. Later, in 1972, the bridge was opened to traffic. The bearings were inspected in
1974 and no failures were reported. In November 1976, some of the rollers were found failed.
In Switzerland in the late 1980s and early to mid-1990s, failures of many road, highway and railway high strength stain-
less steel bridge roller bearings were reported by Schindler and Morf [6].
In the UK in 2002, failures of high strength stainless steel roller bearings on the Thelwall viaduct, a major bridge over the
Manchester ship canal and the river Mersey near to Warrington, was reported by Edwards et al. [7]. The viaduct was orig-
inally erected between 1959 and 1963. In the early 1990s the viaduct was no longer considered adequate for the then traffic
flow. A new and separate viaduct was erected alongside the old one, the work being completed in 1996. After some 6 years in
service, a total of 52 out of 136 of the bearings were found to contain cracks.
In Turkey in 1994, the renovation and widening project of the Golden Horn Bridge in Istanbul started in order to reduce
traffic congestion by the construction of two new bridges and repair of the existing bridge, erected in 1974 [8]. A preliminary
inspection of the existing bearings made of stainless steel showed that one of the rollers had split in half along a longitudinal
plane and that another roller was cracked. All the bridge bearings were therefore replaced because they were all of the same
design and material.
Recently in Sweden, failure of stainless steel roller bearings has been reported in many bridges from the north, in the Skel-
lefteå Bridge, to the south, in the Öland Bridge [9–11]
In the following, the referred failures have been reviewed and a possible failure scenario is proposed.

2. Design aspects of roller bearings

2.1. Standards and codes

The most common national standards and codes for roller bearings are listed in Table 1.
The current standard in Europe for roller bearings is EN 1337-4. This standard specifies the requirements for the design
and manufacture of single and multiple roller bearings, in which the roller axis is horizontal. The AASHTO-LRFD is the cur-
rent standard in the US for bridge design and Section 14 of this standard contains requirements for the design of roller
bearings.

2.2. Design resistance

The design resistance of a roller bearing per unit length N’Rd as a function of roller length L, radius R and ultimate strength
fu, according to EN 1337-4, is given by
2
fu 1
N0Rd ¼ 23  R 
E c2m

Table 1
Standards for bridge roller bearings.

Standard/Code Region
EN 1337-4 Structural bearings-Roller bearings. Europe
DIN 4141 Lager im Bauwesen, Teil 1 bis 14. Germany
BS 5400 Steel, Concrete and Composite Bridges. United Kingdom
AASHTO-LRFD American Association of State Highway Officials, Bridge Design Specifications. United States
TRVK BRO 11 Trafikverkets tekniska krav bro, G.6.2 och G.6.3. Sweden
318 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Table 2
Bearings design resistance, allowable load and service loads.

Bridge (Region) Ea GPa fu MPa R mm L mm N’Rd N/mm Allowable load N/mm Permanent load N/mm Transient load N/mm
Unknown (Switzerland) 210 3000 70 500 69000 7960 – –
Thelwall (The UK) 210 1778 85 490 29429 – 13204 5530
Öland (Sweden) 210 1828 65 523 23789 9063 – –
Skellefteå (Sweden) 210 1586 65 553 17907 8336 4345 2714
a
E = 210 GPa for steels is recommended in EC EN 1993.

where cm = 1 is recommended. The design resistance of the roller bearings failed is calculated and listed in Table 2. Analo-
gous approaches are also followed by other standards. The allowable load recommended by the roller manufactures and the
permanent and transient loads calculated by the designers of the bridges [6,7,9,10,16] are also shown in Table 2 for compar-
ison. The design resistance values of the rollers are well in excess of the total of the corresponding permanent and transient
loads.

2.3. Tolerances

According to EN 1337-4, flatness, surface profile, surface roughness and parallelism of the contact surfaces must be con-
sidered for the design of a roller bearing. Among those tolerances, flatness and parallelism of the contact surfaces are mainly
affected by

(a) out-of-squareness of the lower girder flange at a structural bearing,

(b) out-of-flatness of the flanges, and
(c) temperature differences during erection of the bridge superstructure.

The out-of-squareness imperfection can cause severe stress concentration at a roller end sufficient for failure, Fig. 2(a).
The out-of-flatness imperfection results in either severe contact stress concentration at the roller ends or bending along
the axis of the roller. The latter can cause failure at the middle, weaker part, of the roller, Fig. 2(b). Finally, the deflection
of the girder due to temperature differences between primary and secondary structural parts can also result in out-of-
squareness imperfection, Fig. 2(c).

2.4. Material

Only ferrous materials as specified in Table 3 shall be used in the manufacture of rollers and roller plates, according to EN
1337-4. Stainless steel shall be in accordance with EN 10 088-2 and the minimum tensile strength shall be 490 N/mm2 for
any component. According to AASHTO, roller bearings shall be made of stainless steel conforming to ASTM A240.
The rollers failed were manufactured from a quenched and tempered martensitic stainless steel which is similar in com-
position to X40Cr13 (1.4034) or AISI 420. The chemical composition, uniaxial tensile test and hardness requirements for
these materials are given in Table 4 and Table 5, respectively.

3. Experiments

3.1. Preliminary examination of the failed rollers

The fractures of the rollers appeared most likely to have initially started at the end face of the rollers close to a contact
interface and extended in a radial plane through the roller centre and longitudinally along the axis of the roller (see Fig. 3).
However, in some cases multiple cracks were found at the end face of a roller (e.g. see Fig. 3(b) RH), which subsequently

Fig. 2. Bearing failure as a result of (a) out-of-squareness, (b) out-of-flatness [9] and (c) temperature differences [9].
 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 319

Table 3
Ferrous material classes.

Material Tensile strength Yield strength Impact/at Surface hardness Elongation Friction
class (minimum) [N/mm2] (minimum) [N/mm2] temperature (maximum) [HV 10] (minimum) [%] coefficient
(minimum) [J] (maximum)
A 340 240 27/0 °C 150 25 0.05
B 490 335 27/20 °C 250 21 0.05
C 600 420 27/20 °C 450 14 0.02
D 1350 1200 11/20 °C 480 12 0.02

Table 4
Chemical composition requirements.

Standard Number % by mass (maximum values)

C Si Mn P S Cr Ni Mo
EN 10088-2 1.4034 0.43 to 0.50 1.00 1.50 0.040 0.015 12.5 to 14.5 – –
ASTM A240 420 0.15 min 1.00 1.50 0.040 0.030 12 to 14 0.75 0.50

Table 5
Uniaxial tensile test and hardness requirements.

Standard Number Heat treatment Hardness Yield strength MPa, min. Tensile strength MPa Elongation in 50 mm%, min.
EN 10088-2 1.4034 Q&T 510 to 610 (HV) – – –
ASTM A240 420 Q&T 217 (HBW) 690 – 15

Fig. 3. Roller bearings failed on (a) the Öland Bridge in Sweden [5], (b) a bridge in Switzerland [6], (c) the Thelwall viaduct in the UK [7] and (d) the
Skellefteå Bridge in Sweden [9,10].

merged into a main crack. The most probable fracture origin is assumed situated on the diameter through the 6 and 12
o’clock positions at the end face (see Fig. 3(a,c), and A1 and A2 in Fig. 3(d)). This assumption was confirmed by performing
magnetic particle inspection [7]. Fig. 3(a,c) and B1 and B2 in Fig. 3(d) show further beach-marks on the fracture surface indi-
cating progressive crack extension with alternating periods of growth and arrest. Fig. 3(d) shows the final fracture surface C
and also the area along the contact surface of the roller where the paint coating layer was removed by wear in service D.

3.2. Chemical composition

The chemical composition of the failed rollers conformed to a 0.40–0.45% carbon and 13–14% chromium martensitic
stainless steel. The failed rollers were manufactured of two different materials, denoted X40Cr13 (1.4034) and AISI 420,
but almost similar in composition.

3.3. Mechanical testing

Mechanical properties of the rollers are summarized in Table 6.

As shown in Table 6, the hardness measured through the thickness of the rollers varies very little. However, in the vicinity
of the most probable fracture origin A1 (see Fig. 3(d)) the average hardness across the loading contact line was 640 HV and
some 10% greater than the specification requirement [15] of 580 HV.
320 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Table 6
Mechanical properties.

Bridge Region Hardness RP0.2 Rm Elongation% KIc KIIc Cv Ref.

HV N/mm2 N/mm2 – N/mm3/2 N/mm3/2 J
Öland Sweden 570–600 – – – – – – [5]
a
Unknown Switzerland 460–500 1750 >3000a,b >5a,b 1150c 2500c – [6,12]
Thelwall The UK 491–610 1385 1778 3 – – – [7]
Öland Sweden 576–588f 1266 1828 3.63 792.6d – 4.25e [10]
Skellefteå Sweden – 1478 1586 1.11 792.6d – 3.14e [10]
a
The uniaxial stress-strain curve has been determined from strain-gauge instrumented four-point bend tests following an experimental procedure
described in [13].
b
The strain gauge failed at a strain level of approximately 2.5%. The Rm was calculated from the measured ultimate load and the elongation% was
determined from the elongation of the aligned specimen parts after fracture.
c
The KIc and KIIc were determined at room temperature following the procedure described in [12].
d
The KIc was determined according to [14] at 30 °C.
e
Standard Charpy V-notch toughness testing was at 20 °C.
f
The hardness was measured across a cross-section surface far away from the end face of the roller.

3.4. Metallurgical and fractographic examinations

The microstructure of the rollers was tempered martensitic with finely dispersed carbides [5–11], Fig. 4.
The globally flat lustrous fracture surface, neither shear lips nor lateral contraction on an Öland Bridge Charpy impact test
specimen at the testing temperature 20 °C indicated a brittle fracture mechanism, Fig. 5(a). However, at high magnification
a mixture of micro-voids, quasi-cleavage and secondary cracking revealed a failure mechanism of very low ductility in [6,10],
Fig. 5(a,b).
The metallurgical examination of the fractured rollers in [7] indicated that the failures were associated with intergranular
corrosion typical of hydrogen-assisted stress corrosion cracking (HASCC), Fig. 6(a). The microstructural examination of the
secondary cracks, on a cross-section perpendicular to the fracture surface, in the failed roller of the Öland Bridge also showed
a crack pattern typical of stress corrosion cracking (SCC), Fig. 6(b).
Extensive SEM examination of fracture surfaces of the failed roller bearings on the Thelwall viaduct [7] and the Öland
Bridge [10] at high magnifications showed no evidence of fatigue striations. However, beach-marks were observed on the
fracture surface, Fig. 7.
A stress corrosion test was carried out in [7] in order to examine the susceptibility of the roller material to HASCC. After
several weeks the sample loaded in bending failed and the fracture surface showed the presence of beach-marks, even
though the loading on the sample was static, not cyclic.
Energy dispersive spectroscopy (EDS) analysis on the cross-section surface across the cavity and at the final crack front on
the fracture surface indicated the presence of chloride ions in those regions, Fig. 8. The results revealed the presence of some
4.5 (wt%) chloride inside the cavity at C and at the crack front shown in Fig. 8.
EDS analysis of the carbides in Fig. 9 indicated precipitation of chromium-rich carbides with up to 55% chromium which is
much greater than the matrix mean chromium content of 13%.

Fig. 4. Optical micrograph of the microstructure [10].

 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 321

Fig. 5. (a) Charpy specimen fracture surface of the Öland Bridge and (b) the fracture surface of the Öland Bridge [10].

Fig. 6. Optical micrograph of (a) HASCC of the roller failed on the Thelwall viaduct [7] and (b) the secondary cracks on the roller failed on the Öland Bridge
[10].

Fig. 7. Fracture surface of the failed roller on the Öland Bridge: No evidence of fatigue striations was found [10].

Fig. 8. (a) Line spectrum EDS analysis across a cavity and (b) EDS analysis at the crack tip on the fracture surface [10].
322 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Fig. 9. SEM micrograph of the chromium carbides [10].

4. Stress analysis

As a curiosity, the first attempt of roller stress analysis was made by Martson and Carandall [17] already in 1893 using the
photoelastic method, Fig. 10.
Using present day methods Bushell and Prinja [18] performed a two-dimensional plane strain finite element analysis of a
section of the bearing away from the roller ends consisting of a roller held in contact with two plates, with a further base
plate beneath the roller, Fig. 11(a).
The results showed that along a roller radius through the centre of the contact interface, the roller is under pure compres-
sion both in the vertical and the hoop directions, Fig. 11(b). S11 and S22 are normal stresses in the vertical and the hoop
directions, respectively. Off the radius through the centre, a shear stress, S12, in the vertical-hoop plane, develops in the tran-
sition regions towards unloaded material. It was also found that the material begun to yield in radial compression around the
centre of the contact interface followed by shearing of the surrounding material.
Noury and Eriksson [10] performed a three-dimensional finite element analysis in order to clarify the mechanism causing
the Skellefteå Bridge bearing failure, Fig. 12. The results showed that the hoop stress is compressive within the first 3 mm
from the contact interface, Fig. 13(a), but becomes tensile at greater distances from the contact interface. At the allowable
applied load, a maximum tensile value of the hoop stress, of about 150 MPa, occurs in a region some 5–6 mm above the con-
tact interface. The maximum shear stress appear along lines ± 45° from the loading line, also some 5–6 mm above the con-
tact interface, with the extreme values of ±440 MPa, Fig. 13(b). As the roller rotates on the contact plate with expansion and
contraction of the bridge, the region at 5–6 mm above the contact interface where experiences the maximum tensile value of
the hoop stress, of about 150 MPa, moves through the shear stress area. Since, the shear stresses reverse on either side of the
contact area, the maximum shear stress range experienced as the roller moves through the position immediately under the
instantaneous contact zone is of the order of 880 MPa.
Koltsakis et al. [9] used three-dimensional finite element models to investigate the accuracy of the traditional roller bear-
ing design rules regarding abutment and girder deformation and misalignment. Here, the same finite element model of the
Skellefteå Bridge as used in [9] has been employed in order to address the influence of misalignment of contact surfaces,
Fig. 12. A misalignment (deviation from ideal parallelism) of maximum 1/1000 of the contact surfaces along the roller axis
is allowed in EN 1337-4, and this condition can only be met through careful adjustments during the construction work.
Therefore, a perfectly aligned roller and with misalignment of 1/1600, 1/800 and 1/400 are considered. Fig. 14 shows the

Fig. 10. Stress pattern for a roller with increasing load (1 lbs  4.45 N) [13].
 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 323

Fig. 11. (a) Model mesh and (b) stress system associated with contact interface [18].

Fig. 12. (a) Arrangement of the bearing: roller (blue), roller plates (red), outer plates (green), shear keys (yellow), flange, web stiffeners and web (orange)
and abutment (grey) (b) single roller bearing, and (c) mesh detail [10]. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

Fig. 13. Stresses associated with the contact region: (a) hoop stresses, and (b) shear stresses [10].
324 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Fig. 14. Maximum contact stress distribution along the contact interface between the roller and the upper roller plates for (a) elastic material and (b)
elastic-plastic material.

maximum contact stress, which appears along the centre of the contact interface between the roller and the upper roller
plate, over the length of the roller at the allowable roller load, see Table 2. A linear elastic material with elastic modulus
E = 210 GPa and Poisson’s number m = 0.3 has been assumed, as well as a nonlinear elastic-plastic material model, for the
Skellefteå Bridge bearing. The dashed line indicates the allowable contact stress of 2500 MPa defined in [15] for the roller
bearings made of X40Cr13 loaded by permanent plus transient loads.
The maximum contact stress distribution along the contact line of the roller for the linear elastic material is shown in
Fig. 14(a). It is noted that due to stress concentration at the roller ends the maximum contact stress exceeds the allowable
contact stress for all misalignments considered. Also, stress concentration increases with misalignment. The corresponding
stress distribution for the elastic-plastic material is shown in Fig. 14(b). In this case, the maximum contact stresses is signif-
icantly smaller than the allowable contact stress, even at the roller ends.
The maximum contact stress attains a global maximum at the centre of the contact area some 2 mm away from the roller
end face, in the axial direction of the roller, Fig. 15(a). Due to plastic deformation in a small region extending from the roller
ends and within the contact area (see Fig. 15(b)) the contact stress was reduced and redistributed in this region, Fig. 14(b).

5. Fracture mechanics

Perhaps the best approach towards an explanation of fractures observed in practice is that by Schindler [19,20] introduc-
ing the idea of a swinging Hertzian stress-field: due to daily and seasonal thermal deformation of a bridge girder (superstruc-
ture), a radial crack rotates around a mean position under predominantly Mode II loading.
Stress intensity factors (SIF) for an edge-cracked disk subjected to crack surface loading have been obtained by several
authors using the weight function method and ad hoc manipulations [26–29], while solutions for compressive radial loading,

Fig. 15. (a) Contact stress distribution and (b) equivalent plastic strain distribution for the perfectly aligned roller.
 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 325

which occurs in practice, have not been found in the literature. However, none of the weight function method solutions
appear to give satisfactory results for the cracked roller.
Considering a cracked roller bearing as an edge-cracked disk and the resultant of compressive stresses acting on the disk
as a point load along a diameter, relative to which the crack is rotating, Schindler derived an approximate closed form solu-
tion for KI using the principle of superposition [19,20]. Using a similar procedure Noury and Eriksson [11] re-derived a some-
what less approximate solution for KI in [19]. For KII, Schindler [19,20] suggested the SIF for an edge crack in a strip of width
2R, where R is disk radius, loaded by two coplanar opposite forces acting on each side of the crack-mouth [21]. However,
none of these solutions permits calculation of SIFs for the relevant angle of rotation. Recently, Noury and Eriksson [11] con-
sidered the cracked roller as a two-dimensional edge-cracked disk subjected to a radial compressive line load and deter-
mined KI and KII for the relevant load configuration and disk rotation angle using the finite element method, Fig. 16.
Fig. 17 shows dimensionless KI and KII as functions of the disk rotation angle a. For a = 0, the crack is subjected to pure
Mode I loading. As a increases, crack opening in general decreases. It was also found that Mode II SIF dominates with a max-
imum value at a certain rotation of the crack in which position, the most unfavourable, material on only one side of the crack
is loaded by the support plate and the other is free to slide.
Noury and Eriksson [10] further performed three-dimensional finite element analyses in order to gain an understanding
of the crack growth mechanism of a bridge roller bearing in service. Fig. 18 shows the finite element model used to compute
the SIFs of an initial quarter-circle radial corner crack of 20 mm radius for (i) a perfectly aligned bearing, and (ii) a bearing
with a misalignment of 1/600 applied. Koltsakis et al. [9] adding the maximum allowable misalignment, and the
temperature-induced misalignment, calculated 1/600 as the total misalignment for the Skellefteå Bridge. The shape and
the size of the crack were chosen in accordance to the preliminary examination of the fracture surface, see A1 in Fig. 3(d) RH.
Calculated SIFs showed that the crack loading for all values of a were predominantly in Mode II and Mode III at node 1
(the crack front node on the roller end face) and node 16, respectively, Fig. 19(b,c). However, the KI and KII for the perfectly
aligned roller did not exceed the corresponding fracture toughness values KIc and KIIc of the material, Fig. 19(a,b).
The stress intensity factors increased significantly for a misalignment of 1/600, Fig. 20. KI in the range 0  a < 1° and KII for
a  2.8° roller rotation exceeds KIc and KIIc, respectively, and are alone thought sufficient for fracture. Keff and KII, the predom-
inant SIF, are in the a  2.8° region so much greater than KIIc that a mixed-mode fracture condition would most likely also be
satisfied, or alternatively, fracture might ensue for a much smaller crack and/or a larger misalignment.
Schindler et al. [22] measured residual stresses over the diameter of a roller (see Fig. 21(a)) by determining the SIF of a
crack cut along a diameter of a cross-section using the cut compliance (CC) method [23], Fig. 21(b). In short, the CC method is
destructive and the residual stress distribution is determined by making a narrow cut or slit of progressively increasing
length and measuring the corresponding strain change at a suitable location. Using linear elastic fracture mechanics (LEFM)
relations, the corresponding SIF for a certain crack length is obtained directly for the residual stress from the strain change at
the location of the strain gauge [22].
As shown in Fig. 21(a) the maximum tensile residual stress measured is about 170 MPa, or about 10% of the yield stress,
1750 MPa, and occurred near the contact interfaces. The residual stresses alone may cause the SIF to exceed KIc of the roller
material considered, thereby resulting in unstable crack extension, Fig. 21(b). Considering the SIF due to the residual stresses,
a region of unstable crack growth was determined for crack depths from a = 0.05 D to about a = 0.35 D, depending on the
crack orientations with respect to the applied load, Fig. 21(b).

Fig. 16. (a) Cracked roller loaded by compressive radial force and (b) finite element mesh used for the roller containing a crack depth ratio of a/(2R) = 0.2
[11].
326 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Fig. 17. Variation of SIFs: (a) KI and (b) KII [11].

Fig. 18. (a) Mesh detail and (b) node positions on the crack front [10].

6. Discussion

A very small fracture elongation of 1 to 5% in uniaxial tensile testing, a Charpy V-notch toughness of about 4 Joules at
20 °C, a very low fracture toughness KIc = 792.6 N/mm3/2 at 30 °C and 1150 N/mm3/2 at room temperature for the roller
materials considered, Table 1, indicate all a rather brittle material. Considering the stainless steel rollers as class D in Table 3,
neither fracture elongation nor impact toughness complied with the corresponding requirements.
The hardness measured over the thickness of the Skellefteå roller complied with the specification requirements [15].
However, the hardness measured in the vicinity of the most probable fracture origin of this roller, shown in Fig 1(d), was
some 10% higher than the specification requirements [15]. The increased hardness may be induced by plastic deformation,
Fig. 15(b).
Visually observed beach-marks on the fracture surfaces revealed progressive crack growth, Fig. 3(a, c, d), but SEM exam-
ination of fracture surfaces showed no evidence of fatigue striations. This may be related to the very hard and/or brittle roller
material.
Initial multiple cracking at the end face of the roller shown Fig. 3(b) RH, is in agreement with a high stress concentration
in this region.
EDS analysis of carbides in the microstructure showed the precipitation of chromium-rich carbides with a chromium con-
tent of up to 55%, Fig. 9. Heat treatment [24,25] during manufacturing of the rollers and/or local friction heating due to crack
faces sliding relative to one another in service affect the material hardness and results in chromium-depleted regions in the
microstructure. Exposure of such regions to an environment sufficiently aggressive to cause corrosion creates a risk of SCC.
Road de-icing salt, salt aerosol from the coastal environment and hydrogen chloride gas from nearby industrial plants also
most likely contribute to a corrosive atmospheric environment.
 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 327

 12
K2
Fig. 19. Variation of SIFs at node 1, 11 and 16 for a perfectly aligned roller: (a) KI, (b) KII, (c) KIII and (d) K eff ¼ K 2I þ K 2II þ 1IIIm [10].

The fractographic examination of the fractured rollers indicated that the initiation of the failures were associated with
SCC or intergranular corrosion cracking typical of HASCC, Fig. 6. EDS analysis of the failed Öland Bridge roller in Fig. 8 showed
the presence of chloride ions in the microstructure. Intergranular corrosion cracking may be promoted by chloride ions: the
roller of a high strength and high hardness martensitic stainless steel, was embrittled due to presence of atomic hydrogen
generated by intergranular corrosion. Then, the roller embrittled by atomic hydrogen is susceptible to SCC when subjected to
the hoop tensile stress region some 5–6 mm above the contact interface on the end face of the roller, Fig. 13(a), and the resid-
ual stress distribution over the roller diameter, with a maximum value of about 10% of the yield stress, Fig. 21(a). The hoop
and residual stresses are assumed sufficient for crack initiation through SSC in an aggressive environment. The associated
initial crack is assumed to be a surface crack along the load line at mean ambient temperature and of size sufficient for fur-
ther growth through foremost fatigue.
The analysis of the original dimensioning of the bearings showed that both the allowable load and the design resistance
were greater than the applied load (permanent + transient) in service (e.g. see the Skellefteå Bridge and the Thelwall viaduct,
respectively, Table 2). Therefore, it was unlikely that the bearings failed due to static overloading. It should be noted that
even for the misalignment considered, the contact stresses alone are not sufficient for failure, Fig. 14(b).
Perhaps the most likely failure scenario of a roller is that crack initiation in a tensile hoop stress region at the end face of
the roller is facilitated by an aggressive environment and a susceptible microstructure. Due to daily and seasonal tempera-
ture differences within a bridge superstructure, the roller rotates somewhat and the initial crack travels back and forth
328 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329

Fig. 20. Variation of KI, KII, KIII and Keff at the end face crack front position (node 1) for a misaligned roller plate (imperfection = 1/600) along the roller axis
[10].

Fig. 21. (a) Residual stresses measured along a diameter and (b) the corresponding SIF of a crack emanating from the surface in comparison with fracture
toughness KIc. [22].

around a mean position, leading to growth of the crack through fatigue. The hoop stress governs crack growth in the roller
both radially and longitudinally in Mode I and Mode III, respectively, but the shear stress governs growth only radially in
Mode II, Fig. 13. After a certain number of load cycles and for a certain misalignment, failure through sudden fracture occurs
as the crack attains such a critical size that the Keff at some point along the crack front exceeds the effective fracture tough-
ness of the roller material, Fig. 20. Failure is expected to occur for the angle a in the range of about 0.05D to 0.35D depending
upon the influence of residual stresses, Fig. 21(b). However, very low fracture toughness of the material and misalignment
allows failure to occur for even smaller than 0.05 D cracks.

7. Conclusions

The present investigation concerns the mechanics and failure mechanism of single bridge roller bearings made of high
strength stainless steel. The mechanisms involved in the roller failure scenario can be described as the following:

 The initiation of the failures was associated with SCC or HASCC due to exposure of the susceptible material of the roller to
an environment sufficiently aggressive.
 The susceptible microstructure of the roller material was subjected to the hoop tensile stress region some 5–6 mm above
the contact interface on the end face of the roller and the residual stress distribution over the roller diameter. The hoop
stress range and the residual stress are assumed sufficient for crack initiation through SSC in an aggressive environment.
 Once initiated, the crack grows in a radial plane through fatigue under multiaxial loading mainly by longitudinal thermal
deformation of the bridge.
 P. Noury, K. Eriksson / Engineering Fracture Mechanics 180 (2017) 315–329 329

 Finally, the service loading in conjunction with certain (or a critical) roller misalignment can be sufficient for sudden
fracture.

In summary, the primary cause of the failure is considered to be due to inadequate attention of the material specifications
to corrosion resistance of the high strength stainless steel rollers. The failure is primarily assumed related to SCC: the SCC
susceptible material of the roller exposed to a sufficiently aggressive environment was subjected to the hoop tensile stress
region some 5–6 mm above the contact interface on the end face of the roller and residual tensile stresses. Then, the cracks
caused by SCC, grew in multiaxial fatigue. Finally, due to thermal deformation of the bridge superstructure, the fatigue crack
combined with certain wedge imperfection was sufficient for final failure. However, in the absence of the wedge imperfec-
tion, roller failures seem to be unlikely. Wedge imperfection causes inelastic deformation at the contact interface at the end
edges of a roller, resulting in a hardness level in excess of the specification requirements.

8. Recommendations

Accurate assessment of steel bearings presently in service for renovation, widening, maintenance, and strengthening of
existing bridges is possible only if reliable fundamental knowledge of related issues is maintained. Although replacing of
steel bearings may appear a straight forward solution, in practice such a procedure often proves to be expensive, complicated
and impractical. To prevent or delay roller failures, the parameters which are most suitably adjusted are the roller diameter
and the material properties, in particular the material strength and toughness. A larger roller diameter and a higher strength
allow larger contact stresses, but they result in a greater bearing which increases the risk of lateral instability and a lower
toughness, respectively. Using mating surface or some other surface modification methods which allow a material with high
strength and hardness at the surface and a high toughness bulk material can also increase susceptibility for failure. Signif-
icant improvement of high strength stainless steel bearings is relatively difficult to achieve. Therefore, if possible it is rec-
ommended to replace such bearings with a new type of bearing.

Acknowledgments

This work was sponsored by the Swedish Transport Administration (Trafikverket).

References

[1] König G, Maurer R, Zichner T. Spannbeton: Bewährung im Brückenbau; Analyse von Bauwerksdaten. Berlin, Heidelberg, New York, London, Paris,
Tokyo: Schäden und Erhaltungskosten. Springer-Verlag; 1986.
[2] Karig J. Vorschlag für ein einheitliches Brückenlager. Bautechnik 1923;1(357–60):416–33.
[3] Wetzk V. Brückenlager. 1850–1950. Dissertation. BTU Cottbus. Online: urn:nbn:de:kobv:co1-opus-20692, 21. 12; 2010.
[4] Block T, Kauschke W, Eggert H. Lager im Bauwesen. Berlin: Ernst & Sohn; 2013.
[5] Linder HG, Mårtensson J. I sprucken lagerrulle av stål till Ölandsbron. Stockholm: Statens Provningsanstalt; 1977.
[6] Schindler HJ, Morf U. Schadens- und Sicherheitsanalysen an betriebsgeschädigten Stahlrollen von Brückenlagern. In: Bundesamt für Strassenbau
(Hrsg.): Forschungsauftrag 87/90 auf Antrag der Arbeitsgruppe Brückenforschung; 1995.
[7] Edwards T, Schofield M, Burdekin M. Failure of roller bearings on the Thelwall Viaduct. In: 4th International Conference on Forensic Engineering: From
Failure to Understanding, London; 2009.
[8] Yanagihara M, Matsuzawa T, Kudo M and Turker T. Replacement of Bearings in the Golden Horn Bridge, Turkey. Structural Engineering International
2000;10:121–123(3).
[9] Koltsakis EK, Noury P, Veljkovic M. The contact problem of roller bearings: investigation of observed failures. Struct Eng Int 2016;26:207–15.
[10] Noury P, Eriksson K. Failure analysis of martensitic stainless steel bridge roller bearings. Engineering Failure Analysis (in press).
[11] Noury P, Eriksson K. Determination of stress intensity factors for cracked bridge roller bearings using finite element analyses. Engng Fract Mech
2017;169:67–73.
[12] Veidt M, Schindler HJ. On the effect of notch radius and local friction on the Mode I and Mode II fracture toughness of a high-strength steel. Engng Fract
Mech 1997;58:223–31.
[13] Mayville RA, Finnie I. Uniaxial stress-strain curves from a bending test. Exp Mech 1982;22:197–201.
[14] ASTM-E399. Standard test method for linear-elastic plane-strain fracture toughness of metallic materials. American Society for Testing and Materials,
2009.
[15] Stahlbau DER. Zulassungbedingungen für Kreutz-Edelstahl-Lage. 1967:254–5.
[16] Vectura Report: Bro24-364-1: ‘‘Bärighetsutredning inför broreparation”, Uppdragsnr: 111726, 2012–07-02.
[17] Crandall CL, Marston A. Friction Rollers. Transactions of the American Society of Civil Engineers 1894;32:99–129.
[18] Bushell JM, Prinja NK. Prediction of residual stresses in bridge roller bearings using Abaqus. In: Abaqus Users’ Conference, 2008.
[19] Schindler HJ. Crack Growth in Rollers Due to Moving Hertzian Compression. In: 8th European Conference on Fracture, Torino; 1990.
[20] Schindler HJ and Morf U. Load Bearing Capacity of Cracked Rollers Containing Residual Stresses. In: l0th European Conference on Fracture; 1994.
[21] Tada H, Paris PC, Irwin GR. The stress analysis of cracks handbook. 3rd ed. New York: ASME Press; 2000.
[22] Schindler HJ, Cheng W, Finnie I. Experimental determination of stress intensity factors due to residual stresses. Exp Mech 1997;37:272–7.
[23] Cheng W, Finnie I. Measurement of residual hoop stresses in cylinders using the compliance method. ASME J. Eng Mat. 1986;108:87–92.
[24] Bhambri SK. Intergranular fracture in 13 wt% chromium martensitic stainless steel. J Mater Sci 1986;21:1741–6.
[25] Nasery Isfahany A, Saghafian H, Borhani G. The effect of heat treatment on mechanical properties and corrosion behaviour of AISI420 martensitic
stainless steel. J Alloy Compd 2011;509:3931–6.
[26] Wu XR, Carlsson AJ. Weight function and stress intensity factor solutions. BPCC Wheatson Ltd, Exeter: Great Britain; 1991.
[27] Fett T and Munz D. Proceedings 23. In: Vortragsveranstaltung des DVMArbetskreises Bruchvorgnge, Berlin, 1991:249–259.
[28] Schindler HJ. Weight functions for deep cracks and high stress gradients. In: ICF8, Kiev; 1993.
[29] Cheng W, Finnie I. Residual stress measurement and the slitting method. Springer Science+Business Media: LLC, Berkeley, CA; 2007. p. 177–80.