J Engfailanal 2016 12 023

EFA-03038; No of Pages 16
Engineering Failure Analysis xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal
Failure analysis of a pull rod actuator of an ATOX raw mill used in the cement
production process
Ruben Barata a, Rui F. Martins a,⁎, Tiago Albarran b, Tiago Santos b, A. Mourão a
a
UNIDEMI, Faculty of Sciences and Technology, Universidade Nova de Lisboa, Department of Mechanical and Industrial Engineering, Campus de Caparica,
2829-526 Monte de Caparica, Portugal
b
SECIL, Maintenance Engineering, Secil Outão Plant-Apartado 71, 2901-864 Setúbal, Portugal
a r t i c l e i n f o a b s t r a c t
Article history: A failure analysis of a forged and machined pull rod made of a quenched and tempered high
Received 3 November 2016 strength steel – namely DIN 34CrNiMo6 – is presented in this paper. The pull rod under
Received in revised form 19 December 2016 study is one of the three rods that connect the three existing hydraulic cylinders to the roller
Accepted 19 December 2016
mill and it is used to exert grinding pressure on the roller in an ATOX raw mill. The study of
Available online xxxx
the failure involved the mechanical system's simulation, as well as the fatigue damage assess-
ment based on the application of the Rainflow Method and Fracture Mechanics principles. The
Keywords: pull rod's stress state was determined from in situ strain gages rosettes measurements, and sys-
ATOX raw mill
tem's simulation showed that the pull rod was designed to support axial loadings only. To en-
FKM-Guideline
sure this stress state, a proper functioning of the plain bearings and a good condition of the
Fatigue life prediction
Strain gages elastomers of the horizontal rod buffer are required. According to the stress-life approach
Failure analysis given in the FKM guideline, the current design solution should provide a useful life of
50 years to the pull rod. However, fatigue crack growth calculations made with the strain spec-
tra registered demonstrated that the presence of an initial surface crack is extremely harsh to
the lifespan of the rod, which will reach a critical value in a much shorter period of time. Sur-
face finish, corrosion protection and maintenance actions to ensure the correct functioning of
the plain bearings, could be important to ensure the desired longevity of the pull rod.
© 2017 Elsevier Ltd. All rights reserved.
1. Introduction
The main goal of the work herein presented was to analyse a mechanical failure that occurred on a pull rod of an ATOX raw
mill. The ATOX raw mill under study is shown in Fig. 1a and b, in which are represented the most important components of the
raw mill, as well as the recurrent location of failure of the pull rod. The ATOX raw mill is used to prepare kiln feed for the pro-
duction of cement clinker, and it is one of the most important equipments in the production process of cement, whereby a failure
could have a negative impact in the production process and, consequently, in the incomes of the enterprise. The ATOX raw mill
uses compression and shear generated between the rollers and the rotating table to crush and grind raw materials (Fig. 1b). The
raw materials are directly feed onto the grinding table by the feed chute (Fig. 1b) and the rotation of the grinding table brings the
materials towards the grinding track and under the rollers. The three rollers, each one weighting about 50 ton, are kept centred
on the grinding table by a triangular structure and by three horizontal rods that are connected to the mill body in order to
⁎ Corresponding author.
E-mail address: rfspm@fct.unl.pt (R.F. Martins).
http://dx.doi.org/10.1016/j.engfailanal.2016.12.023
1350-6307/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: R. Barata, et al., Failure analysis of a pull rod actuator of an ATOX raw mill used in the cement production
process, Engineering Failure Analysis (2017), http://dx.doi.org/10.1016/j.engfailanal.2016.12.023
2 R. Barata et al. / Engineering Failure Analysis xxx (2017) xxx–xxx
Nomenclature
2N number of reversals
a crack length
ao initial crack length
af critical crack length
CD size correction factor
CD,FKM size correction factor according to FKM-Guideline
CL load correction factor
CR reliability correction factor
CR,FKM reliability correction factor according to FKM-Guideline
CT temperature correction factor
CE,T,FKM temperature correction factor for the endurance limit according to FKM-Guideline
Cσ,R roughness correction factor for normal stress
Cσ,R,FKM roughness correction factor for normal stress according to FKM-Guideline
Cσ,E endurance limit factor for normal stress
Cσ,E,FKM endurance limit factor for normal stress according to FKM-Guideline
C ,m coefficient and exponent of the Paris law, respectively
da
dN
crack growth rate
ERT elastic modulus at room temperature
EET elastic modulus at elevated temperature
fw finite width correction factor
Gk,σ relative normal stress gradient
K stress intensity factor
Kf fatigue notch factor
Km stress intensity factor due to misalignment
Kt stress correction factor
Kt elastic stress concentration factor
Ktb bending stress concentration factor
Ktm membrane stress concentration factor
Y fracture mechanics geometric factor
Mm,t,tb,tm stress intensity magnification factors
Mσ mean stress sensitivity factor in normal stress
NE endurance cycle limit
nk,σ supporting factor
R stress ratio
S applied stress
Sa stress amplitude
Sar equivalent fully reversed normal stress amplitude
Se corrected endurance limit of a notched rod-shaped component under fully reversed loading
Sf fatigue strength coefficient
Sm normal mean stress in a stress cycle
Smin normal minimum stress in a stress cycle
Smax normal maximum stress in a stress cycle
ε strain
ν Poisson's ratio
σ1 maximum principal stress
σ2 minimum principal stress
φ direction of maximum principal stress
ΔS applied stress range
ΔSb bending stress range
ΔSm membrane stress range
prevent the roller's rotation (Fig. 1b, c). The grinding compression is axially applied by three pull rods, each of which is attached
to the roller shaft at one end side (Fig. 1b, c, d) and to a hydraulic cylinder at the opposite side (Fig. 1c) [1].
The ATOX raw mill entered in service in the 80s of the 20th century and since then maintenance has been periodically carried
out. The analysis of the most recent failures registered allowed to conclude that the pull rod failed four times between 2014 and
R. Barata et al. / Engineering Failure Analysis xxx (2017) xxx–xxx 3
Fig. 1. a) Main view of the ATOX raw mill under study; b) overall view of the most important components of an ATOX raw mill - adapted from [1]; c) pull rod
system modeled in a CAD 3D software and identification of recurrent location of failure; d) main dimensions (in millimeters) of the pull rod; e) typical fatigue
fracture surface obtained for the pull rod under study.
2015, namely on 09/04/2014, 20/06/2014, 11/08/2014, 05/08/2015, and failures systematically occurred at the top of pull rod no.
1 (Fig. 1b, part no. 5), either at the circular cross section of the rod with a constant diameter equal to ∅ 150 mm, Fig. 1d, or near
the transition between the rod and the flange (detail view, Fig. 1d).
An example of a typical fracture surface due to fatigue failure is presented in Fig. 1e. In fact, fracture surfaces of the pull rods
that failed in the past always revealed to be planar, with little sign of plasticity during crack propagation, and fatigue was always
the mechanism of crack propagation, as can be clearly seen by the observation of fatigue beachmarks present in Fig. 1e. In addi-
tion, fracture surface denotes a fatigue crack initiation point located at the outer cylindrical surface of the rod (Fig. 1e), and a crack
propagation typical of an axial tensile-compression state involving high to moderate nominal stresses, and starting from an
unnotched surface. Moreover, some inclusions are visible in the cross section of the rod, but were not the cause of initiation of
Table 1
Chemical composition of pull rod's material: comparison between experimental data and specifications given in [4], and in technical sheet [2].
C [%] Mn [%] Si [%] Cr [%] Ni [%] Mo [%]
EN 10083-3 [4] 0.30–0.38 0.50–0.80 Max 0.40 1.30–1.70 1.30–1.70 0.15–0.30

Technical Sheet [2] 0.36 0.64 0.23 1.62 1.60 0.28
Experimental data – 0.62 – 1.60 1.59 0.20
Table 2
Hardness values of pull rod's material: comparison between experimental data and specifications given in [2].
Depth from the treated surface [mm] 25 75

Hardness values [2] [HB] 279 264
Experimental data [HB] 277 270
the crack. Finally, failure reports did not indicate the failure of any bolts that connects the pull rod's flange to the joint head,
which, in case of occurrence, could indicates the application of bending or overloads to the pull rod.
As mentioned previously, forged pull rods were made of a high-strength steel, namely ref. DIN 34CrNiMo6, and were subjected
to quench (830 °C) followed by temper (630 °C) [2]. This type of steel is a versatile engineering material frequently employed in
the manufacture of gears, couplings, pull and torsion bars, aircrafts components, etc. In general, this type of steel covers an exten-
sive spectrum of applications due to both good strength-to-weight ratio and high toughness [3]. The chemical composition and
the hardness of the pull rod's material were determined by optical emission spectrometry (portable spectrometer Thermo
Niton XLT 898D XRF Analyzer) and a portable hardness tester (EQUOTIP 2), respectively. These analyses were made in different
points of a pull rod's cross section, in order to compare experimental data with chemical composition specified in EN 10083-3 [4]
and with technical data sheet material [2]. Results are shown in Tables 1 and 2.
Additionally, monotonic mechanical properties of steel ref. DIN 34CrNiMo6 were adopted from the literature [2], taking into
account the material, manufacturing process, heat treatments and the cross section diameter of the pull rod (Table 3).
2. Theory and calculation
This section provides an overall view of the functioning of the pull rod under study and describes the main theoretical con-
cepts that ensure its normal operation in order to avoid failure. In fact, fatigue is an important mechanism of crack propagation
that should be considered every time that a structural component is subjected to constant or variable amplitude loading [5], and it
is essential to understand the physics behind this phenomenon [6] in order to reduce or even eliminate the fatigue failures. The
type of failure herein presented started to be studied by the ATOX raw mill manufacturer [7], and aim to be developed in the
present research.
2.1. Cinematic of the mechanical system
The compression generated between the rollers and the rotating table in order to crush and grind raw materials is possible
due to the force applied by hydraulic cylinders. The pull rod is the mechanical element that connects the hydraulic cylinders to
the joint head of each roller. The assembly of the three rollers is kept centred in relation to the grinding table due to a triangular
structure connecting rollers and due to the three horizontal rods that avoid rotation. The buffer represented on Fig. 2(a), which
consists of two elastic plates arranged in series, allows accommodating some horizontal displacement resulting from the friction
developed between the roller and the rotating grinding table; in addition, if the grinding table remains horizontal and the centre
of gravity of the roller system only moves vertically, all impact forces and dynamic reactions would be mainly vertical. In this the-
oretical condition, the pull rods would be only subjected to axial forces, once the system has a set of plain bearings that couple the
pull and the horizontal rods, as can be seen in Fig. 2(a). In addition, in Fig. 2(b) it is represented a quite probable movement of
the system, caused by the change of the raw material layer level (material between the rotating table and the rollers). In this case,
the movement of any rigid element of the system causes a shift in the others (one degree of freedom), and angles θ and β are a
direct consequence of the ascendant movement of the hydraulic cylinder while offset Δ is mostly related to the different forces
that the rotation of the grinding table causes on the horizontal rod (Fig. 2b).
The simulation carried out in a computer aided design software allowed to realize that the system was designed in order that
the pull rods would ideally be subjected only to axial forces (Fig. 2a, b). However, it is important to realize that any mechanical
system has efficiency losses throughout their lifetime, and the malfunction of any plain bearing (universal joints) could cause re-
sistance to the natural movement of the system, forcing the pull rod to incur in a bending stress state (Fig. 3a). Fig. 3(a) shows
what the deterioration on the plain bearings can cause to the pull rod. In fact, the malfunction of the marked plain bearing will
not allow the pull rod to follow the desired movement. Thus, the pull rod will be forced to bend, due to the restriction to move-
ment of the joint head; in addition, the proper functioning of the plain bearings depends on some external factors, namely on the
correct positioning of the roller joints, on the right position of the base of the hydraulic cylinder in relation to the metallic shoe,
Table 3
Monotonic mechanical properties of quenched and tempered steel ref. DIN 34CrNiMo6 (adapted from [2]).
Tensile yield strength, SYS (MPa) 700

Ultimate tensile strength, SUTS (MPa) 900
Young's modulus, E (GPa) (20 °C) 220
Young's modulus, E (GPa) (200 °C) 195
Poisson's ratio, v 0.296
Reduction in area [%] 15.4
Fig. 2. (a) – Simplified representation of the system; (b) – simplified representation of the roller in ascendant movement.
and the elastic condition of elastomers of the buffer that will influence the tilt angle, θ, and the displacement allowed, Δ. In Fig.
3(b) is represented the case of an incorrect centring position between the base of the hydraulic cylinder and the metallic shoe.
This misalignment will result in improper functioning of the plain bearing and will not allow the pull rod to rotate, causing the
effect shown in Fig. 3(a).
2.2. Strain gages – calculation of principal stresses
Strain gages are used as sensing elements to measure indirectly load, pressure, displacement and torque. Therefore, when a
structure deforms due to the application of an external force, the strain gages attached to that structure will also deform and
its electrical resistance will change [8,9]. For a precise measurement of such change of resistance, an electric bridge circuit -
the Wheatstone Bridge - is used to convert the change of electrical resistance in strain variation that can be read using data log-
gers. Generally, if the direction of the principal strain is unknown, a strain gage rosette, which is an arrangement of two or more
closely positioned gage grids separately oriented to measure the normal strains along different directions in the underlying
Fig. 3. a) Simplified representation of the cinematic of the system in case of a spoiled plain bearing at the base of the hydraulic cylinder; b) representation of the
connection between the base of the hydraulic cylinder and the metallic shoe.
Fig. 4. a) Overall view of the region where the two strain-gages rosettes were installed; b) detail view of the electrical wires connected to the strain gages terminals.
surface of the test par, is bonded to the structure and the maximum and minimum principal stresses, σ1 and σ2, respectively, can
be calculated using Eqs. (1)–(3) [8,9].
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
E
σ1 ¼ ð1 þ v Þ ðεa þ εc Þ þ ð1−v Þ 2 ð ε a −ε b Þ 2
þ ð ε b −ε c Þ2
ð1Þ
2 ð1 þ vÞ2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
E
σ2 ¼ ð1 þ v Þ ðεa þ εc Þ−ð1−v Þ 2 ð ε a −ε b Þ 2
þ ð ε b −ε c Þ 2
ð2Þ
2 ð1 þ vÞ2
Fig. 5. a) National Instruments NI PXI-1052 device; b) Microlog CMVA 65 device.
The principal direction, φ, can also be calculated from the measured strains, εa, εb and εc [7,8] using Eq. (3). However, due to
the ambiguity related to the trigonometric function (tangent), it is necessary to determine the signs of the numerator and of the
denominator of the fraction before carrying out the final calculation of the quotient. This is important because it determines the
quadrant of the circular arc in which the angle φ is located [8,9].
1 ð2 εb −εa −εc Þ
tagðφÞ ¼ ð3Þ
2 εa −εc
In order to determine the principal stresses induced in the pull rods during its normal functioning, pull rod no. 1 was instru-
mented with two 45°-rectangular strain rosettes, which were installed near the local where fracture usually occurs (Fig. 4).
3. Materials and methods
The research presented in this paper uses two different fatigue design rules, namely the North American approach and the
FKM-Guideline [10], in order to characterise the fatigue resistance of the pull rod under analysis. Each of these methods has its
own strengths and domain of applicability, and the fatigue analyses of the pull rod were carried out using the experimental
data collected by the strain gages installed on the surface of pull rod no.1 that is actuated by the nitrogen-filled accumulators con-
nected to a hydraulic cylinder (Fig. 1b).
3.1. Experimental strain data
As previously mentioned, the instrumentation of the pull rod was carried out with two 45°-rectangular rosettes that were
installed according with procedures given in [8,9], and were angularly displaced one from the other by 90°, as can be seen in
Fig. 4(a). The two HBM rosettes were placed approximately at the middle of the pull rod length. Ideally they should be installed
at top of the rod; however, that zone is already inside of the raw mill and data collection became impossible due to the harsh
environmental conditions of work. In fact, in-service ambient conditions were so adverse, that it was necessary to repeat strain
gage data logging several times to have consistency in the results and lifetime of the strain rosettes was relatively short. Addition-
ally, in order to avoid the aliasing phenomenon, the pull rod working frequencies (10–60 Hz) were previously monitored by a SKF
Microlog CMVA 65 (Fig. 5a) and data acquisition frequency was set to 200 Hz in a NI PXI-1052 logger from National Instruments
(Fig. 5b).
The calculation of the principal strains and directions were carried out using Formulas (1)–(3) given in [8,9].
3.2. Fatigue analysis
The fatigue life analysis of the pull rod under study was carried out using two different design methodologies, namely the so
called North American solution and the FKM-Guideline approach [10]. Both are based on the concept that fatigue life of any struc-
tural non-cracked component can be estimated from S-N curves correspondent to the metallic material used to build the compo-
nents; those S-N curves are typically obtained from fatigue tests carried out on small polished standardized specimens (in
alternative, S-N curves can also represent the component under study), to which fatigue strength reductions factors are applied
afterwards. In fact, the two methods differ on the strength reduction factors applied in the high cycle fatigue regime. Nevertheless,
FKM method seems to include the latest scientific research published in the fatigue field.
Using the two techniques described above, the fatigue strength of a notched component can be determined either by Eqs.
(4) or (5), depending if the North American solution or the FKM-Guideline is used, respectively [10]:
0 0:504 C L C D C σ ;R C R C T SUTS
Se ¼ ð4Þ
Kf
0
C σ;E;FKM C D;FKM C R;FKM C T;FKM SUTS
Se;FKM ¼ ð5Þ
1
K f ;FKM þ −1
C σ;R;FKM
The differences between each strength reduction factor and the stress concentration factors, for the case under study, are pre-
sented in Section 4.
Both methods use sets of experiments under constant amplitude loading (CAL) conditions. However, most of service loading
histories are variable over time. For the estimation of fatigue lives under complex service loading, the variable amplitude loading
histories (VAL) must be decomposed into CAL cycles by means of a cycle counting method. Cycle counting is used to summarize
irregular and long load histories giving the number of cycles and the corresponding amplitude [11]. Definition of cycles depends
on each counting method. According to Dowling [12], the Rainflow Counting method is generally the one that leads to better fa-
tigue life time prediction. Therefore, with the use of a damage accumulation method, the overall damage caused by the VAL block
can then be calculated based on the sum of the individual amounts representing the damage caused by each cycle counted. In this
article, the Rainflow Counting method and the Palmgren-Miner damage rule were considered.
In addition, engineering components and structures are often subjected to cyclic loads with mean stress values different from
zero. Mean stress, which is the mean value of the maximum and the minimum stresses applied during one load cycle, is an im-
portant factor that affects the fatigue life, and it is very likely to be found in VAL histories [12]. In the high cycle fatigue regime,
mean stresses have a significant effect on the fatigue behaviour of components [12,13], and the fatigue limit might be negatively
influenced by other factors like periodic overloads, elevated temperatures or corrosion, and depending on the damage rule applied
in a VAL – Miners's rule or Miner-Haibach model, stress cycles with amplitudes below fatigue limit could cause damage to the
mechanical structure [10].
3.3. Fatigue crack propagation analysis
The aim to extend fatigue lives in almost all structures and mechanical components are stretching the limit of acceptance
levels of flaws, and, consequently, crack growth control becomes critical and it might provide an assessment tool for fatigue life-
time predictions. Therefore, if a crack or flaw is present in a mechanical component, a fatigue crack growth approach might be
considered in order to understand the number of cycles it takes to propagate from an initial crack size to a critical one. In
order to determine appropriate stress intensity factor ranges,ΔK, references such as [14,15] were used during research herein pre-
sented. This standard (BS 7910) provides a general procedure to assess the acceptability of flaws in metallic materials and carry
out fatigue crack growth simulations based on the Paris Law, which relates crack growth rates to the stress intensity factor range.
In BS 7910 [14,15], the general applied stress intensity factor range, ΔK, is given by Eq. (6):
pffiffiffiffiffiffi
ΔK ¼ ðYΔSÞ πa ð6Þ
and for fatigue assessments the following equation applies:
ðYΔSÞ ¼ M f w ½ktm M km Mm ΔSm þ ktb M kb M b ðΔSb þ ðkm −1ÞΔSm Þ ð7Þ
where subscript “m” relates to membrane stresses and “b” to bending stresses applied. According to the stress state characteristics
described in Sections 2.1 and 4, the loads applied to the pull rod are mainly axial. Hence, Eq. (7) can be simplified in order to
consider axial loading only (Eq. (8)):
ðYΔSÞ ¼ M f w ½ktm M km Mm ΔSm ð8Þ
where, for a semi-circular surface flaw with crack depth, a, present in a round bar of radius r: M = Mkm = fw = 1, with Mm as
follows:
a πa3
Mm ¼ g 0:752 þ 2:02 þ 0:37 1− sin ð9Þ

2r 4r
Fig. 6. Strain data acquisition at the early stage of mill's functioning. a) Rosette A; b) Rosette B.
where:
1:84 h πa πai0:5

tan =
g¼ π 4r
πa 4r for the range of application : a=2rb0:6 ð10Þ
cos
4r
4. Results and discussion
4.1. Strain data
The mechanical system of the mill is complex, with a large number of variables involved in its functioning. So, in order to val-
idate experimental data, strain data acquisition started with the grinding table rotating freely and the pull rods supporting the
dead weight of the rollers mill together with some weight of the central rigid triangular structure responsible by centring the
three rollers, which implies a compression stress state applied in the rods. As can be observed in Fig. 6(a) and (b), the strain mea-
surements registered at this initial stage were negative and almost constant. In fact, the weight of the whole system mentioned
above (dead weight of the rollers mill together with some weight of the central rigid triangular structure) is about 55 tons. Hence,
axial stress applied due to this axial force and is about −27.8 MPa according to the calculations presented in Eq. (11).
−F 49895; 175 9:81

S;theory ¼ ≅ −27; 8 MPa ð11Þ
Asection π 0:0752
This theoretical value (−27.8 MPa) should be compared with the experimental values recorded by the strain gage of the two
45°-rectangular strain rosettes (Fig. 6, strain2A and strain2B) aligned along the axial direction of the rod (Eqs. (12) and
(13)).
−4 9
Sstrain2A ¼ εaxial E ¼ −2; 0 10 220 10 ≅−44; 0 MPa ð12Þ
−4 9
Sstrain2B ¼ εaxial E ¼ −1; 3 10 220 10 ≅−28; 6 MPa ð13Þ
As can be seen, experimental axial stresses measured from rosette B (−28.6 MPa) are very close to the theoretical value cal-
culated with Eq. (11). For this reason, strain data collected by rosette B will be analysed further on.
In Figs. 7 and 8 are shown the maximum principal stress values induced in the instrumented pull-rod, during one-hour of con-
tinuous grinding, together with its principal direction, respectively. The values ranged from zero to 140 MPa (approximately) dur-
ing about 7.4 × 105 cycles.
From the analysis of Fig. 8, it can be seen that the average direction of maximum principal stresses is about 44° with a vari-
ation of approximately 1°. From the physical point of view, this means that the maximum principal stress direction is aligned with
the axial direction of the pull rod, and it points to the conclusion that the pull rod is mainly subjected to axial loading.
Table 4 resumes the application of the Rainflow counting method [16] to the experimental data represented in Fig. 7.
Fig. 7. Max. principal stresses calculated from registered strain data acquired by rosette B during 1 h of continuous grinding.
Fig. 8. Principal direction of maximum principal stress registered by rosette B (Fig. 7).
4.2. Fatigue analysis considering the North American solution and the FKM-Guideline
The fatigue calculations had into account some pull rod characteristics and the following preliminary considerations:
• Material: DIN 34CrNiMo6, quenched and tempered (Tables 1–3)
• Machined surface finish with a mean surface roughness of 12.5 μm;
• Operating temperature: 150 °C;
• The endurance limit was defined as the stress amplitude for fully reversed loading at NE = 106 cycles;
• Design should be based on a probability of survival rate of 97.5%;
• The fatigue notch factor, Kf,Total (Eq. (14)), resultant from superimposed notches, Kf1 and Kf2 (Fig. 1d, detail view), was estimated
as referred in [10]:

K f ;Total ¼ 1 þ K f 1 −1 þ K f 2 −1 ð14Þ
• Accordingly, the pull rod is considered a shoulder shaft containing two geometric discontinuities that will originate
superimposed stress gradients. To calculate the fatigue notch factor, Kf, the first step for both methods is to calculate the static
stress concentration factors, Kt, which, for a shouldered rod subjected to axial stresses, according to Peterson [17], is given by Eq.
(15):
2 3
2t 2t 2t
kt ¼ C 1 þ C 2 þ C3 þ C4 ð15Þ
D D D
Table 4
Loading characteristics and cycle counting (Fig. 7) by the Rainflow Method [16].
Block no. Sm max ½MPa Sa max ½MPa Stress ratio, R Stress range [MPa] Cycles
13 70.80 69.00 −0.05 138.00 124.5

12 67.80 62.10 −0.05 124.00 407.0
11 70.50 55.20 −0.06 110.00 263.5
10 76.40 48.30 −0.05 96.60 109.0
9 90.90 41.30 −0.04 82.80 76.5
8 104.00 34.40 −0.01 69.00 181.0
7 112.00 27.70 −0.03 55.20 449.5
6 122.00 20.70 −0.04 41.40 917.0
5 126.00 13.80 −0.04 27.60 952.5
4 128.00 10.30 −0.05 20.70 1575.0
3 128.00 6.90 −0.05 13.80 4130.0
2 134.00 3.45 −0.04 6.90 6573.5
1 138.00 1.72 −0.05 3.45 45,985.5
Tables 5 and 6 provide a comparison of the different coefficients values considering the North American factors calculated ac-
cording Budynas et al. [18] and Yung-Li Lee et al. [19], whereas for FKM methodology the document provided by Sen A. McKelvey
et al. [10] was used as reference.
FKM Guideline defines the slope factor (k) that is as the negative inverse of the typical slope (b) of an S-N curve in the high
cycle fatigue (HCF) regime, which results in the equation k = − 1/bk = −1/b [10], and according to the North American solution
[18], the fatigue strength exponent can be estimated through Eq. (16):

Sf S þ 345 900 þ 345
log log UTS log
Se 0:5 SUTS 0:5 900
b¼− ¼− ¼− ¼ −0:0701 ð16Þ
logð2NE Þ log 2 106 log 2 106
In addition, according with [10], for a surface hardened made of steel, k = 15. Hence, the fatigue life for the pull rod under fully
reversed loading can be calculated based on Eqs. (17) and (18), which represent S-N curves for the North American solution and
FKM-Guideline in the HCF regime, respectively:
−0:0701
Sa ¼ 280:376ð2NÞ ð17Þ
−1=15
Sa ¼ 302:912ð2NÞ ð18Þ
Schematic representation of Eqs. (17) and (18) is shown either in Figs. 9 or 10, respectively.
Additionally, mean stress effect needs to be taken into account to carry out fatigue life assessments when applying a stress-
based approach. Many numerical models were developed to consider this effect, namely Goodman's, Morrow's, Walker's or
Smith-Watson-Topper's, just to mention the most commonly used. For the North American solution, the ASME elliptic mean stress
correction is typically applied for ductile metallic materials [18]. Hence, the equivalent fully reversed stress amplitude according to
ASME's relation is shown in Eq. (19) [18]:
Sa
Sar ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ð19Þ
1− SSm
YS
Table 5
Coefficients used in the North American methodology [17–19].
Parameter North American
CL 0.900
CD d 1.000
Cσ, R (Rz = 12.5 μm) 0.700
CR 0.843
CT 1.017
Kt1 (r1 = 15 mm) 2.003
Kt2 (r1 = 56 mm) 1.424
Kf1 (q = 0.992) 1.995
Kf2 (q = 0.998) 1.423
Kf, Total 2.420
S′e [MPa] (Eq. (4)) 101.341
Table 6
Coefficients defined by FKM-Guideline [10. 17].
Parameter FKM
Cσ, E,FKM 0.400

CD,FKM (deff = 150; deff .min = 250) 1.000
CR, FKM 0.843
CE,T, FKM 0.930
Cσ, R,FKM (ar = 0.22; St. u.min = 400) 0.860
Kt1 (r1 = 15 mm) 2.003
Kt2 (r1 = 56 mm) 1.424
Kf1 (Gk, σ = 0.153; nk,σ = 1.059) 1.892
Kf2 (Gk,σ = 0.041; nk,σ = 1.020) 1.397
Kf, Total 2.289
S′e FKM [MPa] (Eq. (5)) 115.146
Alternatively, FKM-Guideline specifies the mean stress sensitivity factor (Mσ) (Eqs. (20) and (21)), which will depend on load-
ing condition (stress ratio) and the type of material. Considering the stress ratio values, R, given in Table 4, then regime II applies
(− ∞ ≤ R≤ 0) and the fully reversed stress amplitude could be calculated as follows (Eqs. (20) and (21)) [10]:
Sar ¼ Sa þ Mσ Sm ð20Þ
Mσ ¼ am St:u þ bM ðam ¼ 0:00035; bM ¼ −0:1 for steel ½10Þ ð21Þ
where St.u is the corrected ultimate tensile strength of a real component based on the probability of 97.5% survival rate (CR,FKM),
which leads to CD,FKM·CR,FKM·SUTS = 1 × 0.843 × 900 = 758.7 MPa (Tables 3 and 6) [10].
Fatigue damage results are shown on Table 7. For each stress block (Table 4), fatigue life, N, was calculated and the damage
correspondent to the number of cycles applied was determined according to the Palmgren-Miner linear damage rule. In addition,
to include the effect of periodic overloads and the presence of oxidation, an S-N curve with the slope factor, k, was extended be-
yond the fatigue limit value in order to approach zero stress amplitude load [10].
Palmgren and Miner found the critical damage value of 1.0. However, it has been recently shown that the critical damage
value could vary from 0.15 to 1.06 for mechanical designs. FKM [10] recommends 0.3 for the critical damage value for steels,
and this value was used for calculations herein presented (Table 7).
Considering that the ATOX raw mill works 5200 h per year, according to the damage values calculated (Table 7) and the crit-
ical damage value mention above (0.3), the pull rod will have a useful lifetime of approximately 57 years according to the North
American procedure and 50 years according to the FKM-Guideline. Hence, as long as functional requirements described in
Section 2.1 are fulfilled, it seems that failures registered for the pull-rod might be explained by the presence of surface defects
on the pull rod and its propagation during in service loading.
Fig. 9. Representation of the pull rod's S-N curve according to the North American approach (Eq. (17). Table 5).
Fig. 10. Representation of the pull rod's S-N curve according to the FKM approach (Eq. (18), Table 6).
4.3. Fatigue crack growth
In mechanical components, crack initiation can occur from inherent flaws existing in the materials, or from defects introduced
during assemblage, or after maintenance, or during normal/abnormal functioning (cyclic loadings, corrosion, etc.), and experimen-
tal practice shows that these situations cannot always be avoided. Hence, it is important to predict fatigue crack growth that could
start from an initial crack length, ao, frequently associated to a typical initial defect size and to a minimum probability of detection
of a non-destructive inspection method, until a final critical crack size, af. Therefore, stress under variable amplitude loading rep-
resented in Fig. 7 was converted into constant stress amplitude blocks (Table 7), also represented by histogram presented in Fig.
11.
The fatigue crack growth assessment, which was implemented in a Matlab routine (Fig. 12), followed the guidelines indicated
in BS 7910 [14–15] and considered the following hypotheses and values:
• The existence of a single semi-circular superficial crack, a0, with an initial size equal to 0.15 mm. This value did not result from
any measurement done in actual condition, but it was considered a reliable minimum threshold value if a common non-
destructive testing technique is used;
• The stress intensity factor range at the crack tip was determined by equations given in BS 7910 (see also Section 3.3, Eqs.
(6)–(10));
• For steels with yield or 0.2% proof strengths up to 600 MPa, frequencies higher than 1 Hz, and in-service temperature higher
than 100 °C and up to 600 °C, the recommended values for the constants m and C of Paris Law are as follows [15]:
m¼3
Table 7
Calculation of fatigue damage applied during 1 h of continuous grinding based on experimental strain data (Table 4).
Block North American solution FKM-Guideline
Sar [MPa]) N (cycles) Damage Sar [MPa]) N (cycles) Damage
13 69.356 2.23E+08 5.58E−07 80.721 2.06E+08 6.04E−07

12 62.393 1.01E+09 4.04E−07 73.324 8.71E+08 4.67E−07
11 55.482 5.37E+09 4.90E−08 66.871 3.47E+09 7.60E−08
10 48.590 3.56E+10 3.06E−09 60.948 1.39E+10 7.82E−09
9 41.653 3.20E+11 2.39E−10 56.348 4.52E+10 1.69E−09
8 34.786 4.18E+12 4.33E−11 51.617 1.69E+11 1.07E−09
7 28.062 8.93E+13 5.03E−12 46.241 8.78E+11 5.12E−10
6 21.022 5.49E+15 1.67E−13 40.896 5.54E+12 1.66E−10
5 14.029 1.75E+18 5.44E−16 34.659 6.63E+13 1.44E−11
4 10.477 1.13E+20 1.40E−17 31.490 2.79E+14 5.64E−12
3 7.018 3.41E+22 1.21E−19 28.090 1.55E+15 2.66E−12
2 3.515 6.51E+26 1.01E−23 25.633 6.12E+15 1.07E−12
1 1.754 1.31E+31 3.52E−27 24.565 1.16E+16 3.97E−12
Ʃ=1.01E−06 Ʃ=1.20E−06
Fig. 11. Histogram with constant amplitude stress blocks calculated from the experimental strain data using the Rainflow counting method.

−13 E100 3:
C ¼ 5:21 10
EET
where, E100 is the elastic modulus at 100 °C and EET is the elastic modulus at elevated temperature, for da/dN in mm/cycle and ΔK in N/
mm3/2. For the present case, a value of C equal to 5.21 × 10−13 was considered.
Fig. 12. Matlab routine procedure for fatigue crack growth calculations.
• Fatigue crack propagation threshold: ΔKth = 63 N/mm3/2;

• Critical stress intensity factor: KIc =3000 N/mm3/2 [20];
• For each of the eight more relevant block of stress range (Table 6, from block no. 6 to block no. 13), crack growth, Δa, was cal-
culated cycle by cycle (ΔN= 1Δ cycle) and stress intensity factor range values, ΔK, had to be updated. This procedure was re-
peated until critical crack length or critical stress intensity was reached (Fig. 13);
• Two different simulations were carried out, namely assuming the presence of a crack in the stress concentration region near the
flange and the presence of a crack where this effect is not present.
Crack growth in the transition region between the rod and the flange (stress concentration zone) can be seen in Fig. 13a, cor-
responding to approximately 1340 h of milling (3 × 106 cycles), which makes about three months until the crack would reach a
depth that will cause the failure of the pull rod. Instead, the presence of the same initial defect in an area far away from the stress
concentration region of the rod will result in the fracture of the pull rod after 10,200 h of operation, which correspond to circa
twenty-four months of functioning (Fig. 13b).
5. Conclusions
This work presents a fatigue failure analysis of a pull rod from an ATOX raw mill used in the cement production process. A
description and simulation of the working principles of the mechanical system under study was carried out during this research
and a fatigue damage assessment was performed according to the stress-life method and also assuming crack growth propagation.
The following conclusions could be drawn:
• Cinematic analysis carried out and presented in section two of this paper allowed to conclude that the mechanical system was
theoretically designed in order to the pull rods be actuated only by axial forces. This condition cannot be succeeded if the set of
plain bearings do not perform their function properly. In fact, the proper functioning of the plain bearings depends on some
Fig. 13. Crack growth (a vs. N) through the cross section of the pull rod due to the application of in-service loading (Table 7, Fig. 11). a) In the transition region
between the flange and the pull rod (kt = 2, a0 = 0.15 mm); b) in a region far away from the stress concentration influence.
external factors, namely, on the correct positioning of the roller joints, on the right position of the base of the hydraulic cylinder
in relation to the metallic shoe, and on the elastic condition of the elastomers placed inside the buffer. The failure of any com-
ponents mention above, particularly noticed on pull rod no. 1, may cause a bending stress state, thus reducing the useful life of
the pull rod. Hence, periodical visual inspection and maintenance of the mechanical system is essential to assure the proper
functioning of the ATOX raw mill;
• From the observation of fracture surface, fatigue crack propagation, typical of an axial tensile-compression stress state, could be
clearly perceived. The moderate area occupied by the fatigue region also denotes the application to the pull rod of high to mod-
erate nominal stresses and fatigue crack initiation point was located at the outer cylindrical surface of the rod and started from
an unnotched surface;
• Failure reports did not indicate the failure of any bolts that connects the pull rod's flange to the joint head, which, in case of
occurrence, could indicates the application of bending or overloads to the pull rod. In addition, the analyses carried out to
the experimental strain data acquired allowed to realize the absence of bending stress state or overloads;
• According to the stress-life approach, the current design solution should provide a useful life of 50 to 57 years to the pull rod.
However, it should be noted that these results were obtained based on the assumption that the pull rods are only subjected to
axial loading requests and there is not any superficial or internal defects on the pull rod;
• Crack growth analysis demonstrated that the presence of an initial crack depth size of 0.15 mm on the surface area, in the stress
concentration region, is extremely harsh to the life span of the rod, which will reach a critical condition in a much shorter pe-
riod of time, namely three months, for a critical damage accumulation value equal to 0.3; instead, in case of crack propagation
outside of that zone, twenty-seven months till fracture could be expected. Additionally, the harsh environmental conditions
where the pull rods operate could induce corrosion pits at the superficial cylindrical surfaces of the pull rod, resulting on poten-
tial nucleation points that could results in premature failure as indicated by the results obtained, and periodical visual inspection
will be required to avoid this happening.
Acknowledgements
The authors would like to thank the Portuguese Foundation for Science and Technology through project ref. UID/EMS/00667/
2013.
References
[1] FLSmith, ATOX Raw Mill, http://www.flsmidth.com/~/media/Brochures/Brochures%20for%20raw%20grinding%20and%20silos/ATOX_Raw_Mill.pdf 2011

(accessed 12.07.16).
[2] L.Group, http://www.lucefin.com/wp-content/files_mf/0934crnimo6.pdf (accessed 13.09.16).
[3] R. Branco, et al., Low-cycle fatigue behaviour of 34CrNiMo6 high strength steel, Theor. Appl. Fract. Mech. 58 (2012) 28–34.
[4] DIN EN 10083-3, Steels for quenching and tempering, http://members.marticonet.sk/jkuba/normy/EN-10083-3-Technical-delivery-conditions-for-alloy-steels.
pdf 2007 (accessed 20.07.16).
[5] E.L. Menachem, P. Weiss, Fatigue of metals – what the designer needs? Int. J. Fatigue 84 (2015) 80–90.
[6] J. Schijve, Fatigue of structures and materials in the 20th century and the state of the art, Int. J. Fatigue 25 (2003) 679–702.
[7] B. Karakkunnummal, A case study of a pull rod failure in ATOX raw grinding mill, Int. J. Mater. Sci. Eng. 3 (2015) 90–103.
[8] K. Hoffmann, An introduction to stress analysis and transducer design using strain gauges, HBM, http://www.kk-group.ru/help/Strain_Gauge_Measurements_
Book_2012_01.pdf 2012 (acessed 19.12.16).
[9] MICRO-MEASUREMENTS, Tech-Note TN-515, Strain gage rosettes: selection. Application and data reduction, http://www.vishaypg.com/docs/11065/tn-515.pdf
(acessed 19.12.16).
[10] Sean A. McKelvey, et al., Stress-based uniaxial fatigue analysis using methods described in FKM-Guideline, J. Fail. Anal. Preven. 12 (2012) 445–484.
[11] Gabriel Marsh, et al., Review and application of Rainflow residue processing techniques for accurate fatigue damage estimation, Int. J. Fatigue 82 (2016) 757–765.
[12] N.E. Dowling, Fatigue at notches and the local strain and fracture mechanics approaches, Fracture Mechanics, ASTM 1979, pp. 247–273.
[13] T. Wehner, A. Fatemi, Effects of mean stress on fatigue behaviour of a hardened carbon steel, Int. J. Fatigue 13 (1991) 241–248.
[14] C.S. Wiesner, et al., Engineering critical analyses to BS 7910 - the UK guide on methods for assessing the acceptability of flaws in metallic structures, Int. J. Press.
Vessel. Pip. 77 (2000) 883–893.
[15] BS 7910, Guide on Methods for Assessing the Acceptability of Flaws in Structures – Draft for Public Comment (PD6493), 1997.
[16] ASTM E 1049-85, Standard Practices for Cycle Counting in Fatigue Analysis, ASTM International, 2011.
[17] W. Pilkey, Peterson's Stress Concentration Factors, second ed. John Wiley-Interscience, 1997.
[18] Richard Budynas, J. Keith Nisbett, Shigley's Mechanical Engineering Design, tenth ed. McGraw-Hill Series in Mechanical Engineering, 2015.
[19] Yung-Li Lee, et al., Fatigue Testing and Analysis, Theory and Practice, Butterworth-Heinemann, Elsevier, 2005.
[20] J. Schijve, Fatigue of Structures and Materials, second ed. Springer, 2009.

J Engfailanal 2016 12 023

Uploaded by

Copyright:

Available Formats

You might also like

J Engfailanal 2016 12 023

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

J Engfailanal 2016 12 023

Uploaded by

Copyright:

Available Formats

EFA-03038; No of Pages 16

Engineering Failure Analysis xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Engineering Failure Analysis

C [%] Mn [%] Si [%] Cr [%] Ni [%] Mo [%]

EN 10083-3 [4] 0.30–0.38 0.50–0.80 Max 0.40 1.30–1.70 1.30–1.70 0.15–0.30

Depth from the treated surface [mm] 25 75

2. Theory and calculation

2.1. Cinematic of the mechanical system

Tensile yield strength, SYS (MPa) 700

2.2. Strain gages – calculation of principal stresses

Fig. 5. a) National Instruments NI PXI-1052 device; b) Microlog CMVA 65 device.

3. Materials and methods

3.1. Experimental strain data

3.2. Fatigue analysis

3.3. Fatigue crack propagation analysis

and for fatigue assessments the following equation applies:

ðYΔSÞ ¼ M f w ½ktm M km Mm ΔSm þ ktb M kb M b ðΔSb þ ðkm −1ÞΔSm Þ ð7Þ

ðYΔSÞ ¼ M f w ½ktm M km Mm ΔSm ð8Þ

Mm ¼ g 0:752 þ 2:02 þ 0:37 1− sin ð9Þ

1:84 h πa πai0:5

4. Results and discussion

4.1. Strain data

−F 49895; 175 9:81

13 70.80 69.00 −0.05 138.00 124.5

Parameter North American

Cσ, E,FKM 0.400

Mσ ¼ am St:u þ bM ðam ¼ 0:00035; bM ¼ −0:1 for steel ½10Þ ð21Þ

4.3. Fatigue crack growth

Block North American solution FKM-Guideline

Sar [MPa]) N (cycles) Damage Sar [MPa]) N (cycles) Damage

13 69.356 2.23E+08 5.58E−07 80.721 2.06E+08 6.04E−07

• Fatigue crack propagation threshold: ΔKth = 63 N/mm3/2;

[1] FLSmith, ATOX Raw Mill, http://www.flsmidth.com/~/media/Brochures/Brochures%20for%20raw%20grinding%20and%20silos/ATOX_Raw_Mill.pdf 2011

You might also like

1:84 h πa πai0:5