Engineering Fracture Mechanics 181 (2017) 1–6

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

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

A critical insight on the use of external load cells for fatigue

tests in pressurized systems
R.R.A. Garcia a, A.S. de Lima a, Oscar Rosa Mattos a,⇑, Carlos Augusto b, Miguel Maia b
a
LNDC/EP/COPPE/UFRJ, Av. Pedro Calmon s/n, Cidade Universitária, Rio de Janeiro, RJ 21945970, Brazil
b
Petrogal Brasil S/A, Av. Pres. Vargas, 1001 – Centro, Rio de Janeiro, RJ 20071-004, Brazil

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

Article history: For fatigue studies, load cells are generally used to measure the cyclic loading applied to
Received 16 April 2017 the specimen. However, the literature reports that an internal force arises when fatigue
Received in revised form 19 June 2017 tests are performed inside pressurized vessels, acting directly on the vessel shaft.
Accepted 22 June 2017
Another component, the friction force, arises due to the sealing system. These two forces
Available online 23 June 2017
impose substantial differences between the values measured by external load cells and
the true loading applied on the specimen. Faced with this problem, this study aims to ana-
Keywords:
lyze the influence of friction in pressurized testing of corrosion-fatigue, the sealing system
Crack tip opening displacement
Test standards
of which has been specially designed with reduced stick-slip effect to avoid nonlinearities
Fatigue crack growth in the cyclic load. A new methodology based on a strain gauges system installed on the CT
specimen was developed and tested, solving the problems of cyclic load and crack length
measurements for corrosion-fatigue tests under high pressure. The innovation of this sys-
tem is that the specimen itself simultaneously acts as a crack length transducer and a load
cell. Such a system was properly designed to be used under high pressures and is resistant
in aggressive environments. Therefore, it is possible to perform corrosion fatigue tests,
obtaining Fatigue Crack Growth Rates (FCGR) without all the measurement problems
caused when a load cell is used. In addition, the resulting errors obtained when using a
classical load cell in the tests can also be measured by the present experimental setup.
In this paper errors were verified in corrosion-fatigue tests using a conventional load cell.
These errors could be greater than 700 percent, not meeting the requirements of ASTM E4
and E647 standards. Consequently, the crack length measurements do not agree with the
values obtained by an optical microscope, even when tunneling does not occur.
Ó 2017 Published by Elsevier Ltd.

1. Introduction

In 1979, Deans designed a system using a single strain gauge to monitor crack lengths called BFS, Back Face Strain [1].
Recently, Newman improved the accuracy and increased the range of crack length measurement using BFS, showing a suit-
able methodology for low-pressure fatigue tests [2–6]. This improvement was included in the ASTM E647 [7]. However, its
application is not suitable for tests under high pressure, since it depends on an external load cell for load measurements.
Indeed, problems with measuring loads applied to test specimens located inside pressurized vessels using conventional load
cells have been reported since 1979 [1] and recently emphasized [8]. Consequently, for pressured systems the whole

⇑ Corresponding author.
E-mail address: omattos@metalmat.ufrj.br (O.R. Mattos).

http://dx.doi.org/10.1016/j.engfracmech.2017.06.015
0013-7944/Ó 2017 Published by Elsevier Ltd.
2 R.R.A. Garcia et al. / Engineering Fracture Mechanics 181 (2017) 1–6

propagation curve, da/dN vs. DK, will not be accurate if an external load cell is used. In order to reduce inaccuracy caused by
this measurement error, it is commonly proposed in practical works a mathematical compensation, or a safety factor. How-
ever, the use of these factors could be excessively conservative or even unsafe, because a sealing system can cause at the
same time systematic and random errors.
Recently, the LNDC (Non-destructive Testing, Corrosion and Welding Laboratory) developed an integrated system of crack
length and load measurements simultaneously, which proved to be adequate to meet the requirements of the standard [7–
9]. This strain-gauge system transforms the test specimen itself into a load cell and a crack length transducer. As a result, the
use of an external load cell can be disposed of in this system, named here as EFS, Extreme Face Strain Gauges [1,8–10]. In this
paper, experimental comparisons between the loading measured by the new system, better detailed in [8], and the external
load cell were performed [9]. For this, two different kinds of tests were executed: one with and one without a pressurized
vessel. The results showed that in the second case there is a good agreement between the measures; however, in the first
case the difference between the results can reach values much higher than the requirements of the ASTM E647, which is
two percent. This difference is associated with the error caused mainly by friction force, which increases with the higher
pressures and it becomes more important when the load applied under the specimen is decreasing in cycle [11–13]. The
results obtained in this work showed that the load error can be higher than 700% in a fatigue test under high pressure.

2. Experimental

This work used the compact tension (CT) specimen, with a 60 mm width (W) and 10 mm thickness (B). Three strain
gauges were assembled on it: one uniaxial type placed on the back of the sample, traditionally used for crack length mea-
surement and called BFS, and the two others were biaxial type (XY), forming the so-called EFS system. Both of them were
3 mm length and had resistance of 350 X. The two biaxial strain gauges are useful to eliminate discrepancies in load mea-
surement, such as those caused by a slight misalignment, see Fig. 1.
After the strain gauge system was installed in the specimen, it was coated maintaining free the region near the fatigue
crack. It is worth noting that it is possible to use various types of coatings to protect strain gauges. However, it is important
to consider the following aspects: high corrosion resistance; chemical inertia; relatively lower young’s modulus; and good
resistance to high pressures. In the present work, a polymeric coating with these properties was used. Shore A - DIN 53505
and Tensile strength at 190 bar, 130 lb per linear inch [6] was chosen [8]. If this coating were very rigid, it would cause prob-
lems in the strain gauges measurements, since it would prevent the strain gauge from deforming correctly.
Load versus deformation calibration curves was obtained under two different conditions: without and with the coating. In
both conditions, the crack length measured by the strain gauges showed a very good agreement compared with the real
value, obtained by optical microscopy, which means that there was no significant restriction of the deformation. In addition,
if the coating did not protect the strain gauges well, there would be an incorrect measurement of the crack size during the
test, since corrosion would modify the electrical resistance. After the specimen preparation with strain gauges and coating, a
pre-cracking was made and then the test started with or without the vessel developed by LNDC. The temperature was always
room temperature (23 ± 2 °C). However, the strain gauges used in the present work allow temperatures up to 120 °C.

Fig. 1. Positioning of the strain gauges on a C(T) specimen.

 R.R.A. Garcia et al. / Engineering Fracture Mechanics 181 (2017) 1–6 3

A compensation system was specially designed to eliminate or, at least, attenuate the ejection force that comes out on the
shaft when the vessel is pressurized. Its principle of operation is based on the difference of area between the compensation
and the test chambers, Fig. 2 [8]. The compensation chamber has a piston diameter approximately 10 times greater than the
diameter of the hole for the shaft crossing the chamber test. This means, for example, that the ejection force on the shaft, due
to a pressure in the test chamber of 600 bar, can be eliminated by a pressure of approximately 6 bar in the compensation
chamber. The pressure in this chamber is guaranteed during the test by means of a 80 L reservoir connected to it by means
of four ¾-inch hoses. In this way, this complete system of the pressure vessel remains suspended during the test, eliminating
or at least reducing the effects of the pressure in the test chamber on the shaft.
Fig. 3 shows the vessel mounted on the hydraulic testing machine, which was developed by the LNDC and has a capacity
of up to 125 kN.
As it is well known, in any current dynamic machine operating under high pressures, the friction factor is what makes it
possible to seal the vessel. Dimensional and operational aspects contribute directly to the normal force, for example, when
the pressure is raised the o-ring is crushed, increasing the friction force, since the normal is larger. Furthermore, the friction
coefficient also increases with the pressure, because the contact area between the o-ring and the vessel expands when the
internal pressure is raised. Considering this problem, the entire vessel sealing and compensation systems used in this work
have been optimized to reduce the friction force for pressures up to 600 bar, as well as the stick-slip effect. A homemade
software written in LabViewÒ embedded in a real-time/FPGA industrial controller, Compact-RioÒ, manufactured and dis-
tributed by National InstrumentsÓ recorded the data acquisition of load, crack length, stress intensity factor, strain gauge
measurements and number of cycles.

Fig. 2. Force exerted on internal devices due to internal pressure.

Fig. 3. Homemade hydraulic testing machine and vessel designed and manufactured for corrosion-fatigue tests under high pressure.
4 R.R.A. Garcia et al. / Engineering Fracture Mechanics 181 (2017) 1–6

Using this described system, two sequences of fatigue tests were performed, the first without the pressure vessel and the
other with it. The test results were discussed as follows.

3. Results and discussion

In order to systematically evaluate the results of the tests, three variables were defined: PEFS as the value of the load mea-
sured by the EFS system, PLC as the load value obtained directly by a conventional load cell and the friction force – FF. The
difference between the PEFS and PLC values was considered as the FF, with the intention to simplify the problem. Using Eqs. (1)
and (2) the PEFS was obtained for 0:20 6 W a
6 0:65 [9].

a
¼ C 0 þ C 1 M þ C 2 M2 þ C 3 M3 þ C 4 M4 þ C 5 M5 ð1Þ
W

2
1  Wa  B  W  E  j j
PEFS ¼


ð2Þ
a 2 a 3 a 4
1:41  1:462 Wa þ 20:45 W
 26:83 W
þ 11:45 W

where
 ;
m ¼ EFS  and EFS is the strain measured respectively by BFS and EFS system; M ¼ 1
1 ; a is the crack length; E is the
m2 þ1
Young modulus; B ¼ 10 mm; W ¼ 60 mm; C 0 ¼ 1:1845; C 1 ¼ 21:7500; C 2 ¼ 91:8220; C 3 ¼ 179:1; C 4 ¼ 169:79; and
C 5 ¼ 60:5590.
Firstly, three tests (T1, T2, T3) were performed without the vessel, aiming to evaluate the accuracy of the EFS system com-
pared to the load cell measures. These results are presented in Fig. 4 showing an excellent agreement between them, which
ensures that the system developed by LNDC is able to measure the load applied directly on the specimen. For all these three
tests, the applied conditions were R ¼ P min =P max = 0.5, frequency f ¼ 0:25 Hz and loading ranges DPLC ¼ P maxLC  PminLC ¼ 4000,
5000 and 8000 N, respectively, see table 1. After these first tests, another sequence of eight more tests (T4–T11) was per-

Fig. 4. Comparison of the results obtained by the load cell, EFS and friction force in tests using a pressurized vessel.
 R.R.A. Garcia et al. / Engineering Fracture Mechanics 181 (2017) 1–6 5

Table 1
Test conditions with the pressure vessel.

Test number Vessel use DP LC (N) Pressure (barg) Compensation system Average dissipative force (kN)
T1 4000 0 –
T2 5000 0 –
T3 8000 0 –
T4 8000 0 1
T5 5000 0 1.5
T6 8000 180 4
T7 8000 180 2.2
T8 5000 180 6
T9 5000 180 4
T10 4000 180 6
T11 4000 180 4

Table 2
Errors in the tests.

1st bath DPLC DPEFS % Error 2nd bath DPLC DPEFS % Error 3rd bath DPLC DPEFS % Error
Without vessel T1 3.98 4.01 1% T12 4.01 3.97 1% T23 3.99 4.03 1%
T2 4.98 5.00 0% T13 5.01 4.94 1% T24 5.02 5.00 0%
T3 7.98 8.08 1% T14 8.02 7.98 1% T25 7.99 8.09 1%
With vessel T4 7.94 6.63 20% T15 8.01 6.62 21% T26 8.02 6.10 31%
T5 5.02 3.46 45% T16 5.01 3.62 38% T27 5.02 3.15 59%
T6 8 3.90 105% T17 8.01 3.95 103% T28 7.97 4.35 83%
T7 7.99 5.43 47% T18 7.98 5.68 40% T29 8.00 5.03 59%
T8 5 0.76 558% T19 4.99 0.68 634% T30 4.99 0.58 760%
T9 4.96 2.21 124% T20 4.98 1.97 153% T31 5.00 2.38 110%
T10 3.96 0.49 708% T21 3.99 0.62 544% T32 3.98 0.41 871%
T11 4.06 0.88 361% T22 4.00 1.02 292% T33 4.00 0.83 382%

formed. This time a pressurized vessel was used with R ¼ 0:5 and f ¼ 0:25 Hz. Table 1 correlates all the test conditions and
the respective results are presented in Fig. 4.
In Fig. 4 it is possible to see that the load measured by EFS (red1 circle) and the load cell (black square) without the vessel
are very close (see the charts T1, T2 and T3). However, when a vessel is used (T4–T11), the difference between the Load Cell and
EFS curves is very significant and is considered as the FF (blue triangle). It is possible to verify that in the test T4, for example, a
load range measured by load cell was applied, DPLC, of approximately 8.00 kN. However, the load value really applied on the
specimen was lower than this value, i.e., the specimen received the actual load range (DPEFS) determined by the difference
between 14.00 and 7.37, that is 6.63 kN. Therefore, in the test T4 an error of approximately 20% was found, see Eq. (3). This error
is far from the requirements of ASTM E647, which tolerates a maximum of 2 percent.

   
apparent v alue  true v alue DPLC  DPEFS 
% Error ¼  ¼  ð3Þ
true v alue DPEFS
Similar problems were found in all other pressurized tests, and in fact, these errors are larger when the pressure increases
and the load cell measurement decreases. It is possible to see, for example, in test T10 that an apparent load range of
DPLC = 4.00 kN was applied on the specimen, but the true load range was DPEFS = 1.15–0.66 = 0.49 kN. Therefore, the error
in this case was approximately 716%, which is excessively high and unacceptable, and it will be higher for higher pressure,
as is necessary in the Brazilian pre-salt, for example [14].
A triplicate of tests was performed, intending to compare tests under exactly the same condition, i.e., T1, T12 and T23; T2,
T13 and T24; and so on. Their error results were compared and they can be seen in Table 2, for the first batch (T1–T11), the
second (T12–T22) and the third (T23–T33).
It is very easy to see, in all fatigue tests using the vessel, T4–T11, that the actual loads directly applied on the specimen are
different from the ones measured by a load cell. A corrosion fatigue testing system generally contains a sealant subject to the
influence of the following parameters: internal pressure, composition of the solution containing gas and corrosive liquid, the
moving of the shaft, lubrication of the sealing system, and so on. These parameters can cause considerably systematic and
random errors, so it is not possible to get satisfactory repeatability and reproducibility, as can be seen by the results pre-
sented in Table 2. That is the main support to affirm that it is not possible to use correction factor to handle this problem
as is proposed in the literature.

1
For interpretation of color in Fig. 4, the reader is referred to the web version of this article.
6 R.R.A. Garcia et al. / Engineering Fracture Mechanics 181 (2017) 1–6

In addition, by the results shown in Fig. 4, it is possible to see that the friction force during a cycle is not constant, getting
the upper value in the maximum load measured by the load cell and, in contrast, the opposite also occurs. If the friction force
were constant during the cycle, the problem would be trivial for whatever the value of the frictional force, since for any DPLC
applied, it would have the same value received by the specimen, i.e., DPLC = DPEFC. The only difference would be the value of
R that would tend to be lower. However, what really occurs is that the intensity of the friction force in P max is higher than in
P min, consequently, the load range directly applied on the specimen is always lower than the value measured by the load
cell. It is worth noting that the friction force regarding the sealing system detailed in this paper was optimized to be close to
the minimum value required, following the requirements of [12].

4. Conclusions

The EFS system is able to measure loads with or without a pressurized vessel in fatigue testing. Experimental results in
this work show that, in non-pressurized environments, its measurements are equivalent to the ones obtained with a load
cell, transducer traditionally used in fatigue testing. In addition, as the EFS system transforms the specimen into a transducer
itself, effects of friction or pressure do not bring changes to its load measurements. The use of an external load cell to mea-
sure the load applied to the test specimen in pressurized fatigue tests does not meet the requirements of ASTM E4 and E647
standards. This is even more expressive when the pressure inside the vessel is increased and a load balancing system is not
used. Therefore, it is crucial to use the EFS technique for load measurements in pressurized tests.

Acknowledgements

The authors thank PETROGAL, CNPq, CAPES and FAPERJ.

References

[1] Deans WF, Richards CE. A simple and sensitive method for monitoring crack and load in compact fracture mechanics specimens using strain gages. J
Test Eval 1979;7:147–54.
[2] Newman JC, Yamada Y, James Ma. Back-face strain compliance relation for compact specimens for wide range in crack lengths. Eng Fract Mech
2011;78:2707–11.
[3] Shaw WJD, Zhao W. Back face strain calibration for crack length measurements. J Test Eval 1994;22:512–6.
[4] Riddell WT, Piascik RS. A back face strain compliance expression for the compact tension specimen, NASA TM; 1998
[5] Maxwell DC. Strain based compliance method for determining crack length for a C(T) specimen, AFWAL-TR-87-4046; 1987.
[6] Garcia RRA, Beserra A, Dias DP, Assis KS, Mattos OR. Back-face strain compliance relation for SEN (B) specimens for wide range in crack lengths. NACE;
2015.
[7] ASTM, E-647. Standard test method for fatigue crack growth rates (West Conshohocken. PA: ASTM; 2008.
[8] Garcia RRA, Caldas FCC, Mattos OR. Corrosion fatigue tests using strain gauges for measuring load and crack length. Corrosion 2016;72(12):1547–55.
[9] ASTM, E4. Standard practices for force verification of testing machines; 2014.
[10] ASTM, E1820. Measurement of fracture toughness 1; 2001.
[11] Ludema KC. Friction, wear, lubrication: a textbook in tribology. New York: CRC Press; 1996.
[12] Trelleborg. <http://www.trelleborg.com> [accessed July 21, 2014].
[13] ASM. ASM handbook – volume 19 – fatigue and fracture. Ohio, ASM International; 1996.
[14] Henriques CCD, Joia C-JB, Guedes F, Baptista IP. Material selection for Brazilian presalt fields. Offshore technology conference; 2012.