Metrol. Meas. Syst., Vol. XXI (2014), No. 3, pp. 553–560.

METROLOGY AND MEASUREMENT SYSTEMS

Index 330930, ISSN 0860-8229
www.metrology.pg.gda.pl

ESTIMATING THE UNCERTAINTY OF TENSILE STRENGTH MEASUREMENT

FOR A PHOTOCURED MATERIAL PRODUCED BY ADDITIVE
MANUFACTURING

Stanisław Adamczak1), Jerzy Bochnia1), Bożena Kaczmarska2)

Kielce University of Technology, Al. 1000-lecia P. P. 7, 25-314 Kielce, Poland,
1) Department of Machine Technology and Metrology (jbochnia@tu.kielce.pl, * adamj@tu.kielce.pl, +48 22 432 7721)
2) Department of Production Engineering (bozenaka@tu.kielce.pl )

Abstract
The aim of this study was to estimate the measurement uncertainty for a material produced by additive
manufacturing. The material investigated was FullCure 720 photocured resin, which was applied to fabricate
tensile specimens with a Connex 350 3D printer based on PolyJet technology. The tensile strength of the
specimens established through static tensile testing was used to determine the measurement uncertainty. There is
a need for extensive research into the performance of model materials obtained via 3D printing as they have not
been studied sufficiently like metal alloys or plastics, the most common structural materials. In this analysis, the
measurement uncertainty was estimated using a larger number of samples than usual, i.e., thirty instead of
typical ten. The results can be very useful to engineers who design models and finished products using this
material. The investigations also show how wide the scatter of results is.

Keywords: static tensile test, measurement error, measurement uncertainty.

© 2014 Polish Academy of Sciences. All rights reserved

1. Introduction

Additive manufacturing, originally known as Rapid Prototyping, is becoming increasingly

popular, with applications ranging from industry through design and architecture to medicine.
The technology is now used to produce not only models and prototypes but also finished and
semi-finished products. The present state of knowledge and future potential of additive
manufacturing have been discussed, for instance, by Campbell [1]. He reports on the
advancement of additive manufacturing materials, analyzes design possibilities and overviews
industrial applications.
The advances in additive manufacturing processes and materials, being concurrent with the
development of new materials characterized by better physical and mechanical properties,
imply that extensive research is required to fully understand the behaviour of materials used
for 3D printing. The key findings on the subject are presented in Ref. [2], which explains how
environmental conditions, aging and build orientations, i.e., the arrangement of models on the
build tray, affect the mechanical properties of type-1 specimens produced by
stereolithography according to ASTM D638 [3]. The materials manufactured via
stereolithography showed anisotropy and their mechanical properties decreased when they
were exposed to environmental conditions. Another work on the mechanical properties of
materials fabricated by stereolithography [4] is concerned with the effects of layer thickness.
The paper suggests that the thinner the layer, the greater the maximum load to be carried by a
specimen.
There are no papers reporting on the measurement uncertainty related to additive
manufacturing materials. Even manufacturers’ catalogues miss this information; they only
_____________________________________________________________________________________________________________________________________________________________________________________

Article history: received on Mar. 04, 2014; accepted on Jul. 19, 2014; available online on Sep. 15, 2014; DOI: 10.2478/mms-2014-0047.
 St. Adamczak, J. Bochnia, B. Kaczmarska: ESTIMATING THE UNCERTAINTY OF TENSILE STRENGTH MEASUREMENT …

provide the nominal value or range of parameters characterizing a material property, for
example, tensile strength. Studies of the mechanical properties of polymers produced by
additive manufacturing have been conducted using standards applicable to plastics. Most
American companies and research institutes produce specimens and perform tensile testing in
accordance with ASTM D638 [4]. In Europe, the ISO 527 standard is followed, with tensile
specimens being produced in conformity with ISO 527-1 [5]. These standards recommend
that results should be written as arithmetic means and, if necessary, as standard deviations.
The estimation of measurement uncertainty for metallic materials is performed following
Annex J to the ISO 6892-1 standard [6]. Section 23.3 of ISO 527-1 [5] concerning
measurement results suggests that the evaluation of measurement uncertainty should not be
combined with the evaluation of measurement results to determine the agreement with
product specifications unless otherwise agreed with the buyer. In Section 23.1, however, we
read that the analysis of measurement uncertainty is useful in order to identify the main
sources of discrepancy between measurement results. More details on the evaluation of
uncertainty of stress and strain measurement during a static tensile test can be found in [7].
This study attempts to estimate measurement uncertainty for a photocured resin, FullCure
720, taking into account the parameters established through static tensile tests. The specimen
preparation, the static tensile testing and the analysis of results are described in the next
sections.

2. Materials and methods

The specimens were made of photocured resin, FullCure 720, by means of an Objet
Connex 350 3D printer based on PolyJet technology (http://objet.com/3d-printing-materials).
The specimens were produced in accordance with ASTM D638, with dimensions being the
same as those of type-I tensile bars used for testing plastics [3] width of the narrow section,
13 ±0.02 mm; length of the narrow section, 57 ±0.02 mm; overall width, 19 ±0.025 mm;
overall length, 165 ± no max; gauge length, 50 ±0.01 mm; distance between the grips,
115 ±0.02 mm; fillet radius, 76 ±0.04 mm; and thickness, 4 ±0.4 mm.
The study was conducted according to the procedure presented in Fig. 1.
A solid 3D model of a specimen was created in a CAD program and saved as an .stl file.
The triangulation parameters to be exported include: resolution – adjusted, deviation – 0.016
mm tolerance, and angle – 5° tolerance. It is important that the values of the triangulation
parameters should not be too low to ensure rounded edges (in this case, the fillet radius, R);
they should not be too high, either, so as not to increase the volume of the file (.stl).
Subsequently, the specimen models were virtually placed on the build tray of the Connex350
printer using the Objet Studio program. The specimens were produced in the Glossy mode to
ensure a smooth surface. Figure 2 shows the virtual arrangement of the specimen models over
the build tray in the Objet Studio program.

554
 Metrol. Meas. Syst., Vol. XXI (2014), No. 3, pp. 553–560.

START

CREATE A SOLID 3D MODEL OF THE SPECIMENS

AND SAVE IT AS AN .stl FILE

IMPORT THE .stl FILE INTO THE OBJET STUDIO PROGRAM

AND SET THE PRINTING PARAMETERS

PRINT THE SPECIMEN MODELS

WITH THE CONNEX 350

PERFORM STATIC TENSILE TESTS

ANALYZE THE RESULTS

FINISH

Fig. 1. Test procedure.

Fig. 2. Specimen models on the build tray in the Objet Studio program.
After the printed specimens and the support material were removed off the build tray, the
specimens were prepared for static tensile tests. The testing was performed using an Ispekt
mini universal testing machine [8]. The test speed was set at 5 mm/min in the LabMaster
program [9] that the Inspekt mini universal testing machine is equipped with.

555
 St. Adamczak, J. Bochnia, B. Kaczmarska: ESTIMATING THE UNCERTAINTY OF TENSILE STRENGTH MEASUREMENT …

3. Results and discussion

The example cumulative plot in Fig. 3 shows changes in the load acting on the specimens
in the function of displacement drawn by the computer connected to the universal testing
machine.

Fig. 3. Load vs displacement for FullCure 720.

Table 1 shows values of the maximum tensile load obtained for each test.

Table 1. Values of the maximum tensile load.

Test/specimen Maximum load (Fmi – Fm )2 Test/specimen Maximum load (Fmi – Fm )2

number Fmi [N] [N2] number Fmi [N] [N2]
Test 1 2396 4542 Test 16 2438 11883
Test 2 2253 5716 Test 17 2428 9818
Test 3 2292 1340 Test 18 2166 26373
Test 4 2165 26810 Test 19 2331 6
Test 5 2406 6045 Test 20 2262 4466
Test 6 2335 41 Test 21 2413 7188
Test 7 2453 15406 Test 23 2373 1959
Test 8 2237 8301 Test 25 2276 2809
Test 9 2274 3003 Test 24 2333 19
Test 10 2452 15203 Test 26 2391 3942
Test 11 2391 3890 Test 27 2411 6864
Test 12 2356 747 Test 28 2189 19577
Test 13 2384 3106 Test 29 2281 2306
Test 14 2288 1684 Test 30 2271 3346
Test 15 2354 645 Test 31 2260 4704
X 2329
S 201741

Specimen No 22 was excluded from the calculations, because it failed before the maximum
tensile load was reached.
The measurement uncertainty was determined for each quantity using the following
formulae [10]:
- type-B evaluation was applied to calculate the standard uncertainty

556
 Metrol. Meas. Syst., Vol. XXI (2014), No. 3, pp. 553–560.

a ,
uB = (1)
n
where: a – half of the width of the interval containing a boundary error,
n – number of measurements;
- type-A evaluation was applied to establish the standard uncertainty
n
1
uA = å
n(n - 1) i =1
( xi - x )2 , (2)

where: n – number of measurements,

x – arithmetic mean of all the quantities measured,
- the combined standard uncertainty was estimated on the basis of a series of measurement
results obtained for each quantity
2
æ ¶y ö 2
n
u y = å çç ÷÷ ui , (3)
i =1 è ¶xi ø

where: ui – standard measurement uncertainty obtained for the input data calculated using the
type-A or type-B evaluation.

The width of the specimens, b0, was measured with an accuracy of 0.05 mm, whereas the
thickness, a0, was established with an accuracy of 0.01 mm, as recommended by the ISO 527-
1 2012 standard. The average thickness and width were:
– a 0 = 3,97 mm and
– b0 = 13 mm, respectively.
The standard uncertainty obtained by type-B evaluation using formula (1) was:
– uaB = 0.0009 mm for the thickness of the specimens tested,
– ubB = 0.0046 mm for the width of the specimens tested.
The standard uncertainty calculated by means of formula (2) using type-A evaluation was:
– uaA = 0.0021 mm for the thickness of the specimens tested,
– ubA = 0.02 mm for the width of the specimens tested.
When thickness and width were measured, the predominant uncertainty was calculated using
type-A evaluation. Further calculations were based on these values.
The uncertainty for the average maximum load measured Fm was calculated from formula
(2) and the data included in Table 1 as u Fm = 15.2 N.
The uncertainty for the indirectly measured tensile strength Rm was calculated from
formula (3) using the following transformation:
2 2 2
æ ¶R ö æ ¶R ö 2 æ ¶Rm ö 2
u Rm = çç m ÷÷ u F2m + çç m ÷÷ uaA + çç ÷÷ ubA , (4)
¶F
è mø è ¶a0 ø è ¶b0 ø
and after more transformations, we had:
2 2 2
æ 1 ö 2 æ - Fm ö 2 æ - Fm ö 2
u Rm = çç ÷÷ u Fm + çç 2 ÷÷ uaA + çç ÷ ubA ,
2 ÷
(5)
è a0b0 ø è a0 b0 ø è a0 b0 ø
where: Fm = 2329 N – average maximum tensile load on the basis of Table 1,
u Fm = 15.2 N – uncertainty of the average maximum tensile load,

557
 St. Adamczak, J. Bochnia, B. Kaczmarska: ESTIMATING THE UNCERTAINTY OF TENSILE STRENGTH MEASUREMENT …

a 0 = 3.97 mm – average thickness of the specimens,

uaA = 0.0021 mm – uncertainty of the average thickness of the specimens,
b0 = 13 mm – average width of the specimens,
ubA = 0.02 mm – uncertainty of the average width of the specimens.

Substituting the calculated values into formula (5), we obtain the uncertainty of tensile
strength measurement performed for thirty samples uRm = 0.295 MPa. Tensile strength was
calculated from the formula:
Fm . (6)
Rm =
a0b0

The data from Table 1 and the calculated values of the average thickness and width of the
specimens tested were substituted into formula (6) to calculate tensile strength Rm = 45.1
MPa. The final result was written as: Rm = 45.1 ±0.295 MPa.
The standard deviation is the parameter that is most commonly calculated by computer
programs of universal testing machines. The parameter is determined from the formula:
n
1
s= å
(n - 1) i =1
( xi - x )2 . (7)

Table 2 shows data obtained with the LabMaster software operating the universal testing
machine.

Table 2. Data used to calculate the standard deviation.

Test/specimen Tensile strength (Rmi – Rm )2 Test/specimen Tensile strength (Rmi – Rm )2

number Rmi [MPa] [MPa2] number Rmi [MPa] [MPa2]
Test 1 46.54 2.16 Test 16 47.23 4.66
Test 2 43.65 2.01 Test 17 47.10 4.12
Test 3 43.91 1.36 Test 18 42.13 8.64
Test 4 41.74 11.14 Test 19 44.88 0.04
Test 5 46.21 1.30 Test 20 43.94 1.30
Test 6 45.42 0.12 Test 21 46.64 2.47
Test 7 47.40 5.44 Test 23 45.75 0.45
Test 8 43.24 3.34 Test 25 43.98 1.19
Test 9 44.17 0.82 Test 24 45.32 0.06
Test 10 47.33 5.08 Test 26 46.34 1.59
Test 11 46.44 1.88 Test 27 46.72 2.73
Test 12 45.76 0.48 Test 28 42.57 6.26
Test 13 45.85 0.61 Test 29 44.08 0.99
Test 14 44.32 0.56 Test 30 44.11 0.93
Test 15 45.61 0.29 Test 31 43.79 1.65
X 45.07
S 73.63

Substituting the data from Table 2 into formula (7) gives the value of the standard deviation,
s = 1.59 MPa. Knowing the expected value and the standard deviation, we can represent the
result in the form of normal distribution using the formula:
-( x - Rm)2
1
f ( x) = e 2s2 , (8)
s 2p
where: f(x) – the probability density function.

558
 Metrol. Meas. Syst., Vol. XXI (2014), No. 3, pp. 553–560.

Figure 4 shows the probability density function of the normal distribution of tensile strength.

Fig. 4. Probability density function of the normal distribution of tensile strength.

If we divide the value of the standard deviation by n , we obtain the value of measurement
uncertainty u¢Rm = 0.291 MPa, which differs by 0.004 from the uncertainty calculated from
formula (5).

4. Innovation-related risk vs measurement uncertainty

Additive manufacturing is an innovative approach to production of machine elements and

other finished products. This technology can be used to fabricate small batches or one-of-a-
kind items. However, it would be difficult to determine the actual properties of finished
products, i.e., strength properties, for every manufacturing process. To reduce innovation-
related risk, it is vital to calculate measurement uncertainty on the basis of tensile test data.
Determining measurement uncertainty is a certain solution. The calculations take into
consideration changes in the properties of each piece measured.
Tensile tests are destructive tests and their results can be used for an entire population of
samples, i.e., items produced. In the case considered here, the problem of uncertainty refers to
the geometrical dimensions of the specimen cross-section and the maximum tensile load
causing failure. That is why it was necessary to estimate combined standard uncertainty. The
assessment was based on one parameter, i.e. the indirectly measured tensile strength, Rm. We
took into account the changes in the specimen dimensions and their material structure, with
the latter affecting the load transfer capacity. Determining the measurement uncertainty for
this type of innovative manufacturing techniques is essential as it provides design engineers
with a tool to support decision-making on how to practically use additive manufacturing.
Similarly to the application of other modern measurement methods described in elaborations
[11, 12], extends the applicability of different manufacturing techniques.

5. Conclusion

This analysis has confirmed good repeatability of results at the maximum tensile load, as
shown in the cumulative plot in Fig. 3. After the maximum tensile load was exceeded, i.e., in
the plastic region, the specimens failed at different elongations.
Model materials used in additive manufacturing have not been studied as extensively as the
most common structural material, i.e. metal alloys and plastics. Hence, the necessity to better
understand them. In this study, the performance of FullCure 720 was analyzed using a larger
number of samples, i.e., thirty, whereas a typical number is ten.

559
 St. Adamczak, J. Bochnia, B. Kaczmarska: ESTIMATING THE UNCERTAINTY OF TENSILE STRENGTH MEASUREMENT …

The results will be useful to engineers designing models made of FullCure 720. The
measurement uncertainty determined on the basis of the tensile test data testifies to a high
repeatability of results. The standards concerning the testing of composites, which were
adopted for the material used in additive manufacturing, do not include requirements for the
evaluation of measurement uncertainty. They include, however, requirements for the
calculation of the standard deviation. Evaluation of measurement uncertainty of new, not
commonly known, materials seems absolutely justified.

Acknowledgements

The study was performed using the equipment purchased for The Equipment Support for the
Innovative Research Laboratories of the Kielce University of Technology (LABIN) Project
co-funded by the European Union within the Development of Eastern Europe Operational
Programme 2007-2013, the Innovative Economy Priority Axis, Measure I.3: Support for
Innovation.

6. References

[1] Campbell, I., Bourell, D. and Gibson, I. (2012), Additive manufacturing: rapid prototyping comes of age,
Rapid Prototyping Journal, Vol. 18 Issue 4, 255–258.
[2] Puebla, K., Arcaute, K., Quintana, R., Wicker, R.B., (2012), Effects of environmental conditions, aging,
and build orientations on the mechanical properties of ASTM type I specimens manufactured via
stereolithography, Rapid Prototyping Journal, Vol. 18 Issue 5, 374–388.
[3] ASTM, Standard 638 (2010), Standard test method for tensile properties of plastics.
[4] Chockalingam, K., Jawahar, N., Chandrasekhar, U., (2006), Influence of layer thickness on mechanical
properties in stereolithography, Rapid Prototyping Journal, Vol. 12 Issue 2, 106–113.
[5] ISO, Standard 527-1 (2012), Plastics - determination of tensile properties - Part 1: General principles.
[6] ISO, Standard 6892-1 (2009), Metallic materials – Tensile testing - Part 1: Method of test at room
temperature.
[7] Adamczak, S., Bochnia, J., Kundera, Cz. (2012), Stress and strain measurements in static tensile tests,
Metrology and Measurement Systems, No. 3, Vol. XIX, 531–540.
[8] Inspekt Mini (2011), Universal testing machine Inspekt mini 3kN, Hegewald & Peschke MPT GmbH.
[9] LabMaster software (2011), Version 2.5.3.21.
[10] Adamczak S., Makieła W. (2010), Fundamentals of metrology and quality engineering for mechanical
engineers, WNT.
[11] Stępień K., Makieła W. (2013), An analysis of deviations of cylindrical surfaces with the use of wavelet
transform, Metrology and Measurement Systems, No. 1, Vol. XX, 139–158.
[12] Cedro L., Janecki D. (2011), Determining of Signal Derivatives in Identification Problems -FIR
Differential Filters, Acta Montanistica Slovaca, R 16, cz. 1, 47–54.

560