Investigating the factors that influence the cold crushing strength of shaped refractories

Abstract

Cold crushing strength, commonly used to characterise the mechanical strength of refractory materials, is determined according to standards in different parameters of the test. In this paper we present the results of investigations conducted within the framework of the ReStaR project, which is aimed at determining the influence of the testing parameters on CCS determination of dense-shaped refractories. The Plackett-Burman design of experience and variance analysis was applied to determine the most significant factors. In the first stage of investigation, 14 influencing factors were tested in one laboratory, and in the second stage the influence of the chosen factors was analysed on the basis of inter-laboratory tests.

Full text

ABSTRACT
Cold crushing strength, commonly used to characterize the
mechanical strength of refractory materials, is determined ac-
cording to standards in different parameters of the test. In this
paper are presented the results of investigations conducted within
the framework of the ReStaR project, which is aimed at determin-
ing the inuence of the testing parameters on CCS determina-
tion of dense-shaped refractories. The design of experiment and
variance analysis was applied to determine the most signicant
factors. In the rst stage of investigation, 14 inuencing factors
were tested in one laboratory, in the second stage the inuence of
the chosen factors was analyzed on the basis of inter-laboratory
tests and in the last stage reproducibility and repeatability of CCS
results obtained in round robin test were calculated.
1. INTRODUCTION
One of the important parameters characterizing refractory
products is cold crushing strength (CCS), which is always pro-
vided in datasheets and is a prerequisite for the technical ac-
ceptance of products. It is determined at room temperature by
a uniaxial compression test and calculated from the maximum
force that a test piece can withstand before failure divided by the
loaded surface.
Due to the fact that refractory materials are heterogeneous and
their properties are not constant but can differ between samples
even when extracted from the same brick, it is particularly im-
portant to eliminate or reduce the factors caused by differences in
sample preparation, as these conditions of the test can inuence
the results and their dispersion.
Different standards describe procedures for CCS testing for
dense-shaped refractories (EN 993-5, ISO 10059-1, ISO 10059-2
and ASTM C133). According to the existing standards, different
requirements apply to the shape, size and dimensional tolerance,
loading rate, hardness and roughness of the loading plates and
packing (Table 1). The above requirements are not fully dened
in all of the standards, e.g. in ASTM, standard parallelism and
perpendicularity tolerance as well as hardness and roughness of
the loading plates are not specied. The packing is applied ac-
cording to the ASTM and ISO 10059-2 standards.
As a result, inuencing the testing parameters related to the CCS
test may vary in different laboratories and may affect the obtained
results. In order to establish the best conditions of tests that will
give results with high repeatability and reproducibility, it is nec-
essary to identify the most signicant parameters connected with
the testing procedure which inuence the CCS measured results.
In the investigations conducted within the framework ReStaR
project [1] the design of experiments and variance analysis was
applied to determine the most signicant testing parameters (fac-
tors) in order to minimize the number of experiments.
In the rst stage of investigations a large number of inuencing
factors was tested in one laboratory according to Plackett-Bur-
man plan [2]. In the second stage, a limited number of signicant
testing parameters was tested in inter-laboratory tests using full
factorial design of experiment, and in the last stage, repeatabil-
ity and reproducibility of the results was investigated in inter-
laboratory tests.
Proceeding 374
INVESTIGATING THE FACTORS THAT INFLUENCE THE COLD CRUSHING
STRENGTH RESULTS OF SHAPED REFRACTORIES
Jerzy Czechowski, Jacek Podwórny, Anna Gerle
Institute of Ceramics and Building Materials, Refractory Materials Division, Gliwice, Poland
Emilie Dahlem
Forschungsgemeinschaft Feuerfest e.V. Höhr Grenzhausen, Germany
Tab. 1: Comparison of the standards’ requirements for dense-shaped refractory products.
Standard EN 993-5 ASTM C 133 ISO 10059-1 ISO 10059-2
Shape and size of the
test piece
Cylinder: ϕ=h=50 mm,
Cube: 50x50x50 mm or
Cylinder: ϕ=h=36 mm,
Cylinder: ϕ=h=2in. (51 mm) As in EN 993-5 As in EN 993-5
Surfaces of the test piece Plane Parallel
(tolerance 0.2 mm)
Plane Parallel
(conditions not specied)
Plane Parallel
(tolerance 0.2 mm)
Plane Parallel
(tolerance 0.2 mm)
Perpendicularity Tolerance 0.5 mm Not specied Tolerance 0.5 mm Tolerance 0.5 mm
Orientation
of loading
Parallel to press direction Parallel to press direction Parallel to press direction Parallel to press direction
Loading rate 1.0 MPa/s ± 0.1 MPa/s 0.5 MPa/s 1.0 MPa/s
±0.1 MPa/s
1.0 MPa/s
±0.1 MPa/s
Loading plates:
Hardness
Roughness, Ra
58-62 HRC
0.8-3.2 μm
Not specied
Not specied
58-62 HRC
0.8-3.2 μm
(Tolerance of parallelism:
0.03 mm)
58-62 HRC
0.8-3.2 μm
(Tolerance of parallelism:
0.03 mm)
Packing No Cellulose bre wall board
0.25 in. (6.4 mm)
No Cellulose bre wall board
3-7 mm

2. MATERIALS AND METHODOLOGY
Two kinds of shaped refractory materials were used for the
tests, i.e. grade HA 75 (High Alumina >75% Al2O3 with high
strength, i.e. 70 MPa according to the data sheet) and grade MC
95/10 (Magnesia Carbon >95% MgO, ~10% carbon character-
ized by low strength, i.e. 35 MPa according to the data sheet),
both manufactured by RHI for the ReStaR project. In order to
limit the inuence of structural defects of the tested materials on
the CCS results, all bricks and prepared test pieces were checked
by using an ultrasonic tester. The average ultrasound velocity (4.6
± 0.2) × 103 m/s for product HA 75 and (4.0 ± 0.2) × 103 m/s for
product MC 95/10 were measured. Small dispersion of the results
indicates good homogeneity of both materials.
In the rst stage of investigations, fourteen relevant factors on
two levels – high (+) and low (-) – were chosen for the Plackett-
Burman factorial design (Table 2).
Tab. 2: Factors levels for dense refractory materials.
No Factor Label + -
X1Shape Sh cube cylinder
X2Extraction Ex corner middle
X3Press direction Pd // ·
X4Load rate Lor 1,0MPa 0.2MPa
X5Preload Prel 2000Pa No
X6Grinding Gr Yes No
X7Height H 36mm 50mm
X8Dimensions D 36mm ø50mm
X9Parallelism Para // Δh=0.5mm
X10 Perpendicularity Per ┴ΔI=2mm
X11
Roughness of
plates Rou Ra
3.0-3.2μm polished
X12 Hardness of plates Har 60HRC 50HRC
X13 Packing Pack No 7mm
X14 Materials Mat A C
In order to determine the inuence of the test pieces preparation
quality on the CCS results the samples with deviation of paral-
lelism and perpendicularity were made. The values of the devia-
tions, given as low levels in Table 1, were equal to the limits of
tolerance in the EN and ISO standards. The upper carrying steel
plates, with the hardness and roughness also given Table 1, were
specially prepared. A 7 mm-thick cellulose ber wall-board was
used in the tests with packing. The thickness of the board was
close and equal to the upper limits applied in the ASTM C133 and
ISO 10059-2 standards, respectively.
In the rst stage (ReStaR phase 1), the experimental design with
14 factors at two levels consisted of 16 experiments with three
repeating ones. Together, 48 test pieces were tested.
The obtained CCS results were used to calculate the regression
equation, and the Student’s t-test was applied to determine the
level of signicance for each regression coefcient.
Determining the signicance of each factor was expanded on
variation analysis (ANOVA) using the OptiVAL v.3.3.2.4 soft-
ware from Quodata. ANOVA are based on the F-test: variation
due to non-experimental effect divided by variation due to experi-
mental error. The factor is statistically signicant if the calculated
F is larger than critical F value on signicance level (here) 0.05.
In the second stage of investigations, the round-robin tests were
performed in order to check the signicance of the factors chosen
from the previous tests. In the scope of these investigations, four
laboratories throughout Europe were involved in the testing us-
ing the same material of HA 75 grade and the same established
conditions of the tests. Repeatability and reproducibility of the
experiments were analyzed in the third stage after round robin
tests in 9 laboratories.
3. RESULTS AND DISCUSSION
The average CCS results for each experiment according to the
Factorial Design Matrix are shown in Table 3.
The sign (+) in the matrix corresponds to the upper level of the
factor and (-) corresponds to the lower level of the factor as it
is given in Table 2. The calculated regression equation for the
obtained results was:
Tab. 3:. Factorial Design Matrix and average CCS results from three tests obtained in each experiment.
Run Sh
X1
Ex
X2
Pd
X3
Lr
X4
PI
X5
Gr
X6
H
X7
D
X8
Para
X9
Per
X10
Rou
X11
Har
X12
Pac
X13
Mat
X14
Average CCS
(MPa)
1 + + + - - - + - - - + + + - 39.7 ± 7.1
2 + + + + - - - + - - - + + + 84.1 ± 19.7
3 - + + + + - - - + - - - + + 90.9 ± 15.3
4 + - + + + + - - - + - - - + 53.5 ± 11.3
5 + + - + + + + - - - + - - - 35.5 ± 8.3
6 + + + - + + + + - - - + - - 21.5 ± 4.3
7 - + + + - + + + + - - - + - 40.1 ± 6.1
8 - - + + + - + + + + - - - + 52.1 ± 11.8
9 - - - + + + - + + + + - - - 33.9 ± 2.2
10 + - - - + + + - + + + + - - 41.8 ± 2.8
11 - + - - - + + + - + + + + - 34.2 ± 6.9
12 - - + - - - + + + - + + + + 94.4 ± 21.3
13 - - - + - - - + + + - + + + 102.9 ± 25.6
14 + - - - + - - - + + + - + + 85.0 ± 17.6
15 + + - - - + - - - + + + - + 66.0 ± 17.6
16 - - - - - - - - - - - - - - 30.9 ± 4.6

y=56.9xo - 3.1x1 - 4.9x2 + 3.3x3 + 4.5x4 - 4.8x5 - 15.6x6 – 11.6x7 +
1.6x8 + 10.8x9 + 1.9x10 – 3.2x11 + 4.4x12 + 14.6x13 + 21.8x14
(1)
The critical value of the Student’s t-test (tcrit) was 2.03.
Fig 1 shows the determined levels of signicance of equation re-
gression coefcients (tcal) on the signicance level a=0.05.
0,0
2,0
4,0
6,0
8,0
10,0
12,0
14,0
16,0
x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 x13 x14
tcal
tcryt
Significance
Fig. 1: CCS testing signicance factors
The results presented in Fig. 1 revealed that the type of material
(x14), packing (x13), test piece height (x7), parallelism of the loaded
surface (x9) and grinding (x6) have a strong inuence on the CCS
results and a minor inuence was exerted by extraction (x2), load
rate (x4), preload (x5) and hardness of the plates (x12).The calcu-
lated coefcients for the shape (x1), direction of pressing (x3) and
roughness of the plates (x11) are close to tcrit, so their inuence is
very small whereas the dimension (x8), which is rather surprising,
and perpendicularity (x10) in the tested range of 2 mm tolerance
have a negligible inuence on the CCS results.
The tests conrmed the earlier remarks of A. Majdič et al. [3] that
the application of packing has a negative inuence on the CCS
results. Therefore, the application of cardboard spacers should be
eliminated. As for the steel plates used as spacers, the conditions
connected with their hardness and roughness of surface as dened
in the current testing standards can be considered sufcient. In
order to avoid the impact of the test pieces’ quality, especially the
inuence of surface non-parallelism, it is necessary to pay special
attention to this element during sample preparation.
Taking into consideration the obtained results, in the next stage
of testing the number of factors was limited to four, as given in
Table 4, and the series of tests were performed in four different
European laboratories.
Tab. 4: Factors and levels chosen for the CCS design of
experiment in the second stage of the tests.
Factor High level (+) Low level (-)
Shape Cube Cylinder
Geometry 36 x 36 mm 50 x 50 mm
Load rate 1.0 MPa/s 0.2 MPa/s
Grinding Yes No
The “extraction” factor was additionally included. Part of the test
pieces were cut from the centre and part from the outer part of the
examined shapes. In total, 96 test pieces were examined. All tests
were performed using HA 75, i.e. the same grade material.
The design of the experiments and the CCS average results ob-
tained in four laboratories for each experiment are shown in Table
5.
Tab. 5: Design of experiments and CCS results.
Exp. Size
(mm) Grind. Load rate
[MPa/s] Format CCS
[MPa]
750x50 No 0,2 cylinder 58,9±4,6
1 50x50 No 1,0 cube 63,6±9,2
350x50 Yes 0,2 cube 65,4±10,1
850x50 Yes 1,0 cylinder 66,7±8,9
2 36x36 Ye s 1,0 cube 62,1±11,5
4 36x36 Ye s 0,2 cylinder 64,1±10,6
6 36x36 No 1,0 cylinder 61,1±8,4
536x36 No 0,2 cube 58,5±11,3
Fig. 2: Box plots of measurements for results obtained in all
laboratories
The summarized results from all of the laboratories for each run
(Fig. 2) showed the positive inuence of the 50 x 50 mm test-
piece size on the dispersion of results, whereas different levels of
factors such as grinding, shape and load rate had no inuence on
the dispersion.
The inuence of the possible heterogeneity of material on the
cold crushing strength results was analyzed on the basis of tests as
shown in Table 6 taking into account the place of extraction. The
results obtained for test pieces cut out from the outer part (64 test
pieces) and the pieces cut out from the centre of the shapes (32
test pieces) were separated. The results in Table 6 showed slightly
higher values for the samples from the centre of the shapes. How-
ever, the differences were smaller for test pieces in the form of
cylinders, for which, dispersion was also smaller. The higher CCS
values in the centre of the shapes are in good agreement with the
results of the bulk density and open porosity tests obtained for the
same material [4].

Tab. 6: Inuence of sample extraction on the CCS testing results.
CCS [MPa]
All tested samples 62.4 ± 9.4
All samples from the outer part
Cylinders
Cubes
61.9 ± 8.4
62.1 ± 8.3
61.3 ± 8.5
All samples from the centre
Cylinders
Cubes
63.5 ± 11.3
62.8 ± 8.7
64.3 ± 13.7
Variance analysis was used to determine the inuencing factors
in each of the laboratories separately (Table 7). The signal effect
provides information about the major factors, whereas the noise
effect represents factors which have an impact on the standard
deviation of the testing results (relevant if p < 0.05).
Tab. 7: Identied signicant factors inuencing the CCS results.
Lab. Signal effect Noise effect
1Grinding* Load rate Grinding * Geometry
2Grinding* Shape Load rate
3Shape* Load rate,
Geometry* Shape Geometry, Load rate
4Grinding Shape, Geometry
The signal effects differ for each laboratory and mutual inuences
were found in most cases. The grinding seems to be an important
factor as it was the main factor in one laboratory and its interac-
tion was identied in two other laboratories. Also, noise effects
were varied for each laboratory. The difference was most prob-
ably caused by an insufcient number of repetitions in each run
in individual laboratories. It is important to emphasize that the
statistical evaluations presented here do not take into account the
heterogeneity of the material.
When all of the results were treated as one population (there were
results for 12 samples for each experiment: 3 samples x 4 labora-
tories) and the laboratory was treated as an inuencing factor, it
was found that none of the investigated factors was identied as
signicant at a signicance level a=0.05.
The results of the round robin tests conducted in 9 European labo-
ratories are showed on Fig.3 and 4. The test pieces for tests were
prepared in two laboratories. To minimize the inuence of hetero-
geneity on the test results a special method of selecting samples
for each laboratory from the shapes (16 test pieces from each
shape) was used. In all tests xed procedure of CCS testing was
applied: grinded cylinders 50 X 50mm and 1MPa/s loading rate.
Fig.3: Round robin tests results for brick HA 75
Fig.4: Round robin tests results for bricks MC 95/10
Repeatability sr, which means closeness of agreement between
results obtained using the same method on identical test material
under the same conditions, and reproducibility sR, which means
closeness of agreement between the results using the same meth-
od on identical material but under different conditions (different
laboratory operators, apparatus), are shown in Table 8.
Tab. 8: Repeatability and reproducibility of results obtained in
round robin tests
Brick HA 75 Brick MC 95/10
Mean CCS, MPa 75,0 ±9.1 50.2 ±4.3
Reproducibility, sR25.5 MPa (34.0%) 12.1 MPa (24.2%)
Repeteability, sr19.2 MPa (25.6%) 7.5 MPa (15.0%)
Nbr. of test pieces 88 83
There was not outliers and Z-scores in all tests were below 2. The
test showed lower dispersion of results and better reproducibility
and repeatability for magnesia carbon bricks than for red HA
75 bricks. It is probably related to the greater heterogeneity of
the latter.
4. CONCLUSIONS
1. Factorial designs were affectively applied to determine the
factors that inuence the cold crushing strength results of
dense shaped refractory materials.
2. The application of packing should be eliminated because us-
ing a cardboard spacer has a negative inuence on the CCS
results.
3. Conditions related to the hardness and roughness of the sur-
face of steel plates dened in current standards are sufcient.
4. A load rate of 1.0 MPa/s as required by the EN 993-5 standard
is adequate.
5. Test pieces in the form of cylinders or cubes with dimensions
of 50x50 mm with grounded surfaces are recommended.
When smaller test pieces, i.e. 36x36 mm, are used, dispersion
of the results is higher.
6. The CCS results depend on the place of extraction. The higher
CCS values in the centre of the shapes are in good agreement
with the results of the bulk density and open porosity tests
obtained for the same material
7. The round robin tests showed higher reproducibility and re-
peatability values for red HA 75 bricks than for magnesia
carbon MC 95/10 bricks as result of differences in their ho-ho-
mogeneity. It is proposed to include information on reproduc-
ibility and repeatability to the European standard.
5. ACKNOWLEDGEMENT
The ReStaR project has received funding from the European
Union’s Seventh Framework Programme for research, techno-
logical development and demonstration under grant agreement
no 314884.

Data citation:
- Proceeding title, see above
- Proceeding no., see above
- Authors’ names, see above
- Source: USB-Stick UNITECR2015 – 14th Biennial Worldwide Congress
- ISBN 978-3-9815813-1-7
- urn:nbn:de:101:1-201506294612
6. REFERENCES
[1] Brochen E., Quirmbach P., Volckaert A., ReStaR –concert-
ed effort in the European refractory sector to consolidate
and made EN testing standards future-proof. Refractories
WORLDFORUM 3 (2015).
[2] Plackett P, Burman JP. The design of optimum multifactorial
experiments. Biometrika. 1946.37: 305-325.
[3] Majdič A., Hagemann L., Lichomski H. Einuss der gute der
probekorperdruckachen und der Druckplattenrauheit auf
Mittelwert und Streubreite der Kaltdruckfestigkeit feuerfester
Steine. Tonind. Zeit. 97, 1973(9):237 – 243.
[4] Dahlem E., Clasen S., Dannert Ch. Which factors inuence
the bulk density and open porosity testing results for dense
shaped refractory materials. XVI Int. Conf. on Refract., Wis-Conf. on Refract., Wis-
la, May 2015.