THE FRACTURE STRENGTH OF FIBRE-CEMENT CORRUGATED

SHEETS: A STATISTICAL APPROACH USING WEIBULL ANALYSIS

KRUNOSLAV VIDOVIČ1, MILAN AMBROŽIČ 2, KRISTOFFER KRNEL 3, TOMAŽ KOSMAČ2 and
STEPHEN AKERS4
1
Esal d.o.o. Anhovo, Vojkova 9, SI-5210 Deskle, Slovenia; 2 Faculty of Natural Sciences and Mathematics,
University of Maribor, Koroška 160, SI-2000 Maribor; 3Jožef Stefan Institute, Jamova 39, SI-1000
Ljubljana, Slovenia; 4 Eternit (Schweiz) AG, CH-8867 Niederurnen, Switzerland

ABSTRACT
The breaking load and the breaking moment during transversal and longitudinal loading of fibre-cement
profiled (corrugated) sheets were measured on products manufactured on an industrial Hatschek machine,
and the corresponding bending strengths were calculated. The influence of two processing parameters, i.e.,
the pressure applied to the green sheet directly during the manufacture and the content of the reinforcing
organic fibres within the portland cement matrix, on the fracture strength of the products was studied and
statistically analyzed assuming Weibull statistics. The Weibull parameters were compared for different
fabrication conditions. While the fibres content influences considerably the Weibull modulus for both
transversal and longitudinal loading, there is no correlation between the shaping pressure and Weibull
modulus.

KEYWORDS:
Cement; Fibre reinforcement; Bending strength; Weibull statistics.

INTRODUCTION
In recent years there have been numerous successful investigations relating to the development of fibre-
cement composites using various organic or synthetic fibres (Studinka 1989; Savastano et al. 2003; Coutts
2005; Agopyan et al. 2005; Ma et al. 2005; Peled and Mobasher 2005). The candidate fibres for such
composites need to meet several requirements, such as high strength and appropriate elastic modulus,
chemical, temperature and dimensional stability, resistance to an alkaline environment, the appropriate
morphology and compatibility with the cement matrix, good dispersion properties in an aqueous cement
slurry, and long-term durability. One of the synthetic fibre types with appropriate properties for use in fibre-
cement composites is polyvinyl alcohol (PVA) and it is used in our manufacture of corrugated sheets.
However, new composite materials have to be tested rigorously using classical mechanical tests as well as
various accelerated ageing tests to predict the long-term behaviour of the products under severe climatic
conditions (Akers and Studinka 1989; Akers 1989; Purnell and Beddows 2005). Different testing
requirements should be applied to different fibre-cement products, e.g., profiled (corrugated) roofing sheets,
facades, pipes, etc., and also with respect to different climatic conditions.
Improvements and/or cost reductions relating to the fibre-cement fabrication process and the quality of the
final products (in our case corrugated roofing sheets) can be achieved by varying different processing
parameters (Ma et al. 2005; Vidovič et al. 1996; Negro et al. 2005; Negro et al. 2006; Beaudoin 1990). With
regard to the fibres, their type, volume fraction and length are typical parameters that influence the
mechanical properties (Ma et al. 2005). It is also known that increasing the volume fraction of the reinforcing
fibres increases the composite tensile strength because of the partial transfer of the load from the cement

116
matrix to the fibres. However, at very large fibre contents some detrimental effects occur, such as increased
porosity and a weaker interface between the fibres and the matrix; in addition, the production costs are
increased. With regard to other processing parameters, the pressure used to shape the green sheet products is
of great importance. To optimize the production the influence of variable processing parameters on the
mechanical properties of the composite materials/products should then be measured and analyzed. Usually,
mechanical tests are performed on laboratory-made samples with a simple geometry, rather than on final
products from serial production.
The measured strengths in typical mechanical tests for many brittle and quasi-brittle materials result in a
Weibull statistical distribution (Weibull 1949; Weibull 1951; ReliaSoft 1992; Kosmač et al. 1999; Setien
2005; Lewis et al. 2005). This statistical distribution has been widely used in many fields, with many
examples from the literature given (Reliasoft 1992). Weibull statistics have also been used in civil
engineering (Anton et al. 1998; Toutanji 1999; Caliskan 2003; Li et al. 2003; Huang and Cheng 2004; Cooke
2005). Usually, simple 2-parameter Weibull statistics are used, although in many cases the 3-parameter
Weibull statistics or the Weibull statistics corresponding to two or more different modes of fracture works
better (Reliasoft 1992; Li et al. 2003).
This investigation has the aim of checking the relevance of the two of the production parameters with regard
to the strength of fibre-cement corrugated roofing sheets, i.e., the volume fraction of the fibres and the
shaping pressure on the green sheet. The experiments were performed on industrial products in order to
avoid the difficult translation of know-how from a research laboratory to industry. In this article the results
on measurements/calculations of four mechanical properties of the corrugated roofing sheets are presented:
the breaking force during transversal loading, the breaking bending moment during longitudinal loading, and
the calculated bending strengths for transversal and longitudinal loadings. The experimental data were fitted
to the 2-parameter Weibull probability distribution function. The aim of this paper is to show that the 2-
parameter Weibull statistics can be used successfully in this case and that it reveals significant differences in
the statistical parameters when the production parameters are varied. Some evidence is also given for the
influence of the alignment of PVA fibres on the difference in the effective bend strength of the material for
transversal and longitudinal loading, respectively.

EXPERIMENTS AND STATISTICAL PROCEDURE

Materials and processing

The mix composition of the final profiled sheet material (after 21 days of maturing) is given in volume
fractions as follows: 40% of hydrated portland cement, which represents the matrix; 30% of air captured in
pores; 12% of water; 11% of inorganic additives, such as fillers and pigments; 5% of processing cellulose
fibres; and 2% of reinforcing PVA fibres.
The manufacturing process for the corrugated sheets used in this work is essentially the Hatschek procedure
(Studinka 1989). All the material components are first dispersed and homogenized in water, producing a
highly diluted suspension. In the filtration process a filter cake is formed on rotating sieve cylinders, and it is
then transported by the felt conveyer to the forming drum. When the desired thickness of deposited material
on the drum is achieved, it is cut and transported on a conveyer belt for corrugating and pressing, which is
followed by primary maturing of the product at an elevated temperature. Finally, the products are matured
for 21 days under ambient conditions in a closed warehouse.
The mechanical testing was performed on standard products from production, i.e., a fibre-cement corrugated
roofing sheet, referred to as V5, with the following dimensions (Fig. 1): width W = 920 mm, length L = 1250
mm, corrugation pitch P = 177 mm, corrugation height H = 51 mm. The typical thickness, T, of the V5 sheet
is between 5.95 mm and 6.20 mm.
In the first part of the experiment the mass fraction, cf, of the PVA and cellulose fibres was varied in the
initial products. For three different material compositions the mass fractions of the fibres with respect to the
total mass of solid components (without water and air) were as follows: cf = 5.4%, 5.7% and 6.1%,

117
respectively; the remaining portion being the cement, additives and fillers. The mass ratio of the reinforcing
PVA fibres to the processing cellulose fibres was 1:2 in all three cases. The corrugation shaping pressure, ps,
of the hydraulic press during the shaping of the V5 sheets was 10 MPa. In the second part of the experiment
the shaping pressures ps = 8 MPa, 9 MPa and 10 MPa, respectively, were used, keeping the mass fraction of
the fibres at 6.1%. The reason for the relatively narrow span of the chosen variable parameters was the
following. In the ordinary industrial production process we have cf = 6.1% and ps = 10 MPa. As already
mentioned, the trial change of parameters was made in massive serial production with many products where
we could not afford changing the parameters too widely, with the risk of degrading the quality of the
products to the degree that they have to be discarded. On the other hand, even slight reducing the PVA-fibres
content, for instance from 6.1% to 5.4% without measurable degradation of mechanical properties, would
lower the production cost.

Measurement of the mechanical properties

For the determination of the mechanical properties of the V5 sheets the methods prescribed by the European
standard EN 494 were used (EN 2004). Before the mechanical tests were conducted the test sheets were
soaked in water for 24 h and then briefly dried according to the standard. The focus for the discussion in this
paper was the measurement of the breaking force (transversal loading) and the bending moment (longitudinal
loading); in addition, the bending strengths for both the transversal and longitudinal loadings were
calculated. For these measurements a BP-10 laboratory press-machine (Walter+Bai AG, Switzerland), with
the measuring scales 2 kN and 10 kN, and equipped with the corresponding software, was used. The
geometry of the transverse and longitudinal loadings, according to the EN 494 standard, is shown in Fig. 1.
While the breaking force (load) was measured directly for both types of loading (transversal and
longitudinal), the three remaining properties were calculated from the breaking force and the geometrical
parameters of the test.
TRANSVERSAL LOADING: According to EN 494 this is essentially a 3-point bending test (Fig 1a). The
width of the horizontal transverse supports is 50 mm, and the inner span is 1100 mm. On the upper side of
the sheet a transverse loading beam with a width of 230 mm is placed symmetrically with respect to the
supports. It is necessary to insert 10-mm-thick strips of soft material (possibly felt) between the sheet, the
supports and the loading beam. The breaking load (force) is usually defined per unit width of the sheet, but in
this case a constant width W = 920 mm was used; therefore, the results for the breaking force are expressed
in Newtons (N) for the actual width of 920 mm (Table 1). The transversal breaking force (load) is denoted by
FT, where the index T refers to the transversal loading.
LONGITUDINAL LOADING: This is again a 3-point bending test, but with the supports in the longitudinal
directions of the sheet (in Fig. 1b, one of the possible variants is shown). A sample of the sheet, at least L’ =
300 mm long, should be cut for the longitudinal loading test. The supports should be rounded, while a soft
material strip (possibly felt) should be used for the contact between the sheet and the loading beam, as is the
case for the transversal loading. The span S between the supports is determined by the corrugation pitch. The
bending moment ML at the moment of rupture is expressed per unit length L’ of the test piece and is given by
the expression:
FS
ML 
4L'
(1)
where F is the breaking load. The index L in the term ML refers to the bending moment for longitudinal
loading, and the moment itself as a vector has the longitudinal direction, i.e., parallel to the corrugations.

118
Fig. 1 The geometry of transverse (a) and longitudinal (b) loading according to the European standard EN 494.
Lengths are given in millimetres.

A comment should be given why the force and the moment of force are relevant for the transversal and
longitudinal bending loading according to the standard EN 494. In bending loadings in the tests and in every-
day use it is the moment of force (and not the force itself) which determines when the sample/product breaks
since the span between the supports is also relevant. While in the transversal loading the moment of force
could be used as the criterion for the quality of the corrugated sheet the span between the supports is
uniquely prescribed by the standard (Fig. 1) and thus the force can be simply used as a testing property. On
the other hand, in the longitudinal loading, the span cannot be uniquely defined (the corrugation pitches
differ for different products) and so the loading moment must be kept as a standard.
TRANSVERSAL AND LONGITUDINAL BENDING STRENGTHS: While the force FT and the moment
ML are the most interesting properties with regard to the application of the final V5 fibre-cement products,
rather than the material bending strength, the latter is calculated and included in the discussions for
comparison. For both the transversal and longitudinal loadings of the V5 sheets the calculation of the
bending strengths, T and L, respectively, from the load at rupture and the geometric parameters of the sheet
and the bending test is rather complicated, due to the geometry of the corrugated sheets. The problem of
transversal loading is easier since the applied stress can be treated as uni-axial since the cross section of the
sheet is uniform through its length, thus the main problem is calculating numerically the cross-section inertia
moment for the given sheet profile. On the other hand, for longitudinal loading the direction of stresses
varies according to the corrugations between the supports, and the entire stress analysis is necessary. Here
special software developed by Eternit, Switzerland, was used to calculate the strengths from the input data
(breaking load, type of loading, geometry and dimensions of the sheets). The formulas for the evaluation of
the transversal bending strength can also be found in the DIN 274/1 standard (DIN 1972).
The relative measurement uncertainties were roughly as follows: 0.5% for the force, 0.1% or less for the
length scales (other than the sheet thickness), and 1% for the sheet thickness. Since the relative uncertainties
for multiplied or divided quantities are added together an estimate can be made of the relative uncertainties
of ML, T and L. Using Eq. (1) for the bending moment, we obtain its relative uncertainty: 0.5% + 2*0.1% =
0.7%. Calculations of the bending strengths scale with the factors/quotients F, L/L and T2 (where F, L and T
are the breaking force, the geometrical lateral dimensions and the sheet’s average thickness, respectively),

119
thus the relative uncertainty of the strengths is 0.5% + 2*0.1% + 2*1% = 2.7%. The actual uncertainties may,
however, be somewhat larger, due to misalignments and variability of the sheet thickness.

Statistical evaluation
It has been shown that for brittle ceramic materials at least 30 measurements of the mechanical strength are
needed in order to obtain sufficiently reliable Weibull statistics (Quinn 1990). The fibre-cements behave as
quasi-brittle materials in the bending test for which we have found no direct recommendations about the
number of samples in the literature, thus facing the experience of ceramic materials we adopted the number
of samples N = 30 samples for each set of processing parameters. The individual measurements were then
statistically evaluated. First, the mean values and the standard deviations were calculated. Then the data were
fitted to the 2-parameter Weibull statistics as follows.
Let the statistical variable be denoted by x (x being one of the measured/calculated quantities, FT, ML, T,
L.) and let its failure probability function be P(x), where the letter P refers to the probability. The function
P(x) is the probability of finding the value for the measured variable to be less than x. If we assume all
possible values of x between zero and infinity, then at both limits, P(0) = 0, and P() = 1 holds. In the
literature, this function can also be called the unreliability function (ReliaSoft 1992). In the case of 2-
parameter Weibull statistics, P(x) is equal to:
x m
P( x)  1  exp( ( ) ) (2)
x0
with the pair of Weibull parameters m and x0. The dimensionless Weibull modulus, m, refers to the slope of
the line in the special Weibull plot diagram (Reliasoft 1992), while the scale parameter x0 determines its
position, as shown in Fig. 3. Here, a rough sketch of the procedure is given to find the Weibull parameters;
the reader can find details elsewhere (Reliasoft 1992; Ritter et al. 1981; Johnson 1951; Lloyd and Lipow
1962).
The N measured values xi of the quantity x (i = 1 to N) are first sorted in increasing order. Then the value Pi
is attributed to the i-th value xi (Pi being independent of the value xi) by solving the equation:
N
N
  k  P i
k
(1  Pi ) N k  0.5
k i   (3a)
in accordance with the binomial distribution (Johnson 1951). This equation for Pi has to be solved
numerically. Alternatively, a much simpler equation can be used, which leads to similar calculated values for
Pi:
i  0.3
Pi 
N  0.4
(3b)
Other simple formulas are also used in the literature, instead of Eqs. (3a) and (3b) (Reliasoft 1992; Li et al.
2003; Li et al. 1993; Wu et al. 2006). Using Monte Carlo simulations, Wu et al. showed that simple formulas
like (3b) give slightly biased values of Weibull moduli, either too high or too low values; for instance, Eq.
(3b) results in an about 5% underestimated value of m, for N between 15 and 50 (Wu et al. 2006). The
ordered data pairs (xi, Pi) for i = 1 to N are finally fitted to the function (2) by the usual fitting procedures, in
order to obtain the Weibull parameters m and x0, as shown in Fig. 3. Because there is a limited number of
measurements (in this case N = 30) the Weibull parameters obtained need not be equal to their true values for
a statistical assessment of many samples; rather, these are estimated values of m and x0. Knowing the
estimated Weibull parameters one can re-evaluate various statistical parameters, for instance, the mean value
of the variable x and its standard deviation, and compare the calculated values with the values obtained from
experimental data.

120
Once the parameters of the Weibull statistics are obtained the reliability of their estimated values should be
evaluated. This is usually done with the help of the Fisher matrix of the system (Lloyd and Lipow 1962).
One can then predict with some confidence level CL, say 90%, that the actual value of the parameter lies
within some interval. ReliaSoft's commercial computer program package Reliasoft-Weibull++ for the
computation and visualization of the statistical distribution was used (Reliasoft 1992).

RESULTS AND DISCUSSION

Typical deflection-load curves for transversal and longitudinal loading are schematically sketched in Figs. 2.
Two typical points on the curves are the limit of proportionality (LOP) and the modulus of rupture (MOR),
where the load reaches its maximum value. At LOP the first sign of matrix cracking is detected. Between
LOP and MOR multiple matrix cracking occurs together with fibre pull-out and fibre fracture. After MOR
the fibre pull-out and fibre fracture is continued and is believed to gradually become the dominant failure
mechanism. The difference in the shape of deflection-load curves for both types of loading (rather
complicated curve with more maxima for transversal loading in comparison to simple curve for longitudinal
loading) is attributed to the cascade failure for transversal loading (not all of the corrugations fail
simultaneously). Since the deflection-load curves are shown in arbitrary units in order to directly compare
the curves for two physically different quantities (force and moment of force), we mention here some
numerical data for the reader’s evidence. The MOR for the transversal loading is approximately at the
average deflection of 40 mm (the actual span of values is between 20 mm and 70 mm) and the force between
5 kN and 6 kN. The MOR for the longitudinal loading is at the deflection between 20 mm and 30 mm and
the moment of force per unit length between 90 N m/m and 100 N m/m.
The average values and the standard deviations (in parentheses) of the quantities FT, ML, T and L for N =
30 samples for each group are shown in Table 1. The standard deviations for all quantities are of the order of
10% of their mean values, which is much larger than the measurement uncertainties mentioned in subsection
Measurement of the mechanical properties. It has been shown by Monte Carlo simulations that the influence
of measurement errors on the statistics may be ignored unless they become comparable to the natural
deviations of the measured quantities due to intrinsic differences between the samples (Ambrožič and
Vidovič 2007).
It can be seen from Table 1 that while the mean values of all the quantities increase with increasing fibre
content, the shaping pressure has a significant effect on the properties only during longitudinal loading. It is
also evident that although the given composition and pressure are the same (cf =6.1%, p = 10 MPa) for the 3rd
and 6th data rows in the table, the measured results are not completely equal; but the results are similar
anyway, indicating a good consistency with control parameters. There was a 3-month period between the
production and measurement of the mechanical properties of the series corresponding to the 3rd and 6th rows.

121
Fig. 2 Typical schematic deflection-load curve for transversal (a) and longitudinal (b) loading

Table 1. The dependence of the mechanical properties of the V5 sheets on the mass fraction of fibres cf and the
shaping pressure ps: average values and standard deviations (in parentheses).
Data cf ps FT ML T L
row (%) (MPa) (N) (N m/m) (N/mm2) (N/mm2)
1 5.4 5056 (589) 92.8 (10.6) 19.16 (1.78) 13.31 (0.86)
2 5.7 10 5550 (548) 99.2 (11.4) 20.95 (2.03) 15.02 (1.48)
3 6.1 5794 (395) 104.3 (9.1) 22.34 (1.30) 16.78 (1.02)
4 8 5682 (485) 85.0 (8.5) 21.88 (1.60) 13.36 (1.12)
5 6.1 9 5508 (375) 91.9 (7.6) 20.97 (1.32) 15.29 (1.05)
6 10 5652 (441) 102.9 (11.1) 21.74 (1.53) 16.62 (1.06)

The estimated Weibull moduli, m, and the scale parameters (x0  FT0, ML0, T0 and L0) are shown in Tables
2a-2d, together with the correlation coefficient, , which indicates how the experimental data fit to the 2-
parameter Weibull statistics. The value  = 1 means an exact (idealized) fit. It can be seen that the correlation
coefficient is in most cases almost equal to 1, indicating a good fit of experimental data to the 2-parameter
Weibull statistics. The Weibull moduli are of the order of 10, comparable to brittle ceramic materials. For
comparison with ceramic materials and with our results for corrugated roofing sheets, Huang and Cheng
obtained Weibull moduli for the measured fracture toughness mostly between 7 and 10 for foamed alumina
cements of different relative densities (Huang and Cheng 2004), while Li et al. reported moduli of about 7
and 12 for the body and surface crack mode, respectively, in their bimodal Weibull statistical distribution for
the strength of their concrete specimens (Li et al. 2003).

122
Table 2a The estimated Weibull statistical parameters for the force FT
Data cf ps m FT0 
row (%) (MPa) (N)
1 5.4 9.9 5310 0.986
2 5.7 10 11.6 5793 0.948
3 6.1 17.2 5970 0.976
4 8 13.6 5896 0.974
5 6.1 9 17.1 5677 0.973
6 10 14.9 5849 0.961

Table 2b The estimated Weibull statistical parameters for the moment ML

Data cf ps m ML0 
row (%) (MPa) (N m/m)
1 5.4 9.9 97.5 0.983
2 5.7 10 10.1 104.2 0.931
3 6.1 13.4 108.3 0.985
4 8 11.6 88.8 0.952
5 6.1 9 13.9 95.4 0.938
6 10 10.8 107.7 0.976

Table 2c The estimated Weibull statistical parameters for the bending strength T
Data cf ps m T0 
row (%) (MPa) (N/mm2)
1 5.4 12.3 19.95 0.978
2 5.7 10 11.7 21.87 0.935
3 6.1 19.9 22.93 0.983
4 8 15.8 22.60 0.982
5 6.1 9 18.4 21.57 0.966
6 10 16.6 22.42 0.976

Table 2d The estimated Weibull statistical parameters for the bending strength L
Data cf ps m L0 
row (%) (MPa) (N/mm2)
1 5.4 17.8 13.70 0.962
2 5.7 10 11.8 15.67 0.969
3 6.1 18.7 17.26 0.936
4 8 13.9 13.86 0.948
5 6.1 9 16.3 15.79 0.892
6 10 18.1 17.11 0.960

The following discussion will be mainly focused to the Weibull parameters relating to the quantities FT and
ML. The modulus m increases with the mass fraction of fibres cf (but more significantly for FT), while there is
no evident correlation between m and the shaping pressure, ps. Similarly, the scale parameters FT0 and ML0
evidently increase with increasing cf, but for increasing pressure ps only an increase of ML0 is evident.
However, to be certain of the explanations given above, one needs to have information about the confidence
of the estimated Weibull parameters, as well as the mean values and the standard deviation calculated either
directly from the experimental data or from the Weibull parameters. Table 3 gives an example of the

123
boundary values of 90%-confidence level intervals (shortly 90%CL intervals) for both the Weibull
parameters and the calculated mean value of the force FT in the case of variable cf and fixed ps = 10 MPa,
corresponding to the data in rows 1-3 in Tables 1 and 2a. For instance, the values m = 7.9 (LB = lower
boundary value), 9.9 (EST = estimated value) and 12.5 (HB = higher boundary value) in the first row of
Table 3 mean that the estimated value m = 9.9 (rewritten from Table 2a) fits the experimental data best, but
there is a 90% probability that the actual (unknown) value of m lies within the interval 7.9 < m < 12.5. Note
that the estimated parameter is the geometrical (but not arithmetical) average of both boundary values of
90% CL intervals.

Table 3 Estimated values and 90%CL intervals for the Weibull parameters m and FTO, and the corresponding
calculated mean values <FT> for the transversal breaking force.
Data cf M FT0 <FT>
row (%) LB* EST* HB* LB* EST* HB* LB* EST* HB*
1 5.4 7.9 9.9 12.5 5142 5310 5483 4863 5050 5244
2 5.7 9.4 11.6 14.4 5632 5793 5959 5362 5545 5735
3 6.1 14.0 17.2 21.2 5857 5970 6086 5657 5789 5925
*LB = lower boundary value according to 90%-confidence level intervals
*EST = estimated value
*HB = higher boundary value according to 90%-confidence level intervals

The 90%CL intervals for m are relatively wide, typically about 40% of the estimated value. The 90%CL
intervals for the scale parameters are about 5% of the estimated value, in accordance with Li et al. (1993). It
should, however, be remembered that the relative differences in the estimated Weibull parameters are larger
than the differences for scalar parameters for different processing parameters, as well. Therefore, with regard
to both Weibull parameters, in some cases the 90%CL intervals for different processing parameters overlap
and in some cases they do not. For instance, the 90%CL interval for m in the case cf = 6.1% overlaps only
slightly with the interval for cf = 5.7 and is clearly distinct from the interval for cf = 5.4%.
The Weibull statistics for different cases from Table 2 are visualized in Figs. 3a-3d, which show the fitting of
data pairs (xi, Pi) to the P(x) functional dependence. The horizontal and vertical scales of the diagram are
arranged in order to present the P(x) function as a straight line (solid lines in Figs. 3a-3d), where the larger
Weibull modulus corresponds to the steeper line. The 90%CL intervals for the Weibull parameters are taken
into consideration, so that each distribution line is surrounded by a pair of dashed lines, inside which most
data points should lie. In Fig. 3a the P(FT) distributions are compared for cf = 5.4% and 6.1% (both the 4th
and 6th rows in Table 2a for the latter fibre content). Fig. 3b makes a similar comparison of the P(ML)
statistics for cf = 5.4% and 6.1%. Figs. 3c and 3d compare the statistics for FT and ML, respectively, for two
different pressures: ps = 8 MPa and 10 MPa. The Weibull moduli are similar for both pressures, but there is
still an increase of the scale parameter in the case of ML. Graphs for intermediate values of processing
parameters, i.e., cf = 5.7% in Figs. 3a and 3b, and ps = 9 MPa in Figs 3c and 3d were not included to keep the
diagrams clear.

124
Fig. 3 Fitting of the N = 30 data pairs to the 2-parameter Weibull statistics: a) P(FT) for cf = 5.4% (circles) and
6.1% (triangles and squares), rows 1, 3 and 6 from Table 2a, ps = 10 MPa; b) P(ML) for cf = 5.4% (circles) and
6.1% (triangles and squares), rows 1, 3 and 6 from Table 2a, ps = 10 MPa; c) P(FT) for ps = 8 MPa (circles) and
10 MPa (triangles), rows 4 and 6 from Table 2a; d) P(ML) for ps = 8 MPa (circles) and 10 MPa (triangles), rows 4
and 6 from Table 2b.

125
The mean values of the transversal bending strengths are higher than for the longitudinal bending strengths.
One of the main reasons for this is most probably the preferred orientation of fibres in the longitudinal
direction of the V5 sheets (Cooke 2005), as shown in scanning electron microscope (SEM) microstructures
on Figs 4a-4b. Fig. 4a shows a polished fibre-cement-sheet surface containing the production direction of the
machine which is the longitudinal direction of the sheet while Fig. 4b is a cross-section perpendicular to that
direction. It is clearly seen that the fibres (dark wavy lines) are preferably (although not completely) oriented
in the sheet-longitudinal direction. A greater strength of the material when loading perpendicular to the
orientation of the fibres compared to longitudinal loading was confirmed by bending tests on simple flat
specimens removed from the production line before corrugation took place. As already mentioned the main
failure mechanism is first cracking the cement matrix followed by pull-out and rupture of the reinforcing
fibres. It is natural to expect that the pull-out and the rupture of PVA fibres is more pronounced in
transversal loading compared to longitudinal loading. This expectation would need a quantitative
verification, but nevertheless supports the findings from the bending test results that the contribution of the
fibres to strength properties is obviously dependent on the preferential orientation of the synthetic fibres in
particular. It needs also to be pointed out that although the synthetic fibres have the predominant contribution
to strength properties, the cellulose fibres do provide additional strength to the products as well. Figs. 5a and
5b show the SEM pictures of the fracture surface of the sheet in transversal and longitudinal loading,
respectively.

(a) (b)
Fig. 4 SEM micro-structure of the polished surface of the PVA-fibre cement corrugated sheet containing the
longitudinal direction of the sheet (a) and perpendicular to it (b).

126
 (a) (b)
Fig. 5 SEM micro-structure of typical fracture surfaces of the sheets tested in the transversal (a) and the
longitudinal (b) loading direction.

CONCLUSIONS
A study of the statistical distribution of repeated measured values of some mechanical properties of V5
corrugated roofing sheets showed that the measurements can be fitted well to the 2-parameter Weibull
statistical distribution. The fitted Weibull moduli are high, and they show a correlation with the volume
fraction of the reinforcing and processing fibres in the cement matrix. Higher volume fractions result in
higher strength of the products. There is no conclusive correlation with the shaping pressure for the range of
pressures under consideration in the present study. The statistical distributions can be conveniently
visualized in special Weibull plots, where the failure probability function is represented as a straight line,
with the Weibull modulus determining its slope.

ACKNOWLEDGEMENT
The authors are grateful to ESAL, d.o.o. Anhovo, and ETERNIT (SCHWEIZ) AG, Niederurnen, for support
and permission to publish these results, especially to Mr. Terčič and Mr. Holte, managing directors.

REFERENCES
Agopyan V., Savastano H., John V. M., Cincotto M. A. 2005. Developments on vegetable fibre-cement
based materials in Sao Paolo, Brasil: an overview, Cem Concr Compos 27 (5) 527-536.
Akers S. A. S., Studinka J. B. 1989. Ageing behaviour of cellulose fibre cement composites in natural
weathering and accelerated tests, The International Journal of Cement Composites and Lightweight Concrete
11 (2) 93-97.
Akers S. A. S. 1989 Micromechanical studies of fresh and weathered fibre cement composites, The
International Journal of Cement Composites and Lightweight Concrete 11 (2) 117-131.
Ambrožič M., Vidovič K. 2007. Reliability of the Weibull analysis of the strength of construction materials,
J Mater Sci 42 9645-9653.

127
Anton N., Ruiz-Prieto J. M., F. Velasco, J. M. Torralba. 1998. Mechanical properties and wear behaviour of
ceramic matrix composites based on clinker portland doped with magnesia, J Mater Processing Tech 78 12-
17.
Beaudoin J. J. 1990. Handbook of Fiber-Reinforced Concrete – Principles, Properties, Developments and
Applications, Noyes Publications, New Jersey, US.
Caliskan S. 2003. Aggregate/mortar interface: influence of silica fume at the micro- and macro-level, Cem
Concr Compos 25 (4-5) 557-564.
Cooke A. M. 2005. The measurement and significance of green sheet properties for the properties of
hardened fibre cement, Cem Concr Compos 27 604-610.
Coutts R. S. P. 2005. A review of Australian research into natural fibre cement composites, Cem Concr
Compos 27 (5) 518-526.
DIN 274/1. 1972. Asbestzement-Wellplatten – Masse, Anforderungen, Prufungen, April 1972.
EN 494. 2004. Fibre-cement profiled sheets and fittings for roofing – Product specification and test methods,
December 2004.
Huang J . S., Cheng C. K. 2004. Fracture toughness variability of foamed alumina cements, Cem Concr Res
34 (5) 883-888.
Johnson L. G. 1951. The median ranks of sample values in their population with an application to certain
fatigue studies, Industrial Mathematics 2.
Kosmač T., C. Oblak, P. Jevnikar, N. Funduk, L. Marion. 1999. The effect of surface grinding and
sandblasting on flexural strength and reliability of Y-TZP zirconia ceramic, Dental Mater 15 426-433.
Lewis G., van Hooy-Corstjens C. S. J., Bhattaram A., Koole L. H. 2005. Influence of the radiopacifier in an
acrylic bone cement on its mechanical, thermal, and physical properties: Barium sulfate-containing cement
versus iodine-containing cement, J. Biomed Mater Res B 73B (1) 77-87.
Li G. Q., Cao H., Li Q. S., Huo D. 1993. Theory and its Application of Structural Dynamic Reliability,
Earthquake Press, Beijing.
Li Q. S., Fang J. Q., Liu D. K., Tang J.2003. Failure probability prediction of concrete components, Cem
Concr Res 33 (10) 1631-1636.
Lloyd D. K., Lipow M. 1962. Reliability: Management, Methods and Mathematics, Prentice Hall,
Englewood Cliffs, New Jersey.
Ma Y. P., Zhu B. R., Tan M. H. 2005. Properties of ceramic fiber reinforced cement composites, Cem Conc
Res 32 (2) 296-300.
Negro C., Blanco A., Fuente E., Sanchez L. M., Tijero J. 2005. Influence of flocculant molecular weight and
anionic charge on flocculation behaviour and on the manufacture of fibre cement composites by the
Hatschek process. Cem Concr Res 35 (11) 2095-2103.
Negro C., Blanco A., Pio I. S., Tijero J. 2006. Methodology for flocculant selection in fibre-cement
manufacture, Cem Concr Compos 28 (1) 90-96.
Peled A., Mobasher B. 2005. Pultruded fabric-cement composites, ACI Mater J 102 (1) 15-23.
Purnell P., Beddows J. 2005. Durability and simulated ageing of new matrix glass fibre reinforced concrete,
Cem Concr Compos 27 (9-10) 875-884.
Quinn G. 1990. Flexure strength of advanced structural ceramics: A round robin, J Am Ceram Soc 73 (8)
2374-2384.
ReliaSoft's Weibull ++. 1992. Life Data Analysis Reference, ReliaSoft Publishing.

128
Ritter J. E., Bandyopadhyyay N., Jakus K. 1981. Statistical reproducibility of dynamic and static fatigue
experiments, Ceram Bullet 60 798-806.
Savastano H., Warden P. G., Coutts R. S. P. 2003. Potential of alternative fibre cements as building materials
for developing areas, Cem Concr Compos 25 (6) 585-592.
Setien V. J., Armstrong S. R., Wefel J. S. 2005. Interfacial fracture tougness between resin-modified glass
ionomer and dentin using three different surface treatment, Dent Mater 21 (6) 498-504.
Studinka J. B. 1989. Asbestos substitution in the fibre cement industry, The International Journal of Cement
Composites and Lightweight Concrete 11 (2) 73-78.
Toutanji H. A. 1999. Evaluation of the tensile strength of cement-based advanced composite wrapped
specimens, Comp Sci Tech 59 (15) 2261-2268.
Vidovič K., Lovreček B., Hraste M. 1996. Influence of surface charge on sedimentation and filtration
behaviour of fibrous material, Chem Biochem Eng Q 10 (1) 33-38.
Weibull W. 1949. A statistical representation of fatigue failure in solids, Transactions of the Royal Institute
of Technology, No. 27, Stockholm.
Weibull W. 1951. A statistical distribution function of wide applicability, J Appl Mech 18 293-297.
Wu D., Zhou J., Li Y. 2006. Unbiased estimation of Weibull parameters with the linear regression method, J
Eur Ceram Soc 26 1099-1105.

129