LONG-TERM BEHAVIOUR OF GFRP PIPES: REDUCING
THE PREDICTION TEST DURATION
Hugo Faria, Rui Miranda Guedes
Faculdade de Engenharia da Universidade do Porto
Porto, Portugal
ABSTRACT
The certification of glass-fibre reinforced plastic (GFRP)
piping systems is regulated by normative standards in which test
series of 10000 hours are required to estimate the residual
properties at the end of the expected life (normally, 50 years).
In this paper, the possibility to reduce the tests duration,
maintaining an equivalent prediction of long-term properties, is
discussed. Experimental results from standard test procedures
conducted on GFRP pipes of four different types and respective
data analysis support this possibility. The estimation error when
using only data from shorter tests is consistently less than 10%
if compared to the standard methods.
INTRODUCTION
Glass-fibre reinforced plastic (GFRP) pipes have been
increasingly introduced in piping systems. They find attractive
applications in chemical industry, ducts, offshore, water supply
and sewage systems. However, the lack of fully understanding
the failure mechanisms and long-term materials performance
necessarily leads to over-design, in-service prototype
evaluations and, furthermore, inhibits greater utilization.
Moreover, the existence of different types of GFRP pipes
construction, namely filament wound, centrifugal cast and
hybrid ones, makes this task even more difficult. Thus, the
mechanical behaviour of GFRP pipes under ring deflection
and/or internal pressure is assessed in experimental procedures.
Two typical in-service load conditions are internal pressure
and ring deflection. For these, empirical test methods have been
developed and are described in the European Standards
EN1447 [1] (based on ISO10471-2) and EN1227 [2] (based on
ISO7509), respectively. The very long testing periods stated in
these standards, in addition to the factors mentioned above,
strongly discourage the industrial improvement and innovation
of the products and also prevent the end users from performing
confirmation tests. Shorter but reliable tests and/or better
predictive models, capable of being standardized, have been
required by the GFRP piping industry.
The few research works conducted specifically on long-
term properties of GFRP pipes between 1970 and 2000 gave
relevant information but the conclusions could only be applied
to classical filament wound GFRP pipes since all the
experiments have been only performed on these [3-7]. More
recently, several alternative short-term test methods for the
various loading conditions and different GFRP pipes
construction types have been studied in a co-normative
European research project [8]. Ultimate elastic wall stress
(UEWS) and strain at failure tests have been studied as
alternative methods for estimation of long-term pressure and
dynamic loading, UEWS and relaxation for the case of long-
term ring deflection. In some of these procedures it was
additionally considered a period of preconditioning under water
previously to the tests. In fact, the slowness of the liquid
diffusion at room temperature is a major aspect that is not taken
into account in the existing standard test methods. This
phenomenon has been recently investigated [9-10]. Other
limitations of the existing standard methods are the implicit
assumption that the mechanisms responsible for the long-term
failure are the same at different levels of load and the non
inclusion of material variability parameters that would damp the
scatter typically observed in these tests. Although these
problems are not explicitly considered in the standards, the
experimental evidence of the whole failure phenomenon and the
lack of adequate information was the main reason to extend the
testing periods over 10000 hours. In a logarithmic time scale,
this period is only 1.5 decades distant from 50 years and this
makes the existing test and prediction methods seem
reasonable. Any eventual phenomena occurring after 10000
hours contributing to unpredictable decrease in properties are,
however, ignored in the actual standard analyses.
Polymeric materials exhibit a time dependent behaviour
whose degree of linearity or nonlinearity is, often, not clearly
determined and several modelling approaches have been
1
presented through the years [11-15]. In most of these predictive see two different setups used in the experimental tests for ring
models, the governing laws include parameters to be taken from deflection property analysis. The same ring loading condition is
experimental data. However, there are no established methods assured in two different admissible ways.
to predict long-term properties of GFRP pipes based on
experimental data from short-term tests.
The European Standard EN705 [16], which is based on
ISO/TC138/SC6/WG1/N197, ISO10928, describes procedures
for the regression analysis of test data, normally with respect to
time. These extrapolation techniques are used to extend the
trend from data gathered over a period of approximately 10000
hours, to a prediction of the property at 50 years. The method
used in this analysis is transcribed in Appendix A. This work
follows methods described in standards EN1447, EN1227 and
EN705 to predict the long-term pressure and ring
load/deflection properties from the tests data.
In the next sections, tests results from two groups of Fig. 2 – Apparatus for EN1227 test procedures
standard experimental test series according to EN1447 and
EN1227 led on GFRP pipes specimens of four different types
are presented. Test procedures are briefly described, results data In these test series GFRP pipes of four different
and their regression analyses are summarized. Conclusions are construction types, from four main European manufacturers,
retrieved from this study. were used. The characteristics of the series of specimens used
are indicated in table 1.
TEST PROCEDURES AND RESULTS
Both test procedures specified in EN1447 and EN1227
relate to creep behaviour of GFRP pipes and state constant Table 1 – Main specifications of GFRP pipe specimens used
loading conditions to be imposed at different levels and during in the experimental test series
different periods of time to each specimen, from few minutes up GFRP pipe type
to 10000 hours. The determination of the long-term resistance A B C D
to internal pressure (EN1447) is done by imposing hydrostatic internal pressure specimens
internal pressure to the specimens using axial constrained-free nominal diameter (DN) [mm] 300 300 300 300
end sealing devices. Failure must occur within a determined nominal stiffness (SN) [N/m2] 5000 5000 5000 5000
nominal pressure (PN) [bar] 10 10 10 10
valid failure zone (in the middle of the specimen) for the test to
specimen length (L) [mm] 1300 1300 1300 1300
be validated. Time to failure is registered. In figure 1 the testing hybrid
apparatus used in the experimental tests for internal pressure filament filament (filament
centrifugal
property analysis is depicted. construction type winding winding winding +
casting
(epoxy) (UP) mat
deposition)
ring deflection specimens
nominal diameter (DN) [mm] 500 *** 500 500
nominal stiffness (SN) [N/m2] 10000 *** 10000 10000
nominal pressure (PN) [bar] 10 *** 10 10
specimen length (L) [mm] 300 *** 300 300
hybrid
filament (filament
centrifugal
construction type winding *** winding +
casting
(epoxy) mat
deposition)
Missing matrix and reinforcement materials!!!!
Fig. 1 – Apparatus used in the EN1447 test procedures
The properties to analyze in each of the testing procedures
are the internal pressure (EN1447) and ring load or initial
For the analysis of the ring deflection creeping behaviour deflection (EN1227) as function of the respective failure times.
(EN1227) the pipe specimens are subjected to a constant ring Additional interesting data may be assessed in these test series.
load and the evolving ring deflection is measured during the The exhaustive list of results of the test series conducted on
test. The complete setup is immersed in water at room different GFRP pipe types are tabled in Appendix B. The results
temperature. Time to failure is registered. In figure 2, one can from testing procedures according to EN1447 are in table B1,

2
presenting pressure vs failure time. There were valid tests for GFRP pipe types A, C and D. Trend lines are also shown for
all four sets of GFRP pipes. The results from testing procedures each set of tests.
according to EN1227 are in tables B2 and B3, presenting ring
load vs failure time and initial relative ring deflection vs
failure time, respectively. In those, only tests ran on GFRP 1,8
pipes of types A, C and D were validated. Within these, initial A
C
relative ring deflection data was only available for types C and D
1,6
D.
All tests showed considerable scatter in results which is

Ring Load [log kN]

1,4
mainly justified with the typical material variability of these
structures. Different levels of admissible load and different
trends of results were also observed for the different types of 1,2

GFRP pipes construction. However, the experimental procedure

have shown to be applicable to all these different types. Graphic 1
representation of data is shown in next section together with
their regression analysis.
0,8
-2 -1 0 1 2 3 4 5
DATA REGRESSION ANALYSIS Time to Failure [log h]

The objective here is to predict the value of internal Fig. 4 - Results of test series according to EN1227: ring
pressure (EN1447) and ring load or initial ring deflection load vs failure time
(EN1227) that lead to failure times of 50 years (438000 hours).
These are called the long-term properties and are assessed by
extrapolation of the test data. In figure 5 ring-deflection tests results are presented in
The European Standard EN705, which is referred in the terms of initial relative ring deflection vs failure time in a log-
standards EN1447 and EN1227 to which these experimental log scale for the two GFRP pipe types C and D. Trend lines are
data presented relate, specifies the regression and extrapolation also shown for each set of tests.
methods to apply. Namely, for experimental data showing a
skewed distribution, the covariance method (Method A of
EN705) is the one to use. This method is transcribed in 1,4
Appendix A and fits a linear trend line to the experimental data C
D
when represented in a logarithmic time scale.
Initial Relative Ring Deflection [log %]

In figure 3 hydrostatic pressure tests results, pressure vs 1,2

failure time, are represented in a log-log scale for the four

GFRP pipe types. Least squares based trend lines are shown for
1
each set of tests, respecting to each type of specimen.

0,8
2
A
B
1,8 C 0,6
D -2 -1 0 1 2 3 4 5
Pressure [log bar]

Time to Failure [log h]

1,6

Fig. 5 - Results of test series according to EN1227: initial

1,4
relative ring deflection vs failure time

1,2
Suitability of each set of tests data for extrapolation was
evaluated accordingly to the covariance method for regression
1
analysis (Method A of EN705) and intermediate statistical
-2 -1 0 1 2 3 4 5
Time to Failure [log h]
parameters were calculated. The full listing of these is presented
Fig. 3 - Results of test series according to EN1447: in Appendix B. The lower the values of the property being
pressure vs failure time analyzed are, greater is the relative scatter allowed in this
In figure 4 ring-deflection tests results are presented in method , due to the method applied to calculate the coefficients
terms of ring load vs failure time in a log-log scale for the three of correlation.

3
 The linear relations fitted to each set of tests data fulfilling type C ( )
log F = 1.14545 − 0.02636 log tu ( ) (14)
the standard method are given in the following equations
type D log (F ) = 1.23424 − 0.0931 log (tu ) (15)
initial rel. ring deflection vs failure time regression lines1
pressure vs failure time regression lines
type A log (P ) = 1.76567 − 0.00836 log (tu ) (1)
type C ( )
log y0 d m = 1.27016 − 0.04019 log (tu ) (16)
log ( y0 d m ) = 1.20304 − 0.10116 log (tu )
type B ( )
log P = 1.89891 − 0.04132 log tu( ) (2)
type D (17)

type D log (P ) = 1.59978 − 0.03158 log (tu ) (3)

Missing type C, please justify!!! The prediction of the long-term properties using these two
estimation approaches - the standard procedure and the
alternative procedure considering only data from shorter tests
( tu ≤ 1000h ) - present results of the same order that can be
ring load vs failure time regression lines directly compared. The maximum error found was of 10.5%
type A log (F ) = 1.50787 − 0.0377 log (tu ) (4) (overestimation or what??). These comparisons are summarized
type C ( )
log F = 1.15358 − 0.032 log tu( ) (5) in tables 2, 3 and 4.
In order to retrieve the maximum information from the
type D log (F ) = 1.23758 − 0.0955 log (tu ) (6) available test data, extrapolation and analysis of data was
applied to all the series that collected sufficient information.
initial relative ring deflection vs failure time regression This included series that did not accomplish the standard
lines criteria for suitability of data for analysis and/or extrapolation
type C ( )
log y0 d m = 1.2739 − 0.04037 log (tu ) (7)
mainly because of the low number of tested specimens.

type D ( )
log y0 d m = 1.23758 − 0.0955 log (tu ) (8)
Table 2 - Comparison of estimations on pressure leading to
50 years life by two methods: a) using complete data from
where P is the applied pressure in bar, F is the applied ring
valid tests and b) using data only from tests shorter than
load in kN, y d is the initial relative ring deflection in % 1000 hours
0 m
GFRP pipe type
and tu is the failure time in hours.
A B C D
Following the alternative methodology suggested in this suitability of data
work, the covariance method (Method A of EN705) was applied for analysis2 yes yes no 3 no 3
to the experimental data but considering only the valid tests that (EN705)
suitability of data
presented failure times, tu , shorter than 1000 hours. Again, for extrapolation4 yes yes n/a n/a
intermediate statistical parameters were calculated and are fully (EN705)
extrapolated long-
listed in Appendix B. term value for
The linear relations between the same parameters fitted to internal pressure 52.3 bar 46.3 bar *** 26.4 bar
each set of tests considering only the shorter tests are given in using complete data
the following equations from valid tests
extrapolated long-
term value for
internal pressure
pressure vs failure time regression lines1 using only data from
53.6 bar 44.1 bar 7.5 bar 29.5 bar
type A log (P ) = 1.76212 − 0.00578 log (tu ) (9) tests shorter than

( ) ( )
1000h
type B log P = 1.8976 − 0.0449 log tu (10) relative error 2.4 % 5.0 % *** 10.5 %
type C log (P ) = 1.32149 − 0.13189 log (tu ) (11)
type D log (P ) = 1.57806 − 0.01932 log (tu ) (12) One must note that no special attention was given to the
distribution of failure times within the results of shorter tests. In
ring load vs failure time regression lines1
type A log (F ) = 1.50398 − 0.04343 log (tu ) (13)
2
evaluated in function of correlation coefficient (see Appendix A)
3 2
minimum value for r extrapolated from table A1
4
1
considering only data from valid shorter tests (tu≤1000h) evaluated in function of Student’s t (see Appendix A)

4
the alternative methodology, all valid tests with tu ≤ 1000h would maintain equivalent prediction of the long-term
properties. (This is not very clear, please explain better!)
were considered. In each set of the experimental tests conducted
Further research on higher order relations to be fitted to
these were the majority of the tests.
experimental data by regression analysis could bring even more
confidence to the shortening of the tests duration. Other long-
Table 3 - Comparison of estimations on ring load leading to
50 years life by two methods: a) using complete data from
term properties should also be studied in the same scope.
valid tests and b) using data only from tests shorter than After analysis of the experimental data gathered from
1000 hours extensive standard testing procedures led on four different types
GFRP pipe type of GFRP pipes it was observed that shorter tests could be
A C D reliably considered in the certification process of these
suitability of data for analysis composite structures. This certainly would promote faster
(EN705)
yes yes yes product developments and innovation in this industry.
suitability of data for
extrapolation (EN705)
yes yes yes
ACKNOWLEDGMENTS
extrapolated long-term value for
ring load using complete data 19.7 kN 9.8 kN 5.0 kN The test series hereby presented were conducted within two
from valid tests research programmes supported by the Growth Programme of
extrapolated long-term value for
the European Comission and the Portuguese Foundation for
ring load using only data from 18.2 kN 9.9 kN 5.1 kN
tests shorter than 1000h Science and Technology through projects G6RD-CT2000-
relative error 8.2 % 1.0 % 2.0 % 00259 and POCTI/EME/47734/2002, respectively. I
acknowledge André Vieira and João Reis for the design of two
testing apparatus that supported experimental tests conducted at
Table 4 - Comparison of estimations on initial relative ring INEGI. I also acknowledge Catherine Hervé (CETIM) for the
deflection leading to 50 years life by two methods: a) using compilation of pressure tests data from project partners.
complete data from valid tests and b) using data only from
tests shorter than 1000 hours REFERENCES
GFRP pipe type
[1] EN1447:1996 - Plastics piping systems - Glass-reinforced thermosetting
C D plastics (GRP) pipes - Determination of long-term resistance to internal
suitability of data for analysis (EN705) no yes pressure, CEN, 1996.
suitability of data for extrapolation (EN705) n/a yes [2] EN1227:1995 - Plastics piping systems - Glass-reinforced thermosetting
extrapolated long-term value for initial relative plastics (GRP) pipes - Determination of the long-term ultimate relative
ring deflection using complete data from valid 11.1 % 4.2 % ring deflection under wet conditions, CEN, 1995.
tests [3] H. J. M. A. Mieras, Irreversible creep of filament-wound glass-
reinforced resin pipes, Plastics & Polymers, Vol.41, pp.84-88, 1973.
extrapolated long-term value for initial relative
[4] W. S. Carswell, The behaviour of glass filament wound pipes under
ring deflection using only data from tests shorter 11.1 % 4.3 % internal pressures, Proceedings of the 2nd International Conference on
than 1000h
Composite Materials, pp. 472, 1976.
relative error 0.6 % 1.8 % [5] J. C. Lenain, A. R. Bunsell, The behaviour of GRP pipes under cyclic
pressurization, Composites, Vol.9, no.2, pp.77-82, 1978.
[6] I. Ghorbel, P. Spiteri, Durability of closed-end pressurized GRP pipes
under hygrothermal conditions. Part I: Monotonic tests, Journal of
CONCLUSIONS Composite Materials, Vol.30, no.14, pp.1562-1580, 1996.
[7] I. Ghorbel, Durability of closed-end pressurized GRP filament wound
The alternative methodology for the prediction of the long- pipes under hygrothermal conditions. Part II: Creep tests, Journal of
Composite Materials, Vol.30, no.14, pp.1581-1595, 1996.
term properties (pressure and ring deflection) of GFRP pipes [8] GRP Design & Test, Project Contract No. G6RD-CT2000-00259,
that was presented here showed results that are considerably European Comission - Growth Programme, 2000-2005.
similar to those achieved by the actual standard methods. Errors [9] J. Yao, G. Ziegman, Water Absorption Behavior and Its Influence on
within the intervals of 2.4%-10.5%, 1.0%-8.2% and 0.6%-1.8% Properties of GRP Pipe, Journal of Composite Materials, Vol.41, no.8,
pp.993-1008, 2007.
were found in the predictions of the long-term pressure, the
[10] R. M. Guedes, A. Sá, H. Faria, Influence of moisture absorption on creep
long-term ring load and the long-term initial relative ring of GRP composite pipes, Polymer Testing, Vol.26, pp. 595–605, 2007.
deflection, respectively. [11] Yu. N. Rabotnov, L. K. Papernik, Nonlinear creep of TS8/3-250 glass-
Despite the considerable scatter in the tests direct results, reinforced plastic, Translated from Mekhanika Polimerov, No.3, pp.391-
the statistical regression analysis determined no relevant 397, 1971.
[12] W. N. Findley, J. S. Lai, K. Onaran, Creep and relaxation on nonlinear
differences between using complete data (including tests up to viscoelastic materials, Dover Publications, New York, 1989.
10000 hours) or using only data from shorter tests (up to 1000 [13] R. Schapery, A theory of mechanical behaviour of elastic media with
hours). This suggests that the standard procedures could growing damage and other changes in structure, Journal of Mechanics
consider shorter test times and include weighted factors that and Physics of Solids, Vol.38, no.2, pp. 215–253, 1990.

5
[14] Li Zhang, Time-dependent behaviour of polymers and unidirectional
polymeric composites, PhD Thesis, Delft University of Technology,
1995.
[15] F. Richard, D. Perreux, The safety-factor calibration of laminates for
long-term applications: bevior model and reliability method, Composites
Science and Technology, Vol. 61, pp. 2087-2094, 2001.
[16] EN 705:1994, Plastics piping systems - Glass-reinforced thermosetting
plastics (GRP) pipes and fittings - Methods for regression analysis and
their use, CEN, 1994.

6
 APPENDIX A

STANDARD METHOD FOR THE ANALYSIS OF TEST DATA

In this Appendix, a standard procedure for the analysis of The variables Q y , Q x and Q xy are assessed from the
data gathered from experimental test series led on GFRP pipes
is transcribed. following relations:
The European Standard EN 705 [16] specifies procedures
suitable for the analysis of data which, when converted into
logarithms of the values, have either a normal or a skewed
Qy =
(
∑ yi − Y )2 (A4)
distribution. It is intended for use with the test methods for the
n
analysis of properties as a function of, usually, time. These
extrapolation techniques typically extend the trend from data
Qx =
(
∑ xi − X )2 (A5)
gathered over a period of approximately 10000hours to a
n
[( )( yi − Y )]
prediction of the property at 50 years. The covariance method
(method A of EN705), transcribed hereafter, is applicable to ∑ xi − X
Q xy = (A6)
experimental data with a skewed distribution, which is the case n
under analysis.

Covariance Method Using a First Order Relationship
This method is applied to tests results having a skewed
distribution to fit a straight line of the form
where Q y is the sum of the squared residuals parallel to the y
axis divided by n , Q x is the sum of the squared residuals
parallel to the x axis divided by n , Q xy is the sum of the

y = a + bx (A1) squared residuals perpendicular to the line, divided by n , xi

and yi are individual values, n is the total number of results
(pairs of readings for xi , yi ). Y and X are, respectively, the
where y is the logarithm (log) of the property being
investigated, a is the intercept on the y axis, b is the slope, arithmetic means of the y and x data, i.e.
x is the logarithm (log) of the time, in hours.
The suitability of data for analysis is evaluated by
2 ∑ yi
calculating the squared, r , and the linear coefficient of Y = (A7)
correlation, r , as follows: n
∑ xi
X = (A8)
n
2
2 Q xy
r = (A2)
Qx Q y
NOTE: If the value of Q xy is greater than zero the slope of the line is positive
2 Q xy and if it is less than zero then the slope is negative.
r= r = (A3)
Qx Q y
2
If the value of r or r is less than the applicable minimum
value given in table A1 as a function of n , data should be
considered unsuitable for further analysis.

7
 Table A1 - Minimum values for correlation coefficients ' '
where xi and yi are the best fits for each true xi and yi ,
respectively.
The suitability of data for extrapolation is evaluated by
comparing the absolute value of T , T , which is given by

b
T = (A10)
C

with the applicable value for Student’s t , t v , shown in table A2

for n − 2 degrees of freedom. If T ≥ t v then data is considered
suitable for extrapolation.

The coefficients a and b of equation (A1) are calculated Table A2 - Percentage points of Student’s t distribution
as follows:
(upper 2.5% points; two sided 5% level of confidence; tv
for 97.5%)

b=− Γ (A9)
a = Y − bX (A10)

where Γ is the ratio between Q y and Q x , i.e.

Qy
Γ= (A11)
Qx

The variance, C , of the slope, b , is calculated as follows:

2
2 Γb σ s  bσ s2 
C = 1 +  (A12)
 2 Q xy 
n Q xy
 

2
where σ s is the error variance for x , i.e.

( ' 2
) ' 2
2 ∑ yi − yi + Γ ∑ xi − xi
σs =
( ) (A13)
(
n−2 Γ )

8
 APPENDIX B

EXPERIMENTAL TESTS RESULTS AND DATA SUITABILITY ANALYSIS FOR REGRESSION AND EXTRAPOLATION

In this Appendix, the exhaustive lists of tests results from 26,817 1930 11,4113 80,5 12,4 144
the standard test series, accomplishing the procedures stated in 28,092 1303,4 11,1398 341 12 43
EN1447 and EN1227, conducted on four different types of 28,822 448,7 12,4008 55 11,8 299
GFRP pipes, are presented. Intermediate data for regression and 31,6 0,19 12,4008 94 11,1 403
32 0,16 13,0883 28,9 10,5 531
extrapolation suitability analyses, calculated accordingly to
30,588 548,9 13,9885 119,5 9,8 840
method A of EN705 (transcribed in Appendix A), is also
26 110 9 2226
presented. 26 129 8,5 1210
21,9 198 8 6007
22,6 829 9,5 1178
Table B1 – Results of the test series according to EN1447 9 1641
led on four GFRP pipe types
GFRP pipe type
A B C D Table B3 – Results of the test series according to EN1227
P P P P led on two GFRP pipe types considering the initial relative
tu [h] tu [h] tu [h] tu [h]
[bar] [bar] [bar] [bar] ring deflection applied
56 13 84,12 0,06667 52 0,1 36,3 0,112
GFRP pipe type
58 0,1 58,12 30,87 45 1,17 36,3 4,5
57,5 19 95 0,03333 40 0,52 37,6 6 C D
57 125 95 0,06667 35 15,53 38 23 y0/dm y0/dm
tu [h] tu [h]
[%] [%]
56 216 78,5 1,88 23 144,45 36,1 53
16,0389 314 11 7,2
55 830 80 2,24 20 273 34,3 178
18,1021 178 9,5 87
56,5 436 79,7 1,34 20 493 32,4 794
14,158 3162 8,84 190
55,5 4218 76,78 8,29 16 321,47 36,7 1072
17,2047 120 6,29 1700
54,5 1973 72,6 0,03333 32 7943
20,3543 0,1 18,197 0,22
56 392 68,43 29,75 25,1 10000+
12,1114 5623 15,3109 0,62
57 13 75,24 27,37 29,9 10000+
17,5256 892,5 15,417 3,2
54 6463 65 522
16,4144 1512 14,7911 4,75
53 2781 60 1793
13,5728 10 12,5893 9,64
54 9136 60 1882
18,6024 0,03 13,8038 13
53 10000+ 57 2578
14,7244 174 11,6145 21,16
57 3258
15 174 10,9648 62,37
16,7717 47 10,2329 144,1
15,4134 173 10,4713 43,05
Table B2 – Results of the test series according to EN1227 14,2087 2162,7 9,66051 299,7
for three GFRP pipe types considering the ring load applied 17,815 2,5 8,91251 402,57
GFRP pipe type 13,576 3162 8,41395 531,22
A C D 14,4789 80,5 7,94328 839,91
F [kN] tu [h] F [kN] tu [h] F [kN] tu [h] 13,732 341 7,4131 2226,4
41,049 0,04 12,1963 314 9,5 7,2 14,5779 55 7,07946 1209,96
36,72 0,1 10,6439 3162 9 87 15,5466 94 6,0256 6007,4
34,98 0,1 12,7199 120 8,5 190 16,1668 28,9 11,6145 1177,9
34,98 0,1 13,9703 0,1 7,1 1700 15,3176 119,5 11,6145 1641
28,86 0,05 14,4204 0,03 17,1 0,22
33,24 0,23 11,8314 174 17 0,62
32,1 0,43 12,087 174 16,5 3,2
31,11 3,63 12,8309 47 16 4,75 Please change the commas to points. The number of significant
28,95 31,4 12,6928 173 15 9,64 digits must be uniform!!!
31,11 8,25 10,3806 2162,7 15 13
28,606 4,84 13,6383 2,5 13 21
25,16 4300 10,4312 3162 12,5 62

9
 Table B4 - Variables calculated accordingly to Method A Table B6 - Variables calculated accordingly to Method A
(Covariance Method) of EN705 for the four types of pipes (Covariance Method) of EN705 for GFRP pipes of types A, C
using complete data from valid long-term pressure tests and D using complete data from valid long-term creep in
(EN1447) wet conditions tests (EN1227); the testing parameter
Value considered is the constant ring load
Parameter Value
A B C D Parameter
n 14 16 *** 11 A C D
X 2.37325 1.05838 *** 2.22955 n 22 18 23

Y 1.74583 1.85518 *** 1.52936 X 0,91955 1,99992 1,95044

Qx 1.742368 3.08708 *** 2.67062 Y 1,47321 1,08957 1,05138

Qx 2,92073 1,74719 1,39083

Qy 0.00012 0.00527 *** 0.00266
Qy 0,00415 0,00179 0,01268
Qxy -0.01201 -0.11213 *** -0.06419
Qxy -0,08558 -0,04569 -0,1133
r2 0.67974 0.77272 *** 0.57914
2
r 0,60403 0,66776 0,72813
r -0.82446 -0.87905 *** -0.76101
Γ 6.99E-05 0.00171 *** 0.001 r -0,7772 -0,81717 -0,8533

b -0.00836 -0.04132 *** -0.03158 Γ 0,00142 0,00102 0,00911

a 1.76567 1.89891 *** 1.59978 b -0,0377 -0,03200 -0,0955

σ 2
0.35683 0.42674 *** 0.78008
a 1,50787 1,15358 1,23758

σ s2
s
0,71583 0,35938 0,22346
C 2.79E-06 3.62E-05 *** 8.3E-05

T 5.00669 6.86765 *** 3.46731

C 4,7E-05 3,23E-05 0,00016

T 5,48921 5,63545 7,47144

Table B5 - Variables calculated accordingly to Method A

(Covariance Method) of EN705 for the four types of pipes Table B7 - Variables calculated accordingly to Method A
using only data from valid tests (EN1447) with failure times (Covariance Method) of EN705 for GFRP pipes of types A, C
under 1000 hours and D using only data from valid tests (EN1227) with failure
Value times under 1000 hours; the testing parameter considered
Parameter is the constant ring load
A B C D
n 9 12 8 7 Value
Parameter
A C D
X 1,676651 0,290134 1,221404 1,245
n 19 15 18
Y 1,752425 1,884574 1,460406 1,554
X 0,53663 1,59632 1,58234
Qx 1,319587 1,751594 1,838253 1,340136
Y 1,48067 1,10337 1,087
Qy 4,41E-05 0,003531 0,031974 0,0005 Qx 2,29928 1,43890 1,13509

Qxy -0,005829 -0,058817 -0,232016 -0,016216 Qy 0,00434 0,0001 0,00983

2 0,583959 0,55941 0,915877 0,392198
r Qxy -0,07746 -0,02806 -0,0833
r -0,764172 -0,747937 -0,957015 -0,626257
Γ 3,34E-05 0,002016 0,017394 0,000373 r2 0,60161 0,54738 0,62168

b -0,00578 -0,044896 -0,131885 -0,019321 r -0,77564 -0,73985 -0,7885

a 1,762116 1,897599 1,621491 1,578055

Γ 0,00189 0,0007 0,00866
b -0,04343 -0,02636 -0,0931
σ 2
s
0,400109 0,529814 0,105357 0,701213
a 1,50398 1,14545 1,23424
C 3,53E-06 0,000163 0,000268 0,000126
σ s2 0,57657 0,43191 0,27012
T 3,076329 3,512962 8,052912 1,718936
C 7,5E-05 4,52E-05 0,00033

T 5,0295 3,920253 5,09012

10
 Table B8 - Variables calculated accordingly to Method A NOMENCLATURE
(Covariance Method) of EN705 for GFRP pipes of types C
and D using complete data from valid long-term creep in
wet conditions tests (EN1227); the testing parameter a - coefficient of equation of the linear regression line
considered is the initial relative ring deflection b - slope of the linear regression line
Value C - variance of the slope b
Parameter F - applied ring load [kN]
C D
n 23 23 Γ - ratio between Q y and Q x
X 1,99992 1,95044 n - number of tested specimens
Y 1,19317 1,01444 P - applied pressure [bar]
Q x - sum of the squared residuals parallel to the x axis divided by n
Qx 1,64652 1,39083
Q xy - sum of the squared residuals perpendicular to the regression line,
Qy 0,00268 0,01554
divided by n
Qxy -0,04339 -0,12806
Q y - sum of the squared residuals parallel to the y axis divided by n
2 0,42604 0,75882
r r - coefficient of correlation
r 2
-0,65272 -0,87110 σ s - error variance for failure times data
Γ 0,00163 0,01117
T - ratio between the slope, b , and the square root of its
b -0,04037 -0,10570
tu - failure time [h]
a 1,27390 1,22061

σ s2 0,62626 0,19635
X - arithmetic mean of the failure times, tu , data
Y - arithmetic mean of the property value
C 0,00011 0,00017
y d - initial relative ring deflection [%]
0 m
T 3,90926 8,10188

Table B9 - Variables calculated accordingly to Method A

(Covariance Method) of EN705 for GFRP pipes of types C
and D using only data from valid tests (EN1227) with failure
times under 1000 hours; the testing parameter considered
is the initial relative ring deflection
Value
Parameter
C D
n 18 18
X 1,59632 1,58234
Y 1,20601 1,04298

Qx 1,3445 1,13509

Qy 0,00217 0,01162

Qxy -0,02945 -0,11033

r2 0,297 0,92329
r -0,54498 -0,96088
Γ 0,00162 0,01023
b -0,04019 -0,10116
a 1,27016 1,20304

σ 2
s
0,68825 0,04996

C 0,00025 5,33E-05

T 2,55333 13,86006

11