Las Vegas 2009. Weep Failure

Last revised on December 22, 2009
Structural failure of composite pipes – a trilogy
Part 2 – Weep failure
Presented at the 2009 ACMA conference in Las Vegas

Antonio Carvalho
Abstract – The long-term structural design of composite pipes for infrastructure applications is addressed from
the perspectives of rupture and weep. The rupture mode of failure is controlled by the glass fibers – not by the
resin - and typically occurs in pipes operating under low strains. The weep failure is controlled by the resin –
not by the fibers – and typically occurs in pipes under high strains. These two modes of failure are driven by
separate mechanisms and are quite distinct and independent. There is no interaction between the weep and
rupture modes of failure. There is still a third kind of structural failure, known as strain-corrosion rupture, which
occurs when the composite is subjected to bending strains while immersed in corrosive environments. The
strain-corrosion failure, however, can be fully explained as a combination of the weep and rupture modes. The
structural design of composite pipes for long-term durability in corrosive or non-corrosive environments must
meet both rupture and weep criteria.
The commercial standards for composite pipes, like AWWA C 950 and ISO 14692, are focused exclusively on
the weep and strain-corrosion modes of failure, and ignore the long-term burst failure. In addition, these
standards are based on long-term HDB values that may be misleading for long-term design. We have developed
a series of three papers addressing each of these modes of failure. The first paper recognizes the burst type
failure which is caused by hydrolysis of the UD glass fibers. The second paper addresses weep failure by
introducing the concept of a threshold strain in place of the traditional HDB. And finally the third paper
combines these two concepts to explain and quantify the elusive phenomenon of strain corrosion rupture.
Introduction – This is the second paper in the trilogy. It opens with a discussion of the classical weep
regression line obtained from pressure testing water filled pipes. Next it identifies the resin as the controlling
factor for weep and introduces the concept of threshold strain. The threshold strain and its measurement are the
focus of this paper. The concept of threshold strain introduces two substantial innovations in composite pipe
design. First, it indicates that the weep resistance of the composite pipes is higher than implied by the classical
HDB. This improved performance is noted mostly in high temperature and cyclic applications. And last but not
least, it relieves the pipe manufacturers of the burden to do the expensive tests required by ASTM D 2992.
Composites pipes may fail by weeping or by rupture. Weep failure is controlled by the resin, while rupture is
controlled by the glass fibers. These modes of failure have distinct causes. Weeping results from the growth of
resin cracks, which coalesce and merge to allow the passage of water. Rupture results from the slow
deterioration of the glass fibers in contact with water. Which event occurs first will depend on the hydrolytic
stability of the glass, the toughness of the resin and the operating strain. As a general rule, if the strain is low the
pipe will fail by burst instead of by weep. This statement may sound heretic and preposterous, but we hope to
make it clear as we go. The burst failure was addressed in the first paper in this series (reference 7) under the
assumption that the pipe will not weep throughout its structural life. This condition is fulfilled if the operating
strain is low.
Av. Amazonas, 1100, Brás Cubas, Mogi das Cruzes, SP, Brasil, 08744-340, tel.: 55-11-4795-8205, www.reichhold.com.br
This paper, the second in the trilogy, will focus on the strain that leads to weeping, known in the industry as
HDB, or Hydrostatic Design Basis. The test protocol currently used to estimate the HDB is described in ASTM
D 2992. It consists essentially in pressure testing a minimum of 18 water filled pipes under different sustained
high strains. The strains and times to failure are annotated and combined on a regression line that is used to
predict the long-term failure strain, which could be 20 years (ISO 14692) or 50 years (AWWA C 950). The
extrapolated long-term strain is known as Hydrostatic Design Basis, or HDB. This paper will show that the
regression line established this way is valid to predict the short-term (3 to 5 years) weep failure of pipes, but
cannot be used to make long-term predictions. The pages that follow build a strong case against the long-term
HDB and its use in the design of composite pipes.
This paper will argue that under low strains the resin will not crack and the pipes will not weep. The strain that
defines the onset of resin cracks is called the threshold strain. The concept of threshold strain is in sharp contrast
to the current practice of extrapolating the weep regression line to predict the long-term HDB. The classical
assumption that the resin cracking observed in testing pipes at high strains occurs also at low strains is simply
untenable. It is our contention that the classical weep line developed for pipes under high strains is not valid to
predict weep failures at low strains. Specifically we will show that (a) the regression line changes its slope and
becomes horizontal for strains below the threshold limit and (b) no weeping occurs if the strains are kept below
that limit. In other words, we propose that the actual weep line is bi-linear and not mono-linear as implied in the
classical approach.
Although originally defined for pipes designed to transmit water under static pressure, the concept of the HDB
can be expanded to include cyclic loads as well, implying the existence of a static as well as of a cyclic HDB.
This paper will deal with both the static and cyclic cases at room temperature and at elevated temperature.
The current test method – The current test method to develop the weep regression line of composite pipes is
described in ASTM D 2992. The method consists in subjecting several water filled pipe specimens, each to a
constant pressure (cyclic or static) and annotating the times to weep. To accelerate the testing, the pipe
specimens are subjected to very high strains, much higher than those found in use. The data points consisting of
the strains and the times to weep are plotted on log x log space to produce a regression line which is
extrapolated to yield the HDB. Equation (1) shows the weep regression line on log x log space. The Greek letter
“Є” denotes the hoop elongation that leads to weep in the time “t”. The intercept A is related to the short-term
elongation at weep of the pipe and the slope G measures the rate of coalescence of the cracks that leads to weep
in the time “t”. The parameters (A and G) depend on the temperature and the toughness of the controlling resin
(the interphase resin in oil pipes and the matrix resin in sanitation pipes). Details on the interphase and matrix
resins are coming soon.
log ε = A − G log t (1)
As an example, the constant hoop strain that fails the pipe in 50 years of continuous use, known as the HDB for
50 years, is obtained by entering the time 50 years in equation (1).
log HDB = A − G log 50 (1A)
To properly understand and interpret the weep line, it is necessary to know how the commercial pipes are
constructed. Figure 1 is a representation of typical longitudinal cross-sections of pipes used in sanitation and oil
applications, showing the sequence of layers that are used in their construction. What follows is a description of
those layers.
Liner: Both sanitation and oil pipes have an inner resin rich layer which, if not broken or otherwise punctured,
assures full water tightness. The liner of sanitation pipes is made by saturating a special lightweight veil with
polyester or vinyl ester resins. As a rule, the liners of oil pipes do not have this veil.
Weep barrier: All sanitation pipes have a layer of chopped glass, also saturated with polyester or vinyl ester
resins, placed immediately on top of the liner. This layer consists of one or more plies of chopped glass fibers
and serves two functions. First, it works as a corrosion barrier in pipes used to convey chemical products and
second it helps in the water tightness. This chopped glass layer is known as “corrosion barrier” or “weep
barrier” according to the application.
Structural layer: The structural layer of both types of pipes is made of several plies of high modulus
unidirectional, continuous glass fibers which are laminated by a process known as filament winding. The
unidirectional fibers (referred to as UD fibers) in the structural layer provide the modulus and strength required
to handle the pressure and other loads on the pipe. The UD plies crack easily along the fiber direction and are
not very efficient as a weep barrier.
Core: The sanitation pipes used underground may have a core of sand-filled resin which builds up the wall
thickness to provide ring stiffness. The core ply also cracks easily and is not good as a weep barrier.
OIL PIPES
SANITATION PIPES
UD ply
Core
UD ply
UD ply
Weep barrier Liner
Figure 1
Typical wall construction for sanitation and oil pipes.
Figure 2 shows typical weep regression lines for the two kinds of pipes just discussed. The line with the flatter
slope is typical of epoxy oil pipes that do not have a weep barrier of chopped glass. The line with the steeper
slope is typical of polyester sanitation pipes used to convey water, sewage and chemical effluents. All sanitation
pipes have a weep layer of chopped glass on top of the liner.
Figure 2
Typical weep regression lines for sanitation and oil pipes. When compared to the sanitation pipes, the oil pipes
have a flatter line.
We proceed to explain how the weeping mechanism relates to pipe construction. The first thing to have in mind
is that for the water to leak and the pipe to weep, the liner must be broken. This statement is obvious and seems
to indicate that pipes having highly resilient liners are protected against weep. Unfortunately this is not so,
because the cracking of the liner is controlled not by the liner itself, but by the ply immediately above it. For oil
pipes, in which the UD plies are laminated directly on top of the liner, it is the innermost UD ply that has the
control. And for sanitation pipes weeping is controlled by the chopped glass ply. These very unusual and
unexpected remarks are contrary to common sense and require explanation.
We need three hypotheses to explain why the cracks in liner are initiated and controlled by the substrate ply.
First, we assume perfect bonding of the liner to the substrate lamina. Second, we assume that the stiffness of the
liner is small when compared to the stiffness of the rest of the laminate. And third, we assume that the substrate
ply cracks before the liner itself. Stated in other words, we assume that the liner (a) cannot separate from the
substrate, (b) is too thin to control the deformation of the substrate and (c) develops its own micro cracks after
the substrate. These are reasonable assumptions that are certainly met in all practical situations.
Suppose now that the pipe is subjected to an increasing tensile force that stretches it until the first cracks
develop in the critical UD or chopped glass ply. The liner, made of a resilient resin, would crack after the UD or
chopped ply. However, even if the liner cracks last, it is the first to rupture. In the discussion that follows we
will try to explain why that is so. The cracking sequence is a direct consequence of our earlier hypotheses:
1. When the micro cracks first appear and then begin to widen in the substrate, two things happen.
2. First, the liner is not stiff enough to prevent these cracks from opening
3. And second, given its perfect bonding to the substrate, the opening cracks in the substrate induce
infinitely large strains on the liner.
4. Any liner, no matter how resilient, cracks under infinite strains.
5. Since the liner has no glass fibers to check the crack growth, the cracks grow and eventually fail the
liner.
6. The cracks that first originated in the UD or chopped ply are checked in their growth by the glass fibers.
7. The above explains how, regardless of its resiliency, the liner fails before the substrate.
This argument explains why the rupture of the liner is controlled by micro cracks that initiate in the substrate
ply immediately above it. This substrate ply is, for obvious reason, called the critical ply. For oil pipes the
critical ply is the innermost UD layer. And for sanitation pipes the chopped glass is critical. The argument just
presented has been observed and reported by many authors. For a quick review of this subject the reader is
referred to references 3 and 8. The elongations at break of liners embedded in laminates are sometimes called
“in-situ” elongations to differentiate them from those measured on isolated plies in which the liner is tested by
itself, not bonded to any substrate.
The “in-situ” elongation of the liner is determined by two properties of the substrate. The first property is the
resiliency of the resin. The reader will have to make an effort to remember that it is the resin in the substrate,
and not the resin in the liner, that controls the weep process. The second property, applicable to oil pipes only, is
the winding angle of the critical UD ply. We have said that the UD laminas are prone to cracking along the
glass fibers and that those cracks cause oil pipes to weep. Since we have defined the weep strains as strains in
the hoop direction, not in the direction transverse to fibers, the weep line of oil pipes depends on the winding
angle of the UD layers. The sanitation pipes, with their critical ply of isotropic chopped glass, always have the
same weep line, regardless of construction.
We close this section with a few remarks regarding the parameters A and G in the regression equation (1). The
slope G reflects the rate of crack growth under constant strain. It increases with the operating temperature and
decreases with the resiliency of the resin. Figure 3 shows the slope of weep lines of oil pipes increasing from G
= 0.03 at 25C to G = 0.07 at 65C.
Figure 3
The slope of the regression line depends on the resiliency of the resin and the temperature
Table 1 shows some typical values for the static intercepts A and slopes G of commercial sanitation and oil
pipes at several temperatures. The intercept A has been adjusted to give the elongation in % when the time is
expressed in hours. Note the large dependence of G on the operating temperature. The HDB values in table 1
were extrapolated by equation (1) to 50 years (438 000 hours).
log ε = A − G log t A G HDB

(50 years)
Sanitation pipe @ 25 C 0.124 0.055 0.65%
Oil pipe @ 25º C -0.096 0.030 0.54%
Oil pipe @ 65º C -0.032 0.073 0.36%
Oil pipe @ 82 C 0.087 0.102 0.32%
Table 1
Weep parameters and 50 years HDB for typical oil and sanitation pipes.
These are the old ideas behind the classical weep regression lines and the HDB. The next section will introduce
a new approach to this subject.
A new approach – We start this section with arguments that indicate the inadequacy of the classical approach
to predict the long-term HDB. Specifically, this section will show that the ASTM protocol is good only to
predict short-term (3 to 5 years) weep failures. Let us see the reason for that. As indicated earlier, the high
strains required by the ASTM test protocol cause the resin to crack. Once started, and under sustained strain,
these resin cracks grow and eventually coalesce to form pathways that allow the passage of water. The lower
the strain, the smaller is the number of cracks, the slower is the rate of crack growth and the longer is the time
for coalescence and weeping. If the strain is low enough, below a certain threshold limit, the resin will not
develop new cracks and the pipe will not weep. Every resin has a limiting strain which marks the onset of
cracking. We call this onset strain as the threshold strain. The threshold strain is indicated by the first acoustic
signals emitted by the resin as the cracks start to form. The threshold strain separates weep from non-weep. The
regression line developed by the traditional ASTM D 2992 test method is good to predict weep above the
threshold strain. Below the threshold strain, however, that regression line is not applicable and the pipes would
run out indefinitely without weeping.
The existence of a threshold strain has been suggested without proof by many authors. The lack of proof can be
justified by the short duration of the ASTM D 2992 test protocol, which stops at 10 000 hours, just short of
capturing the threshold strain. Had the tests been extended a little longer, say to 30 000 hours instead of 10 000
hours, the threshold limit would have become apparent. Many authors have advanced the suspicion that the
weep regression line should turn horizontal in the long-term, but no test results have been offered to support that
claim. The first author to reject the ASTM protocol was Frank Pickering (reference 6, 1983) who proposed that
it be replaced by a parameter that he referred to as PEL, or Proportional Elastic Limit, determined by the strain
at which the pipe would give off the first acoustic signals. That is almost the same idea advocated in the present
paper, except that we propose to measure the first acoustic signals on the critical ply, not on the pipe itself. It is
best to test isolated plies instead of pipes, to eliminate the possibility of contamination by spurious signals
coming from non-critical plies. We will have more to say about this in the following paragraph. For now, it
suffices to say that the data generated by short-term weep tests are questionable to predict long term weep.
The obvious method to determine the crack onset strain is by monitoring the acoustic signals emitted by pipe
specimens under increasing pressure levels. The onset strain is reached when the first acoustic signals are heard.
One problem with this procedure is the difficulty to separate the signals coming from different plies. As
explained earlier, we are interested in the crack onset of the ply that is immediately on top of the liner. To make
sure the measured signals come from the critical ply, the test specimens must be made exclusively with that
type of ply. The use of these specially constructed test specimens eliminates the spurious emissions from other
plies. This argument holds for both sanitation and oil pipes. For example, the threshold strain for sanitation
pipes must be measured on pipe specimens made exclusively of chopped glass laminas. And that of oil pipes
measured on specimens made exclusively of UD plies wound with the same angle as the pipe itself.
It may be argued that the acoustic signals measured on special pipes, instead of on commercial pipes, ignore the
cracks from other plies that might grow and eventually reach the liner. This fear can be put to rest, since the
fibers on the plies would stop any inter-ply transfer of cracks. Cracks that originate in one ply never “jump” to
adjacent plies. The only case known of inter-ply crack transfer in multilayered pipes occurs when substrate
cracks transfer to the liner. And that is possible because, as we know, the liner does not have glass fibers.
We propose that the strain at the first acoustic emissions marks the initiation of the cracking process in the
critical ply. Below the threshold strain, defined by the crack onset, the resin does not crack and the pipe does
not weep.
Figure 4 shows the threshold strain superimposed on the classical weep line of sanitation pipes. The first thing
to notice is that the classical ASTM line holds for strains above the threshold line. This is in line with what we
have said earlier, that the classical line is applicable to short-term predictions. However, figure 4 also shows that
for strains below the threshold value, the regression line flattens out and merges with the threshold strain. The
classical line is obviously not valid to make long-term predictions.
Figure 4
Weep regression line of sanitation pipes with the inclusion of the threshold line. The threshold strain is assumed
at 0.8%, which is reasonable for a matrix resin with elongation of 3.0%.
The transition point, where the classical line turns horizontal and merges with the threshold strain, is controlled
by the resin. Both the short-term regression line and the long-term threshold strain are dominated by the resin.
In the case of oil pipes the threshold strain is controlled by the interphase resin. And for sanitation pipes, the
resin matrix itself has the control. In either case the resin supplier – not the pipe manufacturer - should measure
and report the threshold strain.
Note: The HDB is the “initial hoop strain” that weeps the pipe in the long term. Often times we think of the
HDB as a “final hoop strain” implying that creep may be the ultimate cause of the weeping. According to our
model, however, weeping is caused not by creep, but by the slow growth of cracks in the resin. The error
incurred in neglecting creep cannot be quantified at this time. However, given the low magnitude of the working
strains in commercial pipes, creep may not be an important factor.
Suggested values for the threshold strain – The values of the threshold strain depend on the toughness of the
resin that forms the critical ply, i.e. the ply in direct contact with the liner. For oil pipes the cracking is
controlled by the interphase resin around the UD fibers, whereas for sanitation pipes, the matrix resin itself has
the control. Neither of these threshold strains has ever been systematically measured. The few reports available
on this topic were written not to measure the threshold strain, but to explain the cracking of gelcoats backed by
chopped glass laminas. Such reports are of limited value to the design engineer. Reference 1 reports threshold
strains of 2.0% and 1.0% respectively for isolated chopped glass plies saturated with vinyl ester and polyester
resins. These results, however, are not reliable as they were obtained by testing gelcoated specimens submitted
to three point flexural loading. Better results would have been obtained by testing non-gelcoated plies in
tension. Table 2 shows the threshold strains reported by Norwood and Millman (ref 3) for laminas (not pipes)
made of chopped glass and woven roving.
Elongation at break of the Threshold strain measured by acoustic emission on isolated

polyester resin lamina
Chopped glass lamina Woven roving lamina
2.5% 0.4% 0.3%
3.8% 0.8% 0.6%
Table 2
Threshold strains reported by Norwood and Millman from tensile tests performed on isolated plies of chopped
glass and woven roving.
The critical plies of sanitation pipes are made of chopped glass saturated with a polyester resin having a
minimum elongation of 3.0%. From table 2 such pipes would have a threshold strain in the neighborhood of
0.8%. We will assume the value of 0.8% for the threshold strain of sanitation pipes made of polyester resins
with elongation of 3.0%. Unfortunately there is no study available to indicate the threshold strain of oil pipes.
For the sake of this paper we will assume a threshold strain of 0.6% for oil pipes. The threshold strain, we say it
again, is the hoop strain at which the first acoustical emissions, or first cracks, appear on the critical lamina. For
oil pipes, as we know, the threshold strain depends on the winding angle as well as on the resiliency of the
interphase resin. We will have more to say about the interphase resin and how it relates to the matrix resin and
the sizing that is applied on the glass fibers.
The effect of temperature – The influence of the operating temperature on the weep line has never been
systematically studied. Data developed by a leading oil pipe manufacturer indicate a strong increase of the slope
G with temperature, as shown in table 1. Similar results are expected for sanitation pipes, but there is no study
available to quantify that. The higher slopes indicate that the rate of crack growth increases with the
temperature. This is valid for the short-term. The influence, if any, of the temperature on the threshold strain is
yet to be determined.
It is unlikely (but still unproven) that a temperature increase would lower the threshold strain. Higher
temperatures are expected to increase the rate of crack growth and therefore to increase the slope of the
regression line. However, there is no overriding reason why higher temperature should deplete the threshold
strain. It is more likely (but still unproven) that the plasticization of the resin at high temperature would
increase, not decrease, the threshold strain. Figure 5 shows how the weep lines of oil pipes would look like at 25
C and 65 C under the assumption that the threshold strain is unaffected by the temperature. The figure indicates
that the temperature is an important design factor for the short-term, which is dominated by the classical weep
line. For the long-term, however, the temperature appears to be irrelevant. This is an unexpected conclusion
that, if true, would have an enormous impact on the composites pipe industry.
Figure 5
The operating temperature affects the short-term, not the long-term weeping.
Cyclic loading – The mechanism of crack growth under cyclic loadings is different from that under static loads.
The static crack growth is a continuous process dominated by the strain and the temperature. Under cyclic
loads, however, the cracks grow in a stepwise fashion, one little bit at a time, depending on the energy that is
delivered in each cycle. The process of crack growth under cyclic loading is similar to the hammering of nails
into wood. The nail penetrates the wood one bit at a time depending on the energy of the blow. Again, as the
nail is driven faster by cyclic hammering than by static pushing, the slopes of cyclic lines are steeper than those
of static ones. Figure 6 indicates a slope G = 0,083 for oil pipes under cyclic loads @ 25 C. This is much higher
than the static value G = 0,030. This, however, is valid only for the sloped part of the regression line. The effect
of cyclic loads on the threshold strain is a different story.
Cyclic loads are not expected to affect the threshold strain. There is no apparent reason why the resin
(interphase or matrix) would “decide” to crack under cyclic and not under static strains. Figure 6 indicates that
the long-term behavior of pipes is dominated by the same crack threshold strain, regardless of the cyclic or
static nature of the loading. This is another unexpected conclusion which, if proven true, would have an
enormous impact on the composites pipe industry.
Figure 6
The horizontal part of the weep line (the threshold strain) is the same for both cyclic and static loadings.
Rupture versus weep – The preceding discussions have shown that the weep failure is controlled exclusively
by the interphase resin (oil pipes) or the matrix resin itself (sanitation pipes) and has nothing to do with the glass
fibers. The glass fibers control the rupture or burst failure, not the weep failure. This section will compare the
weep line of sanitation pipes with the rupture lines of UD laminas made of E glass and of boron-free glass.
The rupture lines of UD laminas, developed by Mark Greenwood (ref. 2), were extensively discussed in the part
1 of this trilogy. Figure 7 shows the dramatic impact of the type of glass on the long-term rupture strains. We
see that the boron-free UD plies extrapolate to a strain at rupture of 0.92% at 50 years, versus 0.41% for those
made of standard E glass. The weep line, being dominated by the resin, is the same for both types of glass.
Figure 7 shows that for E glass pipes the transition from weep failure to rupture occurs at circa 5000 hours,
indicating that the long-term failure of these pipes is by rupture, not by weep. The opposite is true for boron-
free glass pipes which, for a design lifetime of 50 years, would fail by weep instead of by rupture.
A closer examination of figure 7 reveals the following:
1 – For strains above the threshold, the boron-free pipes would weep before bursting.
2 – The HDB line is controlled by the resin and does not depend on the type of glass.
3 – The rupture line of E glass pipes crosses the HDB line somewhere between 1000 hours and 10000 hours,
indicating that E glass pipes may have mixed modes of failure in this interval.
4 – Above 10000 hours and for strains below the threshold, the E glass pipes would burst before weeping.
Note: When the test time is too short, as in ASTM D 1599, the pipe may burst before the water has time to weep.
This failure is reported as burst, instead of weep, even if the mechanism of weep has not been violated.
Figure 7
Weep and rupture lines for E glass and boron-free glass pipes.
Conclusions – There are two mutually exclusive approaches to account for the weep failure of composite
pipes. These approaches agree on the short-term and conflict on the long-term.
The classical approach proposes that the weeping caused by high strains occurs also at low strains. This
approach produces regression lines that are valid to make short-term predictions (3 to 5 years), but are
unrealistically over conservative for long-term design. Also, the test method to determine the regression line by
this approach is expensive and time consuming.
The new approach, introduced in this paper, proposes that all pipes have a threshold strain below which
weeping never occurs. The new approach accepts all procedures and conclusions of the classical approach for
short-term predictions. The disagreement is in the long-term. And last, but not least, the new approach makes
bold predictions for the HDB of pipes under cyclic loadings and high temperatures.
A conclusive experiment to discriminate between these theories would require tests performed on water filled
pressurized pipes at low strains over long times. Such tests have never been performed. The best approximation
to these long-term pressure tests has been published just recently, early in 2009, by Hogni Johnson, who
reported the results of 30 years of tests on deflected pipes under strain-corrosion in acidic solution. See
reference 4. Figure 8 shows the flattened strain-corrosion line published by Hogni in exactly the same fashion as
proposed in this paper.
Figure 9
The long-term strain-corrosion tests of composite pipes in acidic media shows the flattening of the regression
line. The threshold strain of 0.8% has been superimposed on the graph. (Courtesy Amiantit)
It may be argued that strain-corrosion tests performed on deflected pipes are not valid to vindicate a hypothesis
for weep failure of pressurized pipes. A rebuttal of this argument will be presented in connection with the
discussion of strain corrosion rupture, in the third paper of this trilogy. This third paper will show that the
phenomenon of strain corrosion rupture is a special case of the weep failure that we have just discussed.
Therefore, the Hogni’s results in table 9 are valid as proof of weep failure. A more detailed discussion of this
topic will be presented in the forthcoming third paper of this trilogy.
Biography: Antonio Carvalho is an engineer with 39 years dedicated to composites. Past experience includes 30
years with Owens Corning and 9 years as a full time consultant for Reichhold. His current responsibility
includes technical support and market development for Reichhold`s DION resins in Brazil and Latin America.
For direct communication please contact Antonio.carvalho@reichhold.com
Acknowledgement: The author recognizes the invaluable help, comments and suggestions from Mark
Greenwood and Gabriel Gonzales. Many blind alleys have been avoided thanks to their insights and advice.
References:
1 – Environmental stress corrosion of chemically resistant polyester resins and glass reinforced laminates in
acids, by G. Marshall, M. Kisberyi, D. Harrison and R. Pinzelli, in the 37th Annual Conference of the SPI, 1982.
2 – Pultruded composites durability: A key value, by Mark Greenwood, in CFA (now ACMA) Composites
Conference, 2001.
3 – Strain limited design criteria for reinforced plastic process equipment, by L. Norwood and A. Millman SPI
1980
4 – Chemical resistance of GRP pipes measured over 30 years. By Hogni Jonsson in Munich, march 2009.
5 – Mode of failure of hydrostatically overstressed Reinforced Plastic pipe, by W. G. Gottenberg, R.C. Allen,
W. V. Breitigan and C. T. Dickerson, in the 34th Annual Conference of the SPI, 1979.
6 – Design of high pressure Fiberglass downhole tubing: A proposed new ASTM specification, by Frank H.
Pickering, in the 38th annual conference of the SPI, 1983.
7 – Burst failure of composite pipes. Antonio Carvalho, ACMA, Las Vegas, May 2007.
8 – Surface cracking of RP repeated bending. J. Malmo in the 36th annual conference of the SPI, 1981.
9 – Historical background of the interface; studies and theories, by Porter W. Erickson in the 25th annual
conference of the SPI, 1970.
Appendix A. Further comments on the weep lines
This appendix explains the differences between the weep regression lines of oil and sanitation pipes. From
figure 2 the differences requiring explanation are: (a) the different slopes between the two lines and (b) the
reason why the line for sanitation pipes is above that for oil pipes.
We begin with the relative position of the weep lines. The random chopped fibers present in the weep barrier of
sanitation pipes stop crack growth and prevent the formation of large cracks. The weep barrier of sanitation
pipes requires a large number of small cracks to form the pathways that allow the passage of water. And a large
number of cracks require high strains. In contrast, the UD laminas in oil pipes require lower strains to develop a
small number of long cracks parallel to the fibers. The strains required to weep the UD plies are less than those
required to weep the plies of chopped fibers. That explains why the weep line of sanitation pipes plots above
that of oil pipes. As a corollary, this argument shows that the regression line of oil pipes could be substantially
improved by introducing a thin ply of chopped glass under the first UD lamina.
Next we address the difference in slopes. We begin by recognizing that the time to weep is obtained by adding
the time to coalescence of the cracks with the time for the water to travel the pathway from the liner to the outer
surface of the pipe.
t = tc + tt (A1)
Equation (A1) indicates that the time to weep “t” is actually a sum of the time to coalescence plus the time for
the water to travel the pipe thickness. We know that the slope “G” of the weep line is the ratio of changes in
strain to changes in the time to weep. Therefore, different slopes indicate different relations between the change
in the applied strain and the change in the total time to weep.
∆ log ε
G=
∆ log (tc + tt )
(A2)
According to equation (A2), the slope G decreases when, for any given change in strain, the change in weep
time is increased. We next show that for any given change in strain, the change in weep time is higher for oil
pipes than it is for sanitation pipes.
First, the coalescence time is longer for the cracks growing in UD plies than it is for those in chopped glass
plies. This follows from the fact that the UD cracks run parallel to each other and never coalesce on the same
ply. The coalescence occurs when a crack in one ply intersects a crack in another ply. Second, the travel time
for the water is longer in UD plies. This is explained in figure A1. The only points where the water can pass
from one UD ply to the next are those at the intersection, or crossing, of the cracks. The water in this case must
travel long distances, meaning long travel times, to reach the outer surface. By contrast, the pathways in the
chopped glass plies (a) are formed on the same ply and (b) are direct and shorter. The long coalescence and
travel times in oil pipes stretch out the regression line and explain its flat slope.
It may be argued that the sanitation pipes also have UD plies and should therefore have the same long travel
times as the oil pipes. This argument ignores that the high weep strains of the sanitation pipes (a) dramatically
increase the number of UD cracks and (b) open up these cracks to facilitate the passage of water. The weep
strains in sanitation pipes are simply too high. And high strains negate the retarding effect of UD plies on the
travel time.
Figure A1
The crack pattern of UD plies explains the flat weep lines of oil pipes. The innermost critical ply cracks and
provides an inlet for the water. The pathway followed by the water is shown by the arrows. The water moves to
the next ply at the intersection points “I”. Eventually the water exits the outermost ply at the point shown as
outlet. The flat slope is explained by this long, tortuous pathway.
Appendix B. Matrix resin and interphase
Sanitation and oil pipes are manufactured using a single resin, usually referred to as the resin matrix. The glass
fibers used in the critical ply can be of two kinds. The critical UD laminas of oil pipes are formed by a
homogeneous distribution of a large number of closely spaced individual fibers. By contrast, the critical
chopped glass plies of sanitation pipes are made of a collection of widely separated strands. Within the strands
the chopped fibers are in close proximity and show strong interaction, like they do in oil pipes. But outside the
strands their wide separation isolate them and they do not interact. This minor feature, this special detail in the
construction of the critical laminas, explains the difference in weeping behavior between oil and sanitation
pipes.
When the resin matrix impregnates the fibers it dissolves the sizing that is applied on the glass, forming with it a
new resin that we will call “interphase resin”. The interphase resin is a blend of the glass sizing and the resin
matrix. This blend is a new resin, with properties that differ from those of the original resin matrix. The new
resin fills the space within the glass strands and we can say that, as far as the glass fibers are concerned, they are
in fact surrounded by the interphase resin, not by the resin matrix. All phenomena involving the glass-resin
interface, such as fiber blooming, glint, debonding, etc, are controlled by the interphase resin, not by the matrix
resin.
The fibers in UD plies are located in very close proximity, like in a strand. In fact, the glass manufactures refer
to the type of reinforcement they sell for the filament winding process as a “single strand” roving. The
properties of UD plies should therefore be strongly dependent on the interphase resin, especially those that are
dominated by the resin, like the threshold strain. The weeping behavior of oil pipes should be strongly
dependent on the matrix resin and the glass sizing.
By contrast, the low glass loading and the wide separation of the strands in the chopped glass plies suggest that
the threshold strain of sanitation pipes is controlled essentially by the resin matrix. It is therefore the matrix
resin, and not the interphase resin, that controls the threshold strain in sanitation pipes. The sizing on the glass
fibers is not expected to affect the weeping behavior of sanitation pipes.
The idea of an interphase layer surrounding the glass fibers is not new (ref 9). In spite of its obvious merit to
explain the behavior of UD plies, however, this idea has never been pursued by researchers. To my knowledge
the interphase layer has never been characterized. It is hoped that the concept of threshold strain introduced in
this paper, and its dependence on the interphase layer, would revive the interest in this topic. The glass fiber
manufacturers should benefit greatly from this concept.
Appendix C. Is there a threshold strain?
This appendix answers an interesting objection regarding the validity of the threshold strain. The argument
against the existence of the threshold strain can be divided in four parts:
1 – Any mass of matrix material has pre-existing “flaws”, or “hot” spots, that serve as starting points for cracks.
2 – Under any applied strain “Є” these pre-existing cracks would grow and eventually cause the resin to weep.
3 – The pre-existing flaws will therefore weep the pipe regardless of the applied strain.
4 – Therefore, there is no threshold strain.
This is an impressive argument. Fortunately it is valid only for homogeneous materials, not for composites. It is
very true that homogeneous materials, like metals or neat polymeric matrices, develop large “flaw-induced”
cracks and are for that reason not expected to have threshold strains. This situation does not occur in
composites, because the fibers that are present in them would not allow the growth of large cracks. Composites
weep not because they develop large cracks, but because a large number of small cracks coalesce to allow the
passage of water. For this coalescence to take place, the number of small cracks must be very large, far larger
than those from pre-existing “flaws”.
To develop the required large number of cracks, the resin (interphase or matrix) must be under high strain. That
is the justification for the threshold strain.
Appendix D. Total strains and safety factors.
In real life the pipes are subjected to tensile and bending strains in both the longitudinal and hoop directions.
Take, for instance, those pipes that are installed underground and deflect under the weight of the soil. The pipes
in this condition are subjected to strains that are tensile hoop (from the pressure), flexural hoop (from the
deflected condition) and tensile axial as well. In addition to these mechanical strains, the pipe may also be
subjected to thermal strains. Often times the pipes operate in temperatures other than the room temperature at
which they were qualified. In situations like those the critical ply may be subjected to thermal strains in addition
to the mechanical tensile and bending strains. The total strain (thermal + mechanical) on the critical ply is
obtained from the following equation.
(total strain ) = tensile + bending + thermal (D1)
Equation (D1) recognizes the equivalence of thermal and mechanical strains.
The strains that are calculated or measured in practice are usually referred to the global “x” (longitudinal) and
“y” (hoop) directions of the pipe. To obtain the safety factors against long-term weep, we must rotate the global
strains to the local reference frame of the critical lamina. The chopped glass plies are isotropic and require no
such rotation. For the UD plies the rotation is done by equation (D2).
 ε1   cos 2 α sen 2α 2(cos α )( senα )  ε x 

   
 ε2  =  sen α
2
cos α2
− 2(cos α )( senα ) × ε y  (D2)
 1 γ  − (cos α )( senα ) (cos α )( senα ) cos 2 α − sen 2α   0 
 2 12     
Where
α is the angle of the UD fibers with respect to the longitudinal (axial) direction of the pipe.
Є1 is the strain in the fiber direction
Є2 is the strain transverse to the UD fibers
Єx is the strain in the axial direction of the pipe
Єy is the strain in the hoop direction of the pipe
γ12 is the shear strain on the UD lamina
Expanding (D2) we obtain the strains expressed on the reference frame of the UD lamina.
ε1 = ε x cos2 α + ε y sen2α (D3)

ε 2 = ε x sen α + ε y cos α
2 2
(D4)
γ 12 = 2 senα cos α (ε y − ε x ) (D5)
Equation (D3) gives the strain in the direction of the UD fibers. This strain controls the long-term rupture of the
pipe as discussed in the part 1 of this trilogy. Equation (D4) gives the strain in the direction transverse to the
UD fibers. This transverse strain controls the weep failure of oil pipes.
The safety factors against long-term weep are obtained by matching the total strains and the threshold strains
(HDB) in the principal directions of the critical lamina.
For sanitation pipes:
 HDB    threshold strain  

     
 (SF ) x   tensile + bending + thermal  x   tensile + bending + thermal  x 
  = =
(SF ) y  


HDB   
  
threshold strain  
 

 tensile + bending + thermal  y   tensile + bending + thermal  y 
For oil pipes
 (See Note )   (See Note ) 
 (SF )1    
(SF )  =  HDB   =  threshold strain  
  
 2
 tensile + bending + thermal  2   ε x sen 2α + ε y cos 2 α  
 2
Note: As explained in the part 1 of this trilogy, the UD plies strained in the fiber direction fail by rupture, not by
weep. Therefore the weep mode of failure is not applicable in the fiber direction. As indicated in the above
equation the oil pipes would weep when the critical UD ply fails in the transverse direction. This is a very
interesting conclusion that would allow the specification of oil pipes by reference to the threshold transverse
strain, regardless of the winding angle.
The effect of the operating temperature on the pressure rating of the pipe can be estimated by entering the
appropriate thermal strains in the above equation. Note that the strains in the UD plies are referred to the ply
principal directions 1 and 2. Strain rotation is not required for the isotropic chopped glass lamina.
Appendix E: The effect of temperature on the regression lines
This appendix addresses the effect of temperature from a general standpoint, taking into account the issues of
rupture, weep and strain corrosion. To facilitate the reading, this same appendix is repeated verbatim in all three
parts of this trilogy. The comments that follow are based on the assumption that the rupture modes of failure of
commercial pipes (RDB and CDB) are controlled by the glass, while the weep mode (HDB) is controlled by the
resin.
Static effect on the glass fibers – Temperature affects the rate of chemical attack on the glass, which would
affect the slopes of the rupture static regression lines. Therefore, the slopes of the static RDB and CDB lines are
expected to increase with increases in temperature.
Cyclic effect on the glass fibers – Chemical attack is not an issue in cyclic rupture and therefore temperature
should have no effect on the cyclic CDB or RDB lines.
Static effect on the resin – The static regression line that predicts short-term weep failure is controlled by the
interphase resin (oil pipes) or the matrix resin (sanitation pipes). The rate of crack growth in resins increases
with the temperature. Therefore, the slopes of the static weep lines, or HDB lines, should increase with
temperature.
Cyclic effect on the resin – As in the static case, the rate of crack growth in resins increases with temperature.
Therefore, the slopes of the cyclic weep lines should increase with temperature.

Las Vegas 2009. Weep Failure

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Las Vegas 2009. Weep Failure

Uploaded by

Copyright:

Available Formats

Last revised on December 22, 2009

Structural failure of composite pipes – a trilogy

Part 2 – Weep failure

Presented at the 2009 ACMA conference in Las Vegas

log ε = A − G log t (1)

log HDB = A − G log 50 (1A)

log ε = A − G log t A G HDB

Elongation at break of the Threshold strain measured by acoustic emission on isolated

A closer examination of figure 7 reveals the following:

Appendix A. Further comments on the weep lines

Appendix B. Matrix resin and interphase

Appendix C. Is there a threshold strain?

Appendix D. Total strains and safety factors.

Equation (D1) recognizes the equivalence of thermal and mechanical strains.

 ε1   cos 2 α sen 2α 2(cos α )( senα )  ε x 

ε1 = ε x cos2 α + ε y sen2α (D3)

For sanitation pipes:

 HDB    threshold strain  

For oil pipes

Appendix E: The effect of temperature on the regression lines

You might also like