Pipeline Integrity Management is the
management of pipeline assets
such that availability is maximised at optimum cost, without compromising
environmental, safety and legislative standards. This is achieved when, under
specified operating conditions, risks of
failure endangering the safety of personnel, environment or asset value are as
low as reasonably practicable.
Oil and gas pipelines can become
susceptible to a whole variety of threats throughout operational life, which if
not adequately mitigated against, may eventually compromise pipeline integrity. One of
the primary life-limiting threats is internal corrosion and therefore the
requirement for effective corrosion management is vital. In essence, a
corrosion management strategy should be risk-based and should take account of all aspects of
asset maintenance, corrosion rate activity, historical and future operational
parameters and the management and business requirements.
Risk-based pipeline corrosion management strategies,
developed subsequent to the performance of a corrosion risk assessment, deliver benefits
in that pipeline inspections
and corrosion control/mitigation activities may be targeted specifically at
those pipeline assets
that are assessed as high risk (i.e.
high probability of failure and high consequence). Corrosion management is
essentially a "closed loop" (iterative) process where the
corrosion risk assessment is
central to the management process. The purpose of this paper therefore, is
essentially a critique of the methods commonly used to assess pipeline corrosion risks.
Risk
assessment
Risk is defined as the combination
of the probability of the occurrence of an event and the magnitude of the
consequences of the occurrence (The Engineering Council, 1993). Risk analysis is a
structured process that identifies both the likelihood and extent of adverse
consequences arising from a single given activity. A Risk Assessment is the
integrated analysis of risks inherent
to a pipeline and
their significance in an appropriate context. Therefore, a
"Corrosion Risk"
is the probability of corrosion failure and the consequences, in terms ofhazard to personnel and
asset availability, which would result should a loss of containment occur.
There are essentially two ways of
assessing pipeline corrosion risks: deterministically, where
this can be approached cither qualitatively and/or semi-quantitatively, and
probabilistically, where quantitative risk assessment methods arc employed.
The traditional deterministic
approach to the assessment of pipeline corrosion risks is typically based on
the judgement of "competent engineering personnel" as the paradigm
for identifying risk.
Semi-quantitative (deterministic) methods essentially substitute the analytic
of science for the fallible judgement of "competent personnel", with
the explicit notion that scientific treatment provides a superior basis for
reliable prediction; an opinion which must surely be true. Probabilistic
approaches deal with uncertainties in the input data by employing probability
density distributions. Implicit in the assessment approach is the claim that rigorous
application of probability theory will yield a superior conceptual framework
for understanding and managingrisk;
this notion is clearly untrue, however this does not mean that probabilistic
treatments should never be used in the management of risk. It is clearly the case that
the way the various risk
assessment techniques are applied is a matter of more subtlety than
is often thought.
During this work the comparison of
deterministic and probabilistic methods was made using the same basic inputs,
where the former approach made use of mean values and the latter probability
density functions. For probabilistic assessments, statistical analyses of the input data were
performed in order to discern the form of each probability density function.
However, limited knowledge of certain of the primary input variables did
preclude statistical analysis. In these, instances a normal distribution of
values was assumed (e.g. the natural spread of pipeline nominal wall thickness was assumed to
be normally distributed, where the limits of the distribution were based on the
API specification (American Petroleum Institute, 2000)).
Fourteen in. main oil export pipeline
The pipeline chosen for this review is a 16.1 km
long, 14 in. main oil export line in the North Sea. It is a submarine pipeline constructed in API
5L X60, with a nominal wall thickness of 14.3 mm (0.559 in.), and was just over
three years into its period of operation, with a design life of 25 years. The
service conditions are relatively warm with typical crude oil inlet
temperatures of the order of 65°C and outlet temperatures of approximately
30°C; the pipeline bathymetry
and thermal profile are shown in Figure 1. The pipeline is currently operated at a pressure of
20 bara where the maximum allowable operating pressure (MAOP) is 140 bara.
Various initial screening assessments were performed
in order to determine the susceptibility of the pipelineto a variety of internal corrosion threats
which included microbial corrosion, sulphide stress corrosion cracking, CO^sub
2^ corrosion and fluid and solids erosion. In particular, the initial
screening assessments had
shown that the crude oil velocity (approximately 0.4m/s (1.3ft/s)) was too low
to prevent water drop out in thepipeline even
when present in small quantities. Calculations had yielded a critical crude oil
velocity in the region of 1.1 m/s (3.6ft/s) for the avoidance of water drop out
in the line; this indicates that the internal surface of the pipeline would likely
experience water-wetting during normal operation. The initial screeningassessments had shown that
the primary corrosion threat to the pipeline was internal CO^sub 2^ corrosion, where all other
potential failure threats were deemed to be negligible.
The pipeline is dosed continually with a corrosion
inhibitor in order to mitigate against any corrosion that could occur where
water accumulations develop in the line. Indications from both laboratory-based
and field-based trials were that inhibition in the region of 85-90 per cent was
achievable. The corrosion risk
assessments therefore were performed with CO^sub 2^ corrosion as
the dominant threat and that inhibition at 85-90 per cent was achievable.
Corrosion rates were predicted during this work using the model of (deWaard and
Lotz, 1993), the current CO^sub 2^ concentration (2.25 mole per cent) and
calculated thermal profile data.
The consequences of failure for
this pipeline, a main
oil export line, were assessed as "High" and therefore when
considered in the context of risk,
it is clear that the probability of failure for this pipeline was the primary
driver. For pipelines which
are used to transport liquid and/or gaseous hydrocarbons, the consequences of
failure arc likely to remain unchanged ("High") and therefore when
managing "risk"
this invariably means managing pipeline failure
probabilities.
Deterministic approach
The above approach therefore based
the probability of failure on a predicted time to failure - the pipeline remaining life. In
the present case, the remaining life values (R(z)) were used as the basis for
determining whether pipeline failure
probabilities were either "High", "Medium" or
"Low". The threshold values of R(z) together with the correspondingpipeline failure
probabilities are as shown in Table I.
Probabilistic methods
Probabilistic assessment methods are
considerably more labour-intensive than are deterministic methods, therefore
given the level of effort required it was not feasible to assess the entire
length of the pipeline using
probabilistic analysis. It was decided that, based on the pipeline bathymetry and
thermal profile data (Figure 1),assessments would
be performed at two specific locations along the pipeline: in the region of kilometre point KP0 to
KP0. 1 (i.e. the first 100m of the sealine), and within KP5 to KP6. Both
locations were "low" points in the line where temperatures were
warmer and the propensity for water "pooling" was deemed to be high.
The approach made use of probability density functions of the basic input
variables as opposed to mean values as used in the deterministic approach.
The first probabilistic method used,
that of the "Service Limit"assessment,
was essentially an assessment of pipelineremaining life performed
probabilistically. The analysis involved estimating the respective
distributions of cumulative corrosion damage at various specified future time
intervals and comparing each with regard to the corrosion allowance; the
approach used is shown in Figure 2.
Quantitative failure probabilities
were determined in this paper using the first order reliability method (FORM).
The methodology involved in the application of this technique is presented in
considerable detail elsewhere (Cizeli et al., 1994; WS Atkins Science &
Technology for HSE, 1997; WS Atkins Science & Technology for HSE, 1999;
BOMEL Ltd for HSE, 2001; WS Atkins Consultants Ltd for HSE, 2001) and therefore
only a brief synopsis is presented in the following.
In the case of Monte Carlo
simulation, there is an obvious relationship between the number of trials N and
the degree of accuracy on P^sub f^. By performing a large number of iterations,
the ratio of the number of failure outcomes to the total number of iterations
tends to the exact probability of failure. The primary limitation is the number
of iterations required, and although Monte Carlo simulations are easily applied,
they are more or less limited to determining failure probabilities in the
vicinity of 10^sup -5^ per year and above (10^sup -5^ per year failure
probability would require approximately 10^sup 7^ iterations to be performed).
Monte Carlo simulations were used
during this work as the primary yard-stick upon which to determine the accuracy
and validity of FORM. Monte Carlo simulations make no assumptions about the
nature of the limit state equation, unlike FORM where this is assumed to be
linear; inaccuracies resulting from FORM are often manifested as a direct
consequence of the non-linear variation of all points along G(x) = 0; in such
cases, the more detailed second-order reliability method (SORM) is often
required. Indeed, statistical errors may also result in FORM or SORM when
transforming the basic variables to equivalent standard normal variables.
Transformation of variables and the assumptions in respect of the nature of the
limit state equation are avoided in the Monte Carlo simulation; these numerical
simulations therefore, were used during this work to validate the results
obtained using FORM.
Deterministic analysis had yielded a
corrosion rate profile for the entire 16.1 km length of the sealine. The
corrosion rate algorithm yielded a base corrosion rate which was then modified
for inhibition at 85 per cent and 90 per cent - the profiles for which are as
shown in Figure 5.
Pipeline remaining life was determined
using the present operational envelope as the basis for the assessment; the calculation is as
illustrated above in equation (1). This yielded a remaining life profile for
the entire length of the pipeline which
varied from approximately 12-18 years at the pipeline inlet and approximately 47-70 years at
the outlet; the spread of remaining life values at each point along the pipelinereflected the perceived
corrosion inhibitor effectiveness.
In terms of determining a likely
probability of failure, the approach is as detailed below in Table I and this
yielded a worst-case failure probability of "Medium". A corrosion
management strategy would then be developed on the basis of a
"Medium" probability of failure and a "High" consequence,
but this approach is clearly a subjective classification of risk.
Results - probabilistic assessments
The results from the "Service
Limit" assessment are
shown in Figure 6 where the illustration shows the variation of the probability
of exceeding the corrosion allowance in the vicinity of the first 100m of the
sealine with time. The variation of probability with time indicates that in
approximately 8 years from the time of thisassessment the probability of exceeding the corrosion
allowance starts to become significant; with a probability of approximately
10^sup -2^. These data mean that that the pipeline would still be operable even when
relatively high probabilities are encountered, but significantly its integrity
will become increasingly impaired. Unfortunately, however, it is merely an
indicator that pipeline failure
may occur at some point when the corrosion allowance is exceeded.
Using the Service Limit assessment data clearly
requires care; a target probability can be chosen, and a strategy for corrosion
management developed on the basis of the time to reach this target. However,
choosing an appropriate level is difficult as it cannot be guaranteed that
the pipeline would
not be excessively at risk when
at or near the chosen value. The probability of exceeding the corrosion
allowance is not a failure probability and this is its primary weakness.
The variability in pipeline failure probability
with time, predicted using both the FORM and Monte Carlo simulation methods is
shown in Figure 7, and as can be observed the failure probability increases
over the time periods considered, consistent with an increased level of
corrosion damage with time. Notably, the agreement between failure
probabilities predicted by both FORM and Monte Carlo simulation (within the
more limited range of the latter) was exceptionally good, suggesting that
predictions from either method were reliable. As before, the spread of
predicted failure probabilities reflects the perceived corrosion inhibitor
effectiveness.
Using probabilistic output data for
corrosion management purposes is generally a matter of choosing an acceptable
maximum failure probability and developing a strategy to ensure that it is not
over-stepped. In this exercise the DNV criteria (Submarine Pipeline Systems, 2003) were
considered appropriate, where failure probabilities of 10^sup -5^ and 10^sup
-4^ for "High" consequence and "Normal" consequencepipeline segments,
respectively, were used as the basis for determining the schedules for the
performance of thorough inspections (where the integrity status of the pipeline would be formally
established through inspection). Indications were that within the first 100m of
the sealine (which was considered a high consequence section) the time to reach
the target failure probability (i.e. 10^sup -5^ per year) was of the order of
8.5 to 13 years, where the variation as indicated was dependent upon the
effectiveness of corrosion inhibition; for KP5-KP6 (which was considered as a
normal consequence section) this was of the order of 15.5 to 23 years (again,
the variation as indicated was dependent the effectiveness of corrosion
inhibition). Clearly the strategy for managing pipeline corrosion would be expected to be
driven by the most susceptible section of the pipeline; although notably, quantitative estimates
of failure probability were determined based on the natural variability of the
basic input parameters, where this had not been possible using the
deterministic approach.
Summary comparison
Comparing the output results from
the two assessmentapproaches
was particularly interesting. The deterministicassessment had yielded a worst-case failure
probability of "Medium" with a corresponding consequence of
"High"; classifications which are clearly subjective. The
"Medium" failure probability classification in reality actually covers
a relatively large window of remaining life (in this case 5-15 years) and
therefore actions based purely on a "Medium" failure probability and
"High" consequence (to reflect risk), may mean being overly conservative if
the pipeline is
at the top-end of the remaining life window, and the converse for those pipelines located at the
lower end. Therefore in addition to the assessed failure probability,
consideration should also necessarily be given to remaining life when
developing a strategy based on deterministic assessments of risk for managing corrosion. Generally, a
reasonable approach would be to make use of the half-life rule in particular
for scheduling pipeline inspections;
indications from this work would suggest that scheduling within the next 6 to 9
years would appear appropriate, although this depends on how effective
corrosion mitigation activities are in practice.
The probabilistic assessments had
quantified pipeline failure
probabilities, though it is important to note that more effort was required
when performing such an assessment.
Using target probabilities for "High" and "Normal"
consequence pipelinesegments
(Submarine Pipeline Systems,
2003), indications were that between 8.5 and 13 years was the time period for
which the target (predicted) failure probabilities would be reached, again
depending on how effective corrosion mitigation activities are in practice.
Basing pipeline inspections
in particular on the outputs from the deterministic assessment would therefore
be conservative in this instance; but this may not necessarily always be so.
That the probabilisticassessment indicates
that inspections justifiably may be extended beyond that suggested by the
deterministicassessment is
a clear benefit, in that it affords the opportunity to defer expenditure
on pipeline inspections
to a later date, but it may be the case that the converse may be required. It
may be argued therefore, that probabilistic assessment provides a superior basis for
driving pipeline corrosion
management activities given that the approach deals with the uncertainties in
the basic input data.
Conclusions
The deterministic approach has the
distinct advantage of simplicity and the capability of being applied to an
entire pipeline or
collection of pipelines relatively
easily. The deterministic approach therefore lends itself to ease of the
application. The disadvantages of the deterministic approach may often, but not
entirely be linked to inaccuracies in the input data, but notably it is the
inability to deal with uncertainties in the input data that is the primary
weakness. This may lead to an underestimation of the likelihood of failure or
overestimating the true "risk"
associated with current pipeline operations.
In other words, the outputs from deterministicassessments can be rather uncertain. In certain
circumstances more conservative approaches may be employed which make use of
more extreme values of the basic input variables and this may make the outputassessments of risk overly conservative.
The primary advantage of probabilistic assessment is that it
facilitates quantification of failure probability on the basis of uncertain
data. The probabilistic approach overcomes the seemingly arbitrariness of
selecting contingencies in the deterministic method and causes attention to be
focused on the degree of "risk"
associated with pipeline operations.
The probabilistic approach reduces the weakness in the deterministic method
concerning the assumptions made with regard to the input variables, but this
does not necessarily remove the possibility that important parameters are
omitted, or perhaps even misjudged. Nevertheless, the use of probabilistic
methods should allow better management decisions to be made based on evaluation
of the primary threats to a given pipeline. There are disadvantages in that it is intensi
Sumber : Lawson, K. "Pipeline corrosion Risk Analysis - An Assessment of Deterministic and Probabilistic Methods". 27 Januari 2014. http://search.proquest.com/docview/218907935?accountid=31562
Tidak ada komentar:
Posting Komentar