INTRODUCTION
Corrosion is recognized as one of the most important degradation mechanisms
that affect the longterm reliability and integrity of metallic pipelines (Li
et al., 2009). Pipeline operators throughout the world are challenged
with the expensive and risky task of operating aged pipelines because of corrosion
and its potential damaging effects (Ahammed, 1998).
Loss of the metal cross section is one of the major effects of corrosion and
these results in a reduction of pipeline carrying capacity and safety and as
the pipelines are ageing and corrosion may develop, the economical consequences
of reduced operation pressure, repairs or replacements may become adversely
high.
Assessment method is required to determine the severity of such defects when
they are detected in pipelines (Cosham and Hopkins, 2003).
The assessment condition of existing oil and gas pipeline is necessary in order
to protect the public, financial. The assessment condition of existing oil and
gas pipeline is necessary in order to protect the public, financial, investment
and environment from such failures. Systematic and optimised regular inspections
of pipelines with stateofthe art tools and procedures can reduce significantly
the risk of any undue accident caused by a lack of unawareness of the integrity
of the line (Cosham et al., 2007).
Pipeline corrosion assessment: As inline inspection technology advances
and tool resolution and accuracy increases, the traditional methods of dealing
with InLine Inspection (ILI) data are quickly becoming unfeasible, both from
an economic and a practical point of view. Corrosion growth analysis provides
a proactive method of analysing large quantities of ILI data, prioritizing pipeline
repair programs and optimizing reinspection intervals. It enables operator
to fully understand the potential future risk to a pipeline due to corrosion
(Desjardins, 2002). There is thus a need to provide the
pipeline industry with an effective and affordable approach to assess corroding
pipeline and in particular, a need to effectively manage the vast amounts of
collected data relating to pipeline condition so that operators can maintain
and prolong pipeline integrity. An initial review of pipeline integrity based
on the operating history and the results of any past survey or inspections can
potentially identify the likely problems and consequently the additional data
that may needed to make a decision about the future of the pipeline (Jones
and Hopkins, 2005). In order to achieve such optimization through full utilization
of ILI data in structure maintenance scheme, the improvised of best practice
assessment and prediction tools need to be developed.
Availability of assessment tools: The availability of pigging data in
vast amount is just the first piece of solution towards pipeline sustainable
and effective inspection, repair and maintenance scheme (IRM). The challenge
is how to build a system capable of collecting data and turn it into information
in the context of managing pipeline integrity (Wiegele et
al., 2004). Recognizing the value of numerous of pigging data in industry,
an inherent understanding of how to fully utilize pigging data for the sake
of structure integrity assessment has become a necessity. A systematic approach
to automate the assessment process by the means of software specially tailored
to assess corroding pipeline with great capability of data processing and sampling
as well as deterministic and reliability assessment engine is hardly available
in the industry since existing tools are rather not comprehensive (Dawson
et al., 2001).
In general, a good number of the existing pipeline assessment methods are purely
deterministic and lacking reliability element as well as data sampling tool,
hence under utilises the inspection data for the estimation of the variation
of corrosion growth rate (Li et al., 2009; Melchers
and Jeffrey, 2007). Available assessment tools for assessing corroding pipeline
in industry such as Electronic Corrosion Engineer (ECE), ENCIPDA and CORROLINE
is designed for internal pipeline and heavily rely on material and operation
properties such as temperature, liquid flow velocity, hydraulic diameter of
the pipe and others (Kvernvold et al., 1992).
All the above mentioned parameters if not measured correctly on site can mislead
the calculation of pipe remaining strength. While existing software such as
RSTRENG used for determining the remaining strength of externally corroded pipe
normally take into account of maximum depth of defect from the numerous number
of inspection data merely (Bjornoy and Marley, 2001).
Hence, neglecting the effect of defect variation upon pipeline failure probability.
The assessment procedure frameworks: The mainframe of PICA consists
of four independent stages namely data sampling, data analysis and pipeline
assessment and integrity prediction. Every single stage is developed as a standalone
component with specific output.

Fig. 1: 
Flowchart of data sampling and data analysis stages 

Fig. 2: 
Flowchart of pipeline assessment stage 
Even so, all stages are later integrated to give a comprehensive assessment
option to pipeline owner if failure probability of the line becomes the preferable
main output. Figure 1 to 3 depict the architecture
of software. Data sampling and data analysis require large volume of inspection
data which in practiced are often underutilised due to lack of understanding
in handling the data statistically. Key to the effectiveness of PICA is the
ability to assess with a measurable level of confidence, the corrosion rates
for all defects on the pipeline. When multiple inline inspections are available,
corrosion rate determination is enhanced by software that automatically correlates
defects from one inspection to the next. The appropriate level of confidence
in growth rates and corrosion severity predictions is obtained by incorporating
the error associated with inspection tools into all observations and subsequent
calculations.

Fig. 3: 
Flowchart of integrity prediction stage 
The data optimisation can be used effectively to assess the integrity of the
corroding pipeline so as to determine the level of fitnessforservice. The
structure integrity assessment and prediction can be based on either deterministic
or reliability method or combination of both.
Data sampling: In this study, historical data representing metal loss dimension (depth and length) measured through repeated pigging inspection are utilised to determine the corrosion growth rate. The calculation of growth rate is based on linear model. The sampling process is intended to match corresponding inspection data from previous inspection with the later one. When two or more inspections database available, individual defect growth rates can be determined with a decent degree of confidence. Corrosion rates are then calculated from the change in defect size between two or more inspections.
Determining the change in size however, presents the significant challenge
of matching every defect from multiple ILI data sets.

Fig. 4: 
Interfaces of data matching 
Manual checks are conducted throughout the process to ensure data accuracy.
This process takes full advantage of the ILI data and gather corrosion growth
information, enabling the future repair and reinspection needs to be assessed
based on economy and safety issues (Dawson and Walker, 2005).
The flowchart of data sampling procedure and its software interface are dipict
in Fig. 4.
On the other hand, an empirical model of the deWaard and Milliams is included
in the framework to give an option to the user when no repeated data available
for the estimation of corrosion growth rate based on metal loss (Nyborg,
2006). The aforementioned model capable of estimating the averaged growth
rate of internal corrosion by just relying on operational and flow parameters.
Data analysis: Data analysis provides useful information pertaining
to corrosion mechanism such as defect dimensions and defect growth rate based
on successful matched data. This process consists of two stages which are statistical
and probabilistic analysis. Statistical method is used to estimate the average
and variation of the corrosion parameters.

Fig. 5: 
Data analysis procedure 
Throughout several systematic probability methods, a proper distribution can
be verified. The whole procedure is summarized in Fig. 5.
Under deterministic approach, histogram of the corrosionrelated parameter is more than sufficient to provide vital information for calculating the remaining pressure of corroding pipeline. However, if the operator tends to use reliability method in the assessment process, it is requisite to enhance data analysis so as to cover the proper steps of probabilistic technique. Data analysis can be effectively used to locate data problem areas, measure data changing over time and increase the overall understanding of complex statistics. The information obtained from such analysis can have a significant effect on decision making.
Pipeline assessment: The evaluation of remaining strength and reliability
assessment of corroded pipelines can be carried out using both deterministic
and reliability methods. Deterministic assessment is a straightforward approach
based on codes or established capacity equation such as the ASME B31G, modified
ASME B31G and DNV RPF101 (Table 1). The failure pressure
equation adopted in these codes is used to estimate the maximum allowable pressure
for corroded pipeline. The available assessment codes can be categorised into
two categories namely fully deterministic code and semiprobabilistic code.
Fully deterministic code such as B31G equation does not cater the variation
of the corrosion parameters as it is already represented by safety factor, hence
poor capability of projecting the future remaining life of the pipeline.
DNV (2004) has introduced a deterministic capacity equation with partial
safety factors through its new assessment code known as RPF101.
Table 1: 
Capacity equation of pipeline remaining pressure 

The partial safety factor was developed with more advance probabilistic technique
to improve the representation of uncertainties associated with defect dimension,
hence the term semiprobabilistic.
The traditional design code which focused in deterministic methods are unable
to predict the failure probability of corroded pipelines at given time and reference
back to design code or specifications is likely to produce unduly conservative
assessment (Li et al., 2009). As an alternative,
the methods require a reliability engineering technique to assess future risk
based on calculating the Probability Of Failure (POF). This method is based
on the principle of loadresistance interference distribution. It has great
capability of taking into account inherent uncertainties that govern the variation
of corrosion parameters in order to improve the accuracy of pipeline assessment.
As. previously mentioned, the outcome of this approach is the failure probability
estimated using limit state function derived from pipeline capacity equation
from the assessment codes.
Integrity prediction: The concept of integrity or reliability means
that any attempt to quantify it must involve the use of statistical and probabilistic
methods. Therefore to evaluate the future remaining life of corroding pipelines,
the use of statistical and probabilistic approaches are necessary.

Fig. 6: 
Procedure of probability of failure estimation using monte
carlo simulation method 

Fig. 7: 
System architecture 
Since the outcome of the pipeline assessment based on reliability approach
is the Probability Of Failure (POF), the loadresistance interference principle
must be adopted.
Monte Carlo simulation is one of the methods for iteratively evaluating a deterministic model using sets of random numbers as inputs. This method is often used when the model is complex, nonlinear, or involves more than just a couple uncertain parameters. The variation in failure probability can be plotted against time and compared with the target probabilities for the limit state considered, the critical hazard and the time of pipeline failure can be identified. Monte Carlo simulation is preferable since this technique is less complicated and it can simulate a large number of experiments and the variables can be in any type of distribution.
This technique involves sampling of random variables from respective distribution
and evaluating the numbers of failure attempt (violation of limit state function)
over the number of simulation cycles, N. Limit state function shows how the
pipeline fails (leakage or bursting). It is repeated many times with a new random
vector of variables (Fig. 6). Therefore for N trials, the
probability of failure is determined as;
n [G[x]<0 
= 
number of trials which violated limit state function 
N 
= 
number of trials. 
P_{f} 
= 
probability of failure 
Requirement and development of software system: The next stage is to
integrate all of the standalone stages and later automate the assessment process
using IT tools (Table 2, Fig. 7).

Fig. 8: 
Integration of assessment stages 
Table 2: 
Types if programming platform and system utiliez in software
development 


Fig. 9: 
Tolerance of orientation and wheel count for data matching
in stage 1 

Fig. 10: 
Example of result from data matching in percentage 
The automation of the procedure is meant to boost the effectiveness and speed
of the assessment and make it more marketable.
Integration of the assessment tools: The integration of four standalone stages as formerly mentioned in section IV: Methodology can assist pipeline operator to assess their assets integrity effectively at a much cheaper cost. Assessing structure integrity, pipeline in this case, involves multiple levels that requires combination of multidimensional knowledge covering variety of disciplines. As such, the overall assessment process becomes complex and impractical unless these multiple levels of assessment methodology can be integrated seamlessly.
The proposed assessment procedure meet adverse challenge to combine four standalone
stages of data sampling, data analysis, structure assessment and integrity prediction
under one single framework with flexibility of producing different types of
output according to the selected stages.

Fig. 11: 
PICA intrface of pipeline assessment 
For instance, the user can choose to assess the pipeline remaining life at
the time of inspection rather than projecting to the point where the structure
might fail in the future. This selection only involves data sampling, data analysis
and pipeline assessment without having to go through integrity prediction stage.
Every single stage is designed to produce its intended output independently,
yet the output can be utilized in other stage to achieve better output. In a
simple word, the stages as mentioned in Fig. 8 are independent
of each other unless the user decides to run multiple analyses under several
stages.
Software demonstration: Data sampling procedure was conducted to match corresponding inspection results from different years manually in Stage 1. To locate the corresponding defects, information of girth weld, wheel count, and defect orientation is referred to. The existence of distancerelated errors may cause difficulties in locating the pair of defect based on wheel count distance between two databases of pigging data (Fig. 9). Therefore, a reasonable margin of error /tolerance in regard to relative distance was allowed until sufficient numbers of data can be successfully matched to produce a proper distribution. Margin of errors/tolerance will alleviate difficulties in finding correct pair of data between two historical databases of pigging data. Figure 10 illustrates example of result from data matching in percentage. Result from data matching is then stored and ready for access in next stage to calculate probability of failure of the pipeline. Figure 11 shows the flow of the procedure in assessing the pipeline using this tool. The outcome from this procedure is a probability of failure either at the time of inspection or in the future.
CONCLUSION
The study will contribute to a betterment of pipeline integrity assessment as presently practiced by pipeline operators. It will encourage pipeline operators to optimise the large volume of inspection data gained through precipitouscost pigging inspection for assessment purposes. The integration of different processes related to pipeline assessment is hardly feasible due to the complexity of reliability method and the issue on the availability of pipeline inspection data.
By integrating two different principles of deterministic and reliability engineering to form a comprehensive assessment package, pipeline operators can have a great option to choose either deterministic or reliability approaches according to the targeted requirement. Moreover, the proposed seamless integrated approach as a software package may simplify yet maintain the practicality aspect of pipeline fitnessforservice evaluation, hence minimise the future maintenance cost that may arises due to structure failure associated with corrosion defects.
ACKNOWLEDGMENTS
The authors are pleased to acknowledge the Ministry of Science, Technology and Innovation, Malaysia (MOSTI) and the Ministry of Higher Education (MOHE) for the support of providing the research funds and scholarship (NSF).