Reliability engineering application to pipeline design

Background: importance of pipelines

The world has over 2.5m km of oil and gas (O&G) pipeline with several key routes transporting various products (CIA, 2017); amongst these are O&G products that are transported between regions, countries and continents using these pipelines. The USA has over 2,000,000 km, Europe has over 400,000 km and Africa has around 50,000 km of pipelines. Consequently, with increasing energy demands around the world and with O&G being the primary source of energy, pipelines are becoming increasingly important. The US Energy Information Administration’s World Energy Outlook predicts a global liquid and natural gas consumption rise from about 55 to 80bn barrel of oil equivalent by 2050 as demonstrated in Figure 1 (Energy Information Administration (EIA), 2017). According to ExxonMobil’s (2017) energy outlook forecasts, global demand for energy is expected to climb about 25 per cent by 2040 and would soar significantly higher – closer to a 100 per cent increase – but for anticipated efficiency gains across the economy. To support the growth in demand, pipeline infrastructure is anticipated to grow 7 per cent in the next 15 years, which translates to 8,000 km/year and on an international level, 32,000 km of new pipelines are constructed annually with a value of $28bn (Hopkins, 2007). In the USA, there was a pipeline investment of $101bn between 2012 and 2016 and a capital expenditure of $2.3–$3.7bn per year till 2035 is projected to meet energy demands (American Petroleum Institute, 2017).

Background: pipeline failure

However, the safety of O&G pipelines remains a sticky point due to the effect of pipeline failure on human life, the environment and other infrastructure. Figure 2 shows the effects of pipeline failure over the last 20 years in terms of fatalities and injuries in the USA alone. Figure 3 shows the monetary cost of these incidents in terms of the value of the properties damaged. From Pipeline and Hazardous Materials Safety Administration’s (PHMSA) data, it can be concluded that pipeline integrity remains a critical issue because pipeline incidents cost an average of $414m plus the fatalities and injuries due to its failure every year. More so, Europe had a total of 1,309 incidents between 1970 and 2013 for gas pipelines alone (European Gas Pipeline Incident Group, 2015).

Despite these short comings, O&G pipelines remain the most efficient and effective mode of fluid transportation when compared with rail or road transport. Statistics from the CEPA (2014) shows that their network of pipelines transports an equivalent of 15,000 lorries and 4,200 rail cars in a day. From a safety aspect, pipelines remain the safest mode of transportation of O&G products, when comparing pipeline failure statistics with other modes of transportation of O&G, it is shown that natural gas pipelines have an incident rate of 0.89 incidents per billion tonne-miles compared to road and railway which have a rate of 19.95 and 2.08, respectively (Furchtgott-Ruch, 2013).

Data from the USA show that more than 80 per cent of pipeline incidents are man-made (PHMSA, 2018). As seen in Figure 4, out of 11,761 pipeline incidents that happened between 1998 and 2017, only 18 per cent are due to natural causes and other unknown causes. The causes of failure include material/weld/equipment failure, corrosion, incorrect operation and excavation damages, which account for 33, 18, 8 and 15 per cent, respectively. This paper focusses on how these man-made forms of failure of O&G pipeline systems can be prevented or reduced by applying reliability tools at different stages of the pipeline life cycle.

Reliability engineering tools for oil and gas pipelines

There are several reliability tools that can be applied to O&G pipeline systems (Birolini, 2007), it is an engineer’s responsibility to understand each tool and how they can be applied to improve the reliability of a product. Ossai (2013) uses life data analysis (Monte Carlo simulation) to predict the time to failure for corroded pipelines, Ahmed et al. (2016) propose the use reliability block diagrams to improve the accuracy of Monte Carlo simulation, Petrovskiy et al. (2015) propose the application of failure mode effect and criticality analysis (FMEA and FMECA) for identifying both risks and hazards, hazard operability analysis (HAZOP) and hazard identification (HAZID) are applied across several industries including O&G and Anjumna et al. (2012) propose the use of fuzzy logic along with bow tie analysis, Abaqus and other software tools are used as simulation tools, non-destructive testing (NDT) techniques are used for accelerated testing and measurement, structural reliability assessment is used to determine the structural reliability of a pipeline system in DNV codes and other pipeline codes. Other tools include root cause analysis (RCA), failure reporting, analysis and corrective action systems (FRACAS), to mention a few.

Reliability optimisation

Applying these tools comes at a cost to the system; this means that the engineer must find the optimal solution between designs based on performance, cost, safety and maintainability of pipeline systems. Figure 6 shows the relationship between reliability and cost; the initial cost of a product increases with increasing reliability but the post-implementation cost decreases with increasing reliability. In other words, a product with both low and high reliability will have a high total cost because either the initial cost or post-implementation cost will be too high. Mathematically, this relationship is shown in the two cases below:

Therefore, the optimal solution between cost and reliability will be at the lowest point on the total cost curve; the optimum reliability is therefore not necessarily the highest achievable reliability but a balance of cost and reliability, finding the optimum reliability for the system which combines the multiple objectives of design (design, performance, safety and maintainability).

Case Study 1: the failure expansion tree (FET)

The FET (Lin et al., 2014) aims to improve the traditional FTA methodology, which has been shown to have various application problems. The FET is built on six main principles, which are created to avoid the problems of the traditional FTA which includes subjectivity, mutually inclusive events (repetition of failure events at different levels) and collectively in-exhaustive failure events (exclusion of some possible failure events).

The FTA provides quantitative (ranking of failure events) and qualitative analysis (determining what basic events can lead to a top event) of a system. With a well-modelled FTA, an engineer can determine the root causes of failure and quantitatively determine which failure events are most likely to occur and thus focus on them during design with the aim of preventing them. The FET solves the problem of inexhaustible failure modes by accounting for events which cannot be identified as “other” failure modes, this occurs across several levels of the FET. This leaves room for the FET to be expanded to include such rare events which cannot be done with the traditional FTA. This ensures that every level of the FET is collectively exhaustive. In addition, it solves the problem of subjectivity of the traditional FTA. When multiple numbers of engineers are trying to build an FTA, they are most likely going to come up with multiple models of FTA for the same product or system. The FET solves this problem by focussing on the physics of failure; by doing this, the multiple numbers of engineers are likely to come up with the same FET because the physics of failure is the same across all levels. Furthermore, the FET solves this problem by making the events mutually exclusive; thus, at whatever level the event is on the FET, it will produce the same probability of occurrence (failure probability). This is not the case with the FTA where the failure probability for an event depends on the level of the event. A practical example is presented in the paper (Lin et al., 2014).

Grouping the pipelines under several phases (useful life, infant mortality and wear out) as used in the FET will help in understanding the cause of pipeline failure. An example is the analysis done by Azevedo (2007) in which the pipeline was said to have corroded because of the wrong selection of the insulation used in the pipeline, a problem of manufacturing classified under infant mortality.

The FET is a more effective tool than FTA because it focusses on the physics of failure; this ensures that considerable attention is given to the system itself rather than the method used in creating the system. Hence, an engineer can identify the possible failure modes of the system and is more objective. Also, it provides the advantage of being collectively exhaustive, mutually exclusive at all levels.

Case Study 2: the condition-based fault tree analysis (CBFTA)

The CBFTA is a special FTA that improves on the FTA by applying a condition monitoring (CM) system that makes it a design tool as well as an operation and maintenance tool. Figure 8 compares the CBFTA and the FTA. The tool was created by Shalev and Tiran (2007) and applied to a two-pump system. The tool updates reliability values for a system and calculates the residual life; this is done periodically to ensure reliability is above control limit. However, the first question is whether the tool can be applied to O&G pipelines. The conditions stated for its application are as follows:

1. the critical components failure is a step-by-step deterioration process, which can be divided into predefined recognisable stages;

2. the detection of each stage is possible by using a measurement device and observation; and

3. for each failure stage, the residual time to failure is definable.

The authors believe that the tool can be used because all these conditions are fulfilled by an O&G pipeline system. Pipeline systems are composed of several components and the failure of its components is responsible for most pipeline failures (PHMSA, 2018), corrosion is the second highest cause of pipeline failures. These modes of failure (equipment failure and corrosion) can be monitored to identify the step-by-step deterioration of the entire pipeline system and the detection of each stage of deterioration is possible by using measurement devices. The evaluation of the residual time to failure is also possible, as shown by Zhou (2010), where he evaluates the failure probabilities of a pipeline with corrosive defects. From this, the reliability of the pipeline system and the time before failure can be determined using textbook reliability methods.

Two things are key in the application of this tool, CM and predictive maintenance; these are the solutions applied to the FTA to form the CBFTA making it an applicable real-time operative system analysis (Shalev and Tiran, 2007). Furthermore, the CBFTA does not only help determine the reliability of the system but also helps to determine the optimal path for operation of the system with spare components as illustrated in the paper by Shalev and Tiran. This makes it both an operation and maintenance tool. The CBFTA combines the statistical data of the FTA and the CM data for reliability monitoring and control.

However, one of the biggest issues with O&G pipelines is their accessibility for CM. This leaves design and operations engineers with the work of coming up with a list of variables which can be monitored and used to calculate the reliability of the pipeline system. Shalev and Tiran (2007) provide an example of how this can be carried out on a two-pump system in their study. The problem of availability of historical data about a system for determining the failure probability of basic events of the FTA remains.

Case Study 3: reliability prediction tool, split system approach (SSA)

This tool predicts the reliability of a complex system by using a SSA, as reported by Sun et al. (2009). It identifies the complexity of pipeline system and solves the problem by applying the SSA, it also predicts the reliability of a pipeline system that has been maintained using preventive maintenance (PM). The model was created to solve the problems of reliability models that do not account for the complexity of a pipeline system and the impact of PM actions on system reliability (Sun et al., 2009). Some questions from the analysis of this case study are follows:

• Does the repair of the repaired part affect the unrepaired part of the pipeline?

• What is the methodology for determining the recovery coefficient?

• How is the reliability determined at any time, T, when there is no PM work carried out?

• What is the difference between time-based preventive maintenance (TBPM) and reliability-based preventive maintenance (RBPM)?

First, the exposed pipeline is assumed to be the only part that is preventively maintained in all PM actions and the reliability functions of the repaired pipes are assumed to be known (Sun et al., 2009). This is a limitation of this tool because it needs to be modified for cases when the buried part of the pipeline is repaired or both parts of the pipeline are repaired because at some point in the life of a pipeline, PM actions might need to be carried out on buried sections. However, the advantage of this tool is that it provides the SSA which distinguishes the repaired section from the unrepaired section of the pipeline system.

Another limiting assumption is that the failure of the repaired part is assumed to be independent of the unrepaired part (Sun et al., 2009). While the authors understand the need for this assumption in simplifying the model, this is not the case practically. The failure of a section of the pipeline will impact the reliability of the other sections of the pipeline, for example, a pressure increase due to failure of a section of the pipeline will increase stresses on the other sections which will affect the strength of that section and other conditions, hence the reliability. This is a problem because not knowing the reliability of the unrepaired section of the pipeline means that the reliability of the entire pipeline is assumed to be constant (Sun et al., 2009). This shows the importance of having CM methods for knowing the reliability of any pipeline section. However, knowing the reliability of the pipeline in operation remains one of the biggest issues of reliability engineering (Blanks, 1992). It should be noted that the analysis done by Sun et al. (2009) indicates that if the recovery factor is reduced to zero the entire pipeline was still in a state of imperfect repair. This makes the model a valid one because it is known that maintenance work never restores the pipeline to its original reliability, this means that repair work is always imperfect.

Another important point is that it can be applied to a system with a known reliability function. This must be determined during the design phase. This shows the importance of a reliability-centred approach to design; DFR and EDBR models are at the foundation of reliability engineering application in operation and maintenance. Hence, the SSA model can also be used in the design phase to give a realistic reliability of the system that takes the complexity of a pipeline system into account. It serves as a model that can help a designer as well as operation and maintenance personnel to accurately predict the reliability of the entire pipeline system and to determine what kind of PM strategies can be utilised to prolong the life of the pipeline. The application of the tool to a real-life system gives more credibility to it, the result of this shows that a RBPM strategy is more effective for pipeline maintenance when compared with TBPM.

Case Study 4: a system for corroded pipelines

This case study presented the “system reliability for corroded pipelines”, a suitable tool for brownfield engineering in O&G pipelines (Zhou, 2010). Most issues of pipeline integrity and safety come from already existing pipelines because they are old designs in which improvements in pipeline design from failure analysis cannot be applied. It is a simulation tool. The initial size of the defects (e.g. a large initial defect size lowers the reliability in comparison to a small initial defect size), the defect growth rates and the spatial variability of the internal pressure were found to affect the reliability of the pipeline. Hence, in a brownfield system for a given pipeline with limited resources, this would provide a maintenance programme aiming at maintaining a minimum acceptable reliability level (PHMSA, 2018). This model proves that there are certain physical variables which can be monitored to determine the stages of deterioration of the pipeline system (a requirement of Case Study 2). Zhou (2010) applies this tool to an existing pipeline and finds the relationship between various variables. The failure probabilities gotten from the application of this tool can be used to determine the reliability of the entire pipeline system. One of the solutions this tool provides is that it is modelled for a small leak, rupture and large leak. This makes it versatile for different modes of failure known to pipelines.

Pipeline engineers and operators need a method for evaluating the reliability of an existing pipeline system. The corrosion of a pipeline system can only be slowed down but cannot be prevented, old pipelines have multiple defects and this tool will serve to help detect probability of failure, and hence a reliability threshold below which the pipeline should not be used. Combined with other tools presented in this paper, pipeline operators will be able to predict the reliability of the pipeline system and provide effective and efficient maintenance strategies which help to prevent pipeline failure.

The question for this tool is whether the pressure of the system is the only important variable in failure via leakage and rupture? Electrochemical properties of the system also affect the Time Needed before Action, i.e. the time before which maintenance work needs to be done. An improvement would be to include all possible variables which are affected with the fluid flow that could increase the rate at which the pipeline leaks or ruptures. Another improvement on this tool will be to find ways of determining the reliability of pipeline components which account for majority of pipeline failures. After this is done, it can be combined in a series or parallel system to determine the reliability of the entire pipeline system.

Review and analysis

When these four tools are combined, they cover the design, operation and maintenance of the pipeline. The FET (Case Study 1) can be applied at the design phase to determine the most critical failure event that is likely to occur with the help of risk ranking and focussing on it during design. Using the structure of the FET, the engineer identifies failure modes which need to be condition monitored during operation of the pipeline system which makes the CBFTA (Case Study 2) applicable by monitoring various stages of failure for the pipeline system. With this, the CBFTA becomes useful for operation and maintenance personnel because they can evaluate the reliability of the pipeline system at any given time and maintain the pipeline above a reliability threshold. Using the data collected, a modified form of the reliability prediction tool (Case Study 3) can be used to determine what the best maintenance strategy will be, the SSA model can also be used in design and operation to account for the complexity of the pipeline system and the most effective PM strategy. The CBFTA also helps to determine the optimal operation path for the various components of the pipeline system and in the final stages of the life of the pipeline, the system for corroded pipelines (Case Study 4) can be used in evaluating the correct failure probabilities of the pipeline system that can be used to keep the reliability of the pipeline system above a specified control limit. Table I shows the relevance of each reliability tool at different stages in the life of the pipeline system. However, these tools do not cover the other stages such as installation and decommissioning of the pipeline system.

Pipeline design, operation, maintenance and reliability tools

Is the integrity of pipelines a design, operation or a maintenance issue? If pipeline failure occurs despite the application of reliability tools in design, does this make the failure of pipelines a maintenance and operation issue? These are questions that need to be answered. Are O&G pipeline failures a result of poor design methods or bad maintenance or operational tools? Answers to these questions can help determine how reliability tools will be applied to each phase of the pipeline life cycle from design, to construction, to installation, to operation and to maintenance.

The results of survey and review of literature show that the design, maintenance and operation phases of the pipeline life cycle apply some reliability tools like the FMEA, FRACAS RCA, fatigue assessment, SRA and others. However, there are certain problems that affect the effective application of these tools which require further research work by all stakeholders.

First is the quality of data available (e.g. lack of reliable, accurate and representative data). Reliability tools and models require both historical and real-time data sets; historical data help in creating models which represent new and existing pipeline systems, while real-time data help to predict future failure events. The survey results show that getting these data is a problem because of the length, size as wells as conditions in which the pipeline functions and past practices which do not require collection of such data. In this case, reliability engineers need to use models that combine expert knowledge and other knowledge areas that have been applied to risk assessment models (Wei-Shing et al., 2015).

Furthermore, reliability models which are true representations of the pipeline system remain a questionable because they depend on reliable and representative data. There is a need to build standard approaches which can be used as minimum requirements for determination of reliability of any system that is generally accepted by stakeholders. In the reliability studies for lateral buckling of O&G pipelines, this study is used to predict the probability of lateral buckling occurring in a pipeline. This method is accepted because there is a standard approach to determining the probability. This shows that the O&G industry is open to methods which can improve the reliability of the systems despite the uncertainties if there is a standard approach (Beele and Denis, 2012).

Pipeline standards and codes focus on failure prevention, while there are few sections that try to add reliability concepts to complement them. Reliability engineering tools should be applied in an integrated manner throughout the lifetime of the pipeline with design methods and integrity management systems focussing on reliability going from DFR to EDBR to RCM till end of life; for example, DeSanctis et al. (2016) describe the integration of RCM methodology with RAM in the O&G industry as an application with high priority level, also a reliability-based system is a much more effective strategy as shown from the results of Case Study 3 and the Time to Action can be determined with Case Study 4, tools like the CBFTA (Case Study 2) and the reliability prediction tool (Case Study 3) should be effective in detecting failure before they occur. A focus on failure prevention methods is analogous to treating in humans the symptoms without taking care of the cause of the disease. By focussing on the reliability of the system rather than solely on the known failure modes, reliability tools and models should be more effective in preventing the failure of the pipeline system.

Legislation is required to improve the application of reliability models to pipeline integrity management. The survey results show that reliability models are applied more in countries with such legislation, as most legislation still focus on a time-based approach to integrity management. However, the effectiveness of these models in reducing pipeline failure is another matter. Hence, it is recommended that reliability engineering is applied to pipelines the same way quality assurance and quality control methods have become the standard in the industry.

Cost of reliability tools application

One of the questions asked by pipeline operators and designers is: “What is the cost of these tools compared to the cost of the pipeline failure itself?” It is evident from Figure 9 (created based on PHMSA data) that the cost of failure of O&G pipelines is very high. When reliability tools are applied effectively, they will be able to reduce the number of incidents and their costs (as pointed out earlier in Reliability optimisation section), the cost of a low reliability leads to the high post-implementation cost, of an average of over $414m annually being lost as damage to property in the USA alone (PHMSA, 2018). Therefore, pipeline operators should be willing to take these measures into consideration. However, only appropriate standards and regulations that incorporate the use of reliability prediction and RCM tools will be able to ensure that pipeline designers and operators apply efficiently the reliability tools. A standard approach to reliability tools application needs to be developed for O&G pipeline application by regulators. An in-depth research is also recommended into the cost of applying reliability tools vs the cost of damages due to pipeline failure.

Focussing on research for frequent failure modes

As mentioned in the Introduction section, O&G pipelines are a crucial infrastructure to the energy industry, as there is a high dependency on this type of energy. Legislation set by countries and disruptive innovation in the area of technology is pushing for a move to different types of energy; however, the dependency now and in the next half century is deeply rooted in the O&G industry. The integrity of pipeline systems remains a challenge of pipelines, especially in brownfield engineering; already installed pipelines will continue to fail if engineers do not find ways to predict the reliability of these aged pipelines, the same errors will be repeated if pipeline design, operation and maintenance are not reliability centred.

Data from Figures 4 and 9 indicate that the focus of research should also be on corrosion, material/weld/equipment failure and third-party damage. Since corrosion cannot be eliminated, implementing reliability monitoring methods for pipelines that maintain reliability above reliability thresholds is necessary. Also, the integrity of equipment used in a pipeline system accounts for more than half of material/weld/equipment failure (PHMSA, 2018) and a focus should be put on their integrity. To be able to achieve this could be by developing DFR and reliability testing models for equipment which can be part of procurement requirements apart from current NDT methods.