In a Nutshell:

Operators in the U.S. adhere to prescriptive regulations to manage the integrity of their pipeline networks. These regulations are explicitly minimum safety standards, and there is an expectation that operators will go beyond them – routinely. However, in well detailed areas, such as acceptable dimensions of a crack-like anomaly, the rules are clear and comprehensive, to the point that justifying additional repairs or actions becomes challenging. In this article, U.S.-based ROSEN expert Mark Wright explores the role of risk assessment – and probabilistic risk assessment specifically – to support crack assessment decisions alongside EMAT (Electro-Magnetic Acoustic Transducer) inspection technology.

Long-established in integrity management, risk assessment is also required by regulation. In a prescriptive regulatory environment, the potential scope of risk-based decisions is somewhat limited, as many tasks are requisite, but there remains an expectation that risk results can be linked to decisions and actions. In principle, risk assessment is a method by which to make a decision, and the type of decision that needs to be made should inform the way the assessment is conducted. This article will explore the role of risk assessment – and probabilistic risk assessment specifically – to support crack assessment decisions alongside EMAT inspection technology.


The key to making risk-based decisions is to ensure the risk results are meaningful and can support those decisions. One question to ask, for example: Does the proposed action sufficiently impact the risk the action may justify? Given the potential impacts of these decisions and associated costs, it is imperative that risk results be accurate and reliable.

The probability of failure of detected and postulated anomalies may be calculated using an engineering model that predicts burst pressure. While this method infers exceedance of a limit rather than actual failure, it is analogous in the context, and the process of incorporating uncertainty within input variable distributions removes a significant amount of conservatism. Such methods are considered robust and are relatively well established.

Consequences may be calculated at points along the pipeline route, which can be obtained from in-line inspection technology. Calculating the potential costs of failure is a significant challenge due to the number of possible outcomes and the variation in conditions. A limited number of representative events, large and small, may be modelled to cover the majority of possible scenarios. Availability of data and establishment of processes in geographic information systems now means that detailed probabilistic calculations can be performed for a range of scenarios at regular intervals.

Failure events are often relatively small and occur in inconsequential areas, meaning associated costs are relatively low; nevertheless, if a large failure were to occur in a highly populated area, the consequences would be catastrophic. Since risk-based decisions account for both of these elements, it follows that the highest risks will be the confluence of elevated conditions in both.


Probabilistic assessments differ from deterministic assessments in two main ways: they convert the binary pass/fail into probability – i.e. the likelihood something has failed – which allows uncertainty to be accounted for and conservatism to be largely eliminated. Moreover, since risk equals the product of probability and consequence, equal weight is given to the consequences of an event. Thus, an anomaly in a high-consequence area may carry more risk than a more severe anomaly in a low-consequence area. By calculating an accurate representation of the probability of failure along with a representative range of consequences and their associated costs, the risk becomes an effective liability and may be compared directly to the costs associated with remediation or mitigation. This provides a very insightful metric and can be used to direct available resources to the highest-risk areas. Additionally, a number of other results become available that can also offer insights. A comparison of the baseline and ongoing risk, for example, provides a measure of the impact of the entire process.


To demonstrate these points, a study was conducted on a series of pipelines subject to in-line inspection using EMAT crack detection technology. In all cases, a probabilistic risk assessment was carried out post-inspection, which included all reported anomalies, and post-remediation once all U.S. regulatory responses had been addressed.

Figure 1 – Results of the study using EMAT crack detection technology

Figure 1 – Results of the study using EMAT crack detection technology


The results of the analyses are displayed cumulatively i.e. as the sum of all risks as they accumulate along the distance of the pipeline, with large increases denoting localized areas of higher risks. The monetary values denote the annual liability associated with each pipeline and represent reasonable direct costs. The analyses reveal significant differences in the risk profiles: moderate risks are accumulated throughout the length of Pipeline 1 and cumulate to annual liabilities of around $200,000, measured post-inspection. At this level, it is marginal whether the annual liability and associated costs of the inspection and remediation activities may be justified. In the case of Pipeline 2, the vast majority of risk is concentrated in a specific area and totals almost $9 million, a significant amount that clearly justifies reduction. The results demonstrate where risk is useful and meaningful: anomalies reported in Pipeline 2 were not significantly worse than those reported in Pipeline 1. Neither were there higher-consequence areas. Indeed, the most significant consequence areas were located along the route of Pipeline 1. The most significant risks cumulated where there was a confluence of anomalies and high-consequence areas.


Initial repair decisions are dictated by prescriptive regulatory response criteria; examining the post-remediation results shows the impact of these regulations upon risk reduction. There was not a significant difference in the number of repairs, but the difference in the impact of those repairs was clear. As may be expected, repairs deemed necessary by regulation reduce risk significantly but do not eliminate risk completely. The critical question is whether the residual risk is acceptable. In the case of Pipeline 1, the answer is affirmative: the residual liability is around $70,000 per year, distributed throughout the line. It would be difficult to justify further expenditure on this pipeline based on risk. Pipeline 2 is different: the residual risk level remains at $1 million per year and is largely concentrated in a single area. Additional analyses found that by repairing four further anomalies – not required by regulations – annual liability could be reduced to less than $100,000 per year, which is broadly equivalent to the level in Pipeline 1. When measured against the costs, these actions are clearly justified. There is a clear trend towards probabilistic assessments in the pipeline industry, but finding where they fit best has been challenging. This application shows a clear use case and augmentation of the process and subsequent decisions.


Identifying pipelines at risk of cracking threats is challenging. Even in cases where threats like stress corrosion cracking can be deemed non-credible, there remains the risk of long-seam cracks. Since in-line inspection data is fundamental for this type of analysis, a first inspection and assessment cannot be avoided. But the level of insight, impact and information can be used to make more informed decisions in the future. The risks identified on Pipelines 1 and 2 have not been eliminated, even though they were reduced significantly to the point where future actions, including inspection and assessment, may be considered marginal, at least in the shorter term. Other pipelines where the risk is even lower provide even greater justification for deferment.

Risk assessment is a powerful tool, and probabilistic risk in particular provides accurate and objective results. New capabilities in software enable greater range and repeatability of complex routines, meaning such analyses can now be performed with relative speed and ease. Addressing the risk of cracks and moving beyond regulation will rely on having these sorts of analyses to justify decisions.