Part 2 of this article, featuring case studies, can be found here.

Quantitative risk models based on statistics of past failures have the limitation that future failures are not reflected by historical averages and cannot be predicted.

Additionally, statistically repeatable studies cannot be performed on low probability-high consequence events, and are difficult to perform in complex systems when the failure frequencies are low and the interactions within the system are poorly understood.

Failures often appear perfectly predictable in hindsight. For that reason, DNV GL has developed a smart risk assessment methodology, Multi-Analytic Risk Visualization (MARV), to improve decision-making related to risk assessment of pipelines.

Development Of MARV

Conventional risk assessment methodologies for pipelines involve a risk-indexing approach that does not capture the fundamental processes that pose threats to the pipelines. They are subjective, and therefore can be inadequate in forecasting risks. Other major challenges of any pipeline risk assessments are:

  • The inherent uncertainty in the data or even an absence of data at many locations of a pipeline;
  • Many threat models use past inspection data to forecast future performance. To forecast, one should understand the underlying mechanisms of degradation, which for pipelines are complex;
  • Data collection can be expensive and therefore prioritizing what type of data is “really” needed is important; and
  • Visualisation of complex information on all risk assessments all in one data platform.

The MARV tool was developed over a seven-year period to improve decision-making related to risk assessment of pipelines. DNV GL’s strategic research and innovation unit initiated the development of MARV based on its existing knowledge of Bayesian modeling techniques and their extensive involvement in pipeline assessments.

The aim was not only to develop a more flexible, yet rigorous probabilistic analysis, but to deliver risk results in a touchscreen user friendly platform, thus enhancing the decision-making process.

MARV™ can combine diverse data sources, e.g., sensor networks, databases, expert information, inspection data and assessment methods. It can be categorized as a digital twin with a visual touch-screen interface. The main differentiation of MARV is that it utilizes the well-established mathematical tool of Bayesian networks to infer statistical relationships, enabling refined risk assessments.

Digitalization Enabling Smarter Assessments

We are in a digital age, one in which decision makers are demanding smarter assessments using more complex and diverse data, without increasing any operational cost. It was determined that a smart risk model of pipelines must include a minimum of three characteristics:

  • It should incorporate an understanding of failure mechanisms including interactions between different mechanisms. This can be in the form of physical models or subject matter expert input;
  • It should account for uncertainties in the data (including gaps) and in physical models used to predict failures; and
  • It should be cognitive and able to learn from erroneous predictions.

The three characteristics of a smart risk model are handled by MARV. The Bayesian network uses cause-effect relationships, including resolving the confusion of multiple interacting threat mechanisms. It accommodates uncertainties allowing ambiguous data (including gaps) to be incorporated into the analysis. It analyses complex cause-effect relationships through failure models and expert input. It displays risk information in a visual touch-screen mode that combines geographical maps, physical models, and risk outputs in one user-friendly interface.

The user sees how the system is represented and moves along a virtual representation of the asset, and observes how the risk changes from one location to another; it can move back and forth in real-time.

The system provides a quantifiable and verifiable way to incorporate the effects of mitigative actions and monitors activities on risk. The algorithms are fully transparent, allowing anyone to see the built-in logic. It allows the user to bring in a variety of information in a quantitative and objective fashion. It smartly allows automatic updating of risk, based on inspection, incidents, maintenance, or mitigation activities.

Using the same tool, the user can: perform fully quantitative risk assessments, evaluate life extension strategies, prioritize data gathering, plan mitigative actions, and explain the hidden root-cause of risks.

The tool prioritizes data gathering, recognizing that every risk assessment has uncertainties. The origin of these uncertainties comes from unreliable data, not fully understood physics, or simply variability in the threat mechanism. With MARV, the uncertainties in the calculated risk drive the data gathering activities.

Figure 1: (1) a high-level overview of the risk assessments, data on a map with simplified risk model, (2) a more detailed view of each part of the model and (3) access the probability of every state of every parameter This is done by comparing the cost of data with the effectiveness of that data on the risk results' uncertainty. It is possible to run the tool with limited and readily available information. As more information becomes available by the operators, it can be added to the model to increase the accuracy of the data. This unique stepwise approach in MARV saves resources that would be spent gathering unnecessary information.

MARV is presented as an interactive tool (see Figure 1) that extracts information from the models and displays the results on a touch screen interface. This allows stakeholders other than subject matter experts the opportunity to utilize complex physics based models, understand individual risks, assess aggregated risk and optimize the inspection interval. It significantly improves decision-making related to risk assessment of pipelines.

Ali Mirzaee-Sisan is business development manager – pipelines; Francois Ayello is ‎principal engineer, strategic research and innovation; Shan Guan is principal consultant, strategic research and innovation; and Narasi Sridhar is program director, materials technology and development, all of DNV GL – Oil & Gas.