Pipeline Risk models can be categorized into four general categories – Qualitative, Relative/Index, Quantitative & Probabilistic. While PHMSA indicates that Probabilistic modeling is the most effective, here’s a breakdown of each category & its ability to support pipeline operator decision-making.
Qualitative Models use qualitative inputs and outputs. The models translate any quantitative inputs into ranges or qualitative outputs (e.g., high, medium, low). Algorithms in these models are a direct mapping of inputs to outputs, often represented by a matrix.
Relative Assessment or Index Models use quantitative or qualitative inputs to derive quantitative outputs using a scoring algorithm. Scores assigned to inputs are combined to obtain a unit-less quantitative output “index” score. The most common method of combining inputs and obtaining model outputs is to sum the individual and sometimes weighted risk factor scores.
The quantitative outputs are not expressed in risk assessment units like probability, frequency, or expected loss. Instead, they are unit-less index scores for likelihood, consequence, and risk. This method of combining risk factor inputs and producing outputs distinguishes this model from a quantitative system or probabilistic models. Index models were widely used by pipeline operators to establish priorities for integrity assessments as part of the baseline integrity assessment requirements of the pipeline IM rules.
Quantitative System Models also have quantitative inputs and outputs. However, they are distinguished from Relative Assessment/Index models in significant ways, including:
- Use of quantitative inputs and outputs that are expressed in risk assessment units like probability, frequency, expected loss, etc. Usage of risk assessment units is an important distinction from numerical/quantitative values used in Relative Assessment/Index models that are unit-less values and only can be used to compare if they are higher/lower than other values within the model.
- Algorithms that model the physical and logical relationships of the pipeline system risk factors, the threats to system integrity, and the potential consequences of a product release from the system. This approach aims to combine risk factors in ways that more directly reflect physical reality (e.g., corrosion rates applied to effective wall thicknesses). The outputs from these models are likelihood, consequence, and risk measures expressed in recognizable units, such as probability or frequency of failure and expected loss.
Probabilistic Models are a specific type of Quantitative System model. They are distinguished from other such models by using probability distributions to represent uncertainties in model inputs. Input distributions are propagated through the model to obtain probability distributions that represent uncertainty in the model outputs, such as failure probability, the severity of consequences given a failure, or expected loss.