Failures due to external interference are often the most significant contributor to probability of failure in risk assessments completed by operators, and hence drive significant decisions on risk control expenditure. External interference is very challenging to predict where and when it may happen. However, top-of-line deformation damage (dents) reported by in-line inspection (ILI) are a clear indicator of past external interference. Machine learning models trained on ILI data from ROSEN’s Integrity Data Warehouse (IDW), have been used to estimate a “hit rate” i.e. the frequency of external interference damage (per km/year), and their variation with relevant factors such as population density. These results can then be combined with the probability of degradation to failure to estimate the failure frequency for input to pipeline risk assessment.

The degradation to failure is assessed using industry standard engineering models. However, the prediction of the hit rate can often be subjective or based on statistics, which may not always be applicable to the pipeline under assessment.

The following factors may all influence the likelihood of external interference damage occurring:

• local population density;
• land use;
• frequency of crossings;
• pipeline burial depth;
• additional impact protection;
• pipeline markers and warning tape;
• surveillance frequency; and,
• pipeline material properties.

The United Kingdom Onshore Pipeline Operators Association (UKOPA) publishes data on the total number of reported individual external interference defects that result in a loss of containment and those that do not cause immediate failure, and operating exposure. This data can be used to estimate a probability of damage but it is specific to onshore pipelines designed and operated in the UK. This paper presents corresponding statistics that have been generated using European ILI data and additional predictor variables, based on pipeline exposure, resistance and mitigations to predict more accurate and justifiable hit rate estimates than current methods can deliver.