Anomaly Detection

Anomaly detection, in a pipeline integrity context, is the general class of methods - statistical, rule-based, or machine-learning - used to flag a measurement or an observation that deviates from an expected baseline, whether that baseline is a prior inspection cycle, a physical model, or a learned pattern across similar segments. The hard part of anomaly detection is rarely finding deviations; sensors and models routinely find them everywhere. The hard part is ranking which deviations are worth an engineer’s attention, which is why anomaly detection on its own is a component, not a finished product - it needs to feed into prioritisation and evidence, not just a list of flags.