Using ILP to Detect Anomalies in Pipelines

Simulated data generated from an accurate modelling tool can demonstrate real-life events. This approach mimics pipeline opera- tion without the need for maintaining the original anomaly record that is already scarce in the pipeline industry. This synthetic data carries precise signatures and the shape of the curve depicts the type of alarm. Learning rules can be inferred from this parametric data and the examples are con- strued from the threshold levels. The issue is addressed by considering the method that lessens data handling and the associated complexity of the problem. Probabilistic ILPs can be the most appropriate candidate for classifying anomalies of this nature. A logic program addresses this issue by learning the parameters of a program given the structure (the rules), using the ability to incorporate probability in logic programming, interpreting the examples for target predicate, and refining background knowledge for the dissemination of discoveries. This synthetic data also develops a direct link with the ILP parameter learning for competing hypotheses. The modelling tool will receive feedback, check and tune the hypothesis until exclusivity is evolved. This will fall in the domain of closed-loop active learning