Pipeline Rupture Detection Using Real-Time Transient Modelling and Convolutional Neural Networks

Demonstrating the ability to reliably detect pipeline ruptures is critical for pipeline operators as they seek to maintain the social license necessary to construct and upgrade their pipeline systems. Current leak detection systems range from very simple mass balances to highly complex models with real-time simulation and advanced statistical processing with the goal of detecting small leaks around 1% of the nominal flow rate. No matter how finely-tuned these systems are, however, they are invariably affected by noise and uncertainties in a pipeline system, resulting in false alarms that reduce system confidence. This study aims to develop a leak detection system that can detect leaks with high reliability by focusing on sudden-onset leaks of various sizes (ruptures), as opposed to slow leaks that develop over time. The expected outcome is that not only will pipeline operators avoid the costs associated with false-alarm shut downs, but more importantly, they will be able to respond faster and more confidently in the event of an actual rupture. To accomplish these goals, leaks of various sizes are simulated using a real-time transient model based on the method of characteristics. A novel leak detection model is presented that fuses together several different preprocessing techniques, including convolution neural networks. This leak detection system is expected to increase operator confidence in leak alarms, when they occur, and therefore decrease the amount of time between leak detection and pipeline shutdown.