A global gas distributor.
Underground gas pipelines are monitored with sensors in order to detect leaks and unusual flow behaviour. If the sensors break, there is a chance that pipeline leaks will go undetected. Since leaks are extremely damaging to the environment, preventative detection is essential. Faculty was asked by a gas distribution company operating long-distance pipelines to help them detect faults in the ultrasonic flow meters they use to monitor their network.
Faculty partnered with Palantir to build new capability to detect and categorise pipeline sensor anomalies before they became failures, effectively combining engineering knowhow with data science expertise. The model built was able to detect unusual signal behaviour, from which potential faults could be surfaced.
Our model provides valuable decision-support to engineers, flagging a select number of more likely faults for review, and underpinning future predictive capability.