
A neural network will help find pipeline leaks more accurately than the standard
Specialists from the Perm National Research Polytechnic University have created an innovative system for accurately detecting leaks in pipelines. The development combines two independent methods and has already been successfully tested on an operating oil pipeline in the Perm region, the PNIPU press service reported.
The system is based on hydraulic location and a neural network. Pressure sensors installed several kilometers along the length of the pipe continuously measure parameters and transmit data to the server via a wireless network. The hydraulic location method monitors changes in the hydraulic profile, and a neural network trained on real data analyzes pressure dynamics and distinguishes emergency situations from normal operations - for example, planned switches or repair work. This allows you to filter out false signals. When a leak is confirmed by the neural network, the system automatically marks a point on the map with the coordinates of the incident and transmits a signal to the repair team. There is no need to stop the pipeline operation when installing sensors. Tests on an oil pipeline more than 15 km long showed high development efficiency: the system determined the location of the control leak with an error of 76 meters. This is 4 times more accurate than the standard, which allows a deviation of up to 300 meters.