Times are displayed in (UTC+02:00) Amsterdam, Berlin, Bern, Rome, Stockholm, ViennaChange
Session: 11-13-01 Digitalization of Subsurface, Well Systems, Subsea Systems and Operations
Paper Number: 79757
79757 - A Deep Learning Approach for Underwater Leak Detection
The increase in costs in the exploration and production of oil and gas in deep waters has led companies in the sector to invest in innovative technologies to detect, locate and correct faults in their production systems.
Although the monitoring systems available on the industry already meet technical requirements such as robustness and real-time location, the vast majority still have the problem of producing false alarms.
The use of advanced Artificial Intelligence techniques has the main benefit of reducing the risk of false alarms during the monitoring of equipment, in addition to providing greater reliability of the monitored systems.
This research aims to develop a methodology for monitoring and detecting leaks in subsea structures based on deep neural networks, allowing automated, efficient, and less costly monitoring than conventional monitoring methodologies.
A set of monitoring data will be pre-processed for noise elimination, resolution improvement and resizing, to obtain a better performance of the algorithm. The next step consists of extracting relevant characteristics from the dataset to clearly identify the leak.
The results show the metrics used to evaluate the performance of the neural network as the accuracy and efficiency of the algorithm to detect leaks in the underwater structures and equipments.
