Speaker
Description
With the establishment of radioxenon detector networks, such as the IMS, a long-standing challenge has been to accurately correlate individually detected samples that share a common source. Traditional methods, relying on classifications by human operators or simplistic time-based connections, can be time-consuming and prone to biases and oversimplifications. To address these issues, we present a machine learning-based approach designed to automate the classification and association of xenon samples across multiple stations. This approach incorporates a range of parameters, including sample strength, timing, station location, and backward weather trajectories, to group observations from the same source. The model is trained using ~10,000 simulations, and then applied on two years of data from the Swedish Radioxenon Q$_B$ array. Results demonstrate the model's ability to quickly classify samples, offering a scalable solution for plume identification, which is critical for pinpointing potential source sites in environmental and nuclear monitoring.
[email protected] |