November 2025 Profile: Ahmad Tourei
Enhancing Large-Scale Distributed Acoustic Sensing (DAS) Data Management with Unsupervised Anomaly Detection
Distributed acoustic sensing (DAS) systems can generate massive datasets due to their high spatial and temporal resolution, creating significant challenges for data management and interpretation. Using this technology, a fiber-optic cable is turned into thousands of seismic or strain sensors. Ahmad Tourei, a Ph.D. candidate at Colorado School of Mines, is working under the supervision of Dr. Eileen Martin to address this challenge through the development of das-anomaly, an open-source Python toolkit that applies deep learning to streamline large-scale DAS data processing.
As illustrated in the figure, the das-anomaly framework utilizes autoencoder-based unsupervised learning to detect and classify anomalous signals from power spectral density (PSD) representations. By compressing DAS data into a low-dimensional latent space and applying a probabilistic metric for anomaly detection, the model distinguishes anomalies from typical background noise. Transient or impulsive events such as microseismic and cryoseismic activity, traffic signals, and interrogator noise can be detected without prior need for labeled data. This enables researchers to focus computational and human resources on segments of interest while significantly reducing overall data volume.
Tourei’s work establishes a scalable foundation for distributed sensing big data, which can facilitate near-real-time DAS data analysis for a range of applications. The das-anomaly toolkit is freely available on GitHub at https://github.com/ahmadtourei/das-anomaly, promoting open and reproducible research in distributed sensing.
For more information, connect with Ahmad Tourei directly at tourei@mines.edu or https://www.linkedin.com/in/ahmad-tourei/.