Anomaly Detection in Multisensory Systems

Anomaly detection and diagnosis in multi-sensor systems refer to pinpointing abnormal status in specific periods and identifying the root causes. However, building such a diagnosis tool is challenging since it requires encoding both temporal and inter-sensor dependencies such as spatial sensor information. Also, the detector system should reflect the severity of different incidents.

This webinar will focus on these points:

• State-of-the-art feature extraction methods for multi-sensor systems
• Multi-score paradigms for different levels of anomaly scores
• Signature matrices to encode inter-sensor correlations
• Synthetic and real benchmarks for anomaly detection

About the speaker

Dr. Behzad Nobakht
Data Scientist

He has a PhD in Petroleum Engineering from Heriot-Watt University. Before that he completed a research internship at Total E&P Pau, France. He also has a Master’s degree in Petroleum Engineering from Polytechnic University of Turin, and a Bachelor’s degree in Chemical Engineering from Sharif University of Technology.
Dr. Nobakht is also responsible for automating workflows using machine learning and advanced statistical methods for CBM modelling and Uncertainty Quantification (UQ).