Farid Abdalla

Summary

DL4Space is a pioneering project initiated by the European Space Agency (ESA) with the aim of exploring the potential of Deep Learning technology to enhance spacecraft operations (OPS) and Assembly, Integration and Testing (AIT) activities. Our team at HE-Arc has been entrusted with two critical use cases: one pertaining to OPS and the other to AIT. I have been actively involved in the AIT use case, which involves identifying anomalies in satellite time series data and determining their root cause.

Traditionally, anomalies were detected based on a predefined set of rules, which often resulted in a tedious and time-consuming process. With the increasing complexity of satellite systems, it has become challenging to identify the root cause of an anomaly using this approach. The objective of this project is to leverage the power of Deep Learning to classify events accurately based on available data and use Explainable AI (XAI) techniques to pinpoint the root cause of anomalies.

One of the biggest challenges we faced was working with a massive and unstructured database. Our team had to clean and process the data to make it usable for Deep Learning models. We used PyTorch to develop a Deep Learning model and incorporated XAI techniques like SHAP and LIME to gain insights into the factors contributing to anomalies.