Softcom led the design, implementation, and validation of the ProPML (Proactive Performance Monitoring Engine) project, completed in 2020, which employed machine learning techniques to predict failures in ground systems. This initiative focused on collecting and analyzing large volumes of data from ESOC’s operational ground segment systems—including SCOS, LMS, GFTS, ARES, MATIS, NIS, SIMSAT, FARC, DARC, PARC, IFMS, SMF—and across different missions, such as GAIA and Sentinel-2. The project aimed at performing root cause analysis, predicting system failures, and delivering a standardized dataset for training predictive machine learning algorithms.

ProPML also benchmarked various algorithmic approaches, comparing models trained on data from space missions with those trained on avionics industry data, and assessed their effectiveness in supporting a proactive, real-time operator warning system. The project included complex visualizations of predictions and cross-checked feedback from mission teams with anomaly databases like ARTS and Überlog.

As the prime contractor, Softcom was responsible for the project’s concept development, selection of use cases, data preparation, dataset creation, and the selection, implementation, and evaluation of machine learning algorithms. The team leveraged technologies such as Python, the ELK stack, Beats framework, RedHat Log Anomaly Detector, Docker, and Jupyter Notebooks to deliver a reliable solution.

This project highlights Softcom’s expertise in predictive machine learning and its commitment to advancing proactive performance monitoring for ESA’s ground systems.

End Customer:  ESOC   

Team:   SoftCom-Int, Etamax space       

Position: Prime

under a programme of, and funded by, the European Space Agency