Towards Predictive Maintenance at the European XFEL Using Machine Learning and Software Engineering Techniques

DASHH Doctoral Researcher: Amna Majid

Supervisors: Dr. Steve Aplin (EuXFEL), Prof. Walid Maalej (UHH)

Research facilities based on particle accelerators, like the European XFEL (EuXFEL), are multifaceted systems that require thousands of interconnected components and devices for their control and operation. Failure in any single component may result in unexpected shutdown and performance degradation. Therefore, early detection of irregularities and predictive maintenance routines are crucial for such facilities. Accelerators generate a huge amount of data that, if analyzed systematically, can be a viable indicator of their condition. We propose a software framework that uses machine learning to detect anomalies in the EuXFEL and its control system beforehand. We also explore how to increase the prediction reliability by using recent advances in engineering adaptive systems as well as Human-In-The-Loop paradigm.