Board Level Fault Prediction for the Embedded Sensor Systems at the European XFEL

Supervisors: Dr. Holger Schlarb (DESY), Prof. Thomas C. Schmidt (HAW), Prof. Görschwin Fey (TUHH)

Huge particle accelerators such as the European XFEL are complex facilities that cannot deliver service without a large ecosystem for compute and control. In particular, sensors and actuators guide operations, and management controllers monitor the system – many of which are small, embedded devices made of commodity hardware. Such hardware components undergo an aging process in regular operation, which may  accelerate in regimes of oscillating temperatures or enhanced background radiation. Unexpected hardware failures and early instabilities may occur in consequence. Like in XFEL also in many applications related to the Internet of Things, electronic boards are the most important hardware and processing components.
The main goal of the project is to search for early indicators of hardware deterioration and to design a board level fault prediction for embedded sensor systems. Based on an extensive data analysis of operational records, environmental measurements, and from artificial aging experiments, we want to design a simple model and implementation that concurrently runs on the microcontrollers at low processing cost for performing fault prediction with high accuracy. The main scientific challenge of this work is to extract such a simple, lightweight, yet expressive prediction model which can be deployed in the field without interfering with the operational tasks of the nodes.