A real-time FPGA based monitoring and fault detection processing system for the Beam Wire Scanner instruments at CERN

The CERN Beam Instrumentation group (BE-BI) is designing a new generation of an instrument called Beam Wire Scanner (BWS). This system uses an actuator to move a very thin wire through a particle beams, back and forth with a movement stroke of pi [rad]. To achieve very fast speed when touching the particle beam with such a small stroke, large torque is applied while the expected smoothness of the displacement is crucial. This system relies on a resolver to determine the angular position and power correctly its Permanent Magnet Synchronous Motor (PMSM). In 2016, a failure of the position acquisition chain has highlighted the severe consequences of such problem, which resulted by 24 hours downtime of the Super Proton Synchrotron (SPS) accelerator with significant financial consequences. This work mitigates this single failure point by taking advantage of the existing redundancy in the sensors embedded on the system. The resolver is compared to two Incremental Optical Position Sensor (IOPS) developed in-house and redundant motor currents sensor are allowing failure detection of this part. The processing capability of the large Field-Programmable Gate Array (FPGA) present on the actuation controller is also taken advantage of. The resulting system is more robust against sensor failures and provides more reliable measures. An emphasis is made to the methodology of the design and development to unsure the high confidence in the results. The prototyping algorithms have been developed using the software MatLab prior to their implementation in VHSIC Hardware Description Language (VHDL). The simulation vectors are coming from a field measurements recorder prior to this thesis and extra failure modes have been simulated by creating artificial data starting from these existing sets. The output produced by the algorithms written in the two languages have been finally compared to detect differences due to eventual implementation issues or number representation limitation. The processing architecture developed during this project is able to detect numerous sensor acquisition chain failures. By processing and monitoring each sensor individually, many issues can already be detected. The final detection is made by comparing the value delivered by these sensors. In addition, one sensors fusion architecture is implemented to provide position and speed based on redundant information. In addition to the work on the algorithms, during the design for the processing of the IOPS, a new type of optical disk was developed which could overcome the limitation of this sensor. By embedding the rotation direction and the absolute position directly into the unique mirrors track, this incremental encoder could be turned into a absolute encoder and improved the system reliability. The outcome of this work will allow to deliver a more reliable and robust system based on complex technologies when compared to previous Beam Wire Scanner (BWS) generations. Giving to the accelerators operators and users high performance with better availability.