High-performance accelerators have higher requirements for operational reliability and stability. By analyzing the historical data that is routinely saved during accelerator operation, most failures can be judged. However, when some rapid failure processes occur, due to the insufficient granularity of the historical data stored conventionally, it is impossible to effectively analyze such rapid failure processes. When a failure occurs in a particle accelerator, fast acquisition techniques are needed to collect large amounts of data from various devices with precise timestamps. The failure occurrent process can be rapidly reconstructed by using these data to locate and judge the root cause of the failure. In order to obtain data accurately when a failure occurs, hardware devices with data cache function can be used at the front-end devices, and data can be locked and obtained in the synchronous trigger mode. That is, after receiving the synchronous trigger signal, data in the cache area of the front-end hardware device can be locked, and then read and stored.
This study aims to design a failure analysis system prototype based on the event-timing technique.
Two core parts of the prototype were implemented: global high-precision timestamp implementation and data assembly and acquisition analysis. As one of the key factors, the global high-precision time stamping of failure data was applied to analyzing failure causes. Based on a high-performance rubidium atomic clock and the event-timing system, high-precision time stamps were implemented in this prototype with synchronization accuracy better than 16 ns to provide global high-precision time stamps for time data. Structured data based on the normative type of EPICS 7 was adopted for assembling and publishing the data. Essential information, including the system name, the subsystem name, the device name, the device card number, the data sampling frequency, the event timestamp, and the latched data, was obtained from the structured data.
The prototype experiment results show that the failure sequence of different equipment can be distinguished by the obtained high-precision time data, confirming the high feasibility of our proposed failure analysis system.
The prototype designed in this study meets the requirements for rapid failure analysis of particle accelerators. And this prototype will be applied to the CSNS accelerator in the near future. In addition, it can also be applied to EPICS-based and event-timing based accelerator control systems.
