Speaker
Description
At MAX IV, accelerator downtimes are classified strictly according to their root cause. When no root cause can be identified after exhaustive investigation, the downtime is labeled as unknown. Historically, there have been several periods during which a short-term increase in unknown downtimes has been observed. Such increases can for example be caused by a component malfunctioning in a way that is not detected by traditional diagnostics.
In this poster, we present a statistical model of the unknown downtimes at MAX IV as a Poisson process with a baseline downtime rate and periods with an elevated downtime rate. We show how this method can be used to quantify the statistical evidence that a given period has an elevated rate. This helps identify periods in which the accelerators may be affected by an issue that is not captured by existing diagnostics, supporting the detection of problems that are difficult to diagnose.