Automated Availability Statistics

The availability of a particle accelerator or any large machine with users is not only of paramount importance but is also, at the end of the day, an oft quoted number (0 to 100%) which represents (or is taken to represent) the overall health of the facility in question. When a single number can somehow reflect on the maintenance, operation, and engineering of the machine, it is important to know how this number was obtained. In almost all cases, the officially quoted availability is generated by hand by a machine coordinator, who peruses the operation statistics over the period in question. And when humans are involved in such a calculation there might be a latent tendency to avoid the stigma of low availability. So, not only might there be scepticism at 'impossibly high' availability, but comparing quoted availability from one machine with another might turn out to be virtually meaningless.In this paper we present a method for calculating the machine availability automatically, based on the known (and archived) machine states and the known (and archived) alarm states of the machine. Ideally this is sufficient for a completely automatic determination of the availability. This requires, however, a perfect representation of all possible machine states and a perfect representation of all possible fatal alarms (those leading to down time). As achieving perfection is ever an ongoing affair, the ability for a human to 'post-correct' the automated statistics is also described