Pixel Advisor: An Expert System for the ATLAS Pixel Detector Control System

Since December 2009, the Large Hadron Collider (LHC) and its four main experiments are taking physics data. Each of those four experiments consists of sub-detectors, which are responsible for energy, momentum and particle trajectory measurements. As the innermost sub-detector of the ATLAS experiment, the Pixel detector provides precise measurements of momenta and trajectories of charged particles, which are for example needed to identify and separate the primary and secondary vertices. The ATLAS Pixel detector makes high demands on its control system, as more than 30000 parameters like voltages, currents and temperatures have to be monitored and controlled “twenty-fourseven”. Several software control and automation layers including a finite state machine are employed to filter the incoming information and provide a clear detector user-interface. In addition, these software layers allow for detector operation by the use of just a few buttons under default conditions. While the multi-layered user-interface of the control system software, which is based on the PVSS supervisory control and data acquisition system, enables normal detector operation by trained shifters, expert knowledge is still needed to solve problems that arise. In addition to the available on-call experts, a Java JBoss-Rules based expert system called “Pixel Advisor” has therefore been developed to provide help in solving error conditions. The presented paper describes the basic principles of this expert system and the first experience from its use.