A neural-network method for diagnosing beam-position monitors in storage ring

Abstract

We propose a neural-network method to diagnose beam-position monitors in a storage ring. Since a circulating beam in a storage ring passes through many beam-position monitors, there is a correlation among measurements of the orbit at the monitors. A perceptron is trained to predict the orbit at a particular beam-position monitor based on orbit measurements at other beam-position monitors. If the perceptron's prediction is significantly different from the actual measurement, the corresponding beam-position monitor is considered to be malfunctioning. Experimental results from the storage ring of the Pohang Light Source indicate that the scheme could be used for fault diagnosis of beam-position monitors in the storage ring. The accuracy of the perceptron's prediction is found to be better than that of the numerical approach based on the measured sensitivity matrix. The neural-network approach is more economical, less disruptive and less time consuming than the numerical approach.