Neural Network based Fault Diagnosis Procedure for the Detector System of CFDF

Abstract

This paper outlines and deals with the problem of fault detection, isolation and identification of the four-elements detector system attached to the Cairo Fourier diffractometer facility (CFDF) used for neutron time-of-flight (TOF) spectrum measurements. A feed forward neural network and error back propagation training algorithm are employed to diagnose four commonly occurring faults of the detector system: preamplifier, amplifier, discriminator and the high voltage. The diagnostic system processes the acquired data to determine whether the detector system state is normal or not. The experimental results showed that the trained network has the capability to detect and identify various faults which can make one of the detector units to be out of order.