Ultrasonic location of inner crack defects in a compound elastic cylinder using an artificial neural-network apparatus

Abstract

The identification of inner defects was performed using ultrasonic control methods and an artificial neural network device. Models of a pipe cross section with an inner ring pad and the propagation of an ultrasonic wave were developed. The influence of different geometric parameters of the studied objects on the indicators of neural-network learning was studied. The optimal structure of a neural network and the form of training vectors for some ultrasonic control methods and their combinations were established.