Neural Network for Processing Ultrasonic Signals in Flaw Detection Control Systems

Abstract

The basic methods of non-destructive testing of metallic media and their compounds are considered. The expediency of using ultrasonic testing to identify various types of defects is substantiated. It is shown that, unlike other methods, its application does not lead to the destructive consequences of a material or compounds of metallic materials. It is noted that the principle of operation of ultrasonic devices is based on the analysis of the shape and amplitude of the emitted and reflected waves from the boundary of two media. Based on the established differences in forms and amplitudes, it is possible to identify the presence of defects and determine its type. Decryption of defects is carried out by the person who decides on the danger of the defect. To make a reliable decision, he needs information about the value of the speed of propagation of ultrasound in a specific material. The speed of ultrasound in different materials differs significantly in value. It is also necessary to perform an analysis of the forms of ultrasonic waves. Neural networks make it possible to find solutions to complex problems that require analytical calculations similar to those performed by the human brain. It was found that the use of a neural network for signal processing and calibration of ultrasonic sensors reduces the calibration time. The results of the developed neural network are presented.