Ultrasonic Nondestructive Evaluation Of Steel Aging Microstructure Based On Wavelet And Neural Networks

Material nondestructive evaluation (NDE) is a subject evaluating integrity of objects without damaging the properties of materials. By interaction between ultrasonic waves and elastic medium, it carries much information of material, which is received in the form of ultrasonic echo signals. Recently, the digital signal analysis and processing is applied in NDE extensively with the development of computer and information technology, so the existing test methods can be used effectively and sufficiently to find the rich and undiscovered information in signals and gain the greatest benefit with the least cost. With the unused information and the automatic recognition for signal character parameters, the ability of quantify and reliability of traditional ultrasonic nondestructive evaluation are advanced, the method has been accepted by workers in the filed of ultrasonic nondestructive testing.Since there is a close relationship between microstructure and properties of material, the testing and characterization of material microstructure is the important branch of material science, as well as the major part of ultrasonic nondestructive evaluation. Different microstructure has different mechanical properties, which will result in the variation of sonic velocity and modulus decay, the information parameter relative with microstructure in echo signal will also have a change. Based on above theory, the information parameters relative with microstructure of material are extracted and interpreted exactly by wavelet transformation, different microstructures were classified based on these parameters by BP neural networks to accomplish ultrasonic nondestructive evaluation of material microstructure.The samples are 20 steel, ultrasonic testing technique based on water immersion line-focusing is used in this experiment, the wavelet coefficients are extracted and optimized to characterized different microstructure based on the theory of wavelet transform ,then BP neural networks is established to classified different microstructure according to optimized parameters. For sake of certifying the generalization of ultrasonic nondestructive evaluation, Cr5Mo containing four sorts of high-temperature aging microstructure as well as 30Mn2SiV structural steel containing three sorts of rolled microstructure are selected as samples. The discrimination of three sorts of materials are above 80% by this method.