The Application of Artificial Neural Networks to Non-destructive Identification of Microstructure of 20 Steel

The wavelet coefficient of ultrasonic signal was extracted as the characteristic parameter of microstructure.The application of BP neural networks to the non-destructive identification of spheroidized microstructure of 20 steel and rolled microstructure of 30Mn2SiV was discussed.The BP neural networks is available for non-destructive identification and classification of 20 steel spheroidized microstructure as well as 30Mn2SiV rolled microstructure.The average reliability is 86.6% and 88.3% respectively.It provides an advanced and valid method for analyzing the microstructure of steel.