Grain Size Characterization of TA1 with GA-BP Neural Network Using Laser Ultrasonics

In this study, TA1 was selected as the research object, and EBSD was used to determine the microstructure information of the samples, then grain size considering the distribution was calculated according to the expected value and standard deviation of the log-normal distribution. The attenuation coefficients at different frequencies were extracted. Based on the Rayleigh scattering, the grain size prediction model was established by fitting the attenuation coefficient at different frequencies to the grain size. Taking the attenuation coefficients at different frequencies as inputs and grain size as output, the GA-BP neural network model was established. The average prediction error of the GA-BP model was 10.48%, which verified the feasibility of GA-BP neural network in grain size characterization and prediction.