Abstract: Considering that the characteristic information of grain size that reflects from single ultrasonic response characteristic parameter is not comprehensive enough, a combination of multiple ultrasonic parameters is proposed to construct a multi-parameter ultrasonic evaluation method for the non-destructive quantitative characterization of the grain size of GH4169. According to the correlation metric, the effective parameters are selected from 8 ultrasonic parameters, such as sound velocity, attenuation coefficient and nonlinear coefficient. A quadratic polynomial mapping model is constructed to reduce the selected multidimensional parameters to single-dimensional parameters and normalizing. In the process of fitting single-dimensional parameters and grain size, an optimization problem aiming at minimizing the average absolute error between them is constructed and solved by evolutionary algorithm to find the best mapping function and fitting function coefficient. Finally, a multi-parameter ultrasonic evaluation model is given. Compared with single-parameter sound velocity model, attenuation coefficient model and backscattering EMD model, the test results show that the model evaluation results have high accuracy, stable performance, small error and good evaluation effect. Due to the integration of multiple ultrasonic detection parameters, the response information to the grain size is retained, thereby improving the measurement accuracy and anti-interference ability.

Key words: evolutionary algorithm, grain size, mapping model, multi-parameter, ultrasonic evaluation