基于RFR的Ti-6Al-4V激光超声微观组织特征检测

Importance analysis of Ti-6Al-4V microstructure characteristics on laser ultrasonic longitudinal wave velocity based on Random Forest Regression

基于激光超声检测系统,研究了Ti-6Al-4V钛合金板材微观组织特征对纵波速度的影响。根据电子背散射衍射(EBSD)技术得到不同热处理条件下试样的微观组织特征。采用变分模态分解(VMD)方法对激光超声信号进行降噪,以微观组织特征作为自变量,纵波速度作为因变量建立随机森林回归(RFR)模型,得到了对超声速度影响的特征重要性排名。研究结果表明:VMD信号处理方法能够很好地保留激光超声纵波信号,以完全再结晶比例和纵波速度进行拟合,小角晶界含量和平均晶粒尺寸的变化能够较好地解释偏离点,当特征重要性小于0.3311时,微观组织特征和声速相关性很弱。

Based on the laboratory laser ultrasonic testing system,the effect of the microstructure characteristics of Ti-6 Al-4 V titanium alloy on the longitudinal wave velocity was studied.Samples of the same specification were cut from the same batch of Ti-6 Al-4 V titanium alloy plate and different heat treatments were carried out.The microstructure characteristics of heat treated samples were obtained by electron backscatter diffraction(EBSD)technique.The variational modal decomposition(VMD)method is used to denoise the laser ultrasonic signal.A random forest regression(RFR)model was established with the microstructure characteristics as the independent variable and the longitudinal wave velocity as the dependent variable,and the importance ranking of the characteristics of the influence on the ultrasonic velocity was obtained.The results show that the VMD signal processing method can well retain the laser ultrasonic longitudinal wave signal.The variation of small-angle grain boundary content and mean grain size can better explain the deviation point by fitting the ratio of complete recrystallization and longitudinal wave velocity.When the feature importance is less than 0.3311,the correlation between microstructure characteristics and sound velocity is very weak.