【摘要】 目的 借助机器学习算法对超声波衰减系数-平均等效晶粒尺寸模型（AC-AGS）进行优化，以实现高效准确的铜材晶粒尺寸预测。方法 通过超声检测设备获得不同温度热处理后的TP2铜材的界面波幅值以及一次底面回波幅值，并通过幅值数据计算衰减系数，采用截点法获得铜材的晶粒尺寸。为优化铜材晶粒尺寸的预测模型，引入3种机器学习优化算法，即粒子群算法（PSO）、遗传算法（GA）以及差分进化算法（DE），对TP2铜材的晶粒度预测模型进行优化。结果 耦合机器学习算法的AC-AGS模型可以实现精密铜材平均晶粒尺寸的评测，PSO算法在本文优化结果中表现最佳，其验证结果的决定系数R~2为0.904 5，均方根误差（Root Mean Square Error,RMSE）为20.85，平均绝对误差（Mean Absolute Error,MAE）为20.569。结论 结合机器学习优化算法的超声波衰减系数与平均晶粒尺寸模型，实现了对TP2铜材晶粒尺寸的精确预测。该模型不仅为铜材晶粒尺寸的无损检测提供了一个高效、准确的预测工具，还为其他材料的晶粒尺寸预测提供了新的思路和方法。更多还原

【Abstract】 The work aims to optimize the ultrasonic attenuation coefficient-average equivalent grain size model(AC-AGS) with the help of machine learning algorithms, so as to achieve efficient and accurate prediction of copper grain size, and provide key technical support in industrial production. The interface amplitude and the primary bottom echo amplitude of TP2 copper after heat treatment at different temperature were obtained by ultrasonic testing equipment, the attenuation coefficient was calculated by amplitude data, and the grain size of copper was obtained by the intercept method. In order to optimize the prediction model of copper grain size, three machine learning optimization algorithms, namely particle swarm optimization(PSO), genetic algorithm(GA) and differential evolution algorithm(DE), were introduced to optimize the grain size prediction model of TP2 copper. Results showed that the AC-AGS model coupled with the machine learning algorithm could realize the evaluation of the average grain size of precision copper. It was found that the PSO algorithm performed the best in the optimization results in this paper, and its coefficient of determination R~2 was 0.904 5, the Root Mean Square Error(RMSE) was 20.85, and the Mean Absolute Error(MAE) was 20.569. In conclusion, combined with the ultrasonic attenuation coefficient and the average grain size model of the machine learning optimization algorithm, the accurate prediction of the grain size of TP2 copper is realized. This model not only provides an efficient and accurate prediction tool for the non-destructive testing of copper grain size, but also provides new ideas and methods for the grain size prediction of other materials.更多还原