【摘要】 针对磨削表面粗糙度声发射预测精度和可靠性较低的问题,对球墨铸铁磨削表面粗糙度的声发射智能预测进行了研究。在球墨铸铁QT700-2平面磨削表面粗糙度声发射预测实验200组数据的基础上,提取了包含磨削声发射信号经验模态分解4个本征模函数的相关系数,和磨削声发射信号波形幅值、均方根值、方差、峰值频率、频谱峰值、功率谱峰值、峭度、偏度、AE信息熵等13个磨削声发射信号特征参数;建立了遗传优化的支持向量回归机GA-SVR和粒子群优化的支持向量回归机PSO-SVR这2个预测模型;在这200组磨削表面粗糙度声发射实验数据中,把随机提取的13个声发射信号特征参数输入到这2个预测模型中,进行了反复训练和预测,以提高其可靠性。研究结果表明:GA-SVR和PSO-SVR的磨削表面粗糙度声发射预测精度较高;这为磨削声发射在线智能监测汽车发动机球墨铸铁QT700-2曲轴磨削表面粗糙度打下了基础。更多还原

【Abstract】 Aiming at the problem of low accuracy and low reliability of acoustic emission prediction of grinding surface roughness, 200 sets of experimental data of acoustic emission prediction of surface roughness of nodular cast iron QT700-2 were obtained in surface grinding experiments. 13 characteristic parameters of grinding acoustic emission signals, such as the correlation number of four Intrinsic Mode Functions including empirical mode decomposition of grinding acoustic emission signal and waveform amplitude, root mean square value, variance, peak frequency and spectrum peak value of grinding acoustic emission signal, peak value of power spectrum, kurtosis, skewness and acoustic emission information entropy of grinding acoustic emission signal were obtained. Two prediction models, genetic algorithm support vector regression machine GA-SVR and particle swarm optimization support vector regression machine PSO-SVR, were established. The 13 acoustic emission signal characteristic parameters extracted from the 200 sets of acoustic emission experimental data of grinding surface roughness were input into the two prediction models, GA-SVR and PSO-SVR, for repeated training and prediction to improve their reliability. The results show that GA-SVR and PSO-SVR have higher prediction accuracy. It lays a foundation for on-line intelligent monitoring of grinding surface roughness of nodular cast iron QT700-2 crankshaft in automobile engine by grinding acoustic emission.更多还原