采用遗传算法优化神经网络的铸铁表面粗糙度声发射预测
Acoustic Emission Monitor Grinding Surface Roughness of Cast Iron via BP Neural Networks and Genetic Algorithm

表面粗糙度是汽车发动机曲轴精密磨削加工中的一个非常重要的指标，在线监测表面粗糙度是曲轴智能磨削成功的标志。应用美国声学物理公司PAC的PCI-2声发射实验仪器测量磨削声发射信号，采用遗传算法优化BP神经网络，以磨削声发射信号均方根和快速傅里叶变换峰值为特征值，对平面磨削曲轴球墨铸铁材料QT700-2表面粗糙度成功进行了预测。与表面粗糙度的实测结果表明相对误差可控制在6.22%以下。
Surface roughness is a very important index in precision grinding of automotive engine crankshaft. On-line monitoring of surface roughness is a sign of intelligent grinding of the crankshaft. The PCI-2 acoustic emission experimental instrument made by PHYSICAL ACOUSTICS CORPORATION(PAC) was used in grinding test. By adopting a genetic algorithm for optimization of BP neural network and using grinding the acoustic emission signal root mean square(RMS) and fast Fourier transform peak as the characteristic values, the surface grinding crankshaft spheroidal graphite cast iron QT700-2 material surface roughness were predicted successfully. The relative error between the tested results and predicted is below 6.22%