| 摘 要: 为满足航天航空、电子封装、光学精密仪器等领域对工件加工质量的高要求,针对提高中高体积分数铝基碳化硅复合材料(SiCp/Al)铣削表面粗糙度的声发射智能在线监测精度的问题,通过小波包技术对铣削声发射信号进行分解,对分解后的特征值与表面粗糙度进行相关性分析,确定了最相关频段为375~406.25 kHz,筛选出相关特征矩阵,并利用GA-BP(Genetic Algorithm-Back Propagation)神经网络进行训练。研究结果表明,该方法能够实现对45%SiCp/Al铣削表面粗糙度的较小预测误差,通过成功构建的声发射预测模型,将平均预测误差控制在0.050 4左右,相比未经特征提取的BP神经网络模型,该方法的平均预测误差减小了0.072 8,为工程实践提供了可行且有效的方法。 |
| 关键词: SiCp/Al;铣削声发射;小波包分解;相关性分析;GA-BP神经网络 |
|
中图分类号: TP311
文献标识码: A
|
|
| Prediction of Surface Roughness of SiCp/Al Milling Based on Acoustic Emission and GA-BP Neural Network |
|
JIANG Houwei1, CHEN Zongyu1, LIU Deliang2, LIU Suyang2
|
(1.University o f Shanghai for Science and Technology, Shanghai 200082, China;
2.Beijing Xinghua Machinery Factory Co., Ltd., Beijing 102600, China)
1071288964@qq.com; 869558205@qq.com; ldl230@163.com; liusuyang631@163.com
|
| Abstract: To meet the high requirements for workpiece processing quality in fields such as aerospace, electronic packaging, and optical precision instruments, this paper aims to improve the intelligent online monitoring accuracy of acoustic emission for the milling surface roughness of aluminum matrix silicon carbide particles (SiCp/Al) composites with medium to high volume fractions. Wavelet packet technology is used to decompose the milling acoustic emission signals, and a correlation analysis is conducted between the decomposed feature values and the surface roughness. The most relevant frequency band is determined to be 375~406.25 kHz. The relevant feature matrix is screened out and trained by using the GA-BP (Genetic Algorithm-Back Propagation) neural network. The research results show that this method can achieve a small prediction error for the surface roughness of 45% SiCp/Al miling. Through the successfully constructed acoustic emission prediction model, the average prediction error is controlled at approximately 0.050 4. Compared to the BP neural network model without feature extraction, the average prediction error of the proposed method is reduced by 0.072 8, providing a feasible and effective method for engineering practice. |
| Keywords: SiCp/Al; milling acoustic emission; wavelet packet decomposition; correlation analysis; GA-BP neural network |
