基于多传感器融合的微晶刚玉砂轮磨削性能在线检测系统的研究

发布时间：2018-03-30 01:18

本文选题：多传感器融合　切入点：在线检测系统　出处：《山东大学》2017年硕士论文

【摘要】：磨削加工作为一种常用的精密和超精密加工工艺,是决定工件加工精度和表面质量的关键因素之一,在实际生产中占据着重要的地位。面对小批量、多品种、高质量的产品生产需求,对砂轮和磨削工艺参数的选用要求越来越高,这与砂轮磨削性能的检测息息相关。而由于磨削加工的复杂性与磨削性能评价指标的多样性,采用单个传感器无法满足在线检测的要求,基于多传感器融合的在线检测已成为智能加工的必然发展方向。本文基于激光测微仪、多分量测力仪、声发射传感器和功率计,借助LabVIEW的开发平台和MATLAB的工具包设计开发了一套多传感器融合系统,完成了信息融合各流程的自动处理,实现了包括砂轮磨损状态的在线识别、磨削工件表面粗糙度和磨削比的在线预测在内的砂轮磨削性能的在线检测。对于多传感器融合的在线检测系统,针对多个传感器提取的特征种类众多的问题,提出了采用基于主成分分析的人工神经网络信息融合算法(PCA-ANN),通过提取的特征的关联性采用主成分分析降维,然后经人工神经网络完成信息的表达。该算法减少了提取的特征的冗余信息,加快了系统的运行速度,提升了系统在线检测的精度。针对在线检测系统的设计和开发,本文通过FFT频谱、功率谱和小波包分析等方法对信号进行频域和时频域的分析,根据分析结果选定信号时频域段落进行基于统计量的特征提取,获取与砂轮磨削性能有关的信号特征,之后对照磨削性能表征量通过PCA-ANN方法完成特征融合和信息表达。以此流程在LabVIEW平台根据信号分析、特征提取、磨削性能表征量提取的方法和信息融合算法的原理,完成了整套系统的具体编程开发。通过微晶刚玉砂轮磨削淬硬20CrMnTi工件的样本实验,对提取的特征与砂轮磨削性能表征量进行分析。避免了磨削加工时功率信号存在延迟的影响,提出了融合系统提取的特征对磨削加工工艺的指导作用,验证了系统选用特征的有效性.并通过对系统进行测试与调试,证明了 PCA-RBF方法在本领域的优异性,设计开发的系统能够实现高精度的砂轮磨损状态在线识别,高精度的磨削加工工件表面粗糙度和磨削比的在线预测。
[Abstract]:Grinding, as a common precision and ultra-precision machining technology, is one of the key factors to determine the machining accuracy and surface quality of workpiece, and occupies an important position in practical production. The requirement of high quality product production, the selection of grinding wheel and grinding process parameters is more and more high, which is closely related to the grinding performance testing. However, because of the complexity of grinding processing and the diversity of grinding performance evaluation index, The on-line detection based on multi-sensor fusion has become the inevitable development direction of intelligent machining. This paper is based on laser micrometer, multi-component dynamometer, acoustic emission sensor and power meter. With the help of LabVIEW development platform and MATLAB toolkit, a set of multi-sensor fusion system is designed and developed. The automatic processing of each flow of information fusion is completed, and the on-line recognition of grinding wheel wear state is realized. On-line inspection of grinding wheel grinding performance, including on-line prediction of grinding workpiece surface roughness and grinding ratio. An artificial neural network information fusion algorithm based on principal component analysis (PCA) is proposed. The correlation of extracted features is reduced by principal component analysis (PCA). Then the information is expressed by artificial neural network. The algorithm reduces the redundant information of extracted features, speeds up the running speed of the system, and improves the accuracy of on-line detection. In this paper, FFT spectrum, power spectrum and wavelet packet analysis are used to analyze the signal in frequency domain and time-frequency domain. According to the analysis results, the time-frequency segment of the signal is selected for statistical feature extraction. The signal features related to grinding performance of grinding wheel are obtained, and then the feature fusion and information expression are completed by PCA-ANN method compared with the grinding performance characterization. The process is based on signal analysis and feature extraction in LabVIEW platform. The method of extracting the characteristic quantity of grinding performance and the principle of information fusion algorithm have been completed. The sample experiment of grinding hardened 20CrMnTi workpiece with microcrystalline corundum grinding wheel is carried out. The feature extraction and grinding performance characterization of grinding wheel are analyzed. The influence of delay in power signal in grinding process is avoided, and the guiding effect of feature extraction by fusion system on grinding process is put forward. By testing and debugging the system, the PCA-RBF method is proved to be excellent in this field. The system designed and developed can realize the on-line identification of grinding wheel wear state with high precision. High precision on-line prediction of workpiece surface roughness and grinding ratio.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212;TG743

【相似文献】