基于PCA和SVM的汽车涂装线机电设备智能诊断

发布时间：2018-06-05 06:15

本文选题：PCA + SVM　；参考：《浙江工业大学》2012年硕士论文

【摘要】：随着新的汽车涂装生产技术、生产工艺以及大型复杂设备的不断涌现和迅速发展,为了保证生产和设备高效、可靠的运行,因此对汽车涂装系统各机电设备运行状态的准确诊断提出了更高的要求。由于智能理论的发展,设备状态判别进入了智能化发展阶段。本文详细研究了主成分分析法和支持向量机在涂装线设备诊断中的应用,并结合虚拟仪器进行了涂装线设备监控与智能诊断系统的设计。论文主要内容如下: 1.根据汽车涂装线生产工艺,对各子系统主要设备的故障机理进行了分析研究,指出了设备经常发生的故障类型和征兆,在各系统内建立了数据采集系统。 2.研究了故障征兆提取技术。在实际环境下,众多传感器采集到的信号,一方面,并不是所有变量都反映设备状态的重要信息,有些会干扰诊断;另一方面,设备信号特征的输出有一定相关性。因此,论文讨论了主成分分析方法和改进的主成分分析法的应用。通过烘房燃烧加热系统设备的实例对比分析,验证了此方法的优势。 3.较深入研究了核函数类型及核参数对分类器精度的影响。通过双螺旋数据样本仿真试验,分别分析了高斯核和多项式核对分类精度的影响,以及高斯核宽度系数和惩罚参数对分类精度的影响,表明高斯核参数和惩罚参数在某个范围时,分类器精度最好。 4.提出了一种基于主成分分析和支持向量机的设备状态分类识别方法。结合主成分分析法的特征提取和向量机的识别优势,采用网格搜索交叉验证法寻求最优核参数,来建立向量机训练模型。通过烘房燃烧加热系统4种设备的12种状态进行了验证。分别分析了主成分分析法改进前后的分类精度,识别率都基本达到85%以上。 5.结合虚拟仪器进行了涂装线设备监控智能诊断系统的设计,对数据采集、时域和频域分析、特征提取和智能诊断等模块进行了介绍。 6.最后,对全文进行了总结,并对进一步的研究提出一些展望。
[Abstract]:With the new auto painting technology, production technology and the rapid development of large and complex equipment, in order to ensure the efficient and reliable operation of the production and equipment, the accurate diagnosis of the operating state of the mechanical and electrical equipment of the automobile coating system is higher. The application of principal component analysis (PCA) and support vector machine (SVM) in the diagnosis of coating line equipment is studied in detail, and the design of monitoring and intelligent diagnosis system for coating line equipment is carried out with virtual instrument. The main contents of this paper are as follows:
1. according to the production process of the automobile coating line, the failure mechanism of the main equipment of each subsystem is analyzed and studied. The types and signs of the equipment often occur, and the data acquisition system is set up in each system.
2. the fault symptom extraction technology is studied. In the actual environment, the signals collected by many sensors, on the one hand, not all the variables reflect the important information of the equipment state, some will interfere with the diagnosis; on the other hand, the output of the signal features of the equipment has some relevance. The application of the sub analysis method is compared with the example of the heating system in the drying room, and the advantages of this method are verified.
3. the effect of kernel function type and kernel parameter on classifier precision is studied in depth. The influence of Gauss kernel and polynomial nucleation on classification accuracy is analyzed by double helix data sample simulation test, and the influence of Gauss kernel width coefficient and penalty parameter on classification precision are analyzed, and Gauss kernel parameter and penalty parameter are shown in a certain range. The classifier has the best accuracy.
4. a new method of equipment state classification and recognition based on principal component analysis and support vector machine is proposed. Combining the feature extraction of the principal component analysis and the recognition advantage of the vector machine, the grid search cross validation method is used to find the optimal kernel parameters, and the training model of the vector machine is established. The 12 states of the 4 equipment of the heating system are through the drying room combustion heating system. The classification accuracy is analyzed before and after the improvement of principal component analysis. The recognition rate is basically over 85%.
5. the design of the intelligent diagnosis system for the monitoring and control of the coating line equipment is carried out with the virtual instrument. The modules of data acquisition, time domain and frequency domain analysis, feature extraction and intelligent diagnosis are introduced.
6. finally, we summarize the full text and make some prospects for further research.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：U468.22;TH165.3

【参考文献】