基于模糊C均值及粒子群参数优化的支持向量机故障诊断方法研究

发布时间：2018-06-22 14:18

本文选题：支持向量机 + 故障诊断　；参考：《电子科技大学》2011年硕士论文

【摘要】：支持向量机是一种在统计学习理论基础上发展起来的新型的机器学习方法。在解决小样本问题、高维问题和局部极值的问题时表现出了十分优良的特性,另外,支持向量机还具有十分简单的结构,这两点决定了支持向量机在人工智能领域的特殊性,且适于应用在故障诊断领域。由于支持向量机的训练样本具有冗余性,有些训练样本距离分类面很远,并且还不会对分类结果产生很大的影响,这些样本是可以剔除的。另外,支持向量机的参数选择对分类结果也会有影响,因此本文以减少支持向量机训练时间,提高分类器准确率为出发点,对支持向量机训练样本的预处理和参数优化进行了研究和改进。故障诊断问题是十分复杂的,故障诊断技术的发展是向着智能信息处理技术前进的。本文主要是研究基于模糊C均值的支持向量机训练样本预处理及粒子群参数优化的支持向量机的故障诊断方法,探索新的、更高效、精度更高的故障诊断分类方法。本文首先为了对支持向量机的冗余训练样本进行预处理,研究了基于模糊C均值算法的支持向量机的实现,并通过数值试验来验证了模糊C均值算法对支持向量机训练样本的处理效果,结果表明,在分类准确率相差不大的情况下,使用模糊C均值算法对支持向量机的训练样本进行预处理的分类时间被大大缩短了。接着研究了基于粒子群算法的最小二乘支持向量机的参数优化问题,给出了算法的步骤并进行了数值验证,得出采用粒子群算法对参数进行优化可以提高分类准确率的结论。在分别验证了使用这两种方法对算法的改进作用之后,最后将两者结合,采用基于模糊C均值和粒子群参数优化的最小二乘支持向量机来对滚动轴承故障数据进行了分类测试,通过与传统向量机、使用上述两种方法改进的支持向量机的比较,本文所提出的算法在保证较高的分类准确率的情况下,可以有效地减少训练时间,提高分类效率。
[Abstract]:Support vector machine (SVM) is a new machine learning method developed on the basis of statistical learning theory. In solving the problem of small sample, high dimension and local extremum, support vector machine (SVM) has a very simple structure, which determines the particularity of support vector machine (SVM) in the field of artificial intelligence. It is suitable for fault diagnosis. Because the training samples of SVM are redundant, some of the training samples are far from the classification surface and will not have a great influence on the classification results. These samples can be eliminated. In addition, the selection of support vector machine parameters will also have an impact on the classification results, so this paper takes reducing the training time of support vector machine and improving the accuracy of classifier as the starting point. The pretreatment and parameter optimization of SVM training samples are studied and improved. The problem of fault diagnosis is very complex, and the development of fault diagnosis technology is advancing towards intelligent information processing technology. In this paper, the fault diagnosis method of SVM based on fuzzy C-means training sample preprocessing and particle swarm optimization is studied, and a new, more efficient and accurate fault diagnosis classification method is explored. Firstly, in order to preprocess the redundant training samples of SVM, the realization of SVM based on fuzzy C-means algorithm is studied. The numerical results show that the fuzzy C-means algorithm can deal with the training samples of SVM. The results show that the classification accuracy is not different from each other. The classification time of training samples of support vector machine (SVM) is greatly reduced by using fuzzy C-means algorithm. Then the parameter optimization problem of least squares support vector machine based on particle swarm optimization algorithm is studied. The steps of the algorithm are given and the numerical results are given. The conclusion is drawn that the particle swarm optimization algorithm can improve the classification accuracy. After the improvement of the algorithm is verified by the two methods, the least squares support vector machine based on fuzzy C-means and particle swarm optimization is used to classify and test the rolling bearing fault data. Compared with the traditional vector machine, the proposed algorithm can effectively reduce the training time and improve the classification efficiency by using the improved support vector machine.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TP18;TH165.3

【引证文献】