基于改进BFA的旋转机械故障诊断核参数优选研究

发布时间：2018-01-02 07:33

本文关键词：基于改进BFA的旋转机械故障诊断核参数优选研究　出处：《湖南科技大学》2011年硕士论文　论文类型：学位论文

【摘要】：旋转机械是现代工业生产中的核心设备,开展旋转机械故障诊断技术研究,对于确保此类设备安全、高效运行,避免巨大的经济损失和灾难性事故的发生,具有极大的经济、社会意义,同时也是对机械设备状态监测与故障诊断技术的丰富与发展。论文以旋转机械为对象,开展基于典型核方法——核主元分析(Kernel Principal Component Analysis,KPCA)和支持向量机(Support Vector Machine,SVM)的故障诊断方法与技术研究,针对KPCA、SVM性能受核函数及其参数影响很大,而最优参数难以选取的问题,论文提出了一种改进的细菌觅食算法(Bacterial Foraging Algorithm,BFA),展开旋转机械故障诊断中核参数优化选取研究。主要研究内容如下: 1、分析了标准细菌觅食算法中存在的问题,如种群的大小、运动步长、其迭代次数完全由各种操作所设定的最大次数决定、没有引入收敛准则,难以保证求解的精度并增加不必要的迭代过程,提出了一种改进细菌觅食算法,二维连续函数仿真实验证明了改进后的细菌觅食算法不仅提高了优化速度,而且提高了求解的精度。 2、分析了核参数对KPCA特征降维的影响,设计了基于改进细菌觅食算法的KPCA特征提取算法,旋转机械故障特征实例表明,该方法能够快速、准确的对KPCA核参数进行优化。 3、设计了基于改进细菌觅食算法的SVM参数优化算法,分析和比较改进后的细菌觅食算法与传统的优化算法包括遗传算法、粒子群算法和交叉验证法之间的寻优性能,结果表明改进后的细菌觅食算法优与其他算法。 4、将基于改进细菌觅食算法的KPCA和基于改进细菌觅食算法的SVM应用到旋转机械故障诊断中,实现了基于基座多传感信息融合的滚动轴承故障诊断和基于多传感器信息融合的齿轮故障诊断。实验结果证明了本文所提方法的优越性,同时,实验中基于基座的传感器安装方式可克服现场传感器安装不便等问题,为故障诊断中振动测试提供了一种有用的参考方案,具有极大的应用推广前景。
[Abstract]:Rotating machinery is the key equipment in modern industrial production, to carry out the research of fault diagnosis of rotating machinery, to ensure the equipment safety, efficient operation, to avoid huge economic losses and catastrophic accidents, has great economic and social significance, but also for the machinery and equipment condition monitoring and fault diagnosis technology for rich and development.
This paper takes the rotating machinery as the objects, carry out based on typical kernel methods - kernel principal component analysis (Kernel Principal Component Analysis, KPCA) and support vector machine (Support Vector Machine, SVM), and the technology of fault diagnosis method based on KPCA, the performance of SVM is much influenced by kernel function and its parameters, and the optimal parameters are difficult to select problems, this thesis proposes an improved bacterial foraging algorithm (Bacterial Foraging Algorithm, BFA), launched a selection of optimization of rotating machinery fault diagnosis of nuclear parameters. The main research contents are as follows:
1, analysis of the standard bacterial foraging algorithm in the existing problems, such as population size, movement step length, maximum number of iterations by various operations set by the decision, without introducing the convergence criteria, it is difficult to guarantee the accuracy and increase the unnecessary iterative process, and proposes an improved bacterial foraging algorithm, two-dimensional continuous the function simulation experiment proves that the improved bacterial foraging algorithm not only improves the optimization speed, but also improve the accuracy of solution.
2, the influence of nuclear parameters on KPCA feature reduction is analyzed. A KPCA feature extraction algorithm based on improved bacterial foraging algorithm is designed. The example of rotating machinery fault feature shows that this method can optimize KPCA kernel parameters quickly and accurately.
3, the design of SVM parameter optimization algorithm of the improved bacterial foraging algorithm based on the analysis and comparison of the improved bacterial foraging optimization algorithm and traditional algorithm including genetic algorithm, particle swarm algorithm and cross validation between the optimization performance, the results show that the improved bacterial foraging algorithm and other algorithms.
4, the improved bacterial foraging algorithm KPCA and SVM using the improved bacterial foraging algorithm based on fault diagnosis of rotating machinery based on the gear fault diagnosis of rolling bearing fault diagnosis base based on multi sensor information fusion and based on multi-sensor information fusion. Experimental results prove the superiority of the proposed method in this paper. At the same time. The sensor installation base, based on the experiment can overcome the inconvenience of installation of field sensor, provides a useful reference scheme for fault diagnosis of vibration test, and has a great application prospect.

【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3

【引证文献】