基于自适应增量LLE和SVM的滚动轴承健康状态评估方法研究

发布时间：2018-05-13 19:35

本文选题：滚动轴承 + 高维特征集　；参考：《哈尔滨理工大学》2017年硕士论文

【摘要】：滚动轴承作为旋转机械设备中的重要基础元件之一,整个机械设备运行的安全性和可靠性都与其健康状态息息相关,一旦出现故障不能及时监测并排除,就有可能造成整个机械设备的失效瘫痪,甚至造成巨大的灾难性后果。因此,如何准确评估滚动轴承的健康状态,以真正达到智能定量评估,变定时或事后维修为视情维修,实现滚动轴承的主动维护具有十分重要的意义。论文以建立滚动轴承健康状态评估方法为主线,从特征提取、维数约简、健康状态评估模型以及优化算法方面逐步展开研究。特征提取方面,为有效利用时域、频域、时-频域中各类具有显著差异的特征,更全面准确的刻画轴承的健康状态。论文对实际轴承振动信号进行时域、频域的特征提取,并将集合经验模态分解(EEMD)结合奇异值分解(SVD)获取时-频特征。在此基础上构建高维多域特征集。维数约简方面,深入研究局部线性嵌入(LLE)算法。基于振动信号提取的特征具有数据流形的局部几何结构分布不均匀的特点,同时为避免出现新增样本时需要重复计算的缺点,本文对局部线性嵌入算法进行自适应增量改进。通过对比实验证明改进后的局部线性嵌入算法的有效性。优化算法方面,由于人为选取支持向量机(SVM)中的参数具有一定的盲目性,传统的参数优化算法难以快速找出全局最优解。因此本文深入研究了鸡群优化(CSO)算法,并将混沌理论引入到鸡群算法中,即混沌优化鸡群算法(C-CSO)。通过对基本函数进行优化实验对比,详细分析各优化算法的性能。健康状态评估方面,用混沌优化鸡群算法优化支持向量机中的参数,构建自适应增量局部线性嵌入-支持向量机分类模型。并提出一致性相对补偿距离作为健康状态评估指标,实现对滚动轴承健康状态的有效评估。
[Abstract]:As one of the important basic components in rotating machinery, the safety and reliability of the whole machinery and equipment are closely related to its health condition. Once the fault occurs, it can not be monitored and eliminated in time. May cause the failure of the entire mechanical equipment paralysis, and even cause huge catastrophic consequences. Therefore, it is of great significance to accurately evaluate the health status of rolling bearings, to truly achieve intelligent quantitative evaluation, to change the timing or after maintenance to maintenance according to the situation, and to realize the active maintenance of rolling bearings. The main line of this paper is to establish a method for evaluating the health status of rolling bearings, and the research is carried out step by step from the aspects of feature extraction, dimension reduction, health state evaluation model and optimization algorithm. In the aspect of feature extraction, in order to utilize the different features in time domain, frequency domain and time-frequency domain effectively, the healthy state of bearing can be described more comprehensively and accurately. In this paper, the real bearing vibration signals are extracted in the time domain and frequency domain, and the time-frequency features are obtained by using the set empirical mode decomposition (EEMD) and singular value decomposition (SVD). On this basis, the high dimensional multi-domain feature set is constructed. In the aspect of dimension reduction, the local linear embedding (Lle) algorithm is studied in depth. The feature extracted based on vibration signal has the characteristic of uneven distribution of local geometric structure in data stream shape, and in order to avoid the shortcoming of repeated calculation when new samples are added, the adaptive incremental improvement of local linear embedding algorithm is carried out in this paper. The experimental results show that the improved local linear embedding algorithm is effective. In the aspect of optimization algorithm, it is difficult for traditional parameter optimization algorithm to find the global optimal solution quickly because of the blindness in the artificial selection of the parameters in support vector machine (SVM). Therefore, this paper deeply studies the CSOs algorithm, and introduces the chaos theory into the chicken swarm optimization algorithm, that is, the chaos optimal chicken swarm algorithm (CSOO). The performance of each optimization algorithm is analyzed in detail by comparing the basic function optimization experiments. In the aspect of health state evaluation, chaotic optimization algorithm is used to optimize the parameters of support vector machine (SVM), and an adaptive incremental local linear embedded-support vector machine (SVM) classification model is constructed. The consistent relative compensation distance is proposed as an index to evaluate the health status of rolling bearings.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【相似文献】