基于ISOMAP的机械故障诊断方法研究与应用

发布时间：2018-03-08 01:22

本文选题：ISOMAP　切入点：数据降维　出处：《华南理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：智能机械故障诊断方法研究一直是机械诊断领域研究的热点问题。随着人工智能、计算机软件技术、现代传感器技术以及现代信号处理技术的飞速发展，大型机械设备的故障诊断信号数据集呈现出维数高、随机性强、数据量大的典型特点。在保证数据间的几何关系和距离测度不变的前提下，将原始数据对应的高维空间的流形映射至低维空间，可以减少相关计算量，找出关键特征，全面提高故障诊断效率。针对机械故障信号高维数、时变性、非线性和非高斯分布特征，本文利用流形学习理论中改进之后的ISOMAP算法对机械故障信号数据集进行非线性降维处理，使得故障数据更加易于分类。论文研究的重点是深入分析经典ISOMAP算法的原理和计算过程，明确经典算法应用到机械故障领域的局限性，提出了一种适用于机械故障诊断有监督的快速ISOMAP算法，利用改进的算法对故障数据进行非线性降维，将降维之后的数据分为训练数据集和测试数据集，用训练数据集对支持向量机进行训练，然后利用训练之后的支持向量机对测试数据集进行预测，实现故障诊断和分类。论文的主要内容包括： (1).探讨ISOMAP算法应用在机械故障诊断领域中存在的问题，包括噪声问题，，参数的优化选择问题以及算法的泛化能力。 (2).针对ISOMAP算法应用到机械故障诊断领域存在的问题，对经典ISOMAP进行改进，提出有监督的快速ISOMAP算法，采用美国西储大学的电机轴承故障数据，对提出的新算法进行验证。 (3).利用汽车传动试验台对汽车变速箱进行无故障、齿轮点蚀和齿轮剥落模拟试验，在时域分析、频域分析和小波分解无法准确迅速进行故障诊断的情况下，将提出的改进ISOMAP算法应用到齿轮故障诊断中，并与传统方法进行对比，证明了该方法在齿轮故障诊断中有效性和优越性。改进之后的ISOMAP算法能够有效约简故障数据维数、找出本征维数，这将会大大缩短计算时间，利于数据分类，提高故障诊断效率和正确率。
[Abstract]:The research of intelligent mechanical fault diagnosis method has been a hot issue in the field of mechanical diagnosis. With the rapid development of artificial intelligence, computer software technology, modern sensor technology and modern signal processing technology, The fault diagnosis signal data set of large mechanical equipment presents the typical characteristics of high dimension, strong randomness and large amount of data. Under the premise of ensuring the geometric relationship between data and the invariance of distance measure, Mapping the manifold of the high-dimensional space corresponding to the original data to the low-dimensional space can reduce the relevant computation amount, find out the key features, and improve the efficiency of fault diagnosis in an all-round way. Aiming at the characteristics of high dimension, time variation, nonlinearity and non-#china_person0# distribution of mechanical fault signal, the improved ISOMAP algorithm in manifold learning theory is used to deal with the nonlinear dimensionality reduction of mechanical fault signal data set in this paper. The emphasis of this paper is to analyze the principle and calculation process of the classical ISOMAP algorithm, and to clarify the limitation of the classical algorithm in the field of mechanical fault. A fast ISOMAP algorithm for mechanical fault diagnosis is proposed. The improved algorithm is used to reduce the nonlinear dimension of the fault data. The reduced dimension data is divided into the training data set and the test data set. The support vector machine is trained with the training data set, and then the test data set is predicted by the training support vector machine to realize fault diagnosis and classification. The main contents of the thesis include:. This paper discusses the problems existing in the application of ISOMAP algorithm in the field of mechanical fault diagnosis, including the problem of noise, the optimization of parameters and the generalization ability of the algorithm. In view of the problems existing in the application of ISOMAP algorithm in the field of mechanical fault diagnosis, the classical ISOMAP is improved, and a supervised fast ISOMAP algorithm is proposed. The new algorithm is verified by using the fault data of motor bearings from the University of Western Reserve in the United States. Using the automobile transmission test bench to carry out the fault-free, pitting corrosion and spalling simulation test of the automobile gearbox, when the time domain analysis, the frequency domain analysis and the wavelet decomposition can not accurately and quickly carry on the fault diagnosis, The improved ISOMAP algorithm is applied to gear fault diagnosis, and compared with the traditional method, it is proved that this method is effective and superior in gear fault diagnosis. The improved ISOMAP algorithm can effectively reduce the dimension of fault data and find the intrinsic dimension, which will greatly shorten the calculation time, facilitate data classification, and improve the efficiency and accuracy of fault diagnosis.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【相似文献】