基于CEEMD样本熵的柴油机故障诊断研究

发布时间：2018-10-13 10:26

【摘要】：柴油机作为一种常见的复杂的动力机械,广泛应用于汽车、飞机、船舶等交通工具中,它具有着效率高、比功率大等特点。整体动力系统能否安全、可靠的运行,受到很多因素的影响,柴油机的工作状态就是其中之一。由此可见,通过研究不断完善柴油机监测和故障诊断技术具有重大的实践价值。柴油机工作时其内部零件的状态信息会经过某种渠道反映在缸盖振动当中,所以通过缸盖振动信号对柴油机进行故障诊断是一种有效方法。本课题的研究,主要包含了如何有效地从柴油机缸盖振动信号中提取故障特征信息以及诊断识别柴油机故障状态,提出了结合CEEMD-样本熵的柴油机故障诊断新方法。本文所开展的工作有:(1)设计并构建了能够测取柴油机缸盖振动信号的实验平台。选择CZ4110柴油机为例展开测试,对该型号柴油机在不同工作状态(包括正常的以及各种异常的状态)下的缸盖振动信号数据进行采集,在数据上为后面的柴油机缸盖振动信号特征的提取与故障诊断研究提供支持。(2)研究了柴油机出现故障的原因以及传播渠道。采用理论分析结合实验验证的方式,对柴油机不同故障下的缸盖振动信号特征进行分析,以时域和频域入手,揭示了柴油机缸盖振动信号在这两方面上的特征。(3)对经验模式EMD分解原理在信号分解领域中的应用展开了研究;针对EMD对信号分解的过程中会出现模态混叠的问题,引入具有噪声辅助功能的EEMD以及CEEMD分解方法,通过实验验证了两者能够在一定程度上抑制模态混叠,方法有效,实验还证明CEEMD能够将信号分解到不同时间尺度上从而对信号的局部信息进行提取。提出结合CEEMD和小波的降噪方法,即先用CEEMD对信号进行分解,再用小波对分解出来的每一阶IMF分别进行去噪,最后再重构各降噪后的IMF作为最终降噪的信号,实验验证了其对信号进行了有效的降噪处理。(4)引入专门用于度量信号复杂性和非线性的样本熵,将它们应用于柴油机出现故障时振动序列复杂性的度量当中,分析表明了样本熵有着一致性和受参数影响等特点。提出在挑选IMF分量时,选择依据应为完成分解的IMF分量和原始信号彼此间相关性的大小。对柴机缸盖信号,使用样本熵量化CEEMD分解出的IMF分量从而获得缸盖振动信号在不同频带上的信息,将其作为模式识别的输入向量,为柴油机故障识别提供依据。(5)将CEEMD分解后的各IMF样本熵作为特征向量输入支持向量机训练并对柴油机故障样本进行识别,对比其他一些诊断方法,其提高了准确率。研究了主元分析PCA原理在故障特征降维处理中的应用,通过进一步的对比诊断实验,证明了该方法在有效保留了故障特征信息的同时去除了冗余成分,获得了更精准的诊断信息,用之结合CEEMD-样本熵方法,能够在柴油机故障诊断中取得良好识别效果。
[Abstract]:As a kind of common complex power machinery, diesel engine is widely used in vehicles, aircraft, ships and other vehicles. It has the characteristics of high efficiency and high specific power. Whether the whole power system can operate safely and reliably is affected by many factors, and the working condition of diesel engine is one of them. Therefore, it is of great practical value to improve the monitoring and fault diagnosis technology of diesel engine. The state information of the internal parts of a diesel engine is reflected in the vibration of the cylinder head through a certain channel, so it is an effective method to diagnose the fault of the diesel engine by the vibration signal of the cylinder head. The research of this subject mainly includes how to extract the fault characteristic information from the vibration signal of the cylinder head of diesel engine effectively and to diagnose and identify the fault state of the diesel engine. A new method of diesel engine fault diagnosis based on CEEMD- sample entropy is put forward. The main works of this paper are as follows: (1) an experimental platform for measuring vibration signals of diesel engine cylinder head is designed and constructed. Taking CZ4110 diesel engine as an example, the vibration signal data of cylinder head of the diesel engine under different working conditions (including normal and abnormal states) are collected. On the basis of the data, it can be used to extract the vibration signal characteristics of diesel engine cylinder head and to study the fault diagnosis. (2) the cause of diesel engine failure and the propagation channel are studied. The characteristics of cylinder head vibration signal under different faults of diesel engine are analyzed by means of theoretical analysis and experimental verification, starting with time domain and frequency domain. The characteristics of cylinder head vibration signal in diesel engine are revealed in this paper. (3) the application of empirical mode EMD decomposition principle in signal decomposition field is studied, and the problem of mode aliasing in the process of EMD signal decomposition is discussed. The EEMD and CEEMD decomposition methods with noise auxiliary function are introduced. The experimental results show that the two methods can suppress the mode aliasing to a certain extent and the method is effective. The experiment also proves that CEEMD can decompose the signal to different time scales to extract the local information of the signal. A denoising method combining CEEMD and wavelet is proposed, that is, the signal is decomposed by CEEMD, then each IMF is de-noised by wavelet, and then the IMF is reconstructed as the final de-noising signal. Experimental results show that the proposed method is effective in noise reduction. (4) sample entropy is introduced to measure signal complexity and nonlinearity, and it is applied to measure the complexity of vibration sequence in diesel engine when fault occurs. The analysis shows that the sample entropy is consistent and affected by parameters. In selecting the IMF component, the selection is based on the magnitude of the correlation between the IMF component and the original signal. For the cylinder head signal of diesel engine, the information of cylinder head vibration signal in different frequency bands is obtained by quantizing the IMF component decomposed by CEEMD with sample entropy, which is regarded as the input vector of pattern recognition. (5) the entropy of IMF samples decomposed by CEEMD is used as feature vector to input support vector machine to train diesel engine fault samples, and compared with other diagnosis methods, the accuracy is improved. The application of principal component analysis (PCA) principle in fault feature dimensionality reduction is studied. Through further comparative diagnosis experiments, it is proved that this method not only effectively preserves fault feature information, but also removes redundant components. More accurate diagnosis information is obtained, and the method of CEEMD- sample entropy can be used to identify diesel engine fault.
【学位授予单位】：江苏科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TK428

【相似文献】