基于混沌神经网络的故障诊断方法研究

发布时间：2018-05-03 00:26

本文选题：故障诊断 + 轴向柱塞泵　；参考：《燕山大学》2012年硕士论文

【摘要】：随着液压技术不断地发展，液压系统被广泛地应用于许多重要的领域。在液压系统的功能不断增强的同时，其结构变得越来越复杂，这也增大了液压系统发生故障的可能性。液压泵作为整个液压系统的动力源，它所处的工作环境恶劣，并且结构复杂，导致其很容易发生故障。它的工作状况将成为影响整个液压系统乃至整个设备的正常工作的关键，因而对液压泵进行状态监测和故障诊断尤为重要。近年来，液压泵的故障诊断技术成为研究的热点，正向着智能化、自动化的方向发展。本文采用了一种将混沌理论和神经网络相结合的方法完成液压泵的故障诊断过程。混沌理论是当今非线性科学研究非常活跃的一个方面，将混沌理论和神经网络相结合构成性能更为优越的混沌神经网络成为研究的热点之一。本文在前向神经网络的基础之上，，建立了一种基于Logistic映射的前向混沌神经网络，并研究了该网络的学习算法。通过引入混沌机制，使得该种混沌神经网络能够有效地避免神经网络在训练过程中易陷入局部极小值的缺点，并对微小区别的模式具有更好的识别效果，该网络具有良好的寻优能力、泛化能力以及模式识别能力。为了验证该方法的有效性，本文以实验室材料试验机的斜盘式轴向柱塞泵为诊断对象，对液压泵进行状态监测，采集了泵在不同工作状态下在其端盖处的振动信号。以垂直于端盖的振动信号为研究信息，采用短时最大熵谱分析的方法得出了各故障状态的共振频带范围，为小波包带通滤波提供依据。利用小波包理论和希尔伯特变换的包络解调的方法完成信号的处理，并进行了功率谱分析。提取包络信号的幅值域特征指标作为特征向量，以多组特征向量作为混沌神经网络的训练和测试输入。应用MATLAB软件进行编程，证明了前向混沌神经网络应用在液压泵故障诊断中是切实可行的，并且与目前应用广泛的BP神经网络的诊断结果相比较，得出前向混沌神经网络比L-M优化的BP神经网络的收敛速度更快、诊断正确率更高，体现了混沌神经网络应用于液压泵故障诊断方面的优越性。
[Abstract]:With the development of hydraulic technology, hydraulic system is widely used in many important fields. At the same time, the structure of hydraulic system becomes more and more complex, which increases the possibility of hydraulic system failure. As the power source of the whole hydraulic system, the hydraulic pump is in a bad working environment and complex structure, which leads to its failure easily. Its working condition will be the key to affect the whole hydraulic system and even the whole equipment, so the condition monitoring and fault diagnosis of hydraulic pump is very important. In recent years, the fault diagnosis technology of hydraulic pump has become a hot spot, and is developing towards the direction of intelligence and automation. In this paper, a method combining chaos theory and neural network is used to complete the fault diagnosis of hydraulic pump. Chaotic theory is one of the most active aspects of nonlinear science nowadays. The combination of chaos theory and neural network to form a chaotic neural network with better performance has become one of the research hotspots. In this paper, based on the forward neural network, a forward chaotic neural network based on Logistic mapping is established, and the learning algorithm of the network is studied. By introducing chaos mechanism, this kind of chaotic neural network can effectively avoid the shortcoming that the neural network is easy to fall into the local minimum value in the training process, and has better recognition effect to the pattern of small difference. The network has good optimization ability, generalization ability and pattern recognition ability. In order to verify the effectiveness of this method, this paper takes the oblique disc axial piston pump of the laboratory material testing machine as the diagnostic object, carries on the condition monitoring to the hydraulic pump, and collects the vibration signal of the pump at the end cover of the pump in different working state. Taking the vibration signal perpendicular to the end cover as the research information, the resonance frequency band range of each fault state is obtained by using the method of short-time maximum entropy spectrum analysis, which provides the basis for the wavelet packet bandpass filtering. The wavelet packet theory and the envelope demodulation method of Hilbert transform are used to complete the signal processing, and the power spectrum analysis is carried out. The amplitude range feature index of the envelope signal is extracted as the feature vector, and the multi-group eigenvector is used as the training and test input of the chaotic neural network. The application of forward chaotic neural network in hydraulic pump fault diagnosis is proved to be feasible by using MATLAB software, and compared with the result of BP neural network, which is widely used at present. It is concluded that the forward chaotic neural network has faster convergence speed and higher diagnostic accuracy than the L-M optimized BP neural network, which shows the superiority of chaotic neural network in hydraulic pump fault diagnosis.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;TP183

【参考文献】