基于能量耗损的机械设备故障诊断理论与方法研究

发布时间：2018-05-28 10:47

本文选题：故障诊断 + 能量耗损　；参考：《华南理工大学》2013年博士论文

【摘要】：目前，机械设备故障诊断方法主要有振动分析和油液分析等方法。无论是基于振动分析的设备故障诊断，还是基于油液分析（磨损信息）的设备故障诊断，它们有一个共同点，就是设备故障时都会伴随有系统能量耗损变化。针对机械设备在发生故障时都伴随能量耗损变化这一特征，开创性地提出了一种基于能量耗损的机械设备故障诊断理论与方法。这种方法通过获取摩擦学系统的能量耗损信息，建立能量耗损信息的相关性，提取能量耗损信息特征并进行故障模式识别，建立基于能量耗损的故障规则。首先，论文提出基于能量耗损的机械设备故障诊断新方法。研究摩擦学系统的能量耗损理论与摩擦过程的能量耗损信息流，研究输入能量耗损信息特征、磨损信息特征、振动信号特征。提出了能量耗损信息的相对标度和能量耗损信息的累计相对标度，输入能量耗损采用功率或者油耗等特征量，磨损能量耗损采用光谱元素指标；振动耗能采用振动速度信号时域均方值与振动加速度的峭度等指标，建立能量耗损信息的特征集。研究能量耗损信息的相关性，建立基于能量耗损的机械设备相关性模型。其次，论证了基于能量耗损的机械设备故障诊断方法是可行的。齿轮模拟故障实验研究表明，齿轮发生点蚀、剥落、断齿等不同故障时，输入的功率耗损波动特性不同；磨损能耗信息磨损量和严重磨损指数表明故障的剧烈程度，振动时域信号通过小波包分析提取了各频带的能量分布。齿轮疲劳故障诊断相关性研究表明，输入功率耗损与磨损特征信息与振动特征信息变化规律具有一致性，具有较强的相关性。柴油机活塞缸套疲劳性实验研究表明，瞬时油耗随着活塞磨损故障程度的增加而增加，磨损能耗信息磨损量和严重磨损量指数一直递增。振动能量的变化具有随故障程度增加而增大的趋势，三者能量耗损信息表现规律具有一致性。从而验证本文提出的基于能量耗损的机械设备故障诊断理论和方法是可行的。再次，提出一种基于流形学习算法与支持向量机结合的故障模式识别方法。研究局部线性嵌入LLE、局部切空间排列算法LTSA流形学习算法，并对算法进行了改进。采用流形学习算法对齿轮和柴油机能量耗损数据降维，然后采用多类分类器进行分类，通过分类识别率来判断模式识别的效果。仿真和实验表明流形学习是一种有效的非线性特征提取方法，改进的算法使邻域较好保持了曲面数据的原有对应关系，使得投影后的特征保持了样本间的差异信息和同类样本之间的相似信息。改进的流形学习算法的识别率得到了提高。流形学习与支持向量机结合的模式识别方法是一种有效的特征提取和模式识别方法。然后，建立了能量耗损信息的故障诊断规则。研究了粗糙集与模糊理论的故障规则提取方法，利用粗糙集理论中的不可分辨关系把齿轮能量耗损信息的故障论域划分等价类，生成粗糙集的上近似关系和下近似关系，通过属性重要性分析和属性约简导出故障决策知识和故障分类规则，建立了齿轮能量耗损信息的故障规则；采用模糊理论与神经网络结合的方法，应用自适应模糊控制规则提取方法，输入柴油机能量耗损信息的模糊量，能自动对模糊控制规则进行修改，建立了能量耗损信息的柴油机活塞磨损模糊的故障规则。最后，，研究了能量耗损信息监测与诊断系统的基本结构。设计了基于虚拟仪器技术的能量耗损信息监测与诊断系统结构，使用LabVIEW虚拟化图形化用图标代替文本创建应用程序的计算机编程语言，开发了能量耗损信息在线故障诊断监测系统，包括数据采集系统，信号分析系统，实现了能量耗损信息的采集与分析，初步建立了能量耗损信息的监测与诊断系统。
[Abstract]:At present , the fault diagnosis method of mechanical equipment is mainly characterized by vibration analysis and oil - liquid analysis . Whether it is based on vibration analysis equipment fault diagnosis or oil - liquid analysis ( wear information ) equipment fault diagnosis , they have a common point , which is the system energy consumption change when equipment failure occurs .

First , a new method for fault diagnosis of mechanical equipment based on energy consumption is put forward . The energy dissipation theory and energy loss information flow of the friction process are studied , and the characteristics of the input energy loss information , the characteristics of the wear information and the characteristics of the vibration signal are studied .
In this paper , the characteristic set of energy loss information is established by using the time - domain mean square value of the vibration velocity signal and the frequency of the vibration acceleration , and the correlation of energy consumption information is studied , and the correlation model of the mechanical equipment based on energy consumption is established .

Secondly , it is proved that the fault diagnosis method based on energy consumption is feasible . The experimental study of gear simulation shows that the power consumption fluctuation is different when the gear has different faults such as pitting , spalling , tooth breaking , etc .
The wear and wear index of wear energy consumption information indicates the severity of failure , and the vibration time domain signal is used to extract the energy distribution of each frequency band by wavelet packet analysis . The research on the correlation between input power consumption and wear characteristic information is consistent with the change rule of vibration characteristic information .

This paper presents a fault pattern recognition method based on manifold learning algorithm combined with support vector machine . The local linear embedding LLE , local tangent space arrangement algorithm LTSA manifold learning algorithm is studied , and the algorithm is improved . The method of manifold learning is used to classify the energy consumption data of gears and diesel engines , and the recognition rate of pattern recognition is improved by classification recognition rate . Simulation and experiment show that manifold learning is an effective nonlinear feature extraction method . The improved algorithm makes the recognition rate of the improved manifold learning algorithm improved . The pattern recognition method combined with the manifold learning and support vector machine is an effective feature extraction and pattern recognition method .

Then , a fault diagnosis rule for energy loss information is established . The fault rule extraction method based on rough set and fuzzy theory is studied , and the fault of gear energy consumption information is divided into equivalent classes by the non - discrimination relationship in rough set theory . The fault decision knowledge and fault classification rule are derived by attribute importance analysis and attribute reduction , and fault rules of gear energy consumption information are established .
Based on fuzzy theory and neural network , an adaptive fuzzy control rule extracting method is applied to input the fuzzy quantity of energy consumption information of diesel engine , and the fuzzy control rule can be modified automatically , and the fuzzy fault rule of piston wear of diesel engine with energy loss information is established .

In the end , the basic structure of the energy consumption information monitoring and diagnosis system is studied . The system structure of energy consumption information monitoring and diagnosis based on virtual instrument technology is designed . The computer programming language of energy consumption loss information online fault diagnosis is developed by using LabVIEW virtualization graphical icon instead of the computer programming language of the text creating application program . The system includes data acquisition system and signal analysis system , and the acquisition and analysis of energy consumption information is realized , and the monitoring and diagnosis system of energy consumption information is established .
【学位授予单位】：华南理工大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TH165.3

【参考文献】