【作者】 冯光烁；

【导师】 杨海青；

【作者基本信息】 南京航空航天大学 ， 车辆工程， 2009， 硕士

【摘要】 发动机失火故障是一种常见的发动机故障,对失火故障的监测和诊断是发动机故障诊断系统的必要组成部分。本文主要研究基于小波分析理论、粗糙集理论的发动机失火故障智能诊断方法。论文主要内容由六部分组成:(1)、发动机故障的振动诊断机理分析。介绍发动机振动的激振力和发动机振动的类型,建立每种振动类型的动力学模型,总结发动机缸盖振动的信息模型;(2)、发动机缸盖振动信号的滤波算法研究。首先介绍数学形态滤波器的理论基础,然后建立仿真信号,对经典形态滤波器进行仿真研究,在此基础上,提出基于数学形态滤波器的组合滤波方案,最后利用该滤波方案对发动机缸盖振动信号进行处理;(3)、发动机缸盖振动信号的特征提取算法研究。首先简要介绍小波分析理论,然后研究三种特征提取算法:小波域近似熵特征提取算法、基于图像处理的小波时频分析特征提取算法、小波包能量谱特征提取算法,最后将三种算法应用于发动机缸盖振动信号特征提取,得到可以用于故障模式识别的特征向量;(4)、基于粗糙集理论的故障特征约简算法研究。首先简要介绍粗糙集理论,然后给出粗糙集理论中的连续属性离散化方法和属性约简算法,最后提出发动机缸盖振动信号故障特征的约简算法,并计算得到根据小波包能量谱特征提取算法提取出的16个故障特征量的一个属性约简P = { E0 , E1 , E3};(5)、基于支持向量机的故障模式识别算法研究。首先简要介绍支持向量机的基本概念,然后提出基于支持向量机的发动机失火故障智能诊断算法,并通过实验数据验证该算法的正确性和有效性;(6)、介绍发动机状态监测与故障诊断系统的软硬件系统设计过程。在简要介绍虚拟仪器技术和NI LabVIEW语言的基础上,详细阐述大众2VQS电喷发动机失火故障智能诊断系统的软硬件系统设计过程,以及发动机实验台架状态监测与故障诊断系统的软硬件系统设计过程。更多还原

【Abstract】 Engine misfire event is a main internal combustion engine fault. Monitoring and diagnosing engine misfires are indispensable functions of engine fault detection systems. The main research of this paper is to develop intelligent engine misfire detection methods based on wavelet analysis and rough set theory.The main contents of this thesis consist of six parts as follows. (1) The mechanism of engine fault vibration diagnosis methods is analyzed. The excitation forces of the engine vibration are presented. The classification of the engine vibration is introduced and the dynamic models of every kinds of engine vibration are constructed. Finally, the information model of engine cylinder head vibration is presented. (2) The filtering algorithm of engine cylinder head vibration signal is studied. Firstly, the theoretical basis of mathematical morphological filters is introduced. Then, the classical morphological filters are studied using a simulation signal and an integrated filtering algorithm based on morphological filters is presented. Finally, the cylinder head vibration signals are processed based on this filtering solution. (3) Feature extraction algorithms of cylinder head vibration signal are investigated. Firstly, wavelet analysis theory is introduced. Then, three kinds of feature extraction algorithms are studied. The three algorithms are approximate entropy feature extraction algorithm in wavelet domain, wavelet time-frequency analysis feature extraction algorithm based on image processing and wavelet packet power spectrum feature extraction algorithm. Finally, apply the three algorithms to extract features of cylinder head vibration signal and feature vectors that can be used for fault pattern recognition are gained. (4) The fault vector reduction algorithm based on rough set theory is studied. Firstly, rough set theory is introduced. The discrete method of continuous attribute values and attributes reduction algorithms are investigated. Finally, attributes reduction algorithm of cylinder head vibration signal fault vectors is investigated and reductive attributes P = { E0 , E1 , E3} are gained from 16 feature values that are calculated according to the wavelet packet power spectrum feature extraction algorithm. (5) The fault pattern recognition algorithm based on support vector machines is studied. Firstly, the basic concepts of support vector machines are introduced. Then, intelligent engine misfire detection algorithm based on support vector machines is presented, and the experiment results show that this algorithm is correct and valid. (6) The design of hardware and software of the engine fault monitoring and diagnosing system is detailed. Firstly, the virtual instrument technology and NI LabVIEW programming language are introduced. Then, the design of hardware and software of the intelligent misfire detection system of VOLKSWAGEN 2VQS electronic controlled engines is presented. Finally, the design of hardware and software of monitoring and diagnosing system of engine test bench is presented.更多还原