基于振动分析的柴油机燃油系统与配气机构故障诊断研究

发布时间：2019-02-08 19:34

【摘要】：柴油机是一种最常见的动力装置,在国民经济和人们的日常生活中起着举足轻重的作用,但复杂的结构和恶劣的工作条件导致柴油机故障时有发生,给人们的生命和财产造成了很大的威胁。据统计,在导致柴油机故障的各种原因中燃油系统故障与配气机构故障所占的比例较高,分别为27%和15.1%。此外,燃油系统与配气机构直接影响着发动机的燃烧情况,它们的工作状况对整个发动机能否正常高效运转具有重要影响。因此,本文从柴油机故障诊断的实际需求出发,选择柴油机中故障概率较高的燃油系统与配气机构作为研究对象,以分析缸盖振动为切入点,对柴油机燃油系统和配气机构的故障诊断方法开展深入研究,具体包括以下几个方面:分析了柴油机缸盖的主要激励源,建立了柴油机缸盖振动信息模型,并对柴油机缸盖振动的时域和频域特性进行了研究。在两台柴油机上进行了故障实验,分别模拟了5种典型配气机构故障和12种常见的燃油系统与配气机构故障,为后续的研究工作提供了数据支撑。研究了固有时间尺度分解(Intrinsic Time-scale Decomposition,ITD)方法的基本原理,提出了一种基于ITD边际谱和马氏距离的故障诊断方法,并利用该方法对配气机构故障进行了诊断,诊断结果表明该方法可以对配气机构故障进行粗略识别。针对有向无环图的误差累积问题,提出了一种基于类别可分性的改进有向无环图(Separability based Directed Acyclic Graph,SDAG),并利用该方法构建了相关向量机(Relevance Vector Machine,RVM)多分类模型。此外,提出了一种RVM多分类通用框架概率性纠错输出编码(Probability-based Error Correcting Output Codes,PECOC),该框架实现了RVM的概率性输出与纠错输出编码多分类方法的有机结合。针对ITD方法的插值方法、终止条件等方面的不足,提出了一种改进固有时间尺度分解(Improved Intrinsic Time-scale Decomposition,IITD)方法,并将ITD方法和SDAG-RVM相结合对柴油机配气机构故障进行诊断。实验分析结果表明该方法可以获得较高的故障诊断精度,优于其它常规方法。研究了IITD方法的模态混叠问题,提出了一种完备总体改进固有时间尺度分解(Complete Ensemble Imporved Intrinsic Time-scale Decomposition,CEIITD)方法,该方法通过向原信号添加正负相反的噪声并且每产生一个旋转分量进行一次集总平均,消除了分解结果中的噪声残留以及由每次分解获得的旋转分量个数不同造成的难以进行集总平均的问题。仿真分析结果表明CEIITD方法较好地解决了IITD方法的模态混叠问题,优于其它常规方法。针对Wigner分布的交叉项问题,提出了一种基于IITD方法的Wigner分布交叉项处理方法即自适应Wigner分布,并利用仿真信号对该方法进行了验证。仿真分析结果表明该方法不仅可以消除交叉项的干扰,而且可以确保时频图具有较高的分辨率和时频聚集性。针对原快速相关滤波(Fast Correlation-based Filter,FCBF)算法未考虑候选特征与已选特征子集间冗余的问题,提出了一种改进FCBF算法。结合自适应Wigner分布、改进FCBF和PECOC-RVM提出了一种新的柴油机燃油系统与配气机构故障诊断方法,该方法将柴油机故障诊断问题转化为时频图像分类问题,从而可以将在图像分类领域得到大量成功应用的不变矩、灰度统计特征、纹理特征和差分分形盒维数等图像特征作为故障特征,降低了特征提取的难度和工作量。实验分析结果表明基于自适应Wigner分布、改进FCBF和PECOC-RVM的柴油机燃油系统与配气机构故障诊断方法可以获得较高的故障诊断精度,优于其它常规方法。针对同时诊断许多种故障时单一故障识别模型精度明显下降的问题,提出了一种基于多级Adaboost-RVM的柴油机燃油系统和配气机构故障诊断方法。该方法利用KFCM算法将包含较多种故障类型的复杂分类问题分解为若干个简单的问题,通过对每个简单分类问题输出结果的整合便可以获得复杂分类问题的输出结果,同时采用Adaboost技术对弱分类器进行提升,进一步增强了分类算法的诊断能力。实验分析结果表明该方法对于故障类型较多的诊断问题具有明显的优势,优于其它常规方法。
[Abstract]:The diesel engine is the most common power device, plays a very important role in the national economy and people's daily life, but the complicated structure and the bad working conditions cause the failure of the diesel engine to happen frequently, which poses a great threat to people's life and property. According to the statistics, the proportion of the failure of the fuel system and the gas distribution mechanism in the various causes leading to the failure of the diesel engine is high, 27% and 15.1%, respectively. In addition, the fuel system and the gas distribution mechanism directly affect the combustion of the engine, and their working conditions have an important influence on the normal and efficient operation of the whole engine. Therefore, based on the actual demand of the diesel engine fault diagnosis, the fuel system and the gas distribution mechanism with higher fault probability in the diesel engine are selected as the research object to analyze the cylinder head vibration as the starting point, and the fault diagnosis method of the diesel engine fuel system and the gas distribution mechanism is further researched. The main excitation source of the cylinder head of the diesel engine is analyzed, the vibration information model of the cylinder head of the diesel engine is established, and the time domain and frequency domain characteristics of the vibration of the cylinder head of the diesel engine are studied. The failure of five typical gas distribution mechanism and 12 common fuel system and gas distribution mechanism were simulated on two diesel engines, and data support was provided for the follow-up research work. In this paper, the basic principle of Intrisic Time-scale Decomposition (ITD) method is studied, a fault diagnosis method based on the ITD marginal spectrum and the horse's distance is proposed, and the fault of the gas distribution mechanism is diagnosed by the method. The results show that the method can be used for rough identification of the fault of the gas distribution mechanism. In order to solve the problem of error accumulation for acyclic graph, a kind of improved non-ring graph (SDAG) based on the division of the class is put forward, and the multi-classification model of the Reliance Vector Machine (RVM) is constructed by this method. In addition, an RVM multi-classification general-purpose frame probability error correction output code (PECOC) is proposed, which realizes the organic combination of the probability output of RVM and the multi-classification method of error correction output coding. In view of the deficiencies of the interpolation method and the termination condition of the ITD method, an improved Intro-time-scale Decomposition (IITD) method is proposed, and the ITD method and the SDAG-RVM are combined to diagnose the fault of the gas distribution mechanism of the diesel engine. The experimental results show that the method can obtain higher fault diagnosis precision than other conventional methods. In this paper, we study the mode mixing problem of the IITD method, and put forward a complete and improved complete-improved Intro-scale Decomposition (CEIITD) method. The method adds positive and negative noise to the original signal and makes a total average of one rotation component at a time. the problem that the noise residue in the decomposition result and the number of rotation components obtained by each decomposition are different due to the difference in the number of rotation components obtained by each decomposition are eliminated. The results of the simulation show that the CEIITD method better solves the problem of the IITD method and is superior to the other conventional methods. In order to solve the cross-item problem of the Wigner distribution, a Wigner distribution cross-item processing method based on the IITD method is proposed, that is, the adaptive Wigner distribution, and the method is verified by using the simulation signal. The simulation results show that the method can not only eliminate the cross-term interference, but also ensure that the time-frequency graph has higher resolution and time-frequency aggregation. In order to solve the problem of redundancy between the candidate feature and the selected feature subset, an improved FCBF algorithm is proposed for the Fast Correlation-based Filter (FCBF) algorithm. Combined with the adaptive Wigner distribution, the improved FCBF and PECOC-RVM provide a new method for fault diagnosis of the fuel system and the gas distribution mechanism of the diesel engine. The method converts the fault diagnosis of the diesel engine into the time-frequency image classification problem, so that the invariant moment of a large number of successful applications in the field of image classification can be obtained. Image features such as gray-scale statistical features, texture features, and differential fractal box dimension numbers are used as fault features, and the difficulty and workload of feature extraction are reduced. The experimental results show that the improved FCBF and PECOC-RVM diesel engine fuel system and the gas distribution mechanism fault diagnosis method can obtain higher fault diagnosis accuracy than other conventional methods based on the adaptive Wigner distribution. In order to solve the problem of the accuracy of single fault identification model at the same time, a new fault diagnosis method for diesel engine fuel system and gas distribution mechanism based on multi-level Adaboost-RVM is proposed. the method uses the KFCM algorithm to decompose the complex classification problem containing a plurality of fault types into a plurality of simple problems, the diagnostic capability of the classification algorithm is further enhanced. The results of the experiment show that the method has obvious advantages over the diagnosis of more fault types, and is superior to other conventional methods.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TK428

【相似文献】