不完备信息下基于流向图的齿轮故障诊断方法研究

发布时间：2018-01-07 18:34

  本文关键词：不完备信息下基于流向图的齿轮故障诊断方法研究　出处：《哈尔滨工业大学》2017年博士论文　论文类型：学位论文

  更多相关文章： 故障诊断 齿轮 不完备信息 流向图 知识获取 模式识别

【摘要】：齿轮作为旋转机械的重要组成部分,已被广泛应用于电站、直升机和重型卡车等设备中。由于齿轮的运行环境恶劣、工况复杂,使得齿轮本身状态信息表露不完备。而且人类实践活动总是受到客观环境和条件的限制,所获得的描述齿轮故障模式的诊断信息常有某种程度的不完备。本文首先将量化特征关系用于解决齿轮不完备诊断信息处理的问题。然后,以流向图理论为基础,提出了不完备信息下齿轮故障诊断知识表示方法、不完备信息下齿轮故障诊断知识获取方法,以及齿轮齿故障模式识别方法。最后,以齿轮箱实验组件为研究对象进行验证。本文主要包括以下几个方面的内容:现有的广义不可分辨关系大多仅能采用一种语意来理解、分析和处理不完备信息;无法理解、分析和处理由多种原因造成的齿轮不完备故障诊断信息。为此,本文在特征关系的基础上,给出了一种实例间相似度的计算方法。对特征关系进行改进,给出了一种量化特征关系。针对齿轮不完备故障诊断信息,提出了基于量化特征关系的齿轮不完备故障诊断信息处理方法。采用自动变速箱故障诊断信息的处理实例验证该方法的实用性、有效性和准确性。现有的不完备信息下齿轮故障诊断知识表示方法无法直观地表示各故障属性值,以及故障征兆属性值和故障决策属性值之间的依赖关系。难于定量描述属性间的依赖程度。为此,本文在流向图的基础上,给出了不完备流向图的定义。根据不完备流向图的定义,给出了一种不完备流向图的构建算法。针对齿轮不完备故障诊断信息的知识表示问题,提出了一种基于流向图的齿轮不完备故障诊断信息的知识表示方法。结合知识表示实例验证该方法可直观地表示包含三种未知属性值不完备故障诊断信息,定量描述属性值之间的依赖程度,便于用户理解和分析。现有不完备信息下齿轮故障诊断知识获取方法大多仅能从包含一种未知属性值的不完备信息中获取故障诊断知识。个别方法能从同时包含两种未知属性值的不完备信息中提取故障诊断知识,但知识获取过程抽象难于理解。为此,本文在量化特征关系的基础上,给出了一种基于量化特征关系的分配约简算法。借鉴基于量化特征关系的分配约简算法,给出了一种不完备流向图的属性约简算法。针对齿轮不完备故障诊断信息的知识获取问题,提出了一种基于流向图的齿轮不完备故障诊断信息的知识获取方法。通过实例验证了此方法能够从同时包含三种未知属性值的不完备信息系统中直接获取故障诊断知识;且知识获取结果直观、简洁与清晰。现有的齿轮故障模式识别方法无法直观地表示出故障征兆属性值,以及属性值之间的依赖关系。有些方法的识别模型过于复杂,识别过程依然较晦涩难于理解。为此,本文给出了流向图中节点条件概率的定义、先验概率的定义,以及完整路径后验概率的定义,实现流向图的概率化。在流向图概率化的基础上,给出了一种基于流向图的模式识别算法,进而提出了一种基于流向图的齿轮故障模式识别方法。实验结果表明该方法可准确地识别齿轮故障模式,而且模式识别模型的结构简单,模式识别策略清晰。以Spectra Quest公司开发的齿轮实验组件为研究对象,利用振动分析法和油样分析法提取的齿轮不完备故障诊断信息,验证了本文提出的不完备信息下基于流向图的齿轮故障诊断方法。建立齿轮不完备故障诊断信息系统,采用量化特征关系对该信息系统进行处理。构建齿轮不完备故障诊断流向图,并从中获取故障诊断知识。通过基于流向图的模式识别算法判断待诊样本的故障类型。实验结果表明此方法具有非常高的准确率,而且故障诊断过程直观,故障诊断策略清晰。
[Abstract]:Gear is an important part of the rotating machinery has been widely used in power plants, helicopters and heavy trucks and other equipment. Because of the operation environment of gear bad conditions are complex, so that the gear state information disclosure is not complete. But human activities have always been the objective environment and conditions, the fault diagnosis of gear model description information obtained is incomplete to some degree. This paper will be used to solve the relationship between the quantitative characteristics of gear incomplete diagnosis information processing. Then, the flow graph theory, the incomplete information of gear fault diagnosis knowledge representation method, incomplete information of gear fault diagnosis knowledge acquisition method, and the gear fault the pattern recognition method. Finally, the gear box test module as the object of study is verified. This paper mainly includes the following aspects: the existing The generalized indiscernibility relation most can only adopt a semantic analysis to understand and deal with incomplete information; to understand, analyze and deal with by a variety of causes of gear incomplete fault diagnosis information. Therefore, based on the characteristics of the relationship, a method of calculating the similarity between the examples are given. The characteristics of the relationship improved, gives a quantitative relationship. According to the characteristics of gear fault diagnosis of incomplete information, put forward the relationship between the quantitative characteristics of incomplete gear fault diagnosis method based on information processing. Based on an example of fault diagnosis information processing automatic transmission to verify the practicability of the method. The accuracy and effectiveness of the existing gear under incomplete information fault diagnosis knowledge representation methods cannot directly represent the fault attribute value and attribute value relation between fault symptoms and fault decision attribute values to quantitative. Describe the degree of dependence between attributes. Therefore, based on the flow pattern, gives the definition of incomplete flow graph. According to the definition of incomplete flow graph, gives the algorithm to construct a complete flow chart. According to incomplete information of the gear fault diagnosis knowledge representation problem, this paper proposes a new method flow chart of the incomplete gear fault diagnosis information based on knowledge. Combined with the knowledge representation example shows that this method can intuitively represent contains three unknown attribute values of incomplete fault diagnosis information, quantitative description of dependence between attribute values, allowing users to understand and analyze the existing incomplete information of gear fault diagnosis knowledge acquisition methods are only from contains a fault diagnosis knowledge acquisition in incomplete information of unknown attribute values. The individual method can also contain two kinds of extraction from incomplete information of unknown attribute values Fault diagnosis knowledge, but the knowledge acquisition process is abstract and difficult to understand. Therefore, based on the quantitative characteristics of relations, presents a distribution reduction algorithm based on the relationship between the quantitative characteristics. From distribution reduction algorithm based on the relationship between the quantitative characteristics, gives an incomplete flow diagram. Attribute reduction algorithm to extract gear incomplete fault diagnosis information knowledge, proposes a method of obtaining the flow graph of gear incomplete fault diagnosis information based on knowledge. Through the example proves the method can directly obtain the fault diagnosis knowledge from also contains three unknown attribute values in incomplete information system; knowledge acquisition and intuitive results, concise and clear. Gear fault pattern recognition of the existing methods can not clearly shown that the fault attribute value and attribute value dependencies between. Some methods of identifying model In the complex, the recognition process is still obscure and difficult to understand. Therefore, this paper gives a definition of the Condition node flow graph probability, defined a priori probability, and the definition of the full path of the posterior probability of the flow graph. Based on the probability of the flow chart, a pattern recognition algorithm based on flow graphs given, then put forward a kind of gear fault pattern recognition method based on flow graphs. The experimental results show that this method can accurately identify the gear fault pattern and structural pattern recognition model, pattern recognition strategy clear. In gear test component development Spectra of Quest company as the research object, analysis extraction using vibration analysis method the gear oil sample and incomplete fault diagnosis information, gear fault diagnosis method based on flow graphs are verified under incomplete information is proposed in this paper. The establishment of incomplete gear fault diagnosis Fault information system, the relationship between the quantitative characteristics of the information system construction of gear processing. Incomplete fault diagnosis flow chart, and obtain the fault diagnosis knowledge from it. By judging the fault type diagnosis for sample pattern recognition flow graph based algorithm. Experimental results show that this method has very high accuracy rate, and the process of fault diagnosis intuitively, the fault diagnosis strategy is clear.

【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TH132.41
，

本文编号：1393759