基于粗糙集的分层有向图故障诊断方法研究及其应用

发布时间：2018-03-22 21:38

本文选题：热力系统　切入点：符号有向图　出处：《太原理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：在火电机组故障中,热力系统发生故障的机率很大,因此热力系统的安全性会对机组整体运行的安全经济性产生巨大影响。由于热力系统结构复杂、运行工况多变,在进行故障诊断过程中难以建立和完善故障知识库,因此对热力系统进行故障诊断研究十分必要。本文将图论和粗糙图理论与符号有向图(SDG)相结合,使得SDG模型更加易于诊断,从而提高SDG故障诊断的准确性。首先针对SDG多义性推理导致的分辨率低,大系统诊断速度慢、实时性差等问题,将故障传播图与SDG两种方法结合起来描述系统,提出了一种分层符号有向图(SDG)故障诊断模型。该分层符号有向图故障诊断模型既具有SDG自身的完备性,又具有故障传播图的固有属性,可以反映系统部件之间的连接关系,同时将分层技术、节点压缩技术、约束传播技术和对中寻优技术四种优选测试点技术加入改进的符号有向图模型来构建分层SDG模型,并采用回溯搜索策略搜索独立相容通路进行故障诊断。以上几种优选测试点技术利用了系统设备内部各个部件之间故障传播的特性,在推理过程中可以减少候选测试点集合,提高搜索故障源的效率。该方法能够区分相同故障模式下诊断不明且难分辨的故障及可能出现的“组合爆炸”问题,所得结论与传统SDG诊断结果进行对比,表明该方法较传统方法优越。最后将该方法应用到离心泵与液位系统,验证了其方法的正确性和有效性。其次,针对使用SDG进行故障分析时所获信息不完备、不精确的特征,将粗糙集挖掘不确定性问题的优点加入SDG故障诊断过程中,提出了一种基于粗糙图的分层SDG故障诊断模型。该模型包括部件节点粗糙关系图、粗糙分层SDG的故障图生成算法及故障最大流分析算法三部分。实例验证和模型分析表明该模型较传统SDG故障诊断模型更能反映实际情况,所获诊断结论更加准确。最后,将基于粗糙图的分层SDG故障诊断模型用于热力系统,对其不同关系层面之间的故障进行诊断,结果表明该模型能够降低热力设备动态特性故障知识获取的复杂度,及时准确识别热力系统早期故障和轻微故障,这些改进都能在很大程度上提高热力系统诊断效率,促进热力系统故障诊断技术不断的向前发展。本文将系统分层理论和粗糙图、粗糙网络理论引入SDG故障诊断,提出了一种基于粗糙图的分层SDG故障诊断模型,使得SDG搜索推理发生冲突的概率在一定程度上减少甚至避免,提高了传统SDG故障诊断搜索推理的分辨率。该方法用于热力系统,实例结论表明其行之有效、具有良好的通用性,对提高机组运行的安全性和经济性起到一定作用,推动了故障诊断智能化的不断发展。
[Abstract]:In the fault of thermal power unit, the probability of thermal system failure is very large, so the safety of thermal system will have a great impact on the safety and economy of the whole operation of the unit, because the structure of the thermal power system is complex and the operating conditions are changeable. In the process of fault diagnosis, it is difficult to establish and perfect the fault knowledge base, so it is necessary to study the fault diagnosis of thermal system. In this paper, graph theory and rough graph theory are combined with symbolic directed graph (SDG) to make the SDG model easier to diagnose and improve the accuracy of SDG fault diagnosis. Firstly, because of the low resolution caused by SDG polysemy reasoning, the diagnosis speed of large system is slow. In this paper, the fault propagation graph and SDG are combined to describe the system, and a hierarchical symbolic directed graph (SDG) fault diagnosis model is proposed, which has the completeness of SDG itself. It also has the inherent properties of the fault propagation graph, which can reflect the connection relationship between the system components, and at the same time, it will be stratified technology, node compression technology, The constraint propagation technique and the four kinds of test point optimization techniques, which include the improved symbolic directed graph model, are added to construct the hierarchical SDG model. And the backtracking search strategy is used to search the independent compatible path for fault diagnosis. The above techniques make use of the characteristics of fault propagation between the components of the system equipment, and can reduce the set of candidate test points in the reasoning process. To improve the efficiency of searching fault sources, this method can distinguish the unknown and difficult faults in the same fault mode from the "combined explosion" problem that may occur. The results obtained are compared with the traditional SDG diagnosis results. It is shown that this method is superior to the traditional method. Finally, the method is applied to the centrifugal pump and liquid level system, and the correctness and effectiveness of the method are verified. Secondly, aiming at the incomplete and inaccurate information obtained in fault analysis with SDG, the advantages of rough set mining uncertainty problem are added to the process of SDG fault diagnosis. A hierarchical SDG fault diagnosis model based on rough graph is proposed. There are three parts of fault graph generation algorithm and fault maximum flow analysis algorithm of rough layered SDG. The example verification and model analysis show that the model can reflect the actual situation better than the traditional SDG fault diagnosis model, and the diagnosis results are more accurate. The hierarchical SDG fault diagnosis model based on rough graph is applied to the fault diagnosis of thermal system. The results show that the model can reduce the complexity of fault knowledge acquisition of thermal equipment dynamic characteristics. These improvements can improve the efficiency of thermal system diagnosis to a great extent and promote the continuous development of thermal system fault diagnosis technology. In this paper, the hierarchical theory of system, rough graph and rough network theory are introduced into SDG fault diagnosis, and a hierarchical SDG fault diagnosis model based on rough graph is proposed, which can reduce or even avoid the probability of collision of SDG search reasoning to a certain extent. The resolution of traditional SDG fault diagnosis search reasoning is improved. The method is used in thermal system. The example shows that the method is effective, has good generality, and plays a certain role in improving the safety and economy of unit operation. It promotes the development of fault diagnosis intelligence.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TM621;TH165.3

【参考文献】