下载文档原格式
电力故障诊断方法研究的一些参考文献Research on the method of power failure diagnosis has been a crucial area of focus in the field of electrical engineering. One important reference in this area is the paper "Power Proxy: Anomaly Detection in Power Usage Data" by Zhang et al. This paper proposes a novel approach using deep learning techniques to detect anomalies in power usage data, which can aid in the diagnosis of power failures. The authors demonstrate the effectiveness of their method through experiments on real-world power usage datasets.电力故障诊断方法的研究一直是电气工程领域的一个重要研究方向。
张等人的论文《Power Proxy: Anomaly Detection in Power Usage Data》是这方面的一个重要参考文献。
这篇论文提出了一种新颖的方法，利用深度学习技术来检测电力使用数据中的异常，有助于诊断电力故障。
作者通过对真实电力使用数据集的实验验证了他们方法的有效性。
In addition to Zhang et al.'s work, another valuable reference is the paper "A Survey on Fault Diagnosis Techniques Through EE Stream Processing" by Wang et al. This survey paper provides a comprehensive overview of fault diagnosis techniques in the contextof electrical engineering stream processing. The authors discuss various methods such as Bayesian networks, neural networks, and decision trees, highlighting their applications in diagnosing power failures. This paper serves as a useful guide for researchers interested in exploring different fault diagnosis techniques.除了张等人的工作，王等人的论文《A Survey on Fault Diagnosis Techniques Through EE Stream Processing》也是一个很有价值的参考文献。
高一英语科学发现词汇应用高级单选题40题1. In the process of scientific research, a(n) ____ is often put forward first, which is a proposed explanation for a phenomenon.A. experimentB. hypothesisC. observationD. conclusion答案：B。
解析：本题考查科学发现相关词汇的辨析。
“hypothesis”意为假设，在科学研究中通常首先提出一个假设来解释某种现象，这是科学研究的常见步骤。
“experiment”是实验，是用来验证假设的手段，而非首先提出的内容。
“observation”是观察，虽然观察也是研究的一部分，但不是这种对现象提出解释的概念。
“conclusion”是结论，是在经过一系列研究之后得出的结果，不是一开始就提出的。
2. Scientists made a careful ____ of the strange phenomenon before they started their research.A. experimentB. hypothesisC. observationD. discovery答案：C。
解析：这里考查词汇的语境运用。
“observation”表示观察，科学家在开始研究之前会对奇怪的现象进行仔细观察，这是符合逻辑的。
“experiment”是进行实验，在还未开始研究时不会先进行实验。
“hypothesis”是假设，此时还未到提出假设的阶段。
“discovery”是发现，这里强调的是对现象的观察过程，而不是发现这个结果。
3. The ____ they designed was very complicated but it could test their hypothesis effectively.A. experimentB. observationC. conclusionD. theory答案：A。
第39卷第1期2018年1月白动化仪表PROCESS AUTOMATION INSTRUMENTATIONVol. 39 N o. 1Jan.2018融合多种算法的堆垛机诊断专家系统吕婷\杨涛2(1.安徽电子信息职业技术学院信息与智能系，安徽蚌埠233000;2.西南科技大学信息工程学院特殊环境机器人技术四川省重点试验室，四川绵阳621010)摘要：针对堆垛机设备在运行过程中呈现的复杂性、不确定性等问题，设计了基于故障树和贝叶斯网络的混合诊断专家系统。
采用故障树分析技术对堆垛机进行故障建模，得到最小割集，建立了以规则为知识表示形式的规则库。
根据输入的故障征兆系统 自动寻找匹配的故障事实库，建立了以该事件作为顶事件的故障树，并转化得到相应的贝叶斯网络，形成了基于规则的推理和贝叶 斯网络的概率计算混合诊断机制。
该方法有效利用了故障树分析和贝叶斯网络两种算法的优势，为复杂机器的故障诊断提供了一 种新途径。
试验表明，该系统有效解决了传统诊断专家系统存在的推理模式单一、知识获取困难等问题。
概率计算混合诊断机制是 一种快速诊断堆垛机的可行方式。
关键词：堆垛机；故障诊断；贝叶斯网络；专家系统；故障树分析中图分类号：TH165 ;TP182 文献标志码：A DOI：10.16086/j. cnki. issnlOO O -0380.201801003 Fault Diagnosis Expert System with Multiple Algorithms for StackersLYU Ting1,YANG Tao2(1. Department of Information and Intelligence, Anhui Vocational College ofElectronics and Information Technology ,Bengbu 233000, China ；2. Robot Technology Used for Special Environment Key Laboratory of Sichuan Province,School of Information Engineering, Southwest University of Science and Technology ,Mianyang 621010, China)A bstract：According to the complexity and uncertainty of the stackers during operating,the hybrid fault diagnosis expert system based on fault tree and Bayesian network are designed. By using fault tree analysis technology, the model of stacker is established, and the minimum cut set is obtained, thus the rule base is constructed with the rules as the knowledge representation. In accordance with the input failure symptom system, the system automatically searches the exactly matched fault fact, and builds the fault tree that with the fact as the top event. Then the corresponding Bayesian network istransfromed based on the fault tree. The hybrid diagnosis mechanism is realized based on the rule -based reasoning and Bayesian network -based probability computation. The method is simple and flexible ；the advantages of both the fault tree analysis and Bayesian network are used to provide a new way for fault diagnosis of complex machines. The tests show that the system effectively resolves the difficulty of single reasoning mode of traditional expert system. The probability calculation mixed diagnosis mechanism is a feasible way to diagnose trackers quickly.K eyw ords：Stacker；Fault diagnosis；Bayesian network；Expert system；Fault tree analysis〇引言作为立体仓库的核心设备，堆垛机直接影响整个 物流系统的正常运作。
我钦佩钟南山院士英语作文{z}Title: My Admiration for Professor Zhong NanshanProfessor Zhong Nanshan, a renowned respiratory disease expert in China, has won the hearts of millions with his courage, expertise, and unwavering dedication to public health.His significant contributions to the field of medicine, especially during the COVID-19 pandemic, have made him an inspiration for many.Born in 1936 in Guangzhou, China, Professor Zhong completed his medical education at Peking Union Medical College in 1960.He then went on to specialize in respiratory diseases at the same institution.Over the years, he has held various academic and administrative positions, and has been instrumental in establishing several respiratory disease research centers in China.One of Professor Zhong"s most notable contributions is his research on tuberculosis and chronic obstructive pulmonary disease (COPD).His work has not only enhanced our understanding of these diseases but has also led to improved diagnostic and treatment methods.His efforts in promoting public health awareness and advocating for policies to combat respiratory diseases have had a significant impact on the health of the Chinese population.In 2003, during the severe acute respiratory syndrome (SARS) epidemic, Professor Zhong played a pivotal role in containing theoutbreak.He led a team of researchers in investigating the virus and developing effective treatment strategies.His courage and expertise in the face of the crisis earned him the nickname "the people"s doctor."Fast forward to 2020, when the world was once again gripped by a respiratory disease pandemic - COVID-19.Professor Zhong, despite being in his late 80s, was at the forefront of the battle against the virus.His calm and authoritative demeanor during interviews and press conferences provided much-needed reassurance to the public.He was one of the first to raise the alarm about the severity of the pandemic and called for strict measures to control its spread.Throughout the pandemic, Professor Zhong has been a beacon of hope, providing accurate information and guidance based on scientific evidence.His commitment to truth and his unwavering dedication to public health have earned him the respect and admiration of people worldwide.In recognition of his exceptional contributions, Professor Zhong has received numerous awards and honors.He was awarded the Republic of China"s highest honor, the Order of Merit, in 2020.His work has also been recognized globally, and he has been conferred honorary doctorates by several universities around the world.In conclusion, Professor Zhong Nanshan is a true hero in the field of medicine.His selflessness, expertise, and unwavering commitment topublic health have made him an inspiration for generations to come.As we continue to battle the ongoing pandemic, his legacy will serve as a reminder of the importance of science, truth, and unity in the face of adversity.。
在多年研究成果的基础上，本文利用Delphi7及ODBC 数据库开发出了基于Access 数据库的电力变压器故障诊断系统。
故障诊断基于专家知识库，而专家知识库又由专家经验构成。
现场通常采用潜伏性故障诊断和绝缘预防性诊断对变压器状 态进行诊断。
其中，潜伏性故障诊断以油中溶解气体色谱分析为基础，结合外部检 查、绝缘油诊断来综合分析判断运行中变压器的潜伏性故障；而绝缘预防性诊断则 由绝缘电阻、直流电阻、介损、直流泄漏、套管诊断构成。
通过潜伏性故障诊断和绝缘预防性诊断，专家系统能够综合判断变压器的整体绝缘水平，并为现场操作人空一行 关键词：电力变压器；专家系统；故障诊断；数据库；知识库黑体小四号提示：论文摘要是学位论文的缩影， 要以浓缩的形式概括课题的研究内容， 文字要简练、明确。
内容要包括目的、方法、结果和结论。
单位制一律换算成国际标 准计量单位制，除特别情况外，数字一律用阿拉伯数码。
文中不允许出现插图。
除封面，页边距上下2.54左3.17右2.5若无特别说明，则不空行插入分页符空一行 黑体小三号，居中，1.5倍 行距，段前0、段后0.5员提出建议。
本专家系统在大量调研的基础上完成考虑到现场的需要，有较强的实用价值。
j 以试验报告等功能，充分 正文300-500字，宋体小四号，1.5倍行距，段前、 段后均为0行，每段的首行缩进 2汉字用；号分隔Power Transformer Fault Diagnosis Expert System字体：Times New Roman 小四号行间距：固定值 25磅Based on researching for several years, Delphi7 and ODBC are utilized to developTran sformer Fault Diag nosing Expert System (TFDES) based on Access database. FaultDiag no sis is based on expert kno wledge base, composed by experts Fault Diagnostic (LFD) and Insulation Precautionary Test (IPT) are popularly used inreality ,so as to diagnose faults of transfo rmers ' insulation. LFD, is used to indicate the late nt faults of tran sformers, based on Dissolved Gas An alysis (DGA), and assistedby External Exam in ati on (EE), In sulati on Oil (IO).IPT, con sisted of In sulati on Resista nee ,Ohmic Resista nee, Oil Dielectric Loss ,DC leak ing Curre nt and Bush ing. Through LFD and ITP, TFDES can judge synthetically the whole insulation level of tran sformer, and give out proper expert suggestio ns to operators. This TFDES , completed through a lot of investigation andresearching, affiliated with Test Report and some other fun cti ons, th in ki ng completely of on-the-spot n eed, is very worthy practically.—四号 Times New Roman 加粗------ -------------- 」Key Words: Electric Tran sformer; Expert System; Faults Diag no sis; Database; Kno wledge Base提示：英文摘要应与中文摘要对应experie nces. Late AbstractTimes New Roman 小三号加粗居 中，1.5倍行距，段前0、段后0.5插入分页符第二章电力变压器绝缘故障诊断模型 ........................2.1.电力变压器结构简介 .................................................. 5 2.2. 电力变压器绝缘基本知识 . (5)2.2.1电力变压器的绝缘 (5)2.2.2油浸式变压器常用绝缘材料 (6)2.3. 电力变压器的故障及检测手段 (7)2.3.1故障原因及其种类 (7)2.3.2电力变压器常规试验项目 (8)2.4. 电力变压器异常情况的分析 (12)2.4.1声音异常 ....................................................... 12 2.4.2油温异常 . (13)2.4.3油位异常 (14)2.4.4外表异常 (15)2.3.1气温、颜色异常 ................................................. 15 2.5. 油中气体色谱分析法 (15)2.5.1油中气体组成的规律 ............................................. 16 2.5.2油中气体含量与故障性质的关系 . (16)2.5.3 油中气体分析过程 (17)2.5.4气体分析方法 (17)第三章电力变压器绝缘故障诊断专家系统 (26)3.1. 专家系统的概念 (26)3.2. 专家系统的结构 (28)3.2.1 知识库 (28)3.2.2数据库 (29)3.2.3推理机 (29)3.2.4数据管理 (30)黑体小三号一级标题格式目录目录自动生成标题按此格式对齐，正文格式 第一章引言3.2.5 人机界面 (30)3.3.专家系统中模糊问题的处理 (30)331 模糊知识的获取 (31)332模糊综合评判的处理方法 (33)34专家系统各模块介绍 (34)3.4.1气体色谱跟踪试验模块 (34)3.4.2潜伏性故障诊断模块 (35)3.4.3绝缘预防性诊断模块 (35)第四章专家系统的完善与应用开发 (36)4.1前期准备 (36)4.1.1经验总结 (36)4.1.2开发方案 (37)4.2完善与应用开发 (38)4.2.1诊断系统 (39)4.2.2管理系统 (43)4.2.3查询系统 (44)4.2.3界面风格 (45)4.2.4系统说明 (45)4.3诊断程序函数说明 (46)4.3.1诊断函数 (46)4.3.2主要辅助函数 (48)4.4诊断实例 (50)第五章结论 (52)5.1结论 (52)5.2电力变压器故障诊断专家系统的展望 (52)致谢 (54)参考文献 (55)＜附录1潜伏性故障诊断结论＞ (56)＜附录2绝缘预防试验报告＞ (58)＜附录3试验报告制作对应机制＞ (60)最近十多年来，我国的国民经济一直以 10%左右的速度稳定发展，为满足国民 经济对电能需求的迅速增长，我国电网的规模日益扩大。
Recent Progress on Mechanical Condition Monitoring and Fault diagnosis Abstract:Mechanical equipments are widely used in various industrial applications. Generally working in severe conditions, mechanical equipments are subjected to progressive deterioration of their state. The mechanical failures account for more than 60% of breakdowns of the system. Therefore, the identification of impending mechanical fault is crucial to prevent the system from malfunction. This paper discusses the most recent progress in the mechanical condition monitoring and fault diagnosis. Excellent work is introduced from the aspects of the fault mechanism research, signal processing and feature extraction, fault reasoning research and equipment development. An overview of some of the existing methods for signal processing and feature extraction is presented. The advantages and disadvantages of these techniques are discussed. The review result suggests that the intelligent information fusion based mechanical fault diagnosis expert system with self-learning and self-updating abilities is the future research trend for the condition monitoring fault diagnosis of mechanical equipments.Keywords: Condition monitoring; Fault diagnosis; Vibration; Signal processing1. IntroductionWith the development of modern science and technology, machinery and equipment functions are becoming more and more perfect, and the machinery structure becomes more large-scale, integrated, intelligent and complicated. As a result, the component number increases significantly and the precision requirement for the part mating is stricter. The possibility and category of the related component failures therefore increase greatly. Malignant accidents caused by component faults occur frequently all over the world, and even a small mechanical fault may lead to serious consequences. Hence, efficient incipient fault detection and diagnosis are critical to machinery normal running. Although optimization techniques have been carried out in the machine design procedure and the manufacturing procedure to improve the quality of mechanical products, mechanical failures are still difficult to avoid due to the complexity of modern equipments. The condition monitoring and fault diagnosis based on advanced science and technology acts as an efficient mean to forecast potential faults and reduce the cost of machine malfunctions. This is the so-called mechanical equipment fault diagnosis technology emerged in the nearly three decades [1, 2]. Mechanical equipment fault diagnosis technology uses the measurements of the monitored machinery in operation and stationary to analyze and extract important characteristics to calibrate the states of the key components. By combining the history data, it can recognize the current conditions of the key components quantitatively, predicts the impending abnormalities and faults, and prognoses their future condition trends. By doing so, the optimized maintenance strategies can be settled, and thus the industrials can benefit from the condition maintenance significantly [3, 4].The contents of mechanical fault diagnosis contain four aspects, including fault mechanism research, signal processing and feature extraction, fault reasoning research and equipment development for condition monitoring and fault diagnosis. In the past decades, there has been considerable work done in this general area by many researchers. A concise review of the research in this area has been presented by [5, 6]. Some landmarks are discussed in this paper. The novel signal processing techniques are presented. The advantages and disadvantages of these new signal processing and feature extraction methods are discussed inthis work. Then the fault reasoning research and the diagnostic equipments are briefly reviewed. Finally, the future research topics are described in the point of future generation intelligent fault diagnosis and prognosis system.2. Fault Mechanism ResearchFault Mechanism research is a very difficult and important basic project of fault diagnosis, same as the pathology research of medical. American scholar John Sohre, published a paper on "Causes and treatment of high-speed turbo machinery operating problems (failure)", in the United States Institute of Mechanical Engineering at the Petroleum Mechanical Engineering in 1968, and gave a clear and concise description of the typical symptoms and possible causes of mechanical failure. He suggested that typical failures could be classified into 9 types and 37 kinds [7]. Following, Shiraki [8] conduced considerable work on the fault mechanism research in Japan during 60s-70s last century, and concluded abundant on-site troubleshooting experience to support the fault mechanism theory. BENTLY NEV ADA Corporation has also carried out a series experiments to study the fault mechanism of the rotor-bearing system [9].A large amount of related work has been done in China as well. Gao et al. [10] researched the vibration fault mechanism of the high-speed turbo machinery, investigated the relationship between the vibration frequency and vibration generation, and drew up the table of the vibration fault reasons, mechanism and recognition features for subsynchronous, synchronous and super-synchronous vibrations. Based on the table they proposed, they have classified the typical failures into 10 types and 58 kinds, and provided preventive treatments during the machine design and manufacture, Installation and maintenance, operation, and machine degradation. Xu et al. [11] concluded the common faults of the rotational machines. Chen et al.[12] used the nonlinear dynamics theory to analyze the key vibration problems of the generator shaft. They established a rotor nonlinear dynamic model for the generator to comprehensively investigate the rotor dynamic behavior under various influences, and proposed an effective solution to prevent rotor failures. Yang et al. [13] adopted vibration analysis to study the fault mechanism of a series of diesel engines. Other researchers have done a lot in the fault mechanism of mechanics since 1980s, and have published many valuable papers to provide theory and technology supports in the application of faultdiagnosis systems [14-18]. However, most of the fault mechanism research is on the qualitative and numerical simulation stage, the engineering practice is difficult to implement. In addition, the fault information often presents strong nonlinear, non stationary and non Gaussian characteristics, the simulation tests can not reflect these characteristics very accurately. The fault diagnosis results and the application possibility may be influenced significantly. As a result, the development of the fault diagnosis technique still faces great difficulties.3. Advanced Signal Processing and Feature Extraction MethodsAdvanced signal processing technology is used to extract the features which are sensitive to specific fault by using various signal analysis techniques to process the measured signals. Condition information of the plants is contained in a wide range of signals, such as vibration, noise, temperature, pressure, strain, current, voltage, etc. The feature information of a certain fault can be acquired through signal analysis method, and then fault diagnosis can be done correspondingly. To meet the specific needs of fault diagnosis, fault feature extraction and analysis technology is undergoing the process from time domain analysis to Fourier analysis-based frequency-domain analysis, from linear stationary signal analysis to nonlinear and nonstationary analysis, from frequency-domain analysis to time-frequency analysis. Early research on vibration signal analysis is mainly focused on classical signal analysis which made a lot of research and application progress. Rotating mechanical vibration is usually of strong harmonic, its fault is also usually registered as changes in some harmonic components. Classical spectrum analysis based on Fourier transform (such as average time-domain techniques, spectrum analysis, cepstrum analysis and demodulation techniques) can extract the fault characteristic information effectively, thus it is widely used in motive power machine, especially in rotating machinery vibration monitoring and fault diagnosis. In a manner of speaking, classical signal analysis is still the main method for mechanical vibration signal analysis and fault feature extraction. However, classical spectrum analysis also has obvious disadvantages. Fourier transform reflects the overall statistical properties of a signal, and is suitable for stationary signal analysis. In reality, the signals measured from mechanical equipment are ever-changing, non-stationary, non-Gaussian distribution and nonlinearrandom. Especially when the equipment breaks down, this situation appears to be more prominent. For non-stationary signal, some time-frequency details can not be reflected in the spectrum and its frequency resolution is limited using Fourier transform. New methods need to be proposed for those nonlinearity and non-stationary signals. The strong demand from the engineering practice also contributes to the rapid development of signal analysis. New analytical methods for non-stationary signal and nonlinear signal are emerging constantly, which are soon applied in the field of machinery fault diagnosis. New methods of signal analysis are main including time-frequency analysis, wavelet analysis, Hilbert-Huang transform, independent component analysis, advanced statistical analysis, nonlinear signal analysis and so on. The advantages and disadvantages of these approaches are discussed below.4. Research on Fault ReasoningAt present, many methods are adopted in the process of diagnostic reasoning. According to the subject systems which they belong to, the fault diagnosis can be divided into three categories: (1) the fault diagnosis based on control model; (2) the fault diagnosis based on pattern recognition; (3) the fault diagnosis based on artificial intelligence. Among them, the fault diagnosis based on control model needs to establish model through theoretic or experimental methods. The changes of system parameters or system status could directly reflect the changes of equipments physical process, and hence it is able to provide basis for fault diagnosis. This technology refers to model establishment, parameters estimation, status estimation, application of observers, etc. Since it requires accurately system model, this method is not economically feasible for the complicated devices in the practice. Pattern recognition conducts cluster description for a series of process or events. It is mainly divided into statistical method and language structure method. The fault diagnosis of equipments could be recognized as the pattern recognition process, that is to say, it recognizes the fault based on the extraction of fault characteristics. There are many common recognition methods, including bayes category, distance function category, fuzzy diagnosis, fault tree analysis, grey theory diagnosis and so on. Recent years, some new technologies have been also applied in the field of the fault diagnosis of rotary machines, such as the combination of fuzzy set andneural network, the dynamic pattern recognition based on hidden markov model, etc.5. Research and Development of Fault Diagnosis DevicesFault diagnosis technology ultimately comes down to the actual devices, and at present research and development of fault diagnosis devices is in the following two directions: (1) Portable vibration monitoring and diagnosis (including data collector system), and (2) On-line condition monitoring and fault diagnosis system. Portable instrument is mainly adopted single-chip microcomputers to complete data acquisition, which has certain ability for signal analysis and fault diagnosis. On-line monitoring and diagnosis system is usually equipped with sensors, data acquisition, alarm and interlock protection, condition monitoring subsystem, etc. And it is also fitted with rich signal analysis and diagnosis software. These software include America BENTLY Corporation 3300, 3500 and DM2000 systems, America Westinghouse Company PDS system, the 5911 system developed by ENTECK and IRD Company, Japan Mitsubishi MHM system, the Danish B&K Company B&K 3450 COMPASS system, etc. China has also successively developed large on-line monitoring and fault diagnosis system, which has been put into use on steam turbine and other important equipments. Based on the realization of condition monitoring of equipments, network diagnostics center can monitor and diagnose the operation of equipments at any time through the network to achieve the long distance information transmission. The remote monitoring system can also achieve the collaborative diagnosis of production equipments, multiple diagnostic systems serve the same piece of equipment, and multiple devices share the same diagnostic system.6. ConclusionsTo achieve a dynamic system condition monitoring and fault diagnosis, primary task is the need to get enough reliable characteristic information from the system. Due to the fluctuation of the system itself and the environment disturbance, reliable signal collection is seriously affected. It is therefore very urgent for advanced signal processing technology to eliminate noise to get true signal. No matter classical or advance fault diagnosis techniques, they have achieved great progress in various applications. In the point of systematic view, every technology is a part of the whole diagnostic system, and the efficient fusion of these parts willprovide best performance for the condition monitoring and fault diagnosis. Thus, the fault mechanism research, signal processing and feature extraction, fault reasoning research and equipment development will connect even tighter to form an effective fault diagnostic expert system in the future. To realize the expert system, the core issue is to break through the bottleneck of knowledge acquisition, update the data model in a reliable manner and provide good generalization ability of the expert system. By doing so, the fault diagnostic expert system can offer accurate estimation of the potential abnormalities, and prevent them before breaking out to ensure the normal operation of the machines. Hence, the loss caused by the machine breakdowns can be minimized significantly.References[1] Wu XK. The fault diagnosis based on information fusion theory and its application in internal combustion engine. Ph.D. thesis, Wuhan University of Technology, 1998.[2] Chen YR. Modern signal processing technology in the application of vibration diagnosis of internal combustion engine. Ph.D. thesis, Wuhan University of Technology, 1998.[3] Qu LS, He ZJ. Mechanical fault diagnostics. Shanghai: Shanghai Science and Technology Press, 1986.[4] Huang WH, Xia SB, Liu RY. Equipment fault diagnosis principle, technology and application. Beijing: Science Press, 1996.[5] Jayaswalt P, Wadhwani AK. Application of artificial neural networks, fuzzy logic and wavelet transform in fault diagnosis via vibration signal analysis: A review. Australian Journal of Mechanical Engineering 2009; 7: 157-172.[6] Daneshi-Far Z, Capolino GA, Henao H. Review of failures and condition monitoring in wind turbine generators. 19th International Conference on Electrical Machines. Rome, Italy; 2010.[7] Sohre JS. Trouble-shooting to stop vibration of centrifugal. Petrop Chem. Engineer 1968; 11: 22-23.[8] Shiraki T. Mechanical vibration lectures. Zhengzhou: Zhengzhou Mechanical Institute; 1984.[9] Bently DW. Forced subrotative speed dynamic action of rotating machinery. USA: ASME Publication, 74-pet-16.[10] Gao JJ. Research on high speed turbine machinery vibration fault mechanism and diagnostic method. Ph.D. thesis, Xi'an Jiaotong University, 1993.[11] Xu M, Zhang RL. Equipment fault diagnosis manual. Xi’an: Xi'an Jiaotong University Press, 1998.[12] Chen YS, Tian JY, Jin ZW, Ding Q. Theory of nonlinear dynamics and applied techniques of solving irregular operation of a large scale gas turbine in a comprehensive way. China Mechanical Engineering 1999; 10: 1063-68.[13] Yang JG, Zhou YC. Internal combustion engine vibration monitoring and fault diagnosis.Dalian: Dalian Maritime University Press, 1994.[14] Wang Y, Gao JJ, Xia SB. The study of causes and features of faults in supporting system for rotary machinery. Journal of Harbin Institute of Technology 1999; 31:104-6.[15] Liu SY, Song XP, Wen BC. Catastrophe in fault developing process of rotor system. Journal of Northeastern University (Natural Science) 2004; 17:159-162.[16] Han J, Zhang RL. Rotating machinery fault mechanism and diagnostic technique. Beijing: China Machine Press, 1997.[17] Chen AH. Research on some nonlinear fault phenomenon of rotating machinery. Ph.D. thesis, Central South University of Technology, 1997.机械状态监测和故障诊断的最新进展摘要：机械设备被广泛的使用在各种工业应用中。
每转产生一次脉冲电压的传感器，这个电压脉冲叫做键相器信号。
这个信号主要用来测量轴的转数，并可对所测振动相位角提出参考坐标。
4．6．2．9 电涡流传感器eddy current probe一种非接触式器件，以电涡流原理工作，能用来测量位移运动以及待测表面相对于安装点的距离。
4．6．2．10 转速表tachometer测量转轴角速度的仪表。
4．6．2．11 汽轮机(旋转机械)故障诊断系统automated diagnostics for steam turbine(rotating equip— ment)(ADRE)是一个能采集汽轮机(旋转机械)各轴承的振动数据，通过专家系统软件综合分析汽轮机(旋转机械)运行状况，对存在的隐患进行判断、预告或处理的专用装置。
它包括振动检测及计算机数据采集系统和专用的应用软件。
4．6．3 联锁interlock有两种含义：a)在相互关联设备间的生产流程中(如输煤系统、锅炉的燃烧系统)，当某一设备故障时，为保证安全运行按先后次序安全停止故障设备前或后的自动操作。
b)为防止超出极限状况或不适当的操作程序，因而危及设备安全，采取关停设备中造成故障的有关设备或防止进入不恰当的操作程序，以避免危险工况的联动操作，如高压加热器切旁路。
4．6．4 机组快速甩负荷fast cut back(FCB)当汽轮机或发电机甩负荷时，使锅炉不停运的一种控制措施，根据FCB后机组的不同运行要求，可分为两种不同的运行方式：a)5％FCB，是机组带厂用电单独运行的方式。
b)0％FCB，是停机不停炉的运行方式。
4．6．5 辅机故障减负荷run back(RB)是针对机组主要辅机故障采取的控制措施。
即当主要辅机(如给水泵、送风机、引风机)发生故障、机组不能带额定负荷时，快速降低机组负荷的措施。
4．6．6 联锁控制interlock control某一参数到达规定值或某一设备启停时，同时控制另一设备的控制。
基于智能技术的电力变压器故障诊断系统在多年研究成果的基础上，本文利用Delphi7及ODBC数据库开发出了基于Access数据库的电力变压器故障诊断系统。
故障诊断基于专家知识库，而专家知识库又由专家经验构成。
现场通常采用潜伏性故障诊断和绝缘预防性诊断对变压器状态进行诊断。
其中，潜伏性故障诊断以油中溶解气体色谱分析为基础，结合外部检查、绝缘油诊断来综合分析判断运行中变压器的潜伏性故障；而绝缘预防性诊断则由绝缘电阻、直流电阻、介损、直流泄漏、套管诊断构成。
通过潜伏性故障诊断和绝缘预防性诊断，专家系统能够综合判断变压器的整体绝缘水平，并为现场操作人考虑到现场的需要，有较强的实用价值。
关键词Based on researching for several Transformer Fault Diagnosing Expert System (TFDES) based on Access database. Fault Diagnosis is based on expert knowledge base, composed by experts’ experiences. Latent Fault Diagnostic (LFD) and Insulation Precautionary Test (IPT) are popularly used in reality ,so as to diagnose faults of transfo rmers’ insulation. LFD, is used to indicate the latent faults of transformers, based on Dissolved Gas Analysis (DGA), and assisted by External Examination (EE), Insulation Oil (IO).IPT, consisted of Insulation Resistance ,Ohmic Resistance, Oil Dielectric Loss ,DC leaking Current and Bushing. Through LFD and ITP, TFDES can judge synthetically the whole insulation level of transformer, and give out proper expert suggestions to operators. This TFDES , completed through a lot of investigation and researching, affiliated with Test Report and some other functions, thinking completely of on-the-spot need, is very worthy practically.Key System; Faults Diagnosis; Database; Knowledge Base目录1第二章电力变压器绝缘故障诊断模型 (5)2.1. 电力变压器结构简介 (5)2.2. 电力变压器绝缘基本知识 (5)2.2.1电力变压器的绝缘 (5)2.2.2油浸式变压器常用绝缘材料 (6)2.3. 电力变压器的故障及检测手段 (7)2.3.1 故障原因及其种类 (7)2.3.2 电力变压器常规试验项目 (8)2.4. 电力变压器异常情况的分析 (12)2.4.1 声音异常 (12)2.4.2 油温异常 (13)2.4.3 油位异常 (14)2.4.4 外表异常 (15)2.3.1 气温、颜色异常 (15)2.5. 油中气体色谱分析法 (15)2.5.1 油中气体组成的规律 (16)2.5.2 油中气体含量与故障性质的关系 (16)2.5.3 油中气体分析过程 (17)2.5.4 气体分析方法 (17)第三章电力变压器绝缘故障诊断专家系统 (26)3.1.专家系统的概念 (26)3.2.专家系统的结构 (28)3.2.1 知识库 (28)3.2.2 数据库 (29)3.2.3 推理机 (29)3.2.4 数据管理 (30)3.2.5 人机界面 (30)3.3.专家系统中模糊问题的处理 (30)3.3.1 模糊知识的获取 (31)3.3.2 模糊综合评判的处理方法 (33)3.4.专家系统各模块介绍 (34)3.4.1 气体色谱跟踪试验模块 (34)3.4.2 潜伏性故障诊断模块 (35)3.4.3 绝缘预防性诊断模块 (35)第四章专家系统的完善与应用开发 (36)4.1前期准备 (36)4.1.1 经验总结 (36)4.1.2 开发方案 (37)4.2完善与应用开发 (38)4.2.1 诊断系统 (39)4.2.2 管理系统 (43)4.2.3 查询系统 (44)4.2.3 界面风格 (45)4.2.4 系统说明 (45)4.3诊断程序函数说明 (46)4.3.1诊断函数 (46)4.3.2 主要辅助函数 (48)4.4诊断实例 (50)第五章结论 (52)5.1结论 (52)5.2电力变压器故障诊断专家系统的展望 (52)致谢 (54)参考文献 (55)<附录1 潜伏性故障诊断结论> (56)<附录2 绝缘预防试验报告> (58)<附录3 试验报告制作对应机制> (60)最近十多年来，我国的国民经济一直以10%左右的速度稳定发展，为满足国民经济对电能需求的迅速增长，我国电网的规模日益扩大。
神经网络与专家系统的结合及其应用研究第l0卷第2期江八一农垦大学……醚盎塑'王智敏(黑龙江八一农垦大学工程学院密山158308)摘要在分析神经罔络与专家系统相结合的优点基础上.探讨了神经阿络与专家系统的几种常见结合方式,■述了该方法的典型应用一一基于神经阿络的故障诊断系统,并以发动机故障诊断为倒精出了两者结台的具体宴现.中国分类号TP3191引言专家系统和神经网络是两种主要的人工智能应用技术.将专家系统与神经网络有机结合,两者取长朴短,充分发挥各自特长.再加入模糊理论等先进技术.是当前智能系统发展的基本特征和必然趋势.如何把它们结合得更合理更巧妙已成为有关专家共同探讨的新兴前沿课题.本文对神经网络和专家系统的结合方式进行了初步探索.神经网络为现代机器的故障诊断提供了新的理论方法和技术手段,具有很大的发展潜力和应用前景.利用神经网络与专家系统技术相结合.提高了系统的智能水平.可实现诊断的准确,快速和高效性,也为汽车发动机的故障诊断提供了一种新的方法和思路.2神经网络与专家系统的互补神经网培可以弥补解决传统专家系统在应用中遇到的问题.比如,(1)专家系统的.脆弱性印知识和经验不全面.遇到没解决过的问题就无能为力;利用神经网络的自学习不断丰富知识库内容,从而解决知识更新的同题.《2)对于E8"知识获取的困难这.瓶颈问题,利用ANN的高效性和方便的自学习功能,只需用领域专家解决问题的实例来训练ANN.使在同样的输入条件下,ANN能获得与专家给出的解答尽可能接近的输出.(3)推理中的匹配冲突.组合爆炸及.无穷递归使传统鹤推理速度慢,效率低,主要是由于E8采用串行方式,推理方法简单和控制策略不灵活.而ANN的知识推理通过神经元之间的作用实现,总体上ANN的推理是并行的.速度快.一般来说.ANN是基于精人～输出的一种直觉性反射,也叫形象思维经验思维,适于发挥经验知识的作用,进行浅层次的经验推理E8是基于知识,规则匹配的逻辑知识的作用,进行深层次的逻辑推理.鹤的特色是符号推理,ANN擅长数值计算.由此可见.传统鹤与^NN科学地加以综合,并加人探层次知识,取长补短,充分发挥各自的特长,将会提高智能系统的智力水平.1998—04—28l趺稿?中国农业大学东区2l4信箱孙永厚?男,31岁?讲师.中国农业大学(东区)硬士研究生毕业.第2期孙永厚等:神墅哩鳖童塞墨堡塑堕全垦基堡旦里塞3神经网络与专家系统结合的方式神经网络与专家系统结合的方法多种多样,常见的有以下几种.首先,按连接方式分为:(1)并列协同法:并列使用神经网络,专家系统和算法库等作为各自独立的模块,执行系统的某些功能,最后经过组合,得到问题的解答.(2)串行法:将专家系巍租神经阿络串联相接来求解问题.例如:专家系统1用于帮助神经网络进行训练及复杂的人机交流;神经网络用来进行决策和问题求解;专家系统2用来解释神经网络的输出结果,并驱动有关执行机构.上述两种方法根据被求解问题的需要把系统分为若干个模块?每个模块分别用专家系统或人工神经网络技术实现.这两种方法通称为模块相接法或集成法.其次,按两者的地位分为:(1)专家系统为主,神经网络为辅(见图1).专家系统在必要时调用神经网络文件.例如嵌人法,即在专家系统内嵌人神经网络,用于执行在专家系统周期中耗费时间最多的工柞模式匹配,以加快专家系统的执行速度.(2)神经网络为主,专家系统为辅(见图2).神经网络在必要时调用专家系统文件,由专家系统给出解释.进行界面臂理.例如功能模拟法,神经网络模拟专家系统来实现某种功能,以追求系统性能的改善.图1Bs为主的结构图2ANN为主的结构图3两院制结构此外,还有指导式和两院制结构等.其中,两院制结构(见图3)将使Bs和ANN两种形式的知识可以共事.虽难以实现却最具发展前景.所谓两院制结构.就是在整个系统中.大多数知识同时以神经网络和符号形式两种方式表示,每部分以各自独特的推理机制工柞.岿要时可从一种形式中抽取知识并将其转化成另一种形式.实质上两种形式的知识是共事的.例如用神经网络构造一个符号化模型.～Bs和ANN的结合在具体应用时,可以不拘一格,将上述各种方式混合运用.以便更挥此种方法的优越性,实现更多的功能.本文后面实例中对神经网络和专家系统的结合方式进行了初步探索.总体上将神经网络嵌人到专家系统中,具体诊断推理时主要采取两者的串型或并型等连接方式.4应用实例:故障诊断系统4l基于神经网络的故障诊断专家系统神经网络与专家系统技术相结合比较适用于故障诊断.基于神经网络的故障诊断专家系统,将利用神经网络的自学习功能联想记忆功能和分布式并行信息处理功能等来解决诊断系统的知识表述,知识获取和并行推理等问题.神经网络与专家系统的集成可以发挥各自的优势.非常适合于表达故障诊断及处理系统的知识.48黑龙江八一农垦大学第10卷该系统的知识表述分两种:一种是将专家经验形式化成规则,并存储于知识库中:另一种是通过现场历史数据对神经网络进行训练,将难以形式化的专家经验以非线性映射的形式存储于神经网络的结点上,由协调机构针对不厨情况用规则和神经网络对系统故障进行诊断,得出相应的诊断结果.神经网络系统在完成一个诊断实例之后,可以记忆诊断过程和结果,从而归纳出新的诊断规则,不断扩充知识库的内容,使知识库具有自学习功能,这是本系统与普通诊断系统的重要区别.系统的推理主要包括ANN的浅层经验知识推理和Es的深层逻辑知识推理.ANN采用数据驱动的正向推理策略,从韧始状态出发,向前推理,到目标状态为止.这种推理方式对同一层处理单元来说是并行的,不需要进行规则的前提匹配,克服了传统推理中的匹配冲突等困难.这种推理过程只与网络自身参数有关,其参数可通过学习算法进行自适应训练,因此具有自适应能力.4-2发动机故障诊断系统的螭构特点笔者研审I了一种用于汽车发动机的故障诊断系统,采用了神经网络与专家系统相结合的方法.一般地,神经网络用于对故障进行分类,给出韧步诊断结果,专家系统通过人机对话进行推理.最后给具体诊断结果并解释诊断过程,用户通过人机界面对系统进行操作和管理.系统总体上采用神经网络嵌人的方式,在具体的子模块中包含很多个神经网络和专家系统文件,根据要实现的不同功能要求,分别采用神经网络与专家系统的串受,并受或混合型等方式连接.具体解决某一问题时.系统各子模块有些以专家系统为主,也有些以神经网络为主,更多的情况是将两者有机结合来进行混合求解;有些子模块中利用神经网络和专家系统可以分别求解.供用户参考选择,再通过人机对话确定最后结果.4.3典垂的诊断过程诊断系统子模块的典型结构如图4所示,采用串型连接方式,将Es和^NN两者结合运用.其中,专家系统1用来进行复杂的人机交流;神经网络1用于问题求解;专家系统2用来解释神经网络的输出结果,并进一步推理,得出具体诊断结果.实时专家围4诊断系统子模块的典型螭构/第2期孙永厚荨:神经网培与专i隧统些堕鱼垦基堕旦塞!!以*发动机内部机械一故障为倒简介其诊断过程.判断汽车发动机内部机械部分有无故障最简便的方法就是测量各汽缸压缩终了压力利用神经网络分析这些数据与正常相比偏高或偏低,从而对其进行故障分类.再由专家系统推断出相应故障原因,给出诊断结果.对来自接口由传感器测出的汽缸压力数据值.由镦机内部进行分析处理,井进入内部机械台勺相应子模块.该模块中首先由神经网络进行计算.得出故障分类结果.再进入专家系统中进一步推理.专家系统首先解释神经阿络的输出结果:(例如)某一汽缸压力偏高进一步诊断推理(人机对话):(屏幕显示)请问:行驶中还出现过热或突爆吗?(用户选择)回答:YBs{回车)(弹出窗口)诊断结果:积炭过多或经几次修理后压缩比有了变化,请及时修整15结束语根据要实现的不同功能要求,将神经网络与专家系统结合时可以采用多种方式,如串型,并型或混台型连接等等.这些结合方式各具特色,可以充分发挥神经网络和专家系统各自的优点.从而组合成各种薪型的智能化实用系统.采用神经用络和专家系统相结合构造新型的神经网络专家系统,是智能系统发展的必然趋势.神经网络方法模拟了人类的形象思维,是一种非逻辑,非语言,非静态,非局域和非线性的信息处理方法.它与传统人工智能方法不是简单取代而是互为扑充,辩证统一的关系;此种方法与专家系统结合的发展和应用将给人工智能,计算机科学与信息科学荨领域带来历史性的变革.参考文献1蔡自必等人工智能及其应用北京:清华大学出版杜.1998.6—112张际先.盛霞神经阿培及其在工程中的应用北京:机械工业出版社.1996.1—193衰泉.何募荨专家幕境与神经罔终结合的油机故障诊断系统.中国农业大学.l998.a(2)4戚扬.韩北山汽车教障诊断北京:人民交通出版社.1988383—38689一g2STUDY oNTHEeOMBINA TIONoFNEURALNETWORK WITHEXPERTSYSTEMANDITSAPPLICATIONSSunY onghouYuanQuanWangZhiminABSTRACT:Inthispaper.6omccoⅢmonwaystocombineArtificialNeuralNetwork(ANN)withExpertSystem(E8)areprovided.basedoaanalysingthebenefitsofthe combination.Thetypicalapplicationinfaultdiagnosticexpertsystembasedon ANNisindicated. Thepracticeofthesecombiningwayispresentedbyanexample aboutenginefaultdiagnostic.Keywords:Neuralnetwork~ExpertsystemIFaultdiagnostic。
