【作者】 杨刚；

【导师】 汪道辉；

【作者基本信息】 四川大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 随着各种新的应用出现，数字系统的集成度不断攀升，为满足这些数字系统的故障检测需求，出现了许多智能故障检测和诊断的方法，然而在将这些方法应用到实际的故障诊断系统中后，遇到了一些难题，其中如何区分并解决多线桥接故障和固定逻辑故障之间的征兆混淆问题一直是智能故障诊断系统设计的瓶颈，为解决这一难题，本文深入研究目前国内外最新取得的一些理论成果，如布尔矩阵理论、故障检测和紧凑性定理、故障诊断完备性定理和抗误判定理，通过这些研究为本文的撰写提供了坚实的理论依据。 本文作者还对测试优化算法进行仔细研究，研究结果表明测试优化算法可分为两类，第一类是在满足给定故障定位精度的前提下寻找能够实现最优紧凑性指标P的PTVs(并行测试向量集)，如True／Compliment算法；第二类是在满足给定紧凑性指标的前提下寻求能够实现最优故障定位精度的PTVs，如极小权值优化算法，研究结果还显示这些常规的算法存在测试时间长、故障覆盖率低的缺点。因此，本文提出一种属于第一类的新型测试优化算法：NAAC算法(No Aliasing and Anti-Confounding Algorithm：无误判抗混淆算法)，并将其应用于实际的智能故障诊断系统中。采用NAAC算法生成的测试矩阵能够确保无征兆误判和无征兆混淆，同时该算法还能将大的测试矩阵压缩成小规模的测试矩阵。在其生成的测试矩阵中各PTV互不相同，但都具有相同的权值，而权值可以由故障诊断系统的设计者予以指定。当这些PTVs从PC机中依次加载到被测电路板后，相应的PRVs(并行响应向量集)可以获取供诊断系统进行测试响应分析，从而检测出固定逻辑故障和桥接故障。实践表明，NAAC算法生成的测试矩阵具有较高的故障覆盖率。 本文还讨论了两种智能故障诊断技术，即专家系统智能故障诊断技术和边界扫描测试智能故障诊断技术。对于专家系统智能故障诊断技术而言，构建专四川大学硕士学位论文家系统规则库和设计推理机制是重点，本文采用C语言构建规则库，然后通过精心设计的专家推理机，从而使得固定逻辑故障和大多数多线桥接故障能够被专家诊断系统检测出来。与此同时，本文采用C语言构建规则库的方法也为其他智能故障诊断系统的设计提供一个良好的范例。至于本文采用边界扫描测试故障诊断技术，是考虑到本系统的通用性和简洁性，因为对于大多数数字系统而言，具有边界扫描结构的器件己广泛应用，本文只需4条或5条信号线就能将PC机和被测边界扫描电路连接起来，由此极大地简化了智能故障诊断系统中为实现PTVs加载和PRVs获取而专门设计的接口板电路。 在智能故障诊断系统中，虚拟仪器技术是关键，本文采用Delphi 6.0设计本系统虚拟仪器界面和驱动程序，整个系统按功能划分模块，符合软件工程设计要求。系统中一些模块采用MCS一51汇编语言和VHDL硬件描述语言编写，因而涉及到单片机、EDA等新技术。此外由于采用强大的GUI界面，使得该系统操作简单且易于移植到其他应用中。 本文精心设计该诊断系统的硬件接口板，使其能够检测和诊断MCU、PLD、存储器、数据传输和接口器件、工/0口，而这些都是数字系统的常见部件，故该系统有良好的通用性。 此外本文还探讨自适应测试及其算法，并结合NAAC算法提出了自适应NAAC算法。同时考虑到将来的工作，本文也探讨了第二类测试优化算法:极小权值—极大相异性算法，该算法的故障检测能力与NAAC算法相似，但生成测试矩阵的步骤较复杂。限于时间，作者仅就基本原理做了阐述，更多的工作留待以后完成。更多还原

【Abstract】 Driven by some emerging applications, digital system integration is on the rise. To serve digital system fault detection requirements, a lot of intelligent fault detection and diagnostic methods have been found. However, in most of fault diagnosis system, it is a difficult thing to classify multi-line short circuit fault from fixed logic fault. To solve the puzzle, the paper author widely study in Boolean Matrix theories , fault detection and condensing theorem, fault diagnostic soundness theorem and anti-aliasing theorem. In addition, the author also research in general detection optimal algorithm used by fault diagnosis system, these detection optimal algorithm can classify into two types, the first type is that algorithm must be able to find the PTVs (Parallel Test Vector)which has the best condensing index P under fault location precision request, these algorithm include WOA optimal algorithm, True/Compliment optimal algorithm; The second type is that algorithm must be able to find the PTVs which has the best fault location precision under the condensin index P request. These algorithms include minimum weight value optimal algorithm. But these general detection optimal algorithms have a lot of flaw in test time cost or low fault coverage rate. Therefore, the paper brings up a new detection optimal algorithm that belongs to the first type: NAAC (No Aliasing and Anti-Confounding) detection optimal algorithm, and put the NAAC detection optimal algorithm into a practical intelligent fault diagnosis system. The NAAC detection optimal algorithm can assure to generate test matrix with no symptom aliasing and no symptom confounding, at the same time, it can compress a large of scale test matrix to verysmall of scale test matrix. The PTV of test matrix is different for each other, but has same weight value. And the weight value can be appointed by fault diagnosis system designer. When those PTVs are sequentially loaded to CUT (Circuit Under Test) from intelligent fault diagnosis system, homologous PRVs (Parallel Response Vector) are taken back. Through test response analysis by expert system , Intelligent fault diagnosis system can detect all fixed’1’logic fault, fixed’0’logic fault, and detect the majority of multi-line short circuit fault. Practical analysis and experimental results show that the NAAC detection optimal algorithm is characterized by almost full error coverage.The paper also discusses two intelligent fault diagnostic methods: expert system intelligent fault diagnostic method and boundary-scan technique intelligent fault diagnostic method. Benefit from designer experience, Expert system regulations are built by C language. Through the inference, all fixed logic fault and most multi-line short circuit fault are detected by expert system. On that account, the design process built by C language will be a mode that can be imitated by other intelligent fault diagnosis system. On the other hand, Boundary-scan technique intelligent fault diagnostic method was applied to practice. For most digital system, devices with boundary-scan architecture are broadly used. Only using four line or five line to connect PC parallel port with CUT TAP (Test Access Port), all the PTVs can be loaded to CUT and all homologous PRVs can be taken back to intelligent fault diagnosis system. The method extremely simplifies diagnosis system design. .In this intelligent fault diagnosis system, virtual instrument technology is a key point. "Software is instrument" isn’t a slogan but a fact. The author designs diagnosis system software with Delphi 6.0 , the whole system are divided according to functional block. Some of the blocks are written by MCS-51 ASM language and VHDL hardware description language. With the powerful GUI (Graph User Interface), this diagnosis system is ease to use, and can be conveniently transplanted into other applications.In the hardware part of the intelligent fault diagnosis system, a circuit board for interface is design meticulously by author, the interface board connects PC with CUT, all l更多还原