【作者】 周虹；

【导师】 左洪福；

【作者基本信息】 南京航空航天大学 ， 载运工具运用工程， 2013， 博士

【摘要】 随着国内民机产业的飞速发展，飞机故障诊断和故障风险评估是亟待解决的问题。对于民用飞机尤其是新研制飞机的各种疑难故障,由于其引发因素繁多且相互耦合,故障征兆与最终故障原因之间无明确对应关系，因而在诊断过程中，存在对象的精确定量模型难以建立，同时先验知识又匮乏的问题。本文以飞机多工况系统为研究对象，提出了一种新的定性描述模型—过程动态有向图TMSDG (Test-Maintenance Signed Directed Graph)。一方面充分利用符号有向图(signed directedGraph，sDG)模型包容深层知识的能力和处理复杂因果关系独特的优势，另外一方面弥补SDG在诊断诊断和维修领域的知识表示和处理方面的不足，突破SDG实际应用的局限性，从而为复杂系统的故障诊断与故障风险分析提供有效支持。论文主要研究四个问题：(1)给出了TMSDG模型的定义。TMSDG模型将SDG模型覆盖在系统结构模型上，增加描述测试与状态参数的关联关系，进一步引入了工况条件。TMSDG模型不但继承了SDG模型的特性，更重要的是包容结构、状态、测试、维修更多潜在信息,因而不仅具有自动诊断和而且还为交互诊断能力提供了支持，并满足多工况过程动态系统模型随工况变化调整的实际需求。根据复杂系统故障在层间传播的特点，讨论了一种基于分层思想的建模方法，降低系统级TMSDG模型建模的难度，使模型具有良好的封装性、可重用性，也使模型具备分层诊断能力，适应复杂系统诊断和民用航空分级维修体制。(2)研究了TMSDG“分而治之”的分层诊断策略。基于TMSDG的诊断推理，首先确定发生故障的具体工况，重构单稳定工况SDG模型。在单稳定工况诊断过程中引入故障--故障关联矩阵，使用故障--故障关联矩阵和可达性理论扩充了相容性理论，进而计算可能的故障源集合。提出的TMSDG分层诊断策略减小故障源搜索空间的大小，避免了未测节点的符号假设、停止假设的判断和修正假设符号的过程，大大提高了诊断算法的效率和诊断结果的准确性。最后，TMSDG诊断策略综合多个工况诊断结果，进一步减少冗余解，提高诊断精度。(3)提出了一种多值AND/OR图描述飞机系统测试序列问题，并研究了基于最大单位信息熵故障诊断树的交互式诊断算法，充分利用了故障概率信息、测试结果及测试费用，得到了具有最大增益测试费用比的测试序列。该算法不用列举整个结果树的集合，运算中不存在反馈，计算结果用多值AND/OR描述,可有效减少计算复杂度和平均测试代价,而且生成的故障诊断策略能有效、快速地指导维修人员进行故障隔离。提出的方法能自适应不同维修级别不同的故障隔离精度要求，生成相应的测试序列。(4)建立了基于灰色聚类方法的故障风险评估方法。对于已识别的故障模式,通过TMSDG模型获取故障因果链和故障树，进行故障劣化趋势预测,明确故障模式的发展变化过程，以故障因果链为分析对象实现整体和部分相结合的故障风险评估。该方法突破了以往FMEA方法的局限，解决了风险评估过程中的不确定性和不一致性问题，从而更为准确地评估故障风险，其分析结果不仅能将故障模式按照风险顺序排列，还详细阐述了具有实用意义的中间过程，能更有效地帮助决策者分析制定维修策略及改进措施，增强在整个寿命周期中由阶段消除、控制这些危险的能力。通过理论分析和气源系统诊断系统实例研究表明，基于TMSDG的故障诊断方法能有效地应用于民用飞机多过程故障诊断。TMSDG的故障诊断方法通用性良好，论文研究结果对于复杂系统的故障诊断与故障风险分析具有重要的参考意义。更多还原

【Abstract】 When fault diagnosis is conducted in system of Civil aircraft，especially in the newlydeveloped aircraft，There are many complex faults with multiple coupling influencing factors, andsymptoms. It is difficult to build accurate quantitative models and lack of prior knowledge.Tosolve these problems and conduct effective fault isolation of ircraft multi-condition process thispaper proposed A new qualitative description model—process dynamic signed directed graph(TMSDG) based on SDG which has the ability to accommodate large-scale deep knowledge andhandle complicated cause-effect relationship, Four Problem are studied as below:(1)The definition of TMSDG model is given. TMSDG model is built by combining SDGmodel to the system structure, introducing the stage conditions, unmeasured nodes in SDG.TMSDG not only inherits the characteristics of SDG model but also has the capacity of containinglarge-scale potential information such as structure, test and maintenance. which can support bothautomatic diagnosis and interactive diagnosis.TMSDG can meet the actual needs of change ofoperation conditions. According the characteristics of transmission between hierarchies of failuremodes a hierarchical modeling method which can make modeling easier is studied. the TMSDGmodel has good encapsulation, reusability, so it can support complex system diagnosis and civilaviation maintenance system(2)A hierarchical diagnostic strategy based on TMSDG is studied．First, the specificconditions of failure are determined and the corresponding single-state stable SDG models arereconstructed. Then, possible fault sources are reasoned based on single-state stable SDG modelby introducing the fault-fault dependence matrix, with accessibility theory and the compatibilitycalculation principle. Last,The final diagnosis algorithm integrated multiple modes of thediagnosis. Hierarchical diagnostic strategy avoids the symbols assumption and symbols correctionof unmeasurement nodes, greatly improving the efficiency of diagnosis and diagnosis accuracy.The further reduce redundancy solution, and to improve the diagnostic accuracy.(3)A multi-value AND/OR graph is presented to describe the test sequencing problem ofaircraft system and a diagnosis tree generation algorithm based on maximum unit informationentropy is studied. In the given algorithm,the complicated status association and fault-testdependency of complex system fault are modeled and multi-value AND/OR graph with maximuminformation gain per unit cost of the test is developed considering fault probability and testing cost.The AND/OR graph is a fault diagnosis strategy which can effectively help maintenance man tolocate the fault at lower maintenance costs. The generated diagnostic strategy ensures faultisolation with desired precision at desired maintenance level，which meets the requirements ofcivil aircraft maintenance system．(4)A fault risk assessment algorithm based on gray clustering is proposed．For theidentified failure modes, the deterioration trend are predicted by isolating the event tree and faulttree from TMSDG model. This method improves accuracy of risk valuation since the event treesare regarded as risk evaluation objects. It solves the uncertainties problem and inconsistenciesproblem in traditional FMEA risk assessment. The assessment results includes not only riskpriority of fault mode, but also more practical detailed analysis, which can help to set upmaintenance strategy, improvement measures and enhance the ability of eliminating or control ingthese dangerous in the whole life cycle. Finally, Based on theoretical analysis，an intelligent Fault diagnosis system is developed andan aircraft bleed system is employed as an object to validate the proposed method. The case resultsproves the effectivity and utility of this model.更多还原