【作者】 张海军；

【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， 2007， 博士

【摘要】 价值高昂、结构复杂的航空发动机构成了飞机的心脏。为了有效地保障其规定的高可靠性、安全性和经济性,国内、外航空公司已普遍采用经济、合理的视情维修(Condition-based maintenance, CBM)策略。然而,由于民航发动机各单元体/子系统之间的复杂性、运行监控信息的不确定性以及维修阈值的模糊性,航空公司当前在发动机运行管理与CBM决策过程中还存在着一些亟待解决的问题。本文基于这一工程背景,结合民航总局重点基金项目“民航发动机性能预测与维修成本控制研究”,利用多元参数建模思想,对发动机随机劣化过程中存在的性能综合评估、随机失效过程建模、换发优化决策及送修等级决策等CBM决策方面的关键和核心问题进行了深入的研究,并经相关航空公司的实证分析,表明所提出的理论建模方法是有效、合理的,具有很大的实用价值。主要研究内容如下:(1)鉴于目前工程实践中所用的单参数评估方法存在的诸多不足,根据客观赋权法的思想,采用粗糙集与模糊数学理论相结合的模糊信息熵权法研究了基于多参数的发动机性能综合评估问题,拓展了传统的综合评估方法;在工程实例计算中通过将模糊信息熵权法与单参数评估法、主成分分析法等传统评估方法加以对比,表明该方法勿需任何先验知识、决策过程中损失的客观信息较少,因此通过该方法对发动机性能的综合评估,可以更加准确地把握发动机机队的运营状况。该方法目前已在国内航空公司得到初步应用,发动机维护工程师经过多个工程实例验证了该方法的有效性。(2)性能衰退是导致民航发动机换发的重要原因,但是在实际中该衰退过程的随机性增加了实际维修决策的难度。因此,针对发动机随机失效过程的维修决策建模问题,首次提出了将发动机性能衰退导致换发的在翼时间视为寿命数据进行生存分析的建模思想。文中分析了数据的截尾机制、协变量相关性及筛选基本原则等问题;进而研究了生存分析中广泛应用的比例强度模型(Proportional intensity model,PIM)的构成特点;在此基础上,综合考虑状态参数、维修历史等协变量对系统的影响效应,提出了应用于复杂系统的扩展型PIM(Gerenalized PIM,GPIM),并详细推导了该模型在全参数和半参数两类情形下的参数估计的求解过程。(3)民航发动机运行过程中执行CBM策略的核心内容是换发决策,优化目标是在满足规定可靠性的前提下实现维修成本的最小化。首先研究了状态、维修等协变量对发动机系统劣化的影响效应,并采用GPIM求解出系统的失效强度函数和可靠性函数;其次,以长期的平均维修成本为目标函数进行维修阈值的优化,以CF6型发动机的历史故障数据为例,得到了相应的发动机换发决策控制限曲线,从而为航空公司实际维修决策提供了准确的量化指标。该类建模方法还可以推广应用到其他型号发动机的维修决策中,因此具有重要的实用价值。(4)当在翼发动机整机性能衰退、需要立即换发送修时,对送修等级的决策是航空公司目前面临的一个难题。基于民航发动机的状态监控信息,提出了采用变精度粗糙集(Variable Precision Rough Set,VPRS)理论方法来挖掘反映送修等级和状态参数之间内在关联性的决策规则。其中,针对连续型条件属性数据的离散化问题,本文利用自组织神经网络和不相容评估相结合的方法来处理连续数据的离散化问题。通过实例验证表明:由该方法所获取的知识规则为发动机维护工程师制定合理的送修等级提供了积极的辅助作用,因而具有较大的参考价值。更多还原

【Abstract】 Aero-engine, a valuable and complex systems in aviation industry, is the heart of an aircraft. In order to ensure the better reliability, safety and economy of aero-engine, condition-based maintenance (CBM) policy is populated for airlines in the world. However, for the complexity among the modules or subsystems in the engine, the uncertainty of operation monitoring information, and the fuzziness of maintenance threshold, there are some problems needed to be solved between the engine operation management and maintenance decision-making. Based on this engineering background, combining the key fund project“Study on predicting performance and optimalizing maintenance cost of aeroengine”of civil aviation administration, the thesis deeply discussed the key problem of synthetic performance estimation, modeling of stochastic failure, decision-making of engine removal and forecasting of maintenance levels, which were appeared in the engine stochastic deterioration process, based on the theory of multi-parameters modeling. And these models and methods were proved to be very effective, reasonable, and valuable by some practical examples in airlines. The main contents of this dissertation are as follow:(1) According to the various shortages of previous single-parameter assessment method for engines, the feasibility of applying fuzzy information entropy weights method, which combines rough theory and fuzzy math theory, was studied in the thesis, based on the subjectively weighted method and multi-status parameters. In the example, after comparing with the traditional ways of single-parameter assessment and principle component analysis, fuzzy information entropy weights method showed satisfied result, which needed no prior knowledge with less lose of objective information. After estimating the engine performance synthetically, people can accurately know the health of the whole engine fleet. And this method, which extended the traditional synthetic estimation theory, has been initially applied in airlines, and is proved to be very effective by some engineering examples.(2) Performance deterioration is an important reason of engine removals. But since the randomness of performance deterioration increases the difficulty of maintenance decision-making, it’s significative to study the modeling of maintenance decision-making considering engine stochastic failure process. According to the difference between classical hypothesis and practice conditions in modeling of CBM decision-making, the engine times on wing (TOW) due to performance deterioration were looked as life data creatively. And the coherent censoring theory, the correlation of covariats, and basic screening principle were analysed. Furthermore, the characteristics of proportional intensity model (PIM) were studied, which was populated in survival data analysis. Based on thought of PIM, a generalized PIM was put forward, which considered the affects of condition monitoring parameters, maintenance history, etc, and could be applied in complex systems. The resolve process of parameters estimation in this model was detailedly deduced too.(3) The key part of CBM policy during commercial aero-engine operation is engine removal decision-making, whose optimization object is to minimize the maintenance cost under some reliability requirements. Firstly, in the case of fleet’s history failure data of CF6 type engine, the generalized PIM was established, which studied the affects of covariates such as condition parameters, maintenance effect, and so on. Then the maintenance threshold was optimized under the object function of long-term average maintenance cost, and the engine removal decision-making control curve was created, which could be an accurate factor of engine maintenance decision-making in airlines. This modeling theory can be generalized to the maintenance decision-making of other type engines, and has important practising value.(4) When the engine on wing is deteriorating greatly and need to remove immediately, the estimation of maintenance levels is very important for airlines. Based on the engine condition monitoring information, a method about variable precision rough Set (VPRS), was introduced to mine the relevance and decision-making rules between maintenance levels and condition parameters. According to the problem of discretization for continuous condition attributes, a discretization method combining self-organization maps(SOM)neural network algorithm and incompatibility estimation was designed. And the certification of some examples showed that this method had obvious dominance, could control the maintenance cost effectively, and had good value for practice.更多还原