【作者】 汤新民；

【导师】 钟诗胜；

【作者基本信息】 哈尔滨工业大学 ， 机械设计及理论， 2007， 博士

【摘要】 与离散制造车间不同,航空发动机维修车间具有维修等级的不确定性、发动机构型的动态性、维修时间的随机性以及过程数据采集困难和维修设备柔性差等特点。为提高维修车间作业的自动化和决策科学化水平,本文采用基于Petri网的离散事件动态系统建模理论,开展了航空发动机维修车间逻辑层次、时间层次和统计层次模型的建模与应用研究。在逻辑层次建模方面,针对发动机维修等级推理中的不确定性,建立基于模糊Petri网的发动机故障评估、性能评估知识表达模型,研究知识库的定性和定量推理算法。为使等级决策知识库系统具有自适应能力,提出了基于元模型的发动机性能评估知识库学习算法。为实现发动机分解装配工艺制定的自动化,采用基于优先约束关系的装配Petri子网对航空发动机部件的装配序列进行了建模,并采用离散时间最小值原理,将最优装配序列规划和选择性序列规划问题分别转化为固定端点和自由端点的最优控制问题,降低了计算复杂度。在时间层次建模方面,为识别与预测维修工作流时间计划可能存在的时间违反问题,提出了面向维修任务分解结构的“自顶向下”层次细化工作流建模方法,分析了层次细化后保持原网可调度性不变的等价变换原则,在此原则下给出了基于极大极小代数的工作流执行时间矩阵验证算法。针对发动机总装作业中不同维修工艺对设备资源的需求冲突,进行了面向资源冲突的Petri网建模。将离散模型转化为连续模型,并对该模型的周期性和稳态性进行了分析。还采用粒子群优化算法以资源最大利用率为目标对资源路由策略进行了优化。在统计层次方面,采用扩展随机Petri网对发动机在维修车间的转移过程进行了建模,并引入随机Petri网平均值分析算法,利用单维修等级下系统的性能指标对多维修等级性能指标进行修正,还采用平方变差系数方法实现服务时间从指数分布向正态分布的逼近,进而实现对多等级、服务时间为正态分布的车间维修能力进行评价,找出影响维修能力的主要环节。最后,在上述理论研究的基础上,完成了航空发动机计算机辅助维修系统的建立。该系统包含维修等级推理、维修工作流分析验证、发动机构型及零部件跟踪、维修资源调度和维修系统能力评价五个模块。在北京飞机维修工程有限公司的应用效果表明,该系统提高了维修决策与分析的智能化水平,降低了航空维修差错的可能性,提高了维修资源的利用率,缩短了发动机的维修周期。更多还原

【Abstract】 Different from discrete manufacturing system, aero-engine workshop maintenance procedure is nondeterministic, aero-engine configuration is dynamic, maintenance procedure data acquisition is difficult, maintenance equipments lack flexibility, maintenance time distribute randomly. To improve automatization of production and rationalization of decision for maintenance workshop, Petri nets is introduce to logic, temporal and statistical layers modeling respectively.As for logic layer modeling, the maintenance level decision knowledge base and disassembly and assembly sequence model is proposed. To deal with uncertainty of aero-engine maintenance level decision, a fault and performance evaluation knowledge representation model based on fuzzy Petri nets (FPN) is proposed. Qualitative and quantitative reasoning algorithm of knowledge base is researched. To make decision knowledge base more adaptive, learning algorithm for performance evaluation knowledge base based on metal fuzzy Petri nets (MFPN) model is presented. To eliminate maintenance errors during assembly and construct disassembly process automatically, assembly sub Petri Nets is used to modeling aero-engine parts assembly sequence (PAS) based on assembly precedence relationship. Discrete-time Pontryagin’s minimum principle (DTPMP) is introduced to find OFS and DTPMP states that an OFS must minimize the Hamiltonian function which can be treated as the heuristic information to find the PAS. After complexity analysis, the OFS algorithm complexity proved to be polynomial.As for logic layer modeling, the maintenance workflow and resource scheduling modeling is proposed. To identify and forecast time consistency, the hierarchical refinement workflow net (HRWF-net) is introduced to“top-down”aero-engine maintenance break-down structure workflow modeling. An equivalent transformation principle which keeps the schedulability of HRWF-net unaltered is analyzed. To verify the executive time of maintenance workflow, a matrix verifying algorithm based on max-min algebra is proposed. Because of the resource contention among different processes, a theory that synthesized resource sharing Petri net is introduced. To optimize the schedule performance of the discrete Petri net model, discrete model is transformed into determinate time continuous Petri nets (DTCPN) and whose stationary and periodic behavior is analyzed. Particle swarm optimization is used to optimize routing strategy. Furthermore, Dynamic factors affect scheduling is considered and the event-driven rescheduling strategy is presented.As for statistical layer modeling, the aero-engine transferring model in workshop based on extended stochastic Petri Nets is built and mean value analysis technique for product form stochastic Petri Nets is introduced. To deal with normal distribution of maintenance time and diversity of maintenance level, the approximate mean value analysis(AMVA) applied to extend stochastic Petri Nets is introduced, which evaluate multi-level maintenance service performance indexes using single level performance indexes. In addition, the sojourn time is divided to three parts and squared coefficient of variation is used to correct service time with normal distribution. It’s proved that AMVA is an efficient method evaluating air-engine fleet maintenance capacity for maintenance workshop.Based on modeling for aero-engine workshop, Computer aided maintenance system (CAMS) is developed which includes five modules: maintenance level reason and decision, assembly and disassembly sequence planning, maintenance workflow modeling and procedure supervising, maintenance history tracing and performance evaluating. Applied to Beijing Aircraft Maintenance Engineering Ltd. Co, it is proved that CAMS can aid maintenance workshop transferring information automatically, making decision more intelligentized, reducing maintenance errors, improving maintenance resource utilization and cutting down on maintenance cost.更多还原