【作者】 董学军；

【导师】 陈英武；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2012， 博士

【摘要】 航天器发射是航天工程建造的最后阶段，其可靠性与灾难评估在避免工程灾难中处于重要地位。近年来，航天产品可靠性不断改善，但由于航天工程固有的开拓性和复杂性，工程灾难时有发生。当前，航天器发射工程可靠性与灾难评估仍处于成长阶段，关于全系统的灾难评估文献报道较少，尤其是关于人因、组织和过程可靠性的研究成果几乎检索不到，因此，综合利用相关学科研究的新成果，结合航天器发射工程技术和管理的特点，研究适合工程本身的理论和方法，构建能全面融各类灾难症候的可靠性和灾难评估模型具有重大理论和现实意义。本文在总结分析国内外相关研究成果的基础上，以工程实践为背景，结合数据仿真结果，构建并验证了人因可靠性、组织可靠性、过程可靠性和工程灾难评估模型，主要工作、创新与结论如下：（1）在人因可靠性和组织可靠性的论域上存在着不可替代因素和补偿因素，通过集成这两类因素的状态可以达到考察人因可靠性和组织可靠性的目的。基于航天器发射工程实践数据和知识，定义补偿和不可替代因素，提出补偿和不可替代因素状态合成方法（the Composition for State of Irreplaceable and CompensatoryFactors，CSICF），讨论航天器发射工程人因和组织可靠性的因素空间，构建人因可靠性和组织可靠性模型，并用X-51发射工程飞控指令发送活动的数据分析检验模型的可行性与有效性，初步解决了航天器发射工程人因与组织可靠性量化评估问题。（2）从航天产品进场至运载器点火发射，航天器发射工程各子过程的状态变化呈现出Markov性，且状态转移概率只与系统在状态上的停留时间相关，因此，可以使用连续时间马尔科夫链（Continuous-time Markov Chain，CTMC）来描述航天器发射工程组织过程。在综合考察航天产品、地面设备和工程组织的任务可靠性基础上，依据航天器发射组织过程约束关系，构建航天器发射工程组织过程的CTMC模型并给出其性能参数的计算方法。利用互模拟等价关系（BisimulationEquivalence Relation，BER），对航天器发射组织过程的多个CTMC并发模型进行状态空间压缩，提出工程组织全过程的可靠性计算模型，并通过对X-51任务计划的分析，进一步说明所提模型的使用方法和实际用途。（3）对航天器发射工程灾难症候的有效获取和综合利用，可以提高工程灾难评估的准确性，降低灾难预测的模糊性。依据航天器发射工程各系统间的依存关系构建故障依存关系树（Failure Dependency Tree，FDT），并以此为框架，分别利用证据理论和Bayesian网络推理讨论症候信息融合和症候信息推理两类方法，构建航天器发射工程灾难评估模型。基于证据理论的症候信息融合方法可使不同领域专家采用不同观察手段所获取的各种局部的、不完整、不确定的灾难症候相互补充、消除冗余，达到降低不确定性目的。基于Bayesian网络的症候信息推理方法可有效避免症候信息融合方法易忽略少数正确意见的弊端，二者结合使用可以有效保证航天器发射工程灾难评估导向清晰、正确的结论。STS-51L灾难评估的实例分析，进一步验证了本文所提灾难评估模型的有效性。更多还原

【Abstract】 The reliability and disaster assessment of spacecraft launch which is in the finalstage of the construction of aerospace engineering plays an important role in avoidingengineering disaster. In recent years, the reliability of space products continues to beimproved，but the natural pioneering and complexity of aerospace engineering result indisasters from time to time. Moreover, the study of reliability and disaster assessment ofspacecraft launch engineering (SLE) is still under growth, and we have yet to seeconcrete results from the research for human, organization and process reliability.Therefor, comprehensive using recent research results from related disciplines,researching theory and method for the spacecraft launch engineering, establishingmodel with all-round combining symptoms of extreme events in order to effectivelyestimate the reliability and disaster in a SLE is viteal important in theory and reality.On the basis of the summary and analysis of relevant research results at home andabroad, considering the engineer practice background, associating the simulation data,this paper establishs and verifies the human reliability, the organization reliability, thescheduler reliability and the engineering disaster assessment model for SLE. The mainwork, contribution and conclusion are as follows:(1) There exist irreplaceable and compensatory facts in the discourse domain ofhuman reliability and organization reliability. The purpose of examining human andorganization reliability by the composition of state of irreplaceable and compensatoryfactors can be attained. Based on field data and knowledge on SLE practice, this paperdiscusses the factor space of human and organization reliability in SLE, and proposesthe method of the composition of state of irreplaceable and compensatory factors(CSICF), and establishes assessment models of human and organization reliability inSLE, and validates the effectiveness and feasibility of the proposed method through theworking team reliability assessment for the flight control command delivery activity inthe mission XL-51, and initially solves the methodological issues for how toquantitatively assess human and organization reliability in SLE.(2) From aerospace products into the launch site to the vehicle ignition, Changesabout the states of sub-processes present the Markov property and state transfer ratesonly relate to residence time of system in the state. Therefor, we can applying the modelof continuous-time Markov chain (CTMC) to simulate organization processes of a SLE.On the basis of the comprehensive study the mission reliability of aerospace productsand ground equipments and organization reliability, according to the constraintrelationship between sub-processes, this paper constructs the model of CTMC fororganization processes of SLE and gives the calculation method of the model parameters.Whereafter, the paper discusses the problems of how to compress the state space of multi-CTMC concurrency model by using bisimulation equivalence relation (BER), andpresents a model of how to calculate the reliability of the engineering organizationoverall process. Finally, through the X-51mission schedul analysis, the paper furtherillustrates that the model is of practical use and method of use.(3) Effective access and comprehensive utillization of disaster symptoms in SLEcan improve the accurancy of the disaster assessment and reduce the fuzziness of thedisaster forecast. According to the dependencies between systems in SLE, this paperdesigns failure dependency tree (FDT). And to take the FDT as a framework, the paperdiscusses the symptom information fusion method (SIFM) based on evidence theory,and studies the symptom network inference method (SNIM) by using Bayesian networkinference, and constructs the disaster assessment model of SLE. The SIFM based onevidence theory can make all kinds of partial, incomplete, uncertain disaster symptomsacquired by experts in different fields with different methods mutual compensation, toeliminate redundant, to reduce uncertainty. The SNIM based on Bayesian networkinference can effectively avoid the drawback that the SIFM easily leads to the minorityopinion being ignored. A combination of both can effectively guarantee the clear andcorrect result for disaster assessment in SLE. The case analysis for STS-51L disasterassessment verifies the effectiveness of disaster assessment model proposed by thispaper.更多还原