测试性增长试验理论与方法研究

发布时间：2018-01-01 19:39

本文关键词：测试性增长试验理论与方法研究　出处：《国防科学技术大学》2016年博士论文　论文类型：学位论文

【摘要】：测试性增长是通过逐步纠正测试性设计缺陷,不断提高装备测试性水平,从而达到预期测试性设计指标的过程。在研制阶段开展测试性增长试验是保证装备测试性指标达到设计要求的强有力手段。相对于可靠性增长试验已基本成熟应用的现状,尽管测试性设计与改进已存在大量研究成果,但测试性增长试验尚未引起足够重视,试验开展尚未上升至系统工程和科学的高度。究其原因主要是承制方缺乏科学有效的理论与技术用于指导测试性增长试验。论文针对装备测试性增长试验中存在的理论和实际问题,系统的研究了测试性增长试验理论框架、技术方法与模型以及试验管理方法,并有针对性的开展了测试性设计与改进方法研究。研究内容对提高装备测试性增长试验的科学性具有很好的实用价值。论文的主要研究成果如下:(1)针对测试性增长试验相关概念模糊的现状,理清和提出了测试性增长试验的理论框架、基本概念、实施流程与工程管理实施等。首先分析了装备设计、生产与使用全过程各阶段实现测试性增长的方法,以及各阶段实施测试性增长的时效性,指出测试性增长试验在测试性增长中的地位;然后根据测试性设计改进实施的时机对测试性增长试验进行了分类,详细分析了测试性增长试验的实施流程,对试验过程中用到的相关术语进行了定义和解释;最后对测试性增长试验管理的理论框架做了详细阐述,指出了试验规划、跟踪与预计的工作内容,及其对测试性增长试验有序开展的作用,为后文理论与实践研究奠定了基础。(2)针对目前测试性增长试验方案规划方法缺乏的问题,在分析影响测试性增长能力因素、测试性增长试验成本组成以及研制阶段装备测试性指标变化规律的基础上,研究了测试性增长试验规划模型与方法。首先对采用及时纠正策略的试验,提出了以允许失败的故障检测/隔离试验次数为优化变量的试验动态规划方法;然后对采用延缓纠正策略的试验,建立了试验资源优化配置模型,并给出了基于拉格朗日松弛与局部搜索的模型求解办法;最后对分阶段开展的测试性增长系列试验,建立了以总试验成本最优为目标,以测试性指标合同要求值为约束的多阶段测试性增长目标确定模型。案例应用和仿真对比研究证明了所提模型与方法的有效性。(3)针对测试性增长试验管理对测试性增长跟踪与预计方法的需求,在分析测试性增长试验数据特点的基础上,研究了用于测试性增长试验跟踪的测试性指标评估方法,建立了测试性增长概率模型,并利用该模型绘制了测试性增长试验跟踪与预计曲线。首先对采用及时纠正策略的试验,研究了分别考虑试验规划信息和考虑指标序化增长约束的两种不同的Bayes测试性指标评估方法,并利用数值仿真讨论了两种方法的优缺点和应用范围;其次对采用延缓纠正的试验,在总结前人研究成果的基础上,给出了基于多源先验数据的测试性增长试验评估总体技术方案;再次,考虑增长模型对刻画和绘制测试性增长试验跟踪与预计曲线的重要作用,建立了考虑测试性设计改进非理想条件的测试性增长马尔科夫链模型,将测试性指标评估结果作为模型输入,利用混合粒子群和遗传算法的参数估计方法得到了测试性增长模型的参数估计值,绘制了测试性增长跟踪与预计曲线;最后讨论了所提测试性增长模型与参数估计方法的稳健性。仿真结果表明,利用本文所提出的测试性增长模型可以方便的绘制测试性增长跟踪与预计曲线,减小了基于Bayes的测试性增长指标评估误差,并实现了对测试性增长试验的有效预计。(4)针对测试性增长过程中测试性设计改进具体实施问题,在总结已有研究成果的基础上,研究了基于试验数据的测试性设计改进方法初选规则,并重点对基于试验数据的故障诊断决策方法更新问题开展了研究。首先为能够较全面的对测试性设计缺陷进行统一描述,扩展了非完美测试的概念,给出了基于试验数据的测试性设计缺陷定量衡量方法,并根据缺陷的恶劣程度,指出了可能采取的测试性设计改进措施;然后提出了一种由基于人工免疫的数据扩充、基于密度的数据压缩,以及基于代表样本点混合学习等三部分组成的故障诊断决策算法更新方法,解决了基于数据更新的一类故障诊断决策方法存在的的分类器迭代学习、数据量不平衡,以及装备硬件支持限制等问题。研究表明,测试性增长试验数据为测试性诊断决策改进提供了有效的数据支持,本文所提方法能够较好的支持基于数据更新的诊断算法改进与更新,有效提高诊断准确度。
[Abstract]:The test of growth through gradually correct the testability design defects, and constantly improve the testability level, so as to achieve the process of design. The expected test index to carry out the test of growth test in the development stage is to ensure the equipment testing indicators meet the design requirements. The powerful means relative to the status quo of reliability growth test has been basically mature application, although the design and the improved test has many research results, but the test of growth test has not attracted enough attention, has not yet risen to carry out test of system engineering and scientific level. The main reason is the lack of scientific theory and manufacture technology effectively to guide the test of growth test. The test equipment of growth test in theory and the actual problem, systematic research framework to test the growth test theory, method and model test and management method, And to carry out the research design test and improvement method. The research content has good practical value to improve the scientific equipment testing growth test. The main research results of this paper are as follows: (1) according to the present situation of testing growth test fuzzy concepts, clear and puts forward the theoretical framework, test the basic concept of growth test, the implementation process and implementation of project management. First analyzes the design and implementation of production equipment, testing methods to increase the use of each phase of the whole process, timeliness and various stages of implementation of test growth, points out the status of growth test in the test test in the growth; then according to the design test to improve the timing of the implementation of the classification of the test and growth test, a detailed analysis of the implementation process of growth test on testing, terminology used in the testing process of the definition of reconciliation The last release; a theoretical framework to test growth test management in detail and pointed out that the test plan, tracking and predicting the content of the work, and the test of growth test is carried out in an orderly manner, will lay the foundation for the research of theory and practice. (2) due to the lack of the current test of growth test program planning method the problems in the analysis of the impact test of growth factors, the basic composition and the development cost growth test test variation stage equipment testing indicators, research the testing growth test programming model and method. The correct strategy test, proposes a test method of dynamic programming optimization variables to allow failure fault detection / isolation test number; then the delay test strategies, test resource optimization model was established, and based on Lagrange Song Model solution relaxation and local search; growth series test finally phased out, established the optimal total test cost as the goal, to test the index requirements of the contract value determination model for constrained multi stage test of growth target. Case application and simulation results show the effectiveness of the proposed model and method. (3) according to the requirements of testing growth test management for the test of growth tracking and prediction method, based on the analysis of the test data on the characteristics of growth test, test methods for evaluating growth test tracking test index research, establish test growth probability model, and use the model to draw the test growth test tracking and predicting curve. The correct strategy tests were studied considering test planning information and consider the order of two kinds of growth index constraints Bayes test of the same index evaluation method, and use the numerical simulation of the advantages and disadvantages and the application range of the two methods discussed; secondly the delay correction test, based on the previous research results, given the test data of growth prior test based on the assessment of overall technical scheme; thirdly, the growth model of the important role of drawing test characterization and growth test tracking and predicting curve, establishes a model of the Markoff chain growth test considering testability design improvement of non ideal conditions, the results of testing and evaluation indexes as model input, parameter estimation method using hybrid particle swarm optimization and genetic algorithm have been estimated parameter test of growth the model, draw the growth curve and the expected tracking test; finally discusses the test of growth model and parameter estimation method. The robustness of the simulation results The test shows that the growth model presented in this paper can easily draw test of tracking and predicting growth curve, reduce the Bayes test of growth evaluation index based on error and effective, is expected to test the growth test. (4) aiming at the improvement of the specific implementation of test design for test of the growth process, on the basis of summarizing existing research results, research the design of test data of primary rules based on the improved method, and focus on the fault diagnosis decision method based on test data updating is studied. The first is to make a comprehensive unified description of testability design defects, extends the concept of non perfect test. Gives the measure of testability design defect quantification based on test data, and according to the severity of defects, points out the improvement measures of testability design may take it; After a by the expansion of artificial immune based data, density based data compression, and the representative samples of blended learning is composed of three parts, the fault diagnosis decision algorithm based on update method, solve learning a class of fault diagnosis decision method based on iterative classifier data updating, data imbalance, and equipment hardware support restrictions and other issues. The study shows that the growth of test data to provide effective data support for the improvement of diagnostic testing decision, the proposed method can better support the improvement of diagnosis algorithm of data update and update based on, effectively improve the diagnostic accuracy.

【学位授予单位】：国防科学技术大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TH17

【相似文献】