【作者】 蒋荣华；

【导师】 王厚军；

【作者基本信息】 电子科技大学 ， 测试计量技术及仪器， 2009， 博士

【摘要】 随着电子产品集成度提高，电子系统越来越复杂，客观上对其自身的可测性提出了更高的要求。一方面电子系统可测性需要从系统整体考虑测试资源的配置与分配，以保证系统的故障诊断能力；另一方面，由于电子系统的复杂性，必须提高其可测性研究方法的精度和效率，以降低系统的测试成本。因此，单一地采用基于图论的方法、信息熵启发式算法及基于符号分析法的具体电路可测性研究方法已经不能满足电子系统可测性要求，不能全面合理地对电子系统测试资源进行安排和利用。近年来，粒子群算法以其快速收敛及工程容易实现的特点，得到广泛应用。本文根据电子系统的特点，提出一种基于粒子群算法的电子系统可测性研究方法。论文的主要研究工作有：1．电子系统可测性建模方法。本文在比较现有可测性模型的基础上，分析电子系统采用多信号模型的优势，以雷达发射机系统为例详细介绍电子系统可测性建模方法，为电子系统可测性研究奠定了基础。2．基于多维粒子群算法的测试点选取方法研究。本文将电子系统测试点选取转换为多目标优化问题，提出一种多维粒子群算法用于测试点选取及多目标优化问题。该算法将其粒子适应度函数定义为多维的，其维数与测试点选取的目标数目一致，将测试点选取问题的每一个目标定义为多维粒子群算法适应度函数的一维。通过多维粒子群算法粒子的搜索，测试点选取的多个目标同时得到优化，在多维粒子群算法中引入精英集，保证了算法的全局最优性能。实例验证表明，与其他算法相比，多维粒子群算法提高了测试点选取的效率，能较好地保证其算法全局最优性能，为粒子群算法的改进和多目标优化问题提供了新的思路。3．基于粒子群的改进AO~*算法最优测试策略设计方法研究。本文以多维粒子群算法选择的系统测点组成最优测试集为基础，针对当前最优测试策略设计方法的局部最优及当系统太大容易出现“计算爆炸”的缺点，提出一种粒子群算法与改进AO~*算法相结合的电子系统最优测试策略设计方法。该方法通过粒子群算法优选AO~*在系统最优测试策略设计中每一步要扩展的节点，减少了测试数目及AO~*算法的回溯次数，降低了计算复杂度，提高了计算效率。又由于多维粒子群算法保证了AO~*算法扩展根节点的测试集的全局最优，所以基于粒子群的改进AO~*算法能够满足电子系统最优测试策略设计方法的全局最优要求。4．基于粒子群算法的电子系统掩盖故障识别方法研究。本文针对电子系统可测性分析中系统掩盖故障识别难点，将其转换为最小碰集问题，提出一种基于粒子群算法的最小碰集求解方法，通过系统的隐藏故障及求解的掩盖故障最小碰集，能够得到系统所有的掩盖故障集。该方法能够避免原有算法当系统太大，容易出现“组合爆炸”的缺点，尤其适合于识别大型复杂电子系统的掩盖故障集，也为最小碰集的求解提供了一种新的方法。5．基于变异粒子群算法的模拟电路模糊组识别方法研究。模拟电路模糊组识别方法是模拟电路可测性分析的难点，本文提出一种变异粒子群算法识别模拟电路模糊组。该方法以符号分析法为基础，建立模拟电路测试矩阵，将矩阵中的向量分为基向量与非基向量，通过粒子群算法粒子速度变异，快速识别模拟电路模糊组。该方法能够彻底消除原有三角分解方法的误差，提高了计算精度。6．基于可测性分析的电子系统改进与多故障诊断方法研究。电子系统可测性是系统故障诊断能力的保证，本文针对现有电子系统，通过可测性分析方法评价其故障诊断能力，对可测性较差的系统进行改进；针对系统存在隐藏故障和掩盖故障时，容易出现“多故障并发的单故障现象”，通过系统可测性分析进行有效的故障诊断，弥补了原有系统多故障诊断方法的不足。更多还原

【Abstract】 With the increased integration of electronic products,electronic system becomesmore and more complex and needs a higher demand of testability objectively.Forelectronic system testability,test resource must be collocated and distributed in thewhole system;on the other side,the accuracy and efficiency of testability study methodshave to be advanced to reduce test cost because of the complexity of electronic system.Therefore,any single testability study method such as graph theory-based method,information entropy-based algorithm and symbol analysis-based method for specificcircuit can not satisfy testability demand and arrange test resource comprehensively andreasonably.In recent years,a particle swarm optimization algorithm is applied widelyowning to its characteristics of fast convergence and easy to realize in engineering.According to characteristics of electronic systems,this dissertation presents a studymethod of electronic system testability based on particle swarm optimization algorithm.The main work is as follows:1.Study on testability model of electronic system.On the basis of comparingexisting testability models,the superiority of using multi-signal model for electronicsystem testability and detailed steps of constructing electronic system testability modelvia radar transmitter system is presented.It lays the foundation for electronic systemtestability study.2.Study on test points selection of electronic system based on multidimensionalfitness function discrete particle swarm optimization(MDFDPSO) algorithm.Testpoints selection in electronic system is converted to a multi-objective optimizationproblem,a multidimensional fitness function discrete particle swarm optimizationalgorithm is proposed to select test points and solve multi-objective optimizationproblem.This algorithm defined a multidimensional fitness function,the number ofdimensions in its fitness function is the same as the number of optimal goals,eachdimension of the fitness function is a goal in test points selection problem.The multiplegoals can be optimized at one time via the particle searching in MDFDPSO algorithm.An elitist set is added in MDFDPSO to guarantee global optimization capability. Examples in the dissertation validated that MDFDPSO improves efficiency of testpoints selection and has good global optimization characteristic when it is comparedwith other methods.It provides a new way to improve particle swarm optimizationalgorithm and solve multi-objective optimization problem.3.Study on optimal test strategy problem based on particle swarm optimizationand improved AO~*.For the case that current methods have disadvantage in localoptimization and“computation explosion”when the system is too large,a combinationmethod of particle swarm optimization and improved AO~* is proposed in thisdissertation to design the optimal test strategy in electronic system on the foundation ofthe best test sets selected by MDFDPSO.This method uses particle swarm optimizationalgorithm to select test in each node to expand by AO~* in the process of designing thebest test strategy.It decreases the number of test and trace by AO~*,reduces thecomplexity of computation and improves computational efficiency.Because the root ofAO~* is the test set selected by MDFDPSO,this method can satisfy the global optimaldemand in best test strategy design for electronic system.4.Study on identifying masking false failure in electronic system based on particleswarm optimization.Identifying masking false failure is a difficult problem in electronicsystem testability analysis.This dissertation converts it to minimal hitting set matter andpresents a method based on particle swarm optimization algorithm to solve minimalhitting set problem.It can attain all masking false failure in system through hidden faultand the minimal hitting set of masking false failure and avoids“CombinatorialExplosion”for large-scale system compared with existing methods.It is especiallysuitable for identifying masking false failure in large-scale complex electronic systemand provides a new method of solving minimal hitting set problem.5.Study on finding ambiguity groups in analog circuit via variation particle swarmoptimization algorithm.Finding ambiguity groups in analog circuit is difficult,avariation particle swarm optimization algorithm is proposed.Based on symbol analysismethod,it constructs the testability matrix of analog circuit,decomposes the vector intestability matrix into base vector and non-base vector,and can quickly find allambiguity groups in analog circuit via the variation of velocity in particle swarmoptimization algorithm.This method drastically eliminates the errors caused by originaltriangular factorization method and boosts the computation precision. 6.Study on improvement and multiple fault diagnosis of electronic system basedon testability analysis.Testability guarantees the ability of electronic system faultdiagnosis;this dissertation evaluates the fault diagnosis ability via testability analysisand improves the system whose testability is poor.If ther are hidden faults and maskingfalse failures in electronic system,“multiple fault concurrent-single failurephenomenon”can easily appear.The method proposed in this dissertation canefficiently isolates the faulty components in this phenomenon through testabilityanalysis and makes up the shortage in original multiple fault diagnosis method.更多还原