【作者】 袁东辉；

【导师】 刘大有；

【作者基本信息】 吉林大学 ， 计算机应用技术， 2011， 博士

【摘要】 二十一世纪以来,各国空军的发展代表了军队建设的现代化进程,而飞行模拟器为空军的训练机制提供了完整的技术支持,它的引入具有节省训练经费,保障飞行员安全,轻松完成复杂特情训练等优势,因此我国在部队现代化进程中也对飞行模拟器开展了大量的立项研究,目前来讲飞行模拟器由平台系统、计算机系统、视景系统、电子控制系统以及控制台系统组成,其中平台系统、计算机系统以及视景系统随着多年实践经验的积累已经日趋稳定。目前随着对模拟器模拟实战能力要求的提高,如何利用飞行控制台提供集团作战模拟是亟待解决的问题之一,同时大量集成电路的引入也为模拟器的维护增加了困难,针对大规模集成电路的故障诊断一直是国际上研究的热点问题,本文将以上述几点为依据展开研究。蚁群算法源于人们对生物界蚁群觅食行为的观察和模拟,相较于传统的优化方法,蚁群算法在解决各类优化问题时具有很强的适应性和鲁棒性,适用于各类组合优化问题。蚁群算法在使用过程中参数设置简单,易于实现,因此在众多科学研究以及生产制造领域得到了广泛的应用,本文以蚁群算法为基础同时结合遗传算法和模拟退火算法两种仿生算法,对飞行模拟器平台中的多目标数据关联问题,数字电路测试集生成以及测试矢量集合优化问题进行了一定程度的研究,并且通过相关实验将本文算法与现有的一些算法进行了对比,实验结果表明,本文提出的几种算法能有效的提高蚁群算法的收敛速度,克服局部极值的出现,算法在迭代次数和执行效率上都体现出了很大的优势,本文的具体研究内容如下:1.以多传感器多目标数据关联问题为研究对象,提出了一种基于蚁群-遗传算法的AC-GADA算法,该算法结合了遗传算法的染色体交叉变异思想,为蚁群中的个体赋予了单独的染色体,同时对蚁群算法的信息素模型进行了改进,在每一代种群寻优之后,利用交叉变异机制对蚁群进行重新构造,同时结合了控制种群适应度的方法,进一步做到了个体差异化,有效地避免了局部极值的出现,在AC-GADA算法中,信息素强度不仅与途经该路径的蚂蚁数目相关,同时还受蚂蚁自身的信息素编码影响,因此AC-GADA算法具有更强的随机搜索能力,通过大量的实验表明AC-GADA算法在收敛速度、关联的准确率和成功率上都有明显的优势。2.针对大规模集成电路的故障诊断问题,提出了AC-GATSG算法,该算法首先将大规模集成电路的关联矩阵转化为图结构,从而利用蚁群算法解决图着色问题上的优势,对电路测试集的自动生成问题进行探索实现,在处理图节点着色问题时同样引入了交叉变异机制,在每代的遍历过程出现最优解时对蚁群个体进行染色重构,同时更新信息素,并将阈值引入信息素扩散模型中,利用阈值限定了交叉变异过程中产生突变个体的几率,降低了算法执行过程中由一个局部极值陷入另一个局部极值的可能。在对比实验中选取了标准组合电路以及典型的飞行模拟器电路单元为对象,通过与一些经典算法以及以仿生学算法为基础的文献算法对比,验证了AC-GATSG算法的有效性,AC-GATSG算法能够满足模拟器生产实践中对基础电路单元的测试集生成需求,并且具有广阔的应用前景。3.自动生成的大规模集成电路测试集往往存在冗余问题,在进行标准化测试时,这些冗余矢量会造成大量的资源浪费,影响生产和研发的效率,针对上述问题本文提出了AC-SATSO算法对电路测试集进行优化,AC-SATSO算法首先对故障节点与测试矢量节点进行分层建模,并通过故障入度和测试矢量出度对双层节点进行标识,AC-SATSO算法利用模拟退火算法的扰动机制加速种群的收敛速度,并通过Boltzmann机制判断是否接纳新的种群,这种寻优策略在算法执行初期可以避免由于缺乏信息素而导致的寻优速度慢的问题,后期则有效的限制了局部极值的出现并且同时保留了优良的数据集,随着迭代次数的增加,模拟退火算法保留优秀结果的概率也大大增加,同时也加速了算法的收敛。并且针对生产实践中的测试环节本文提出了最优完备测试集的概念,即所需系统开销最少的完全测试集,通过不同规模的测试集优化对比试验证明,AC-SATSO算法在求解质量和稳定性上都优于文献算法。近年来,蚁群算法等仿生算法的研究受到了国内外学者的关注,并且涌现出了大量的改进算法和新的领域应用,本文以蚁群算法为基础,结合遗传算法、模拟退火算法在飞行模拟器平台中展开了一些研究,提出了几种相互结合的优化算法,在蚁群算法的改进以及更丰富领域的应用上,其研究具有一定的实际意义。更多还原

【Abstract】 In twenty-first century, the development of national Air Force represent the level of modernization of the military construction, and the flight simulator provides the complete technical support for Air Force training system, there are a lot of advantages of flight simulator such as saving training fund, protecting pilots’lives and performing special and complex training, and we have carried out a large number of research projects of flight simulator. The flight simulator is composed by platform system, computer system, visualization system, electronic controling system and console system. With the years of experiences, the platform system, computer system and visualization system have become increasingly stable. The group operation in complex situation is a very urgent problem that needs to be solved due to the requirement of higher combat abilities of flight simulator. A increasing number of integrated circuits also increase the difficulty of maintain flight simulator, and fault diagnosis of large-scale integrated circuits has became an international issue.Ant colony algorithm originated from the observation and simulation of the ants’behaviors. Compared with the traditional optimization methods, AC is widely applied in scientific research and industrial production in virtue of their higher adaptability, robustness and parallel processing capability. This paper ccommence the study in multi-target data association, digital circuit test set generation and test set optimization with AC combined with genetic algorithm and simulated annealing algorithm. The result of experiments has shown that AC-GADA, AC-GATSG and AC-SATSO can effectively improve the convergence speed of the ant colony algorithm, and overcome the emergence of local extremum. The main contents of this thesis are as follows:For multi-sensor multi-target data association problem, this paper presents an AC-GADA algorithm which combines ant colony algorithm with genetic algorithm. This algorithm designed difference pheromone for each ant and improves global pheromone increment model, and combined crossover and mutation strategy with fitness of population model in order to improve rate of convergence and avoid the appearance of local extremum. In AC-GADA, the pheromone was desicided not only by the number of ants which have chosen the path but also the pheromone code of each ant. The comparison with ACDA(Ant Colony Data Association) and JPAD(Joint Pobabilistic Data Association) proved its efficiency and superiority.For large scale integrated circuit fault diagnosis problem, this paper presents AC-GATSG algorithm. The algorithm converts scale integrated circuit into a graph, and conducts the study of automatic generation by using the advantage AC algorithm. In AC-GATSG, the individual stained will be reconstructed when optimal solution occurs in each generation, and the pheromone will be updated at the same time. Mutation probability of individual mutations and threshold which limits generated during crossover are useful elements in AC-GATSG which reduce the execution of generating local minimum one by one. We select combinational circuits and typical flight simulator circuit units as experiment objects, and the results show that AC-GATSG algorithm can satisfy the requirement of practical production, and has a widely application prospect.The test set of large scale intefrated circuit, which is generated automatically, has a large number of redundant vectors that cause high waste of resources and low efficiency of production and development in standardized tests. This paper presents AC-SATSO to optimize the sets of the circuit tests. AC-SATSO builds the model of test vector nodes and fault nodes hierarchically, and identify these nodes with in-out degree. AC-SATSO algorithm combines AC with simulated annealing algorithm to accelerate the convergence rate and determine whether to accept the new population with Boltzmann mechanism. Under this optimization strategy, AC-SATSO can make a great convergence rate although lacking of pheromone at the beginning, and avoid the appearance of local minimum in subsequent stage. Besides, this paper puts forward the concept of optimum complete test set which complete the test with the lowest consumption. The result of experiment demonstrated that AC-SATSO method of solution quality and stability of the algorithm are better than literature. In recent years, researches on ant colony algorithm, as well as its applications, have been paid great attentions by many researchers globally. Based on ant colony algorithm, genetic algorithm and simulated annealing algorithm, this paper focuses on the research of flight simulator, and proposes several optimization methods. There are both theoretical and practical significance of this research on the improvement of ant colony algorithm.更多还原