【作者】 林德明；

【导师】 金鸿章；

【作者基本信息】 哈尔滨工程大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 本文基于复杂适应系统理论提出了适应性Agent图的概念,所谓适应性Agent图是指顶点具有适应性的智能图,其中适应性是指复杂适应系统中的适应性Agent主体,它具有适应性能力,能够与其他主体和环境进行交互作用,并且不断学习、总结经验,是复杂适应系统的复杂性根源。围绕着适应性Agent图我们建立了相关的定义,给出了适应性Agent图的动力学模型;给出了它的矩阵表示,使其更易于表示和运算;建立了适应性Agent图的分析方法,包括稳定性分析、频域分析等;这些分析为它的动态行为特性研究提供了理论方法;另外还对适应性Agent图的适应性行为进行了探讨。适应性Agent图融合了复杂适应系统理论的思想,具有适应性、智能性、并行性、涌现性、记忆性、直观清晰、逻辑性强、定性与定量相结合的特点。基于适应性Agent图的种种优点,我们将它应用到复杂系统脆性的研究中。复杂系统脆性是复杂系统的一个重要属性,是一个新的概念。脆性是指系统中某个子系统或某一部分崩溃后,经过连锁性过程,可能导致整个系统崩溃。本论文分析了系统脆性行为与熵的关系,论证了脆性的存在性。在此基础上建立了脆性的元胞自动机模型,它是复杂系统脆性的基本模型。它直观清晰,能够形象地描述系统脆性激发后的几种常见的崩溃形式。然后,建立了复杂系统的子系统脆性联系的图形表示,将其定义为脆性图。针对脆性图的研究,给出了系统崩溃路径的概念,以此描述系统崩溃的连锁性过程。改进了对解决旅行商问题具有很好效果的蚁群算法,来求解系统中最大(小)崩溃路径。崩溃路径的研究对于系统脆性源的辨识以及崩溃的分析与预测有着重要的指导作用。为了更加全面、深入地了解系统的脆性,我们将复杂系统的脆性图升级为适应性Agent图模型。设系统的每个子系统为一个适应性的Agent主体,以熵为顶点之间的交互作用的流函数,以子系统之间的脆性联系为拓扑结构,从而建立了复杂系统脆性的适应性Agent图模型。并且应用适应性Agem图模型,对复杂系统的脆性行为的动态行为特性进行了分析。最后,以舰船电力网络为例子,应用适应性Agent图对它的脆性进行了分析。随着舰船工业的发展,舰船的吨位逐渐升级,舰船电网的规模越来越大,电网结构越来越复杂,所以对它的脆性的研究是很有必要的。因此以舰船电力网络系统为研究对象,应用适应性Agent图进行脆性分析,得到脆性行为的动态演化形式。而且对某船电力网络的全局熵的进行了方差分析,论述了舰船电力网络的脆性行为与系统结构的关系。舰船电力网络有馈线、干线、混合、环形和网形五种常见的配电方式,针对这五种配电方式构成的拓扑结构,建立其适应性Agent图的脆性模型,并且将这五种拓扑结构的脆性进行了比较分析。更多还原

【Abstract】 Adaptive agent digraph based on the complex adaptive theory and digraph theory has been initiated in this paper. It is an intelligent digraph, whose vertex has adaptability. The adaptive digraph’s vertex is as well as an adaptive agent in complex adaptive system. It could store knowledge, study by influencing each other and exchanging with the environment. Adaptability is the cause of complexity.To discuss the adaptive agent digraph, the correlative definition, the dynamic equation of adaptive vertexes and the condition on stability have been proposed. Adjacency matrix of an adaptive agent digraph has been presented to express and calculate the digraph simply. Some methods have been presented to analyze the dynamic behaviors of digraph, such as stability analysis and frequency analysis. Then the adaptability of vertexes has been explored. The adaptive agent digraph based on complex adaptive system theory has the following characteristic: Adaptability, intelligence, parallelism, emergence, memory, clear visualization, good logic and it couldn’t analysis a system quantitatively but analysis it qualitatively. So it is applied to study the brittleness of complex systems.Brittleness is an important character of complex systems. It is a new concept. A system can be collapse after one of its subsystems being collapse by catenulate course, that is brittleness. To prove the being of brittleness, the relation between the entropy and the collapse in a complex system has been analyzed. On this condition, the model of brittleness based on cellular automata has been established. The model is the basic model of brittleness. It can visualize the normal collapse form of complex systems clearly.Furthermore, a digraph for brittleness has been used to describe the brittle link of complex systems. Then the collapse path in the digraph has been presented to express the pipeline of collapse in a system. Ant colony algorithm is good at the TSP, so an improved ant colony algorithm has been proposed to seek for the maximal or minimal collapse path. The research on collapse path is necessary to find out the brittle source and forecast the collapse.To study the brittleness further, an adaptive agent digraph model of brittleness has been established. In the model, the adaptive vertex is the subsystems, the entropy is the flow function, flow is the medium to affect each other among subsystems, and brittle link between subsystems is arc of digraph. Then the brittle behaviors with the adaptive agent digraph model have been studied.In order to illuminate the adaptive agent digraph, the warship power network is looked as the example. The warship power network is more larges and more complex along with the development of ship, so the research on the brittleness of warship power network is important. Moreover the adaptive agent digraph model is applied to study on the brittleness of a warship power network. At first, the being of brittleness in the warship power networks has been discussed. The dynamic brittle behaviors of the warship power network have been simulated. Then using the variance analysis for different adjacency weight matrixes, the relation between adjacency weight matrix and the disorder degree of the whole network has been analyzed.The warship power network has five power distribution modes with different topological structures: feedback mode, main bus mode, mixed mode, ring mode and net mode. In the end, the five topological adaptive agent digraph models of brittleness for the five modes have been built up, and brittleness of the five modes have been analyzed and compared.更多还原