轨道交通运营软件行为动态测评方法研究

发布时间：2018-05-29 08:49

本文选题：软件行为轨迹 + 动态测评　；参考：《苏州大学》2012年硕士论文

【摘要】：随着苏州市地铁土建工程的逐步推进，人们对城市轨道交通运营软件的可靠性、可用性以及安全性等可信性质寄予了很高的期望与要求。为了解决城市轨道交通运营软件日益突出的可信性问题，仅对软件系统做出传统质量保证(如测试、验证)是不够的，更需要对系统交互行为进行有效的分析与态势预测。轨道交通运营软件有别于传统分布式软件，它呈现出松散聚合、规模庞大、行为复杂等特点。本文将软件行为作为切入点，对轨道交通运营软件行为的轨迹分析以及意图预测逐步展开了研究，建立一种以软件行为可信为核心的软件行为动态测评方法。本文将系统监测到的状态变化映射为带有语义的事件序列，利用最精简主要序列提取算法对行为序列进行提炼精简，从被消减掉的重复子序列挖掘有用信息，产生能体现软件行为特征的行为序列，通过序列对比分析行为的可信性，用HMM对交互行为状态进行预测，此为大粒度行为可信性动态测评研究；同时以交互事件为切入点，关注事件的参数值以及经验知识等详细信息，用MEBN工具建立具体事态的贝叶斯网(SSBN)来分析交互行为的复杂过程及效应，此为小粒度行为可信性动态测评研究。本文采用小粒度动态测评分析方法处理复杂情形下群体交互行为可信性分析，采用大粒度行为动态测评分析通用情形。本文的主要工作包括： (1)研究轨道交通运营软件行为的基本特点。明确各智能子系统的信息需求以及它们之间的关系，为打破子系统各自为营的封闭状态，发挥整体优势，实现一个高效运营的分布式集成化的城市轨道交通运营系统提供理论支持。 (2)研究一套软件行为描述方法，将监测系统捕获到的可信相关数据生成能表征软件行为的信息，针对软件实体交互产生的重复子序列的问题，提出最精简主要序列提取算法，从被消除的重复子踪迹中挖掘有用序列，以最简洁而有广泛代表性的格式存储软件行为信息。 (3)给出一种复杂的有标记的大粒度软件行为动态测评方法。使用序列两两比较算法分析行为的可信性。基于EM的数据重构方法训练样本数据，，用HMM预测软件行为安全类别。给出仿真实验，验证此方法具有一定的优势。 (4)针对多实体贝叶斯网MEBN在表示不确定性关系上的优势，本文给出一种复杂情形小粒度行为动态测评方法。此方法有效利用MEBN的一阶逻辑语义化表示能力和概率推理能力，采用片断集有效地描述和分析多个软件实体交互产生的复杂行为，通过仿真实验，验证了小粒度行为轨迹动态测评的可行性和有效性。基于软件行为可信性展开软件行为动态测评方法的研究，不仅能推动动态可信评测理论的发展，对技术实践也有很好的指导意义。
[Abstract]:With the gradual development of subway civil engineering in Suzhou City, people have high expectations and requirements for the reliability, usability and security of urban rail transit operation software. In order to solve the increasingly prominent credibility problem of urban rail transit operation software, it is not enough to make traditional quality assurance (such as testing and verification) for the software system. It is also necessary to effectively analyze and predict the interaction behavior of the system. The rail transit operation software is different from the traditional distributed software. It has the characteristics of loose aggregation, large scale and complex behavior. In this paper, the software behavior is taken as the breakthrough point, the trajectory analysis and intention prediction of the software behavior of rail transit are studied step by step, and a dynamic evaluation method of software behavior based on software behavior trustworthiness is established. In this paper, the state changes monitored by the system are mapped to the event sequences with semantics, and the behavior sequences are refined by the most simplified main sequence extraction algorithms, and useful information is extracted from the subsequences of the subsequences that have been reduced. The behavior sequence which can reflect the behavior characteristic of software is produced, and the credibility of the behavior is analyzed by comparing the sequence, and the interactive behavior state is predicted by HMM, which is the dynamic evaluation research of the large granularity behavior credibility, and the interactive event is taken as the breakthrough point at the same time. Based on the detailed information of event parameters and empirical knowledge, the Bayesian Network (MEBN) is used to analyze the complex process and effect of interaction behavior, which is a dynamic evaluation study of small-grained behavior credibility. In this paper, the small granularity dynamic evaluation and analysis method is used to deal with the credibility analysis of group interaction behavior in complex situations, and the large granularity behavior dynamic evaluation method is used to analyze the general situation. The main work of this paper includes: 1) the basic characteristics of the software behavior of rail transit are studied. To clarify the information needs of each intelligent subsystem and the relationship between them, in order to break the closed state of each subsystem, give play to the overall advantage, To implement a distributed integrated urban rail transit operation system with high efficiency provides theoretical support. (2) A set of software behavior description method is studied. The credible and relevant data captured by the monitoring system are used to generate information that can represent the software behavior. Aiming at the problem of repetitive sub-sequences generated by software entity interaction, the most simplified main sequence extraction algorithm is proposed. Useful sequences are mined from erased repeats to store software behavior information in the most concise and widely representative format. In this paper, a new method of dynamic evaluation of large granularity software behavior is presented. The sequence pairwise comparison algorithm is used to analyze the credibility of behavior. The data reconstruction method based on EM is used to train the sample data, and the HMM is used to predict the behavior security category of software. The simulation results show that this method has some advantages. 4) in view of the advantage of multi-entity Bayesian network (MEBN) in representing uncertain relations, this paper presents a method for dynamic evaluation of small-grained behavior in complex cases. In this method, the first-order logical semantic representation and probabilistic reasoning ability of MEBN are effectively utilized, and the complex behaviors generated by the interaction of multiple software entities are described and analyzed effectively by using the fragment set, and the simulation experiments are carried out. The feasibility and effectiveness of dynamic evaluation of small particle behavior trajectory are verified. The research of software behavior dynamic evaluation method based on software behavior credibility can not only promote the development of dynamic credible evaluation theory, but also have a good guiding significance to technical practice.
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP311.53;F572

【引证文献】