【作者】 周小平；

【导师】 郭开仲；

【作者基本信息】 广东工业大学 ， 管理科学与工程， 2013， 博士

【摘要】 千里之堤毁于蚁穴,小错误引起大损失。如何有效地识别系统中的错误,进而达到纠正或消除错误的目的,是各行各业、各个领域都在努力寻求解决的问题。随着科学技术的发展,尤其是计算机技术、人工智能和自动化理论的发展,计算机及系统工程等领域对错误识别的研究逐渐增多。计算机领域主要致力于错误识别技术的研究,研究内容包括语音、字符、生物特征等,例如语音识别技术的研究及其在发音错误识别系统中的应用[1]、汉字智能工具中的书写错误识别[2]等。系统工程领域主要研究错误的分析法和故障诊断等,期望达到系统优化的目的。本文继承消错理论的研究成果,采用系统工程、数学、逻辑学及计算机工程等理论和方法,构建系统的错误识别理论及方法,包括错误识别对象、识别规则、错误识别方法,及错误识别理论在裁判员决策支持系统中的应用研究。首先,结合错误逻辑等理论,用论域、事物、特征、量值、时间等六个特征或组合表征错误识别的对象；将错误识别对象按活动状态、人们对系统的了解程度、对象的特征T(t)量值L(t)的不同进行分类,并讨论了错误识别对象的分类方法在对象识别中的应用,有利于更全面、系统地认识错误识别的对象；定义了错误识别对象的状态,将错误识别对象的状态分为识别状态、真实状态、应该状态和目标状态等四种状态类型,并讨论了四者之间的关系。运用错误矩阵逻辑方程一类4方程AVX(?)B求解了识别状态与目标状态之间、识别状态与真实状态之间、识别状态与应该状态之间的转化,实现消除和避免错误的目的。其次,研究了错误识别规则建立的科学性、相对完备性和内部无矛盾性等原则；通过对对象系统的条件T，结论J，固有功能GY,目的功能MG,关系R等分析,研究了对象系统规则建立的模式和步骤；根据规则中的指标类型,将规则分为量化规则和非量化规则。可量化规则又分为指标值区间型、指标值固定型、指标值趋小型和指标值趋大型四种类型的规则,并且讨论了在对象集U中的元素u一定的前提下,不同指标类型的规则所对应的错误函数；研究规则函数的差、并和交运算的方法,讨论了规则之间是否有交互作用的情况。再次,结合其它领域对故障、错误识别的研究,根据消错学研究对象的特殊性,提出了几种常见的错误识别方法,包括预先错误分析法、错误假设分析法、错误检查表法、错误与可操作性方法；建立了基于错误系统角度的“PKR”错误识别方法。将系统分解成由最小要素集(可称为点集)、键集和路集构成,再根据各对象的规则,列出“PKR”系统错误识别表,采用相应的规则进行判断,最后达到有效识别错误的目的。最后,研究了基于错误识别理论和方法的南拳动作规格错误识别模型。采用数学模型法、平面方程、解析几何等方法和工具,对南拳套路动作质量(A组)评分规则规定的前扫腿、旋风腿、马步、弓步等十五个动作规则进行形式化描述,建立错误函数和基于计算机视觉的错误动作扣分模型,并举例说明,以期实现基于计算机视觉的武术套路裁判员决策支持系统,减少裁判员人工识别的难度和人为的误差和影响,使裁判工作更加科学、公正。本文以错误识别的对象、规则、识别方法和南拳裁判员动作规格错误识别模型为切入点和重点,系统地构建了错误识别方法和理论,及其在裁判员决策支持系统中的应用,不仅为系统错误识别及系统优化提供了理论支持,同时也为生产和管理实践中的错误识别及错误消除和避免、裁判员决策支持系统等提供了实践方法和工具。更多还原

【Abstract】 Thousands of miles of dikes destroyed the colony, small errors cause large losses. Small mistake can cause major losses of individual, organization and the country. How to effectively recognize errors in the system, and to eliminate errors, is the problem sought by people.With the development of science and technology, especially the development of Computer technology, Artificial Intelligence and Automation. Error recognition is the main issue studied by computer, systems engineering and other fields. They mainly studied error recognition technology in computer industry fields. It contains voice, character, and biometric identification, uch as voice recognition technology research and its pronunciation error recognition system application[1], Chinese intelligence tools typographical errors in the identification[2] and others. They mainly studied error analysis and fault diagnosis to achieve system optimization in system engineering field.The paper construct a complete error recognition system theory and methodology, including error recognition object, error recognition rules, error recognition methods and the application of referees decision support system, to carry forward the Error Recognition Theory, use systems engineering, mathematics, logic and computer engineering theories and methods, etc.First, combine with the study of faulty logic, use six characteristics such as domain, things, features, value and time; Classify the identified objects by active state, the realization of system, features of the object denoted by T(t) and the value of the object denoted by L(t), discuss the application of the identified objects in object identify; And studied the state of the identified object. Divided the state of the identified object into four states, such as the identified state, the true state, the expected state and the target state. We discussed the relationship between the identified state, the true state and the expected state. Finally, we use1class of4equations AV X<B in Matrix logic equations to transform the identified state into the target state.Second, study the rule in error identification. Study the principles of rule set up, such as scientificity, relative completeness and internal non-contradiction etc.; Analyze the object system conditions T, conclusions J, inherent function GY, objective function MG, and relations R etc., set up the model and steps of object rule’s establishment; Rules are divided into quantitative and non-quantitative rules according to the rules of the index type. Quantitative rules are divided into four types of rules, such as interval index value, fixed index value, trend of small index value and large-scale trend index value. And discusses the different index types rules corresponding to different forms of error function, with certain elements in objects set; Study the operational method of intersection, union, set-difference, and discussed whether there is interaction between the rules.Third, combine with the study of fault diagnosis and error recognition, study several error identification methods, including pre-error analysis, false assumptions analysis, error checking table method, errors and operability methods, according to the particularity of object in the Error Recognition Theory; Establish ’PKR’ error identification method based on the perspective of system. The system is decomposed into the minimum feature set (may be called point sets), key set and path set. Then list the error identification table according to the rules of each object. After then judge the error with the corresponding rules. And finally to achieve the purpose of recognition errors effectively.Finally, the study error identification model of Nan Quan action specification based on error identification theory and method. Use mathematical models, the plane equation, analytic geometry and other methods and tools, formal description the fifteen actions rules in quality (A group) of Nan Quan Routine. study rule’s error function, establish error identification model in compute vision, and illustrate, in order to reduce the recognized difficulty and man-made influence, make more scientific and fair.The paper studied identified objects, rules, identification methods and error identification model of Nan Quan action specification as the starting point and focus, constructed error recognition methods and theories systematically, and the application in referees decision support systems. Not only provides a theoretical support for system errors identification and system optimization, but also provide practical methods and tools for error recognition, error elimination, error avoid, and referee decision support systems in production and management practices.更多还原