不确定信息认知对象的仿反馈认知智能机制与计算模型研究

发布时间：2018-07-05 09:46

本文选题：认知智能 + 不确定信息　；参考：《合肥工业大学》2017年博士论文

【摘要】：模拟人的意识、思维过程是一门极富挑战性的科学。本文针对传统认知系统单向开环方式与人类认知事物反复推敲比对的信息交互过程存在显著差异问题,面对认知系统中可识别率与正确率不可在线实时测评的工程难题,探讨了不确定信息认知对象建模、广义认知误差度量和仿反馈调节机制三位一体的仿人认知智能新研究方法。基于粗糙集理论,探索建立不确定信息认知对象的模式化结构模型和运行机制。基于广义误差理论与熵理论,探索建立广义认知误差的多层次变粒度熵函数形式测度指标评价体系。基于反馈控制思想和认知知识粒度计算工具,探索建立具有仿反馈模式的认知智能运行机制。力图对一类不确定信息认知对象和不确定认知过程效果与认知结果的认知智能系统建立仿人反复推敲比对思维信息交互模式的认知智能机制与计算模型,这对于人工智能学科的发展具有拓展意义。本文取得的主要研究成果如下:(1)不确定信息认知对象仿反馈认知智能系统的结构与运行机制研究。构建一种仿人认知事物粗推敲细比对反复交互模式的三层三段互耦合仿反馈认知智能系统的结构模式与运行机制,给出了仿反馈认知智能系统的构建目标,规范了功能要求和运行机制,为建立不确定信息认知对象的仿反馈认知智能系统提供了模式化结构的构建方法。(2)不确定信息认知对象的认知智能决策信息系统模型研究。针对目前学术界普遍认为不存在能够适应于各种数据的通用知识获取方法问题,构建了基于粗糙集的不确定信息认知对象的多信息融合有监督学习认知智能决策信息系统,给出了基于特征简约的认知知识融合充分性和可分类性熵函数形式测度指标,在有限论域的不确定信息条件下,为不确定信息认知对象的认知知识优化表征提供了普适性的认知智能决策信息系统模型。(3)不确定认知过程效果与认知结果的广义认知误差评价体系研究。针对目前学术界普遍采用后验评价认知结果可信度方法不能满足认知系统中可识别率与正确率的在线实时测评问题,构建了不确定信息认知对象和不确定认知过程效果与认知结果的认知误差信息系统,定义了基于熵函数形式测度指标的不确定认知过程效果与认知结果误差熵、不确定认知过程效果与认知结果误差熵序列相似度和认知知识粒度误差三种广义认知误差,建立了不确定认知过程效果与认知结果可信度在线实时测评体系,为仿反馈认知智能调节机制提供了量化标准。(4)不确定信息认知对象的仿反馈认知智能机制研究。针对传统认知系统单向开环方式与人类认知事物反复推敲比对的信息交互过程存在显著差异问题,构建了基于广义认知误差在线实时度量的仿反馈认知智能调节机制,定义了仿反馈认知知识粒度调节规则与计算方法,建立了多层次变知识粒度的层间调节与层内寻优模式,实现对不确定信息认知对象的变知识粒度的仿反馈认知智能调节机制,为模仿人类认知事物反复推敲比对的信息交互过程提供了一种智能认知机制。(5)单目标完全状态的脱机手写体汉字图像的机器认知应用验证研究。针对手写体汉字数量大,字体字型繁多,相似字多,书写风格各异等识别难题,基于本文提出的仿反馈认知智能方法,研究了具有认知知识粒度智能调节机制的脱机手写体汉字图像仿反馈认知智能系统。选用SCUT-IRAC HCCLIB手写体汉字图像样本库的多类汉字样本图像作为认知对象,进行了仿真实验以验证本方法的可行性与有效性。实验结果表明,本文方法较传统单向开环方法具有明显优势,为实现对脱机手写体汉字图像的“机器识字”代替“人工识字”探索了一种实践应用新方法。(6)多目标较完全状态的工业回转窑烧成状态的机器认知应用验证研究。针对工业环境复杂所导致采集的样本数据性能较差问题,基于本文提出的仿反馈认知智能方法,研究了基于火焰图像的工业回转窑烧成状态仿反馈认知智能系统。选用东北大学流程工业综合自动化国家重点实验室提供的某水泥厂2#回转窑烧成带火焰图像样本和测量数据,采用MATLAB仿真实验验证本方法的可行性与有效性。实验结果表明,本文方法能够从宏观到微观挖掘火焰图像特征知识并反复认知决策面附近的样本,为实现水泥烧结过程监控和基于熟料质量闭环控制的“机器看火”取代“人工看火”探索了一种工程应用新方法。(7)多目标不完全状态的人体健康状态评测的机器认知应用验证研究。针对人群样本非均质性、人体特征多样性和人体特征值随机性等问题,基于本文提出的仿反馈认知智能方法,研究了基于变认知知识粒度的人体健康状态仿反馈认知智能系统。选用中国科学院智能机械研究所提供的随机人群生物电阻抗信号样本,采用MATLAB仿真实验以验证本方法的可行性与有效性。实验结果表明,本文方法能够在不同认知知识粒度层次中表征健康知识并模糊匹配健康指标,为客观准确地评测人体健康状态的“机器诊断”取代“人工诊断”探索了一种实际应用新方法。(8)单目标不完全状态的全天候光伏发电智能认知跟踪控制应用验证研究。针对天气环境的非线性、非平稳性、大间歇性和随机性导致的光伏电池跟踪控制难题,基于本文提出的仿反馈认知智能方法,研究了全天候光伏发电智能认知跟踪控制系统,并给出了基于全景图像变粒度仿反馈认知太阳位置的计算方法。采用平板型和聚光型等多种光伏电池组件及其伺服机构,开展了实验研究以验证本文方法的可行性与有效性。实验结果表明,本文的三层三段光伏发电智能认知跟踪控制系统模型能够基于风、雪、白天、黑夜、阴/雨和晴/多云等环境与系统状态的认知知识执行相应的控制策略并在晴/多云跟踪控制中基于准确计算太阳方位提高光伏效能,为实现光伏发电系统的高效、节能和安全运行探索了一种智能化控制应用新方法。
[Abstract]:To simulate human consciousness, the process of thinking is a very challenging science. In this paper, there is a significant difference in the information interaction between the one-way open loop mode of the traditional cognitive system and the comparison of the human cognitive things, and the difficult problem that the recognition rate and the correct rate can not be measured online in real time in the cognitive system, and the inaccuracy is discussed. Based on rough set theory, a model structure model and operation mechanism for establishing a cognitive object of uncertain information are explored based on rough set theory. Based on the generalized error theory and entropy theory, the multiple cognitive errors are explored. Based on the feedback control idea and the cognitive knowledge granularity computing tool, a cognitive intelligence operation mechanism with mimic feedback mode is established based on the feedback control idea and the cognitive knowledge granularity computing tool. The cognitive intelligence mechanism and the calculation model of the thinking information interaction model are compared. The main achievements of this paper are as follows: (1) research on the structure and operation mechanism of the imitative feedback cognitive intelligence system with uncertain information cognitive objects. The structure mode and operation mechanism of the three layer and three segment mutual coupling feedback cognitive intelligent system with repeated interaction mode are given, and the construction target of the imitation feedback cognitive intelligent system is given, the functional requirements and the operating mechanism are standardized, and the model structure is constructed for the imitation and recognition intelligent system of the uncertain information cognitive object. Method. (2) research on cognitive intelligence decision information system model of uncertain information cognitive object. Aiming at the problem that there is no common knowledge acquisition method that can adapt to all kinds of data in the current academic circle, the multi information fusion of uncertain information cognitive objects based on rough set is constructed and the intelligent decision information of supervised learning cognition is constructed. The system provides an index of cognitive knowledge fusion sufficiency and classification entropy function form based on characteristic simplicity. Under the uncertain information condition of finite domain, it provides a universal cognitive intelligence decision system model for the cognitive knowledge optimization representation of uncertain information cognitive objects. (3) the effect of uncertain cognitive process and Research on the generalized cognitive error evaluation system of cognitive results. In view of the fact that the current academic circles generally adopt the method of cognitive results reliability of posterior evaluation to meet the online real-time evaluation of recognition rate and correct rate in the cognitive system, the cognitive errors of uncertain information cognitive objects and uncertain cognitive process effects and cognitive results are constructed. The information system defines the uncertain cognitive process effect and the cognitive result error entropy based on the entropy function measure index. It does not determine the three generalized cognitive errors of the cognitive process effect and the cognitive result error entropy sequence and the cognitive knowledge granularity error, and establishes the online reality of the uncertainty recognition process effect and the cognitive result reliability. The time evaluation system provides a quantitative standard for the feedback cognitive intelligent adjustment mechanism. (4) the study of the imitating feedback cognitive intelligence mechanism of the uncertain information cognitive objects. The mimic feedback cognitive intelligence regulation mechanism which is poor online real-time measurement, defines the rules and calculation methods of the imitation feedback cognitive knowledge granularity, and establishes a multi-level variable knowledge granularity interlayer adjustment and the layer optimization model, and realizes the imitation feedback cognitive intelligent adjustment mechanism for the variable knowledge granularity of the uncertain information cognitive objects. Cognitive things provide an intelligent cognitive mechanism for the interactive process of information interaction. (5) the research and verification of the machine cognitive application of the off-line handwritten Chinese character image with a single goal and complete state. The recognition problem of the handwritten Chinese characters is large, the font font is numerous, the similar characters are many, and the writing style is different. The simulation feedback cognitive intelligence system of off-line handwritten Chinese character image with cognitive knowledge granularity intelligent adjustment mechanism is studied. The simulation experiment is carried out to verify the feasibility and effectiveness of this method by using the multi class Chinese character sample images of the SCUT-IRAC HCCLIB handwritten Chinese character image sample library as the cognitive object. The results show that this method has obvious advantages over the traditional one-way open loop method. A new practical application method is explored for the realization of "machine literacy" instead of "artificial literacy" for the off-line handwritten Chinese character image. (6) the machine cognitive application verification research on the burning state of industrial rotary kiln with multiple targets and complete state. Based on the imitation feedback cognitive intelligence method proposed in this paper, the imitation feedback cognitive intelligence system of the firing state of industrial rotary kiln based on flame image is studied based on the proposed simulation feedback cognitive intelligence method. The 2# rotary kiln firing zone of a cement plant is selected from the State Key Laboratory of the integrated automation of the process industry of Northeastern University. The flame image sample and the measured data are verified by MATLAB simulation experiment. The experimental results show that the method can excavate the characteristics of flame image from macro to micro and recognize the samples near the decision surface repeatedly, so as to realize the monitoring of the cement sintering process and the closed loop control of the clinker quality. A new method of engineering application is explored in place of "fire" instead of "fire". (7) a machine cognitive application verification study of human health status assessment of multi objective incomplete state. The imitation feedback cognitive intelligence system of human health state based on variable cognitive knowledge is studied. The feasibility and effectiveness of this method are verified by using MATLAB simulation experiments to verify the feasibility and effectiveness of this method. The results show that this method can be in different cognitive knowledge. To characterize health knowledge and fuzzy matching health indicators in the hierarchy of granularity, a new method is explored for the objective and accurate evaluation of human health status by "machine diagnosis" instead of "artificial diagnosis". (8) application verification research on intelligent recognition and tracking control of all weather photovoltaic power generation in single target incomplete state. The problem of photovoltaic cell tracking control is caused by nonlinear, non stationary, large intermittency and randomness. Based on the simulated feedback cognitive intelligence method proposed in this paper, the intelligent cognitive tracking control system for all weather photovoltaic power generation is studied, and a calculation method based on panoramic image variable granularity feedback is given. The experimental research is carried out to verify the feasibility and effectiveness of this method. The experimental results show that the model of the intelligent cognitive tracking control system of three layers and three segments of photovoltaic power generation can be based on the environment and system state of wind, snow, day, night, cloudy / rain and cloudy / cloudy. The cognitive knowledge carries out the corresponding control strategy and improves the photovoltaic efficiency based on the accurate calculation of the sun azimuth in the clear / multi cloud tracking control. A new intelligent control application method is explored to realize the high efficiency, energy saving and safe operation of the photovoltaic power generation system.
【学位授予单位】：合肥工业大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP18

【相似文献】