置信规则库专家系统建模方法的研究与应用

发布时间：2018-02-06 00:30

本文关键词： 置信规则库粒子群算法油品检测透气度检测故障诊断　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：为了综合利用人类在决策过程中的不确定信息和定性知识,实现复杂决策问题建模的要求,置信规则库(Belief Rule Base,BRB)专家系统应运而生。和传统IF-THEN规则相比,BRB在规则中加入了置信框架,使得可以更充分的利用各种类型的数据知识,挖掘输入输出之间的非线性信息,从而实现复杂决策问题的建模。置信规则库专家系统以其智能性和多信息知识表达等优势,备受学者们的关注。置信规则库允许专家直接介入,而专家知识具有主观性,加上实际工程中的问题多是复杂的,这些都给准确快速设置置信框架的参数和结构带来了挑战。为了解决以上问题,本论文对置信规则库专家系统的建模方法进行了研究。论文的主要工作如下:(1)针对置信规则库优化模型求解效率差的问题,采用了一种基于粒子群智能算法的BRB参数训练方法。以食用油掺伪检测问题为背景,对所提方法进行了验证。相对于传统参数优化策略,粒子群优化算法明显提高了油品检测BRB模型的求解效率。(2)为了克服在确定BRB结构时专家知识的局限性,通过将前提属性的参考值和输出评价等级的效用作为推理模型中的待估计参数,提出了结构和参数同时优化的 BRB 模型(Optimize Structure and Parameters of BRB,OSP-BRB)。以烟草打孔水松纸透气度为研究对象,与相关BRB结构辨识的方法相比,OSP-BRB更加真实的反映了透气度的实际情况,证明了该方法可以更合理的构建BRB结构。(3)针对多决策因子引起的BRB规模过大的问题,基于属性重要度,增加前提属性权重的优化,提出了 BRB约减模型(BRB-reduction,BRB-R)。以油浸式变压器故障诊断为例,该方法缩减BRB规模的同时,将故障诊断的正确率提高了三个百分点,说明该方法是一种有效的属性约减方法。针对置信规则库专家系统在建模时存在的不足,从置信规则库参数和结构辨识,以及规模约简三个方面进行了研究改进,在三个领域的应用效果说明了文章改进后的建模方法可以有效克服专家知识局限性,较准确的设置置信规则库的置信框架,具有重要的工程实用价值。
[Abstract]:In order to comprehensive utilization of human beings in the decision-making process of uncertain information and qualitative knowledge, realize the modeling of complex decision problems, belief rule base (Belief Rule, Base, BRB) expert system came into being. Compared with the traditional IF-THEN rule, BRB added confidence in the framework of the rules, can make various types of data and make full use of knowledge mining, nonlinear information between input and output, so as to realize the modeling of complex decision problems. The belief rule base expert system to the intelligence and information knowledge expression and other advantages, has been the concern of scholars. The letter rules allowing experts directly involved, and expert knowledge is subjective, and the problem in practical engineering is these are complex, to quickly and accurately set confidence frame parameters and structural challenges. In order to solve the above problems, this paper on the belief rule base of expert system. Model method is studied. The main contents are as follows: (1) according to the belief rule base optimization model of solving the problem of poor efficiency, using a BRB parameter training method based on particle swarm algorithm. The edible oil adulteration detection problem as the background, the proposed method is verified. Compared with the traditional parameter optimization strategy of particle swarm optimization algorithm significantly improves the solving efficiency of oil detection. The BRB model (2) in order to overcome BRB in determining the structure of expert knowledge limitations, the premise of attribute reference value and output rating utility as the reasoning model of the parameters to be estimated, BRB model is put forward to optimize the structure and at the same time the parameters (Optimize Structure and Parameters of BRB, OSP-BRB). The tobacco perforated tipping paper permeability as the research object, comparing with the BRB structure identification, OSP-BRB reflect. The bearing of the actual situation, the result shows that this method can construct BRB structure more reasonable. (3) the decision factor of BRB caused by large scale, based on attribute importance, increase and optimize the weights of attributes, proposed BRB reduction models (BRB-reduction, BRB-R). The oil immersed transformer fault diagnosis an example, the method to reduce the size of the BRB at the same time, the fault diagnosis accuracy is improved by three percentage points, indicating that this method is an effective attribute reduction method. Aiming at the problems in modeling the belief rule base of expert system, from the belief rule base structure and parameters identification are studied and improved size reduction three, the application effect in three areas that the modeling of the improved method can effectively overcome the limitations of expert knowledge, confidence framework accurately set the belief rule base, has important practical engineering Use value.

【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP182

【参考文献】