汽轮发电机组故障诊断的“规则+例外”知识获取模型  

Rule-plus-exception model of knowledge extraction for fault diagnosis of turbine-generator unit

作　　者：黄文涛[1,2] 王伟杰[3] 赵学增[3] 孟庆鑫 

机构地区：[1]哈尔滨工程大学机电工程学院 [2]哈尔滨工业大学机电工程学院,黑龙江哈尔滨150001 [3]哈尔滨工业大学机电工程学院

出　　处：《电力自动化设备》2007年第12期27-31,共5页Electric Power Automation Equipment

基　　金：中国博士后科学基金资助项目(20070410888)~~

摘　　要：在分析了故障诊断样本实例集规律性的基础上,根据认知心理学和机器学习中的"规则+例外"模型的不足,结合粗糙集理论中知识约简的概念,提出了一种适合从包含不一致信息的故障诊断数据集中获取决策规则的改进的"规则+例外"模型,给出了模型的具体描述,研究了模型的基本结构,并结合汽轮发电机组振动故障诊断实例说明了改进的"规则+例外"故障诊断知识获取模型的可行性和有效性,通过与其他方法的比较,表明了改进的"规则+例外"故障诊断知识获取模型由于采用了一小部分例外,将实例集划分为2个部分,使得获取的规则无论在置信度上,还是泛化能力和简洁性上都要优于直接对实例集进行处理的方法。An improved rule-plus- exception model of cognitive psychology and machine learning is proposed based on the analysis of fault diagnosis samples regularity and rough set reduction technique,which is suitable for extracting the decision rules from the fault data containing inconsistent information. It is described in detail with its essential structure,and the example of turbine- generator unit fault diagnosis proves its feasibility and availability in which a short list of exceptions are considered,and the sample set is divided into two groups. Comparing the proposed model with the others,its superiority is proved not only on confidence,but also on the generalization ability and succinctness.

关 键 词：故障诊断 规则+例外 不一致 粗糙集 

分 类 号：TM311[电气工程—电机] TP182[自动化与计算机技术—控制理论与控制工程]