摘要
提出一种基于机器学习的混合知识获取方法,该方法结合了基于历史数据的规则提取方法和基于模型的规则提取方法。使用这两种方法提取规则,将其应用于对原油电脱盐系统的故障诊断中。实验结果表明,该方法能够有效的进行规则的提取,为故障诊断打下了良好的基础。其中基于历史数据的规则提取方法通过基于遗传算法的粗糙集约简来实现;基于模型的规则提取方法利用了符号有向图(SDG)的计算机自动推理结果,将因果图转化为规则。利用两种规则获取方法同时充实专家系统知识库,提供覆盖整个工艺流程的知识。
The objective of this paper is to present a hybrid method for knowledge acquisition which combines rule acquisition methods based on historical data with model-based methods. The former method derives rules through rough set reduction of a genetic algorithm, while the latter transforms the cause-effect graph to rules by using the automatic reasoning result of a signed directed graph （SDG）. Using the rules obtained by the above hybrid method to enrich the rules base provides knowledge covering the whole flow process. An example of the use of the method is given for an electric desalting system.
出处
《北京化工大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第5期89-93,共5页
Journal of Beijing University of Chemical Technology(Natural Science Edition)
基金
北京化工大学青年教师基金(QN0730)
关键词
知识获取
粗糙集约简
符号有向图
故障诊断
knowledge acquisition
rough set reduction
signed directed graph （SDG）
fault diagnosis
