摘要
针对基于证据推理的置信规则库专家系统(RIMER)的学习优化问题,在训练参数仅为规则的置信度、权重的基础上,通过增加前提属性参考值及输出参考值为训练参数来实现局部扩展训练和全局训练,并使用MATLAB中FMINCON函数对参数进行优化求解.分别将该专家系统应用在发动机故障诊断和数据逼近中,并对其进行训练优化.结果表明,与局部扩展优化相比,全局优化时,RIMER系统能更好地模拟实际系统,对参数的优化程度越深,RIMER系统的精度越高.
To solve the optimization problem of belief rule based system based on evidential reasoning （RIMER）, on the basis of the training parameters of only belief degree and weight, referential values for antecedent attribute and output are added to achieve local expansion training and global training, and the FMINCON function in MATLAB software is used to solve the training parameters. Moreover, RIMER expert system is established and trained for fault diagnosis of engine and data approximation. The results show that the RIMER system can simulate the actual system better in the global training optimization. That is, if the degree of optimization of the parameters is deeper, the precision of RIMER system will be higher.
出处
《华北水利水电大学学报（自然科学版）》
2015年第4期72-78,共7页
Journal of North China University of Water Resources and Electric Power：Natural Science Edition
基金
河南省教育厅科学技术研究重点项目(14A120013)
关键词
置信规则库
专家系统
扩展局部优化
全局优化
belief rule base
expert system
local expansion optimization
global optimization
