摘要
符号有向图(SDG)深层知识模型具有好的完备性和较强故障解释能力,将主元统计法(PCA)和SDG两种方法结合起来,用SDG模型包含的过程信息来解释PCA方法产生的残差贡献图,能有效辨识故障,减少诊断时间,增加诊断过程自动化的程度;同时利用PCA分析建模可以消除变量间的非线性关系,降低噪声影响,有效地避免了传统SDG在确定节点状态和阈值时的单变量统计的缺点。案例研究表明:PCA-SDG定性定量方法可以进行有效的诊断。
A Sign Directed Graph＇s （SDG） deep going information model excels in completeness and fault explanation capability. Faults can effectively be identified, diagnosing time saved and the degree of diagnosing process＇ automation raised by combining SDG with the principal component analysis （PCA） method, whereby the process information stored in the former is used to interpret the residual contributions produced by the latter. On the other hand, the PCA＇s analyzing model can cancel the non-linear correlation among variables, reduce noise influences, as well as effectively avoid the shortcoming of single variable statistics in discribminating node conditions and threshold values that appear with traditional SDG models. Case studies show that the PCA-SDG qualitative/quantitative method can effectively serve diagnosing purposes. Figs 8, tables 2 and refs 9.
出处
《动力工程》
EI
CSCD
北大核心
2005年第6期870-875,共6页
Power Engineering
基金
华北电力大学博士学位教师科研基金(20041209)
关键词
自动控制技术
电站
故障诊断
符号有向图
主元统计法
定性定量模型
automatic control technique
power station
fault diagnosis
SDG Graph
PCA method
qualitative/quantitative model
