一种新型融合离群点的稳态检测方法

A New Steady-state Detection Method for Fusing Outliers

针对带有离群点的数据稳态检测,采用分布图法对离群点进行剔除;为了保持数据的完整性,提出用灰色预测值替代离群点值;最后用3δ法则进行稳态检验。如此,数据的稳态与非稳态便会区分开来。与现有稳态检测方法相比,分布图法快速有效地克服了离群点对稳态检测结果不准确的影响,降低了过程中个别异常数据带来的误诊率;灰色预测方法使离群点的替代值更贴近真实值,从而得到的过程数据比现有方法得到的数据更可靠。仿真结果证实了该方法的有效性和优越性。

Considering the detection of data＇s steady state with outliers,the distribution method was proposed to exclude outliers;and in order to maintain integrity of the data,having outliers replaced with predicted data based upon grey theory was proposed and then having their steady state tested with 3未 method so as to distinguish the data＇s steady and unsteady state.Compared with existing detection methods,the distribution method can quickly and efficiently overcome the outliers＇ influence on detection results of steady state and can reduce the rate of misdiagnosis caused by individual abnormal data.The use of gray prediction method can make outlier＇s replacement value more close to the true value,and the obtained process data are more reliable than that obtained through original methods.The simulation results prove effectiveness and superiority of this method.