针对可能含有离群点的过程数据,提出一种融合离群点判别的稳态检测(Steady StateIdentification,SSID)方法,即基于新型3δ法则离群点判别与自适应多项式滤波(Adaptive Poly-nomial Filtering,APF)稳态检测相结合的方法。该方法首先根据历史稳态数据自适应地确定滤波窗口的长度;然后针对过程数据离群点的特点,采用提出的新型3δ法则滤除并替换窗口数据中的离群点;通过对消除离群点的窗口数据进行多项式滤波,得到反映该窗口内数据变化特征的曲线,并根据曲线的特征判断过程是否处于稳态。仿真研究与实际应用表明:融合离群点判别的稳态检测方法克服传统稳态检测方法中离群点对稳态检测结果的影响,检测结果明显优于传统的APF方法。
In this paper, a steady state identification method of containing outliers detection, i. e. , combining polynomial filtering steady state identification with the new 38 formula, is proposed for the process data containing outliers. In the proposed method, the length of filtering window is adaptively searched by means of the history data. And then, the outliers in the process data are filtered and replaced by using 38 formula. Finally, by making polynomial filtering to the present data in the window, the curve that shows the characteristic of the data is obtained. Moreover, the stability of the process is decided according to the characteristic of curve. Both simulation experiment and real application show that the proposed method performs better than traditional methods, e. g. , APF method.