基于非高斯信息的JITL软测量模型

为了有效监控具有非高斯数据特性的工业过程,提出了一种新的基于非高斯信息的JITL(Just-In-Time Learning)软测量模型.首先通过非高斯非相似度测量选择JITL局部建模样本;然后建立局部ICA-PLS回归模型实现工业过程质量变量监控.该方法从局部建模样本选择到局部回归模型建立能够有效处理工业过程数据的非高斯特性,并且保留了JITL建模的优点,能够有效地处理工业过程时变特性以及非线性.通过硫回收处理过程的应用,验证了方法的有效性.

Non-Gaussian Information Based JITL Soft Sensor Model

Abstract： In order to monitor the non-Gaussian industrial process, a novel non-Gaussian information based JITL soft sensor model was proposed in this paper. First, the non-Gaussian dissimilarity measure selects the most relevant local modeling samples of JITL model. Then, an ICA-PLS regression method was established on the most relevant local samples for quality variable prediction. From the local relevant sample selection to the final regression model construction, the proposed method can efficiently extract the higher-order statistical information and is well suited for the non-Gaussian process quality prediction. Meanwhile, the proposed method can well cope with the changes in process characteristics as well as nonlinearity. The validity of the proposed method was verified on the sulfur recovery unit.