基于共同趋势模型的非平稳过程在线监控

发布时间：2018-04-15 16:44

本文选题：共同趋势模型 + 过程监控　；参考：《化工学报》2017年01期

【摘要】：对于非平稳过程监控,传统的基于数据平稳假设的多元统计过程控制方法是不适用的。针对上述问题,提出了一种基于共同趋势模型的非平稳过程监控方法。共同趋势模型从存在协整关系的非平稳多元变量中辨识出共同因子,将各非平稳过程变量分解成非平稳的共同趋势成分与平稳成分之和的形式。不同于现有的基于协整模型的非平稳过程监控方法,共同趋势模型能够获取各非平稳变量中的平稳成分,消除非平稳共同因子的影响并体现变量间全部的动态均衡关系。将对非平稳过程的监控变为应用共同趋势模型,分解得到各非平稳过程变量中的平稳成分,然后应用传统的多元统计方法,估计平稳成分的统计量及相应的控制限进行监测。石油蒸馏过程监控的实例研究结果表明,所提出的方法比基于协整新息变量的方法具有更可靠的监控效果。
[Abstract]:For non-stationary process monitoring, the traditional multivariate statistical process control method based on stationary assumption is not applicable.In order to solve the above problems, a method of monitoring non-stationary processes based on common trend model is proposed.The common trend model identifies common factors from non-stationary multivariate variables with cointegration relationship and decomposes each non-stationary process variable into the sum of non-stationary common trend components and stationary components.Different from the existing non-stationary process monitoring methods based on cointegration model, the common trend model can obtain the stationary components of the non-stationary variables, eliminate the influence of the non-stationary common factors and reflect the dynamic equilibrium relationship among the variables.The monitoring of non-stationary processes is transformed into a common trend model, and the stationary components in the variables of non-stationary processes are decomposed. Then, the statistics of stationary components and the corresponding control limits are estimated and monitored by using the traditional multivariate statistical method.The experimental results of oil distillation process monitoring show that the proposed method is more reliable than the one based on cointegration innovation variables.
【作者单位】：南京航空航天大学机械结构力学及控制国家重点实验室;
【基金】：江苏高校优势学科建设工程资助项目~~
【分类号】：TP277

【相似文献】