一种基于新型数据预测方法的MICA仿真研究  

Simulation of MICA Based on a New Type of Data Prediction

作　　者：白丽娜[1,2] 高翔[2] 苑明哲[1] 曹景兴[3] 

机构地区：[1]中国科学院沈阳自动化研究所,辽宁沈阳110016 [2]沈阳化工学院信息工程学院,辽宁沈阳110142 [3]沈阳新松机器人自动化股份有限公司,辽宁沈阳110168

出　　处：《计算机仿真》2009年第2期134-138,共5页Computer Simulation

基　　金：国家高技术研究发展计划(863计划)(2006AA04Z185);辽宁省教育厅项目(2005320)

摘　　要：在多元统计过程监控中,为解决因未知过程数据统计分布而产生误报漏报的现象,提出一种结合多向独立元分析法(MICA)和广义相关系数(GCC)数据预测的综合方法,进行在线监控过程的仿真。MICA分析方法能有效分解各变量的关联关系,且不需考虑建模数据是否符合正态分布,用此方法计算的独立元变量能更好地描述过程的变化规律。为提高预报未来过程故障的能力,提出用广义相关系数法进行数据预测:确定与运行轨迹相似的监控模型库中的轨迹,并使其相应部分承接于运行轨迹之后。现场采集聚氯乙烯聚合过程的数据进行仿真,仿真结果显示:对于在线监控和在线故障诊断方面,这种新型预测方法优于其它传统处理预测问题的方法。In Multivariate Statistical process monitor, an approach of combining Multiway Independent Component Analysis with Generalized Correlation Coefficients （GCC） is presented to deal with the unknown statistic distribution of the data for overcoming the phenomenon of improper diagnosis and to carry out the simulation of online monitoring process. MICA could separate the correlation of variables without considering whether the model data follow the normal distribution or not, and the independent component variables are able to describe the variation of the process well. Furthermore, to improve the ability of predicting future process fault, GCC method is also presented for predicting unknown future data to ascertain the proper corresponding multi - trajectories in history model library and copy the part from the current time point as supplement for the running trajectories being tested. The data from polymerization process of polyvinyl chloride are sampled for simulation of online process monitoring, and the results show that the new prediction method is more effective than other traditional ones for online process monitoring and fault diagnosis.

关 键 词：广义相关系数法 间歇过程 多向独立元分析法 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]