论文部分内容阅读
传统基于多变量指数移动加权平均的独立元分析(MEWMA-ICA)方法:虽然可以有效监测化工过程中的缓变故障,但对突变故障监测滞后。本文提出了限定记忆模式的MEWMA-ICA方法:。该方法:通过引入基于记忆代数的遗忘因子,构造动态建模数据集。随后利用动态数据集建立ICA统计分析模型。最后采用T~2和SPE统计量对系统进行监测。其中,遗忘因子和记忆代数是通过优化算法进行选择的。通过测试问题和裂解气压缩机的仿真研究,验证了限定记忆模式的MEWMA-ICA方法:在保证传统MEWMA-ICA算法对缓变故障监测能力的同时,减少了对突变故障监测的时滞。
Traditional MEWMA-ICA method based on multivariate exponential moving weighted averaging: Although the slow fault in chemical process can be effectively monitored, the monitoring of catastrophic failure lags behind. This paper presents a MEWMA-ICA approach that defines memory patterns:. The method: By introducing a forgetting factor based on memory algebra, a dynamic modeling data set is constructed. Then use the dynamic data set to establish ICA statistical analysis model. At last, T ~ 2 and SPE statistics are used to monitor the system. Among them, the forgetting factor and memory algebra are selected through the optimization algorithm. The MEWMA-ICA method with limited memory mode is validated by the test and the simulation study of cracked gas compressor. It can reduce the time lag of the monitoring of catastrophic fault while guaranteeing the ability of traditional MEWMA-ICA to monitor the gradual change of fault.