基于改进RPCA的连铸过程数据可靠性研究

Research on the Data Reliability for Continuous Casting Process Based on Improved RPCA

连铸过程数据的可靠性是连铸二冷动态控制系统成功应用的前提。针对正常连铸冷却过程慢时变、非稳态和多输入多输出特性,提出了基于改进递归主元分析方法监控连铸二冷配水过程数据的可靠性,这其中包括了用于建模的协方差阵、主元和两种报警控制限的递归更新,并通过小波除噪消除样本数据的噪声和奇异点。现场数据仿真结果表明应用此方法进行连铸过程监控可以明显减少误报的发生。

The reliability of data in continuous casting process is prerequisite for a successful application of dynamic secondary cooling control system in the continuous casting of steel. In allusion to the characteristic of slow time-varying, non-stationary and MIMO for normal cooling process in continuous casting, an improved recursive principal component （RPCA） was presented to monitor the reliability of data in secondary cooling process. It included the recursive update of covariance matrix, principal component and two kinds of alarm control limits. The samples was pre-processed to remove noise and spikes through wavelet de-noising. Simulation results, which are based on real-world historical data, show that the application of the presented algorithms can lead to a considerable reduction in the number of false alarms.