基于改进KSFA的非线性动态化工过程安全性研究  

Research on the Safety of Nonlinear Dynamic Chemical Processes Based on Improved KSFA

作　　者：刘鹏 张成[2] 邓成龙 Liu Peng;Zhang Cheng;Deng Cheng-long

机构地区：[1]沈阳化工大学环境与安全工程学院,辽宁沈阳110142 [2]沈阳化工大学理学院,辽宁沈阳110142 [3]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142

出　　处：《化工设计通讯》2023年第6期80-82,共3页Chemical Engineering Design Communications

基　　金：辽宁省教育厅基本科研项目(LJKMZ20220792,LJ2020021)。

摘　　要：将核慢特征分析(Kernel slow feature analysis,KSFA)与Wasserstein距离相融合,实现基于改进KSFA算法非线性动态过程故障检测,可以有效提高对非线性动态化工过程的监控效率,从而提高生产过程的安全性。采用KSFA方法提取过程中具有缓慢变化的核慢特征矩阵,利用滑动窗口技术获取慢特征空间系数矩阵Wasserstein距离,通过已获取的Wasserstein距离值作为统计值并进行过程检测,最后,将所提出的方法通过一个非线性动态数值例子进行验证,通过与SFA和KSFA进行对比分析,验证本方法在非线性动态过程中的可行性。Integrating Kernel slow feature analysis(KSFA)with Wasserstein distance to achieve nonlinear dynamic process fault detection based on improved KSFA algorithm can effectively improve the monitoring efficiency of nonlinear dynamic chemical processes,thereby improving the safety of production processes.Firstly,the KSFA method is used to extract the slowly changing kernel slow feature matrix during the process;Next,use sliding window technology to obtain the Wasserstein distance of the slow feature space coefficient matrix;Secondly,use the obtained Wasserstein distance value as the statistical value and perform process detection;Finally,the proposed method is validated through a nonlinear dynamic numerical example,and the feasibility of this method in nonlinear dynamic processes is verified through comparative analysis with SFA and KSFA.

关 键 词：核慢特征分析 非线性过程 动态过程 故障检测 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TQ050.7[自动化与计算机技术—控制科学与工程]