针对化工连续生产过程的时序性及非线性等特征,提出一种新的基于数据驱动的化工过程故障诊断方法:ISOMAP-LDA。首先实行流形学习算法ISOMAP,在保持量测数据几何结构特性下完成非线性降维,然后基于提取的嵌入变量张成的低维空间,选用线...针对化工连续生产过程的时序性及非线性等特征,提出一种新的基于数据驱动的化工过程故障诊断方法:ISOMAP-LDA。首先实行流形学习算法ISOMAP,在保持量测数据几何结构特性下完成非线性降维,然后基于提取的嵌入变量张成的低维空间,选用线性判别分析(LDA)构造故障模式类的判别函数,负责各采样个体故障类型的判定。将该方法用于仿真化工Tennessee East man过程的故障诊断,结果表明,ISOMAP-LDA方法不仅拥有较高的故障诊断能力,而且取得采样在低维空间的可视化表示。展开更多
A visual method of fault diagnosis for a complicated process was developed based on self-organizing map (SOM).Due to the high dimensionality of the complicated process, principal component analysis (PCA) was introduce...A visual method of fault diagnosis for a complicated process was developed based on self-organizing map (SOM).Due to the high dimensionality of the complicated process, principal component analysis (PCA) was introduced to reduce the dimension of the process data.Then the self-organizing map was utilized to project the preprocessed data onto a 2D visualization space in which different process conditions were represented by different regions.Online monitoring could be achieved by the dynamic trajectory in the visualization space.The cause of certain fault could be deduced from the U-matrix of the derived SOM network and the loadings vector of the principal components.The application to the Tennessee Eastman process (TEP) demonstrated that fault detection and diagnosis could be carried out in a more intuitional and practical manner by using the proposed method.展开更多
文摘针对化工连续生产过程的时序性及非线性等特征,提出一种新的基于数据驱动的化工过程故障诊断方法:ISOMAP-LDA。首先实行流形学习算法ISOMAP,在保持量测数据几何结构特性下完成非线性降维,然后基于提取的嵌入变量张成的低维空间,选用线性判别分析(LDA)构造故障模式类的判别函数,负责各采样个体故障类型的判定。将该方法用于仿真化工Tennessee East man过程的故障诊断,结果表明,ISOMAP-LDA方法不仅拥有较高的故障诊断能力,而且取得采样在低维空间的可视化表示。
文摘A visual method of fault diagnosis for a complicated process was developed based on self-organizing map (SOM).Due to the high dimensionality of the complicated process, principal component analysis (PCA) was introduced to reduce the dimension of the process data.Then the self-organizing map was utilized to project the preprocessed data onto a 2D visualization space in which different process conditions were represented by different regions.Online monitoring could be achieved by the dynamic trajectory in the visualization space.The cause of certain fault could be deduced from the U-matrix of the derived SOM network and the loadings vector of the principal components.The application to the Tennessee Eastman process (TEP) demonstrated that fault detection and diagnosis could be carried out in a more intuitional and practical manner by using the proposed method.
