Multidirectional KICA intermittent process fault monitoring method based on independent subspaces

Abstract

The invention relates to a multidirectional KICA intermittent process fault monitoring method based on independent subspaces. The method comprises the following steps: acquiring three-dimensional data X(I*J*K) of an intermittent process, applying multidirectional KICA, and expanding the three-dimensional data in batches into two-dimensional data to be processed, wherein I is the number of batches, J is the number of variables, and K is the number of sampling points; performing offline modeling on the two-dimensional data expanded in batches, adding a kernel trick on the basis of ICA, mapping non-linear data to high-dimensional feature spaces, and then performing linear processing in the high-dimensional spaces; performing online monitoring on the process through applications of statistics of T<2> and SPED; and summarizing fault information obtained by each sub data, and calculating whether the statistics exceed limits. A traditional KICA overall processing method is improved, initial data is divided into the multiple sub spaces to perform detailed KICA analyses, multi-model KICA simultaneous monitoring is established, hidden information is magnified, local information is got effectively, and the fault monitoring rate is improved.

Description

technical field

[0001] The invention relates to the technical field of fault detection and diagnosis, in particular to a multi-directional KICA intermittent process fault monitoring method based on independent subspaces. Background technique

[0002] ICA (Independent Component Correlation Algorithm, Independent Component Analysis) provides a method to solve the non-Gaussian data in the industrial production process, which improves the accuracy and universality of fault monitoring. The batch process is different from the ordinary process and has the characteristics of multiple batches, but there is no obvious distinction between the batches, and the time of each batch is not fixed. The production rate and production status of the batch process change irregularly. This kind of production situation makes it very difficult for workers to deal with, so it is hoped to reduce this irregular change, make the production process as regular as possible, and the monitoring of intermitt...

Examples

Embodiment Construction

[0124]The present invention will be further elaborated below in conjunction with the accompanying drawings of the description.

[0125] A kind of multi-directional KICA intermittent process fault monitoring method of independent subspace division of the present invention, comprises the following steps:

[0126] 1) Collect the 3D data X (I×J×K) of the intermittent process, apply multi-directional KICA, and expand the 3D data into 2D data in batches for processing;

[0127] 2) Carry out offline modeling of the two-dimensional data expanded by batches, add kernel techniques on the basis of ICA, map the nonlinear data to the high-dimensional feature space, and then perform linear processing in the high-dimensional space;

[0128] 3) Apply T 2 and SPE statistics to monitor the process online; summarize the fault information obtained from each sub-data, and calculate whether the statistics exceed the limit.

[0129] In step 1), such as figure 2 As shown, the three-dimensional pr...

technical field

[0001] The invention relates to the technical field of fault detection and diagnosis, in particular to a multi-directional KICA intermittent process fault monitoring method based on independent subspaces. Background technique

[0002] ICA (Independent Component Correlation Algorithm, Independent Component Analysis) provides a method to solve the non-Gaussian data in the industrial production process, which improves the accuracy and universality of fault monitoring. The batch process is different from the ordinary process and has the characteristics of multiple batches, but there is no obvious distinction between the batches, and the time of each batch is not fixed. The production rate and production status of the batch process change irregularly. This kind of production situation makes it very difficult for workers to deal with, so it is hoped to reduce this irregular change, make the production process as regular as possible, and the monitoring of intermitt...

Examples

Embodiment Construction

[0124]The present invention will be further elaborated below in conjunction with the accompanying drawings of the description.

[0125] A kind of multi-directional KICA intermittent process fault monitoring method of independent subspace division of the present invention, comprises the following steps:

[0126] 1) Collect the 3D data X (I×J×K) of the intermittent process, apply multi-directional KICA, and expand the 3D data into 2D data in batches for processing;

[0127] 2) Carry out offline modeling of the two-dimensional data expanded by batches, add kernel techniques on the basis of ICA, map the nonlinear data to the high-dimensional feature space, and then perform linear processing in the high-dimensional space;

[0128] 3) Apply T 2 and SPE statistics to monitor the process online; summarize the fault information obtained from each sub-data, and calculate whether the statistics exceed the limit.

[0129] In step 1), such as figure 2 As shown, the three-dimensional pr...

Images