An Improved Quality-related Statistical Process Monitoring Method Based on Global Plus Local Projection to Latent Structures (GPLPLS)

PLS is widely used in the quality control process system, but it has poor capability in some strong local nonlinear system for fault diagnosis. To enhance the monitoring ability of such type fault, a novel statistical model based on global plus local projection to latent structures (GPLPLS) is proposed. Firstly, the characteristics and nature of quality-related global and local partial least squares (QGLPLS) are carefully analyzed, where its principal components preserve the local structure information in their respective data sheets as large as possible but not the correlation. In order to establish a quality-related model, this paper focuses more attention on the relevance of extracted principal components. Then, a quality-related monitoring strategy is established not only has the ability of PLS to extract the maximum linear relevant information but also the local nonlinear structural relevant information between the process variables and quality variables. Finally, the validity and effectiveness of GPLPLS-based statistical model are illustrated through two sets of artificial three-dimensional data of S-curve and Tennessee Eastman process (TEP) simulation platform. The experimental results demonstrate that the proposed model can be maintained the local property of the original data as much as possible and get a better monitoring result compared with PLS and QGLPLS.