Moving window KPCA with reduced complexity for nonlinear dynamic process monitoring

Area of Science:

• Chemical Engineering
• Data Analysis
• Systems Engineering

Background:

• Nonlinear dynamic systems present significant challenges in modeling and analysis.
• Kernel Principal Component Analysis (KPCA) is a powerful tool for nonlinear dimensionality reduction.
• Existing RKPCA methods may require improvements for dynamic system applications.

Purpose of the Study:

• To propose an enhanced Reduced Kernel Principal Component Analysis (RKPCA) method for nonlinear dynamic systems.
• To introduce the Moving Window Reduced Kernel Principal Component Analysis (MW-RKPCA) technique.
• To demonstrate the effectiveness of MW-RKPCA on a benchmark process.

Main Methods:

• Approximating principal components (PCs) using a reduced dataset for RKPCA model elaboration.
• Implementing a moving window approach to dynamically update the RKPCA model.
• Applying the MW-RKPCA technique to the Tennessee Eastman process.

Main Results:

• The MW-RKPCA effectively approximates system behavior within a moving window.
• The proposed method demonstrates improved handling of nonlinear dynamic systems compared to standard RKPCA.
• Validation on the Tennessee Eastman process showcases the practical relevance of MW-RKPCA.

Conclusions:

• MW-RKPCA offers a robust approach for analyzing nonlinear dynamic systems.
• The moving window update mechanism enhances model adaptability and accuracy.
• This technique provides a valuable tool for process monitoring and control in chemical engineering.