A Novel Dynamic Principal Component Analysis Based on Block Dynamic Mode Decomposition for Chemical Process Fault...

Area of Science:

• Chemical Engineering
• Process Systems Engineering
• Data Science

Background:

• Dynamic Principal Component Analysis (DPCA) is crucial for process monitoring and fault detection.
• Existing DPCA methods struggle with intermittent noise, long-term dependencies, and subsystem interactions in chemical processes.

Purpose of the Study:

• To develop an enhanced methodological framework for DPCA to overcome its limitations in chemical process industries.
• To improve the accuracy and robustness of fault detection in dynamic systems.

Main Methods:

• Frequency-based clustering to filter equipment vibration noise.
• Matrix augmentation to capture long-term time dependencies.
• Block matrix dynamic mode decomposition to model subsystem coupling.

Main Results:

• The enhanced DPCA achieves competitive performance under moderate lag settings.
• Superior fault detection performance is observed under large lag settings.
• The method effectively detects correlated faults across multiple subsystems.

Conclusions:

• The proposed enhancements significantly improve DPCA's applicability and performance in complex industrial environments.
• Each component of the enhanced framework contributes positively to overall performance.
• The integrated approach offers robust fault detection for dynamic chemical processes.