Dynamic Batch Process Monitoring Based on Time-Slice Latent Variable Correlation Analysis.

Area of Science:

• Chemical Engineering
• Process Control
• Data Analytics

Background:

• Batch processes exhibit complex dynamics and data collinearity, making monitoring difficult.
• Accurate fault detection is crucial for ensuring the efficiency and safety of batch operations.

Purpose of the Study:

• To develop a data-driven fault detection framework for dynamic batch processes.
• To address challenges posed by complex dynamics and data collinearity in batch process monitoring.

Main Methods:

• Unfolding three-way batch data into two-way time slices.
• Mapping data to latent variable and residual subspaces for collinearity reduction.
• Utilizing canonical correlation analysis for measurement status determination.
• Generating prediction-based residuals to classify faults as static or dynamic.

Main Results:

• The proposed framework effectively handles variable-wise data collinearity.
• It extracts dominant information from process data.
• Faults are accurately detected and classified as static or dynamic.

Conclusions:

• The developed model predictive fault detection framework is feasible and effective for dynamic batch processes.
• The approach enhances process monitoring accuracy and fault identification capabilities.
• Validated on simulated penicillin production and industrial injection molding processes.