Data fusion strategies to combine sensor and multivariate model outputs for multivariate statistical process control.

Area of Science:

• Chemical Engineering
• Analytical Chemistry
• Process Control

Background:

• Process analytical technologies (PAT) generate multiple outputs from sensors and models for process monitoring.
• Current data fusion strategies for multivariate statistical process control (MSPC) can be improved.
• Combining diverse data sources is crucial for robust process management.

Purpose of the Study:

• To develop and evaluate data fusion strategies for integrating sensor and model outputs in MSPC.
• To demonstrate the application of these strategies using real-world process examples.
• To highlight the benefits of fused information for process control and diagnostics.

Main Methods:

• Exploration of data fusion strategies applied to three distinct industrial processes.
• Utilizing near-infrared (NIR) spectroscopy and temperature sensors.
• Development of MSPC models incorporating fused data, including multivariate calibration and resolution outputs.

Main Results:

• Demonstrated flexibility in combining various model outputs (e.g., property predictions, process profiles).
• Showcased significant improvements in MSPC models using fused information compared to single-sensor models.
• Validated the enhanced capabilities for process control, diagnostics, and interpretation of process deviations.

Conclusions:

• The proposed data fusion strategy is broadly applicable to analytical and bioanalytical processes.
• MSPC models based on fused information offer superior performance for process monitoring and control.
• Data fusion enhances the ability to diagnose and interpret abnormal process situations effectively.