A KPI-based process monitoring and fault detection framework for large-scale processes

Area of Science:

• Industrial Engineering
• Process Control
• Data Analytics

Background:

• Large-scale industrial processes involve complex, interconnected subprocesses requiring robust performance evaluation.
• Existing key performance indicator (KPI)-based approaches lack a unified framework for process monitoring and fault detection (PM-FD).

Purpose of the Study:

• To propose a coherent and consistent framework for KPI-based PM-FD in large-scale industrial processes.
• To address limitations in current methods by considering both static and dynamic relationships between process and KPI variables.

Main Methods:

• Developed a least squares-based approach for static relationships, outperforming partial least squares.
• Utilized kernel representation and instrument variables to simplify dynamic relationships into a static case.
• Applied the framework to the TE benchmark process and the hot strip mill rolling process.

Main Results:

• The proposed framework effectively monitors large-scale industrial processes.
• The method demonstrates superior fault detection capabilities compared to existing techniques.
• Successful application to benchmark and real-world industrial processes validates the approach.

Conclusions:

• The developed framework provides a robust and consistent approach to KPI-based PM-FD.
• The integration of static and dynamic analysis enhances fault detection accuracy in complex industrial systems.
• This research offers a significant advancement in ensuring the reliable performance of large-scale industrial operations.