Invention Reference Number
This technology introduces a novel approach to process control that reduces the time, expertise, and manual effort required to design and calibrate advanced controllers. By combining simulation-based learning with distributed, real-world data, the system continually refines control performance. The approach enables adaptive, data-driven process control that improves over time without the need for constant human oversight.
Description
Conventional methods of developing and tuning controllers for complex process systems demand extensive expertise and hands-on adjustments. This technology leverages modern data-driven strategies to streamline and enhance that process. Initially, a model of the system is developed using previously collected operational data to mimic the system’s dynamics. The controller is first trained in a simulated environment to achieve desired behavior. It is then deployed across multiple real-world systems that each refine the controller based on their own operating data. Through a federated learning framework, these local improvements are securely aggregated to produce a continuously improving global controller model. The approach minimizes human intervention, lowers barriers to advanced process control, and promotes continuous optimization across distributed systems.
Benefits
- Reduces expertise and manual effort required for controller design
- Enables continual performance improvement using real-world data
- Lowers cost and time associated with advanced control development
- Enhances adaptability of industrial systems to process changes
Applications and Industries
- Manufacturing and automation
- Process control and industrial systems
- Equipment and systems integration
- Smart factory and Industry 4.0 environments
Contact
To learn more about this technology, email partnerships@ornl.gov or call 865-574-1051.
