Consistent Optimization of Blast Furnace Ironmaking Process Based on Controllability Assurance Soft Sensor Modeling
Junfang Li1, Chunjie Yang1, Chong Yang1
1State Key Laboratory of Industrial Control Technology, College of Control Science and Engineering, Zhejiang University, Hangzhou 310027, China.
Sensors (Basel, Switzerland)
|June 24, 2022
Summary
This study introduces a new framework for optimizing blast furnace ironmaking by accounting for uncontrollable variables, improving prediction accuracy and providing reliable optimization solutions for steel manufacturing.
Area of Science:
- Metallurgical Engineering
- Process Control
- Data-Driven Optimization
Background:
- Blast furnace ironmaking is crucial for steel production, offering significant economic and environmental benefits through process optimization.
- Existing data-driven methods struggle with uncontrollable variables, leading to uncertainty and reduced optimization effectiveness.
- Addressing this challenge requires a novel approach to predictive modeling in industrial processes.
Purpose of the Study:
- To develop a robust optimization framework for blast furnace ironmaking that explicitly handles variable uncontrollability.
- To enhance prediction accuracy in soft sensor modeling by integrating controllability assurance.
- To provide reliable uncertainty estimates for optimization solutions in complex industrial settings.
Main Methods:
- A consistency optimization framework using controllability assurance soft sensor modeling.
- Information extraction of uncontrollable variables via process supervision.
- An integrated self-encoder regression module to guide encoding and construct latent features.
- Multi-objective gray wolf optimizer coupled with the prediction module.
Main Results:
- Improved posterior distribution prediction accuracy by effectively extracting information from uncontrollable variables.
- Enhanced model prediction accuracy through the self-encoder regression module guiding latent feature construction.
- Successful optimization of the blast furnace ironmaking process using only controllable variables as inputs.
- Demonstrated capability of providing uncertainty estimates for optimization solutions.
Conclusions:
- The proposed framework effectively addresses the uncertainty caused by uncontrollable variables in blast furnace optimization.
- The integrated self-encoder regression module significantly boosts prediction accuracy.
- The method provides a reliable and validated approach for optimizing critical industrial processes like ironmaking.
