Multi-Field Coupling- and Data-Driven-Based Optimization of Cooling Process Parameters for Planetary Rolling Rolls.

Area of Science:

• Materials Science and Engineering
• Mechanical Engineering
• Computational Fluid Dynamics

Background:

• High roll surface temperatures in three-roll planetary rolling cause copper adhesion, degrading roll quality and product yield.
• Efficient roll cooling is critical for maintaining surface integrity and enhancing production efficiency in copper tube manufacturing.

Purpose of the Study:

• To investigate the impact of spray cooling parameters on roll surface cooling performance.
• To develop and validate a fluid-solid-heat coupled model for roll cooling.
• To identify the most effective machine learning model for predicting cooling performance and optimize spray ring geometry.

Main Methods:

• Developed a fluid-solid-heat coupled model and validated it experimentally.
• Evaluated Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and Support Vector Machine (SVM) for predictive modeling.
• Employed Particle Swarm Optimization (PSO) to optimize spray ring parameters based on RF model predictions.

Main Results:

• The coupled model achieved high accuracy (max deviation 4.36%) compared to experimental data.
• The RF model outperformed GBDT and SVM, showing strong predictive capabilities (RMSE=1.73, MAE=1.32, R²=0.91).
• The optimized RF-PSO approach increased the heat transfer coefficient by 44.72%.

Conclusions:

• The developed coupled model accurately predicts roll surface cooling.
• Machine learning, specifically RF, offers an efficient alternative to complex simulations for predicting cooling performance.
• Optimized spray cooling parameters significantly enhance heat transfer, providing a foundation for improved precision tube manufacturing.