An ensemble JITL method based on multi-weighted similarity measures for cold rolling force prediction.

Area of Science:

• Manufacturing Engineering
• Materials Science
• Artificial Intelligence

Background:

• Product quality and yield in cold tandem rolling depend heavily on accurate rolling force prediction.
• Fixed prediction models fail in dynamic, multi-operating condition environments.
• Limited process knowledge hinders effective sample selection using single similarity measures.

Purpose of the Study:

• To propose an ensemble just-in-time-learning modeling method (MWS-EJITL) for accurate rolling force prediction.
• To address limitations of fixed models and single similarity measures in complex rolling environments.
• To enhance product quality and yield through improved prediction accuracy.

Main Methods:

• Developed a multi-weighted similarity measures (MWS) approach for relevant sample selection.
• Constructed local models and estimated query data output.
• Employed an ensemble learning strategy to integrate local model predictions.
• Introduced a cumulative similarity factor for optimizing local model sample size.
• Implemented a similarity threshold for adaptive local model updates.

Main Results:

• The MWS-EJITL method demonstrated effectiveness and accuracy in rolling force prediction experiments.
• The proposed method successfully handles multi-operating conditions.
• Adaptive model updates and optimized sample selection were achieved.

Conclusions:

• The MWS-EJITL method offers a robust solution for rolling force prediction in cold tandem rolling.
• This approach enhances prediction accuracy and adaptability in dynamic manufacturing environments.
• Improved prediction accuracy contributes to better product quality and yield.