Manufacturing & Industrial Engineering Faculty Publications

Adaptive Transfer Learning for Mineral Grade Prediction in Mining Industry 4.0

Document Type

Conference Proceeding

Publication Date

2-2026

Abstract

The mining industry, characterized by dynamic operational environments and fluctuating mineral resources, faces a critical challenge in maintaining the adaptability of artificial intelligence systems for predictive modeling and control. In this context, the need arises for innovative methodologies that can effectively leverage transfer learning and online retraining techniques to enhance the accuracy, reliability, and real-time adaptability of predictive models in mineral processing, specifically in tasks such as flotation monitoring and control. By addressing the complexities of fluctuating mineral grades and process conditions, the research aims to bridge the gap between traditional predictive modeling approaches and the evolving demands of Industry 4.0 in the mining sector. We propose a novel online retraining framework that leverages transfer learning to predict and adapt to the fluctuating conditions of mineral grades in flotation-based mineral processing. Through the evaluation and deployment of several transfer learning models, including ResNet, MobileNet, GoogleNet, and DenseNet, our approach emphasizes ongoing data analysis and model recalibration to confront the intrinsic variability of mineral resources. The findings validate the improved performance and dependability of our model, propelling us towards intelligent and autonomous monitoring systems. Beyond demonstrating the capabilities of adaptive AI in industrial settings, this work paves the way for future advancements in smart, continuous monitoring through intelligent systems.

Comments

https://rdcu.be/e4Wv8

Publication Title

New Technologies, Artificial Intelligence and Smart Data

DOI

10.1007/978-3-032-14964-0_12

Link to Full Text

Share

COinS