Research on the Prediction of Liquid Injection Volume and Leaching Rate for In Situ Leaching Uranium Mining Using the CNN–LSTM–LightGBM Model
Zhifeng Liu, Zirong Jin, Yipeng Zhou, Zhenhua Wei, Huanyu Zhang
Abstract
In traditional in situ leaching (ISL) uranium mining, the injection volume depends on technicians’ on-site experience. Therefore, applying artificial intelligence technologies such as machine learning to analyze the relationship between injection volume and leaching rate in ISL uranium mining, thereby reducing human factor interference, holds significant guiding importance for production process control. This study proposes a novel uranium leaching rate prediction method based on a CNN–LSTM–LightGBM fusion model integrated with an attention mechanism. Ablation experiments demonstrate that the proposed fusion model outperforms its component models across three key metrics: Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Square Error (RMSE). Furthermore, comparative experiments reveal that this fusion model achieves superior performance on MAE, MAPE, and RMSE metrics compared to six extensively utilized machine learning methods, including Multi-Layer Perceptron, Support Vector Regression, and K-Nearest Neighbors. Specifically, the model achieves an MAE of 0.085%, an MAPE of 0.833%, and an RMSE of 0.201%. This attention-enhanced fusion model provides technical support for production control in ISL uranium mining and offers valuable references for informatization and intelligentization research in uranium mining operations.