Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models
Effective management of overbreak (OB) and underbreak (UB) in underground stope production is critical in minimising dilution and ore loss, directly impacting the productivity and efficiency of mining operations. This study aims to predict these phenomena prior to blasting by analysing individual dr...
| Main Authors: | , , , , |
|---|---|
| Format: | Journal Article |
| Published: |
Elsevier
2025
|
| Online Access: | http://hdl.handle.net/20.500.11937/98041 |
| _version_ | 1848766353115185152 |
|---|---|
| author | Taheri, Sina Topal, Erkan Nguyen, H. Jang, H. Shirani Faradonbeh, Roohollah |
| author_facet | Taheri, Sina Topal, Erkan Nguyen, H. Jang, H. Shirani Faradonbeh, Roohollah |
| author_sort | Taheri, Sina |
| building | Curtin Institutional Repository |
| collection | Online Access |
| description | Effective management of overbreak (OB) and underbreak (UB) in underground stope production is critical in minimising dilution and ore loss, directly impacting the productivity and efficiency of mining operations. This study aims to predict these phenomena prior to blasting by analysing individual drillholes and eight key variables including ring sequence, drillhole dip, breakthrough status, powder factor, number of primers, rock quality designation, actual drillhole deviation, and drillhole length. Although the initial linear regression models provided moderate predictive accuracy, advanced Machine Learning (ML) techniques significantly improved results. Boosted tree models and Artificial Neural Networks (ANN) showed significant improvements in predictive performance after hyperparameter tuning, with the ANN model reaching an R2 of 0.94067 and an RMSE of 0.65527. Additionally, eight different hybrid models were developed, producing R2 values ranging from 0.59 to 0.94, achieved using various hybrid modelling techniques, including weighted average, stacking, and combinations of ANN and boosted trees. Meta-models such as Random Forest (RF) and Gradient Boosting Machines (GBM) were utilised to improve accuracy even further. The best-performing hybrid ensemble stacking model, combining RF and GBM meta-models using a weighted average approach (HESM-RF-GBM), achieved the highest performance with an R2 of 0.94502 and an RMSE of 0.45203. This study highlights the potential of ML and advanced hybrid models in optimising OB and UB predictions, leading to reduced costs and improved operational efficiency in underground mining. |
| first_indexed | 2025-11-14T11:49:47Z |
| format | Journal Article |
| id | curtin-20.500.11937-98041 |
| institution | Curtin University Malaysia |
| institution_category | Local University |
| last_indexed | 2025-11-14T11:49:47Z |
| publishDate | 2025 |
| publisher | Elsevier |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | curtin-20.500.11937-980412025-07-23T02:36:59Z Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models Taheri, Sina Topal, Erkan Nguyen, H. Jang, H. Shirani Faradonbeh, Roohollah Effective management of overbreak (OB) and underbreak (UB) in underground stope production is critical in minimising dilution and ore loss, directly impacting the productivity and efficiency of mining operations. This study aims to predict these phenomena prior to blasting by analysing individual drillholes and eight key variables including ring sequence, drillhole dip, breakthrough status, powder factor, number of primers, rock quality designation, actual drillhole deviation, and drillhole length. Although the initial linear regression models provided moderate predictive accuracy, advanced Machine Learning (ML) techniques significantly improved results. Boosted tree models and Artificial Neural Networks (ANN) showed significant improvements in predictive performance after hyperparameter tuning, with the ANN model reaching an R2 of 0.94067 and an RMSE of 0.65527. Additionally, eight different hybrid models were developed, producing R2 values ranging from 0.59 to 0.94, achieved using various hybrid modelling techniques, including weighted average, stacking, and combinations of ANN and boosted trees. Meta-models such as Random Forest (RF) and Gradient Boosting Machines (GBM) were utilised to improve accuracy even further. The best-performing hybrid ensemble stacking model, combining RF and GBM meta-models using a weighted average approach (HESM-RF-GBM), achieved the highest performance with an R2 of 0.94502 and an RMSE of 0.45203. This study highlights the potential of ML and advanced hybrid models in optimising OB and UB predictions, leading to reduced costs and improved operational efficiency in underground mining. 2025 Journal Article http://hdl.handle.net/20.500.11937/98041 10.1016/j.tust.2025.106852 http://creativecommons.org/licenses/by/4.0/ Elsevier fulltext |
| spellingShingle | Taheri, Sina Topal, Erkan Nguyen, H. Jang, H. Shirani Faradonbeh, Roohollah Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| title | Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| title_full | Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| title_fullStr | Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| title_full_unstemmed | Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| title_short | Predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| title_sort | predicting overbreak and underbreak in underground longhole stoping using meta-soft computing models |
| url | http://hdl.handle.net/20.500.11937/98041 |