Grade Control Optimization in Open Pit Mining using Supervised Machine Learning Approach
Loading...
Authors
Potakey, Noble Eyiram
Date
2024-08-28
Type
thesis
Language
eng
Keyword
Grade Control, Ore and Waste classification, Machine Learning, Open Pit Mining
Alternative Title
Abstract
Ore and waste classification in grade control is a critical aspect of mining operations to define which material is considered as ore or waste. Geostatistical techniques such as kriging and conditional simulation are used to estimate grades between sample points, followed by the application of a cut-off grade to discriminate the materials. Misclassification occurs when the estimated grades are not consistent with the actual grade distribution leading to ore loss and waste dilution. With the rapid evolution of computers and technology, including advancements in data processing and predictive capabilities, a machine learning aided classification and a corresponding evaluation system are proposed.
In this research, Support Vector Classification (SVC) is used as a machine learning approach to optimize material classification, alongside an industrial approach based on Ordinary Kriging (OK). Sequential Gaussian Simulation (SGSIM) is used to simulate the ground truth against which the two grade control models (SVC and OK) are benchmarked across two different open pit projects. Both models demonstrated decent results, achieving close to 90% accuracy in their predictions. However, the Kriging model exhibited approximately 8.5% misclassification of blocks, while the Support Vector Machine (SVC) showed a slightly higher misclassification rate of about 10.5% across the two scenarios.
The impact of sampling errors, another source of misclassification, is evaluated by introducing five levels of random noises to the original data. On average, the performances of both Kriging and SVC models reduced by about 4 % when sampling errors were introduced into the data. An economic evaluation of the lost ore due to misclassification in both models showed lower metal losses in OK than in SVC model, which increased in OK as the sampling error level increased.
Description
Citation
Publisher
License
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution-NonCommercial 4.0 International
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution-NonCommercial 4.0 International