<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>The Golden Girl dolomite quarry was selected by the authors to develop predictive artificial neural network (ANN) models and software for optimization of blast fragment size distribution. Blast images from the quarry were analysed using WipFrag©. Seven controllable and two uncontrollable blast parameters, and WipFrag© image analysis results for fifty blasts were used to train ANN models. The reliability of the established models was tested, and the Bayesian regularization algorithm with the architecture of 9-8-3 was found to be superlative. The superlative model was compared with the modified Kuz-Ram model and found to be accurate. The optimum ANN models were translated into mathematical formulas and used to develop user-friendly software called BlastFrag optimizer. The software was validated with R<sup>2</sup> greater than 80% for all models and was found suitable for predicting blast fragment size distribution. The optimized result revealed that percentages for oversize and mean-size fragments were reduced from 68.4% and 418 mm to 27.83% and 101.6 mm, respectively, and undersize fragments increased from 50% to 72.17%.</p>
</abstract>

The Golden Girl dolomite quarry was selected by the authors to develop predictive artificial neural network (ANN) models and software for optimization of blast fragment size distribution. Blast images from the quarry were analysed using WipFrag©. Seven controllable and two uncontrollable blast parameters, and WipFrag© image analysis results for fifty blasts were used to train ANN models. The reliability of the established models was tested, and the Bayesian regularization algorithm with the architecture of 9-8-3 was found to be superlative. The superlative model was compared with the modified Kuz-Ram model and found to be accurate. The optimum ANN models were translated into mathematical formulas and used to develop user-friendly software called BlastFrag optimizer. The software was validated with R2 greater than 80% for all models and was found suitable for predicting blast fragment size distribution. The optimized result revealed that percentages for oversize and mean-size fragments were reduced from 68.4% and 418 mm to 27.83% and 101.6 mm, respectively, and undersize fragments increased from 50% to 72.17%.

Improvement of Blast-induced Fragmentation Using Artificial Neural Network and BlastFrag© Optimizer Software

Materials and Geoenvironment

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Izboljšanje učinka vpliva miniranja na fragmentacijo z uporabo umetne nevtronske mreže in napredne programske opreme

The Golden Girl dolomite quarry was selected by the authors to develop predictive artificial neural network (ANN) models and software for optimization of blast...

Improvement of Blast-induced Fragmentation Using Artificial Neural Network and BlastFrag© Optimizer Software

Article Category: Original Scientific Article

Data publikacji: 04 wrz 2023

Zakres stron: 47 - 59

Otrzymano: 16 paź 2021

Przyjęty: 11 sty 2022

DOI: https://doi.org/10.2478/rmzmag-2021-0010

Słowa kluczowe
blasting, muck-pile, fragment size, ANN model, Bayesian regularization

© 2022 Blessing Olamide Taiwo et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Improvement of Blast-induced Fragmentation Using Artificial Neural Network and BlastFrag© Optimizer Software

Article Category: Original Scientific Article

Data publikacji: 04 wrz 2023

Zakres stron: 47 - 59

Otrzymano: 16 paź 2021

Przyjęty: 11 sty 2022

DOI: https://doi.org/10.2478/rmzmag-2021-0010

Słowa kluczoweblasting, muck-pile, fragment size, ANN model, Bayesian regularization

© 2022 Blessing Olamide Taiwo et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Słowa kluczowe
blasting, muck-pile, fragment size, ANN model, Bayesian regularization