A comparative study of artificial neural networks in predicting blast-induced air-blast overpressure at Deo Nai open-pit coal mine, Vietnam

作者：Hoang Nguyen^*; Xuan-Nam Bui; Hoang-Bac Bui; Mai, Ngoc-Luan

来源：Neural Computing & Applications, 2020, 32(8): 3939-3955.

DOI：10.1007/s00521-018-3717-5

• Artificial neural network
• Multilayer perceptron neural network
• Bayesian regularized neural networks
• Hybrid neural fuzzy inference system
• Air-blast overpressure
• Deo Nai
• Vietnam

Air-blast overpressure (AOp) is one of the undesirable effects caused by blasting operations in open-pit mines. This side effect of blasting can seriously undermine surrounding residential structures and living quality. To control and mitigate this situation, this study using artificial neural networks to predict AOp implemented at Deo Nai open-pit coal mine, Vietnam. A total of 146 events of blasting were recorded, of which 80% (118 observations) was used for training and 20% (28 observations) was used for testing. A resampling technique, namely tenfold cross-validation, was performed with three repeats to increase the accuracy of the predictive models. In this paper, three different types of neural networks were developed to predict AOp including multilayer perceptron neural network (MLP neural nets), Bayesian regularized neural networks (BRNN) and hybrid neural fuzzy inference system (HYFIS). Each type was tested with ten model configurations to discover the best performing ones based on comparing standard metrics, including root-mean-square error (RMSE), coefficient of determination (R-2), and a simple ranking method. Eight parameters were considered for these models, including charge per delay, burden, spacing, length of stemming, powder factor, air humidity, and monitoring distance. The results indicated that MLP neural nets model with RMSE = 2.319, R-2 = 0.961 on testing datasets and a total ranking of 12 yielded the most accurate prediction over BRNN and HYFIS models.

• 出版日期2020-4