当前位置: X-MOL 学术Nat. Resour. Res. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Estimation of Ground Vibration Intensity Induced by Mine Blasting using a State-of-the-Art Hybrid Autoencoder Neural Network and Support Vector Regression Model
Natural Resources Research ( IF 4.8 ) Pub Date : 2021-06-22 , DOI: 10.1007/s11053-021-09890-w
Bo Ke , Hoang Nguyen , Xuan-Nam Bui , Romulus Costache

Abstract

In surface mining, blasting is an indispensable method for fragmenting rock masses. Nevertheless, it can inherently induce many side effects like ground vibrations. At high intensities, the ground vibrations generated because of blasting operations can destroy structures and buildings. Also, in areas with adverse geological conditions, such vibrations can cause bench and slope failures. Therefore, the accurate prediction of ground vibration intensity (GVI) has critical implications in mitigating and controlling the adverse effects along with sustainable development and responsible mining. In this research, a novel intelligent model was proposed to predict GVI based on the hybridization of autoencoder neural networks (AutoencoderNN) and support vector machine regression (SVR), and it was named AutoencoderNN-SVR. Nine input variables were utilized to estimate GVI: borehole diameter, bench height, borehole length, burden, spacing, hardness coefficient, powder factor, maximum explosive charged per delay, and monitoring distance. Two hundred ninety-seven blasting events were collected, analyzed, and evaluated to achieve this aim. Also, the traditional SVR model without the support of AutoencoderNN, an empirical equation (i.e., USBM), and a nonlinear model based on gene expression programing were applied in this research and compared with the proposed AutoencoderNN-SVR model in terms of GVI prediction. Then, the models' obtained results were analyzed and computed through statistical indices, such as root mean squared error (RMSE) and coefficient of determination (R2). The AutoencoderNN-SVR's ensemble model was found to have obtained the highest accuracy and lowest error (i.e., RMSE = 1.232 and R2 = 0.887) compared to the other models and is an insight in predicting GVI in mine blasting with high reliability.

Highlights

  • An autoencoder neural network was investigated to predict GVI in mine blasting;

  • An autoencoder neural network was combined with support vector regression to generate a robust hybrid model (AutoencoderNN-SVR) to predict GVI in mine blasting;

  • The proposed AutoencoderNN-SVR model was compared with the empirical, SVR, and GEP models;

  • The proposed AutoencoderNN-SVR model was introduced as a novel and robust technique for predicting GVI with high accuracy.



中文翻译:

使用最先进的混合自动编码器神经网络和支持向量回归模型估计由矿山爆破引起的地面振动强度

摘要

在露天采矿中,爆破是破碎岩体必不可少的方法。然而,它本质上会引起许多副作用,如地面振动。在高强度下,爆破作业产生的地面振动会破坏结构和建筑物。此外,在地质条件恶劣的地区,这种振动会导致平台和斜坡的破坏。因此,准确预测地面振动强度(GVI)对于减轻和控制不利影响以及可持续发展和负责任采矿具有重要意义。在这项研究中,提出了一种基于自动编码器神经网络(AutoencoderNN)和支持向量机回归(SVR)混合的新型智能模型来预测GVI,并将其命名为AutoencoderNN-SVR。九个输入变量用于估计 GVI:钻孔直径、台架高度、钻孔长度、载荷、间距、硬度系数、粉末系数、每次延迟装药的最大炸药量和监测距离。为了实现这一目标,我们收集、分析和评估了 297 个爆破事件。此外,本研究应用了没有AutoencoderNN支持的传统SVR模型、经验方程(即USBM)和基于基因表达编程的非线性模型,并在GVI预测方面与提出的AutoencoderNN-SVR模型进行了比较。然后,通过统计指标,如均方根误差(RMSE)和决定系数(硬度系数、粉末系数、每次延迟装药的最大炸药量和监测距离 为了实现这一目标,我们收集、分析和评估了 297 个爆破事件。此外,本研究应用了没有AutoencoderNN支持的传统SVR模型、经验方程(即USBM)和基于基因表达编程的非线性模型,并在GVI预测方面与提出的AutoencoderNN-SVR模型进行了比较。然后,通过统计指标,如均方根误差(RMSE)和决定系数(硬度系数、粉末系数、每次延迟装药的最大炸药量和监测距离 为了实现这一目标,我们收集、分析和评估了 297 个爆破事件。此外,本研究应用了没有AutoencoderNN支持的传统SVR模型、经验方程(即USBM)和基于基因表达编程的非线性模型,并在GVI预测方面与提出的AutoencoderNN-SVR模型进行了比较。然后,通过统计指标,如均方根误差(RMSE)和决定系数(本研究将没有AutoencoderNN支持的传统SVR模型、经验方程(即USBM)和基于基因表达编程的非线性模型应用于本研究,并与提出的AutoencoderNN-SVR模型在GVI预测方面进行了比较。然后,通过统计指标,如均方根误差(RMSE)和决定系数(本研究将没有AutoencoderNN支持的传统SVR模型、经验方程(即USBM)和基于基因表达编程的非线性模型应用于本研究,并与提出的AutoencoderNN-SVR模型在GVI预测方面进行了比较。然后,通过统计指标,如均方根误差(RMSE)和决定系数([R 2)。 与其他模型相比,发现 AutoencoderNN-SVR 的集成模型获得了最高的准确度和最低的误差(即,RMSE = 1.232 和R 2 = 0.887),并且是高可靠性预测矿井爆破中 GVI 的见解。

强调

  • 研究了一种自动编码器神经网络来预测矿山爆破中的 GVI;

  • 自编码器神经网络与支持向量回归相结合,生成稳健的混合模型(AutoencoderNN-SVR)来预测矿山爆破中的 GVI;

  • 将提出的 AutoencoderNN-SVR 模型与经验模型、SVR 和 GEP 模型进行比较;

  • 所提出的 AutoencoderNN-SVR 模型被引入作为一种新颖且稳健的技术,用于以高精度预测 GVI。

更新日期:2021-06-22
down
wechat
bug