Blast induced damage to the final wall of rockmass in any civil or engineering application is a major concern to the rock excavation engineers. There are at least four distinct techniques practised by blasting engineers to minimise the damage to the parent rock during blasting that are classified as contour blasting. One of the methods to achieve a smooth wall or rock surface in blasting is pre-splitting, where blastholes in the final row are mildly loaded with explosives and fired first in sequence before the rest of the holes in a round of blast to achieve a split in rockmass. Since blasting is one of the most adopted methods for achieving a final wall in a desired shape, while maintaining the rockmass integrity, the design of pre-split blast assumes importance. There are several variables that determine the final shape of the pre-split face including the rock type, the drill diameter, the explosive characteristics, the orientation of major joints with respect to drill hole, the linear charge density, the spacing of drill holes and inclination of the drill holes. This study attempts to determine the importance and sensitivity of variables in pre-split blasting using historical data from open excavations of a hydroelectric project in India. Keeping in view the multitude of variables, low correlation of independent variables with responses and high level of interactions between the variables that define the blast results, it was found suitable to use ANN to determine the importance and sensitivity of the variables involved. A new method to determine the blast induced damage to rockmass, namely, undamaged area (A_UD%), is introduced. Half Cast Factor (HCF%) and A_UD% were calculated for a high wall and trenches by digital image analysis technique using the Fragalyst software. Joint spacing and linear charge density emerged as most controlling variables in determining the final result of pre-blast. The relative importance of other variables assuming significance in the case of both the designed outputs is also discussed. A comparative analysis of HCF% and A_UD% on the basis of results obtained from ANN analysis is also presented. A futuristic approach concept has also been introduced.

Graphical abstract

中文翻译：

---

使用人工神经网络定义预裂爆破性能的变量的重要性和敏感性

在任何土木或工程应用中，爆破对岩体最终壁的损坏是岩石开挖工程师的主要关注点。爆破工程师至少采用了四种不同的技术来最大限度地减少爆破过程中对母岩的破坏，这些技术被归类为轮廓爆破。在爆破中获得光滑的墙壁或岩石表面的方法之一是预裂，即在最后一排的炮眼中装满炸药，并在一轮爆破中的其余孔洞之前依次发射以实现分裂在岩体中。由于爆破是获得所需形状的最终墙体最常用的方法之一，同时保持岩体完整性，因此预裂爆破的设计非常重要。有几个变量决定了预裂面的最终形状，包括岩石类型、钻头直径、爆炸特性、主要节理相对于钻孔的方向、线装药密度、钻孔间距和钻孔的倾斜度。本研究试图使用来自印度一个水电项目的露天开挖的历史数据来确定预裂爆破中变量的重要性和敏感性。考虑到众多变量、自变量与响应的低相关性以及定义爆炸结果的变量之间的高度交互，发现使用 ANN 来确定所涉及变量的重要性和敏感性是合适的。一种确定岩体爆破损伤的新方法，即引入了_UD % )。使用 Fragalyst 软件通过数字图像分析技术计算高墙和沟槽的半铸系数 ( HCF% ) 和A _UD %。接头间距和线性电荷密度成为决定预爆最终结果的最控制变量。还讨论了在两个设计输出的情况下其他变量的相对重要性。还介绍了基于 ANN 分析结果的HCF%和A _UD %的比较分析。还引入了未来主义的方法概念。

图形概要