基于SSA-BP神经网络爆破参数优选试验研究

Experimental Study on Blasting Parameters Optimization Based on SSA-BP Neural Network

为实现矿山快速准确地选取爆破参数,提出了一种基于麻雀搜索算法(SSA)来优化BP神经网络的数学模型,以抗拉强度、弹性模量、内摩擦角等6项影响矿岩可爆性因素为输入因子,以炮孔间距和炸药单耗为输出因子,基于训练样本建立参数优选模型。以辽阳宏盛镁矿为例,通过优选得到了该矿的爆破参数孔底距为1.5 m,排距为1.2377 m,炸药单耗为0.1603 kg/t。实践证明,此模型有效改善了传统BP神经网络收敛速度慢、精度相对较低等缺陷,相比经验公式得到的炸药单耗降低了27.2%,大块率控制在5%以内,优选的爆破参数能够取得良好的爆破效果。

In order to select blasting parameters quickly and reasonably in mines, a mathematical model of BP neural network optimized by sparrow search algorithm(SSA) was proposed. Six influencing factors of rock explosibility such as tensile strength, elastic modulus and friction angle were taken as input factors, and blast hole spacing and explosive consumption were taken as output factors. Then, a parameter optimization model was established based on training samples. Taking the actual parameters of a mine as an example, the blasting parameters after the optimization were obtained as follows: the hole-bottom spacing of 1.5 m, row spacing of 1.2377 m, and the explosive consumption of 0.1603 kg/t. The practice shows that this model effectively improves the shortcomings of traditional BP neural network, such as slow convergence speed and low precision. Compared with the empirical formula, the explosive consumption is reduced by 27.2%, and the bulk rate is controlled within 5%. The optimized blasting parameters can achieve a good blasting effect.