摘要
抛掷爆破效果直接影响露天矿剥离成本,对抛掷爆破的拉斗铲倒堆工艺生产效率有重要影响,为提高露天煤矿抛掷爆破效果预测准确率,进而提高反馈爆破参数优化设计精度,在分析露天矿抛掷爆破效果影响因素的基础上,提出一种遗传算法(GA)优化最小二乘支持向量机(LSSVM)的抛掷爆破效果预测模型。引入傅里叶级数模型模拟爆堆剖面曲线,利用训练完成的GA-LSSVM预测Fourier级数模型控制参数A_(0),θ,a_(n)及b_(n),进而输出预测的爆堆形态。依据内蒙古黑岱沟露天煤矿抛掷爆破实测数据进行实例分析,选取台阶高度、剖面宽、炸药单耗、最小抵抗线、排距、孔距、坡面角、采空区上口宽、采空区下口宽、松方体积、有效抛掷量作为GA-LSSVM预测模型的输入参数,A_(0)、θ、a_(n)、b_(n)、最远抛掷距离、松散系数和有效抛掷率作为输出参数,建立露天矿抛掷爆破效果GA-LSSVM预测模型,并将偏最小二乘回归模型(PLSR)、LSSVM模型、粒子群算法优化最小二乘支持向量机模型(PSO-LSSVM)与其进行对比。结果表明:(1)通过2~7阶Fourier展开级数对爆堆剖面曲线模拟分析,确定阶数为4时模拟精度与效率达到最优,其误差平方和(SSE)为21.593 4,决定系数(R^(2))与调整后的决定系数(R_(adj)_^(2))为0.999 2,均方根误差(RMSE)为0.479 3;(2)相较于传统LSSVM预测模型,通过GA优化后,最远抛掷距离、松散系数以及有效抛掷率均获得更高的R^(2)(1,1,1)和更小的RMSE(0.180 9,0.000 7,0.000 2),说明改进后的GA-LSSVM具有更好的模拟效果和泛化能力;(3)与PSO-LSSVM、LSSVM、PLSR模型相比,GA-LSSVM模型对抛掷爆破效果的预测精度(R^(2),RMSE)更高且优势明显;(4)结合4阶Fourier级数的GA-LSSVM模型与采用Weibull函数的BP或GA-ELM等模型相比,对爆堆形态的预测具有更高的操作效率及预测精度。
The blast casting effect directly affects the stripping cost of open-pit mine,and has an important impact on the production efficiency of the throwing-blasting-draw bucket dumping process system.Based on the influencing factors of mine blast casting effect,a prediction model of blast casting effect using genetic algorithm(GA)optimized least squares support vector machine(LSSVM)was proposed.The Fourier series model was introduced for the first time to simulate the profile curve of the explosion,and the trained GA-LSSVM was used to predict the Fourier series model control parameters A_(0),θ,a_(n)and bn,and then output the predicted explosion shape.Based on the actual measurement data of blast casting in the Heidaigou Open-pit Coal Mine,the parameters were selected,such as the height of the bench,the width of the section,the unit consumption of explosives,the minimum resistance line,the row spacing,the hole spacing,the slope angle,the widths of the upper opening and the lower opening of the gob,loose square volume,and effective throwing volume,as the input parameters of the GA-LSSVM prediction model,and the A0,θ,an,bn,the farthest throwing distance,loosening coefficient and effective throwing rate were used as output parameters to establish the GA-LSSVM model for predicting the blast casting effect at open-pit mine.In addition,the LSSVM prediction model was compared with the Partial Least Squares Regression Model(PLSR),LSSVM Model,and Particle Swarm Optimization Least Squares Support Vector Machine Model(PSO-LSSVM).The results show that(1)the detonation profile curve is simulated and analyzed by the Fourier expansion series of orders 2 to 7,and it is determined that the simulation accuracy and efficiency are optimal when the order is 4,the sum of squares of errors(SSE)is 21.5934,and the coefficient of determination(R^(2))and the adjusted coefficient of determination(R_(adj)^(2))is 0.9992,and the root mean square error(RMSE)is 0.4793;(2)compared with the traditional LSSVM prediction model,after the GA optimization,the farthest throwing distance,the loose coefficient and the effective throwing rate are all higher R^(2)value(1,1,1)and a smaller RMSE value(0.1809,0.0007,0.0002)are obtained,indicating that the improved GA-LSSVM has better simulation effect and generalization ability;(3)compared with PSO-LSSVM,LSSVM,and PLSR,the GA-LSSVM model has a higher prediction accuracy(R^(2),RMSE)for the blasting effect of throwing and has obvious advantages;and(4)compared with models such as GA-ELM,the prediction of the explosion shape has higher operational efficiency and prediction accuracy.
作者
马力
李天翔
来兴平
许晨
孙伟博
薛飞
徐甜新
MA Li;LI Tianxiang;LAI Xingping;XU Chen;SUN Weibo;XUE Fei;XU Tianxin(College of Energy Engineering,Xi’an University of Science and Technology,Xi’an710054,China;Institute of Surface Mining Technology,Xi’an Uni-versity of Science and Technology,Xi’an710054,China;Heidaigou Open-Pit Coal Mine,Shenhua Zhungeer Energy Co.,Ltd.,Ordos010300,China)
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2022年第12期4455-4465,共11页
Journal of China Coal Society
基金
国家自然科学基金青年科学基金资助项目(51604264)
陕西省教育厅科研计划资助项目(18JK0520)
陕西省国际科技合作计划资助项目(2021KW-37)。