基于ACOR-LSSVM算法的爆破振动速度预测  

Blasting Vibration Velocity Prediction based on ACOR-LSSVM Algorithm

作　　者：郑皓文 赵根[1] 胡英国[1] 柴朝政 ZHENG Hao-wen;ZHAO Gen;HU Ying-guo;CHAI Chao-zhen(Key Laboratory of Geotechnical Mechanics and Engineering of Ministry of Water Resources,Yangtze River Science Research Institute,Wuhan 430010,China)

机构地区：[1]长江水利委员会长江科学院水利部岩土力学与工程重点实验室,武汉430010

出　　处：《爆破》2018年第3期154-158,共5页Blasting

基　　金：国家自然科学青年基金项目(51609017)

摘　　要：由于工程地质条件复杂,传统的方法很难准确预测爆破振动速度。引入连续域蚁群算法(ACOR),对最小二乘支持向量机(LS-SVM)进行参数优化,构建连续域蚁群最小二乘法爆破振动速度预测模型(ACOR-LSSVM)来提高预测精度。结合白鹤滩水电站左岸坝肩槽开挖过程中的40组爆破监测数据,分别采用ACOR-LSSVM、LS-SVM模型与萨氏公式进行爆破振动速度预测,三者的平均绝对相对误差分别为3.16%、10.07%、22.96%。相比之下,ACOR-LSSVM模型预测精度更高,泛化能力更强,在爆破振动速度预测中具有一定的理论意义和工程应用价值。Due to the complex engineering geological conditions,it is difficult to accurately predict the blasting vibration velocity with the traditional methods. In order to improve the prediction accuracy,the ant colony optimization for continuous domains algorithm was used to optimize the traditional least squares support vector machine model,and blasting vibration velocity prediction model was established with ant colony continuous domains least squares support vector machine. Based on 40 groups of blasting monitoring data during the excavation for left bank abutment slot of Baihetan Hydropower Station,the blasting vibration velocity prediction was conducted respectively by using ACOR-LSSVM,LS-SVM and Sodev＇s empirical formula. The average absolute relative error values of the three was3. 16%,10. 07% and 22. 96% respectively. In comparison,the ACOR-LSSVM model has higher prediction accuracy and stronger generalization ability,which has certain theoretical significance and engineering application value in blasting vibration velocity prediction.

关 键 词：连续域蚁群算法 最小二乘支持向量机 坝肩槽开挖 爆破振动速度 

分 类 号：TD235.3[矿业工程—矿井建设]