基于优化IGA-ELM模型的爆破振动特征参量预测研究

Predicting research on characteristic parameters of blast-induced vibration based on optimized IGA-ELM model

为有效预测露天矿爆破振动特征参量,建立基于组合赋权的免疫遗传算法(IGA)优化极限学习机(ELM)预测模型。建立该模型之前,根据爆破振动影响因素确定输入层参数,根据爆破安全规程判据确定输出层参数;然后,应用调和平均数概念整合模糊层次分析法(FAHP)所得主观权重和熵权法所得客观权重,量化输入层参数权重;其次,针对现有ELM输入层权值、隐含层偏差的选择问题,引入IGA对其进行优化选择,并通过逐步增减法探究ELM隐含层最优节点数。该模型曾被应用于某露天矿。研究结果表明:用所构建优化IGA-ELM模型能够更准确地预测露天矿爆破振动特征参量,且所得均方误差、决定系数、仿真误差明显优于其他模型。

To predict the characteristic parameters for blasting vibration of open-mine effectively,an optimal IGA-ELM model was built on the basis of combination weighting.Before building the model,the parameters of input layer was determined in line with the influence factors of blasting vibration.And that of output layer were confirmed according to the safety regulation criterion for blasting.Then,the subjective and objective weights obtained by fuzzy analytic hierarchy process（FAHP） and entropy weight method respectively were integrated by applying the harmonic mean concept.And the weights of input layer parameters were quantified.In addition,IGA was introduced to select the input layer weights and hidden layer deviations of ELM by optimization.The optimal node number of ELM hidden layer was explored by using the stepwise increase-decrease method.The model was applied to a certain open-mine in China.The research results show that the optimal IGA-ELM model can be used to predict the characteristic parameters for blasting vibration of strip mine more accurately,and that the mean square error,determination coefficient,and simulation error are superior to those obtained by other models.