GA改进BP神经网络在抗滑桩孔爆破开挖中的应用

目前桩孔开挖主要依靠工程类比进行,不同设计者设计的爆破参数往往因掌握的爆破理论和经验的不同而有所差异,爆破质量参差不齐。为此,提出基于遗传算法GA改进BP神经网络(GA-BP)建立爆破参数优化设计模型,该法不仅可以利用已有爆破经验数据和工程地质条件,同时,使用遗传算法优化BP神经网络阈值和权值可以弥补BP神经网络不稳定的缺陷,以达到获得更优爆破参数的目的。实践表明,基于遗传算法改进BP神经网络相比一般BP神经网络预测相对误差较小,同时GA-BP神经网络得到的优化爆破参数进行现场试验,取得了良好的爆破效果。因此,GA-BP神经网络模型应用于抗滑桩孔开挖爆破参数设计是可行的,可用于指导爆破施工。

Application of BP neural network improved by GA in blasting excavation design of slide-resistant pile hole

Currently,the pile holes are excavated mainly relying on the engineering analogy,but the blasting parameters designed by different theories and experience and the effect of blasting are different.Therefore,BP neural network improved by genetic algorithm was provided and the model of optimization of blasting parameters was set up.The sample data obtained from good blasting effect and geology of practical engineering could be used,and GA could make up the defect of BP neural network which was not stable through optimizing the threshold and weight of BP neural network,and the purpose of obtaining the better blasting parameters was achieved.The practice showed that relative errors of prediction of GA-BP neural network were fewer than the general BP neural network,and it achieved a good result after field test using the optimized blasting parameters.GA-BP neural network model was well applied in the blasting parameters design of excavation of slide-resistant pile hole,it was feasible and could be used to guide the construction of blasting.