基于BSA-BP神经网络方法的引水隧洞围岩参数反演模型及应用

Inversion Model and Application of Surrounding Rock Parameters in Diversion Tunnel Based on BSA-BP Neural Network

针对地下工程围岩参数取值,提出将回溯搜索优化算法(BSA)与BP神经网络相结合的混合网络(BSA-BP)方法,对隧道围岩参数进行反演研究。通过建立隧道有限元开挖模型,利用反演参数计算监测断面的位移并与现场实测值进行对比,最终对围岩稳定性进行分析预测。结果表明,经BSA算法优化的BP神经网络相对于GA-BP神经网络,具有更快的反演速度与计算效率。利用BSA-BP神经网络反演参数得到的位移计算值与现场实测值相对误差均在5%以内,表明该模型具有较高的反演精度,合理可行,为地下工程参数反演提供了一种新方法。

For the value of the surrounding rock parameters of the underground construction,a hybrid network approach combining backtracking search optimization algorithm(BSA)and BP neural network(BSA-BP)was proposed for the inversion study of the tunnel surrounding rock parameters.By establishing a tunnel finite element excavation model,the inversion parameters were used to calculate the displacement of the monitoring section and compare with the measured values in the field.Finally,the stability of the surrounding rock was analyzed and predicted.Compared with the GA-BP neural network,the results show that the BP neural network optimized by BSA algorithm has faster inversion speed and computational efficiency.The relative errors between the calculated displacement values and the field measured values obtained by using BSA-BP neural network inversion parameters are within 5%,indicating that the model has high inversion accuracy and is reasonable and feasible.The research results provide a new method for the inversion of underground engineering parameters.