基于BP神经网络的瑞雷面波智能优化反演

The Rayleigh Surface Wave Intelligent Inversion Based on the BP Artificial Neural Network

由于人工神经网络具有高度并行性、自组织、自学习和联想记忆能力,它的反演预测能力非常强,能够较准确地预测出目标参数,因此引入BP(Back Propagation)人工神经网络的地震面波智能优化反演。通过快速矢量传递算法计算频散曲线,为人工神经网络训练提供模型,并对层状介质模型试算,取得了好的效果。对实测瑞雷面波记录进行反演计算,并与其他反演方法进行对比,证明了BP人工神经网络反演瑞雷面波时的收敛速度快、对初始模型要求低等优点。

To deal with the existing shortcomings of surface wave inversion, this paper in- troduces the BP Artificial Neural Network （ANN） for seismic surface wave intelligent in version. BP artificial neural network has a high degree of parallelism, self organizing, self learning and associative memory abilities, so it has strong predictive ability of the inver- sion, which can accurately predict the target parameters. The forward dispersion curve pro- gram is written through the fast vector matrix algorithm, which provides the training model for the BP artificial neural network, the layered medium model is preliminary written, and good effects are achieved. Furthermore, the measured Rayleigh wave is inversed by BP ar- tificial neural network. Compared with other inversion methods, BP artificial neural net- work inversion has such advantages as fast receiving velocity and the low requirements on the initial model.