基于神经网络的土层阻尼系数转换频率预测

Prediction on Translation Frequency of Damping Coefficient of Soil Based on Neural Networks

利用神经网络方法研究了地震波激励作用下,土层阻尼系数转换频率与各主要影响因素之间复杂的非线性关系。通过比较时域和频域的计算结果,探讨了在土层时域分析中,如何由滞后阻尼系数形成阻尼矩阵的问题。利用实际地震波的分析结果,建立了阻尼系数转换频率的BP神经网络预测模型,从而为利用滞后阻尼系数在时域中进行土层反应分析提供了依据。

Neural networks technology is applied to investigate the complex non-linear relationship between the translation frequency of damping coefficient of soil and the main related factors under earthquake waves. The method of valuating the damping matrix in the time domain analysis is discussed by comparing the computations of time domain and frequency domain. From the analysis of several earthquake waves, BP （back propagation） neural networks model is built to predict the translation frequency of damping coefficient and thus provide basis for the use of the hysteretic damping coefficient to carry out reaction analysis of soil layers in time domain.