为探究原状土阻尼比D随剪应变γ和土层深度H(上覆压力σ′m)的双向维度特征,本文对同一钻孔自地表至基岩深度范围内原状粉质黏土与粉土开展系列共振柱试验.利用BP神经网络技术深度挖掘、识别、学习原状土D的双向维度耦合规律,建立了D智能预测法.通过预测结果与试验数据的比较,得出基于BP神经网络的智能预测法可较好地适用于原状土D的双向维度特征预测.试验表明:原状土D与γ和H(σ′m)2 个维度存在耦合相关.同一H(σ′m)维度时,D随γ增加呈现非线性上升规律;同一γ维度时,D随H(σ′m)增加呈现相反的降低规律;随着H(σ′m)维度的增加,D～γ 整体非线性关系逐渐下倾伴随着增长速率逐级变缓.本文方法实现了原状土 D 在H(σ′m)和γ双向维度下的智能预测.
Abstract
To investigate the bidirectional characteristics of undisturbed soil damping ratio D with shear strain γ and soil depth H(overburden pressure σ′m),a series of resonant column tests were conducted on undisturbed silty clay and silt from the surface to the bedrock of the same borehole.Back-propagation(BP)neural network technology was utilized to analyze,identify,and learn the coupling law of bidirectional dimensions of undisturbed soil D,and an intelligent D prediction method was established.A comparison of the prediction results with the experimental da-ta revealed that the intelligent prediction method based on BP neural networks could effectively predict the bidirec-tional characteristics of undisturbed soil D.A coupling correlation existed between the undisturbed soil D with two dimensions of γ and H(σ′m).In the same H(σ′m)dimension,D exhibited a nonlinearly increasing regularity with increasing γ,while in the same γ dimension,D showed an opposite decreasing pattern with increasing H(σ′m).Moreover,as the H(σ′m)dimension increased,the overall nonlinear relationship between D and γ showed a down-ward trend,accompanied by a gradual slowdown of the growth rate.This method enabled intelligent prediction of undisturbed soil D in the bidirectional dimensions of γ and H(σ′m).
维普
万方数据