摘要
沉降系数的确定是软土路基沉降计算中的一项重要内容,在对沉降影响因素定性分析的基础上,用前馈型人工神经网络(ANN)模型来计算沉降系数。首先,根据沉降影响因素建立三层型前馈型神经网络模型。然后,利用其非线性映射能力,通过样本学习,建立软土层厚度、硬壳层厚度、填土高度、施工工期等因素与沉降系数之间的定量关系,计算沉降系数,有效地减少了确定沉降系数时的主观性和盲目性。最后,用该方法对某高速公路的沉降系数进行了验算,得出了与实测资料比较一致的结果,表明利用ANN模型的非线性映射能力建立沉降系数ms与影响因素之间的对应关系,确定沉降系数ms是有效而且可行的。用ANN模型确定的沉降系数ms修正分层总和法计算结果,与传统经验方法确定的沉降系数修正沉降量相比,能够更全面地反映各种因素的影响,提高沉降量计算的精度。
Determining the settlement coefficient is an important work in settlement calculation of soft ground embankment.The Back Propagation (BP) Artificial Neural Networks (ANN) is used to calculate the adjustment factor for the settlement based on the settlement influence factors analysis.Firstly, a three layer back propagation artificial neural networks model is established based on settlement influence factors. Secondly, the BP- ANN' s function of nonlinear mapping is used to establish the quantity relationship between the influence factors and settlement values after the samples trained the ANN, which can efficiently reduce subjectivity and blindness in determining the settlement coefficient.Lastly, the ANN method is used to analyze the settlement data of an expressway, and the calculating values are close to the measured values. The results indicate that the method is rational and feasible. Being used to ameliorate the result of multi-element summation method, the settlement coefficient ms by the method is more roundly in reflection the influence of different factors and has higher precision than the traditional empirical method.
出处
《公路交通科技》
CAS
CSCD
北大核心
2007年第4期29-33,共5页
Journal of Highway and Transportation Research and Development
关键词
道路工程
软土地基
人工神经网络
路基沉降
沉降系数
误差逆传播模型
road engineering
soft ground
artificial neural networks
embankment settlement
settlement coefficient
error backpropagation model