客运专线铁路路基粗粒土填料最大干密度的BP神经网络预测

摘要

以沪昆客运专线芷江北站粗粒土填料为研究对象，通过表面振动压实试验，得到不同颗粒级配下试样的最大干密度。考虑各粒径含量与最大干密度的非线性关系，将粒度成分，级配指标，分形指标作为网络输入层，基于误差反向传播算法，以干密度试验结果为训练样本，建立BP神经网络预测模型，实现对不同颗粒级配下粗粒土最大干密度的预测。%Taking the fillers of the coarse-grained soil in Zhijiang north station of Shanghai-Kunming passen-ger dedicated line as the research object,the vibration compaction test was conducted to study maximum dry den-sities under different granular compositions.Considering the non-linear relations between granular compositions and maximum dry densities,a BP neural network prediction model of which the input layer was consisted of gran-ular compositions,grading index and fractal index was established.Based on the backwards error propagation al-gorithm and the result of the maximum dry density test,the model established in this paper performs well in pre-dicting the maximum dry density of the coarse-grained soil of various granular compositions.