BP神经网络与离散GM(1,1)模型在拓宽路堤沉降预测中的应用

Application of BP Neural Network Model and Discrete GM(1,1)Model in Widening Embankment Settlement Prediction

目前,对于拓宽路堤沉降趋势和稳定性的研究很少。为了探究拓宽路堤的沉降变化趋势和沉降规律预测,以益阳市梅林路K0+460~K0+600为基础,运用BP神经网络与离散GM(1,1)模型对拓宽路堤进行沉降预测,并结合已有监测数据,对比分析2种方法的优劣。结果表明:离散GM(1,1)模型能对已有数据进行拟合,输出预测值与实际监测结果相近;BP神经网络模型预测值相对离散GM(1,1)模型更加接近实际监测结果,但需要对阈值与权重进行多次调试。2种方法沉降预测值均与实际监测值相近,均能有效预测拓宽路堤沉降。

At present,there is little research on the settlement trend of widening embankment,and less attention is paid to the stability of widening embankment.In order to explore the trend of settlement change of widening embankment and the law of widening embankment settlement prediction,based on Yiyang Meilin Road K0+460 to K0+600,the settlement prediction of widening embankment is carried out by using BP neural network and discrete GM(1,1)model.And the advantages and disadvantages of the two methods are compared and analyzed in combination with the existing monitoring data.The results show that the discrete GM(1,1)model can fit the existing data,and the output prediction value is similar to the actual monitoring result.The BP neural network model prediction value is closer to the actual monitoring result than the discrete GM(1,1)model,but the threshold and weight need to be debugged several times.The settlement prediction values of the two methods are similar to the actual monitoring values,which can effectively predict the settlement of the widening embankment.