摘 要：为了将BP神经网络应用到黄土湿陷性评价中，在运用因子分析原理消

Abstract: In order to BP network to loess collapsibility evaluation, the factor analysis principle eliminate among physical and mechanical indexes based correlation proposed and established based on the water content, void ratio, plasticity index and compression coefficient Loess four physical and mechanical properties of a sag factor BP neural network model, given the results of soil tests Lintao west highway project for the training and testing samples, comparative analysis of the predicted value and the measured value. The results show that the model is trained, the training results of the coefficient of determination was 0.95, a high degree of interpretation; prediction analysis, the relative error of the predicted and measured values ​​is typically less than 11.5%. DESCRIPTION BP neural network model presented herein may be used to predict loess collapsibility, have utility in engineering.

Abstract: In order to apply BP neural network to the evaluation of loess wet subsidence, on the basis of using factor analysis principle to eliminate the correlation between various physical mechanical indicators, a prediction model of loess wet subsidence neural network based on four physical mechanical indicators, such as moisture content, pore ratio, plasticity index and compression coefficient, is proposed and established. Based on the results of the geotechnical experiment of the Dingxi-Linyi Expressway project, the predicted and measured values were compared and analyzed. The results show that the decision coefficient of the training results is 0.95, which is high lyse, and the relative error of the prediction value and the measured value is generally less than 11.5% when the prediction is analyzed. It is explained that the BP neural network model proposed in this paper can be used for the prediction of wet subsidence of loess, and is practical in engineering.

Abstract: in order to apply BP neural network to the evaluation of loess collapsibility, on the basis of eliminating the correlation between the physical and mechanical indexes by using the principle of factor analysis, a BP neural network prediction model of loess collapsibility coefficient based on four physical and mechanical indexes, such as water content, void ratio, plasticity index and compression coefficient, is proposed and established, which is used for the engineering geotechnical test of Dingxi Lintao expressway The results are training and testing samples. The predicted and measured values are compared and analyzed. The results show that the decision coefficient of the training results is 0.95 and the interpretation degree is high, and the relative error between the predicted value and the measured value is generally less than 11.5%. It shows that the BP neural network model proposed in this paper can be used to predict the collapsibility of loess, and it is practical in engineering.<br>