BP神经网络在道路土壤分离速率模拟中的应用

BP neural network modeling on soil detachment rate of road

土壤分离是土壤侵蚀的重要过程,为坡面径流的搬运过程提供了物质基础,因而对土壤分离速率的准确模拟具有重要的理论和实践意义。采用变坡水槽实验,利用在较大的坡度(8.8%～46.6%)及较大的流量范围(1～5L/s)内测得的黄土高原道路土壤分离速率数据,分别使用BP神经网络模型及回归模型对土壤分离速率进行模拟,并对比上述2种模型的模拟效果。结果表明:BP神经网络模型可以利用实验中较容易测定的坡度、流量等数据对土壤分离速率进行较为准确的模拟(模型效率系数0.952);相对传统回归模型,BP神经网络模型对不同类型道路的土壤分离速率的模拟精度均有所提高;BP神经网络模型可以将道路类型、坡度、流量与土壤分离速率的关系统一为一个模型,可为道路土壤分离的模拟提供新的方法。

Soil detachment is one of the key processes of soil erosion,and it provides material for transportation.It is important to predict soil detachment rate precisely both for better understanding of soil erosion process and soil erosion modeling.This paper used both BP neural network model and regression models to simulate soil detachment rate of road with the soil detachment rate data obtained from flume experiment in a large scale of slope gradient（8.8%-46.6%） and flow rate（1-5 L/s）,and compared the results of two means.The results showed that： BP neural network model can predict soil detachment rate very well with the data which are easily obtained,including slope gradient,flow rate and road type;BP neutral network model improved the accuracy of regression model in predicting soil detachment rate in every type of road.Since BP neutral network model can combine the different road types,different flow rates and different slope gradients into one model,it can improve the efficiency of predicting soil detachment of road and provide a new approach to simulate soil detachment rate of road.