基于BP神经网络的人工盐渍土冻胀预测研究

以季节性冻土区环青海湖段青藏铁路路基冻害为背景，通过现场采集的粉质黏土人工盐渍化后，考虑温度、水分、盐分、压实度4个因素进行室内冻胀试验，测试了不同温度环境下路基土体的冻胀率。根据试验所得参数，建立BP神经网络冻胀预报模型对土体冻胀率进行预测。结果表明:运用BP神经网络的预测结果与试验结果具有良好的一致性，误差为1%～5%。

Study on Frost Heave Forecast of Artificial Saline Soil Based on BP Neural Network

Based on the frost damage of Qinghai-Tibet railway subgrade ringing Qinghai Lake in the seasonal frozen soil region, the Indoor Frost Heaving Test was conducted after the powder clay collected on site and the artificial salinization, considering such four factors as temperature, moisture, salinity and compaction degree to test the frost heaving rate of subgrade soil under different temperature conditions. According to the parameters of the test, the frost heaving rate is forecasted by establishment of frost heave prediction model with BP Neural network. The results show that the predicted results obtained by using the BP neural network method are consistent with the experimental results, and the error is between 1% and 5%.