人工神经网络预测非饱和土的特征参数

针对测定非饱和土的吸力既费时，费钱，又困难，而且往往是通过水分特征曲线间接求取，以致其精度不够理想这一问题，尝试采用人工神经网络进行预测研究，将部分已有的试验数据（含水量，吸力数据）作为输入模式进行训练，然后利用网络参数进行相应的参数预测。结果表明，神经网络预测技术十分有效，为工程应用提供了一种新的工具。

Predicting Characteristic Parameter of Unsaturated Soil with Artificial Neural Network

Because direct measurement of suction is costly time, consuming and difficult, generally, suction is indirectly found with soil water characteristic curve, thus, the precision is not good enough. In order to solve the problem, the artificial neural network method and partial experimental data(water content and suction) are used as input patterns to train and predict corresponding parameters with network parameter. The results show that the method is very effective and is a powerful tool for engineering application.