Abstract: The existing approaches to identify and classify expansive soil have been evaluated. The indices that can reflect and characterize the swell-shrink mechanism and properties are analyzed. Making use of six indices such as liquid limit, plasticity index, less than two microns glutinous granule percent, CBR, free expansion ratio and CBR expansion ratio, the swelling and shrinking grade model on the theoretic base of SOM neural network model is established through determining the limit value of every, index. And the judgment software is also progrannned. Furtbermore, the judgment model is applied to judge and classify the swelling and shrinking grade of expansive soil samples. The model is tested by other two indices besides the six indices of expansive soil mentioned above, that is, swelling and shrinking performance and strength performance. The classification is proved correct and credible.

Key words: expansive soil , grade of expansion and shrink , SOM neural network , experimental validation ;

摘要： 在对现有膨胀土判别分类方法进行评价的基础上,根据公路建设中的常规试验项目选择了液限、塑性指数、 小于0??002mm的黏粒含量、CBR、自由膨胀率和CBR膨胀率6项指标,通过确定每项指标的界限值,建立了以神经 网络中的SOM网络模型为理论依据的膨胀土胀缩等级评判模型,编写了评判软件;并应用评判模型对沿线膨胀土 土样的胀缩等级进行了评判分类;通过应用膨胀土土样6项指标以外的胀缩性能和强度性能指标进行验证,证明 了分类结果是正确可靠的。

关键词: 膨胀土, 胀缩等级, SOM神经网络, 试验验证