利用人工神经网络模型判定膨胀土等级

Determination of Expanded Clay with Artificial Neutral Netwo rk Model

膨胀土是一种吸水后体积产生膨胀 ,失水后体积产生干缩 ,对铁路工程极其有害的土质 ,而其胀缩等级的划分是一个典型的不确定性问题。应用神经网络的联想记忆功能 ,以膨胀土胀缩等级划分标准中三个类别作为训练标本 ,建立膨胀土胀缩等级与分级指标之间的对应关系的BP判定模型 ,以对膨胀土胀缩等级进行划分 ,并结合具体工程进行了应用。应用表明 ,神经网络方法与模糊综合评判法膨胀土胀缩等级判定结果一致 ,神经网络具有在膨胀土胀缩等级判定问题上的可靠和从训练模型中获得划分知识的能力。

The expanded clay is one that expands when a bsorbing water and shrinks when dewatered and thus it is a harmful soil for rail way engineering. The determination of the expanded clay grade is a typical uncer tain issue. With the linked memory function of the artificial neutral network mo del, taking three grades of the expansion and shrinkage of the expended clay as the training samples, the back-propagation (BP) neutral network model is establ ished to identify the corresponding relation between the grades of the expansion and shrinkage of the expanded clay and the graded index. Thus the expansion and shrinkage of the expended clay can be graded. This method is used in the actual engineering project. The results indicate that the grade of expansion and shrin kage of the artificial neutral network model and that of the fuzzy comprehensive classification method are in agreement. It is as reliable as the artificial neu tral network model in grade determination and also has the capability of the tra ining model that is strong in grading knowledge. This provides a new way of clas sifying the grades of the expansion and shrinkage of the expanded clay.