基于BP神经网络的膨胀土判别分级方法研究

以安康膨胀土为研究对象,选用粘粒含量、粉粒含量、液限和塑性指数4个分级指标,建立了两层无偏置的BP神经网络模型,研究膨胀土的判别分级问题。结果表明,该模型学习效果良好,能准确预测未知样本的膨胀性;对于安康膨胀土,粘粒含量和粉粒含量对分级结果影响较大,而液限和塑性指数影响较小;相比于传统的指标分级法,该模型具有较好的容错能力,可有效减小指标测量误差对分级结果的影响;BP神经网络用于膨胀土的判别分级是合理可行的,具有一定的推广与应用价值。

Discrimination and Classification Method for Expansive Soil Based on BP Neural Network

Taking Ankang expansive soil as the research object,four indicators of clay content,powder content,liquid limit and plasticity index are selected as grading indicators,and a two-layer unbiased BP neural network model is established to study the problem of discriminantion and classification of expansive soils.The research shows that:(a) the model has a good learning effect and can accurately predict the expansion of unknown samples;(b) for Ankang expansive soil,the clay content and powder content have a greater impact on the classification results,while the liquid limit and plasticity index have less impact;(c) compared with the traditional index classification method,this model has better fault tolerance and can effectively reduce the impact of index measurement errors on the classification results;and(d) it is reasonable and feasible to use BP neural network to distinguish and classify expansive soil,and it has certain promotion and application value.