基于神经网络模型的预拌流态土剪切特性研究

为充分利用工程弃土，将其与固化剂和水充分拌和，制备一种回填后无需压实且工程性质良好的新型填筑材料。依托南京某基坑肥槽回填工程，对不同配比的预拌流态土进行直剪试验，研究预拌流态土的剪切特性，基于神经网络模型开展了预拌流态土剪应力-剪切位移预测研究。结果表明：预拌流态土剪应力-剪切位移曲线类型受水泥配合比、养护龄期和垂直压力影响；预拌流态土的抗剪强度和粘聚力受水灰比、垂直压力、养护龄期、水泥配合比影响；预拌流态土剪应力-剪切位移神经网络预测模型对抗剪强度的预测具有较高的精度。

Study on Shear Characteristics of Premixed Fluid Soil Based on Neural Network

To fully utilize the engineering waste soil, it is thoroughly mixed with a curing agent and water to prepare a new type of filling material that does not require compaction after backfilling and has good engineering properties. In this paper, depending on a foundation pit fertilizer trench backfill project in Nanjing, direct shear tests were carried out on different proportions of premixed fluid soil to study the shear characteristics of premixed fluid soil. Based on the neural network model, the prediction of the shear stress versus shear displacement curve of premixed fluid soil was carried out. The results indicate that the type of shear stress versus shear displacement curve of premixed fluid soil is influenced by the cement mix ratio, curing age, and vertical pressure. The shear strength and cohesion of premixed fluid soil are influenced by the water cement ratio, vertical pressure, curing age, and cement mix ratio. The neural network prediction model for shear stress versus shear displacement curve of premixed fluid soil has a high accuracy in predicting the shear strength.