Prediction Of Shear Strength Parameters Of Fine-grained Soil In Northern Liaodong Bay By BP Neural Networks

On the basis of collecting abundant investigation material, the paper reveals the characteristics of landforms, sub-bottom layers and sediment in the study area : a reach of estuary and delta , single Landform type; the modern marine and terrestrial sediments are saturated and arrange phase to phase in the vertical direction; the sediments are high water content, low bearing Capacity ,short formation time and weak consolidation.Principally, the paper research the effect of physical property parameters to shear strength ones and make the BP neural networks model to predict the shear strength parameters of fine-grained soil in northern Liaodong Bay. The training and test samples of the BP model are both chose from geotechnical testing data in the study area.The input vectors comprise water content, porosity ratio, density, plasticity index, liquidity index and clay particle content, at the same time, cohesive force and internal friction angle make up of the output vectors. According to the results, the prediction performance is perfect. With the relative error analysis, the prediction performance of cohesive soil is better than the one of silt in cohesive force and the prediction performance of silt is better than the one of cohesive soil in internal friction angle.