Predictive Lime Soil Maximum Strength by Measuring Soil Ingredients Based on Neural Network

摘要

Because there are the complex non-linear relations between the soil ingredient and strength, no mathematics model has been established at present. The experiments of 12 reigns of soil ingredients and the lime soil compressive strength under 180-day saturated water in the optimum lime dosage and the stipulated moisture were carried on. Based on the artificial neural networks technology, the soil ingredient - lime stabilization strength model has been established. The model was used to carry on the strength estimate to the examination samples. The results indicates the average relative error that the model forecasts is 3.97% and the maximum relative error is smaller than 10%, which may satisfy the actual project requirement.