Research on Neural Network Modeling and Computer Simulating Forecasting of Rheological Parameters of Soft-Soil

The rheological parameters of soft-soil are affected by the percentage of clay and water, and the influencing characteristics are nonlinear complex relationship. It is difficult to express the influencing disciplinarian using traditional mathematic functions. The rheological parameters of soft-soil are tested under the different conditions of the influencing factors, and the main affect factors are analyzed. The BP neural network model with three tiers structure is founded by using the theory of neural network and experiment results, the model is made up of two input cells and two output cells, and the structural parameters of the model such as weights value and thresholds value are ascertained by training and self-adjusting after samples have been studied. The rheological parameters of new samples are simulated and forecasted by using the new model, and the result indicates that the veracity is over 90 percent through comparing with actual testing results. So the model can express exactly the influencing disciplinarian of factors. The research results have theory signification and practicality value for us to analyze rheological characteristics of soft-soil, forecast rheological parameters rapidly.