Research On The Final Settlement Of Soft-Clay Ground Based On Artificial Neural Network

With the rapid development of economy, the freeways have been built continuously in our country. Building freeways would meet the problem of soft-clay inevitably; so controlling the settlement is the key problem to build freeways on soft-clay ground. For the sake of dynamic control for the settlement distortion of the embankment in the period of construction, the system of dynamic observation and forecast of the soft-clay ground must be set up. On the one hand, it ensures safety and stability of embankment. On the other hand, it forecasts the post-construction settlement correctly and makes it under the permit of the range of the design.The soft-clay ground has large settlement and it will last for a long time under load. There will be potential post-settlement that maybe made largish harm to traffic and transportation of high-grade highways on soft-clay ground, so it is necessary to calculate and forecast the settlement of soft-clay ground before treatment. There are many factors that influence the settlement of soft-clay ground, and every factor varies in diffident time, furthermore the characters of soft-clay are complex, so it is hard to define parameters of soft-clay. On account of the above, calculating and forecasting the post-settlement of soft-clay ground exactly become a hotspot and a difficultly. Therefore, to study and found the new methods of calculation and forecast post-settlement of soft-clay ground have important theoretical and practical significance for our country's developing highway construction.This paper analyses and generalizes the theory of distortion and calculation of soft-clay ground, expounds the principle and implementation procedure of ANN. Adopts BP-ANN to make the model directly base on measured data of Ji-Feng freeway, and predicts final settlement of soft-clay ground. A perfect effectiveness is obtained through comparing and analyzing the result with some methods of curve fitting method such as hyperbolic method, exponential method, and three-point method. It proved that the neural network method can avoid the mistake due to human factor in traditional methods and can simulate engineering precisely, widely and easily. Therefore, the method has a bright future in practical engineering.