Research On Prediction Of Settlement Of Soft Ground In Expressway By Use Of The Artificial Neural Networks

Building expressways on soft ground is costly and needs high level of technique. On the other hand, the stability and the settlement of the soft ground need to be controlled strictly. Because the soft soil is compressible, moisture-bearing, of low strength and of low permeability etc., the settlement of the soft ground becomes one of the key problem in highway engineering. This problem is important to the success of an entire project. Designers need the ability to predict the settlement of the soft ground in time and accurately.Until now, a prediction method of the settlement used widely in engineering is the imitative method, which needs the settlement curves measured in practical engineering. It needs a lot of data measured in a long period. Thus to predict the settlement as early as possible can not be satisfied in this method. For the artificial neural networks have strong nonlinear mapping and learning ability. In this thesis, a new method is presented for predicting the settlement of soft ground in expressways based on artificial neural networks. The method is of avoiding disturbances from different artificial factors during calculating process by use of measuring data from practical engineering directly. Based on two different manners of function disturbances two network models predicting the settlement of the soft ground, are recommended in the paper. One model is in fact an improvement of the traditional BP network, in which a controller with feedback founction is put into the BP network and forms a dynamic BP network. The controller can be use as the functional disturbances. The other one is nearly the same as the tradstional BP network, in which the disturbances to the network is implemented through the influence function during studying the samples. Examples have shown that the higher accurate and quicker predicted of the settlement after project has been finished can be expected. The method introduced is easy to be used and of widely practical value in engineering projects.