Estimation on concrete trench barrier effect: a hybrid experimental method based on neural network model
LiDaniel1, JunWang2,MingjieLi1, XiLi1, ZhenpingSun3,ZhengLeng1
COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(12D) 14-23
1Concrete Co. Ltd. Of cscec5b, Changsha, Hunan, China
2China West Construction Group Co.,LTD, Chengdu, Sichuang, China
In this paper, the neural network is introduced as the basic model, and the barrier effect of the vertical seismic wave amplitude is estimated by the three dimensional concrete trench. The back propagation training method is adopted in the paper. The main procedure includes inputting the related parameters of trench barrier seismic wave, analyzing by the pre-processed method and factor algorithm, and selecting six parameters such as the trench section, the distance between the epicenter and the trench, the immersed depth in the foundation and the in-filled material property, etc. The disadvantage of the node numbers and the network learning, which is made by the previous trial-and-error method, should be improved by the Cascade Correlation learning procedure, the automatically adjusted learning speed ratio and the inertia factor algorithm. In addition, the neural network should be built up and taken the average vertical amplitude as the output value. The experiment results show that the built neural network model can simulate the barrier effect of the vertical seismic wave amplitude analyzed by the three dimensional concrete trench, and its accuracy of the predicted results is better.