Modeling And Research Of Artificial Neural Network On Bearing Characteristic Of Long Rock-socketed Piles

Now, all are not very intensive to the pressure, distortion and destructive mechanism of the long rock-socketed piles. And systemic testing research on them are not enough. Learning from the main influencing factors of pile bearing capacity and settlement, we know that the soil around pile is important to pile bearing capacity and its settlement, besides the pile intensity and its geometry. Because the characteristic of soil is very complex, it conduces to that estimating pile bearing capacity accurately is very difficult. So, empirical and semi- empirical methods are often applied to pile ultimate bearing capacity and settlement. But all of these have difference with the fact due to the assumption failed short of it. Owing to the ANN has its superiority and adaptability when dealing with empirical, non-linear and complex question, the author want to establish its model and use it to forecast pile ultimate bearing capacity and settlement. In the paper, the author collected the data of the long rock-socketed piles in Hangzhou. It reduced input data serving the ANN utilizing Principal Component Analysis. The author has established the ANN-model about rock-socketed piles in Hangzhou and has improved on it (especially combined the G.A.). It has confirmed the feasibility and correctness of the model through comparing the results. Analyzing on the influencing factors of the ultimate bearing capacity and settlement, it has drawn some conclusion which avail to engineering.