A Self-learning Framework for Estimating Load Transfer Curves from Un-instrumented Pile Loading Tests

摘要

An attempt is made to estimate the load transfer curves from a database of un-instrumented pile loading tests using a self-learning algorithm, an inverse analysis framework capable of handling ill posed problems. The aim is to extract the useful but masked information about load transfer encapsulated in the observed load-deformation response of axially loaded piles. From the results obtained, the estimated load transfer curves tend to produce better load deformation models compared to the predictions based on conventional t-z curves, thus indicating that the self-learning algorithm has succeeded in extracting the load transfer data indirectly from un-instrumented pile tests.