55 / 2019-06-21 22:43:22

Modelling of pile displacement of concrete piles using an evolutionary ANN approach

deep foundations; computational intelligence; sandy soil; concrete pile; Levenberg-Marquardt (LM) algorithm; pile uplift capacity.

水库地质灾害防治与生态环境保护 > 地质灾害大数据与人工智能

This study was performed to examine pile uplift bearing capacity and to develop a novel approach to simulate pile load-settlement response using a new supervised computational intelligence (CI) approach. To reach the planned aim, a series of experimental studies were conducted on concrete piles subjected to uplift loading, with depth-to-diameter ratios of 12, 17, and 25. The concrete piles were penetrated in three sand densities: loose (18%), medium (51%) and dense (83%). According to the statistical analysis, pile effective length (lc), applied load (P), pile flexural rigidity (EA), sand-pile friction angle (δ), and pile aspect ratio (lc/d) were pronounced to play a key role, at different contribution levels, on model output. To evaluate and verify the efficiency of the proposed approach, comparison have been made between the employed training algorithm with experimental pile load-test, and with those specified by a number of design procedures. The results revealed an outstanding agreement between the target and predicted pile-load settlement, thus yielding a coefficient of correlation (R), and a minimum root mean square error (RMSE) of 0.985 and 0.059 respectively, thus, in parallel with a relatively insignificant mean square error level (MSE) of 0.0039.