Modeling and sensitivity analysis of bearing capacity in driven piles using hybrid ANN–PSO algorithm

• Published in: Arabian journal of geosciences Vol. 15; no. 3
• Main Authors: Arjomand, Mohammad Ali, Mostafaei, Yashar, Kutanaei, Saman Soleimani
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.02.2022; Springer Nature B.V
• Subjects: Algorithms; Bearing capacity; Driven piles; Earth and Environmental Science; Earth science; Earth Sciences; Friction; Internal friction; Mathematical models; Original Paper; Parameter sensitivity; Parameters; Piles; Polynomials; Scaling; Sensitivity analysis; Soil; Soil density; Soil layers; Soil settlement; Soil-pile interaction; ANN; Bearing capacity; PSO; Sensitivity analyses
• ISSN: 1866-7511; 1866-7538
• DOI: 10.1007/s12517-022-09557-7

Piles are widely used to transfer load to the underlying soil layers and reduce settlement. It is difficult to determine the exact bearing capacity (BC) of piles due to the large number of effective parameters. This study combined the PSO and ANN algorithms to provide a polynomial relation for the prediction of bearing capacity in driven piles. Sensitivity analysis examined the effect of the input parameters including flap number (FL), pile length (L) and cross-sectional area (A), internal friction angle (ϕ), soil drained cohesion (C), soil density (γ) and soil–pile interaction friction angle (δ) on the output parameter (BC). This study used the data from 100 static loading tests on the piles. The results of this study showed that the quadratic relation obtained from the PSO–ANN and PSO methods for the prediction of BC yielded the R 2 values of 0.912 and 0.957, respectively. The scaling of input data also plays an important role in the ANN performance. The results of sensitivity analyses revealed that γ, A, δ, L, ϕ, C and FL have the greatest effect, respectively. The model presented by PSO–ANN method can be used with a high reliability to predict BC.