A State-of-the-Art Comprehensive Review: Application of Artificial Intelligence in Predicting the Load-Bearing Capacity of Piles

Abstract

Artificial intelligence models have widely replaced traditional statistical modeling tools as reliable alternatives for capacity estimation. Estimating the load-bearing capacity of piles using analytical formulas can yield a range of results, which can be attributed to the simplifying assumptions incorporated into the models. Another innovative method for predicting the load-bearing capacity of piles is the pile dynamic analysis test, which is based on the theory of one-dimensional wave propagation. Simpler methods such as field tests like the cone penetration test and standard penetration test have also been proposed by geotechnical engineers for evaluating pile capacity. This article provides an advanced review by organizing, summarizing, and categorizing the results of the most commonly used machine learning methods, innovative algorithms, and combined approaches-based data from static and dynamic loading tests and in situ tests in determining the load-bearing capacity of piles. To this end, 135 articles published in the past three decades were considered, taking into account reputable journals, top researchers, country contributions, and research institutions to highlight the annual research trends in this field. The paper presents a review of the application of various artificial intelligence algorithms such as ANN, GP, GB, regression algorithms, RF, SVM, DT, and Fuzzy logic in determining the load-bearing capacity of piles. Additionally, the sub-branches related to the ANN and GB algorithms were examined, and finally, studies conducted using optimization algorithms in conjunction with all the artificial intelligence methods were reviewed and presented. Furthermore, artificial intelligence techniques were classified into separate clusters using the VOSviewer clustering program, and their relationships were visualized through the thickness of connecting lines to identify their application domains. These findings can serve as a guide for researchers seeking to utilize artificial intelligence techniques in the field of pile load-bearing capacity in the future.