Estimation of the settlement of pile head using ANN and multivariate linear regression based on the results of load transfer method

• Published in: Open Engineering (Warsaw) Vol. 14; no. 1; pp. 33 - 44
• Main Authors: Al-Awad, Khabeer, Fattah, Mohammed Y., Al-Zuheriy, Ahmed Sh. J.
• Format: Journal Article
• Language: English
• Published: De Gruyter 01.07.2024
• Subjects: artificial neural networks; correlation coefficient; load transfer method; machine learning; multivariate linear regression; top settlement of piles
• ISSN: 2391-5439; 2391-5439
• DOI: 10.1515/eng-2024-0026

Artificial neural networks, machine learning, and data preparation are normally implemented in a wide range of real-world problems, especially in geotechnical applications with optimistic prospects of accurate procedure outcomes. This technique has been utilized to precisely predict the top settlement of piles with various piles and soil parameters. Generally, the pile settlement is an essential requirement to produce a secure structure and has high-performance services. The current article presents the fitting of the artificial neural network (ANN) outcomes by calculating the coefficient of correlation between the predicted and the measured or calculated value of pile settlement. The ANN algorithm is developed using Python 3.9 IDLE and open-source libraries such as Keras, sklearn, Numpy, matplotlib, pandas, and Tensorflow. Because of random training and test performance, the model has been run at least ten times. The ANN model score and are compared for all runs in the testing phase. The higher score and values are chosen. Moreover, the Multivariate Linear Regression with the sklearn library is also offered in this article and utilized to produce a pile settlement formula by applying the same dataset used in ANN. The score and for choosing the first run of the ANN are 99.95% and 0.9631, respectively, while the correlation coefficient for the Multivariate Linear Regression in the training and testing phases is 0.972 and 0.919, respectively. Both techniques illustrate considerable results.