Journal of Materials and Engineering Structures
Volume 12, Numéro 3, Pages 267-280
2025-09-30
A Multi-layer Perceptron (mlp) Neural Network Approach For Predicting Pressuremeter (pmt) Test Results In Geotechnical Engineering
Authors : Aissaoui Soufyane . Bourouis Mohammed El Amin . Zadjaoui Abdeldjalil . Chalabi Youssouf . Mimoune Arezki El Haj .
Abstract
This study proposes a hybrid machine learning approach using a Multi-layer Perceptron Neural Network (MLP) to predict soil pressuremeter parameters, leveraging a database of 120 soil samples from the Algeria railway line project. The MLP model demonstrated superior performance, achieving a high coefficient of determination (R² > 0.93) and low RMSE, indicating strong agreement between predicted and experimental values. Comparative analysis with multiple regression highlighted the MLP’s efficiency in prediction speed and accuracy, though its "black-box" nature limits interpretability of individual parameter influences. While regression provided explicit equations for variable effects, the study underscores the importance of expanding the database with representative inputs to further enhance predictive reliability. The results support MLP as a robust tool for geotechnical forecasting, balancing precision with computational efficiency.
Keywords
Pressuremeter parameters ; Machine learning ; Neural network ; Prediction model
Les articles similaires
Benlaria Ahmed
.
pages 1-8.
Ourdighi Asmaa
.
Benyettou Abdelkader
.
pages 400-414.
Hoang Ha
.
Pham Tuan-dung
.
Nguyen Xuan-tung
.
Ngo Van-minh
.
pages 485-491.
