Applying a novel slime mould algorithm- based artificial neural network to predict the settlement of a single footing on a soft soil reinforced by rigid inclusions

Abstract

Settlements are one of the most important performance indicators for designing footings over soft soils reinforced by rigid inclusions (RI). Although traditional numerical approaches can effectively calculate settlements, the necessary time calculation is important as this problem is a three dimensional one. In this investigation, 369 numerical simulations based on a finite difference (FD) approach were completed to build a database. The collected data include a variation of the footing loading (L), thickness of the load transfer platform (TH), load eccentricity (LE), Young's modulus (E), cohesion (C) and friction angle (F) of the load transfer platform granular material and the compression ratio (CR) of soft soils are input variables. and the settlements are considered as the output variables. Extreme learning machine (ELM), Elman neural network (ENN), generalized regression neural network (GRNN), support vector regression (SVR), Artificial neural network (ANN) and a hybrid model of Slime Mold algorithm- based artificial neural network (SMA-ANN) were used to predict the settlements of a single footing on soft soil reinforced by rigid inclusions. Six performance indicators including the root mean square error (RMSE), the determination coefficient (R²), the mean absolute error (MAE), the prediction accuracy (U₁), the prediction quality (U₂) and the variance accounted for (VAF) are proposed to compare the performance of all the proposed models. The results show that the SMA-ANN was the best model for predicting the settlement of a single footing on a soft soil reinforced by rigid inclusions. The most important input parameters are the friction angle, cohesion and young modulus of the load transfer platform.

Keywords:

• Artificial neural network
• footing
• rigid inclusions
• settlements
• soft soil
• Slime Mold algorithm

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The authors gratefully acknowledge the financial support provided by the China Scholarship Council (ID: 201908070075). This research was complete at the 3SR Laboratory of the University of Grenoble. 3SR-Lab is part of the LabEx Tec 21 (Investissements d’Avenir, Grant Agreement No. ANR-11- LABX-0030).