Prediction of small-scale piles by considering lateral deflection based on Elman Neural Network - Improved Arithmetic...

Area of Science:

• Geotechnical Engineering
• Computational Mechanics
• Artificial Intelligence in Civil Engineering

Background:

• Piles are critical geotechnical structures resisting lateral loads like earthquakes.
• Understanding pile behavior under lateral forces is essential for structural safety.
• Lateral deflection (LD) is a key parameter in pile design and analysis.

Purpose of the Study:

• To investigate the lateral deflection of piles in dry sandy soils under various conditions.
• To develop and evaluate a novel predictive model for pile lateral deflection.
• To compare the performance of different artificial intelligence models for LD prediction.

Main Methods:

• Conducted 192 physical models of piles in dry sandy soils.
• Developed an Elman Neural Network (ENN) model integrated with the Improved Arithmetic Optimizer (IAO) algorithm.
• Utilized comparative models including Elman Neural Network (ENN) and Particle Swarm Optimization-Artificial Neural Network (PSO-ANN) for lateral deflection estimation.

Main Results:

• The ENN-IAO model demonstrated superior reliability in predicting pile lateral deflection.
• Performance evaluation metrics such as Variance Account For, determination coefficient, and Root Mean Squared Error confirmed model validity.
• The proposed ENN-IAO method outperformed the standard ENN and PSO-ANN models in accuracy.

Conclusions:

• The ENN-IAO algorithm offers a reliable and accurate approach for predicting the lateral deflection of small-scale piles.
• This AI-driven methodology enhances the understanding and prediction of pile behavior under lateral loading.
• The findings contribute to improved design practices for piles subjected to lateral forces in sandy environments.