Performance Evaluation and Triangle Diagram of Deep Learning Models for Embedment Depth Prediction in Cantilever Sheet Piles

Pradeep T1

Divesh Ranjan Kumar2

Nitish Kumar3

Warit Wipulanusat2,Email

Suraparb Keawsawasvong4

Jirapon Sunkpho5

1Department of Civil Engineering, National Institute of Technology Andhra Pradesh, Tadepalligudem, Andhra Pradesh 534101, India.

2Research Unit in Data Science and Digital Transformation, Department of Civil Engineering, Faculty of Engineering, Thammasat School of Engineering, Thammasat University, Pathumthani 12121, Thailand.

3Department of Civil Engineering, National Institute of Technology, Patna, Bihar 800005, India.

4Research Unit in Sciences and Innovative Technologies for Civil Engineering Infrastructures, Department of Civil Engineering, Faculty of Engineering, Thammasat School of Engineering, Thammasat University, Pathumthani 12121, Thailand.

5College of Innovation, Thammasat University, Bangkok 10200, Thailand.

Abstract

Sheet piles are essential for maintaining the stability and retention of soil in various applications, including railway and highway embankments, offshore structures, post-excavation sites, and slope stabilization projects. The required depth of sheet piles is contingent upon factors such as soil characteristics, groundwater conditions, and the employed construction method. This study focused on predicting the embedment depth of cantilever sheet pile walls in cohesive soil with a cohesionless soil backfill. Artificial intelligence (AI) techniques, specifically deep neural networks (DNNs), recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and bidirectional long short-term memory (Bi-LSTM) networks, are applied for this purpose. Performance evaluation is conducted through rank analysis, performance parameter determination, and comparison of the actual versus predicted curves, accompanied by an error plot. A triangle diagram is introduced as a graphical representation to assess the performance of different datasets or models. External validation was conducted to evaluate the generalizability of the model. All the proposed models meet these criteria, with the DNN model demonstrating superior performance. This comprehensive evaluation demonstrated the effectiveness and robustness of the DNN model as a practical tool for predicting the embedment depth of sheet piles.

Performance Evaluation and Triangle Diagram of Deep Learning Models for Embedment Depth Prediction in Cantilever Sheet Piles