Reliability and Prediction of Embedment Depth of Sheet pile Walls Using Hybrid ANN with Optimization Techniques

• Published in: Arabian journal for science and engineering (2011) 2022-10, Vol.47 (10), p.12853-12871
• Main Authors: Pradeep, T., GuhaRay, Anasua, Bardhan, Abidhan, Samui, Pijush, Kumar, Sanjay, Armaghani, Danial Jahed
• Format: Article
• Language: English
• Subjects: Ant colony optimization; Artificial neural networks; Civil engineering; Cohesion; Cohesive soils; Criteria; Embedment; Engineering; Evolutionary algorithms; Geotechnical engineering; Humanities and Social Sciences; Internal friction; Machine learning; multidisciplinary; Optimization techniques; Performance prediction; Reliability analysis; Research Article-Civil Engineering; Science; Sheet piles; Soil properties; Swarm intelligence
• ISSN: 2193-567X; 1319-8025; 2191-4281
• DOI: 10.1007/s13369-022-06607-w

Due to the fact that uncertainties in the field of geotechnical engineering are inescapable because this part of civil engineering deals largely with natural materials, the dependability analysis of geotechnical structures has gotten a lot of attention in recent decades. This study looks at the reliability of embedded depth of cantilever sheet pile walls that are embedded in cohesive soil and backfilled with cohesionless soil. For the reliability analysis, the first-order reliability method (FORM) was employed to analyse the cantilever sheet pile. With the use of various optimization strategies, namely artificial bee colony, ant colony optimization, ant lion colony, imperialist competitive algorithm, shuffled complex evolution and teaching–learning-based optimization, the widely used artificial neural network (ANN) is used to forecast the embedment depth of a cantilever sheet pile wall. The input variables for predicting pile embedded depth include soil properties such as cohesiveness (c), angle of internal friction ( φ ) and unit weight (γ). To assess the models` performance, the reliability index, rank analysis, Taylor diagram, accuracy matrix, DDR criterion, AIC criterion, and OBJ criterion are utilised. Based on the experimental results, the hybrid models of ANN and teaching–learning-based optimization (ANN-TLO) perform better in predicting the reliability of cantilever sheet pile walls. The proposed approach can be utilised to quantify the risk associated with such structures in civil engineering projects as an alternative tool.