Fast structural analysis of concrete thin-shells using deep learning

Maxime Pollet; Paul Shepherd; Will Hawkins; Eduardo Costa

doi:10.1016/j.compstruc.2025.108042

Fast structural analysis of concrete thin-shells using deep learning

Maxime Pollet
, Paul Shepherd
, Will Hawkins
, Eduardo Costa

University of the West of England

Research output: Contribution to journal › Article › peer-review

2 Citations (SciVal)

Abstract

The present paper investigates the use of deep learning models as fast structural analysis tools for the design of concrete thin-shells. A dataset of 20,000 thin-shells with various geometric and material properties is generated. The buckling factor and the stress fields of each thin-shell under design loads are determined using Finite Element analysis. Three different types of deep learning models – Multilayer Perceptron (MLP), Convolutional Neural Network (CNN) and Graph Neural Network (GNN) – are then trained for buckling and stress prediction. For both prediction tasks, the MLP and the CNN are found to be the best performing models, reaching errors below 0.31 % for buckling prediction, and below 0.51 % for peak stress prediction. These results demonstrate the ability of such models to act as fast structural analysis tools for concrete thin-shells. Deep learning models could therefore enable faster and wider design space exploration during the shape optimisation of concrete thin-shells.

Original language	English
Article number	108042
Journal	Computers and Structures
Volume	320
Early online date	20 Nov 2025
DOIs	https://doi.org/10.1016/j.compstruc.2025.108042
Publication status	Published - 31 Jan 2026

Data Availability Statement

Data will be made available on request.

Funding

This work was funded by the University of Bath. It is also aligned with the ACORN research project, which was funded by UK Research and Innovation ( EP/S031316/1 ). The authors also gratefully acknowledge the University of Bath’s Research Computing Group (doi.org/10.15125/b6cd-s854) [47] for their support in this work. For the purpose of open access, the authors have applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.

Funders	Funder number
University of Bath
UK Research and Innovation	EP/S031316/1

Keywords

Thin-shell
Finite element analysis
Surrogate
Deep learning
Multilayer perceptron
Convolutional neural network
Graph neural network

Access to Document

10.1016/j.compstruc.2025.108042Licence: CC BY

Cite this

@article{43a6f50cbfcd49518a55a99aa59745f4,

title = "Fast structural analysis of concrete thin-shells using deep learning",

abstract = "The present paper investigates the use of deep learning models as fast structural analysis tools for the design of concrete thin-shells. A dataset of 20,000 thin-shells with various geometric and material properties is generated. The buckling factor and the stress fields of each thin-shell under design loads are determined using Finite Element analysis. Three different types of deep learning models {\textendash} Multilayer Perceptron (MLP), Convolutional Neural Network (CNN) and Graph Neural Network (GNN) {\textendash} are then trained for buckling and stress prediction. For both prediction tasks, the MLP and the CNN are found to be the best performing models, reaching errors below 0.31 \% for buckling prediction, and below 0.51 \% for peak stress prediction. These results demonstrate the ability of such models to act as fast structural analysis tools for concrete thin-shells. Deep learning models could therefore enable faster and wider design space exploration during the shape optimisation of concrete thin-shells.",

keywords = "Thin-shell, Finite element analysis, Surrogate, Deep learning, Multilayer perceptron, Convolutional neural network, Graph neural network",

author = "Maxime Pollet and Paul Shepherd and Will Hawkins and Eduardo Costa",

year = "2026",

month = jan,

day = "31",

doi = "10.1016/j.compstruc.2025.108042",

language = "English",

volume = "320",

journal = "Computers and Structures",

issn = "0045-7949",

publisher = "Elsevier",

}

TY - JOUR

T1 - Fast structural analysis of concrete thin-shells using deep learning

AU - Pollet, Maxime

AU - Shepherd, Paul

AU - Hawkins, Will

AU - Costa, Eduardo

PY - 2026/1/31

Y1 - 2026/1/31

N2 - The present paper investigates the use of deep learning models as fast structural analysis tools for the design of concrete thin-shells. A dataset of 20,000 thin-shells with various geometric and material properties is generated. The buckling factor and the stress fields of each thin-shell under design loads are determined using Finite Element analysis. Three different types of deep learning models – Multilayer Perceptron (MLP), Convolutional Neural Network (CNN) and Graph Neural Network (GNN) – are then trained for buckling and stress prediction. For both prediction tasks, the MLP and the CNN are found to be the best performing models, reaching errors below 0.31 % for buckling prediction, and below 0.51 % for peak stress prediction. These results demonstrate the ability of such models to act as fast structural analysis tools for concrete thin-shells. Deep learning models could therefore enable faster and wider design space exploration during the shape optimisation of concrete thin-shells.

AB - The present paper investigates the use of deep learning models as fast structural analysis tools for the design of concrete thin-shells. A dataset of 20,000 thin-shells with various geometric and material properties is generated. The buckling factor and the stress fields of each thin-shell under design loads are determined using Finite Element analysis. Three different types of deep learning models – Multilayer Perceptron (MLP), Convolutional Neural Network (CNN) and Graph Neural Network (GNN) – are then trained for buckling and stress prediction. For both prediction tasks, the MLP and the CNN are found to be the best performing models, reaching errors below 0.31 % for buckling prediction, and below 0.51 % for peak stress prediction. These results demonstrate the ability of such models to act as fast structural analysis tools for concrete thin-shells. Deep learning models could therefore enable faster and wider design space exploration during the shape optimisation of concrete thin-shells.

KW - Thin-shell

KW - Finite element analysis

KW - Surrogate

KW - Deep learning

KW - Multilayer perceptron

KW - Convolutional neural network

KW - Graph neural network

UR - https://www.scopus.com/pages/publications/105022244540

U2 - 10.1016/j.compstruc.2025.108042

DO - 10.1016/j.compstruc.2025.108042

M3 - Article

SN - 0045-7949

VL - 320

JO - Computers and Structures

JF - Computers and Structures

M1 - 108042

ER -

Fast structural analysis of concrete thin-shells using deep learning

Abstract

Data Availability Statement

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Dataset for "Fast structural analysis of concrete thin-shells using deep learning"

ConcreteShellFEA: A surrogate modelling dataset for the buckling and stress behaviour of concrete thin-shells

Cite this

Fast structural analysis of concrete thin-shells using deep learning

Abstract

Data Availability Statement

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Datasets

Dataset for "Fast structural analysis of concrete thin-shells using deep learning"

ConcreteShellFEA: A surrogate modelling dataset for the buckling and stress behaviour of concrete thin-shells

Cite this