XFEM–ANN approach to predict the fatigue performance of a composite patch repaired aluminium panel

Siddharth Suman; Kishan Dwivedi; Samanvay Anand; Himanshu Pathak

doi:10.1016/j.jcomc.2022.100326

XFEM–ANN approach to predict the fatigue performance of a composite patch repaired aluminium panel

Siddharth Suman
, Kishan Dwivedi
, Samanvay Anand
, Himanshu Pathak^*

^*Corresponding author for this work

Indian Institute of Technology Mandi

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

22 Citations (Scopus)

Abstract

A novel extended finite element method (XFEM)-artificial neural network (ANN) approach is proposed to predict the performance of aluminium panels repaired single-sided with a composite patch. The optimized deep neural network model developed using data sets generated from XFEM—benchmarked against experimental results—has the architecture of 4-17-17-2 and it predicts the stress intensity factor and fatigue life cycle with the maximum error of 0.15 % and 0.25 %, respectively. Global sensitivity analysis of the model is performed based on the Shapley values and it reveals that lamina orientation is the least influencing. A comparison of computational time reveals that 100 % reduction is achieved by using the optimized ANN model. The findings establish the use of ANN for aiding as well as accelerating the composite patch repair process for various engineering applications.

Original language	English
Article number	100326
Journal	Composites Part C: Open Access
Volume	9
DOIs	https://doi.org/10.1016/j.jcomc.2022.100326
Publication status	Published - Oct 2022
MoE publication type	A1 Journal article-refereed

Funding

This research received financial support from Aeronautics Research & Development Board (DRDO), New Delhi, India through grant file no. 1051914 .

Keywords

Composite patch
Failure analysis
Neural networks
Repair technologies
Stress intensity factors

Access to Document

10.1016/j.jcomc.2022.100326Licence: CC BY

Cite this

@article{7341024993af422e9a0ff018c8782e8a,

title = "XFEM–ANN approach to predict the fatigue performance of a composite patch repaired aluminium panel",

abstract = "A novel extended finite element method (XFEM)-artificial neural network (ANN) approach is proposed to predict the performance of aluminium panels repaired single-sided with a composite patch. The optimized deep neural network model developed using data sets generated from XFEM—benchmarked against experimental results—has the architecture of 4-17-17-2 and it predicts the stress intensity factor and fatigue life cycle with the maximum error of 0.15 \% and 0.25 \%, respectively. Global sensitivity analysis of the model is performed based on the Shapley values and it reveals that lamina orientation is the least influencing. A comparison of computational time reveals that 100 \% reduction is achieved by using the optimized ANN model. The findings establish the use of ANN for aiding as well as accelerating the composite patch repair process for various engineering applications.",

keywords = "Composite patch, Failure analysis, Neural networks, Repair technologies, Stress intensity factors",

author = "Siddharth Suman and Kishan Dwivedi and Samanvay Anand and Himanshu Pathak",

note = "Publisher Copyright: {\textcopyright} 2022 The Author(s)",

year = "2022",

month = oct,

doi = "10.1016/j.jcomc.2022.100326",

language = "English",

volume = "9",

journal = "Composites Part C: Open Access",

issn = "2666-6820",

publisher = "Elsevier",

}

TY - JOUR

T1 - XFEM–ANN approach to predict the fatigue performance of a composite patch repaired aluminium panel

AU - Suman, Siddharth

AU - Dwivedi, Kishan

AU - Anand, Samanvay

AU - Pathak, Himanshu

PY - 2022/10

Y1 - 2022/10

N2 - A novel extended finite element method (XFEM)-artificial neural network (ANN) approach is proposed to predict the performance of aluminium panels repaired single-sided with a composite patch. The optimized deep neural network model developed using data sets generated from XFEM—benchmarked against experimental results—has the architecture of 4-17-17-2 and it predicts the stress intensity factor and fatigue life cycle with the maximum error of 0.15 % and 0.25 %, respectively. Global sensitivity analysis of the model is performed based on the Shapley values and it reveals that lamina orientation is the least influencing. A comparison of computational time reveals that 100 % reduction is achieved by using the optimized ANN model. The findings establish the use of ANN for aiding as well as accelerating the composite patch repair process for various engineering applications.

AB - A novel extended finite element method (XFEM)-artificial neural network (ANN) approach is proposed to predict the performance of aluminium panels repaired single-sided with a composite patch. The optimized deep neural network model developed using data sets generated from XFEM—benchmarked against experimental results—has the architecture of 4-17-17-2 and it predicts the stress intensity factor and fatigue life cycle with the maximum error of 0.15 % and 0.25 %, respectively. Global sensitivity analysis of the model is performed based on the Shapley values and it reveals that lamina orientation is the least influencing. A comparison of computational time reveals that 100 % reduction is achieved by using the optimized ANN model. The findings establish the use of ANN for aiding as well as accelerating the composite patch repair process for various engineering applications.

KW - Composite patch

KW - Failure analysis

KW - Neural networks

KW - Repair technologies

KW - Stress intensity factors

UR - https://www.scopus.com/pages/publications/85142180474

U2 - 10.1016/j.jcomc.2022.100326

DO - 10.1016/j.jcomc.2022.100326

M3 - Article

AN - SCOPUS:85142180474

SN - 2666-6820

VL - 9

JO - Composites Part C: Open Access

JF - Composites Part C: Open Access

M1 - 100326

ER -

XFEM–ANN approach to predict the fatigue performance of a composite patch repaired aluminium panel

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this