A machine learning based approach with an augmented dataset for fatigue life prediction of additively manufactured Ti-6Al-4V samples

Jan Horňas; Jiří Běhal; Petr Homola; Radek Doubrava; Martin Holzleitner; Sascha Senck

doi:10.1016/j.engfracmech.2023.109709

A machine learning based approach with an augmented dataset for fatigue life prediction of additively manufactured Ti-6Al-4V samples

Jan Horňas, Jiří Běhal, Petr Homola, Radek Doubrava, Martin Holzleitner, Sascha Senck

Research Center Wels

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

15 Citations (Scopus)

Abstract

The article deals with the prediction of fatigue life using a machine learning (ML) approach. The original dataset is based on the parameters of defects obtained by micro-computed tomography (µ-CT) prior to fatigue tests, stress level and the fatigue life of additively manufactured (AM) Ti-6Al-4V samples. As the original dataset is considered too small to train a comprehensive ML model, the study proposed a novel approach for dataset augmentation. Dataset augmentation is done using inverse transform sampling and multivariate radial basis function (RBF) interpolation with various values of the smoothing parameter (λ). Finally, ML model accuracy is improved up to 0.953 of coefficient of determination (R²).

Original language	English
Article number	109709
Pages (from-to)	1
Number of pages	19
Journal	Engineering Fracture Mechanics
Volume	293
DOIs	https://doi.org/10.1016/j.engfracmech.2023.109709
Publication status	Published - 1 Dec 2023

Keywords

Additive manufacturing (AM)
Dataset augmentation
Fatigue life prediction
Machine learning (ML)
Radial basis function (RBF) interpolation

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.engfracmech.2023.109709

Cite this

@article{a0603d28009245b3905f762e297b7a3b,

title = "A machine learning based approach with an augmented dataset for fatigue life prediction of additively manufactured Ti-6Al-4V samples",

abstract = "The article deals with the prediction of fatigue life using a machine learning (ML) approach. The original dataset is based on the parameters of defects obtained by micro-computed tomography (µ-CT) prior to fatigue tests, stress level and the fatigue life of additively manufactured (AM) Ti-6Al-4V samples. As the original dataset is considered too small to train a comprehensive ML model, the study proposed a novel approach for dataset augmentation. Dataset augmentation is done using inverse transform sampling and multivariate radial basis function (RBF) interpolation with various values of the smoothing parameter (λ). Finally, ML model accuracy is improved up to 0.953 of coefficient of determination (R2).",

keywords = "Additive manufacturing (AM), Dataset augmentation, Fatigue life prediction, Machine learning (ML), Radial basis function (RBF) interpolation, Mikro-Computertomographie, Additive manufacturing, Titanium alloys, Zerst{\"o}rungsfreie Werkstoffpr{\"u}fung, Porosit{\"a}t",

author = "Jan Hor{\v n}as and Ji{\v r}{\'i} B{\v e}hal and Petr Homola and Radek Doubrava and Martin Holzleitner and Sascha Senck",

note = "Publisher Copyright: {\textcopyright} 2023",

year = "2023",

month = dec,

day = "1",

doi = "10.1016/j.engfracmech.2023.109709",

language = "English",

volume = "293",

pages = "1",

journal = "Engineering Fracture Mechanics",

issn = "0013-7944",

publisher = "Elsevier",

}

TY - JOUR

T1 - A machine learning based approach with an augmented dataset for fatigue life prediction of additively manufactured Ti-6Al-4V samples

AU - Horňas, Jan

AU - Běhal, Jiří

AU - Homola, Petr

AU - Doubrava, Radek

AU - Holzleitner, Martin

AU - Senck, Sascha

PY - 2023/12/1

Y1 - 2023/12/1

N2 - The article deals with the prediction of fatigue life using a machine learning (ML) approach. The original dataset is based on the parameters of defects obtained by micro-computed tomography (µ-CT) prior to fatigue tests, stress level and the fatigue life of additively manufactured (AM) Ti-6Al-4V samples. As the original dataset is considered too small to train a comprehensive ML model, the study proposed a novel approach for dataset augmentation. Dataset augmentation is done using inverse transform sampling and multivariate radial basis function (RBF) interpolation with various values of the smoothing parameter (λ). Finally, ML model accuracy is improved up to 0.953 of coefficient of determination (R2).

AB - The article deals with the prediction of fatigue life using a machine learning (ML) approach. The original dataset is based on the parameters of defects obtained by micro-computed tomography (µ-CT) prior to fatigue tests, stress level and the fatigue life of additively manufactured (AM) Ti-6Al-4V samples. As the original dataset is considered too small to train a comprehensive ML model, the study proposed a novel approach for dataset augmentation. Dataset augmentation is done using inverse transform sampling and multivariate radial basis function (RBF) interpolation with various values of the smoothing parameter (λ). Finally, ML model accuracy is improved up to 0.953 of coefficient of determination (R2).

KW - Additive manufacturing (AM)

KW - Dataset augmentation

KW - Fatigue life prediction

KW - Machine learning (ML)

KW - Radial basis function (RBF) interpolation

KW - Mikro-Computertomographie

KW - Additive manufacturing

KW - Titanium alloys

KW - Zerstörungsfreie Werkstoffprüfung

KW - Porosität

UR - http://www.scopus.com/inward/record.url?scp=85176265491&partnerID=8YFLogxK

U2 - 10.1016/j.engfracmech.2023.109709

DO - 10.1016/j.engfracmech.2023.109709

M3 - Article

AN - SCOPUS:85176265491

SN - 0013-7944

VL - 293

SP - 1

JO - Engineering Fracture Mechanics

JF - Engineering Fracture Mechanics

M1 - 109709

ER -

A machine learning based approach with an augmented dataset for fatigue life prediction of additively manufactured Ti-6Al-4V samples

Abstract

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this