Research

17 pages, 6778 KiB

Open AccessArticle

Enhancing the Design of Experiments on the Fatigue Life Characterisation of Fibre-Reinforced Plastics by Incorporating Artificial Neural Networks

by Christian Witzgall, Moh’d Sami Ashhab and Sandro Wartzack

Materials 2024, 17(3), 729; https://doi.org/10.3390/ma17030729 - 03 Feb 2024

Viewed by 527

Abstract

Fatigue life testing is a complex and costly matter, especially in the case of fibre-reinforced thermoplastics, where other parameters in addition to force alone must be taken into account. The number of tests required therefore increases significantly, especially if the influence of different [...] Read more.

Fatigue life testing is a complex and costly matter, especially in the case of fibre-reinforced thermoplastics, where other parameters in addition to force alone must be taken into account. The number of tests required therefore increases significantly, especially if the influence of different fibre orientations is to be taken into account. It is therefore important to gain the greatest possible amount of knowledge from the limited number of available tests. In order to achieve this, this study aims to utilise adaptive sampling, which is used in numerous areas of computational engineering, for the design of experiments on fatigue life testing. Artificial neural networks (ANNs) are therefore trained on data for the short-fibre-reinforced material PBT GF30, and their areas of greatest model uncertainty are queried. This was undertaken with ANNs from various numbers of hidden layers, which were analysed for their performance. The ideal case turned out to be four hidden layers, for which a squared error as small as 1 × 10⁻³ was recorded. Locally resolved, the ANN was used to identify the region of greatest uncertainty for samples of vertical orientation and small numbers of cycles. With information such as this, additional data can be obtained in such uncertain regions in order to improve the model prediction—almost halving the recorded error to only 0.55 × 10⁻³. In this way, a model of comparable value can be found with less experimental effort, or a model of better quality can be set up with the same experimental effort. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

17 pages, 3801 KiB

Open AccessArticle

A Machine Learning Approach for Modelling Cold-Rolling Curves for Various Stainless Steels

by Julia Contreras-Fortes, M. Inmaculada Rodríguez-García, David L. Sales, Rocío Sánchez-Miranda, Juan F. Almagro and Ignacio Turias

Materials 2024, 17(1), 147; https://doi.org/10.3390/ma17010147 - 27 Dec 2023

Viewed by 747

Abstract

Stainless steel is a cold-work-hardened material. The degree and mechanism of hardening depend on the grade and family of the steel. This characteristic has a direct effect on the mechanical behaviour of stainless steel when it is cold-formed. Since cold rolling is one [...] Read more.

Stainless steel is a cold-work-hardened material. The degree and mechanism of hardening depend on the grade and family of the steel. This characteristic has a direct effect on the mechanical behaviour of stainless steel when it is cold-formed. Since cold rolling is one of the most widespread processes for manufacturing flat stainless steel products, the prediction of their strain-hardening mechanical properties is of great importance to materials engineering. This work uses artificial neural networks (ANNs) to forecast the mechanical properties of the stainless steel as a function of the chemical composition and the applied cold thickness reduction. Multiple linear regression (MLR) is also used as a benchmark model. To achieve this, both traditional and new-generation austenitic, ferritic, and duplex stainless steel sheets are cold-rolled at a laboratory scale with different thickness reductions after the industrial intermediate annealing stage. Subsequently, the mechanical properties of the cold-rolled sheets are determined by tensile tests, and the experimental cold-rolling curves are drawn based on those results. A database is created from these curves to generate a model applying machine learning techniques to predict the values of the tensile strength (Rm), yield strength (Rp), hardness (H), and elongation (A) based on the chemical composition and the applied cold thickness reduction. These models can be used as supporting tools for designing and developing new stainless steel grades and/or adjusting cold-forming processes. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Graphical abstract

24 pages, 2092 KiB

Open AccessArticle

Explainable AI for Material Property Prediction Based on Energy Cloud: A Shapley-Driven Approach

by Faiza Qayyum, Murad Ali Khan, Do-Hyeun Kim, Hyunseok Ko and Ga-Ae Ryu

Materials 2023, 16(23), 7322; https://doi.org/10.3390/ma16237322 - 24 Nov 2023

Cited by 3 | Viewed by 957

Abstract

The scientific community has raised increasing apprehensions over the transparency and interpretability of machine learning models employed in various domains, particularly in the field of materials science. The intrinsic intricacy of these models frequently results in their characterization as “black boxes”, which poses [...] Read more.

The scientific community has raised increasing apprehensions over the transparency and interpretability of machine learning models employed in various domains, particularly in the field of materials science. The intrinsic intricacy of these models frequently results in their characterization as “black boxes”, which poses a difficulty in emphasizing the significance of producing lucid and readily understandable model outputs. In addition, the assessment of model performance requires careful deliberation of several essential factors. The objective of this study is to utilize a deep learning framework called TabNet to predict lead zirconate titanate (PZT) ceramics’ dielectric constant property by employing their components and processes. By recognizing the crucial importance of predicting PZT properties, this research seeks to enhance the comprehension of the results generated by the model and gain insights into the association between the model and predictor variables using various input parameters. To achieve this, we undertake a thorough analysis with Shapley additive explanations (SHAP). In order to enhance the reliability of the prediction model, a variety of cross-validation procedures are utilized. The study demonstrates that the TabNet model significantly outperforms traditional machine learning models in predicting ceramic characteristics of PZT components, achieving a mean squared error (MSE) of 0.047 and a mean absolute error (MAE) of 0.042. Key contributing factors, such as d33, tangent loss, and chemical formula, are identified using SHAP plots, highlighting their importance in predictive analysis. Interestingly, process time is less effective in predicting the dielectric constant. This research holds considerable potential for advancing materials discovery and predictive systems in PZT ceramics, offering deep insights into the roles of various parameters. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

18 pages, 6897 KiB

Open AccessEditor’s ChoiceArticle

Estimation of Cyclic Stress–Strain Curves of Steels Based on Monotonic Properties Using Artificial Neural Networks

by Tea Marohnić, Robert Basan and Ela Marković

Materials 2023, 16(14), 5010; https://doi.org/10.3390/ma16145010 - 15 Jul 2023

Viewed by 1147

Abstract

This paper introduces a novel method for estimating the cyclic stress–strain curves of steels based on their monotonic properties and plastic strain amplitudes, utilizing artificial neural networks (ANNs). ANNs were trained on a substantial number of experimental data for steels, collected from relevant [...] Read more.

This paper introduces a novel method for estimating the cyclic stress–strain curves of steels based on their monotonic properties and plastic strain amplitudes, utilizing artificial neural networks (ANNs). ANNs were trained on a substantial number of experimental data for steels, collected from relevant literature, and divided into subgroups according to alloying elements content (unalloyed, low-alloy, and high-alloy steels). Only monotonic properties that were proven to be relevant for the estimation of points on the stress–strain curve were used. The performance of the developed ANNs was assessed using an independent set of data, and the results were compared to experimental values, values obtained by existing empirical estimation methods, and by previously developed ANNs. The results showed that the new approach which combines relevant monotonic properties and plastic strain amplitudes as inputs to ANNs for cyclic stress–strain curve estimation is better than the previously used approach where ANNs estimate the parameters of the Ramberg–Osgood material model separately. This shows that a more favorable approach to the estimation of cyclic stress–strain behavior would be to directly estimate corresponding material curves using monotonic properties. Additionally, this may also reduce inaccuracies resulting from simplified representations of the actual material behavior inherent in the material model. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

17 pages, 4108 KiB

Open AccessArticle

Advanced Tree-Based Techniques for Predicting Unconfined Compressive Strength of Rock Material Employing Non-Destructive and Petrographic Tests

by Yuzhen Wang, Mahdi Hasanipanah, Ahmad Safuan A. Rashid, Binh Nguyen Le and Dmitrii Vladimirovich Ulrikh

Materials 2023, 16(10), 3731; https://doi.org/10.3390/ma16103731 - 15 May 2023

Cited by 4 | Viewed by 1110

Abstract

The accurate estimation of rock strength is an essential task in almost all rock-based projects, such as tunnelling and excavation. Numerous efforts to create indirect techniques for calculating unconfined compressive strength (UCS) have been attempted. This is often due to the complexity of [...] Read more.

The accurate estimation of rock strength is an essential task in almost all rock-based projects, such as tunnelling and excavation. Numerous efforts to create indirect techniques for calculating unconfined compressive strength (UCS) have been attempted. This is often due to the complexity of collecting and completing the abovementioned lab tests. This study applied two advanced machine learning techniques, including the extreme gradient boosting trees and random forest, for predicting the UCS based on non-destructive tests and petrographic studies. Before applying these models, a feature selection was conducted using a Pearson’s Chi-Square test. This technique selected the following inputs for the development of the gradient boosting tree (XGBT) and random forest (RF) models: dry density and ultrasonic velocity as non-destructive tests, and mica, quartz, and plagioclase as petrographic results. In addition to XGBT and RF models, some empirical equations and two single decision trees (DTs) were developed to predict UCS values. The results of this study showed that the XGBT model outperforms the RF for UCS prediction in terms of both system accuracy and error. The linear correlation of XGBT was 0.994, and its mean absolute error was 0.113. In addition, the XGBT model outperformed single DTs and empirical equations. The XGBT and RF models also outperformed KNN (R = 0.708), ANN (R = 0.625), and SVM (R = 0.816) models. The findings of this study imply that the XGBT and RF can be employed efficiently for predicting the UCS values. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

15 pages, 1654 KiB

Open AccessArticle

Multi-Energy and Fast-Convergence Iterative Reconstruction Algorithm for Organic Material Identification Using X-ray Computed Tomography

by Mihai Iovea, Andrei Stanciulescu, Edward Hermann, Marian Neagu and Octavian G. Duliu

Materials 2023, 16(4), 1654; https://doi.org/10.3390/ma16041654 - 16 Feb 2023

Viewed by 1060

Abstract

In order to significantly reduce the computing time while, at the same time, keeping the accuracy and precision when determining the local values of the density and effective atomic number necessary for identifying various organic material, including explosives and narcotics, a specialized multi-stage [...] Read more.

In order to significantly reduce the computing time while, at the same time, keeping the accuracy and precision when determining the local values of the density and effective atomic number necessary for identifying various organic material, including explosives and narcotics, a specialized multi-stage procedure based on a multi-energy computed tomography investigation within the 20–160 keV domain was elaborated. It consisted of a compensation for beam hardening and other non-linear effects that affect the energy dependency of the linear attenuation coefficient (LAC) in the chosen energy domain, followed by a 3D fast reconstruction algorithm capable of reconstructing the local LAC values for 64 energy values from 19.8 to 158.4 keV, and, finally, the creation of a set of algorithms permitting the simultaneous determination of the density and effective atomic number of the investigated materials. This enabled determining both the density and effective atomic number of complex objects in approximately 24 s, with an accuracy and precision of less than 3%, which is a significantly better performance with respect to the reported literature values. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

16 pages, 4374 KiB

Open AccessEditor’s ChoiceArticle

Assessment of Convolutional Neural Network Pre-Trained Models for Detection and Orientation of Cracks

by Waqas Qayyum, Rana Ehtisham, Alireza Bahrami, Charles Camp, Junaid Mir and Afaq Ahmad

Materials 2023, 16(2), 826; https://doi.org/10.3390/ma16020826 - 14 Jan 2023

Cited by 13 | Viewed by 3175

Abstract

Failure due to cracks is a major structural safety issue for engineering constructions. Human examination is the most common method for detecting crack failure, although it is subjective and time-consuming. Inspection of civil engineering structures must include crack detection and categorization as a [...] Read more.

Failure due to cracks is a major structural safety issue for engineering constructions. Human examination is the most common method for detecting crack failure, although it is subjective and time-consuming. Inspection of civil engineering structures must include crack detection and categorization as a key component of the process. Images can automatically be classified using convolutional neural networks (CNNs), a subtype of deep learning (DL). For image categorization, a variety of pre-trained CNN architectures are available. This study assesses seven pre-trained neural networks, including GoogLeNet, MobileNet-V2, Inception-V3, ResNet18, ResNet50, ResNet101, and ShuffleNet, for crack detection and categorization. Images are classified as diagonal crack (DC), horizontal crack (HC), uncracked (UC), and vertical crack (VC). Each architecture is trained with 32,000 images equally divided among each class. A total of 100 images from each category are used to test the trained models, and the results are compared. Inception-V3 outperforms all the other models with accuracies of 96%, 94%, 92%, and 96% for DC, HC, UC, and VC classifications, respectively. ResNet101 has the longest training time at 171 min, while ResNet18 has the lowest at 32 min. This research allows the best CNN architecture for automatic detection and orientation of cracks to be selected, based on the accuracy and time taken for the training of the model. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

15 pages, 1520 KiB

Open AccessArticle

Determination of TTT Diagrams of Ni-Al Binary Using Neural Networks

by Leonardo Hernández-Flores, Angel-Iván García-Moreno, Enrique Martínez-Franco, Guillermo Ronquillo-Lomelí and Jhon Alexander Villada-Villalobos

Materials 2022, 15(24), 8767; https://doi.org/10.3390/ma15248767 - 08 Dec 2022

Viewed by 1431

Abstract

The heat treatment of a metal is a set of heating and cooling cycles that a metal undergoes to change its microstructure and, therefore, its properties. Temperature–time–transformation (TTT) diagrams are an essential tool for interpreting the resulting microstructures after heat treatments. The present [...] Read more.

The heat treatment of a metal is a set of heating and cooling cycles that a metal undergoes to change its microstructure and, therefore, its properties. Temperature–time–transformation (TTT) diagrams are an essential tool for interpreting the resulting microstructures after heat treatments. The present work describes a novel proposal to predict TTT diagrams of the

γ^{'}

phase for the Ni-Al alloy using artificial neural networks (ANNs). The proposed methodology is composed of five stages: (1) database creation, (2) experimental design, (3) ANNs training, (4) ANNs validation, and (5) proposed models analysis. Two approaches were addressed, the first to predict only the nose point of the TTT diagrams and the second to predict the complete curve. Finally, the best models for each approach were merged to compose a more accurate hybrid model. The results show that the multilayer perceptron architecture is the most efficient and accurate compared to the simulated TTT diagrams. The prediction of the nose point and the complete curve showed an accuracy of

98.07

% and

86.41

%, respectively. The proposed final hybrid model achieves an accuracy of

96.59

%. Full article

(This article belongs to the Special Issue Machine Learning Techniques in Materials Science and Engineering)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Machine Learning Techniques in Materials Science and Engineering

Share This Special Issue

Special Issue Editors

Special Issue Information

Keywords

Published Papers (8 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI