Machine Learning-Assisted Investigation of Anisotropic Elasticity in Metallic Alloys

Materials Today Communications, Journal Year: 2024, Volume and Issue: 40, P. 109950 - 109950

Published: July 26, 2024

Language: Английский

---

Guided analysis of fracture toughness and hydrogen-induced embrittlement crack growth rate in quenched-and-tempered steels using machine learning

International Journal of Pressure Vessels and Piping, Journal Year: 2024, Volume and Issue: 210, P. 105247 - 105247

Published: June 18, 2024

Language: Английский

---

FEM-Driven machine learning approach for characterizing stress magnitude, peak temperature and weld zone deformation in ultrasonic welding of metallic multilayers: application to battery cells

Modelling and Simulation in Materials Science and Engineering, Journal Year: 2024, Volume and Issue: 32(8), P. 085009 - 085009

Published: Oct. 14, 2024

Abstract This study investigates the innovative application of machine learning (ML) models to predict critical parameters—stress magnitude (SM), peak temperature (PT), and plastic strain (PS)—in ultrasonic welding metallic multilayers. Extensive numerical simulations were employed develop evaluate three ML models: Radial Basis Function (RBF), Random Forest (RF), Kernel Ridge Regression (KRR). According results, KRR model demonstrated superior performance, achieving lowest RMSE highest R 2 values 0.068 ( = 0.941) for SM, 0.075 0.929) PT, 0.071 0.946) PS, with fewer data samples required. also exhibited low squared bias variance values, ranging from 1 × 10 − 4 3.2 2.2 3.6 variance, indicating its precision in predicting output targets. Moreover, systematic categorization input features, including material properties, geometrical factors, parameters, highlighted their significant influence on predictive accuracy, particularly crucial role parameters at higher values. Finally, a case copper multilayers underscores model’s effectiveness unraveling complex relationships, providing robust tool optimizing advancing processes.

Language: Английский

---

Local elasticity assessment of unidirectional fiber-reinforced polymer composites through impulse excitation and machine learning

Journal of Reinforced Plastics and Composites, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 9, 2024

This study presents a novel methodology that integrates the Impulse Excitation Technique (IET) and machine learning (ML) to predict local elastic properties within isolated regions of unidirectional polymeric composite plates. The proposed model incorporates fiber volume plate thickness as input parameters leverages first resonance frequencies region at different orientations, thus accounting for composite’s anisotropy. Regression results from deep neural network (DNN) demonstrate robust prediction performance across all output targets in both testing training datasets, with R 2 coefficients surpassing 0.9. exhibits particularly strong predicting Young’s moduli. Additionally, each objective shows sensitivity unique balance parameter weight factors achieving optimal ML predictions. Moreover, parabolic trend fundamental orientations is observed rigidity composites changes. Lastly, comparative between carbon-fiber glass-fiber highlights variations constants, emphasizing effectiveness accurately material properties.

Language: Английский

---

Elastic constant analysis of orthotropic steel sheets using multitask machine learning and the impulse excitation technique

Physica Scripta, Journal Year: 2024, Volume and Issue: 100(1), P. 016014 - 016014

Published: Dec. 11, 2024

Abstract This work presents a novel multitask learning approach featuring dual convolutional neural network (CNN) system for determining the elastic constants of orthotropic rolled steel sheets. In proposed model, resonance frequency spectra from impulse excitation technique are converted into 2D image data. The first CNN focuses on detecting and predicting missing peak intensities, while second utilizes features entire spectrum to predict constants, including E 11 , 22 G 12 . input include raw pixel data alongside three key categories enhanced analysis: image-based (such as mean, median, mode, skewness intensity distributions), spatial relations (including frequency, correlations, local contrast), geometric shape descriptors connectivity). results reveal that optimal number peaks (14), resolution (121 pixels), sample size (20 × 20 0.3 cm) maximize model’s efficiency. Under these conditions, model achieves R 2 values 0.952, 0.902, 0.913, RMSE 1.89 GPa, 3.09 1.92 GPa respectively. It is suggested superior prediction accuracy attributed stability Young’s modulus along rolling direction, which less variable in materials. Furthermore, study finds dependency between weight functions—including features, relations, features—as material’s anisotropy changes, underscoring importance accounting process variability predictive modeling.

Language: Английский

---

Machine learning-driven detection of anomalies in manufactured parts from resonance frequency signatures

Nondestructive Testing And Evaluation, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 23

Published: Nov. 24, 2024

This study aims to enhance the detection and characterisation of anomalies in manufactured parts by integrating machine learning (ML) with resonance frequency spectra data. A key contribution this work is development a novel Impulse Excitation Technique (IET)-based method that effectively evaluates material health identifies subtle defects leveraging numerous mathematical physical metrics as input features. Three models – Random Forest (RF), K-Nearest Neighbor (KNN), Multi-layer Perceptron (MLP) were systematically compared determine most effective for classifying defects, specifically focusing on healthy, cracked, dimensionally deviated samples. Among these, MLP model demonstrated highest performance, achieving Receiver Operating Characteristic (ROC) values 0.963, 0.901, 0.942 each class, respectively. Additionally, SHAP (SHapley Additive exPlanations) analysis showed sensitive specific metrics, improving prediction accuracy. Cracked samples exhibited slight peak broadening negative shifts, while positive shifts missing peaks. Dimensional deviations more pronounced than cracks, making them easier identify enhancing predictive

Language: Английский

---