Machine learning pipeline for Structure–Property modeling in Mg-alloys using microstructure and texture descriptors

Research output: Journal contributionsJournal articlesResearchpeer-review

Authors

Identifying the relationships between material structure and mechanical properties has been crucial for accelerating the exploration of the material design space for advanced alloys. However, traditional approaches for magnesium (Mg) alloys often fall short in providing quantitative and broadly applicable structure–property linkages. To address this challenge, a comprehensive machine learning pipeline is presented for structure–property modeling in extruded Mg-alloys, leveraging both microstructure and texture descriptors derived from experimental data. The pipeline encompasses a robust workflow for data extraction from optical microscopy and X-ray diffraction, advanced image processing and deep learning techniques for microstructure binarization and grain statistics, and the computation of statistical descriptors including n-point spatial correlations, gram matrices for microstructure, and generalized spherical harmonics (GSH) for texture. Dimensionality reduction techniques such as principal component analysis (PCA), isomap, and autoencoders are employed to manage the high-dimensionality of the descriptor space. Subsequently, non-linear regression models—Gaussian Process, XGBoost, and Multi-Layer Perceptron regressors—are evaluated to predict mechanical properties, specifically strain hardening exponent (n) and yield stress (σy). Our results demonstrate that XGBoost consistently outperforms other regressors, achieving a notably low mean absolute percentage error (MAPE) of 6.67% for strain hardening exponent and 7.01% for yield stress, using a combination of PCA-reduced 3-point spatial correlations and isomap-reduced gram matrices as microstructure descriptors, and isomap-reduced GSH coefficients as texture descriptors at a 150μm length scale. Shapley Additive exPlanations (SHAP) analysis further reveals that texture descriptors and aspect ratio distribution are the most influential features in predicting mechanical properties. This established ML framework for structure–property modeling in Mg-alloys, surpasses state-of-the-art benchmarks and provides a valuable template for materials design and discovery.

Original languageEnglish
Article number121132
JournalActa Materialia
Volume295
ISSN1359-6454
DOIs
Publication statusPublished - 15.08.2025

Bibliographical note

Publisher Copyright:
© 2025 The Authors

    Research areas

  • Deep learning, Descriptor, Grain boundary, Machine learning, Magnesium alloys, Microstructure, Orientation distribution function, Property prediction, Structure–Property, Texture
  • Engineering

Recently viewed

Publications

  1. Set-oriented numerical computation of rotation sets
  2. Linear Generalised Model Predictive Control to Avoid Input Saturation through Matrix Conditions
  3. Calculation of Average Mutual Information (AMI) and false-nearest neighbors (FNN) for the estimation of embedding parameters of multidimensional time series in matlab
  4. Controlling processing usage at user level
  5. Enhancing Performance of Level System Modeling with Pseudo-Random Signals
  6. Control of a Sun Tracking Robot Based on Adaptive Sliding Mode Control with Kalman Filtering and Model Predictive Control
  7. Age-related differences in processing visual device and task characteristics when using technical devices
  8. Implicit statistical learning and working memory predict EFL development and written task outcomes in adolescents
  9. Tracing exploratory modes in digital collections of museum Web sites using reverse information architecture
  10. A Wavelet Packet Tree Denoising Algorithm for Images of Atomic-Force Microscopy
  11. Machine learning pipeline for Structure–Property modeling in Mg-alloys using microstructure and texture descriptors
  12. A geometric algorithm for the output functional controllability in general manipulation systems and mechanisms
  13. Random measurement and prediction errors limit the practical relevance of two velocity sensors to estimate the 1RM back squat
  14. Homogenization modeling of thin-layer-type microstructures
  15. The learning net - an interactive representation of shared knowledge
  16. Evaluating entity annotators using GERBIL
  17. Supporting the Decision of the Order Processing Strategy by Using Logistic Models
  18. Performance and Comfort when Using Motion-Controlled Tools in Complex Tasks
  19. Perfect anti-windup in output tracking scheme with preaction
  20. Exploiting linear partial information for optimal use of forecasts. With an application to U.S. economic policy
  21. Differences Between Classical and Bayesian Estimates for Mixed Logit Models
  22. Digital Control of a Camless Engine Using Lyapunov Approach with Backward Euler Approximation
  23. A discrete-time fractional order PI controller for a three phase synchronous motor using an optimal loop shaping approach
  24. A Control Scheme for PMSMs using Model Predictive Control and a Feedforward Action in the Presence of Saturated Inputs
  25. Springback prediction and reduction in deep drawing under influence of unloading modulus degradation
  26. Tracking control of a piezo-hydraulic actuator using input-output linearization and a Cascaded Extended Kalman Filter structure
  27. The relationship between long-term memory ability and instructed second language learning
  28. Effective informational entropy reduction in multi-robot systems based on real-time TVS
  29. A Python toolbox for the numerical solution of the Maxey-Riley equation
  30. Robust decoupling through algebraic output feedback in manipulation systems