Hardness prediction using machine learning regression and classification models

What is it about?

The current study focuses on the prediction of metal hardness distribution in upsetting tests for different compositions of ZrO2 embedded with an aluminum matrix using machine learning algorithms and finite element (FE) analysis. The mass fraction of the ZrO2 particles varied from 4% to 8%, and three sets of solid cylindrical rods with Al4%ZrO2, Al6%ZrO2, and Al8%ZrO2 were prepared using the stir casting method. The upsetting process was simulated, and an equation for predicting hardness was developed from the equivalent strain distributions. Artificial neural networks (ANNs), multilinear regression (MLR) along with equations developed from FE analysis were used to train the model for regression analysis, considering the principal stresses, friction factor, anisotropy ratio, effective strain, and hoop strain as input and the Magnitude of hardness as output parameters. Regression analysis reveals that ANN (tri-layer network), XGBoost, and MLR algorithms are the best suitable for the given data sets with a root mean square (R2) greater than 0.95. XGBoost, ANN (narrow), and SVM are linear and are the most recommendable classifier algorithms for the current investigation. Hardness data from ring compression tests were used to validate the results obtained from the trained models with the test results.

Featured Image

Why is it important?

The study described above is important for several reasons: Predictive Capabilities: The ability to predict metal hardness distribution in upsetting tests is crucial for understanding the material behavior under different compositions of ZrO2 in an aluminum matrix. Predictive models developed in the study, particularly using machine learning algorithms and finite element analysis, provide a tool to estimate hardness in scenarios where experimental testing might be challenging, time-consuming, or expensive. Optimization of Material Compositions: The study investigated varying mass fractions of ZrO2 particles in the aluminum matrix. The predictive models can aid in optimizing these compositions to achieve desired hardness levels for specific applications. This optimization can contribute to the development of materials with enhanced mechanical properties, tailored to meet the requirements of specific industries or applications. Process Understanding: Simulating the upsetting process and developing equations based on equivalent strain distributions contribute to a deeper understanding of the material deformation behavior during manufacturing processes. Insights gained from the study can inform process optimization and help researchers and engineers better control and manipulate material properties during fabrication. Machine Learning Applications in Materials Science: The study showcases the application of machine learning algorithms, such as Artificial Neural Networks (ANNs), XGBoost, and Support Vector Machine (SVM), in the field of materials science. Demonstrating the effectiveness of these algorithms in predicting material properties opens up new avenues for the use of machine learning in materials research and development. Validation and Confidence in Models: The validation of predictive models using hardness data from ring compression tests is crucial for establishing confidence in the accuracy of the models. Validated models provide a reliable basis for making predictions and decisions in real-world applications, ensuring that the developed algorithms are robust and applicable. Practical Implications for Industry: The study's findings, particularly the recommendation of specific algorithms based on their performance metrics, have practical implications for industries involved in materials processing and manufacturing. Industries can leverage the recommended algorithms for predicting and controlling hardness, leading to improved product quality and performance.

Perspectives