Modeling of the Sintered Density in Cu-Al Alloy Using Machine Learning Approaches

Area of Science:

• Materials Science
• Mechanical Engineering
• Computational Modeling

Background:

• Sintered density in Copper-Aluminum (Cu-Al) alloys is vital for mechanical properties.
• Experimental determination of sintered density is resource-intensive.
• Accurate prediction models are needed to optimize powder metallurgy processes.

Purpose of the Study:

• To develop and compare advanced machine learning models for predicting Cu-Al alloy powder densification.
• To identify the most accurate predictive model for sintered density.
• To reduce the need for extensive experimental testing.

Main Methods:

• Employed adaptive boosting decision tree, support vector regression, k-nearest neighbors, extreme gradient boosting, and four multilayer perceptron (MLP) models.
• Utilized resilient backpropagation, Levenberg-Marquardt (LM), scaled conjugate gradient, and Bayesian regularization for MLP tuning.
• Input parameters included yield strength, Young's modulus, phase transformation volume variation, hardness, liquid/solid phase properties, sintering parameters, and particle characteristics.

Main Results:

• All developed models outperformed existing approaches in predicting powder densification.
• The multilayer perceptron model optimized with Levenberg-Marquardt (MLP-LM) demonstrated the highest precision.
• MLP-LM achieved an average absolute percent relative error (AAPRE) of 1.292% and a correlation coefficient (R) of 0.989.
• Outlier detection using the leverage technique identified data points outside the MLP-LM model's applicability domain.

Conclusions:

• Advanced machine learning models, particularly MLP-LM, offer a precise and valid alternative to experimental methods for predicting Cu-Al alloy sintered density.
• The developed models can significantly aid in optimizing powder metallurgy processes for Cu-Al alloys.
• Further refinement and validation, including outlier analysis, are important for robust model application.