MechProNet: Machine Learning Prediction of Mechanical Properties in Metal Additive Manufacturing

TL;DR

This study introduces a comprehensive framework for benchmarking ML models for predicting mechanical properties in metal additive manufacturing and explores the Explainable AI method, specifically SHAP analysis, to elucidate and interpret the predicted values of ML models for mechanical properties.

Abstract

Predicting mechanical properties in metal additive manufacturing (MAM) is essential for ensuring the performance and reliability of printed parts, as well as their suitability for specific applications. However, conducting experiments to estimate mechanical properties in MAM processes can be laborious and expensive, and they are often limited to specific materials and processes. Machine learning (ML) methods offer a more flexible and cost-effective approach to predicting mechanical properties based on processing parameters and material properties. In this study, we introduce a comprehensive framework for benchmarking ML models for predicting mechanical properties. We compiled an extensive experimental dataset from over 90 MAM articles and data sheets from a diverse range of sources, encompassing 140 different MAM data sheets. This dataset includes information on MAM processing conditions, machines, materials, and resulting mechanical properties such as yield strength, ultimate tensile strength, elastic modulus, elongation, hardness, and surface roughness. Our framework incorporates physics-aware featurization specific to MAM, adjustable ML models, and tailored evaluation metrics to construct a comprehensive learning framework for predicting mechanical properties. Additionally, we explore the Explainable AI method, specifically SHAP analysis, to elucidate and interpret the predicted values of ML models for mechanical properties. Furthermore, data-driven explicit models were developed to estimate mechanical properties based on processing parameters and material properties, offering enhanced interpretability compared to conventional ML models.

Figures (13)

• Figure 1: Overview of the workflow utilized in our benchmark study, including data acquisition, feature extraction, training of machine learning models, prediction tasks, and model analysis.
• Figure 2: a) Distribution of various MAM sub-processes within our benchmark dataset, b) Distribution of orientation within our benchmark dataset, c) Distribution of different post-processing methods and conditions within our dataset, d) Distribution of surface post-processing methods within our dataset, e) Occurrence of materials studied in our benchmark dataset, and f) Frequency of investigated additive manufacturing machines in our benchmark dataset.
• Figure 3: The dataset processing parameters histograms and their distribution. a) Beam power histogram and occurrence, b) Scanning speed histogram and occurrence, c) Layer thickness histogram and occurrence, d) Beam diameter histogram and occurrence.
• Figure 4: Evaluation of benchmark performances for yield strength, ultimate tensile strength, Elastic Modulus, and Elongation at break: Various machine learning models including 'Random Forest', 'Gaussian Process Regressor', 'Support Vector Regressor', 'Ridge Linear Regressor', 'Lasso Linear Regressor', 'Gradient Boosting', and 'Neural Network' are assessed based on their $R^2$ score (left column) and Mean Absolute Error (MAE) (right column). a) $R^2$ accuracy and MAE results for predicting yield strength, b)$R^2$ accuracy and MAE results for predicting ultimate tensile strength, c) $R^2$ accuracy and MAE results for predicting Elastic Modulus, d) $R^2$ accuracy and MAE results for predicting Elongation at break. (Note: Higher $R^2$ scores and lower MAE values indicate superior performance.)
• Figure 5: Evaluation of benchmark performances for the Vickers hardness, the Rockwell hardness, and $R_z$ Roughness; Various machine learning models including 'Random Forest', 'Gaussian Process Regressor','Support Vector Regressor', 'Ridge Linear Regressor', 'Lasso Linear Regressor', Gradient Boosting', 'Neural Network' are assessed based on their $R^2$ score (left column) and Mean Absolute Error (MAE) (right column). a) $R^2$ accuracy and MAE results for predicting Vickers hardness, b) $R^2$ accuracy and MAE results for predicting the Rockwell hardness, c) $R^2$ accuracy and MAE results for predicting $R_z$ Roughness. (Note: Higher $R^2$ scores and lower MAE values indicate superior performance.)