[Paper Review] Mechanical Property Design of Bio-compatible Mg alloys using Machine-Learning Algorithms

Parham Valipoorsalimi, Yuksel Asli Sari+1 more|arXiv (Cornell University)|May 20, 2023

Magnesium Alloys: Properties and ApplicationsMaterials Science3 citations

TL;DR

This study develops a machine learning model combined with a genetic algorithm to design high-strength, biocompatible magnesium alloys for bone implants. Using a random forest regressor with biocompatible elements only, the model predicted yield strengths with 91% R² accuracy, leading to two synthesized alloys achieving 108 MPa and 113 MPa yield strength—exceeding conventional biocompatible Mg alloys and approaching natural bone's mechanical properties.

ABSTRACT

Magnesium alloys are attractive options for temporary bio-implants because of their biocompatibility, controlled corrosion rate, and similarity to natural bone in terms of stiffness and density. Nevertheless, their low mechanical strength hinders their use as cardiovascular stents and bone substitutes. While it is possible to engineer alloys with the desired mechanical strength, optimizing the mechanical properties of biocompatible magnesium alloys using conventional experimental methods is time-consuming and expensive. Therefore, Artificial Intelligence (AI) can be leveraged to streamline the alloy design process and reduce the required time. In this study, a machine learning model was developed to predict the yield strength (YS) of biocompatible magnesium alloys with an $R^2$ accuracy of 91\%. The predictive model was then validated using the CALPHAD technique and thermodynamics calculations. Next, the predictive model was employed as the fitness function of a genetic algorithm to optimize the alloy composition for high-strength biocompatible magnesium implants. As a result, two alloys were proposed and synthesized, exhibiting YS values of 108 and 113 MPa, respectively. These values were substantially higher than those of conventional magnesium biocompatible alloys and closer to the YS and compressive strength of natural bone. Finally, the synthesized alloys were subjected to microstructure analysis and mechanical property testing to validate and evaluate the performance of the proposed AI-based alloy design approach for creating alloys with specific properties suitable for diverse applications.

Motivation & Objective

• To overcome the low mechanical strength of biocompatible magnesium alloys, which limits their use in load-bearing implants like stents and bone substitutes.
• To reduce the time and cost of experimental alloy development by replacing conventional trial-and-error methods with AI-driven prediction and optimization.
• To design high-strength Mg alloys using only biocompatible elements, ensuring safety for in-vivo applications.
• To validate the AI-predicted compositions through experimental synthesis and mechanical testing.
• To integrate microstructural and thermodynamic analysis to confirm the reliability of the AI-based design approach.

Proposed method

• Trained a random forest (RF) regression model on a dataset of biocompatible Mg alloy compositions and their corresponding yield strength (YS) values, achieving 91% R² accuracy.
• Applied the CALPHAD method and thermodynamic calculations to validate the model’s predictions and ensure phase stability.
• Used the trained RF model as the fitness function in a genetic algorithm (GA) to search for optimal alloy compositions maximizing YS while respecting biocompatibility constraints.
• Set elemental limits to include only biocompatible elements: Ca, Sr, Zn, Mn, and rare earths, excluding toxic or non-biodegradable elements.
• Synthesized two candidate alloys based on GA-predicted compositions and performed microstructural analysis (EDS, SEM) and mechanical testing (hardness, YS) to validate predictions.
• Correlated experimental YS values with Vickers hardness using a 2.4:1 ratio (YS ≈ 2.4 × BHN) to estimate yield strength from hardness measurements.

Experimental results

Research questions

• RQ1 Can machine learning accurately predict the yield strength of biocompatible magnesium alloys using only composition data?
• RQ2 Can a genetic algorithm guided by an ML model identify high-strength Mg alloy compositions that remain biocompatible and thermodynamically stable?
• RQ3 Do the AI-predicted alloy compositions result in experimentally synthesized materials with mechanical properties close to natural bone?
• RQ4 How do microstructural features (e.g., phase distribution, dendrite spacing) correlate with the mechanical performance of the synthesized alloys?
• RQ5 To what extent do predicted YS values align with experimentally measured YS values, and what accounts for discrepancies?

Key findings

• The random forest model achieved a coefficient of determination (R²) of 91% in predicting the yield strength of biocompatible Mg alloys.
• The genetic algorithm identified two optimal alloy compositions: one with 87% Mg, 5% Sr, 8% Zn, and 8% Mn, and another with 96% Mg, 0.5% Ca, 4% Zn, and 4% Mn.
• Experimentally synthesized Alloy 1 exhibited a yield strength of 108 MPa and a Vickers hardness of 47 HV, while Alloy 2 achieved 113 MPa and 52 HV.
• The predicted yield strengths for the experimental alloys were 118 MPa (Alloy 1) and 125 MPa (Alloy 2), showing good agreement with experimental results.
• Microstructural analysis confirmed the presence of Mg(Zn) solid solutions, Mn-rich phases, and intermetallic compounds such as Sr₂Zn₄₃Mg₅₅ and CaMgZn, which contribute to strengthening.
• The hardness of the synthesized alloys, especially Alloy 2 (52 HV), was comparable to that of natural bone and other experimental biodegradable Mg alloys, indicating reduced risk of stress shielding.