Machine learning-based design of biodegradable Mg alloys for load-bearing implants

Abstract

For load-bearing applications, biodegradable Mg alloys require high strength and slow degradation rates to support bone regeneration. This study proposes a design guide of Mg-Zn-Mn-Sr-Ca (ZMJX) alloys for load-bearing Mg implants using machine learning. To this end, it quantitatively investigates the correlation between 4 alloying elements with content of 0–3 wt%, ultimate compressive strength (UCS) and in vitro corrosion rate (CR) in ZMJX alloys. Cascade-forward neural networks predict UCS and CR with high accuracy of over 0.95 for a total of 840 data points. Random forest regression identifies Zn as a major determinant of UCS and CR. Based on this, three chemical compositions are recommended with improved compressive strength and in vitro corrosion resistance by well-verified neural network models. The proposed Mg alloys have UCS of 244–305 MPa and CR of 0.31–0.83 mm/y according to the change of the Zn content. These results can not only provide deep insights into ZMJX alloys, but also recommend a compositional window for load-bearing Mg implants.

Keywords

• Machine learning
• Magnesium alloy
• Biodegradable
• Mechanical properties
• Corrosion