Designing biocompatible titanium alloys: machine learning approach

Abstract

Titanium and its various alloys have been used for decades as for numerous dental and orthopedic devices. What makes it suitable for these applications is the excellent combination of biocompatibility, corrosion resistance, low modulus of elasticity and specific strenght. However, recent reasrches have linked some of the main alloying elements, aluminium and vanadium, and several other elements besides them, with a very harmful effect on human body. Stress shielding is another possible side effect due to the still insufficiently matched elastic modulus of the alloy and bone. These issues have demanded the exploration for alternative alloys, characterized by non-toxic components and low elastic modulus. The design of titanium alloys involves a variety of tehniques, such as Mo equvivalent method, the electron-to-stom ratio (e/a), d electron based alloy design, experimental tehniques, and cutting-edge machine learning approches. The study leverges the Extra Tree Regression from machine learning to analize the most influential parameters for the elastic modulus, identifying the specific heat and shear of the silicon in alloy as significant factors. Multi-component diagrams were subsequently constructed to guide the development of alloys with a low elastic modulus. Also, employing the development model with the Monte Carlo experimental design method we found optimal compositions for high entropy alloys with a low Young’s modulus. These finding provide a solid soundation for future experimental studies on biocompatible titanium alloys.