Artificial neural network based optimization of prerequisite properties for the design of biocompatible titanium alloys

The objective of the current study was to design titanium alloys by optimizing tensile strength, yield strength, modulus of elasticity and biocompatibility using the artificial neural network (ANN) and recursive partitioning (RP) models. The database of 308 titanium alloys served as the basis of the model with alloy composition and processing parameters as inputs and each property to be optimized as the target. Levenberg-Marquardt back propagation was used to train the ANN model. RP model was used to elucidate different thresholds of alloying elements that are decisive of biocompatibility. The sensitivity analysis of ANN model revealed that Ta and Nb when used at very high concentrations (30-35%), their tensile and yield strength are higher. With the increase in the concentration of Ta and Nb, the modulus of elasticity was shown to decrease and at high concentrations, modulus of elasticity reaches to that of cortical bone. The RP model showed that these two alloys are safe. Hence, we have generated ANN simulations of Ti-xNb-yTa ternary system to ascertain the optimal combination of Ta and Nb that can provide higher tensile and yield strengths and lower modulus of elasticity. These simulations were compared with already developed alloys with these combinations. To conclude, ANN-based optimization of prerequisite properties facilitated timely and cost-effective solutions in the design of biocompatible titanium alloys.