Microstructural characterization of Ti/Cu/Ti diffusion bonded system through a micromechanical data-driven neural network approach

Abstract

In this work, a micromechanical data-driven neural network approach, named as adjacent gradient neural network (AGNN), was developed to characterize the microstructural features of Ti/Cu/Ti Diffusion Bonded joints with a wide range of interlayer thicknesses. The AGNN model facilitates the evaluation of microstructure on the basis of Young’s modulus and hardness distributions in the joint zone. The results indicated that the AGNN model accurately classified the microstructural constituents including the stabilized eutectoid structure and the intermetallic compounds; however, there existed some deviations at high-hardness regions, which was originated from the low penetration depth in the indenting process. The predictive outcomes also quantitatively revealed that the increase in the interlayer thickness was accompanied with the generation of intermetallic compounds; however, the proportion of eutectoid structure decreased in the joint zone. In summary, the outcomes of this work indicate that it is feasible to predict the microstructural features of a Ti/Cu/Ti diffusion bonded joint with a preferred interlayer thickness and establish a meaningful correlation between the micromechanical properties and the morphological characteristics.