Reconstructed synchrotron X-ray CT volume of five-phase composite

The reconstructed synchrotron X-ray CT volume of five-phase composite and the result of Deep learning-based segmentation of all reinforcement phases.

Source: BAM, division Micro NDT and DEQM PUC-Rio, Brasil

Multi-phase metal matrix composites (MMC) are advanced materials with excellent mechanical and thermal properties. Nowadays different types of MMC find applications in both automotive and aerospace industries. However, being complex materials with complex microstructure, they still require deep investigations to understand their micromechanical behaviour, the effect of high temperature exposure, and the damage mechanisms under external load. In this study we investigate a hybrid MMC obtained by reinforcing the AlSi12CuMgNi alloy with 7 vol.% of Al2O3 short fibers and 15 vol.% of SiC particles. This material consists of five different phases (Al solid solution matrix, eutectic Si, intermetallics, Al2O3 fibers and SiC particles) and, therefore, has very complex micromechanical response under load condition.

It is well known that the micromechanical behavior and mechanical properties of such composites strongly depend on the orientation of the fibers, the spatial distribution of the particles, the individual volume fractions of all reinforcement phases, as well as on their morphology and interconnectivity. The most suitable tool to provide this kind of information is X-ray computed tomography (CT). However, image segmentation and successive quantitative analysis of the CT data, especially in the case of multiphase materials, remains a highly challenging task.

In this work we showed the successful application of the deep learning approach to a segmentation problem, which could not be solved by any conventional method (apart from lengthy manual approach). We quantitatively assessed the segmentation quality and proved, that even with small amount of training data the neural network can segment complex CT data with high accuracy. The achieved accuracy is sufficient to estimate the volumetric characteristics of every individual phase. This will be used as an input to the FE and analytical models for accurate prediction of mechanical properties and micromechanical characteristics of the composite under different load scenario. Generally, the results obtained in this work open a host of possibilities in quantitative 3D microstructural characterization of complex materials by means of CT.

Advanced Deep Learning-Based 3D Microstructural Characterization of Multiphase Metal Matrix Composites
Sergei Evsevleev, S. Paciornik, Giovanni Bruno, published in Advanced Engineering Materials, Vol. 22, Issue 4, pages 1901197
BAM, division Micro Non-Destructive Testing