Machine learning-based microstructure prediction during laser sintering of alumina

Predicting material’s microstructure under new processing conditions is essential in advanced manufacturing and materials science. This is because the material’s microstructure hugely influences the material’s properties. We demonstrate an elegant machine learning algorithm that faithfully predicts the microstructure under new conditions, without the need of knowing the governing laws. We name this algorithm, RCWGAN-GP, which is regression-based conditional generative adversarial networks with Wasserstein loss function and gradient penalty. This algorithm was trained with experimental SEM micrographs from laser-sintered alumina under various laser powers. The RCWGAN-GP realistically regenerates the SEM micrographs under the trained laser powers. Impressively, it also faithfully predicts the alumina’s microstructure under unexplored laser powers. The predicted microstructure features, including the morphology of the sintered particles and the pores, match the experimental SEM micrographs very well. We further quantitatively examined the prediction accuracy of the RCWGAN-GP. We trained the algorithm with computer-created micrograph datasets of secondary-phase growth governed by the well-known Johnson–Mehl–Avrami (JMA) equation. The RCWGAN-GP accurately regenerates the micrographs at the trained time series, in terms of the grains’ shapes, sizes, and spatial distributions. More importantly, the predicted secondary phase fraction accurately follows the JMA curve.

Machine Learning-based Microstructure Prediction During Laser Sintering of Alumina: A Conditional Generative Approach

Predicting material’s microstructure under new processing conditions is essential in advanced manufacturing and materials science. This is because the material’s microstructure hugely influences the material’s properties. We demonstrate an elegant machine learning (ML) algorithm that faithfully predicts the microstructure under new conditions, without the need of knowing the governing laws. We name this algorithm, RCWGAN-GP, which is regression-based conditional generative adversarial networks (GANs) with Wasserstein loss function and gradient penalty. This algorithm was trained with experimental SEM micrographs from laser-sintered alumina under various laser powers. The RCWGAN-GP realistically regenerates the SEM micrographs under the trained laser powers. Impressively, it also faithfully predicts the alumina’s microstructure under unexplored laser powers. The predicted microstructure features, including the shapes, sizes, and spatial configuration of the grains and the pores, match the experimental SEM micrographs very well. We further examine the statistical features of the micrograph, which were predicted by RCWCAN-GP. We trained the algorithm with computer-created micrograph datasets of secondary-phase growth governed by the well-known Johnson-Mehl-Avrami (JMA) equation. The RCWGAN-GP accurately regenerates the micrographs at the trained time series, in terms of the grains’ shapes, sizes, and spatial distributions. More importantly, the predicted secondary phase fraction accurately follows the JMA curve.

Weld Metal Microstructure Prediction in Laser Beam Welding of Austenitic Stainless Steel

In the present work an approach to weld metal microstructure prediction is proposed, based on an analytical method that allows the evaluation of the thermal fields generated during the laser beam travel on thick plates. Reference is made to AISI 304L austenitic steel as a base material, with the aim to predict the molten zone microstructure and verify the best condition to avoid hot cracking formation, which is a typical issue in austenitic steel welding. The “keyhole” full penetration welding mode, characteristic of high-power laser beam, was simulated considering the phenomenological laws of conduction by the superimposition of a line thermal source along the whole thickness and two point sources located, respectively, on the surface and at the position of the beam focus inside the joint. This model was fitted on the basis of the fusion zone profile, which was experimentally detected on a weld seam obtained by means of a CO2 laser beam, in a single pass on two squared edged AISI 304L plates, that were butt-positioned. Then the model was applied to evaluate the thermal fields and cooling rates, the fusion zone composition and the solidification mode.

Non-isothermal Phase-Field Modeling of Heat–Melt–Microstructure-Coupled Processes During Powder Bed Fusion

Modeling and simulation of powder bed fusion (PBF) remain a great challenge due to the sophisticated and interactive nature of underlying physics. A unified scenario considering interactions among the heat transfer, melt flow dynamics and microstructure evolution (noted as “heat–melt–microstructure-coupled processes”) is therefore essential for a thermodynamically consistent description and thus reliable microstructure prediction. In contrast to the state of the art, where either individual aspects are considered or the thermal history is taken as input from separate numerical scheme, we propose in this work a unified non-isothermal phase-field model for the heat–melt–microstructure-coupled processes during PBF. Simulations on a stainless steel 316L powder bed demonstrate that the model can reproduce well-observed features, but also help to discover new in-process phenomena and reveal the mechanism of the defect formation. Based on massive simulation results, we also present the densification map with respect to beam power and scan speed, and have classified the regions of the parameter combination by the distinct resultant morphology.

The Average Grain Size and Grain Aspect Ratio in Metal Laser Powder Bed Fusion: Modeling and Experiment

The additive manufacturing (AM) process induces high uncertainty in the mechanical properties of 3D-printed parts, which represents one of the main barriers for a wider AM processes adoption. To address this problem, a new time-efficient microstructure prediction algorithm was proposed in this study for the laser powder bed fusion (LPBF) process. Based on a combination of the melt pool modeling and the design of experiment approaches, this algorithm was used to predict the microstructure (grain size/aspect ratio) of materials processed by an EOS M280 LPBF system, including Iron and IN625 alloys. This approach was successfully validated using experimental and literature data, thus demonstrating its potential efficiency for the optimization of different LPBF powders and systems.