Process parameters optimization via machine learning and properties characterization of ALSI10MG-316L multi-materials produced using laser powder bed fusion / Miao Huan

Miao , Huan (2024) Process parameters optimization via machine learning and properties characterization of ALSI10MG-316L multi-materials produced using laser powder bed fusion / Miao Huan. PhD thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (138Kb)
	PDF (Thesis PhD) Restricted to Repository staff only until 31 December 2025. Download (10Mb)

Abstract

AlSi10Mg-316L multi-material parts are expected to be used on spacecraft, thereby reducing the weight of the spacecraft and offering more payload section, since AlSi10Mg provides a lightweight structure. However, achieving metallurgical bonding of the AlSi10Mg and 316L using laser powder bed fusion (LPBF) is extremely challenging due to the large differences in their thermal physical properties. This work aims to produce AlSi10Mg-316L multi-material parts with excellent performance using LPBF. Process parameters optimization for powder-mixed AlSi10Mg-316L multi-materials by employing machine learning method was deeply studied. The electrochemical corrosion behavior of AlSi10Mg-316L multi-materials in NaOH solution was evaluated. Interfacial characteristics of AlSi10Mg-316L multi-materials were determined. The experimental process parameters and properties (density and surface roughness) data were used to train a developed multi-output Gaussian process regression (MO-GPR) model to directly predict the multidimensional output to overcome the limitations of the standard Gaussian process regression (GPR) model. Based on the prediction data, process parameter maps were constructed, and optimum process parameters for different compositions were selected from the maps. Optical microscope (OM), scanning electron microscopy (SEM), transmission electron microscope (TEM), energy dispersive spectroscopy (EDS), electron backscattered diffraction (EBSD) and Vickers microhardness tester were used to investigate microstructure, defects, element diffusion, phase formation, grain orientation and microhardness of AlSi10Mg-316L multi-materials. The electrochemical corrosion tests of AlSi10Mg-316L multi-materials in 5 wt% NaOH solution were performed using an electrochemical workstation. The results revealed that MO-GPR model can accurately predict the properties at any set of process parameters of powder-mixed multi-materials due to the low error ratio for density (1.49%) and surface roughness (9.7%). Laser power, scanning velocity and hatching space had important influence on the density and surface roughness of the parts. Electrochemical corrosion tests show that the corrosion resistance of the LPBFed samples in 5 wt% NaOH solution follows the order: 316L > f=75% > f=50% > f=25% > AlSi10Mg. This indicates that the contributing member 316L significantly improves the corrosion resistance of AlSi10Mg-316L multi-material parts. In addition, using the optimal process parameters, multi-material parts can be produced with a good interface metallurgical bonding without significant defects. The partial Fe- FCC phase in 316L region of multi-materials changed into the Fe-BCC structure, and this shift has also changed the preferred orientation of the grains. Al-Fe icosahedral quasicrystals with five-fold symmetry were found at the boundary of the molten pool, which was caused by an extremely high cooling rate. The microhardness of the Al-Fe interface zone was higher than that of 316L with an average value of 235.57 HV and AlSi10Mg with 124.59 HV, which was caused by the very hard intermetallic compounds Al5Fe2 and AlFe formed at the interface. The metallurgical bonding mechanism of multimaterials was that the dissimilar metals were mixed and in-situ alloyed in the molten pool by the Marangoni convection-induced strong circular flow during LPBF processing. This study provided insights into laser powder bed fusion of multi-materials with dissimilar materials and provided reference for manufacturing functionally graded material parts.

Actions (For repository staff only : Login required)