Effect of the surface condition, defects, residual stresses and microstructure on the fatigue strength of 316L stainless steel additively manufactured by Laser Powder Bed Fusion

Overview & Synopsis

In recent years, the Laser Powder Bed Fusion (L-PBF) additive manufacturing process has attracted much academic and industrial interest. The resulting metal parts exhibit a hierarchical as-built microstructure driven by process parameters and can have both a complex shape and a high mechanical resistance. Additionally, parts can be printed close to the end-use shape, that are net-shape parts. The L-PBF process, however, generates core and surface defects, a rough surface, and tensile residual stresses that adversely affect fatigue life of as-built net-shape parts, thus compromising their industrial viability.The literature is not yet able to clearly explain the coupling of these surface and sub-surface parameters on the fatigue strength, hence the present study aims to decorrelate such effects for a L-PBF 316L stainless steel by conducting a comprehensive experimental campaign.For this purpose, uniaxial fatigue tests are carried out on several batches. Various microstructural conditions (as-built, 900°C-heat-treated) and surface conditions (net-shape, net-shape with artificial defects, polished, polished with artificial defects) are considered. Net-shape specimens are printed with a contour strategy involving a laser pass around the main pattern. The polished condition implies machining and polishing post-treatments. Once tested, critical defects on each specimen are identified. Their influence (i.e. nature, size), as well as the influence of residual stresses and the microstructure under the surface are explored in detail.The microstructure of net-shape specimens is dual, combining a typical core with grain of various sizes and a sub-surface with refined grains. Many pores, from spherical gas pores to lacks of fusion, are found at the contour-core interface, at less than 100 microns from the surface. The rough surface condition presents spatters and valleys. High tensile residual stresses are found under the surface of as-built net-shape parts. The chosen heat-treatment successfully releases such internal stresses but leaves the L-PBF microstructure almost intact.Fatigue tests on net-shape parts, as-built or heat-treated, reveal similar surface and sub-surface killer defects: most often spatters and valleys, sometimes spherical pores and process-induced cracks. Only a few specimens failed at a lack of fusion, the latter being the main killer defect of as-built and heat-treated polished specimens. Such characteristics lead to as-built net-shape parts having a much lower fatigue strength than all of the other batches, including heat-treated net-shape specimens that are closer to polished specimens. Residual stresses are therefore the main parameter governing the fatigue behaviour of as-built net-shape parts. Then, defects bigger than 200 microns, process-induced or artificial, are the second driving parameter, independently of the surface condition (i.e. rough or smooth). Below such sizes, the fatigue tests on heat-treated partially polished specimens (contour microstructure but smooth surface) give similar results as the heat-treated net-shape version, showing that the sub-surface microstructure is the third governing parameter in the conditions of the study. Plotting the defect size relatively to the grain size under the surface as a function of the fatigue limit corrected with the Goodman relation (residual stresses as mean stresses) on a Kitagawa-Takahashi diagram allows to align the data of all batches despite the various microstructural and surface conditions. Lastly, the specific competition between the rough surface and small defects under the surface is studied between net-shape specimens, printed with one or four contours, using high resolution Computed Tomography (CT). Finite element-based numerical simulations from the CT scans confirm that both parameters act as a synergy and contribute to the remaining dispersion between batches after considering the other parameters.