| dc.description.abstract | Laser powder bed fusion (LPBF) is an additive manufacturing (AM) technique to fabricate complex geometries with minimal material waste. Over three hundred printing parameters can be altered to control the final part quality. Some significant factors include the laser power, speed, printing pattern, and part geometry. The parameters must be optimized to suit the type of material used. Accordingly, it is not efficient to use experimental methods as they are costly and time-consuming. This makes finite element analysis (FEA) a viable candidate to identify the optimal printing parameters based on parametric analysis. However, computational cost increases with the necessity for improved resolution in the predicted results (such as capturing stress directionality). This thesis presents a novel multi-scale model based on finite element modelling methods. The term multi-scale model is used because the model allows LPBF stress-temperature prediction at the micron scale (within the laser track) and at the part scales (cm) as additional layers are built. The model also enables simulation in a reasonable time frame for industrial application.
Firstly, a novel hybrid line (HL) heat input model was developed to overcome the step time limitation inherent in the beam-scale model. The HL model was calibrated using mechanical and thermal properties of high gamma-prime Ni-based superalloys and the predicted nodal temperature was compared with results obtained from the exponentially decaying (ED) heat input model. Two sets of design of experiments (DOEs) were used to validate the model both thermally and mechanically. The first DOE was designed to validate the thermal part of the simulation. This includes printed single tracks on a layer of RENÉ 65 powder using six different laser speeds and three different laser powers. The results showed that the HL and ED models are equally accurate in predicting cooling rates and nodal temperatures, essential to simulate the in-process strain and stress. In the second DOE, twelve components were printed using three laser powers and four different printing patterns and residual stress was measured using X-ray diffraction (XRD). The predicted results were in good agreement with the experimental results and the average stress error was below 4% of the yield stress at room temperature. The HL model accurately captures the stress directionality and significantly decreases the computational time. The HL model can become over 1,500 and 30 times faster than the ED model for the thermal and mechanical models, respectively.
Secondly, a multi-scale model was developed using a combination of the HL model and lumping approach. This new lumped HL (LHL) model allows laser track-powder layer lumping, which can be adapted to optimize computational efficiency for the desired accuracy. This enables simulation of the stress directionality in LPBF at the part-scale level. The LHL model was calibrated using the mechanical and thermal properties of high gamma-prime Ni-based superalloys. The model was validated by comparing the final part distortion (measured with a 3D scanner) to the simulation results of LPBF thin-wall components. Two different printing patterns were used for the validation of the model and the final experimental errors were found to be below 10% of the local deformation.
Finally, the two models were used to perform parametric studies on LPBF high gamma prime Ni-based superalloys. The HL model was first used to evaluate the effect of scanning pattern and laser power. The results show that vector length and scanning strategies have stronger effect on the residual stresses. Shorter vector length creates more compressive stresses while longer vector length generates more tensile stresses along the longitudinal directions. The laser power and interlayer rotations are also shown to be beneficial for stress reduction and homogenization. Finally, the LHL model was used to study the effect of printing patterns and part geometry on the limiting build height (LBH). This is when part abortion limits the final part height during LPBF. For this study, a DOE of 34 components was created with lengths between 20 mm and 60 mm (with an increment of 10 mm), 8 different printing patterns (with and without scan rotation), and a fixed wall thickness of 0.5 mm. Buckling mechanisms were identified as the major cause of in-process part failure. The best combination was found to be interlayer vector rotations that maximize the vector lengths. The model showed longer vector length promotes reduction of compressive residual stresses and increase in LBH. The part length has a small effect on the LBH but leads to a change in the buckling mechanism. | en |
