Purpose
Laser powder bed fusion (LPBF) provides the means to produce unique components with almost no restriction on geometry in an extremely short time. However, the high-temperature gradient and high cooling rate produced during the fabrication process result in residual stress, which may prompt part warpage, cracks or even baseplate separation. Accordingly, an appropriate selection of the LPBF processing parameters is essential to ensure the quality of the built part. This study, thus, aims to develop an integrated simulation framework consisting of a single-track heat transfer model and a modified inherent shrinkage method model for predicting the curvature of an Inconel 718 cantilever beam produced using the LPBF process.
Design/methodology/approach
The simulation results for the curvature of the cantilever beam are calibrated via a comparison with the experimental observations. It is shown that the calibration factor required to drive the simulation results toward the experimental measurements has the same value for all settings of the laser power and scanning speed. Representative combinations of the laser power and scanning speed are, thus, chosen using the circle packing design method and supplied as inputs to the validated simulation framework to predict the corresponding cantilever beam curvature and density. The simulation results are then used to train artificial neural network models to predict the curvature and solid cooling rate of the cantilever beam for any combination of the laser power and scanning speed within the input design space. The resulting processing maps are screened in accordance with three quality criteria, namely, the part density, the radius of curvature and the solid cooling rate, to determine the optimal processing parameters for the LPBF process.
Findings
It is shown that the parameters lying within the optimal region of the processing map reduce the curvature of the cantilever beam by 17.9% and improve the density by as much as 99.97%.
Originality/value
The present study proposes a computational framework, which could find the parameters that not only yield the lowest distortion but also produce fully dense components in the LPBF process.