Adaptive toolpath generation for distortion reduction in laser powder bed fusion process

Laser powder bed fusion (LPBF) is one of the most commonly used metal additive manufacturing technologies due to its advantages in fabricating complex structures. The LPBF process utilizes laser scanning to melt and consolidate the powders, in which the laser tool paths are often referred to as the scan pattern. However, traditional scan patterns could cause large distortion due to residual stresses induced during the fabrication process. Although recently some studies have attempted to develop optimization methods instead of using the empirical trial-and-error approach, they are seldomly implemented for fabricating 3D geometries due to various limitations. In this paper, we propose novel adaptive toolpath generation algorithms based on minimizing thermal gradients so that can reduce the part distortion. The objective function of temperature fields is deter-mined by numerical simulation both considering time and distance. Two constraints including collision-free and smoothing constraints are also incorporated during the toolpath patterns generation process. In addition, to improve the efficiency of adaptive toolpath generation algorithms for large or complex structures, island-based strategy is developed and combined with the adaptive toolpath generation algorithms. Furthermore, experiments are conducted to validate the proposed algorithms. The optimal parameters consisting of smoothing coefficient and searching radius are determined by toolpath patterns analysis and characterization of the fabricated cube samples. The fabricated cube samples with optimal adaptive toolpath patterns are compared with traditional zigzag patterns as well and results show that they are at the same level of characterization in surface roughness, microhardness and relative density values. Six different groups of cantilever beams are fabricated and the dis-tortions are measured to compare the effects of toolpaths on the distortion. Results show that the distortion of our proposed algorithms could reduce about 46 % compared with that of the zigzag patterns and reduce about 36 % compared with that of the traditional chessboard-based patterns. Therefore, our algorithms proposed in this paper can significantly reduce the distortion during the LPBF process. In addition, we extend our proposed al-gorithms from 2D to 3D cases to realize fabrication of complex 3D geometries. Finally, we fabricated a "molar tooth" model with freeform surfaces and overhang feature to demonstrate the effectiveness and generality of the proposed algorithms.