Abstract
In mechanical design mass properties, such as center of gravity and total mass, are an important consideration. As such the distribution of mass, i.e., the effective density distribution of a part, is a key consideration. When producing parts with additive manufacturing techniques like Fused Deposition Modeling, parts are typically hollowed to reduce manufacture time and replaced with a uniform infill structure. This process will affect the mass properties and provides only adjustment to the total mass. However, manipulating the center of gravity of a part can beneficial as it can change the total mass, moment of inertia, and vibration response of a part. Methods have been proposed in the past like the works of Chiu et. al., Yamanaka et. al., and Grigolato et. al. with each seeking to overcome certain limitations. In this paper, a method is proposed for the manipulation of the center of gravity of an additively manufactured part. The method first creates a density distribution with the desired mass properties then a novel process-aware continuous variable density infill algorithm is utilized to create the infill structure. With this algorithm it is ensured that a manufacturable functionally graded infill structure can be generated. The algorithm proposed for realizing this density distribution achieves this in two steps: variation of extrusion line spacing and width of the standard rectilinear infill pattern. This paper focuses on a 2D case, as implementing variation in the z axis was shown to be trivial. The methodology has been implemented within MATLAB to first generate the density distribution, then apply the continuous variable density infill generation algorithm, and finally generate the g-code. All sample prints were made with an extrusion-based, fused deposition modeling (FDM) printer. A preliminary test showed inconsistent behavior from the variable extrusion width. As such, each test case was printed with and without the variable extrusion width to further evaluate. While relatively high, the total error from input density gradient to print for the samples with only VLS showed very good consistency, both being ∼10%. This is favorable as a consistent error can be corrected for with pre/post processing. The samples with variable extrusion width did improve the results, however the degree of the improvement was not consistent, varying from ∼3% to ∼8%. This behavior was predicted as the range of extrusion width variation is rather limited at ±50%. The feasibility of the proposed method has been demonstrated and could be implemented for 3D arbitrary shapes in future work.