Improving mechanical properties of additive manufacturing products using novel infill and slicing methods

Abstract

In recent years, additive manufacturing (AM) has drawn significant attention and interest from both academia and industry due to its remarkable advantages. However, one critical challenge in AM is tunability of mechanical properties for AM parts. Therefore, this dissertation focuses on the development of novel infill and slicing methods. G-Lattices. Lattice structures embedding in a solid model play a crucial role in additive manufacturing (AM) for reducing manufacturing cost/time and improving (mechanical, acoustic and etc.) properties of the printed parts. Manually generated lattice structures consist of multiple struts, and their structural properties differ according to the strut shapes and topology. However, there are limited type/numbers of strut-based lattice structures, and therefore, this paper introduces novel lattice structures that are called G-Lattices and a method for generative synthesis of G-Lattices. Given AM, user, and geometrical constraints, G-Lattices can automatically be generated via a particle tracing algorithm, which places/moves particles in a lattice unit cell. Sweeping a sphere along the particle trajectories forms G-Lattices. Two alternative tracing methods are proposed in this work; one using straight and the other via curved struts. Numerous G-Lattices can be created using these techniques in a short time. Users can adjust G-Lattice density in a unit cell, strut thickness, strut shapes (i.e., straight or curvy), and angle between struts in order to control the physical/mechanical properties of G-Lattices (to some extent). As proof of concept, several G-Lattices are manufactured through an AM machine. Additionally, the proposed G-Lattice synthesis method is customized for the models under vertical loading. The G-Lattices obtained in this way are validated through finite element method experiments and have greater strength over volume ratios compared to conventional lattice structures. An extension of G-Lattices (i.e., reinforced G-Lattices) demonstrating better mechanical performance under inclined (compression) loading conditions are also introduced. For different inclined loads, separate reinforced G-Lattices are first optimized, and a G-Lattice library is formed. For a part under loading, displacement vectors in each unit cell (cubic domains within the inner region of the part) are then extracted. Based on these vectors, (pre-optimized) reinforced G-Lattices are selected from the G-Lattice library and utilized (as infills) in the unit cells. This process is called G-Puzzling. As a proof of concept, parts under three different inclined loading conditions are infilled using reinforced G-Lattices and investigated based on stiffness-over-volume ratios. According to these experiments, the resulting parts, onIn recent years, additive manufacturing (AM) has drawn significant attention and interest from both academia and industry due to its remarkable advantages. However, one critical challenge in AM is tunability of mechanical properties for AM parts. Therefore, this dissertation focuses on the development of novel infill and slicing methods. G-Lattices. Lattice structures embedding in a solid model play a crucial role in additive manufacturing (AM) for reducing manufacturing cost/time and improving (mechanical, acoustic and etc.) properties of the printed parts. Manually generated lattice structures consist of multiple struts, and their structural properties differ according to the strut shapes and topology. However, there are limited type/numbers of strut-based lattice structures, and therefore, this paper introduces novel lattice structures that are called G-Lattices and a method for generative synthesis of G-Lattices. Given AM, user, and geometrical constraints, G-Lattices can automatically be generated via a particle tracing algorithm, which places/moves particles in a lattice unit cell. Sweeping a sphere along the particle trajectories forms G-Lattices. Two alternative tracing methods are proposed in this work; one using straight and the other via curved struts. Numerous G-Lattices can be created using these techniques in a short time. Users can adjust G-Lattice density in a unit cell, strut thickness, strut shapes (i.e., straight or curvy), and angle between struts in order to control the physical/mechanical properties of G-Lattices (to some extent). As proof of concept, several G-Lattices are manufactured through an AM machine. Additionally, the proposed G-Lattice synthesis method is customized for the models under vertical loading. The G-Lattices obtained in this way are validated through finite element method experiments and have greater strength over volume ratios compared to conventional lattice structures. An extension of G-Lattices (i.e., reinforced G-Lattices) demonstrating better mechanical performance under inclined (compression) loading conditions are also introduced. For different inclined loads, separate reinforced G-Lattices are first optimized, and a G-Lattice library is formed. For a part under loading, displacement vectors in each unit cell (cubic domains within the inner region of the part) are then extracted. Based on these vectors, (pre-optimized) reinforced G-Lattices are selected from the G-Lattice library and utilized (as infills) in the unit cells. This process is called G-Puzzling. As a proof of concept, parts under three different inclined loading conditions are infilled using reinforced G-Lattices and investigated based on stiffness-over-volume ratios. According to these experiments, the resulting parts, on average, exhibit more than %30 better mechanical performance compared to FBCCZ (a conventional lattice structure). A machine learning procedure is further proposed in this work to predict the mechan- ical properties of G-Lattices under specific loading conditions. 20000 G-Lattices are first generated using a uniform sampling approach. Strength-over-weight ratios for the G-Lattices are obtained using finite element analysis. Furthermore, voxelized data of G-Lattices are exploited as feature vectors in the machine learning step. A linear regression model is then computed using these G-Lattices. However, the model is inaccurate, particularly for the G-Lattices with ratios greater than five. Therefore, 14000 more G-Lattices are further sampled to increase the number of G-Lattices in this range. For each of the two clusters (G-Lattices with ratios greater/equal to or less than five), a separate linear regression models are calculated. According to the experimental results, approximately 70% of G-Lattices have errors less than/equal to 5% prediction errors, and the mean absolute (relative) percentage error for 40000 G-Lattices is 6.4% Parametrized helical printing. AM commonly utilizes slicing techniques to create layers of a model, in which material is deposited layer by layer. However, the slicing method directly affects the mechanical properties of the printed parts. This paper introduces a new AM technique (named Helical5AM), which employs print paths having helical geometry for five-axis AM. Given an object to be printed, a base (supporting helical print paths) with a center curve and helix parameters (such as lead angle and turn direction), a complete volumetric coverage using helical print paths is obtained. Collision-free tool orientation is then generated using a probabilistic roadmap algorithm for depositing material along the helical print paths by avoiding tool interference with obstacles. As a proof of concept, print paths (of models) with orientation information obtained using the proposed algorithms are simulated using a five-axis AM simulation software, and the material deposition process in Helical5AM is demonstrated using a five-axis AM machine. Furthermore, compression tests are performed on the printed parts for evaluating the effects of helix lead angles of the helical print paths on the mechanical properties of the printed parts. It has been confirmed that the mechanical behavior of a printed part is predictable and tunable according to the helix lead angles of the print paths. Helical5AM can potentially empower engineers to obtain AM parts with desirable mechanical properties. Optimization of helix lead angles of helical print-paths for a part under a certain loading conditions is also investigated. The part is first decomposed into layers, which are covered by helical print paths. The layers are meshed conformally using hexahedral elements. According to lead helix angles, material orientations are given to these elements. Finite element analysis (FEA) is then carried out for investigating the mechanical properties of the part. An optimization approach is coupled with FEA to optimize lead helix angles for the helical print paths. The effectiveness of the proposed approach is verified via multiple experiments with various loading conditions. The results indicate that the optimized parts demonstrate better mechanical properties.