Search for: All records

In this work, a tung oil-based thermosetting resin was synthesized via free radical polymerization and reinforced with thirteen different types of sand. The viability of this process inspired the adaptation of the resin for its use as a binder material in binder jetting, an additive manufacturing process. Firstly, it was shown that the resin could have its initial viscosity (~0.33 cP) increased upon heating to attain values compatible to existing printing systems. The curing kinetics of the resin was assessed via dielectric analysis (DEA), combining the utilization of heat and ultraviolet (UV) light, showing that a resin with a viscosity of 10 cP can be fully cured after 250 min at 90 ◦C, or 300 min at 75 ◦C, both under a 365 nm light exposure. Preliminary binder-jet tests successfully provided a solid object, which was post-cured, resulting in a hard specimen. The results presented herein suggest that the tung oil-based resin in question is a suitable bio-based binder for binder-jet 3D-printing applications. The novelty of the work reported lies in the conversion of an already established and effective bio-based thermosetting resin into a versatile photocurable binder that can be irrestrictively used with unsorted sands of different composition, making this technology broadly applicable to different isolated regions, using local resources available. The technology presented herein is potentially transformative and impactful, as binder jetting is typically associated to extremely well-sorted particles. 

In binder jet additive manufacturing (BJAM), uniformity and density of the powder layer impact green part quality. This study investigates the printability of unrefined sand using counter-roller spreading. Altair EDEM, a high-performance software powered by the Discrete Element Method (DEM), was used to simulate the BJAM process to evaluate powder bed homogeneity and density under various operating conditions, including roller rotational speed, traverse speed, powder layer thickness, and roller diameter. Utilizing high-performance computing (HPC) and graphics processing unit (GPU) clusters, time-efficient, and more realistic, simulations were performed simulating 300,000 grains. Detailed DEM simulations were executed by reconstructing representative particle shapes using two-dimensional images obtained using particle characterization equipment. The results highlight roller velocity and powder layer thickness as key determinants of sand spreadability. Optimal powder bed density (PBD) was achieved at a roller velocity of 20 mm/s with minimal deviation. A layer thickness exceeding 200 micrometers was found to prevent jamming and void formation, while percolation led to size segregation. The findings indicate that producing uniform and dense layers of unrefined sand is feasible but may incur trade-offs in print resolution and increased printing times. This work contributes to the advancement of sustainable and/or remote BJAM technologies, ensuring progress in both environmental sustainability and accessibility. 

Recent results suggest that it is possible to grasp a variety of singu- lated objects with high precision using Convolutional Neural Networks (CNNs) trained on synthetic data. This paper considers the task of bin picking, where multiple objects are randomly arranged in a heap and the objective is to sequen- tially grasp and transport each into a packing box. We model bin picking with a discrete-time Partially Observable Markov Decision Process that specifies states of the heap, point cloud observations, and rewards. We collect synthetic demon- strations of bin picking from an algorithmic supervisor uses full state information to optimize for the most robust collision-free grasp in a forward simulator based on pybullet to model dynamic object-object interactions and robust wrench space analysis from the Dexterity Network (Dex-Net) to model quasi-static contact be- tween the gripper and object. We learn a policy by fine-tuning a Grasp Quality CNN on Dex-Net 2.1 to classify the supervisor’s actions from a dataset of 10,000 rollouts of the supervisor in the simulator with noise injection. In 2,192 physical trials of bin picking with an ABB YuMi on a dataset of 50 novel objects, we find that the resulting policies can achieve 94% success rate and 96% average preci- sion (very few false positives) on heaps of 5-10 objects and can clear heaps of 10 objects in under three minutes. Datasets, experiments, and supplemental material are available at http://berkeleyautomation.github.io/dex-net.