skip to main content


Title: Parts beget parts: Bootstrapping hierarchical object representations through visual statistical learning
Award ID(s):
1655300
NSF-PAR ID:
10281092
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Cognition
Volume:
209
Issue:
C
ISSN:
0010-0277
Page Range / eLocation ID:
104515
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract Laser powder bed fusion (L-PBF) additive manufacturing (AM) is an effective method of fabricating nickel–titanium (NiTi) shape memory alloys (SMAs) with complex geometries, unique functional properties, and tailored material compositions. However, with the increase of Ni content in NiTi powder feedstock, the ability to produce high-quality parts is notably reduced due to the emergence of macroscopic defects such as warpage, elevated edge/corner, delamination, and excessive surface roughness. This study explores the printability of a nickel-rich NiTi powder, where printability refers to the ability to fabricate macro-defect-free parts. Specifically, single track experiments were first conducted to select key processing parameter settings for cubic specimen fabrication. Machine learning classification techniques were implemented to predict the printable space. The reliability of the predicted printable space was verified by further cubic specimens fabrication, and the relationship between processing parameters and potential macro-defect modes was investigated. Results indicated that laser power was critical to the printability of high Ni content NiTi powder. In the low laser power setting (P < 100 W), the printable space was relatively wider with delamination as the main macro-defect mode. In the sub-high laser power condition (100 W ≤ P ≤ 200 W), the printable space was narrowed to a low hatch spacing region with macro-defects of warpage, elevated edge/corner, and delamination happened at different scanning speeds and hatch spacing combinations. The rough surface defect emerged when further increasing the laser power (P > 200 W), leading to a further narrowed printable space. 
    more » « less
  2. null (Ed.)
    Abstract 3D printing has been extensively used for rapid prototyping as well as low-volume production in aerospace, automotive, and medical industries. However, conventional manufacturing processes (i.e., injection molding and CNC machining) are more economical than 3D printing for high-volume mass production. In addition, current 3D printing techniques are not capable of fabricating large components due to the limited build size of commercially available 3D printers. To increase 3D printing throughput and build volume, a novel cooperative 3D printing technique has been recently introduced. Cooperative 3D printing is an additive manufacturing process where individual mobile 3D printers collaborate on printing a part simultaneously, thereby increasing printing speed and build volume. While cooperative 3D printing has the potential to fabricate larger components more efficiently, the mechanical properties of the components fabricated by cooperative 3D printing have not been systematically characterized. This paper aims to develop a data-driven predictive model that predicts the tensile strength of the components fabricated by cooperative 3D printing. Experimental results have shown that the predictive model is capable of predicting tensile strength as well as identifying the significant factors that affect the tensile strength. 
    more » « less