In-Situ Verification of 3D-Printed Electronics Using Deep Convolutional Neural Networks

Date
2021
Authors
Ahlers, Daniel
Journal Title
Journal ISSN
Volume Title
Publisher
University of Texas at Austin
Abstract

Printed electronics processes are becoming more stable and evolve into first industrial applications. These industrial applications require proper quality assurance to get a mostly autonomous production process. In this work, we present a new approach to inspect printed electronics and ensure their quality. Our hardware setup extends a fused filament fabrication (FFF) printer with an extruder for direct dispensing of conductive paste, a pick and place unit, and two cameras. The cameras take multiple images during printing. A trained neural network analyzes these pictures to separate the electronic wires from the plastic background. All separated images of a layer are combined to get a full view of the layer. Our algorithms then examine the detected wires to identify printing flaws. The algorithms currently detect connection breaks, shorts, find points that have not been reached, and evaluate the width of the printed wires.

Description
Citation