Automatic in-situ error correction for 3D printed electronics

Abstract

Defects are a major issue in 3D printed electronics because even a tiny inaccuracy will lead to a faulty electronic circuit. This article presents a novel approach to correct printing defects with a neural network based automatic error correction. The errors are detected during printing by recording images of each wire with a high-resolution camera and segmenting the wires using convolutional neural networks. The neural network is trained with a dataset of printed wires with marked wire positions. A novel error detection algorithm then identifies connection breaks and generates repair paths for every connection break in the circuit, which are then executed by the printer. Multiple objects with deliberately inserted defects were printed and automatically repaired on different substrates using a Neotech AMT PJ15X printer to evaluate the performance. The algorithm detected all connection breaks, generated repair paths, and successfully repaired the faulty wires. This article also shows this approach's limitations and areas for future research, like complex circuits printed on 5-axis machines. The automatic error correction is highly reliable and is an important step towards a first-time-right production.

Keywords

• Additive manufacturing
• 3D printing
• Printed electronics
• Process monitoring
• Error correction
• Neural networks