FLEW Solutions

Detection of 3D print job faults using computer vision and machine learning (project completed)

1. QUT and FLEW worked together to develop a computer vision and machine learning system to detect faults in 3D print jobs, enabling rapid intervention in print jobs when faults are detected, saving time and money.
2. This system (including a large corpus of data) has primarily been developed using an existing set of 3D printers maintained by FLEW and has demonstrated over 80% accuracy in detecting faults within print jobs as they occur. The developed system makes use of state-of-the-art deep machine learning architectures, and multi-view data from a pair of stereoscopic cameras, to detect faults.
3. The next phase is further research and development to improve the technology readiness level, application, and commercialisation of the technology (as per project below).

ICSS project “Detecting 3D Printing Faults using Machine Learning” (R&D to increase TRL of solution, including scalability, generalisability, readiness for market, and commercialisation potential)

1. An ACEMS ICSS application was granted, to fund further research and development in relation to the above system, with more than $100,000 to be contributed by FLEW in cash and in-kind support.
2. The system is now being evaluated by FLEW as part of a closed alpha phase, during which it is being exposed to a greater number of printer types, printing materials, and fault types. This increase in complexity as the system is exposed to more diverse printers and materials necessitates further research and development to increase the system’s scalability, generalisability, and readiness for the market.

APR Intern Internship for Ethan Goan

ACEMS PhD student Ethan Goan (left) with Machine Learning Specialist at FLEW Solutions Australia, Samuel Fogarty (right)

1. An internship project has been negotiated, for ACEMS PhD student Ethan Goan to complete a paid internship (commencing 2021) with FLEW Solutions, under the academic supervision of ACEMS AI Simon Denman, working in collaboration with FLEW Solution (and affiliated company CloudForge) on the above project. The Internship should be funded by APR Intern, IMCRC and FLEW Solutions/CloudForge.
2. The system being developed is intended to be used to intervene in 3D printing operations, stopping print jobs if a failure is detected. As such, it is critical that reliable decisions are made. The research to incorporate uncertainty within model predictions will lead to better informed model decisions and will also provide a mechanism to seek human assistance when the model is unsure. The methods developed in this project will seek to develop a reliable and trustworthy machine learning model that can rapidly adapt to new conditions, and by providing an indication of when the model is uncertain enables future efforts that incorporate a human-in-the-loop to drive model adaptation in real-time.

Proposed Cryptography Research Project

1. An initial meeting was held to introduce FLEW Solutions to researchers from ACEMS UNSW and a UNSW cryptography researcher, to discuss a proposed future collaboration between FLEW Solutions and researchers from ACEMS QUT and UNSW.
2. FLEW is interested in proceeding with the collaboration, after the completion of the 3DP R&D project.

Proposed CRC-P / ARC Linkage

1. ACEMS QUT and FLEW/CloudForge aim to build towards larger projects such as a CRC-P and/or ARC Linkage.