Monash Automations

AI powered 3D Printer Error Detection System

Project Lead

Overview

An automated vision-based system designed to detect 3D printing failures in real time and prevent wasted material, time, and operational costs. Camera sensors were deployed across a 3D printer farm to continuously monitor print progress. Images were periodically captured via a Raspberry Pi and passed through a deep learning model to classify print failures. Once a failure was confirmed, the system communicated directly with Prusa XL printers through a REST API to pause or cancel the print automatically.

Technical Approach

• Periodic image polling of active prints using Raspberry Pi–connected cameras
• Failure classification using a ResNet-18–based CNN
• Automated printer control via PrusaLink's REST API
• Endoscopic nozzle monitoring for close-range failure detection

My Contributions

• Customised the ResNet-18 architecture in PyTorch to suit the specific failure classification task
• Designed and implemented the end-to-end training pipeline, using datasets of up to 10,000 labelled images
• Deployed a webcam server to stream an endoscope view of the printer nozzle for enhanced fault visibility

Impact and Outcomes

• Reduced failed print runtime and material waste
• Enabled autonomous intervention without human supervision
• Improved reliability and scalability of printer farm operations

• Learn more

Stroke Rehab Glove

Project Engineer

Overview

In Australia, nearly 50,000 people are affected by stroke each year, with many experiencing partial or complete loss of hand mobility. This project aims to improve the accessibility and affordability of stroke rehabilitation gloves designed to reduce spasticity and support the rebuilding of neural pathways responsible for hand control. Inspired by the Syrebo rehabilitation glove, we are developing a custom controller and rehabilitation program tailored for flexible, software-driven therapy. The project is currently in Phase 1, focused on proof-of-concept development.

Technical Approach

• Reverse engineering of a commercially available Syrebo rehabilitation glove
• Mechanical modification of the glove to improve finger dexterity and range of motion
• Design and development of a custom controller
• Integration with a microcontroller to enable programmable and adaptive rehabilitation routines

My Contributions

• Disassembled and analysed the original Syrebo controller hardware
• Used laboratory equipment, including oscilloscopes and multimeters, to characterise signals and support reverse engineering
• Designed and built a custom controller, enabling full software control of the glove via a microcontroller