Using Infrared Image Analysis to Detect 3D Printing Errors
Session Number
Project ID: ENGN 03
Advisor(s)
Professor Niechen Chen, Northern Illinois University
Discipline
Engineering
Start Date
17-4-2024 9:40 AM
End Date
17-4-2024 9:55 AM
Abstract
This paper investigates a method to detect structural irregularities in 3D prints through infrared image analysis. The study employs an infrared camera, which senses thermal discrepancies in its field of view, to monitor the thread of the material filament as it is deposited during the printing process. Images captured by the camera are processed by Python libraries such as OpenCV and NumPy, using object detection and cropping techniques to differentiate between the hot extrusion material and the hardened plastic for precise error identification. The data collection process involves collecting defective and non-defective prints to train the image processing algorithm. The paper compares various approaches, including cropped and uncropped images, at different zoom levels to determine the most effective method for error detection. Preliminary findings demonstrate promising error detection rates, with varying success based on the image processing strategy. Challenges like false positives are addressed through refined data and algorithmic adjustments. The paper concludes with insights from data analysis and suggests future improvements for 3D printing error detection via image processing and machine learning.
Using Infrared Image Analysis to Detect 3D Printing Errors
This paper investigates a method to detect structural irregularities in 3D prints through infrared image analysis. The study employs an infrared camera, which senses thermal discrepancies in its field of view, to monitor the thread of the material filament as it is deposited during the printing process. Images captured by the camera are processed by Python libraries such as OpenCV and NumPy, using object detection and cropping techniques to differentiate between the hot extrusion material and the hardened plastic for precise error identification. The data collection process involves collecting defective and non-defective prints to train the image processing algorithm. The paper compares various approaches, including cropped and uncropped images, at different zoom levels to determine the most effective method for error detection. Preliminary findings demonstrate promising error detection rates, with varying success based on the image processing strategy. Challenges like false positives are addressed through refined data and algorithmic adjustments. The paper concludes with insights from data analysis and suggests future improvements for 3D printing error detection via image processing and machine learning.