Researchers Develop Machine Learning Method to Monitor 3D Printing Process for Defects
All sorts of issues can occur when a 3D printed part has a defect, and researchers around the world continue trying to find new ways to detect these defects before they cause too many problems. Now, two researchers from the Department of Industrial and Manufacturing Systems Engineering (IMSE) at Kansas State University are taking on the challenge as well.
Not all 3D printers have a designated system for tracking and monitoring 3D printing progress during the job, which means that some parts will continue to print even if there are defects. Ugandhar Delli and Dr. Shing Chang from IMSE have proposed a new method to monitor the 3D printing process by pausing the system at different checkpoints, in order to take stock of things and find any defects; then, corrective actions, such as stopping the print if defects are detected, can be taken if necessary.
Production-scale 3D printing operations especially need automatic process monitoring, as finding defects and halting the prints during the critical stages of the process can not only save money but also cut back on material waste.
Delli and Dr. Chang published a paper on their research, titled “Automated Process Monitoring in 3D Printing Using Supervised Machine Learning,” that details their new process monitoring method in the Procedia Manufacturing journal.
The abstract reads, “Quality monitoring is still a big challenge in additive manufacturing, popularly known as 3D printing. Detection of defects during the printing process will help eliminate waste of material and time. Defect detection during the initial stages of printing may generate an alert to either pause or stop the printing process so that corrective measures can be taken to prevent the need to reprint the parts. This paper proposes a method to automatically assess the quality of 3D printed parts with the integration of a camera, image processing, and supervised machine learning. Images of semi-finished parts are taken at several critical stages of the printing process according to the part geometry. A machine learning method, support vector machine (SVM), is proposed to classify the parts into either ‘good’ or ‘defective’ category. Parts using ABS and PLA materials were printed to demonstrate the proposed framework. A numerical example is provided to demonstrate how the proposed method works.”
The researchers used a LulzBot Mini 3D printer for their research, which involved checking on the part during the 3D printing process at multiple checkpoints; these are defined as critical stages where there is a major change to a part’s geometry.
Delli and Dr. Chang wrote in the paper, “For example, consider a complex part which involves different stages of printing like skirt/base, body and the top. These stages could be considered as the desired checkpoints to inspect quality at.”
Their method used an integrated camera, image processing, and a supervised machine learning model, called a support vector machine (SVM), to automatically assess part quality. The researchers defined three steps in order to implement their proposed 3D printing quality monitoring during production:
- Identify the proper checkpoints for the 3D printed part according to its geometry.
- Take images of the semi-finished part at each checkpoint.
- Perform image processing and analysis.
The researchers concluded that their method could detect both structural or geometrical defects and completion failure detects. However, it’s not 100% foolproof just yet.
The main drawback of the proposed method is that the printing process needs to be paused while the images of a semi-finished part are taken,” Delli and Dr. Chang wrote. “Another drawback is that since only top view images are taken, the proposed method might not be able to detect the defects on the vertical plane which cannot be seen in the top view image. This gives us a direction for future research to incorporate cameras on the sides of the printer as well to detect defects on both the horizontal and vertical planes.”
Delli and Dr. Chang believe they could improve the method by in situ camera mounting on the print head. In the future, they will also work on studying the impacting factors for selecting the appropriate print checkpoints.
Discuss this story and other 3D printing topics at 3DPrintBoard.com or share your thoughts below.
You May Also Like
Alchemite Machine Learning Engine Used to Design New Alloy for Direct Laser Deposition 3D Printing
Artificial intelligence (AI) company Intellegens, which is a spin-off from the University of Cambridge, created a unique toolset that can train deep neural networks from noisy or sparse data. The machine...
Switzerland: Exciting New Technology Multi-Metal Electrohydrodynamic Redox 3D Printing
Researchers from Switzerland explain more about how metals dissolved and re-deposited in liquid solvents can further AM processes by promoting fabrication without post-processing. Their findings are outlined in the recently...
Using Semi-Supervised Machine Learning in Laser Powder-bed Fusion Fault Detection
Researchers from the University of Liverpool outline their findings regarding the automatic detection of faults in additive manufacturing products in a recently published paper, ‘Automatic fault detection for laser powder-bed...
Italy: Experimenting with Coffee in 3D Printed Collectors for Better Solar Absorption
3D printing can be performed in countless ways; in fact, the choices are as infinite as the ability to innovate using this technology with an expanding force of hardware, software,...
View our broad assortment of in house and third party products.