A new machine learning algorithm can detect and correct a wide range of errors that can occur during 3D printing. It also enables printers to learn autonomously – without human intervention – to print new materials. The algorithm is suitable for both existing and new 3D printers.
3D printing enables the production of a very wide range of products. For example, companies use the printers to manufacture custom-made medical implants, but also to manufacture components for aircraft. However, also consider the production of prototypes, which is possible in less time thanks to 3D printers.
Error in the printing process
Despite the wide possibilities of 3D printing, the technique also has its limitations. For example, various errors may occur during the printing process. These range from minor inaccuracies in a print to mechanical weaknesses that affect the usability of the entire print.
Such errors can currently only be detected by a trained employee observing the printing process. This employee must recognize the print, stop the print process, remove the part and adjust settings for a new print. This can be a challenge, especially with long-lasting prints that take hours. Simultaneous monitoring of several 3D printers is also a challenge for employees.
Existing systems have their limitations
The parties have been working on systems for automated monitoring of 3D printing and detection of deviations for some time now. However, existing systems for this purpose have their limitations. For example, they can usually detect a limited number of defects in a specific part produced by a specific 3D printer in a specific material.
A new algorithm developed by a team at the University of Cambridge changes this. “What is really needed is a ‘driveless car system’ for 3D printing,” explains researcher Douglas Brion from the University of Cambridge’s Department of Engineering. “A driverless car would be useless if it only worked on one road or in one village. It must learn to generalize in different environments, cities and even countries. A ‘driveless’ printer should perform similarly for multiple parts, materials and printing conditions.”
Here, Brion and Dr. Sebastian Pattinson from the university’s Department of Engineering has now found a solution. “This means you can have an algorithm that looks at all the different printers you have in use, constantly monitors them and makes adjustments as needed – essentially doing what people can’t,” explains Pattinson.
Dataset with 950,000 images
The algorithm that the researchers use is a deep learning computer vision model. This model was trained using a dataset of 950,000 images created automatically during the printing of 192 objects. Each image is labeled with the printer’s settings, including print speed, nozzle temperature, and the speed at which media moves through the printer. The model is also provided with information on how these settings differ from the ideal settings. In this way, the algorithm can learn how problems arise.
“Once trained, the algorithm can determine by looking at an image which settings are right or wrong – for example, a certain setting is too high or too low, and then make the correct correction,” explains Pattinson. “The cool thing is that printers using this approach can continuously collect data, so the algorithm is constantly improving.”
Brion and Pattinson’s algorithm is widely applicable. It can detect and correct errors in unknown objects or materials. It is also suitable for 3D printers where it has not been used before. Brion explains that with 3D printing with a nozzle, similar problems can arise regardless of the material you print in. “If, for example, the nozzle moves too quickly, you often get blobs of material. Or if you push too much material through the die too quickly, printed lines will overlap, causing creases.”
The researchers trained the algorithm using one type of print material and one 3D printer. Thanks to these trainings, the algorithm is also able to detect and correct errors in various materials, from polymers to ketchup and mayonnaise. The algorithm is also not tied to the 3D printer it is trained on.
According to the researchers, if the algorithm is trained further, it can become more effective and reliable than a human employee in detecting printing errors. They therefore expect that the algorithm can play an important role in quality control in applications where printing errors can have serious consequences.
3D printer Snapper from Inkbit
A few years ago, the American startup Inkbit already developed a 3D printer that can independently correct errors during the printing process. It was about Snapper. Inktbit was founded by electrical engineering and computer science professor Wojciech Matusik and former MIT students Javier Ramos, Wenshou Wang, Kiril Vidimče and David Marini. Snapper uses optical coherence tomography (OCT), a technology traditionally used in ophthalmology. This involves a custom-made OCT scanner which checks each print layer for defects in real time. Will these be discovered? Then Snapper adjusts the printing process in real time using machine learning algorithms to correct the error.
More information about the University of Cambridge algorithm is available here.
Author: Wouter Hoeffnagel
Photo: Lutz Peter from Pixabay