Machine learning for smarter 3D printing

It is widely accepted that while some 3D printing techniques can offer rapid and highly effective artefacts, a high degree of error still remains. Each printer is different, filaments or materials used can behave in unexpected ways and in some cases, manufacturers are required to run several iterations before producing a version of acceptable quality.

A team of researchers from the University of Southern California Viterbi School of Engineering (CA, USA) has been working to combat the financial and economical ramifications of trial-and-error 3D printing, by improving the accuracy by 50% or more.

Described in IEEE Transactions on Automation Science and Engineering, the process of ‘convolution modeling of 3D printing’ includes a new set of machine learning algorithms and the software tool, PrintFixer.

You may also be interested in:

The overarching goal is to develop an AI model that can accurately predict shape deviations in 3D printing processes, ultimately making 3D printing smarter.

What we have demonstrated so far is that in printed examples the accuracy can improve around 50% or more. In cases where we are producing a 3D object similar to the training cases, overall accuracy improvement can be as high as 90%,” Qiang Huang, Associate Professor at the University of Southern California, explained.

It can actually take industry eight iterative builds to get one part correct, for various reasons, and this is for metal, so it’s very expensive,” Huang continued.

The team claims that every 3D-printed object occurs some level of deviation to some degree, due to unexpected behaviors of the materials or the printer itself.

PrintFixer uses data learned from previous print jobs to train its AI to predict the location of distortions, supposedly ‘fixing’ errors before they occur. The team reports that their main aim was to create a model that manufactured artefacts from the input of the minimum amount of 3D printing source data.

As Huang explained:

From just five to eight selected objects, we can learn a lot of useful information. We can leverage small amounts of data to make predictions for a wide range of objects.”

The team has worked to train the model to work with many different materials, appropriate for many applications. One collaboration includes a dental clinic in Australia, as the team works to develop their model for the purpose of 3D printing dental models.

So just like a when a human learns to play baseball, you’ll learn softball or some other related sport much quicker. In that same way, our AI can learn much faster when it has seen it a few times,” added Nathan Decker, leader of the software development effort within the group.

So you can look at it and see where there are going to be areas that are greater than your tolerances, and whether you want to print it,” Decker continued.

The team plan to develop their project for use by commercial partners, to 3D printing hobbyists. Users around the world will also be able to contribute to the software’s improvement and development with the access to a database.

As Decker concluded:

Say I’m working with a MakerBot 3D printer using PLA (a bioplastic used in 3D printing), I can put that in the database, and somebody using the same model and material could take my data and learn from it. Once we get a lot of people around the world using this, all of a sudden, you have a really incredible opportunity to leverage a lot of data, and that could be a really powerful thing.”

Sources: Huang Q, Wang Y, Lyu M, Lin W. Shape deviation generator – a convolution framework for learning and predicting 3D printing shape accuracy. IEEE T Autom. Sci. Eng. doi: 10.1109/TASE.2019.2959211 (Epub ahead of print) (2020);

Source link

Show More

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *