Facing AI challenges for AM together

There are many ways to use Artificial Intelligence (AI) for Additive Manufacturing (AM). Creating a deep machine learning system, however, is not straight forward. Maud Chidiac, AI Program Manager at AMEXCI, explains why AMEXCI started the Rosetta Protocol, and why it is crucial that various stakeholders come together to crack the code.   

Machine learning algorithms require large amounts of data before they can begin to perform powerful results. The Rosetta Protocol initiated by AMEXCI in May of 2020 is about joining forces between AM data users and collecting different types of complementary metal L-PBF monitoring datasets. 
– The main challenge companies have in common is to obtain a diverse, consistent, and robust dataset to train a machine learning model. We hope that our partners would like to join the Rosetta Protocol and take part in a mutual exchange of non-critical monitoring data to cover all possible use cases, says Maud Chidiac. 

Once these raw data are gathered as main material for training an algorithm, the next question is – what to do with it? While most industry players have different strategies for their AM data, one common challenge is to define the right scope based on those datasets. For a year, AMEXCI has done the groundwork – exploring the possible areas of AI applications in the AM workflow, defining a data collection strategy and limiting the scope. Focus areas could be design data, simulation data, testing data, and monitoring data. The key is to think big but start concrete. 
– The scope we chose is to focus on the findings from our in-house monitoring systems to try and figure out what they can say about errors occurring during a print. Since we all struggle in this quest of decrypting our monitoring and AM workflow data, we believe that opening up the dialogue between AM data users could be a positive development for the industry, says Maud Chidiac.

Building on complementary AI approaches 

Participation in the Rosetta Protocol will not be limited to data sharing only but will give participants the opportunity to share experiences based on their specific approach. 
– No matter the way companies explore machine learning for AM, we will build on our complementarities. Participating companies will have access to a forum where they can share their challenges with others. We want to be the driving force in making this joint initiative beneficial to our individual AI journeys, says Maud Chidiac.

Related Posts