January Issue of Additive Manufacturing Pairs AM with Machine Learning

The January issue of Additive Manufacturing magazine looks at three ways machine learning is advancing understanding of 3D printing processes and materials.


Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

Compared to conventional processes, additive manufacturing (AM) lacks the years—and in some cases, centuries—of knowledge building that makes machining or casting predictable. But it’s not just about age. AM also has a greater number of variables that can affect the final outcome of a part. To give just one example, a powder-bed process such as selective laser melting (SLM) depends on variables ranging from laser power and speed to material composition to the pressure of the shielding gas.

Early gains in additive manufacturing were made through trial and error, relying heavily on human judgment and experimentation. But the diversity and number of variables involved in 3D printing makes AM a good match for another emerging technology: machine learning, or the application of computer algorithms to identify patterns in data. When coupled with human judgement, machine learning has the potential to accelerate additive manufacturing’s advance.

The January issue of Additive Manufacturing explores this potential with three stories of machine learning applications within AM:

  • ADAPT, the Alliance for the Development of Additive Processing Technologies, is using machine learning to map variables and outcomes for metal 3D printing, with the goal of developing a predictive model for additive builds.
  • A computer vision system developed at Carnegie Mellon University has learned to identify metal powders for 3D printing with 95 percent accuracy, a capability that could speed material qualification for AM.
  • GE’s Global Research Center is building a digital library of AM parameters that will one day enable closed-loop machine learning at the machine.

Read these stories and more in the digital edition.