Accelerating Electrostatic Plasma Simulations with Reduced-Order Modeling

In this article, we propose a novel approach to optimize process parameters in Directed Energy Deposition (DED) additive manufacturing, also known as metal 3D printing. Our method uses a type of artificial neural network called Autoencoders to learn the relationship between the laser power and scan speed parameters and the resulting temperature history within the build plate.
Think of an Autoencoder like a robot that takes in a complex piece of data, like a 3D print, and then compresses it into a simpler form, called the latent space. The robot then takes this compressed data and reconstructs the original 3D print from it. By training the robot on a large dataset of 3D prints, it can learn to recognize patterns in the data that allow it to make accurate predictions about how the temperature will behave during the printing process.
We use a special type of Autoencoder called a Variational Autoencoder (VAE) to learn the relationship between the laser power and scan speed parameters and the resulting temperature history. The VAE is trained on a large dataset of 3D prints, and it learns to compress the data into a simpler form that captures the most important features of the temperature history.
Once the VAE has been trained, we can use it to make predictions about how the temperature will behave during the printing process for new 3D prints that the robot hasn’t seen before. This allows us to optimize the laser power and scan speed parameters in real-time, without the need for trial and error or extensive computational resources.
Our approach has several advantages over traditional methods of optimizing process parameters in DED additive manufacturing. Firstly, it allows for real-time optimization, which means that we can adjust the parameters during the printing process to ensure that the temperature remains within a safe range. Secondly, it uses machine learning algorithms to learn the relationship between the parameters and the temperature history, rather than relying on empirical formulas or trial and error. Finally, our approach can be used for a wide range of 3D prints and materials, without the need for extensive retraining of the VAE.
In summary, our article proposes a novel approach to optimize process parameters in Directed Energy Deposition additive manufacturing using Autoencoders. By learning the relationship between the laser power and scan speed parameters and the resulting temperature history, we can make accurate predictions about how the temperature will behave during the printing process and adjust the parameters accordingly. This allows for real-time optimization and reduces the need for trial and error or extensive computational resources.