Feature selection

This technique assists in determining the most influencing parameters from the given list of parameters using statistical tools.

Illustration

For instance, in order to find out which parameter is affecting significantly on the 3D printed part e.g., hatch distance, or laser power, or layer thickness. In this situation, a Pearson's coefficient will be determined between two parameters that will figure out the dependency between two parameters. If the Pearson's value (max. = 1) is higher for one parameter as compared to the other parameter, it means it will be affecting my desired output significantly.

Feature combination

This technique helps to carry out dimensionality lessening for the input attributes, and thereby concentrating on the newly generated features. Once the translation regulation is identified, manual manipulations are usually preferred. Mathematical tools such as principle components analysis can be utilized for the same purpose on the basis of attribute.

Illustration

For example, energy density (ED) influences the solidification, metallurgical, microstructure, and mechanical properties of a 3D printed part. Laser power, scanning speed, hatch distance, and layer thickness combines and generates a new feature ED. The aforementioned four features control the ED.