Additive Manufacturing technologies, such as Laser Powder Bed Fusion, have enabled the creation of complex geometrical designs which can be used for lightweighting purposes across multiple industries. One of the most common methods to reduce weight in the design stage is the use of topological optimization or lattice structures. Lattice structures consist of nodes connected with struts in different orientations in space, with the configuration of the unit cell varying depending on the final application. It is an established fact that the mechanical properties of additively manufactured components vary as a function of the size and orientation of the printed part. This inhomogeneity in properties is often neglected by material models implemented in finite element analysis, which normally just consider the mechanical properties of the bulk material. In this work, single struts of different diameters (0.5 mm and 1 mm) and orientations (0° and 45°) were tested to determine the corresponding mechanical properties and use them as an input to construct a material model to predict the mechanical properties of lattice structures. Validation against experimental behavior of two different lattice structures shows improved accuracy over bulk properties material models.

No abstract

Using an unsupervised convolutional neural network classifier, an automated workflow to generate the process map for printing Ti-6Al-4 V with laser powder bed fusion with minimal human supervision is proposed. Single scan vectors using a range of laser powers and scan speeds were printed on a bare Ti-6Al-4 V baseplate, which were then imaged using optical microscopy without further material preparation steps. After resizing and thresholding, the resulting dataset was used to train the neural network into automatically differentiating the tracks into categories. Post-analysis reveals that the model can differentiate between commonly observed track morphologies and map out the viable processing window automatically for the alloy.