A conceptual framework for layerwise energy prediction in laser-based powder bed fusion process using machine learning

Abstract

Additive Manufacturing (AM) is the process of successively joining materials layer by layer to produce a component from a digital 3D model. One of the widely adopted AM processes for producing metal components is laser-based powder bed fusion of metals (PBF-LB/M), in which lasers are used to selectively melt and fuse metallic powder in successive steps. An essential step towards economic industrial application is to minimize the energy consumption of the technology. Therefore, it is important to understand the relation between the process parameters (e.g., scanning speed, layer thickness, build time, laser power), part design, working environment, and the energy required to manufacture a component. In this work, we propose a conceptual framework of multimodal regression (MMR) for layerwise energy and quality prediction in AM process. The MMR model can be implemented using neural networks and used to analyze 2D images and process information of a layer to predict the layer quality and energy required to manufacture the layer. This model can further be used to optimize the part geometry, process design, and machine level parameters to utilize less energy without compromising on quality. Additionally, a PBF-LB/M use case is provided to discuss the feasibility of the proposed framework.