Unveiling Processing--Property Relationships in Laser Powder Bed Fusion: The Synergy of Machine Learning and High-throughput Experiments

TL;DR

This work proposes a methodology embracing the synergy between high-throughput experimentation and hierarchical machine learning to unveil the complex relationships between a large set of process parameters in Laser Powder Bed Fusion (LPBF) and selected mechanical properties (tensile strength and ductility).

Abstract

Achieving desired mechanical properties in additive manufacturing requires many experiments and a well-defined design framework becomes crucial in reducing trials and conserving resources. Here, we propose a methodology embracing the synergy between high-throughput (HT) experimentation and hierarchical machine learning (ML) to unveil the complex relationships between a large set of process parameters in Laser Powder Bed Fusion (LPBF) and selected mechanical properties (tensile strength and ductility). The HT method envisions the fabrication of small samples for rapid automated hardness and porosity characterization, and a smaller set of tensile specimens for more labor-intensive direct measurement of yield strength and ductility. The ML approach is based on a sequential application of Gaussian processes (GPs) where the correlations between process parameters and hardness/porosity are first learnt and subsequently adopted by the GPs that relate strength and ductility to process parameters. Finally, an optimization scheme is devised that leverages these GPs to identify the processing parameters that maximize combinations of strength and ductility. By founding the learning on larger easy-to-collect and smaller labor-intensive data, we reduce the reliance on expensive characterization and enable exploration of a large processing space. Our approach is material-agnostic and herein we demonstrate its application on 17-4PH stainless steel.

Figures (24)

• Figure 1: Schematic flowchart of the proposed framework: High-throughput experimental approaches are coupled with hierarchical learning based on Gaussian processes to design the process parameters that optimize the combination of tensile strength and ductility.
• Figure 2: Designed experimental setups:(a) Schematic of the HT-compatible build design for cuboids with different layer thicknesses, along with the real LPBF generated cuboids, and (b) Tensile specimen dimensions and LPBF printed samples.
• Figure 3: Proposed hierarchical learning framework: Input-output spaces and model structures vary depending on the characteristics of each property.
• Figure 4: Data Fusion via GPs: To fuse the hardness/porosity with tensile data using GP+, we add a two-level categorical variable $t = \{Cuboid, Tensile\}$ (source indicator) to the input space. We use grouped one-hot encoding and matrix multiplication to convert $t$ to its low-dimensional quantitative representation $\boldsymbol{h}$. Then, these mapped values ($\boldsymbol{h}$) are concatenated with the quantitative input features and fed into the mean and covariance functions. To capture more complex relations in the data, we use a FFNN as a mean function and all the model parameters are estimated via MAP.
• Figure 5: Representative microstructural images and hardness maps of cuboids: The top row includes the as-polished surfaces that illustrate the concentration of defects. The second and third row show, respectively, phase formation and hardness maps. This figure highlights the large impact of processing conditions on the microstructure, defect content, and property of the samples.