Session: 02-02-02: Session #2: Measurement Science, Sensors, Non-destructive Evaluation (NDE) and Process Control for Advanced Manufacturing

Paper Number: 96045

96045 - As-Built Mechanical Property Estimation and Control of Laser Powder Bed Fusion Ss-316l Parts 

Laser Powder Bed Fusion (LPBF) additive manufacturing (AM) provides an unprecedented opportunity for achieving high-performance designs from a layer-by-layer material deposition method. Compared with traditional subtractive manufacturing, AM offers larger manufacturability that is not hindered by machine tool accessibility. LPBF provides a rapid fusion and solidification of the metal powders to form 3D models, which enlarges the design and manufacturability compared with the traditional subtractive machining process. In the meantime, AM techniques have not yet been adopted for on-demand manufacturing for their significant amount of human interaction with machines and multiple iterations to achieve successful high-performance components. In other words, there is always an additional quality measurement or post-processing process that comes with AM process to certify or improve the part performance for assuring the part functions. This is particularly difficult for an LPBF-ed AM part needs to be qualified through destructive measurements, such as tensile test, fatigue, and hardness. The as-built mechanical properties are highly correlated to the process parameters. If a set of optimized process parameters that are driven by the as-desired mechanical property can be provided, the as-built property can be assured. In this study, LPBF-ed 316L tensile testing samples are analyzed for investigating the correlation between input process parameters and the tensile strength measures. This study establishes the quantitative comprehensive process-quality modeling from five process parameters (laser power, scanning speed, hatch spacing, building orientation, and scanning direction). With a proposed novel multi-dimensional machine learning framework, as-built tensile strength can be maintained as desired or improved by adjusting the process parameters. The effectiveness and accuracy of the quantitative process-quality model are validated through experiments.

The proposed study seeks to develop and demonstrate the foundational capability for autonomous real-time quality prediction during the LPBF process. New analytical methodologies are needed to (1) characterize process parameters behavior with respect to tensile strength, (2) quantitatively integrate multi-dimensional data with correlated effects to the tensile strength, and (3) the process-quality modeling framework is capable of adapting changes in the dimensions and size of the process data. This study presents our experimental studies on LPBF-ed parts tensile strength, as well as the analysis of multi-process variants characteristics through a novel machine learning framework. The proposed machine learning framework enables the process-quality model (1) can effectively visualize complex multi-dimensional process-quality interactions, (2) predict and measure the as-built tensile strength based on a given set of processing conditions, and (3) reversely design the experiments to select optimal processing features that can significantly deviate the tensile strength.

Presenting Author: Xinyi Xiao Miami University

Presenting Author Biography: Dr. Xiao joined Miami in the Fall 2020 and is an Assistant Professor of Mechanical and Manufacturing<br/>Engineering. She received her Ph.D. and master in Industrial Engineering from The Pennsylvania State<br/>University in 2020 and 2017 respectively.<br/>Dr. Xiao past research is in developing autonomous methods in advanced manufacturing to reduce human<br/>interaction and manufacturing risks, and her work was funded by NSF, and US Army. Her current<br/>research interests mainly include, Process - Quality Prediction and Control in Additive Manufacturing,<br/>Quality Enhancement through Hybrid Additive Manufacturing, and Self-Morphing Structure Design and<br/>Control.

Authors: