Session: 03-08-02: Computational Modeling and Simulation for Advanced Manufacturing

Paper Number: 114152

114152 - Comparison of Machine Learning Models and Analytical Scaling Law for Predicting Melt-Pool Depth in Laser Powder Bed Fusion (LPBF) Additive Manufacturing 

Additive manufacturing (AM) has gained significant attention in recent years as a promising technology for manufacturing complex geometries with superior mechanical properties. Laser powder bed fusion (LPBF) is one of the most widely used AM techniques, which involves the selective melting of powder particles to create a solid object layer by layer. Lattice structures are commonly used to improve the mechanical properties of AM-manufactured parts, for example the structure could significantly reduce mass, tailor stiffness of surface, and improve energy-absorbing characteristics. However, the mechanical properties of parts with different lattice structures, cell sizes, and heat treatment conditions are not well understood. Therefore, there is a need for a robust methodology that can accurately predict the performance of AM-manufactured parts under varying lattice structures and heat treatment conditions. Machine learning (ML) soars rapidly recent years, features in “learning” hidden pattern among large amount of data to improve performance. Researchers from various field apply ML into specific field to solve difficult problem. Especially supervise learning is widely used in AM application, for example melt pool characteristic prediction, closed-loop control and defect detection, property prediction, parameter optimization and cost estimate etc. As a subset of ML, deep learning (DL) passed data through web of different layers structure, like human brain, computing with hundreds of neurons in linear or non-linear methods, which is like how human brain process data information, to map out the hidden pattern and apply to new data set to improve performance. This paper explores the use of deep learning techniques for predicting the capillarity, tensile strength, and compression strength of parts manufactured using laser powder bed fusion (LPBF) with varying lattice structures, cell size, and heat treatment. The lattice structures in the experiment include body-centered cubic (BCC), face-centered cubic (FCC), and octet, with varying cell sizes. The authors present a comprehensive analysis of the influence of these factors on the mechanical performance of the LPBF manufactured parts. The authors propose a deep learning-based model that takes these factors into account and predicts the performance of the manufactured parts to optimizing parameters for the best mechanical-performance lattice structure. The model is trained and tested on a dataset of LPBF manufactured parts with varying lattice structures, cell sizes, and heat treatment conditions. The results demonstrate the effectiveness of the proposed deep learning-based approach in accurately predicting the performance of the manufactured parts under various lattice structures, cell sizes, and heat treatment conditions. The paper concludes by discussing the implications of these findings for the design and optimization of LPBF-manufactured parts, and future research directions in this area

Presenting Author: Siva Surya Prakash Reddy Arikatla University of the DC

Presenting Author Biography: Siva is a graduate student at UDC and he has worked at CAM-STAR since 2021.

Authors:

Feiyang Bai UDC
Siva Surya Prakash Reddy Arikatla University of the DC
Nian Zhang UDC
Fisseha Gebre UDC
Jiajun Xu University of the District of Columbia