Machine-Learned Linear Structural Dynamics

Technical Report · Sat Aug 17 04:00:00 EDT 2024

DOI:https://doi.org/10.2172/2430294· OSTI ID:2430294

Lindsay, Payton ^[1]; Shelton, Timothy Ryan ^[1]; Pierson, Kendall H. ^[1]; Kuether, Robert J. ^[1]; Najera-Flores, David Aaron ^[1]; Wilbanks, James Justin ^[1]; Parish, Eric Joshua ^[2]

---

1. Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
2. Sandia National Lab. (SNL-CA), Livermore, CA (United States)

The tension between accuracy and computational cost is a common thread throughout computational simulation. One such example arises in the modeling of mechanical joints. Joints are typically confined to a physically small domain and yet are computationally expensive to model with a high-resolution finite element representation. A common approach is to substitute reduced-order models that can capture important aspects of the joint response and enable the use of more computationally efficient techniques overall. Unfortunately, such reduced-order models are often difficult to use, error prone, and have a narrow range of application. In contrast, we propose a new type of reduced-order model, leveraging machine learning, that would be both user-friendly and extensible to a wide range of applications.

• Research Organization: Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States)
• Sponsoring Organization: USDOE National Nuclear Security Administration (NNSA); USDOE Laboratory Directed Research and Development (LDRD) Program
• DOE Contract Number: NA0003525
• OSTI ID: 2430294
• Report Number(s): SAND--2023-09694R
• Country of Publication: United States
• Language: English

Related Subjects

97 MATHEMATICS AND COMPUTING