Please log in

Paper / Information search system

日本語

ENGLISH

Help

Please log in

  • Summary & Details

Fundamental Study on Inferring Aerodynamic Characteristics by Large Language Models Using Latent Representations of 3D Shapes

3次元形状の潜在表現を用いた大規模言語モデルによる空力特性推論に関する基礎研究

Detailed Information

Author(J)1) 伊藤 悟, 2) 西口 浩司, 3) 仲谷 幸一郎, 4) 尾嵜 悠利, 5) 金森 敬宗, 6) 則竹 真吾, 7) 田中 博
Author(E)1) Satoru Ito, 2) Koji Nishiguchi, 3) Koichiro Nakaya, 4) Yuuri Ozaki, 5) Hiromune Kanamori, 6) Shingo Noritake, 7) Hiroshi Tanaka
Affiliation(J)1) 名古屋大学大学院, 2) 理化学研究所, 3) 名古屋大学大学院, 4) 名古屋大学大学院, 5) トヨタシステムズ, 6) トヨタシステムズ, 7) トヨタシステムズ
Affiliation(E)1) Nagoya University, 2) RIKEN, 3) Nagoya University, 4) Nagoya University, 5) Toyota Systems, 6) Toyota Systems, 7) Toyota Systems
Abstract(J)LLMの自動車の設計支援への応用のためには,3次元空間の学習が不可欠である.本稿はその初期的研究として3次元形状を潜在ベクトルに圧縮して学習させ,抗力係数を推論する手法を提案する.本研究では潜在ベクトルの次元数が推論精度に与える影響を検証し,次元数の増加による推論精度の向上を確認した.

Translation

Abstract(E)To integrate Large Language Models (LLMs) into the automotive aerodynamic design process, it is essential to accurately capture complex, non-linear geometric features. This study proposes an approach that compresses 3D car shapes into latent vectors using DeepSDF and directly inputs them into an LLM to predict drag coefficients (CD). Specifically, we investigate the effect of expanding the dimensionality of the latent vectors on the inference accuracy. Results using the DrivAerNet++ dataset demonstrate that increasing the dimensionality enhances the LLM's physical inference capability, significantly reducing the prediction error without compromising geometric reconstruction accuracy.

About search

close

How to use the search box

You can enter up to 5 search conditions. The number of search boxes can be increased or decreased with the "+" and "-" buttons on the right.
If you enter multiple words separated by spaces in one search box, the data that "contains all" of the entered words will be searched (AND search).
Example) X (space) Y → "X and Y (including)"

How to use "AND" and "OR" pull-down

If "AND" is specified, the "contains both" data of the phrase entered in the previous and next search boxes will be searched. If you specify "OR", the data that "contains" any of the words entered in the search boxes before and after is searched.
Example) X AND Y → "X and Y (including)"  X OR Z → "X or Z (including)"
If AND and OR searches are mixed, OR search has priority.
Example) X AND Y OR Z → X AND (Y OR Z)
If AND search and multiple OR search are mixed, OR search has priority.
Example) W AND X OR Y OR Z → W AND (X OR Y OR Z)

How to use the search filters

Use the "search filters" when you want to narrow down the search results, such as when there are too many search results. If you check each item, the search results will be narrowed down to only the data that includes that item.
The number in "()" after each item is the number of data that includes that item.

Search tips

When searching by author name, enter the first and last name separated by a space, such as "Taro Jidosha".