生成AIを用いた1D-CAEモデリング要件の対話的抽出とモデル構成の決定

内藤　恭兵・市村　純一・脇本　佑紀（ニュートンワークス）

工数制約により形骸化しがちな，実システムを解析目的に応じて理想化する「モデル化プロセス」に対し，生成AIとの対話を用いた支援手法を提案する．解析目的から適切な仮定や抽象化の指針を対話的に引き出し，工学的に妥当な仕様へと構造化することで，効率的かつ本質を捉えたモデル定義を行うプロセスについて論じる．

生成AIによる1D-CAEモデル自動実装と要求仕様との整合性確認支援

市村　純一（放送大学）・内藤　恭兵・脇本　佑紀（ニュートンワークス）

入力された仕様に基づく1D-CAEモデル自動実装と，要求仕様との整合性や物理的実現性の確認を支援する手法について論じる．実装結果と仕様の乖離要因をAIが考察・レポート化する．設計者がこれを基に仕様適合性を確認し，モデルの修正方針を決定し適正化を図るHuman-in-the-Loop型プロセスを提案する．

クロスドメインモデルを用いたパワーマネジメント，熱性能，車両運動性能等を同時に考慮するMBD手法の開発
-モデル構築と車両開発への適用-

高橋　成光・内海　碧人・松本　佑大・谷　雅之（日産自動車）

本報では，クロスドメインモデルCAEM（Collective Automotive Engineering Model) の概要と車両開発MBDへの適用事例に関して述べる．エネルギーフローに着目し，パワーマネジメント，熱性能，車両運動性能を統合環境で完全に同一の設計情報から予測可能なモデルを開発，ドメインを跨いでトレードオフを考慮したモジュール選定等の検討が可能であることを確認した．

Virtual Validation Methodology and Credibility Assessment for Automated Driving Systems with Applications to NCAP 2029

Prof. Dr. Reza Rezaei･Dr. Christian Lang･Kento Fukuhara･Gnana Prakash Reddy Madduri･Dr. Simon Olma (IAV)･Dr. Dai Araki (Toshiba Digital Solutions Corporation)

This paper presents a methodology for virtual validation and credibility assessment based on EU 2022/1426 and ISO 3540x. The methodology incorporates co-simulation for overall system modeling and validation of automated driving (AD) Level 3+ systems, integrating evidence-based real-world scenarios. Furthermore, it is designed to align with NCAP 2029 requirements, ensuring applicability for virtual testing and crash avoidance safety rating programs. By combining advanced physics-based validation techniques to ensure simulation credibility with a novel LLM-based scenario creation, the framework supports the virtual development and validation of AD systems, addressing future needs for cost reduction, reliability, and extensive mileage testing.

Integrated Model-Based Development Process for Vehicle Thermal Energy System

CHOO BUMSEOK･AN HYUNMO･JO CHANWOONG (Hyundai Motor)

This study introduces an integrated model for automotive thermal energy systems for model-based development (MBD), aligned with ISO 26262 standards. The model consolidates system data such as configurations, functions, and performance into a unified digital platform to enhance collaboration. A structured development process is proposed using tools like MathWorks System Composer and GT-SUITE, with automation for thermal flow analysis. The approach is validated through a Rankine cycle case study, demonstrating improved consistency and efficiency.

Model-Based and Data-Driven Software Architectures for EV Powertrain and Energy Management Systems

Hadi Moztarzadeh (Advanced Propulsion Centre UK)

The shift toward software-defined vehicles (SDVs) is accelerating demand for advanced software architectures that enable higher-fidelity control and optimisation of EV powertrains. This paper examines emerging SDV trends and readiness level, including centralised compute, service-oriented architectures, continuous software deployment, and data-centric control—and their implications for powertrain functionality. A hybrid model-based and data-driven framework is proposed, integrating physics-based control with machine-learning modules for state estimation, thermal prediction, and energy optimisation. System-level analysis shows that major global EV manufacturer integrating more SDV-enabling architectures to enhance computational capacity, cross-domain coordination, and real-time adaptability, enabling more efficient and predictive EV powertrain strategies.