ニアゼロ炭素排出に向けた低温微量潤滑の多目的グラフニューラルネットワーク代替モデリング
Multi-objective graph neural network alternative modeling for low-temperature minimal lubrication towards near-zero carbon emissions (原題)
Mingming Wu, Mingyu Lu
🤖 gxceed AI 要約
日本語
低温微量潤滑（MQL）旋削の脱炭素化を目的に、グラフニューラルネットワークと多目的最適化を統合した意思決定手法を提案。工具・工作物・冷却材系の産業グラフを構築し、炭素効率（CER）を目的関数として改良差分進化法で最適化した。79件の直交実験により表面粗さ予測誤差を11.5%低減、切削力予測精度を20.2%向上、CERを20.0%改善した。チタン合金旋削に限定され、材料汎用性や炭素会計範囲に課題が残る。
English
Proposes a decision-making method integrating graph neural networks with multi-objective optimization to decarbonize low-temperature minimum quantity lubrication turning. An industrial graph of the tool-workpiece-coolant system quantifies coupled cutting thermodynamics, heat transfer, and lubrication. Using 79 orthogonal samples, it cut surface-roughness prediction error by 11.5%, improved cutting-force accuracy by 20.2%, and raised carbon efficiency by 20.0%. Scope is limited to titanium alloy turning.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
製造業のScope 1・2排出削減に資する工程レベル最適化の事例として、SSBJ対応で一次データ収集を迫られる製造業の排出原単位改善に示唆を与える。ただし炭素会計は工程範囲に限定され、開示枠組みとの接続は今後の課題。
In the global GX context
Adds a process-level, AI-driven optimization case to the manufacturing decarbonization literature relevant to Scope 1/2 reduction, though its carbon accounting is confined to the process boundary and does not yet connect to TCFD/ISSB disclosure frameworks.
👥 読者別の含意
🔬研究者:GNNと多目的最適化を工程レベル炭素効率に適用する手法設計の参考になる。
🏢実務担当者:切削工程の炭素原単位改善と品質予測精度向上を両立する実装例として参照可能。
📄 Abstract（原文）
This paper addresses the decarbonisation of low-temperature minimal quantity lubrication turning by proposing a decision-making method that integrates graph neural networks (GNNs) and multi-objective optimisation. An industrial graph of the tool-workpiece-coolant system is constructed to quantify the coupling of cutting thermodynamics, low-temperature heat transfer and fluid lubrication. First, Carbon Efficiency (CER) is used as the standalone optimisation goal. An improved differential evolution algorithm with dynamic scaling factors is used, and a hierarchical constraint mechanism (tool temperature ≤ 600°C, film thickness ratio ≥ 3) for process safety has been set up. Based on 79 orthogonal experiment samples, this method reduced the prediction error of surface roughness by 11.5% (test set RMSE=0.097 μm), improved the accuracy of cutting force prediction by 20.2%, and achieved a 20.0% CER optimisation (0.141 gCO 2 /mm³). Ablation experiments show that the proportion of the deep fusion module is about 3.5%-4.3%. Restricted to titanium alloy turning, this approach has limitations in material generality, dynamic factor inclusion, and carbon accounting scope. Future work will develop adaptive graph updates, build a full-life-cycle carbon model, and explore complex scenario applications to provide new technical references for the low-carbon transformation of manufacturing.
🔗 Provenance — このレコードを発見したソース
- openalex https://doi.org/10.1016/j.sasc.2026.200692first seen 2026-11-07 04:54:54
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。