風力タービンブレード製造における欠陥修理のカーボンフットプリント評価：SGRE工場の事例研究
Assessing carbon footprint of defect repair in wind turbine blade manufacturing: a case study in an SGRE factory (原題)
Léo Staccioli, José Gallego, Milena Nasner, Grégoire Lebreton
🤖 gxceed AI 要約
日本語
風車ブレード製造の欠陥修理に伴うCO2排出を、標準プロセス、最適化注入、センサー＋機械学習の3構成で評価。最適化注入はCO2を30.4%削減、機械学習構成は19.5%削減で、欠陥数削減効果(32.1%)ほどCO2削減に寄与しない。年間490枚生産で約900トンと580トンのCO2回避を見込む。
English
This study assesses carbon footprint of defect repair in wind turbine blade manufacturing across three configurations: baseline, optimized infusion, and sensor+ML. Optimized infusion cuts repair CO2 by 30.4%, ML by 19.5%, less than its 32.1% defect count reduction due to larger remaining defects. Annual savings for 490 blades: ~900 and 580 tCO2e respectively.
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
日本は洋上風力導入拡大に伴い、ブレード製造のサプライチェーン排出削減が重要。本研究成果は製造工程の脱炭素化に寄与し、国内メーカーのScope 1・2排出削減や調達先への環境要求対応に示唆を与える。
In the global GX context
This work contributes to global manufacturing decarbonization, aligning with ISSB/CSRD disclosure expectations for supply chain emissions. It demonstrates how ML can optimize processes, though its carbon benefit may be less than defect reduction suggests, informing similar assessments in other industries.
👥 読者別の含意
🔬研究者:Provides a method for defect-level carbon footprinting and highlights that ML-driven defect reduction may not linearly translate to emissions savings.
🏢実務担当者:Offers quantified carbon savings from process optimization and ML, useful for sustainability reporting and process improvement decisions.
📄 Abstract（原文）
Background Wind turbine blade manufacturing by vacuum resin infusion is prone to defects that trigger repair operations, consuming additional resin, glass fibre and energy beyond the material embedded in the blade. The Horizon Europe Innovation Action TURBO (Grant Agreement No. 101058054) develops technologies to reduce these repair-related impacts. This study reports a defect-level carbon footprint assessment of three manufacturing configurations tested on a blade demonstrator. Methods Three physical castings of a 16 metre section of a 108 metre blade mould were produced and inspected: a standard-process baseline, an optimised-infusion configuration, and a sensors-plus-machine-learning configuration. Repair-related resin, glass fibre and grinding-energy consumption were quantified from post-manufacturing defect inventories and converted to carbon dioxide equivalent (CO 2 eq) emissions using reference factors. Results were extrapolated to a full blade using a surface-area scaling factor, and to annual fleet production. Defects were grouped into four families to assess whether mitigation acted uniformly across defect types. Results Optimised infusion reduced defect-repair CO 2 eq by 30.4% relative to baseline; the machine-learning configuration reduced it by 19.5%, a smaller benefit than its 32.1% reduction in defect count would suggest, because the defects it fails to prevent are on average larger. Both mitigation scenarios almost completely removed root-end air pockets, delamination and miscellaneous defects (50 to 100% reduction), but left dry-spot defects, the largest baseline family, nearly unaffected (4.0% reduction). The optimised-infusion estimate, cross-checked against an independent method, agreed within 3.1%. Scaled to a realistic annual production of 490 blades, avoided impacts of optimised infusion and the machine-learning configuration represent approximately 900 and 580 tonnes CO 2 eq per year. Conclusions These results indicate that both TURBO technologies already deliver measurable reductions in defect-repair carbon footprint, identifying dry-spot defects as the main remaining target for further improvement, with extension to a full impact category suite identified as future work.
🔗 Provenance — このレコードを発見したソース
- openalex https://doi.org/10.12688/openreseurope.24681.1first seen 2026-09-02 04:56:16
🔔 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。