Optimizing Multiple Tier Supplier Networks with Recurrent Neural Network Model in Reducing Scope 3 Carbon Footprint in a Product Supply Chain

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Provides a novel RNN-based methodology for Scope 3 optimization with empirical validation in apparel supply chains.

🏢実務担当者:Demonstrates how regional manufacturing consolidation in Asia-Pacific can significantly reduce logistics emissions, aiding supply chain redesign.

🏛政策担当者:Offers evidence that trade policies favoring regional integration can reduce carbon footprints; useful for designing green trade agreements.

Scope 3 emissions constitute a significant portion of the total product carbon footprint, particularly within globalized supply chains involving cross-border transportation. In response to this challenge, a recurrent neural network (RNN)-based optimization model is developed to optimize the routing of multiple tiers of supplier networks in minimizing logistics-related carbon emissions during the production of standardized products along the global supply chain. The model has been applied to the cotton polo shirts from a well-established segment of the apparel industry. Two routing configurations are assessed: a European Union (EU) model and an Asia–Pacific (AP) model, each comprising five production tiers, from raw material sourcing to final garment assembly. The results indicate that in the EU route, which involves facilities in Peru, Turkey, France, and Morocco, transport-related emissions are estimated to be equivalent to about 0.03 kg per shirt, assuming a TEU capacity of 112,000 units. In contrast, the AP route, which consolidates processing in Vietnam following raw material export from Peru, results in 950 kg of CO2 per TEU, or 0.00848 kg per shirt. This represents a 73.8 percent reduction in transport-related emissions compared to the EU configuration. To support this analysis, the model is trained on several years of empirical logistics and facility-level data sourced from the polo shirt industry. Key input variables include transport modes, regional energy mixes, and the emissions intensity of each production stage. Routing sequences are optimized under both operational and geographical constraints. The findings suggest that regional integration of manufacturing processes, combined with data-informed route planning, can significantly reduce indirect emissions in apparel supply chains. Moreover, the proposed methodological approach may be adapted to other multi-tier networks seeking to quantify and mitigate transport-related Scope 3 emissions.

• openaire https://doi.org/10.3390/proceedings2025131082first seen 2026-05-14 21:10:44