Scholarlink

Product portfolio management recently faces pressure from various stakeholder in the course of the sustainability shift. Consequently, the complexity of decision-making continues to increase, making the integration of sustainability into portfolio management essential to support a producing company’s sustainability goals. However, performing sustainability assessments to support the integration is time-consuming and thus not practical for multi-variant product portfolios. Therefore, this paper aims to develop a scalable method for carbon footprinting using an artificial neural network. The paper is orientated on the Crisp-DM process to understand and prepare the data, select the input and output layers and to design the network architecture. The network uses product characteristics as input and provides an estimated global warming potential. The model is trained, validated and tested with data from environmental product declarations from ABB open library. The work serves as feasibility study for further research in this field.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fast Carbon Footprinting in Complex Product Portfolios Using Machine Learning Approaches

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nikolai Kelbel,
  • Peiran Yang,
  • Michael Riesener,
  • Alexander Keuper,
  • Günther Schuh

摘要

Product portfolio management recently faces pressure from various stakeholder in the course of the sustainability shift. Consequently, the complexity of decision-making continues to increase, making the integration of sustainability into portfolio management essential to support a producing company’s sustainability goals. However, performing sustainability assessments to support the integration is time-consuming and thus not practical for multi-variant product portfolios. Therefore, this paper aims to develop a scalable method for carbon footprinting using an artificial neural network. The paper is orientated on the Crisp-DM process to understand and prepare the data, select the input and output layers and to design the network architecture. The network uses product characteristics as input and provides an estimated global warming potential. The model is trained, validated and tested with data from environmental product declarations from ABB open library. The work serves as feasibility study for further research in this field.