Scholarlink

The increasing demand for sustainable plastic packaging has led to a growing interest in optimizing machine parameters for processing recycled plastics. This paper presents a dataspace-driven demonstrator for machine learning (ML) lifecycle management in machine parameter optimization for thermoforming processes. Our setup involves three key dataspace participants: a thermoforming machine manufacturer, a plastic packaging producer, and an ML service provider. By leveraging dataspaces and a microservice-based architecture, we enable secure data exchange while addressing industry concerns about proprietary data sharing. The implemented demonstrator integrates the ML lifecycle in the form of microservices, facilitating efficient training, deployment, and monitoring of ML models. Our demonstrator highlights the potential of dataspaces in enabling collaborative, data-driven optimization of machine parameters while maintaining data sovereignty.

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

Machine Learning Lifecycle Management Using Dataspaces for Optimized Machine Parameterization in Recycled Plastic Packaging

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Alexander Nasuta,
  • Sylwia Olbrych,
  • Christoph Quix,
  • Tim Kaluza,
  • Florian Schaller,
  • Sabrina Steinert,
  • Hans A. Zhou,
  • Anas Abdelrazaq,
  • Robert H. Schmitt

摘要

The increasing demand for sustainable plastic packaging has led to a growing interest in optimizing machine parameters for processing recycled plastics. This paper presents a dataspace-driven demonstrator for machine learning (ML) lifecycle management in machine parameter optimization for thermoforming processes. Our setup involves three key dataspace participants: a thermoforming machine manufacturer, a plastic packaging producer, and an ML service provider. By leveraging dataspaces and a microservice-based architecture, we enable secure data exchange while addressing industry concerns about proprietary data sharing. The implemented demonstrator integrates the ML lifecycle in the form of microservices, facilitating efficient training, deployment, and monitoring of ML models. Our demonstrator highlights the potential of dataspaces in enabling collaborative, data-driven optimization of machine parameters while maintaining data sovereignty.