Scholarlink

<p>As a sustainable mass individualization paradigm, social manufacturing integrates a variety of distributed resources to meet diversified consumers’ demands through socialized collaboration. However, disturbances and disruptions in social manufacturing processes may result in delays in delivering individualized products. This study tries to predict the delivery capacity of a social manufacturing network. Firstly, we constructed a delivery capability prediction framework for social manufacturing orders.&#xa0;Secondly, based on the production data collected from each social manufacturing node, the order data is localized and filtered using a processing time classification algorithm to generate a dataset. Thirdly, an integrated public file system is used to deal with the secure transmission of private file data; the federated mini-batch gradient descent algorithm is used for distributed data training to predict their delivery capacity. The prediction results demonstrate that the accuracy of predicting the delivery capacity of social manufacturing orders has increased to over 90%. This study is expected to provide theoretical insights to enhance the resilience of social manufacturing network under disturbances and disruptions.</p>

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

Blockchained on-device federated learning-based delivery capacity prediction in a social manufacturing paradigm

  • 期刊论文

  • 发表日期：

    2025年4月16日

查看全文
  • Fuqiang Zhang,
  • Haojie Wang,
  • Xueliang Zhou,
  • Hui Mu,
  • Jiewu Leng

摘要

As a sustainable mass individualization paradigm, social manufacturing integrates a variety of distributed resources to meet diversified consumers’ demands through socialized collaboration. However, disturbances and disruptions in social manufacturing processes may result in delays in delivering individualized products. This study tries to predict the delivery capacity of a social manufacturing network. Firstly, we constructed a delivery capability prediction framework for social manufacturing orders. Secondly, based on the production data collected from each social manufacturing node, the order data is localized and filtered using a processing time classification algorithm to generate a dataset. Thirdly, an integrated public file system is used to deal with the secure transmission of private file data; the federated mini-batch gradient descent algorithm is used for distributed data training to predict their delivery capacity. The prediction results demonstrate that the accuracy of predicting the delivery capacity of social manufacturing orders has increased to over 90%. This study is expected to provide theoretical insights to enhance the resilience of social manufacturing network under disturbances and disruptions.