Scholarlink

<p>Components manufactured through continuous carbon fiber-reinforced nylon additive manufacturing (CCF-RNAM) offer many advantages and have been widely adopted in high-end manufacturing. However, the current processes still exhibit shortcomings, making them susceptible to unpredictable anomalies. Despite having monitoring capabilities, current CCF-RNAM systems are limited in detecting process anomalies, adversely affecting component quality. This study aims to integrate digital twin (DT) models into CCF-RNAM. For this purpose, we deployed a DT model onto a CCF-RNAM device to validate process parameters and detect anomalies. To effectively identify anomalies during manufacturing, we introduced a multi-model deep learning algorithm combining Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Transformer architectures into the DT model. The results demonstrate that the proposed model accurately identifies anomaly sources during manufacturing (achieving an F1 score of approximately 0.90) and effectively validates process parameters in a fully virtual environment. This approach minimizes unnecessary waste of time and materials, thus providing robust support for controllable CCF-RNAM processes.</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

Integrating digital twin models into continuous carbon fiber-reinforced nylon additive manufacturing for process parameters verification and anomaly detection

  • 期刊论文

  • 发表日期：

    2025年6月2日

查看全文
  • Xiao Wei,
  • Yujun Wang,
  • Jian Mao,
  • Man Zhao,
  • Gang Liu

摘要

Components manufactured through continuous carbon fiber-reinforced nylon additive manufacturing (CCF-RNAM) offer many advantages and have been widely adopted in high-end manufacturing. However, the current processes still exhibit shortcomings, making them susceptible to unpredictable anomalies. Despite having monitoring capabilities, current CCF-RNAM systems are limited in detecting process anomalies, adversely affecting component quality. This study aims to integrate digital twin (DT) models into CCF-RNAM. For this purpose, we deployed a DT model onto a CCF-RNAM device to validate process parameters and detect anomalies. To effectively identify anomalies during manufacturing, we introduced a multi-model deep learning algorithm combining Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Transformer architectures into the DT model. The results demonstrate that the proposed model accurately identifies anomaly sources during manufacturing (achieving an F1 score of approximately 0.90) and effectively validates process parameters in a fully virtual environment. This approach minimizes unnecessary waste of time and materials, thus providing robust support for controllable CCF-RNAM processes.