Scholarlink

This paper presents a federated learning-based framework for fault detection in 3D bioprinting, with DenseNet as the proposed deep learning model for classification. The framework is designed to address data privacy concerns in decentralized environments, enabling local model training on clients’ datasets and aggregating updates through the FedAvg algorithm. To evaluate the model’s performance, we compare DenseNet with three other architectures—Multi-Layer Perceptron (MLP), Fully Connected Neural Network (FCNN), and Long Short-Term Memory (LSTM)—in detecting various quality defects in 3D printed objects, such as under-extrusion, nozzle clogging, and material collapse. The federated learning approach is implemented using the Flower framework, ensuring scalability and efficient communication across heterogeneous devices. Experimental results demonstrate that DenseNet outperforms MLP, FCNN, and LSTM in classification accuracy, achieving 96.00% accuracy after 20 communication rounds. In contrast, MLP, FCNN, and LSTM achieve 80.00%, 75.00%, and 70.00% accuracy, respectively. These results highlight the potential of DenseNet for improving fault detection in 3D printing processes while maintaining data privacy. The proposed approach offers promising applications for real-time quality control in 3D bioprinting and other manufacturing sectors.

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

DenseNet-Driven Federated Learning for Enhanced Fault Detection in 3D Bioprinting

  • 章节

  • 发表日期：

    2026年

查看全文
  • V. Ramya,
  • N. Vinothini,
  • S. Levin Rohith,
  • G. Ananthi

摘要

This paper presents a federated learning-based framework for fault detection in 3D bioprinting, with DenseNet as the proposed deep learning model for classification. The framework is designed to address data privacy concerns in decentralized environments, enabling local model training on clients’ datasets and aggregating updates through the FedAvg algorithm. To evaluate the model’s performance, we compare DenseNet with three other architectures—Multi-Layer Perceptron (MLP), Fully Connected Neural Network (FCNN), and Long Short-Term Memory (LSTM)—in detecting various quality defects in 3D printed objects, such as under-extrusion, nozzle clogging, and material collapse. The federated learning approach is implemented using the Flower framework, ensuring scalability and efficient communication across heterogeneous devices. Experimental results demonstrate that DenseNet outperforms MLP, FCNN, and LSTM in classification accuracy, achieving 96.00% accuracy after 20 communication rounds. In contrast, MLP, FCNN, and LSTM achieve 80.00%, 75.00%, and 70.00% accuracy, respectively. These results highlight the potential of DenseNet for improving fault detection in 3D printing processes while maintaining data privacy. The proposed approach offers promising applications for real-time quality control in 3D bioprinting and other manufacturing sectors.