Scholarlink

<p>Digital Twin has emerged as a transformative technology within the Industrial Internet of Things, addressing the pressing need for accurate, real-time fault detection in critical industrial applications. Leveraging Artificial Intelligence, Digital Twin enables near-real data generation from physical assets, enhancing fault detection capabilities in complex environments. This study addresses the challenge of reliable, efficient fault detection by proposing a novel intelligent diagnostic model for Digital Twin systems using a hybrid ensemble machine learning approach optimized through a genetic algorithm. The model combines Voting Ensemble Learning with genetic algorithm-based optimization to improve fault classification performance under real-time conditions. Experimental results reveal that our hybrid Voting-genetic algorithm-ensemble learning model achieves superior fault detection accuracy, with performance metrics reaching 88.2% and 98.2% for datasets A and B, respectively, outperforming traditional methods such as classification and regression tree and random forest, which obtained lower accuracy rates (69.5%, 77.8%, 75%, and 73%). Additionally, the proposed model demonstrates substantial improvements over other advanced models, including those that utilize feature selectionwith metaheuristic algorithms and genetic algorithm-tuned machine learning approach. This framework not only enhances diagnostic accuracy but also offers robust and scalable fault detection in cyber-physical and IoT-enabled industrial environments, establishing a new standard for fault diagnosis in Digital Twin applications.</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

Hybrid voting-GA ensemble learning for multi-class fault detection in digital twin-driven IIoT systems

  • 期刊论文

  • 发表日期：

    2025年1月18日

查看全文
  • Ezz El-Din Hemdan,
  • Samar M. Zayed,
  • Gamal Attiya,
  • Ayman El-Sayed,
  • Amged Sayed

摘要

Digital Twin has emerged as a transformative technology within the Industrial Internet of Things, addressing the pressing need for accurate, real-time fault detection in critical industrial applications. Leveraging Artificial Intelligence, Digital Twin enables near-real data generation from physical assets, enhancing fault detection capabilities in complex environments. This study addresses the challenge of reliable, efficient fault detection by proposing a novel intelligent diagnostic model for Digital Twin systems using a hybrid ensemble machine learning approach optimized through a genetic algorithm. The model combines Voting Ensemble Learning with genetic algorithm-based optimization to improve fault classification performance under real-time conditions. Experimental results reveal that our hybrid Voting-genetic algorithm-ensemble learning model achieves superior fault detection accuracy, with performance metrics reaching 88.2% and 98.2% for datasets A and B, respectively, outperforming traditional methods such as classification and regression tree and random forest, which obtained lower accuracy rates (69.5%, 77.8%, 75%, and 73%). Additionally, the proposed model demonstrates substantial improvements over other advanced models, including those that utilize feature selectionwith metaheuristic algorithms and genetic algorithm-tuned machine learning approach. This framework not only enhances diagnostic accuracy but also offers robust and scalable fault detection in cyber-physical and IoT-enabled industrial environments, establishing a new standard for fault diagnosis in Digital Twin applications.