An Augmented Reality-Assisted Prognostics and Health Management System Based on Deep Learning for IoT-Enabled...

Area of Science:

• Manufacturing Engineering
• Computer Science
• Artificial Intelligence

Background:

• Increasingly complex workshop equipment due to advanced Internet of Things (IoT) technology leads to higher rates of performance degradation and faults.
• The complexity challenges the construction of Remaining Useful Life (RUL) models and outpaces maintenance personnel's ability to keep up with equipment replacement.

Purpose of the Study:

• To propose an augmented reality (AR)-assisted prognostics and health management system for IoT-enabled manufacturing.
• To improve the accuracy of RUL prediction and enhance the efficiency of maintenance operations.

Main Methods:

• Feature extraction using a Convolutional Neural Network-Particle Swarm Optimization (PSO-CNN) model to analyze production data.
• High-accuracy RUL prediction via Gate Recurrent Unit (GRU)-attention to capture time-series dependencies and address gradient disappearance.
• Integration of AR for efficient maintenance guidance and remote expert assistance.

Main Results:

• The PSO-CNN model effectively excavates internal associations within large production datasets.
• The GRU-attention mechanism achieves high-accuracy RUL prediction by capturing long-term and short-term dependencies.
• The integrated AR system provides efficient, visible maintenance instructions and facilitates remote expert support.

Conclusions:

• The proposed AR-assisted deep learning system significantly enhances RUL prediction accuracy and maintenance efficiency in IoT-enabled manufacturing workshops.
• The system effectively addresses the challenges posed by complex equipment and the need for rapid maintenance support.
• Validation in a real-world IoT workshop confirms the practical effectiveness of the developed approach.