Digital Twins Model of Industrial Product Management and Control Based on Lightweight Deep Learning

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Digital twins (DTs) integrate physical and digital information, offering simulation and virtual reality (VR) capabilities.
• DTs are increasingly applied in industrial product management and control due to their versatile functions across product lifecycle stages.

Purpose of the Study:

• To explore the application of digital twins in industrial product management and control.
• To validate the effectiveness of DTs through comparative experiments and develop an intelligent defect detection algorithm.

Main Methods:

• Conceptualization and functional analysis of DTs in product management.
• Application of Workench simulation and SolidWorks for aluminum alloy flange design using DTs.
• Development of a lightweight deep learning algorithm for ultrasonic defect identification.

Main Results:

• A lightweight convolutional neural network (CNN) achieved 94.1% accuracy in ultrasonic defect detection with a 2.9 MB model size.
• Nonlinear finite element analysis results were consistent with experimental test results, confirming method reliability.
• DTs provide a viable technical solution for product management and control within 3D models.

Conclusions:

• Digital twins offer a reliable and efficient approach to industrial product management and control.
• The developed lightweight deep learning model significantly improves ultrasonic defect detection efficiency.
• The study confirms the reliability of finite element analysis and the potential of DTs to shorten design cycles.