Digital Twin-Driven Rear Axle Assembly Torque Prediction and Online Control

Area of Science:

• Manufacturing Engineering
• Artificial Intelligence
• Control Systems

Background:

• Accumulated errors in rear axle assembly torque reduce assembly quality and efficiency.
• Existing methods struggle with the non-stationarity and nonlinearity of assembly torque.

Purpose of the Study:

• To develop an online control method for rear axle assembly torque using digital twin technology.
• To enhance assembly quality and efficiency in intelligent production lines.

Main Methods:

• Proposed a Gray Wolf Optimizer-based Variational Modal Decomposition and Long Short-Term Memory network (GWO-VMD-LSTM) algorithm for assembly torque prediction.
• Developed a digital twin system for visualization and monitoring of the rear axle assembly line.
• Coupled the GWO-VMD-LSTM model with the digital twin system for real-time prediction and online control.

Main Results:

• The GWO-VMD-LSTM algorithm achieved a prediction accuracy of 99.49% for assembly torque.
• The digital twin system enabled real-time monitoring and visualization of the assembly process.
• The integrated system demonstrated a response time of approximately 1 second.

Conclusions:

• The digital twin-based online control method significantly improves rear axle assembly quality and efficiency.
• This approach is crucial for advancing intelligent production lines and addressing assembly torque challenges.