In-Wheel Motor Fault Diagnosis Method Based on Two-Stream 2DCNNs with DCBA Module

Area of Science:

• Engineering
• Artificial Intelligence
• Mechanical Systems

Background:

• Fault diagnosis in four-wheel independent driving systems is critical for safety and reliability.
• Variable driving conditions and harsh environments pose significant challenges to existing diagnostic methods.

Purpose of the Study:

• To propose a novel fault diagnosis method for in-wheel motors in electric vehicles.
• To enhance diagnostic accuracy and robustness under complex operating conditions.

Main Methods:

• A two-stream 2DCNN architecture was developed, integrating a novel DCBA module.
• The DCBA module extracts multi-scale features from vibration signals across different domains.
• Dual-domain learning utilizes both time-domain and time-frequency-domain signals for feature extraction.

Main Results:

• The proposed method achieved high recognition accuracy across various speeds, loads, and road surfaces.
• Comparative analysis showed superior performance and robustness against existing models like SENet, ECANet, and CBAM.
• The integration of DCBA and dual-domain learning effectively improved fault feature extraction.

Conclusions:

• The novel two-stream 2DCNN with DCBA module offers a robust and accurate solution for in-wheel motor fault diagnosis.
• This method demonstrates significant potential for application in real-world electric vehicle systems.
• Effective feature extraction under complex conditions is key to improving diagnostic performance.