Motor Fault Diagnosis Based on Convolutional Block Attention Module-Xception Lightweight Neural Network.

Area of Science:

• Engineering
• Artificial Intelligence
• Automotive Technology

Background:

• Electric motors are critical for self-driving vehicle operation.
• Ensuring motor reliability through effective fault diagnosis is paramount for vehicle safety.

Purpose of the Study:

• To propose an improved motor fault diagnosis method utilizing vibration signals.
• To enhance the accuracy and efficiency of fault detection in electric motors for autonomous vehicles.

Main Methods:

• Vibration signals from motors across different operating states and frequencies were collected.
• Gram image coding transformed time-domain vibration data into grayscale images, highlighting fault features.
• A lightweight neural network, Xception, was enhanced with the Convolutional Block Attention Module (CBAM) for improved feature importance.

Main Results:

• The proposed method demonstrated superior recognition accuracy compared to traditional Convolutional Neural Network (CNN), ResNet, and standard Xception models.
• The integration of CBAM and Gram image coding resulted in faster iteration speeds without compromising computational complexity or accuracy.
• The enhanced Xception model effectively identified motor faults by focusing on critical characteristic information.

Conclusions:

• The developed motor fault diagnosis technique offers a significant advancement in detecting electric motor faults in autonomous vehicles.
• This method provides a more reliable and efficient solution for ensuring the safety and operational integrity of self-driving cars.
• The combination of Gram image coding, CBAM, and lightweight neural networks presents a promising direction for intelligent vehicle maintenance.