In-Wheel Motor Fault Diagnosis Using Affinity Propagation Minimum-Distance Discriminant Projection and...

Area of Science:

• Engineering
• Artificial Intelligence
• Automotive Systems

Background:

• Electric vehicles (EVs) with in-wheel motors require robust operational safety.
• In-wheel motor faults can compromise vehicle performance and safety.
• Existing fault diagnosis methods may lack accuracy and robustness.

Purpose of the Study:

• To propose a novel, effective fault diagnosis method for EV in-wheel motors.
• To enhance the reliability and safety of electric vehicles.
• To improve the accuracy and robustness of fault detection systems.

Main Methods:

• Developed Affinity Propagation-Minimum Distance Discriminant Projection (AP-MDP) for dimension reduction.
• Introduced a multi-class Support Vector Data Description (SVDD) classifier with a Weibull kernel function.
• Modified the SVDD classification rule using the minimum distance from the intra-class cluster center.
• Collected vibration signals from in-wheel motors with bearing faults under various operating conditions.

Main Results:

• AP-MDP demonstrated superior performance over traditional dimension reduction techniques (LDA, MDP, LPP), improving divisibility by at least 8.35%.
• The Weibull kernel-based multi-class SVDD achieved high classification accuracy (over 95%) and robustness for in-wheel motor faults.
• The proposed method outperformed classifiers using polynomial and Gaussian kernels.

Conclusions:

• The novel AP-MDP and Weibull kernel SVDD method offers an effective solution for in-wheel motor fault diagnosis in EVs.
• The enhanced diagnosis system significantly improves operational safety and reliability.
• This approach provides a robust and accurate fault detection mechanism for electric vehicle powertrains.