Severity Estimation for Interturn Short-Circuit and Demagnetization Faults through Self-Attention Network

Area of Science:

• Electrical Engineering
• Machine Diagnostics
• Artificial Intelligence in Engineering

Background:

• Permanent-magnet synchronous machines (PMSMs) are crucial in modern applications.
• Interturn short-circuit faults (ISCF) and demagnetization faults (DF) degrade PMSM performance and reliability.
• Accurate and simultaneous diagnosis of these faults is challenging.

Purpose of the Study:

• To develop a novel fault diagnosis strategy for simultaneous ISCF and DF in PMSMs.
• To estimate the severity of these faults without relying on exact machine models or parameters.
• To validate the proposed strategy under various fault and load conditions.

Main Methods:

• A self-attention-based severity estimation network (SASEN) was developed.
• Positive- and negative-sequence voltage and current were used as input features.
• The SASEN utilizes a self-attention mechanism for enhanced feature extraction and regression.
• The network estimates fault indicators to quantify fault severities.

Main Results:

• The SASEN effectively diagnosed hybrid ISCF and DF occurring simultaneously.
• The proposed method accurately estimated fault severities under different load torques.
• Experimental validation confirmed the strategy's effectiveness and feasibility.

Conclusions:

• The SASEN provides a robust and model-independent approach for diagnosing hybrid faults in PMSMs.
• This strategy enhances the reliability and diagnostic capabilities for critical rotating machinery.
• The self-attention mechanism proves effective for complex fault severity estimation.