Modeling Sensor Reliability in Fault Diagnosis Based on Evidence Theory

Area of Science:

• Engineering
• Computer Science

Background:

• Sensor data fusion is critical for accurate fault diagnosis.
• Dempster-Shafer (D-R) evidence theory effectively combines sensor data but struggles with highly conflicting evidence, leading to counterintuitive results.

Purpose of the Study:

• To develop an improved sensor data fusion method that addresses the limitations of existing techniques, particularly in managing conflicting evidence.
• To enhance fault diagnosis accuracy by incorporating dynamic sensor reliability and evidence conflict management.

Main Methods:

• A new method is proposed that considers both static and dynamic sensor reliability.
• Dynamic sensor reliability is calculated using an evidence distance function and belief entropy.
• A weighted averaging approach modifies conflicting evidence based on sensor reliability and information volume.

Main Results:

• The proposed method demonstrates superior performance in managing conflicting sensor evidence.
• Fault diagnosis accuracy is improved from 81.19% to 89.48% compared to existing methods.
• The approach effectively utilizes the information volume of each sensor report.

Conclusions:

• The novel sensor data fusion method effectively handles conflicting evidence and improves fault diagnosis accuracy.
• Incorporating dynamic sensor reliability and evidence conflict management leads to more robust fault detection.
• The method shows significant promise for practical applications in industrial fault diagnosis.