An Improved Evidential-IOWA Sensor Data Fusion Approach in Fault Diagnosis

Area of Science:

• Information Fusion
• Artificial Intelligence
• Fault Diagnosis

Background:

• Dempster-Shafer evidence theory is crucial for handling uncertain information in fault diagnosis.
• Conflicting evidence in traditional methods can lead to incorrect fault diagnosis and location.
• Existing approaches struggle with effectively managing highly conflicting sensor data.

Purpose of the Study:

• To propose an improved sensor data fusion approach for fault diagnosis using Dempster-Shafer evidence theory.
• To enhance the management of conflicting evidence in fault diagnosis scenarios.
• To improve the accuracy of fault location by optimizing sensor data fusion.

Main Methods:

• Developed an evidential-Induced Ordered Weighted Averaging (IOWA) operator within the Dempster-Shafer framework.
• Determined IOWA parameters (α value) considering both evidence distance and belief entropy.
• Utilized the maximum entropy method to generate a weight vector for sensor data sources.
• Modified sensor evidence using the IOWA operator before applying Dempster's combination rule.

Main Results:

• The proposed method demonstrates superior performance in managing conflicting evidence compared to traditional approaches.
• Incorporating the information volume of each evidence source improves fault diagnosis accuracy.
• Numerical examples and a case study validate the method's rationality and efficiency in fault diagnosis.

Conclusions:

• The improved IOWA-based Dempster-Shafer approach effectively addresses the limitations of conflicting evidence in fault diagnosis.
• This method offers a more robust and accurate solution for sensor data fusion in complex diagnostic systems.
• The enhanced conflict management capabilities lead to reliable fault location and system monitoring.