Multi-Sensor Data Fusion Using a Relevance Vector Machine Based on an Ant Colony for Gearbox Fault Detection

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Sensor reliability is crucial for condition-based maintenance in industrial processes.
• Gearbox vibration signals are often noisy, leading to unreliable fault detection from single sensors.
• Existing methods struggle with noise and feature redundancy in gearbox diagnostics.

Purpose of the Study:

• To develop an intelligent multi-sensor data fusion method for enhanced gearbox fault detection.
• To improve the accuracy and stability of fault diagnosis in rotating machinery.
• To address challenges posed by noisy vibration signals and feature complexity.

Main Methods:

• Proposed an Ant Colony Optimization-Relevance Vector Machine (ACO-RVM) for data fusion and fault detection.
• Utilized Ensemble Empirical Mode Decomposition (EEMD) for noise reduction in vibration signals.
• Employed Distance Evaluation Technique (DET) for dominant feature selection to reduce redundancy.

Main Results:

• The ACO-RVM demonstrated higher fault detection accuracy compared to standard Relevance Vector Machine (RVM) with cross-validation.
• The proposed method effectively preprocesses noisy vibration signals, improving diagnostic stability.
• Feature selection using DET enhanced the efficiency and accuracy of the fault detection model.

Conclusions:

• The intelligent multi-sensor data fusion method using ACO-RVM offers superior performance for gearbox fault detection.
• EEMD and DET significantly contribute to the robustness and accuracy of the diagnostic system.
• This approach provides a reliable solution for condition-based maintenance of critical machinery.