Using a Support Vector Machine Based Decision Stage to Improve the Fault Diagnosis on Gearboxes

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Gearboxes are critical mechanical components in applications like automotive transmissions.
• Gearbox malfunctions can lead to significant economic losses and safety hazards.
• Deep learning offers powerful solutions for gearbox fault diagnosis but requires substantial data and computational resources.

Purpose of the Study:

• To reduce the training time of deep learning-based fault diagnosis systems for gearboxes.
• To maintain or improve the accuracy of fault diagnosis despite reduced training time.
• To address the challenge of training deep learning models when high-performance GPUs are unavailable.

Main Methods:

• Implementation of a decision stage to interpret probability outputs from a classifier.
• Utilizing a classifier with a softmax activation function in its output layer.
• Application of two distinct classification algorithms for the decision-making process.

Main Results:

• Achieved a reduction in training time by approximately 80%.
• Maintained the average accuracy of the fault diagnosis system.
• Demonstrated the feasibility of efficient deep learning model training.

Conclusions:

• The proposed decision stage effectively reduces deep learning model training time for gearbox fault diagnosis.
• The method offers a practical solution for scenarios with limited computational resources.
• This approach enhances the accessibility and efficiency of AI-driven gearbox monitoring systems.