Advanced Bearing-Fault Diagnosis and Classification Using Mel-Scalograms and FOX-Optimized ANN

Area of Science:

• Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Bearing-fault diagnosis is crucial for industrial machinery efficiency and safety.
• Traditional methods may lack accuracy and reliability in detecting diverse faults.
• Advanced signal processing and machine learning are needed for robust fault detection.

Purpose of the Study:

• To propose a novel and highly accurate method for bearing-fault diagnosis.
• To leverage Mel-transformed scalograms and an autoencoder for feature extraction.
• To utilize the FOX optimizer for enhanced artificial neural network performance.

Main Methods:

• Vibrational signals (VS) were transformed into Mel-transformed scalograms.
• An autoencoder with convolutional and pooling layers extracted robust features.
• An artificial neural network (ANN) optimized with the FOX optimizer performed classification.

Main Results:

• The proposed model achieved perfect precision, recall, F1-scores, and AUC of 1.00.
• The FOX optimizer demonstrated superior performance over traditional backpropagation.
• t-SNE plots confirmed clear separability between different fault classes.

Conclusions:

• The novel method offers an efficient and accurate solution for bearing-fault diagnosis.
• This approach is highly reliable for real-time predictive maintenance in industrial settings.
• The findings significantly advance the field of machinery health monitoring.