An Unsupervised Deep Feature Learning Model Based on Parallel Convolutional Autoencoder for Intelligent Fault...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Traditional fault diagnosis relies on manual feature extraction, which is subjective and knowledge-dependent.
• This limits the effectiveness and efficiency of diagnostic frameworks.
• Automating feature generation is crucial for robust machinery diagnostics.

Purpose of the Study:

• To propose an unsupervised deep feature learning model for automated feature generation in diagnostic frameworks.
• To enhance the robustness and accuracy of fault diagnosis systems.
• To address the limitations of manual feature extraction in traditional methods.

Main Methods:

• A parallel convolutional autoencoder (PCAE) model was developed for unsupervised deep feature learning.
• Raw vibration signals were normalized and segmented, with features extracted from time-domain and time-frequency domain data.
• Deep neural networks (DNN) with softmax were used for fault classification, incorporating dropout regularization and batch normalization.

Main Results:

• The proposed PCAE-based framework demonstrated enhanced robustness in fault diagnosis.
• Identification accuracy of operational conditions was improved by approximately 8% compared to traditional methods.
• The model effectively automated feature generation, reducing reliance on manual expertise.

Conclusions:

• The diagnostic framework integrating parallel unsupervised feature learning and deep classification structures is highly effective.
• This approach significantly enhances machinery diagnostic accuracy and robustness.
• The study validates the potential of deep learning for advanced fault diagnosis applications.