Fault Diagnosis of Wind Turbine Based on Convolution Neural Network Algorithm.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Traditional fan fault diagnosis methods struggle with efficiency and intelligent monitoring.
• Existing intelligent diagnosis methods for bearing faults based on vibration signals have limitations.
• Need for improved accuracy and real-time capabilities in fan operation monitoring.

Purpose of the Study:

• To design and validate a deep convolution neural network (CNN) model for intelligent fan bearing fault diagnosis.
• To improve the accuracy and efficiency of automatic inspection and intelligent operation monitoring of fans.
• To compare the effectiveness of different vibration signal transformation methods on diagnostic accuracy.

Main Methods:

• Development of a deep CNN model with three convolution-pooling layers and two fully connected layers.
• Utilized vibration signal analysis, including gray maps, Short-Time Fourier Transform (STFT), and Continuous Wavelet Transform (CWT).
• Experimental verification using a public dataset to assess diagnostic model accuracy.

Main Results:

• The proposed deep CNN model achieved diagnostic accuracy close to 100%.
• Comparison of signal transformation methods showed varying impacts on model accuracy.
• Deep CNN methods outperformed traditional machine learning algorithms based on time-domain statistical features in real-time recognition.

Conclusions:

• The deep CNN model effectively diagnoses fan bearing faults with high accuracy.
• Signal transformation techniques significantly influence the diagnostic performance.
• The study provides a robust method for intelligent fan operation monitoring and fault detection, offering new insights for wind energy verification.