A deep modeling approach based on time-frequency domain feature extraction
ID：163
Submission ID：27 View Protection：ATTENDEE
Updated Time：2025-11-03 11:39:25
Hits：62
Poster Presentation
Start Time：2025-11-09 09:03 （Asia/Shanghai）
Duration：1min
Session：[P] Poster presentation » [P6] 6.AI-driven technology
No files
Abstract
Aiming at the problem of difficulty in extracting fault features of wind turbines under complex operating conditions, this study introduces a method for identifying wind turbine bearing faults based on vibration signals, extracting statistical features in the time domain, then performing a Fast Fourier Transform (FFT) on the original signal, and extracting the frequency domain features as well as statistical features after the FFT. The main features in the time-frequency domain features are then selected using chi-square test. In this study, deep confidence neural network (DBN) is used to classify the bearing faults. Finally, a comparative study is carried out by comparing the classification results with those of Support Vector Machines (SVM) and Extreme Learning Machines (ELM), and the results show that the recognition accuracy of the method proposed in this study is 99.8%, which has a higher classification performance.
Keywords
Wind turbines; FFT; Feature Extractions; Deep Belief Network
Speaker
Comment submit