Answers are grounded in this paper and cite their sources.
Experimental evaluation demonstrates superior fault recognition using STFT and CNN-LSTM in generator equipment, indicating high practical utility for industrial fault diagnosis.
This study proposes a deep learning framework for condition recognition of generator wind-tunnel equipment. The vibration signals from the field generator are captured in real time, producing a time-frequency representation of the signal. In this experiment, the Short-Time Fourier Transform (STFT) is applied to the vibration signal of the experimental equipment for analysis. The remaining section of the framework provides a representation of spatial features in the time-frequency domain using a CNN. Simultaneously, it allows testing a hybrid CNN–LSTM network that captures temporal dependencies along the signal sequence. A comprehensive evaluation of the proposed framework will be performed using the widely recognized CWRU datasets. The proposed method achieves a classification accuracy of 94.64%, outperforming SVM (92.20%) and RF (88.99%). Moreover, the MFPT dataset has cross-dataset evaluation. As shown by cross-dataset evaluation on the MFPT dataset, SVM performance is below chance, whereas our method achieves 75% accuracy under unseen operating conditions. The findings suggest that STFT-based representations offer superior discriminatory power. The findings indicate that STFT-based representations have high discriminative power, and the given framework is an efficient and practical solution to the problem of fault diagnosis in industries.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: