Hub bolt failures in wind turbines can lead to severe equipment damage and significant economic losses due to downtime and maintenance costs. In this study, a classification method named CAMPPlus is proposed. This method introduces an innovative multi-scale feature fusion strategy that integrates Mel-spectrogram, MFCC, and STFT features through weighted feature fusion. The method is validated on the UrbanSound8K dataset, achieving a classification accuracy of 96.2%, with a 95% confidence interval of [96.1%, 96.8%], demonstrating strong robustness in noisy environments. Subsequently, the model is applied to a custom measurement dataset containing hub noise and bolt failure sounds collected from operational wind turbines. On this dataset, the model achieves a classification accuracy of 99.7%, with a 95% confidence interval of [99.5%, 99.8%]. Furthermore, the method’s high classification performance is validated under varying environmental noise conditions. By enabling early fault detection in wind turbines, this method contributes to reducing downtime and maintenance costs while improving operational efficiency. The proposed solution not only enhances the reliability and cost-effectiveness of wind turbine maintenance but also provides a scalable approach for continuous health monitoring, thereby promoting reliable and sustainable wind turbine operation in industrial settings.
AiOCHariharanMYaacobS, et al. (2012) Classification of speech dysfluencies with MFCC and LPCC features. Expert Systems with Applications39(2): 2157–2165.
2.
AslanAÇelebiSB (2024) Improving forecasting performance with random forest regression model in wind power generation. In: 6th International Conference on Applied Engineering and Natural Sciences (ICAENS), April 2024,Sarajevo, Bosnia and Herzegovina.
3.
ChikkerurSCartwrightANGovindarajuV (2007) Fingerprint enhancement using STFT analysis. Pattern Recognition40(1): 198–211.
4.
ChuYWangQZhouE, et al. (2023) Hybrid spectrogram for the automatic respiratory sound classification with group time frequency attention network. In: 2023 IEEE 6th International Conference on Pattern Recognition and Artificial Intelligence (PRAI), August 2023,Xi’an, China. IEEE.
5.
EralievOLeeKHLeeCH (2022) Vibration-based loosening detection of a multi-bolt structure using machine learning algorithms. Sensors22(3): 1210.
6.
FengKJiJCLiY, et al. (2022) A novel cyclic-correntropy based indicator for gear wear monitoring. Tribology International171: 107528.
7.
FengKXiaoHZhangJ, et al. (2025) A digital twin methodology for vibration-based monitoring and prediction of gear wear. Wear571: 205806.
8.
GaoGGuoYZhouL, et al. (2024) Res2Net-based multi-scale and multi-attention model for traffic scene image classification. PLoS One19(5): e0300017.
9.
GreenwoodDD (1997) The Mel scale’s disqualifying bias and a consistency of pitch-difference equisections in 1956 with equal cochlear distances and equal frequency ratios. Hearing Research103(1–2): 199–224.
10.
IttichaichareonCSuksriSYingthawornsukT (2012) Speech recognition using MFCC. International Conference on Computer Graphics, Simulation and Modeling, July 20129.
11.
JiangSZhangJWangW, et al. (2023) Automatic inspection of bridge bolts using unmanned aerial vision and adaptive scale unification-based deep learning. Remote Sensing15(2): 328.
12.
JuradoFSaenzJR (2002) Comparison between discrete STFT and wavelets for the analysis of power quality events. Electric Power Systems Research62(3): 183–190.
13.
KumarRAnandRS (2024) Machine learning approach with multiple feature selection techniques to diagnose the inter-turn winding faults in induction motor. Arabian Journal for Science and Engineering50: 1–17.
14.
KuraniAVakhariaAShahM, et al. (2023) A comprehensive comparative study of artificial neural network (ANN) and support vector machines (SVM) on stock forecasting. Annals of Data Science10(1): 183–208.
15.
LeeHHanHPettitC, et al. (2024) Machine learning approach to residential valuation: a convolutional neural network model for geographic variation. The Annals of Regional Science72(2): 579–599.
16.
LiHWangZ (2025) Anomaly identification of wind turbine blades based on Mel-spectrogram difference feature of aerodynamic noise. Measurement240: 115428.
17.
LiSJiJFengK, et al. (2024) Composite neuro-fuzzy system-guided cross-modal zero-sample diagnostic framework using multi-source heterogeneous non-contact sensing data. IEEE Transactions on Fuzzy SystemsPP(99): 1–12.
18.
LiHXiYChenH, et al. (2025) Anomaly detection of aerodynamic noise from wind turbine blades based on the CNN–LSTM–AS–UNet model. Journal of Civil Structural Health Monitoring1–16.
19.
LiuFGaoSTianZ, et al. (2020) A new time-frequency analysis method based on single mode function decomposition for offshore wind turbines. Marine Structures72: 102782.
20.
LiuSSongZHeL (2024) Improving ecapa-tdnn performance with coordinate attention. Journal of Shanghai Jiaotong University1–7.
21.
LummenDGruberSSchmidtA, et al. (2024) Opcode-based ransomware detection using hybrid extreme gradient boosting and recurrent neural networks. Authorea Preprints.
22.
OzturkSFthenakisV (2020) Predicting frequency, time-to-repair and costs of wind turbine failures. Energies13(5): 1149.
23.
PandeyAWangD (2019) TCNN: temporal convolutional neural network for real-time speech enhancement in the time domain. In: ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), May 12–17,2019, Brighton, UK. IEEE.
24.
RagabMKatibISharafS, et al. (2024) Automated brain tumor recognition using equilibrium optimizer with deep learning approach on MRI images. Scientific Reports14(1): 29448.
25.
RaoASRadanovicMLiuY, et al. (2022) Real-time monitoring of construction sites: sensors, methods, and applications. Automation in Construction136: 104099.
26.
SalamonJBelloJP (2017) Deep convolutional neural networks and data augmentation for environmental sound classification. IEEE Signal Processing Letters24(3): 279–283.
27.
SharmaGUmapathyKKrishnanS (2020) Trends in audio signal feature extraction methods. Applied Acoustics158: 107020.
28.
SongQZhangYChenX, et al. (2024) Real-time ubiquitous radar target classification with 1D ResNet-SE-based multi-channel network. Remote Sensing16(21): 3986.
29.
StevensSSVolkmannJ (1940) The relation of pitch to frequency: a revised scale. American Journal of Psychology53(3): 329–353.
30.
SumanAKumarCSumanP (2022) Early detection of mechanical malfunctions in vehicles using sound signal processing. Applied Acoustics188: 108578.
31.
TariqZShahSKLeeY (2022) Feature-based fusion using CNN for lung and heart sound classification. Sensors22(4): 1521.
32.
TranTLundgrenJ (2020) Drill fault diagnosis based on the scalogram and Mel spectrogram of sound signals using artificial intelligence. IEEE Access8: 203655–203666.
33.
WangYTanJQinX, et al. (2021) Windturbine blade fault diagnosis based on MFCC feature optimization strategy. IOP Conference Series: Earth and Environmental Science675(1): 012074.
34.
ZhangJJiangQChangF (2018) Fault diagnosis method based on MFCC fusion and SVM. In: Proceedings of the 2018 IEEE International Conference on Information and Automation (ICIA), July 6–8, 2018,Beijing, China. IEEE.
35.
ZhangEWuYTangZ (2024) SC-EcapaTdnn: ECAPA-TDNN with separable convolutional for speaker recognition. In: International Conference on Intelligence Science. Cham. Springer Nature Switzerland.
36.
ZhaoXWangLZhangY, et al. (2024) A review of convolutional neural networks in computer vision. Artificial Intelligence Review57(4): 99.
