An integrated monitoring scheme for wind turbine main bearing using acoustic emission
Introduction
With the rise of green energy, wind power has become the fastest developing renewable energy in the world. To achieve Net Zero Emissions by 2050, the annual installed capacity for WTs will be steeper, which is three times the power generation capacity in 2020 [1]. Generally, there are two types of WT in the market, namely the gearbox driven WT and DDWT. Compared with the former WT, DDWT eliminates gearbox failure and transmission losses. Therefore, the DDWT is more effective for higher power ratings, and it potentially becomes the dominant technology [2].
While the rapid development of wind power has brought huge opportunities for the wind power industry, it also presents higher requirements for the reliability of all components of WT [3]. In particular, the MB in DDWT is a critical component supporting heavy loads and handling large moments. As most wind farms are located in remote areas with harsh natural environments, the MB is prone to failure, which can result in unnecessary downtime, expensive maintenance, and even catastrophic accidents [4]. Thus, the condition monitoring of MB is vitally important to guarantee the safe and stable operation of DDWT.
Over the years, various approaches have been adopted to monitor the operating conditions of bearings and have achieved satisfactory results, especially with increasing usage and attention to vibration signals [5,6]. They ranged from the time domain to frequency domain to time-frequency domain, including variational mode decomposition [7], Kurtogram and its improved methods [8,9], wavelet and wavelet package transform [10,11], blind deconvolution [12,13], and cyclostationary approach [14,15], etc.
Nevertheless, different from traditional industrial bearings, the health monitoring of MB still faces many challenges. Firstly, the rotating speeds of MB are very slow (less than 20 rpm), which means that the fault features generated by collision are particularly slight because of low kinetic energy and are easily overwhelmed by noisy interferences [16]. Thus, it is difficult to detect weak faults by vibration signals. Secondly, accurate speed information is the core of periodic feature extraction and enhancement. Although the SCADA system collects the operating information of DDWT, the rotating speed is acquired in an intermittent manner (typically several seconds per sample), thus it is not synchronous with the high-frequency condition monitoring signal [17]. If the SCADA information is directly used for the health monitoring of MB, it will affect the accuracy and reliability of diagnosis results. Therefore, determining how to accurately obtain the rotational speed that synchronized with the monitoring signal is still a challenging problem. Finally, since there is no gearbox in DDWT, a generator with multiple magnetic poles is required to achieve a sufficiently high output frequency [18]. Although such an arrangement can simplify the mechanical structure, a side-effect is that weak damage signatures of MB are often overwhelmed by strong electromagnetic interference (EMI) from the alternating magnetic field. For this reason, it further increases the difficulty of monitoring and may result in rather poor monitoring performance. Therefore, it is urgent to explore an effective health monitoring scheme for the MB.
Recent studies have shown that AE technology is quite sensitive to low-speed bearing fault detection [19,20]. Compared with the vibration signal, the AE signals primarily reflect the rapid release of strain energy in solids material, which presents high sensitivity to identify early bearing defects and monitor the evolution of the bearing fatigue [21]. Furthermore, AE signals have the potential for detecting damage source localization in large-size bearings [22]. Accordingly, many researchers had adopted AE signals to monitor the operating conditions of the low-speed bearing. In [23], the AE energy was proven to be reliable and sensitive for detecting incipient cracks in low-speed bearing. Elforjani and Mba [24] utilized AE signals to diagnose low-speed bearings. Experimental cases indicated that AE signals can be utilized to monitor crack initiation and propagation. Additionally, Fuentes et al. [25] constructed a probabilistic model using AE signals to detect bearing sub-surface damage. But for MB operating under harsh working conditions, the AE signals have the features of low SNR and multi-modulation, which result in challenges to the incipient fault detection of MB.
The narrowband demodulation techniques have been considered as an efficient tool for bearing fault detection [26]. To determine the narrowband with rich failure information, Antoni [8] proposed the Kurtogram, which assumed the FB with maximum Kurtosis as the optimal one for demodulation. Whereafter, a series of Kurtogram-based methods had been constructed to improve the selection strategy of FB [27,28]. Barszcz and Jabłoński [29] pointed out that the kurtosis-guided strategy is susceptible to non-Gaussian noise and random impulses. To tackle this issue, they proposed the Protrugram for bearing diagnosis, where the optimal FB was selected according to the Kurtosis of envelope spectrum rather than the Kurtosis of the filtered signal. Furthermore, Infogram and Autogram were constructed to mitigate the influence of fault-irrelevant impulsive disturbances [30,31]. However, as the AE signals from EMI contamination contains similar and even stronger impulsive characteristics with bearing defects, the aforementioned demodulation methods tend to enhance EMI rather than fault symptoms.
Furthermore, the research on the cyclostationarity of AE signals had been widely employed in bearing fault diagnosis [32,33]. In [34], Spectral Coherence (SCoh) was proposed for bearing damage detection, where the signals can be described from two perspectives, including spectral frequency, which contains the natural frequency components, and cyclic frequency, which contains the modulation frequency components. When the SCoh is integrated along with the spectral frequency, the resulting spectrum of the demodulated signal can be used to detect bearing faults. To suppress noise interference, the improved envelope spectrum (IES) integrated along the spectral FB is proposed to enhance fault features [35]. In the following, the Integrated Spectral Coherence (ISC) was presented to monitor bearing failures under strong EMI and other impulsive disturbances [36]. Similar to Kurtogram, Mauricio et al. [37] constructed an adaptive FB selection mechanism for gearbox bearing diagnostics under time-varying operating conditions. Additionally, to improve the robustness to non-Gaussian impulses, the Generalized Spectral Coherence (GSC) was developed for the health monitoring of bearings operating in crushing machines [38]. Although these works had achieved excellent performance in laboratory environments, the AE signals in real applications are contaminated by strong noise and multi-interference, and their feature enhancement is still a challenging issue.
Additionally, it is necessary to mention that, in practical wind power plants, regular inspections are required for large and expensive equipment such as DDWT. The inspectors often use the endoscope to detect the location of bearing damage. However, the MBs in DDWT are relatively large (diameter greater than 2 m). Detecting the whole bearing is time-consuming and inconvenient in a small nacelle. Consequently, how to locate the damage of MBs based on AE signals is another issue to be addressed.
To address the aforementioned limitations, an integrated methodology based on AE signals is proposed for the health monitoring of MBs. In this paper, the HFEA is first constructed to accurately estimate the operating speed of the MB, which solves the problems caused by asynchronous collection. On this basis, the ASC is then established to adaptively distinguish the optimal FB containing rich failure symptoms. Lastly, the damage localization scheme is developed to improve the detection efficiency of large-size MBs. With the proposed method, the MB could be effectively monitored even under heavy noise and complex interference.
The innovations of this study are mainly summarized as follows:
Accurate rotational speed estimation: As previously mentioned, the rotating speed collected from SCADA and the monitoring information are often asynchronous. Adding an extra tachometer not only increases monitoring costs but also introduces instability. Accordingly, the proposed method directly estimates rotating speed from the AE signals itself, which not only realizes synchronous collection but also reduces the monitoring costs.
Adaptive narrowband selection scheme: On account of the weak fault response and strong interferences, traditional methods are difficult to accurately distinguish fault modulation sources. To address the limitation, the constructed ASC targets the fault source and determines the optimal FB through an adaptive bandwidth selection scheme, which provides a novel solution for narrowband demodulation.
Effective damage localization strategy: This work established an effective damage localization model to greatly improve the health management efficiency of the MBs in wind farms.
In the following, this work is arranged as follows. Section 2 reviews the typical SCoh-based bearing diagnosis methods. Subsequently, the proposed integrated monitoring scheme is elaborated in Section 3.
Next, two industrial cases in Section 4 are employed to validate the effectiveness of the proposed method. At last, Some conclusions are presented in Section 5.
Section snippets
Review of typical spectral coherence for bearing diagnosis
Some studies have also shown that the AE signals of defective bearing are second-order cyclostationary (CS2) [32]. To extract these CS2 components, the SCoh-based methods have been developed for bearing fault diagnosis.
Integrated monitoring methodology
For tackling the challenges mentioned in the above discussion, an integrated methodology based on AE is constructed for the condition monitoring of MB. In this part, to improve diagnostic accuracy, the rotating speed is first estimated by the HFEA. Subsequently, an adaptive FB selection mechanism named ASC is constructed to identify the optimal FB. Finally, the damage localization model is developed to improve inspection efficiency. The complete system structure of the proposed method is shown
Engineering application
As the critical components in DDWT, the health monitoring of MBs is extremely important for the safe and reliable operation of the DDWT. However, traditional vibration-based monitoring methods perform poorly in low-speed MBs. Relying on inspectors to monitor the MBs with an endoscope is time-consuming and difficult to accurately detect early bearing damages. Therefore, to improve the efficiency and accuracy of diagnosis, an integrated scheme based on AE signals is designed to monitor the health
Conclusion
In this study, an integrated methodology using AE signals is proposed for the health management of MBs to ensure the safe and stable operation of DDWT. First, the rotational speed is accurately estimated from the AE signal itself, which not only achieves synchronous collection but also reduces monitoring costs. Whereafter, to identify the optimal FB, an adaptive narrowband selection scheme named ASC is constructed to detect the weak fault of MB. Based on HOAEA and AIC, the damage localization
CRediT authorship contribution statement
Zhipeng Ma: Methodology, Writing – original draft. Ming Zhao: Conceptualization, Supervision, Writing – review & editing. Mourui Luo: Software, Investigation, Validation. Chao Gou: Writing – review & editing. Guanji Xu: Writing – review & editing, Data curation.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
This work is supported by the National Key Research and Development Program of China (No. 2021YFB2011400), and the National Natural Science Foundation of China (Grant No. 51875434), which are highly appreciated by the authors.
References (44)
- et al.
Wind turbine technology battles: gearbox versus direct drive - opening up the black box of technology characteristics
Technol. Forecast. Soc. Chang.
(2020) - et al.
Vibration fault diagnosis of wind turbines based on variational mode decomposition and energy entropy
Energy
(2019) - et al.
Prediction, operations, and condition monitoring in wind energy
Energy
(2013) - et al.
Automatic data-driven spectral analysis based on a multi-estimator approach
Signal Process.
(2018) - et al.
A review on data-driven fault severity assessment in rolling bearings
Mech. Syst. Signal Process.
(2018) - et al.
Adaptive energy-constrained variational mode decomposition based on spectrum segmentation and its application in fault detection of rolling bearing
Signal Process.
(2021) Fast computation of the kurtogram for the detection of transient faults
Mech. Syst. Signal Process.
(2007)- et al.
Application of an improved kurtogram method for fault diagnosis of rolling element bearings
Mech. Syst. Signal Process.
(2011) - et al.
Wavelets for fault diagnosis of rotary machines: a review with applications
Signal Process.
(2014) - et al.
Wavelet based approach to signal activity detection and phase picking: application to acoustic emission
Signal Process.
(2015)
Application of an improved maximum correlated kurtosis deconvolution method for fault diagnosis of rolling element bearings
Mech. Syst. Signal Process.
A novel blind deconvolution based on sparse subspace recoding for condition monitoring of wind turbine gearbox
Renew. Energy
Extraction of second-order cyclostationary sources by matching instantaneous power spectrum with stochastic model – application to wind turbine gearbox
Renew. Energy
Condition monitoring and intelligent diagnosis of rolling element bearings under constant/variable load and speed conditions
Mech. Syst. Signal Process.
Condition monitoring and anomaly detection of wind turbine based on cascaded and bidirectional deep learning networks
Appl. Energy
Application of the ensemble empirical mode decomposition and Hilbert transform to pedestal looseness study of direct-drive wind turbine
Energy
Low speed bearing fault diagnosis using acoustic emission sensors
Appl. Acoust.
Acoustic emission-based condition monitoring methods: review and application for low speed slew bearing
Mech. Syst. Signal Process.
Research on feature extraction algorithm of rolling bearing fatigue evolution stage based on acoustic emission
Mech. Syst. Signal Process.
Defect localization on rolling element bearing stationary outer race with acoustic emission technology
Appl. Acoust.
Natural mechanical degradation measurements in slow speed bearings
Eng. Fail. Anal.
Detection of sub-surface damage in wind turbine bearings using acoustic emissions and probabilistic modelling
Renew. Energy
Cited by (13)
Spectral structure inducing efficient variational model for enhancing bearing fault feature
2024, Signal ProcessingCompound fault diagnosis of wind turbine bearing under ultra-low speed operations using generalized sparse spectral coherence
2024, Mechanical Systems and Signal ProcessingA novel acoustic emission signal segmentation network for bearing fault fingerprint feature extraction under varying speed conditions
2023, Engineering Applications of Artificial IntelligenceA fault transfer diagnosis method for wind turbine bearings based on improved residual neural networks
2024, Zhendong Gongcheng Xuebao/Journal of Vibration Engineering