Bivariate empirical mode decomposition and its contribution to wind turbine condition monitoring

doi:10.1016/j.jsv.2011.02.027

Journal of Sound and Vibration

Volume 330, Issue 15, 18 July 2011, Pages 3766-3782

https://doi.org/10.1016/j.jsv.2011.02.027 Get rights and content

Abstract

Accessing difficulties and harsh environments require more advanced condition monitoring techniques to ensure the high availability of offshore wind turbines. Empirical mode decomposition (EMD) has been shown to be a promising technique for meeting this need. However, EMD was developed for one-dimensional signals, unable to carry out an information fusion function which is of importance to reach a reliable condition monitoring conclusion. Therefore, bivariate empirical mode decomposition (BEMD) is investigated in this paper to assess whether it could be a better solution for wind turbine condition monitoring. The effectiveness of the proposed technique in detecting machine incipient fault is compared with EMD and a recently developed wavelet-based ‘energy tracking’ technique. Experiments have shown that the proposed BEMD-based technique is more convenient than EMD for processing shaft vibration signals, and more powerful than EMD and wavelet-based techniques in terms of processing the non-stationary and nonlinear wind turbine condition monitoring signals and detecting incipient mechanical and electrical faults.

Introduction

As the deployment of onshore wind turbines (WT) becomes established and with a substantial increase in the deployment of WTs offshore to meet the 2020 renewable energy targets [1], [2], the reliability of these machines and their associated equipment has become the focus of increased R&D work within the academic and industrial communities [3], [4]. Due to the harsh operating environments and access challenges future offshore WTs must have higher reliability than onshore WTs to achieve acceptable availabilities and therefore reasonable cost of energy. One of the approaches to meet this requirement is the application of advanced condition monitoring techniques applied to a condition-based operation and maintenance strategy [5], [6], [7], [8], [9]. This approach has proved successful in some isolated cases for the management of onshore wind farms but it will become essential for the next generation of offshore wind farms, particularly the larger ones envisaged for the UK Round 3 and future UK Round 4 projects.

It is being recognized that currently available WT condition monitoring techniques need to be improved to ensure for the long-term reliable operation of machines. For example, they are mostly vibration-based, incorporate a considerable number of sensors, therefore being costly, and they may not be suited to all types of WTs and their faults, in particular as some WT manufactures are considering gearless direct drive designs. Moreover, the current vibration-based condition monitoring techniques applied to WTs are still based upon the experience from other industries where they have achieved success. They have not yet proven their effectiveness in wind industry due to the peculiarities of WTs, which have a slow and variable rotational speed and an unpredictable, rapidly changing torque. The lubrication oil debris counter is a popular tool for monitoring gearbox tooth and bearing wear, but it cannot detect failures outside the gearbox. Blade failure detection has been developed using advanced techniques, such as optic strain measurement, but these are too expensive for economic use in present turbines. Moreover, all these techniques are ineffective for detecting incipient electrical and power electronic faults, which have been shown to have even higher failure rates than the WT mechanical system [3]. In addition, the techniques for processing WT condition monitoring signals require further improvement. Spectral analysis is being widely adopted today and does have computational efficiency advantages, but it is not particularly suited for non-stationary and nonlinear signals as are seen in the WT.

Considering the non-stationary properties of WT signals, the wavelet transform has been recognized as a useful processing tool. For example, the continuous wavelet transform (CWT) was applied for detecting WT blade failure in [10], and diagnosing WT electrical and drive train mechanical faults in [11]. However, the intensive calculation involved in the CWT makes it difficult to apply to processing lengthy online WT monitoring signals. Moreover, it is inconvenient to apply the three dimensional time-scale-amplitude image to machine condition monitoring, as was done in [10], [11]. This is why a new wavelet-based technique namely ‘energy tracking’ was recently developed in [5]. This new technique inherits all the merits of CWT, while significantly reducing the calculation time, giving a more time-efficient algorithm, allowing it to process online condition monitoring signals from variable speed machinery such as WTs. However, this technique is highly dependent on the rotational speed and pre-knowledge of the machine, for example about the specific fault-related frequencies. If either the speed or fault-related frequency information is incorrect, it may lead to error or misdiagnosis and thus cannot give correct predictions on that incipient machine fault. In addition, the ‘energy tracking’ technique developed in [5] has not entirely removed the negative influence of the load from the extracted signal, which is still modulated by the varying load. It can be difficult to distinguish whether the change of energy is due to the presence of a fault or caused by varying load. This is why the ‘energy tracking’ technique may still have limitations in detecting WT incipient faults [5].

By contrast, the empirical mode decomposition (EMD) method [12] overcomes the defects of the wavelet-based ‘energy tracking’ technique. All Intrinsic Mode Functions (IMFs) derived from the signal are characterized as zero-mean oscillations. Therefore, the negative influences of varying load on the IMFs are partially removed. The priorities of EMD in aspect of fault detection have been demonstrated in [13]. In addition, the implementation of EMD is a data-driven process, not requiring any pre-knowledge of the signal or machine. These advantages drive EMD to be a promising tool for delivering improved WT condition monitoring, as indicated by the preliminary research in [14], [15].

However, traditional EMD is only applicable to one-dimensional real-value signals. It is unable to execute information fusion, which considers multiple signals and has been shown to be of importance for reliable condition monitoring. To enhance the capability of EMD, some efforts have been made recently to extend the method to the field of complex numbers using either Complex Empirical Mode Decomposition (CEMD) or bivariate empirical mode decomposition (BEMD) [16], [17], although these efforts were not specially made for machine condition monitoring purposes. In view of the extended EMD's advantages for information fusion, its feasibility in the field of WT condition monitoring will be investigated in this paper, by applying it to the detection of WT mechanical and electrical faults. Considering that BEMD developed in [17] is more powerful for detecting synchronous features in a complex-valued signal than CEMD developed in [16], BEMD is adopted in this paper for processing WT condition monitoring signals. Considering the difficulties of obtaining desired faulty signals from real WTs for research due to confidentiality reasons, in this paper the ‘fault-like perturbation’ signals to which the methods have been applied were collected from the following:

•
Either a centrifugal compressor with rotor-bearing system operating in a Petroleum-Chemical Plant, described in [13]. Although a centrifugal compressor is different from a WT, it is believed that the shaft vibratory signals collected from the former contain similarities of those from the latter with similar rotor-bearing system.
•
Or a specially designed WT condition monitoring test rig comprising a two-stage gearbox and an induction generator with the rotor circuit coupled via slip rings to an external three-phase resistive load bank, so that rotor electrical imbalance could be easily applied by adjusting the phase resistances of the load bank. The details of the test rig were described in [5], [11].

The remaining part of the paper is arranged as follows:

Firstly, EMD and its advantages in nonlinear signal analysis is briefly reviewed incorporating with the comparison to Fourier transform and wavelet transform, which are the most popular tools used today for processing non-stationary signals. BEMD is a further extension of EMD in order to enhance its power of incipient fault detection by introducing information fusion into the technology.

Secondly, the algorithm of BEMD and its comparison with traditional EMD was depicted in order to show the merits of BEMD in signal purification and shaft orbit reconstruction.

Thirdly, following designing a method for constructing complex-valued signals from WT three-phase electrical signals, the comparison of BEMD and EMD as well as the Wavelet-based ‘energy tracking’ technique is presented, showing the benefit of BEMD in detecting the WT incipient electrical fault.

Fourthly, the benefit of BEMD is further demonstrated through detecting a WT rotor mechanical unbalance fault, which cannot be detected successfully by either EMD or the wavelet-based ‘energy tracking’ technique. Besides demonstrating the powerful capability of BEMD on extracting weak features from nonlinear, non-stationary signals, another purpose of this part of work is to further demonstrate the possibility of detecting WT mechanical faults from its electrical signals. The success of this thought will significantly simplify the WT condition monitoring system and thus make it more economically justified. It is recognized that the high cost of an individual system has been one of the bottleneck problems significantly hindering the development of WT condition monitoring technologies and their wide spread in wind industry.

The paper is finally completed by summarizing the work into a few concluding remarks.

Section snippets

Brief review of EMD

EMD was developed as an innovative time series analysis tool by Huang et al. [12] for processing the signals with intra-wave features. EMD has proven to be an important alternative to traditional signal processing techniques such as the Fourier and wavelet transforms and shown great success in dealing with non-stationary, nonlinear condition monitoring signals like those from WTs [14], [15]. The computational algorithm of EMD has been introduced in many papers, so will not be repeated here to

BEMD and its advantages compared to EMD in processing shaft vibratory signals

BEMD is similar to EMD in calculation except in modifications of extrema detection and envelop definition. Its algorithm can be described as follows [17].

Step 1: Determine the number of projection directions N and calculate the projection directions $φ_{n} = 2 n π / N n \in [1, N]$

Step 2: Project the complex-valued signal x(t) on directions φ_n $p_{φ_{n}} (t) = Re (e^{- j φ_{n}} x (t)), j = \sqrt{- 1}$

Step 3: Find all local maxima of $p_{ϕ_{n}}$ (t) and record their locations and values ${(t_{i}^{n}, p_{φ_{n}} (t_{i}^{n}))}$ . Herein, i indicates the No. of individual local

Comparison of BEMD with EMD and the wavelet-based ‘energy tracking’ technique in detecting an incipient electrical fault in a variable speed WT

As mentioned above, the capability of wavelet transform to perform online condition monitoring for WTs has been significantly improved using the ‘energy tracking’ technique [5]. However, the drawbacks of the ‘energy tracking’ technique are that it requires pre-knowledge of the machine fault-related frequencies and is dependent on the measurement accuracy of speed signal. Hence, more powerful condition monitoring techniques are still required, the research in this paper aims at developing such a

Detection of a rotor mechanical unbalance fault in the WT

In this section, the BEMD-based method is applied to detecting a rotor mechanical unbalance fault occurring in the WT described in [5], [11] in order to further validate its power in detecting incipient fault and its feasibility in practical application.

It is necessary to note that in this paper both the mechanical and electrical faults occurring in the WT are being detected via the electrical signals of its generator, which could be of significant advantage when industrially applied because of

Conclusions

A new WT condition monitoring technique has been developed in this paper using the bivariate empirical mode decomposition, and exploiting its advantages for information fusion, which has been shown to be important in the reliable detection of incipient faults in WTs. Following the comparison with traditional EMD-based and the newly developed wavelet-based techniques, the proposed method successfully detected both incipient mechanical and electrical faults in a variable speed WT. From the

Acknowledgment

The authors would like to thank UK National Renewable Energy Centre (Narec), Blyth, for the original provision of the test rig and the staff of Durham University for its commissioning, instrumentation and development into a WT Condition Monitoring Test Rig, which was supported by the UK Engineering and Physical Sciences Research Council Supergen Wind Program EP/D034566/1. The work described in this paper was also supported by the National Natural Science Foundation of China. The Project no. is

References (21)

Y. Amirat et al.
A brief status on condition monitoring and fault diagnosis in wind energy conversion systems
Renewable and Sustainable Energy Reviews
(2009)
Wenxian Yang et al.
Empirical mode decomposition, an adaptive approach for interpreting shaft vibratory signals of large rotating machinery
Journal of Sound and Vibration
(2009)
BWEA, UK Wind Farm Statistics. Available from: 〈http://www.bwea.com/statistics〉 (accessed...
Global Wind Energy Council (GWEC), Global Wind 2008 Report,...
F. Spinato et al.
Reliability of wind turbine subassemblies
IET Renewable Power Generation
(2009)
E. Echavarria et al.
Reliability of wind turbine technology through time
Journal of Solar Energy Engineering—Transactions of the ASME
(2008)
Wenxian Yang et al.
Cost-effective condition monitoring for wind turbines
IEEE Transactions on Industrial Electronics
(2010)
B. Lu, Y.Y. Li, X. Wu, Z.Z. Yang, A review of recent advances in wind turbine condition monitoring and fault diagnosis,...
C.C. Ciang et al.
Structural health monitoring for a wind turbine system: a review of damage detection methods
Measurement Science and Technology
(2008)
R.W. Hyers et al.
Condition monitoring and prognosis of utility scale wind turbines
Energy Materials
(2006)

There are more references available in the full text version of this article.

Cited by (90)

Three-dimensional instantaneous orbit map for rotor-bearing system based on a novel multivariate complex variational mode decomposition algorithm
2022, Mechanical Systems and Signal Processing
Full spectrum and holospectrum are homogeneous information fusion technology developed for the fault diagnosis of rotating machinery, which is extensively exploited in the analysis of the orbits of rotor-bearing systems. However, they are not adapted for non-stationary signals, nor can they be used for fusion analysis of vibrations of multiple bearing sections. By drawing inspiration from the multivariate variational mode decomposition (MVMD) and the complex-valued signal decomposition, we propose a method called multivariate complex variational mode decomposition (MCVMD). It can simultaneously extract the forward and backward components of multiple bearing sections and realize non-stationary complex signal decomposition of multiple bearing sections of the rotor. To achieve the visualization goal of condition monitoring, we propose the three-dimensional instantaneous orbit map (3D-IOM). It enables more features of shaft vibration of a rotor system to be displayed and offers a new way for the fusion analysis of vibration signals of multiple bearing sections of rotating machinery. Furthermore, making the most of the joint information, we also provide a high-resolution time-full spectrum (Time-FS) to display the forward and backward frequency components of multiple bearing sections. The effectiveness of the proposed method through both the simulated experiment and the real-life complex-valued signals is demonstrated in this paper.
Scientometric review of artificial intelligence for operations & maintenance of wind turbines: The past, present and future
2021, Renewable and Sustainable Energy Reviews
Citation Excerpt :
Their technique can help in identifying random wind vibrations and its effects based on the analysed spectrum, facilitating fault diagnosis. Multiple studies [35,38,39] have utilised vibration signals and applied Empirical Mode Decomposition (EMD) for detecting incipient faults in the turbine’s mechanical and electrical sub-components, by decomposing these signals into intrinsic mode functions (IMFs). Feng et al. [35] for instance, proposed demodulation analysis of planetary gearbox vibration signals, by accounting for the IMFs produced using an ensemble EMD method.
Wind energy has emerged as a highly promising source of renewable energy in recent times. However, wind turbines regularly suffer from operational inconsistencies, leading to significant costs and challenges in operations and maintenance (O&M). Condition-based monitoring (CBM) and performance assessment/analysis of turbines are vital aspects for ensuring efficient O&M planning and cost minimisation. Data-driven decision making techniques have witnessed rapid evolution in the wind industry for such O&M tasks during the last decade, from applying signal processing methods in early 2010 to artificial intelligence (AI) techniques, especially deep learning in 2020. In this article, we utilise statistical computing to present a scientometric review of the conceptual and thematic evolution of AI in the wind energy sector, providing evidence-based insights into present strengths and limitations of data-driven decision making in the wind industry. We provide a perspective into the future and on current key challenges in data availability and quality, lack of transparency in black box-natured AI models, and prevailing issues in deploying models for real-time decision support, along with possible strategies to overcome these problems. We hope that a systematic analysis of the past, present and future of CBM and performance assessment can encourage more organisations to adopt data-driven decision making techniques in O&M towards making wind energy sources more reliable, contributing to the global efforts of tackling climate change.
Multichannel fault diagnosis of wind turbine driving system using multivariate singular spectrum decomposition and improved Kolmogorov complexity
2021, Renewable Energy
When wind turbine driving system (WTDS) undergoes abnormal conditions, the fault information hidden in WTDS scatters over multiple signal channels and hence inadequate for fault diagnosis only via fault information extraction of single-channel signal. To make full use of multichannel fault information of WTDS and improve diagnostic accuracy, this paper proposes a new approach based on multivariate singular spectrum decomposition (MSSD) and improved Kolmogorov complexity (IKC). Firstly, based on singular spectrum decomposition (SSD) and the idea of multichannel data processing, a multivariate singular spectrum decomposition (MSSD) method is presented to process multichannel vibration data collected from WTDS, which can obtain adaptively multichannel mode components without extra user-defined parameters. Secondly, through incorporating symbolization process into Kolmogorov complexity (KC), an improved complexity metric abbreviated as IKC is proposed to capture the fault information of multichannel mode components, which can enhance fault feature extraction ability of KC. Finally, IKC-based multichannel fault features are fed into partial least squares (PLS) to automatically discriminate different fault patterns of WTDS. Practical engineering data from WTDS demonstrate the effectiveness of the proposed approach. Additionally, the superiority of the proposed approach has also proven in extracting fault information and health condition identification compared to the other multichannel methods and traditional single-channel approaches reported in the literature.
An EMD-based principal frequency analysis with applications to nonlinear mechanics
2021, Mechanical Systems and Signal Processing
There are time signals of general interest with periodic components in addition to trend and randomness. It is of great importance to identify their frequencies and amplitudes which may be varying with time. In the past, we have seen excellent works on time-frequency analysis of a signal such as short-time Fourier, wavelet, Hilbert and Hilbert-Huang transforms among others. Yet there are still critical and fundamental issues to be addressed. Notably all the previous analyses (tacitly) assume that the signal concerned is a linear superposition of its decomposed components with each of them being Fourier-analyzed its spectrum no matter whether a base set of functions or no base is employed. In this study, we propose to develop a principal frequency analysis (PFA) suitable for general summed linear (or single harmonic) signals and product signals (in the form of a beat or wave-packet). PFA is meant to extract the major frequencies from the phase of a complex signal as well as its amplitude (e.g., defined through Hilbert transform), in particular the product frequencies of a product signal or wave-packet. As an illustration, this approach of analysis, PFA for directly obtaining product frequencies is first applied to several basic examples, then to signals from nonlinear oscillators, and then to time-dependent lift and drag coefficients, related to vortex shedding behind a circular cylinder or a sphere in fluid mechanics.
Damage detection techniques for wind turbine blades: A review
2020, Mechanical Systems and Signal Processing
Blades play a vital role in wind turbine system performances. However, they are susceptible to damage arising from complex and irregular loading or even cause catastrophic collapse, and they are expensive to maintain. Defects or damages on wind turbine blades (WTBs) not only reduce the lifespan and power generation efficiency of the wind turbine, but also increase monitoring errors, safety risks and maintenance costs. Therefore, damage detection for WTBs is of great importance for failure avoidance, maintenance planning, and operation sustainability of wind turbines. This paper provides a comprehensive review of state-of-the-art damage detection techniques for WTBs, including most of those updated methods based on strain measurement, acoustic emission, ultrasound, vibration, thermography and machine vision. Firstly, typical damages of WTBs are comprehensively introduced. Secondly, detection principles, development methods, pros and cons of the aforementioned techniques for blade inspection, and their fault indicators are reviewed. Finally, potential research directions of WTB damage detection techniques are addressed via a comparative analysis, and conclusions are drawn. It is expected that this review will provide guidelines for practical WTB inspections, as well as research prospects for damage detection techniques.
Detection of incipient faults in EMU braking system based on data domain description and variable control limit
2020, Neurocomputing
The performance of braking system strongly affects the safe operation of electric multiple unit (EMU). During the practical operation, it’s of great significance to detect incipient faults in braking system. Since the braking process is a typical non-Gaussian and multi-stage process, it’s difficult to detect these incipient faults. Particularly, the usually occurred overlap of normal and faulty samples during braking process makes the detection more difficult. In this paper, a novel method based on data domain description and variable control limit (VCL) is developed for detecting incipient faults in EMU braking system. The local reachability density (LRD) weighted support vector data description with negative samples (NSVDD) is introduced for offline modeling to get more accurate domain description, while the Gaussian kernel trick is utilized to obtain hypersphere with soft boundary. With the presence of sample overlap, the VCL strategy is adopted for online fault detection, which effectively reduces false alarm rate (FAR) and increases fault detection rate (FDR) simultaneously. A case study of three kinds of incipient faults in EMU braking system fully demonstrates the effectiveness of the proposed method.

View all citing articles on Scopus

View full text

Bivariate empirical mode decomposition and its contribution to wind turbine condition monitoring

Abstract

Introduction

Section snippets

Brief review of EMD

BEMD and its advantages compared to EMD in processing shaft vibratory signals

Comparison of BEMD with EMD and the wavelet-based ‘energy tracking’ technique in detecting an incipient electrical fault in a variable speed WT

Detection of a rotor mechanical unbalance fault in the WT

Conclusions

Acknowledgment

Renewable and Sustainable Energy Reviews

Journal of Sound and Vibration

Reliability of wind turbine subassemblies

IET Renewable Power Generation

Reliability of wind turbine technology through time

Journal of Solar Energy Engineering—Transactions of the ASME

Cost-effective condition monitoring for wind turbines

IEEE Transactions on Industrial Electronics

Structural health monitoring for a wind turbine system: a review of damage detection methods

Measurement Science and Technology

Condition monitoring and prognosis of utility scale wind turbines

Energy Materials