CN103033745B - The method and system of the mechanical breakdown of detection electromotor - Google Patents

Abstract

The invention discloses and a kind of detect the method for fault in wind-driven generator according to current characteristic analysis.Described method includes the one group of signal of telecommunication obtaining the running status representing electromotor.Additionally, process the described signal of telecommunication to generate the normalized spatial spectrum of the signal of telecommunication.According to the analysis of described current spectrum detects fault, described fault and gear-box or bearing or other any parts associated with described electromotor are relevant.

Description

Method and system for detecting mechanical failure of generator

Technical Field

Embodiments of the present invention generally relate to condition monitoring of an electromechanical machine (electromechanical machine). More particularly, embodiments of the invention relate to methods and systems for monitoring the condition of a drive system and bearings of a machine motor based on Current Signal Analysis (CSA).

Background

According to convention, vibration signals are primarily used to monitor mechanical anomalies in electromechanical systems. It has been found that mechanical faults in the drive system will generate vibrations in the radial rotor motion and thereby corresponding torque oscillations at the rotor mechanical rotational frequency. Mechanical faults associated with the drive system may be detected by monitoring and studying the rotor mechanical rotational frequency. However, condition monitoring using vibration signals has many disadvantages such as signal background noise due to externally excited motion, sensitivity to mounting location, and their aggressiveness-measuring characteristics.

Other condition monitoring techniques are based on the observation that load torque oscillations require phase modulation of the stator current, and thereby analyze stator current characteristics to detect mechanical disturbances due to faults. Since such current monitoring techniques can be significantly cost effective and easy to implement, there is increasing interest in detecting mechanical faults in electrical machines. For example, for bearing fault detection in an electromechanical machine, bearing faults may be classified as single point faults or generalized coarse faults. Single point faults have been detected by using Motor Current Signal Analysis (MCSA) through the mechanical natural frequency of the bearing and taking these types of anomalies into account as eccentric faults. However, for generalized asperity faults, the natural bearing failure frequency is not observable or may not exist, particularly in the early stages. Furthermore, regardless of the type of fault, bearing fault signatures are typically subtle relative to dominant components in the sampled stator current, such as power fundamental harmonics, eccentricity harmonics, and slot harmonics. Unlike bearing vibration monitoring, which has evolved to industry standards from long-term field experience, the field experience of stator current monitoring is extremely limited and presents great difficulties. For example, given that the bearing fault signature in the stator current has been extremely subtle, the magnitude of the bearing fault signature may also vary in different applications. Furthermore, while gearboxes are widely used in industrial applications, it is rarely recommended to use stator current signal analysis for gearbox monitoring.

Accordingly, there is a need for improved methods and systems for monitoring the condition of drive systems, particularly gearboxes and bearings, using current signature analysis.

Disclosure of Invention

According to one embodiment of the present invention, a method of detecting a mechanical fault in a generator is provided. The method includes acquiring an electrical signal indicative of an operating condition of the generator. The method also includes normalizing the electrical signal to extract spectral information. The method further includes detecting a fault based on the analysis of the spectral information.

Wherein detecting a fault based on the analysis of the spectral information comprises detecting a bearing fault or a gearbox fault.

Wherein obtaining an electrical signal representative of an operating condition of the generator comprises obtaining an electrical signal for each of a plurality of phases of the generator.

Wherein the electrical signals representative of the operating state of the generator include current and voltage signals.

Wherein the current spectrum is normalized according to at least one of: fast fourier transform, time-frequency analysis, and multi-mode regression analysis.

Wherein normalizing the electrical signal to extract spectral information further comprises: eliminating a symmetric component in the current spectrum; and extracting an asymmetric component that is displayed in the current spectrum during an abnormal operating state.

Wherein removing the identified symmetric components in the current spectrum comprises squaring current signal instantaneous values for each of the multiple phases and summing the squared values.

Wherein eliminating the symmetric component in the current spectrum comprises eliminating a symmetric component that constitutes a generator current spectrum during normal operating conditions of the generator.

The method of detecting a mechanical fault in a generator further comprises: acquiring a second set of electrical signals representative of a non-fault operating condition of the generator; converting electrical signals representative of the operating condition of the generator into a hybrid electrical signal using the second set of electrical signals; normalizing the mixed electrical signal to extract spectral information; and detecting a fault based on the analysis of the spectral information.

Wherein obtaining the second set of electrical signals representative of the non-fault operating condition of the generator comprises obtaining the second set of electrical signals for each of a plurality of phases of the generator.

Wherein normalizing the mixed electrical signal to extract spectral information comprises: dynamically canceling a symmetric component in the mixed electrical signal; comparing the magnitude and phase of the asymmetric component in the mixed electrical signal with the magnitude and phase of the second set of electrical signals; and detecting a fault when the asymmetric component in the mixed electrical signal deviates in magnitude and phase from the second set of electrical signals.

Wherein detecting a fault comprises detecting a fault in a rotor of the generator.

Wherein analyzing the extracted features comprises: the extracted features are compared to a determined threshold.

According to another embodiment of the present invention, a system for detecting gearbox and bearing faults of a generator is provided. The system includes one or more sensors for acquiring electrical signals indicative of an operating condition of the generator. The system also includes a controller for normalizing the electrical signal to extract spectral information. The system further includes a fault detection unit module for detecting one or more faults in the gearbox based on an analysis of the extracted spectral information.

Wherein the fault detection unit module detects a bearing fault or a gearbox fault.

Wherein the one or more sensors acquire electrical signals for each of a plurality of phases of the generator.

The controller is further configured to: eliminating a symmetric component in a current spectrum of the electrical signal; and extracting an asymmetric component that is displayed in the current spectrum during an abnormal operating state.

Drawings

These and other features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

fig. 1 is a schematic diagram of an exemplary embodiment of an electromechanical machine (EMM) according to one embodiment of the present invention.

FIG. 2 is a block diagram of an exemplary fault detection system configuration in accordance with one embodiment.

FIG. 3 is a block diagram of an exemplary fault detection system configuration in accordance with one embodiment.

FIG. 4 is a block diagram of an exemplary fault detection system configuration in accordance with one embodiment.

FIG. 5 is a block diagram of an exemplary fault detection system configuration in accordance with one embodiment.

FIG. 6 is a schematic diagram of a structure of a bearing of an EMM according to one embodiment.

Fig. 7 is a schematic diagram of exemplary waveforms of stator output currents of the EMM of fig. 1 in the frequency domain according to one embodiment of the invention.

FIG. 8 is a schematic diagram of a gearbox gear configuration of EMMs according to one embodiment.

Fig. 9 is a schematic diagram of exemplary waveforms of stator output currents of the EMM of fig. 1 in the frequency domain according to one embodiment of the invention.

FIG. 10 is a flow chart illustrating steps involved in an exemplary method of detecting a fault in a wind turbine.

FIG. 11 is a flowchart illustrating steps involved in an exemplary method of detecting faults in an electromechanical machine (EMM) according to one embodiment of the invention.

Fig. 12 is a flow chart illustrating steps involved in an exemplary method of detecting a fault in an electromechanical machine (EMM) according to another embodiment of the present invention.

Detailed Description

This patent document describes an electrical multiphase imbalance separation technique for highly sensitive detection of faults in an electromechanical machine (EMM), including drive system anomalies and bearing faults. The techniques are capable of distinguishing the state of a degraded EMM component from a normal or acceptable state based on subtle changes that may occur in the magnitude of the fault signature. The dominant components in the stator current of a typical machine motor are power fundamental harmonics, eccentric harmonics, slot harmonics, saturated harmonics, and other components originating from unknown sources such as environmental noise. Since such dominant components exist before and after the occurrence of a bearing fault, most of the information carried by these components is not relevant to the bearing fault. In this sense, they are essentially "noise" to the EMM fault detection problem. In comparison, the component of the stator current injected by the drive system fault is much lower in magnitude than the dominant component. Thus, the sensitivity of the assay will be compromised, even with the best instruments and assays based essentially on single phase assays.

In general, the frequency of the dominant component may be determined by the power supply frequency, the rotor speed, and the machine configuration. If the drive system fault signature is considered as a signal and the dominant component is considered as noise, then the drive system fault detection problem is essentially a low signal-to-noise ratio problem. Furthermore, since the signal frequency may not be predictable, the signal and noise may have the same frequency. Therefore, the noise component should be removed to find the fault signature. Thus, the dominant components of the stator current that are not related to the drive system fault are estimated and then eliminated by the estimation in real time. Thus, the residual component (i.e., the noise-cancelled stator current) is more correlated with the drive system fault.

An embodiment of the invention enables systematic and dynamic cancellation of contributions of symmetric or useful components of the multi-phase system electrical signal, such as current, voltage or power. This elimination allows the "distortion" of the electrical signal due to machine asymmetry and/or faults to be highlighted in the ac spectrum and thus makes the corresponding detection easier. In particular, the invention relates to squaring the electrical signal instantaneous values (current, voltage, power, etc.) for each of a plurality of phases and summing the resulting values. Squaring the instantaneous value of an electrical signal "incorporates" or adds all the symmetrical or balanced (and often useful) component contributions to the equivalent dc signal. Thus, only the unbalanced component of the signal indicative of asymmetry or fault, if any, will be displayed as an alternating current at twice the frequency. Thus, by this elimination of all symmetry terms, the effect of such anomalies, whether inherent machine asymmetry, stator winding failure or drive system failure, or main bearing failure, will be highlighted in the AC spectrum as they do not have to compete with the useful or symmetric components. The resulting ac spectrum, which can be attributed to abnormal conditions only, will therefore be analyzed at a much higher sensitivity level.

Referring to fig. 1, an electromechanical machine (EMM), such as a three-phase generator, is configured to generate electricity. EMM assembly 100 includes rotor assembly 110, main bearing 120, main shaft 130, gearbox 140, electrical sensors (not shown), and multiphase generator 150. The EMM assembly 100 also includes a controller for monitoring and controlling operation of the multiphase generator 150 in response to a generator fault condition. The controller includes a processor for detecting the presence of fault conditions for various components within the EMM assembly 100, including the drive system and bearings. The controller will be described in more detail with reference to fig. 2. The electrical signal sensors may be current and voltage sensors for acquiring current and voltage data associated with the multi-phase generator 150. For example, a current sensor is used to sense current data for one or more of the phases of a multi-phase generator. Specifically, with a three-phase induction generator, current and voltage sensors are used to sense current and voltage data for the three phases of the three-phase induction generator. Although certain embodiments of the invention will be described with reference to a multi-phase generator, other embodiments of the invention are applicable to other multi-phase machines.

In one embodiment of the present invention, the current and voltage sensors detect stator current data and voltage data, respectively, of the multi-phase generator 150. The stator current data and voltage data obtained from the sensors will be transmitted to the controller for further processing and analysis. The analysis includes performing a Current Signature Analysis (CSA) to detect faults within EMMs, including drive system faults and bearing faults. According to one embodiment of the invention, the controller is configured to cancel contributions of symmetrical or useful components in the electrical signal such that only unbalanced components of the signal associated with the fault appear as an amount of alternating current in the alternating current spectrum. In particular, the controller is programmed to remove such non-fault related symmetrical or useful components by squaring the current data instantaneous values for each of the multiple phases and summing the squared values. Thus, the non-fault-related symmetrical component will be converted into a direct current component, while the fault-related asymmetrical component in the current signal will be displayed in the alternating current spectrum at twice the frequency.

Referring now to FIG. 2, a detailed block diagram of the controller is illustrated. As described with reference to fig. 1, the controller 170 includes a processor 180 and an electrical multi-phase imbalance separation technique (eMIST) unit 190. The eMIST unit 190 is connected to the sensor bank 160 and receives stator current and voltage data for each phase of the multi-phase generator 150 and prepares the current and voltage data for processing by the processor 180. The functionality of the eMIST unit 190 will be described in more detail with reference to fig. 3, 4, and 5. Although the illustrated eMIST unit 190 is a separate component, it should also be appreciated that the functions of the eMIST unit 190 may be performed by the processor 180.

The processor 180 functions as an unbalanced splitting system that systematically and dynamically cancels contributions of symmetric or useful components, such as current or voltage or power, in the electrical signals of the multi-phase system. Such cancellation causes electrical signal "distortion" due to asymmetry and/or failure of the machine 100 to be highlighted in the ac spectrum. In other words, the processor 180 is configured to treat the fundamental frequency component as noise and the remaining frequency components as fault-related components. Dynamically eliminating the noise component from the continuously acquired stator current component will result in a fault-only related component, which is introduced into the stator current by the electromechanical fault. For example, the processor 180 may be programmed to eliminate fundamental frequencies (e.g., 60Hz) and low frequency harmonics (particularly fundamental frequency components) in the measured stator current data. Removing the fundamental frequency from the measured stator current data can greatly improve the analog-to-digital conversion resolution and SNR because the fundamental frequency of 60Hz is a large part of the frequency spectrum of the current signal.

The processor 180 can be programmed to further analyze the noise-cancelled stator currents. In particular, the threshold may be calculated such that a measurement frequency falling outside the threshold indicates an EMM component status deterioration/anomaly. One way to calculate the threshold may be to find the highest fault signature level in the normal state and set this value to the threshold.

Referring to fig. 3, in an exemplary embodiment of the invention, an eMIST unit 190 is capable of canceling noise in the stator current and isolating fault signals therein. To accurately cancel noise in the stator current, the eMIST unit 190 is configured to dynamically cancel a non-fault-related balanced component, i.e., a noise component in the stator current. To dynamically cancel noise components in the stator current spectrum, the eMIST unit 190 is configured to square the electrical signal instantaneous values (current, voltage, power, etc.) for each of the multiple phases and sum the squared instantaneous values according to equations (1) and (2), thereby converting the symmetrical or balanced (and often useful) components into equivalent dc quantities and displaying any unbalanced and fault-related components in the ac spectrum at twice the frequency.

I＝Ia2+Ib2+...+In2 ----(1)

V＝Va2+Vb2+...+Vn2 ----(2)

Thus, only any asymmetry terms will be displayed in the ac spectrum as ac traffic for I and V. Thus, by effectively eliminating all symmetry terms in this way, the effects of anomalies, whether inherent machine asymmetry or drive system component or bearing failure, are highlighted as an alternating quantity in the alternating frequency spectrum. The resulting ac spectrum, which can be attributed to abnormal conditions only, will therefore be analyzed with much higher sensitivity for a faulty drive system or bearing condition. The processor 180 analyzes the results of equations (1) and (2) and detects a fault in the presence of any fault-related traffic in the ac spectrum.

It should be noted that mechanical failures associated with the EMM assembly 100 may also include failures in the rotor assembly 110. If a fault occurs in the rotor assembly 100, the stator current measurement will not be helpful in detecting a fault associated with the rotor, as any fault occurring in the rotor will create equal modulation in all n phases of the multiphase generator, where "n" is the number of phases. To detect faults in the rotor assembly, the eMIST unit 190 is configured to calculate baseline measurements as shown in fig. 4 and then "hybrid" stator current data as shown in fig. 5. The baseline stator current and voltage data measurements are current and voltage data associated with a healthy drive system and bearing components and are acquired during a non-fault operating state of the EMM assembly 100. The baseline stator current data associated with the healthy drive system and the bearing condition may include a set of stator current data for each of the acquired multiple phases, wherein the stator current data is acquired shortly after installation of the EMM component 100 including the bearing and the drive system components. In one example, the baseline stator electrical signal is acquired during initial operation of the EMM assembly 100, i.e., during a first operation of the EMM assembly 100 after installation of the EMM assembly 100. The stator current data thus obtained may ensure that no components related to drive system or bearing faults are included in the stator current. Referring to fig. 4, in one example, the eMIST unit 190 receives current and voltage data from the sensor bank 160, and calculates baseline measurements of the current and voltage according to equations (3) and (4) below,

Ibaseline＝Ia-baseline2+Ib-baseline2+...+In-baseline2 ----(3)

Vbaseline＝Va-baseline2+Vb-baseline2+...+Vn-baseline2 ----(4)

wherein,

ibaseline is baseline stator current data measured during healthy machine conditions

Ia-baseline is the baseline stator current data for the first phase

Ib-baseline is baseline stator current data for the second phase

In-baseline is baseline stator current data for the nth phase

Vbias line is baseline stator voltage data measured during a healthy machine state

Va-baseline is baseline stator voltage data for the first phase

Vb-baseline is the baseline stator voltage data for the second phase

Vn-baseline is baseline stator voltage data for the nth phase

Thus, the baseline current (I) and voltage (V) data is the sum of the squares of the instantaneous values of the current and voltage signals for each of the multiple phases. The eMIST unit 190 is further configured to perform a Fast Fourier Transform (FFT) on the baseline stator current and voltage values according to equations (5) and (6) to decompose the non-fault-related components in the current signal (I) into equivalent dc quantities, while the fault-related components are present in the ac frequency spectrum at twice the frequency in the frequency domain. Since all components in the stator current are noise in a healthy bearing condition, the ac spectrum does not contain any fault information during the baseline measurement. Thus, when a fault occurs, only the noise component will be displayed in the ac spectrum at twice the frequency.

FFT(I)＝2*fbaselineI ----(5)

FFT(V)＝2*fbaselineV ----(6)

In the ideal case, assuming that the EMM component 100 is operating properly without mechanical failure, the resulting stator current and voltage data will have non-fault-related components, i.e., balanced components in the ac frequency spectrum that have been eliminated and have no fault-related components. The resulting stator current and voltage data will form baseline data during fault detection. The processor 180 may store the baseline measurements in the memory 185 for use during rotor 110 fault analysis and detection.

Further, referring to fig. 5, the eMIST unit 190 receives current and voltage data from the sensor bank 160 and calculates hybrid stator current data of the current and voltage according to equations (7) and (8) shown below. The combined stator current and voltage data is calculated by replacing current and voltage components of at least one phase of the multi-phase current and voltage data with corresponding baseline current and voltage data for that particular phase, wherein the baseline data is measured during a normal/healthy machine stay condition.

Ihybrid＝Ia-old2+Ib-new2+...+In-new2 ----(7)

Vhybrid＝Va-old2+Vb-new2+...+Vn-new2 ----(8)

Wherein,

ihybrid is stator current data acquired during fault detection

Ia-old is the first phase current data acquired during baseline measurement

Ib-new are current data of the second phase acquired during fault detection

In-new is current data of the n-th phase acquired during fault detection

Vhybrid is stator voltage data acquired during fault detection

Va-old is the voltage data of the first phase acquired during the baseline measurement

Vb-new is voltage data of the second phase acquired during fault detection

Vn-new is voltage data of the nth phase acquired during fault detection.

The eMIST unit 190 is further configured to perform a Fast Fourier Transform (FFT) on the mixed stator current and voltage values according to equations (9) and (10) to decompose the non-fault-related components in the current signal (I) into equivalent dc quantities while presenting the fault-related components in the ac frequency spectrum at twice the frequency in the frequency domain.

FFT(Ihybrid)＝2*fhybridI ----(9)

FFT(Vhybrid)＝2*fhybridV ----(10)

The processor 180 may then compare the magnitude and phase of the mixed ac voltage and current magnitude to the baseline voltage and current measurements. Based on the comparison, any deviation of the amount of mixing from the baseline measurement may indicate a fault in at least one component of the EMM component 100, such as a drive system component or a bearing. The results of the comparison may be stored in memory 185 for further analysis, such as frequency of failure, time of failure, frequency of failure of particular components, and the like.

Furthermore, to improve detection accuracy, a set of thresholds needs to be obtained for each component monitored, including the gearbox 140 and the bearing 120 of the EMM assembly 100. To distinguish between various faults occurring in the components within the EMM assembly 100, multiple samples of the current and voltage signals at the baseline state and fault detection state are collected for each component and corresponding thresholds are set. For example, to distinguish a degraded gearbox condition from other faulty components, a warning threshold needs to be set for the RMS of the noise canceling stator current of the gearbox 140. Possible gearbox faults may be detected by observing uncontrolled deviations in the noise-cancelled stator current from the determined warning threshold. To avoid false determinations due to insufficient data, the processor 180 begins the determination after enough noise-canceled stator current samples are received, e.g., after more than 30 to 50 samples are received. The deviation can be measured by the percentage of out of control samples outside the control limits, e.g. above 10%, and a warning message will be sent about the gearbox status.

Fig. 6 is a schematic diagram of a bearing 600, the bearing 600 having an inner raceway 610, an outer raceway 620, bearing balls 630 located between the inner raceway 610 and the outer raceway 620, and a cage 640 for holding the balls 630 in place within the bearing 600. The outer and inner raceway frequencies develop when each ball 630 experiences a failure. This occurs Nb times during a full raceway cycle, where Nb is the number of balls 630 in the bearing 600. The bearing frequency fbearing can thus be defined according to equations (11) to (13),

fbearing：

outer raceway: $f_0=\frac{N_b}{2}{f}_r(1-\frac{D_b}{D_c}\cos \mathrm{\β})----\left(11\right)$

inner raceway: $f_i=\frac{N_b}{2}{f}_r(1+\frac{D_b}{D_c}\cos \mathrm{\β})----\left(12\right)$

ball bearing: $f_b=\frac{D_c}{D_b}{f}_r(1-\frac{D_b^2}{D_c^2}{\cos }^2\mathrm{\β}).----\left(13\right)$

wherein,

f0 is the frequency of vibration of outer race 620

fi is the vibration frequency of the inner raceway 620

fb is the frequency of vibration of bearing ball 630

Nb is the number of balls 630

fr mechanical wind wheel speed (Hz)

Db is the ball diameter

Dc is bearing pitch diameter

Beta is the contact angle of the ball with the raceway

Since such bearing vibrations will induce a gap adjustment, the vibrations will also be shown in the current spectrum of the generator 150. Specific frequencies in the stator current spectrum may be associated with specific failure modes in the bearing 600 and gearbox 140 components. According to the formula (14),

stator current frequency:

fs＝|ffundamental±k.fbearing| ----(14)

wherein,

ffundamental＝60Hz

in one example, to detect a failure of the bearing inner raceway 610 under a given sample test condition of 800rpm shaft speed, 15KW load, 20Hz rotor excitation frequency, and 60Hz stator output frequency, the bearing vibration frequency according to equation (12) is calculated as,

$f_{\mathrm{bearing}}=\frac{N_b}{2}{f}_{\mathrm{rotr}}(1+\frac{D_b}{D_c}\cos \mathrm{\β})\≈66~67\mathrm{Hz}$

wherein,

N_b＝8

f_rotor＝800/60＝13.3Hz

D_b＝2.3cm

D_c＝9.3cm

β≈0

the stator current frequency according to equation (14) is given by,

$f_{\mathrm{bearing}}=\frac{N_b}{2}{f}_{\mathrm{rotr}}(1+\frac{D_b}{D_c}\cos \mathrm{\β})\≈66~67\mathrm{Hz}$

f_s＝|f_fundamental±k·f_bearing|

f_s＝|f_fundamental-2·f_bearing|≈73.1Hz

wherein,

stator output frequency of 60Hz

Constant k 2

The waveform of the sample stator output current in the frequency domain is shown in fig. 7. The x-axis 710 represents frequency (Hz) and the y-axis 720 represents gain (dB). Curve 730 represents the stator output current for a normal bearing condition and curve 740 represents the stator output current for a failed bearing condition with a failed inner bearing raceway. The peak stator output current due to bearing failure is shown as curve 750. It can be seen from the example that the difference in stator output current for the healthy bearing condition and the failed bearing condition is about 15 dB.

FIG. 8 is a schematic diagram of a gearbox gear with worn teeth 810. For a given sample test condition with a shaft speed of 800rpm, a load of 15KW, a rotor excitation frequency of 20Hz, and a stator output frequency of 60Hz, the stator current frequency of the faulty gearbox gear 800 is calculated according to equation (12) as,

f_s＝|f_fundamental±k·f_rotor|

f_s＝|f_fundamental-56·f_rotor|

≈833Hz

wherein,

f_rotor＝800/60＝13.3Hz

ffundamental＝60Hz

constant k 56

The constant k determines the bandwidth over which the waveform is monitored. Thus, the constant k may be determined to select a particular bandwidth within which the frequency represented by the waveform is most sensitive to fault analysis and detection.

For a failed gearbox, the waveform 900 of the sample stator output current in the frequency domain is shown in FIG. 9. The x-axis 910 represents frequency (Hz) and the y-axis 920 represents gain (dB). Curve 930 represents the stator output current for a healthy gearbox condition and curve 940 represents the stator output current for a failed gearbox condition with a gear failure. The peak stator output current due to gearbox failure is shown as curve 950. As can be seen from the example, the difference in stator output current for the healthy gearbox state and the faulty gearbox state is about 20 dB.

In one embodiment, the present invention is used in a Doubly Fed Induction Generator (DFIG) in which any mechanical fault associated with a generator bearing or gearbox is detected by analyzing electrical signals such as stator voltage and current outputs of the DFIG. In one example, the DFIG is part of a wind turbine assembly, wherein blades of the wind turbine include rotor assemblies of the DFIG. The method includes acquiring a stator output signal, e.g., a stator output current signal, using one or more sensors. The one or more sensors may be condition-based monitoring (CBM) sensors. The method includes monitoring spectral content in the DFIG stator current and correlating spectral characteristics of the stator current with faults in the DFIG bearings and gearboxes. Faults in bearings and gearboxes cause vibrations in the shaft of the DFIG, which in turn propagate as torque oscillations in the DFIG at the rotor mechanical rotational frequency. Such vibrations will also be shown in the current spectrum of the generator, since oscillations will induce a gap adjustment. Thus, a particular frequency of the stator output current of the DFIG may be associated with a particular failure mode in the bearings and drive system components. Although embodiments of the present invention are described with respect to a DFIG in a wind turbine, embodiments of the present invention may also be applied to other dynamoelectric machines and other systems.

FIG. 10 is a flow chart illustrating steps involved in an exemplary method of detecting a fault in a wind turbine generator. In step 1001, an electrical signal indicative of an operating condition of the generator is acquired. In one example, the electrical signal is a current signal corresponding to an operating condition of the generator, such as a fault operating condition. Further, in step 1002, the electrical signal is processed according to a fast fourier transform, a time-frequency analysis, or a multi-modal regression analysis, or a combination of these methods, to provide a normalized electrical signal spectrum. In step 1003, the normalized spectral information is provided to a processor for feature extraction and a conclusion based on spectral enhancement is performed. In step 1003, a fault related to the gearbox or bearings or any other component associated with the generator is detected when the current signal deviates from a determined threshold. The threshold for detecting a particular fault may be determined based on generator rating, field testing, and/or simulation results. In another embodiment, generator failure may be detected by fusing the results of the vibration signal and the electrical signal.

FIG. 11 is a flow chart illustrating steps involved in an exemplary method of detecting faults in an electromechanical machine (EMM) according to an embodiment of the present invention. The method 1100 includes, in step 1101, acquiring one or more electrical signals for each phase of a multi-phase EMM during operation. Examples of electrical signals include current, voltage, power, and the like. Further, the method includes dynamically canceling a symmetric component in one or more electrical signals in step 1102. In one embodiment, the symmetric component is dynamically removed according to the algorithms shown in equations (1) and (2), which square the instantaneous values of one or more electrical signals for each of the multiple phases and sum the squared values of one or more electrical signals. The resulting current and voltage values are decomposed into symmetric and asymmetric components by any normalization technique such as Fast Fourier Transform (FFT). In step 1103, a fault is detected by identifying an asymmetric component in the resulting electrical signal. In one example, the asymmetric components in the resulting electrical signal, i.e., in the resulting AC spectrum, are compared to a determined threshold. When the asymmetric component deviates from a determined threshold, a fault is detected.

Fig. 12 is a flow chart illustrating steps involved in an exemplary method of detecting a fault in an electromechanical machine (EMM) according to another embodiment of the present invention. The method 1200 includes, in step 1201, acquiring a first set of electrical signals for each phase of a multi-phase EMM during an initial run of the EMM. In one embodiment, the first set of electrical signals is acquired during the initial operation of the EMM, i.e., the first time the EMM is operated after installation. The stator current data thus obtained may ensure that no signals related to EMM component failures are included in the stator current. In step 1202, the symmetrical components in the first set of electrical signals are dynamically removed according to the algorithms shown in equations (3) through (6), which square the instantaneous values of the first set of electrical signals for each of the plurality of phases and sum the squared values of the one or more electrical signals to form the baseline electrical signals. In step 1203, a second set of electrical signals for each phase of the multi-phase EMM is acquired during normal operation of the EMM. For example, the second set of electrical signals may be acquired continuously as part of a condition monitoring process in which the EMM may generate a signal indicative of the faulty operating condition of the EMM component. The cause of the faulty operating state may be the presence of a failure or wear of components within the EMM assembly. In step 1204, the second set of electrical signals is converted into a mixed electrical signal.

In one embodiment, the second set of electrical signals is converted to a mixed electrical signal by replacing the electrical signals in the acquired first set of electrical signals for at least one phase with the baseline electrical signals for the respective phase. Any fault occurring in the rotor assembly typically modulates the output stator electrical signal equally in all n phases. Therefore, any malfunction occurring in the rotor assembly can be detected by forming the mixed electric signal as shown in equations (7) to (8). In step 1205, the symmetric component in the mixed electric signal is eliminated as shown in equations (9) and (10). If present, the fault-related asymmetric component in the electrical signal resulting from step 1105 will be displayed in the ac frequency spectrum at twice the frequency. In step 1206, the asymmetric component in the mixed electrical signal is compared to the asymmetric component in the baseline electrical signal. In step 1207, based on the comparison, if the asymmetric component in the mixed electrical signal deviates from the asymmetric component in the baseline electrical signal, a fault signal is generated in step 1208. Otherwise, the method proceeds to step 1201 and continues to monitor the stator output signal to detect a fault.

While only certain features of the invention have been illustrated and described herein, many modifications and changes will occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

Claims (17)

1. A method of detecting a mechanical fault in an electromechanical machine, the method comprising:

acquiring an electrical signal representing the operating state of the electromechanical machine;

normalizing the electrical signal to extract spectral information; and

detecting a fault based on an analysis of the spectral information, including decomposing balanced non-fault related components into direct current quantities and unbalanced fault related components into alternating current quantities.

2. The method of claim 1, wherein detecting a fault based on the analysis of the spectral information comprises detecting a bearing fault or a gearbox fault.

3. The method of claim 1, wherein acquiring electrical signals representative of an operating state of the electromechanical machine comprises acquiring electrical signals for each of a plurality of phases of the electromechanical machine.

4. The method of claim 1, wherein the electrical signals representative of the operating state of the machine motor comprise current and voltage signals.

5. The method of claim 1, wherein current spectra in the spectral information are normalized according to at least one of: fast fourier transform, time-frequency analysis, and multi-mode regression analysis.

6. The method of claim 1, wherein normalizing the electrical signal to extract spectral information further comprises:

eliminating symmetrical components in a current spectrum in the spectrum information; and

an asymmetric component displayed in the current spectrum during an abnormal operation state is extracted.

7. The method of claim 6, wherein removing symmetrical components in the current spectrum comprises squaring current signal instantaneous values for each of a plurality of phases and summing the squared values.

8. The method of claim 6, wherein eliminating symmetric components in the current spectrum comprises eliminating symmetric components that make up an electromechanical current spectrum during normal operating conditions of the electromechanical machine.

9. The method of claim 1, further comprising:

acquiring a second set of electrical signals representative of a non-fault operating condition of the machine motor;

converting electrical signals representative of the operating condition of the machine motor into a hybrid electrical signal using the second set of electrical signals;

normalizing the mixed electrical signal to extract spectral information; and

detecting a fault based on an analysis of the spectral information.

10. The method of claim 9, wherein acquiring the second set of electrical signals representative of the non-fault operating condition of the machine motor comprises acquiring the second set of electrical signals for each of a plurality of phases of the machine motor.

11. The method of claim 9, wherein normalizing the mixed electrical signal to extract spectral information comprises:

dynamically canceling a symmetric component in the mixed electrical signal;

comparing the magnitude and phase of the asymmetric components in the mixed electrical signal with the magnitude and phase of the second set of electrical signals; and

a fault is detected when the asymmetric components in the mixed electrical signal deviate in magnitude and phase from the second set of electrical signals.

12. The method of claim 11, wherein detecting a fault comprises detecting a fault in a rotor of the electromechanical machine.

13. The method of claim 1, wherein analyzing the extracted features comprises:

the extracted features are compared to a determined threshold.

14. A system for detecting a mechanical failure of an electromechanical machine, the system comprising:

one or more sensors for acquiring electrical signals representative of the operating state of the machine motor;

a controller for normalizing the electrical signal to extract spectral information, and analyzing the spectral information, decomposing balanced non-fault related components into direct current quantities, and unbalanced fault related components into alternating current quantities.

15. The system of claim 14, wherein the fault detection unit module detects a bearing fault or a gearbox fault.

16. The system of claim 14, wherein the one or more sensors acquire electrical signals for each of a plurality of phases of the electromechanical machine.

17. The system of claim 14, wherein the controller is further configured to:

eliminating symmetrical components in a current spectrum in the spectrum information; and

an asymmetric component displayed in the current spectrum during an abnormal operation state is extracted.