Robust fault frequency component extraction of motor under variable operating conditions

Abstract

A fault detection system is provided that extracts fault features of a motor operating under variable speed or variable load conditions. The fault detection system includes a sensor interface configured to acquire a sensor signal from a sensor disposed at a predetermined position of the induction motor, where the sensor signal indicates an eccentricity level of a rotor of the induction motor; and the memory is connected with the processor. The instructions include the steps of: a sensor interface configured to acquire an operating signal of the induction motor; and a memory configured to store a computer-implemented fault detection method by extracting a fault signal in a frequency domain from a frequency spectrum formed by a minimum variance beamforming method.

Description

Technical Field

[0001] The present invention relates to a system for extracting fault features of a motor operating under variable speed or variable load conditions based on motor current signature analysis. Background Art

[0002] In the past few decades, motor current signature analysis (MCSA) has become a major method for detecting motor faults such as bearing faults, eccentricity, and broken bar faults. When any of these types of motor faults occur, the rotating flux in the air gap becomes asymmetric, thus causing additional frequency components in the stator current. The fault detection method based on MCSA aims to extract fault features in the frequency domain by analyzing the stator current.

[0003] In fact, due to the following reasons, the extraction of fault frequency components can be very challenging. First, the motor fault frequency components are usually much weaker than the operating frequency components, especially in the early stage of motor fault development. Second, weak fault features may be easily overwhelmed by background noise or interference. For example, when the motor is driven by an inverter, its fault features may be affected by harmonic interference generated by the switching operation of the power electronic device. Third, motors usually operate under variable load and variable speed conditions. The non-periodic time-varying factors will inevitably introduce spectral distortion in MCSA. Therefore, it is desirable to develop a robust fault feature extraction method for motors under variable operating conditions to effectively extract fault features from noisy measurements.

[0004] Spectrum analysis is a typical signal processing problem and is widely used in various applications. The most common spectrum analysis method in fault detection based on MCSA is the Fourier transform because of its simplicity and ease of use. This method works well in most cases when the motor under test operates in a steady state, but it is not very satisfactory for variable load operation. For variable load operation, a straightforward way is to measure multiple time series and take the average, so that the effects of noise and variable operation can be reduced. However, this method requires longer measurement time and may not be able to effectively extract small fault features. Researchers have introduced other advanced signal processing methods such as ESPRIT, MUSIC, and compressive sensing (CS) into the field of motor fault detection to achieve high-resolution spectra. However, these methods are sensitive to noise or highly dependent on the signal model. Low noise levels or small load fluctuations occurring during the measurement period may interfere with the accuracy of fault detection.

[0005] Therefore, there is a need for a robust method for extracting fault features of a motor operating under variable speed or variable load conditions. Summary of the Invention

[0006] Motor Current Signature Analysis (MCSA) has been widely used for motor fault detection, including bearing faults, broken bars, and eccentricity, etc. When the motor fault is in its early stage or when the faulty motor is operating under variable load conditions, the fault features may be submerged in the background noise and interference, making fault detection a very challenging problem. Some embodiments of the present invention provide a method and system for extracting small fault features of frequency components under variable load conditions and with strong background noise. To this end, we segment the time-domain current measurements into overlapping sequences and correspondingly consider these sequences as measurements of a linear sensor array. Then, we adopt the idea of minimum variance beamforming in array signal processing to generate a current spectrum with robust performance under varying load operations. We demonstrate our method using experimental data collected from a motor with a minor eccentricity fault operating under actual changing conditions.

[0007] Some embodiments of the present invention provide a minimum variance-based spectrum analysis method (minimum variance beamforming method) to extract the frequency components of early motor faults with variable operating conditions. The minimum variance beamforming method minimizes the noise variance of the spectrum at each frequency. Therefore, in the final spectrum, the noise-related frequency components will be suppressed, while the frequency components of the signals including the fault signals will be retained. The experimental results verify the effectiveness of this method in extracting fault features in noise measurements.

[0008] The aim of some embodiments is to provide a system and method suitable for extracting the fault features of motors operating under variable speed or variable load conditions. Inspired by the array signal processing method, we propose to use the minimum variance beamforming method for motor current signature analysis. The detailed idea is described as follows. We first segment the measured time-domain stator current into multiple overlapping time sequences with a fixed time shift. Each sequence is considered as an independent measurement of a virtual current sensor. Since the time difference between adjacent sequences is the same, these virtual sensors form a linear virtual sensor array. The spectrum analysis problem is then transformed into a beamforming problem of a linear array in the frequency domain. We borrow the idea of the minimum variance (MV) beamforming method from array signal processing techniques to improve the detection performance under noise measurements. Due to the change of the motor operating conditions, each sensor has a different and unknown frequency-dependent gain factor. The MV spectrum analysis aims to achieve a robust spectrum by using the weighted sum of the spectra of each time sequence, where the weights are frequency-dependent and optimized by minimizing the noise variance of the output spectrum.

[0009] To verify the method we proposed, we take the motor eccentricity problem as an example to conduct fault detection based on MCSA. Specifically, in our experiment, a motor with a slight eccentricity fault was considered, and a magnetic powder brake was installed together with the motor as its load. Then, the stator current of the motor was measured under various load conditions by changing the input current of the magnetic powder brake. By comparing the MCSA results of the experimental data using different methods, we demonstrated that even under strong noise interference, our method can effectively extract the fault frequency components under variable load conditions.

[0010] According to some embodiments of the present invention, there is provided a fault detection system for extracting fault characteristics of an induction motor operating under variable load or variable speed conditions. The fault detection system may include: a sensor interface configured to acquire an operation signal of the induction motor; a memory configured to store a computer-implemented fault detection method; and a digital signal processor configured to perform steps of the computer-implemented fault detection method by using the operation signal acquired via the sensor interface, wherein the steps include: segmenting the operation signal into N segments by means of a time window, wherein the segmentation is performed for each of the N segments to include an overlapping time period between adjacent segmented operation signals; converting each segmented operation signal into N frequency domains; shifting the phase of each segmented operation signal by a preset phase; performing a beamforming method on the converted segmented operation signals, wherein an optimized weight factor is applied to each converted segmented operation signal; and extracting a fault signal in the frequency domain from a spectrum formed by the minimum variance beamforming method.

[0011] In addition, some embodiments of the present invention provide a computer-implemented fault detection method for extracting fault characteristics of an induction motor operating under variable load or variable speed conditions. The method includes the following steps: acquiring an operation signal of the induction motor via a sensor interface connected to a current sensor arranged in the induction motor; segmenting the operation signal into N segments by means of a time window, wherein the segmentation is performed for each of the N segments to include an overlapping time period between adjacent segmented operation signals; converting each segmented operation signal into N frequency domains; shifting the phase of each segmented operation signal by a preset phase; performing a minimum variance beamforming method on the converted segmented operation signals, wherein an optimized weight factor is applied to each converted segmented operation signal; and extracting a fault signal in the frequency domain from a spectrum formed by the minimum variance beamforming method.

[0012] The accompanying drawings, which are included to provide a further understanding of the invention, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention. The drawings shown are not necessarily to scale, but will generally focus on illustrating the principles of the presently disclosed embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1A

[0014] Figure 1A is a schematic diagram of a fault detection system for controlling and monitoring an induction motor according to an embodiment of the present invention.

[0015] Figure 1B

[0016] Figure 1B is a schematic diagram of a fault detection system for controlling and monitoring an induction motor operated by a user (operator) via a network according to an embodiment of the present invention.

[0017] Figure 2A

[0018] Figure 2A is a diagram of a normal motor with a uniform air gap according to an embodiment of the present invention.

[0019] Figure 2B

[0020] Figure 2B is a diagram of an eccentric motor with a non-uniform air gap according to an embodiment of the present invention.

[0021] Figure 3

[0022] Figure 3 is an exemplary graph of stator current in the time domain.

[0023] Figure 4

[0024] Figure 4 shows a block diagram of a system for MCSA-based fault detection for rotor faults according to an embodiment of the present invention.

[0025] Figure 5

[0026] Figure 5 shows a block diagram of a detailed robust spectrum analysis method according to an embodiment of the present invention.

[0027] Figure 6A

[0028] Figure 6A ​​​​​​​​​​​​​​​​is an exemplary graph of the stator current spectrum under one load condition using the Fourier transform spectrum.

[0029] Figure 6B

[0030] Figure 6B is an exemplary graph of the stator current spectrum under one load condition using the average spectrum.

[0031] Figure 6C

[0032] Figure 6C is an exemplary graph of the stator current spectrum under one load condition using the minimum variance (MV) spectrum.

[0033] Figure 7A

[0034] Figure 7A shows an exemplary graph of the stator current spectrum under another load condition using the Fourier transform spectrum.

[0035] Figure 7B

[0036] Figure 7B shows an exemplary graph of the stator current spectrum under another load condition using the average spectrum.

[0037] Figure 7C

[0038] Figure 7C shows an exemplary graph of the stator current spectrum under another load condition using the minimum variance (MV) spectrum.

[0039] Figure 8A

[0040] Figure 8A shows an exemplary graph of the stator current spectrum under another load condition using the Fourier transform spectrum.

[0041] Figure 8B

[0042] Figure 8B shows an exemplary graph of the stator current spectrum under another load condition using the average spectrum.

[0043] Figure 8C

[0044] Figure 8C shows an exemplary graph of the stator current spectrum under another load condition using the minimum variance (MV) spectrum. DETAILED DESCRIPTION

[0045] ​​​​​​​​​​​​​​​​Various embodiments of the present invention will be described below with reference to the accompanying drawings. It should be noted that in all the drawings, the drawings are not drawn to scale, and elements having similar structures or functions are denoted by the same reference numerals. It should also be noted that the drawings are only intended to facilitate the description of specific embodiments of the present invention. They are not intended to be an exhaustive description of the present invention or a limitation on the scope of the present invention. In addition, aspects described in connection with specific embodiments of the present invention are not necessarily limited to that embodiment and can be practiced in any other embodiment of the present invention.

[0046] Figure 1A FIG. is a schematic diagram of a fault detection system 100 for controlling and monitoring an induction motor according to an embodiment of the present invention.

[0047] The induction motor (system) 200 includes a rotor assembly 102, a stator assembly 104, a main shaft 106, and two main bearings 108. In this embodiment, the induction motor 200 is a squirrel-cage induction motor.

[0048] According to an embodiment of the present invention, the controller 220 is powered by a power supply 230 and can be used to monitor and control the operation of the induction motor 200 in response to various inputs. For example, the controller 220 connected to the induction motor 200 can control the speed of the induction motor based on the input received from the fault detection system 100, and the fault detection system 100 is configured to obtain data related to the operating conditions of the induction motor 200 from the current sensor 150, so as to obtain the stator current data of the induction motor 200. For example, the current sensor senses the current data of one or more phases among a plurality of phases of the induction motor. More specifically, in the case where the induction motor is a three-phase induction motor, the current sensor senses the current data of the three phases of the three-phase induction motor. Although certain embodiments of the present invention will be described with respect to a polyphase induction motor, other embodiments of the present invention can be applied to other polyphase electromechanical devices.

[0049] Some embodiments of the present invention describe a system for fault detection in an electric machine, such as induction motor 200. The system configured for detection includes a fault detection circuit module 100 for detecting the presence of a fault condition in rotor 102 within the induction motor assembly, including eccentricity faults, broken bar faults, bearing faults, etc. In one embodiment, the fault detection circuit module 100 is implemented as a subsystem of controller 220. In some cases, the fault detection circuit module 100 may be referred to as a fault detection system. In an alternative embodiment, the fault detection circuit module 100 is implemented using a separate processor. The fault detection circuit module 100 may be a hardware circuit module operably connected to controller 220. In some implementations, the fault detection circuit module 100 and controller 220 may share information. For example, the fault detection circuit module 100 may reuse sensor data used by the controller to control the operation of the induction motor.

[0050] In addition, the fault detection circuit module 100 includes a processor 110, a memory 120, and a fault detection program (computer-implemented fault detection method) stored in the memory 120 when the instructions of the program are executed by the processor 110. The circuit module 100 also includes a sensor interface 130 configured to obtain signals from a sensor 150. The sensor interface 130 includes A / D (analog / digital) and A / D (analog / digital) converters for data communication with the processor 110, the memory 120, the fault detection program, the user interface 140, and the current sensor 150. The processor 110 is a digital signal processor designed to perform the steps of the fault detection program while obtaining a digital signal of the stator current (motor current) of the induction motor from the current sensor 150 via the sensor interface 130.

[0051] The processor 110 may be multiple processors, and the memory 120 may be a memory module including multiple memories. The user interface 140 is configured to be connected to a keyboard and a display unit, and the display unit is configured to indicate the normal / fault status information of the induction motor 200 in response to the output of the fault detection module 200.

[0052] Figure 1BIt is a schematic diagram of a fault detection system for controlling and monitoring an induction motor operated by a user (operator) via a network 250 according to an embodiment of the present invention. In this case, the fault detection module 100 may be included in the operating system of the induction motor located at the user / operator site. When the fault detection module 100 determines a fault based on the sensor signals of the sensors 150 via the network (communication network) 250 during the operation of the induction motor (induction motor) 200, the fault detection module 100 may send a control signal via the network 250 based on a pre-programmed algorithm (not shown in the figure) stored in the memory 120 to slow down or stop driving the induction motor 200. In some cases, the calibration of the current sensor 150 may be performed based on a sensor calibration program (not shown in the figure) stored in the memory of the fault detection module 100. In some cases, the network 250 may be an optical fiber network, a wireless network, an Internet network, or a data communication network composed of at least two combinations of an optical fiber network, a wireless network, and an Internet network.

[0053] This configuration may be a maintenance system managed by the user / client, and the user / client operates an induction motor system 20 located separately from the induction motor site. For example, this system configuration may be used for a power generation system operated by the user and for a transmission system that controls an induction motor of a drive train. In some cases, when the fault detection module 100 detects the separation between the induction motor site and the fault detection system site, data communication is performed between the induction motor site and the fault detection system site via the network. The network 250 may be an optical fiber network, a wireless network, an Internet network, or a data communication network composed of at least two combinations of an optical fiber network, a wireless network, and an Internet network.

[0054] In addition, the fault detection system is included as part of the user's maintenance system. When the determined eccentricity level of the induction motor is equal to or greater than the critical threshold level, the fault detection system uses the sensor interface / control interface 130 to send a control signal to the controller of the induction motor via the network to stop the operation of the induction motor.

[0055] In one embodiment of the present invention, current and voltage sensors respectively detect stator current data from the stator assembly 104 of the induction motor 200. The current data obtained from the current sensor is transmitted to the inverter for control and to the fault detection module for further processing and analysis. The analysis includes performing motor current signature analysis (MCSA) to detect faults within the induction motor 200. In some embodiments, when a fault is detected by using the fault detection module 100, the controller 220 receives a fault detection signal via the interface 130 of the fault detection module 100, and stops the operation of the induction motor by sending an interruption signal of the stator current of the induction motor 200 to the controller 110 for further inspection or repair. In some cases, the current sensor 150 may include a controller interface (not shown), which is configured to receive the fault detection signal from the interface 130 and send a fault status signal to the controller 220, such that the controller 220 interrupts the stator current of the induction motor 200 to stop the operation of the induction motor 200. When the current sensor 150 does not include a controller interface, the interface 130 may be configured to connect to the controller 220, such that the controller interrupts the stator current of the induction motor 200 in response to the fault detection signal from the fault detection circuit module 100 via the interface 130 to stop the operation of the induction motor 200.

[0056] The system further includes a memory for storing the measurement results of the signals and various parameters and coefficients for performing the fault severity detection method.

[0057] Figure 2A and Figure 2B are diagrams of a normal motor with a uniform air gap and a faulty motor with an eccentricity fault indicated by a non-uniform air gap, respectively, according to one embodiment of the present invention. Based on the physical model of the induction motor of the inventors and the fault detection methods using different features, the inventors aim to extract the fault features of the motor operating under variable speed or variable load conditions.

[0058] Motor current signature analysis (MCSA) has been widely used for motor fault detection for decades due to its effectiveness and non-invasive characteristics. Based on the stator current spectrum, the MCSA method aims to extract the characteristic frequency components of different types of faults.

[0059] For example, when there is a broken bar fault in a squirrel-cage induction motor, in addition to the operating frequency components, a new set of frequency components appears in the spectrum of the stator current.

[0060] f bar =(1±2ks)f s , (1)

[0061] where k = 1, 2,...; s is the slip ratio; and f sis the fundamental frequency of the power supply. The broken bar fault detection technology based on MCSA focuses on detecting the dominant frequency component in the stator current, i.e., f b1 =(1 - 2s)f s .

[0062] For bearing faults, depending on the fault location or type in the bearing, some periodic vibration impulses are generated due to impacts, thus inducing characteristic frequency f c in the stator current. For the above different types of bearing faults, f c is listed as follows.

[0063] Cage defect impacts outer race:

[0064]

[0065] Cage defect impacts inner race:

[0066]

[0067] Outer race defect impacts ball:

[0068]

[0069] Inner race defect impacts ball:

[0070]

[0071] Ball defect impacts both races:

[0072]

[0073] where f r is the mechanical frequency of the rotor, d is the ball diameter, D is the pitch or cage diameter, N b represents the number of balls, and θ is the contact angle characterizing the contact point between the ball and the raceway.

[0074] For the eccentricity fault in most induction motors, the fault characteristic frequency in the current signal is

[0075] where R is the number of rotor slots, p is the number of pole pairs, k is any positive integer, n d is the eccentricity order (n d = 0 in the case of static eccentricity, n d = 1, 2, 3,...) in the case of dynamic eccentricity, and ν is the order of the stator time harmonics. In the absence of the number of rotor slots, the simplified version is given by:

[0076]

[0077] Among them, is the rotor frequency related to the rotational speed.

[0078] Therefore, for most motor fault detection problems, the purpose of the MCSA-based method is to extract the corresponding fault feature components via effective spectrum analysis. Once a fault frequency component exceeding a certain threshold is detected, the corresponding fault can be determined. The fault severity level can be further estimated based on the amplitude of the fault frequency component and other operating conditions.

[0079] Spectrum analysis is a typical problem in signal processing. Let i s (t) represent the time-domain stator current of the motor in ideal steady-state operation. Note that the current can be a single-phase current or a combination of three-phase currents after appropriate phase alignment such as Park transformation. For a normal motor, the time-domain stator current can be expressed as

[0080] i sh (t) = I s cos(2πf s t + φ s ); (9)

[0081] where, I s is the stator current amplitude related to the motor load condition, f s is the operating frequency, and φ s is the initial phase of the operating frequency component.

[0082] When there is a motor fault, the motor current signal includes extracted frequency components.

[0083]

[0084] where, I f,k represents the amplitude of the k-th fault frequency component f f,k , and φ f,k is the initial phase of the k-th fault frequency component.

[0085] The spectrum of the stator current i s (t) (depending on the normal condition, = i sh (t) or i sf (t)) can be achieved by Fourier transform as

[0086] I s (ω) = ∫i s (t)e -jωt dt. (11)

[0087] For periodic signals in motor operation, the discrete Fourier transform (DFT) is commonly used to calculate the Fourier spectrum based on discrete-time sampling n. We ignore the detailed correspondence between the frequency and the sampling rate and simplify the expression of the Fourier spectrum to

[0088]

[0089] Note that compared with the operating signal, the fault feature component is usually very weak, and the f,k value of I can be 40 dB to 60 dB lower than the value of Is, depending on the fault type and fault severity. Therefore, it is very common for the fault feature component to be interfered by the operating signal or other noises, especially under variable operating conditions.

[0090] Figure 4 FIG. shows a block diagram of a computer-implemented fault feature / signal detection method / program 121 for MCSA-based fault detection of rotor faults, and Figure 5 FIG. shows a block diagram of a detailed robust spectrum analysis method according to some embodiments of the present invention. The computer-implemented fault feature / signal detection method 121 is stored in a memory 120 or a storage device (not shown) and is configured to be run by a processor 110 when the system performs fault detection on the induction motor 200.

[0091] Under variable load conditions, the stator current amplitude I s is no longer constant but varies with the load condition. In this case, the performance of fault detection via simple DFT will degrade because small fault components may be submerged in the background noise. To effectively analyze the stator current spectrum, we introduce a parameter α(t) to reflect the influence of variable load and express the actual measurement of the time-domain stator current 415 as

[0092]

[0093] where α(t) is a time-domain function showing the influence of variable load, v(t) is the measured noise, and is the convolution operation.

[0094] In the frequency domain, we have

[0095] I(ω) = A(ω)S(ω) + V(ω); (14)

[0096] where I(ω) and S(ω) represent the Fourier spectra of i(t) and i s (t) respectively. Then, our goal is to estimate S(ω) = 425 from the measured i(t) = 415 through spectrum analysis 420 so that the fault features (if any) in S(ω) = 425 can be effectively extracted.

[0097] A direct way to estimate S(ω) is to average multiple spectra to reduce the effect of variable loading. Suppose there is a series of N equally long time series {i1(t), i2(t), …, i n (t), …, i N (t)}520. The spectrum 530 of each series can be expressed as

[0098] I n (ω) = ∫i n (t)e -jωt dt ≈ A n (ω)S n (ω) + V n (ω). (15)

[0099] Note that A n (ω) is the Fourier spectrum of α n (t) due to variable loading as described above. In particular, when the motor is operating in a steady state with a constant load and constant speed, A n (ω) is a constant. Otherwise, A n (ω) varies due to variable operation. The average spectrum is given by:

[0100]

[0101] To utilize array signal processing techniques, we generate N time series {i1(τ), i2(τ), …, i n (τ), …, i N (τ)} by placing a sliding time window on the measured current i(t), where the nth time series i n (τ) can be expressed as

[0102] τ ∈ [0; t w , n = 1, …, N of i n (τ) = i(t + (n - 1)t s ) (17)

[0103] Here, t w is the window size in time, and t s is the time step between two consecutive time windows. The spectrum of each segment can be represented by the spectrum of the first sequence with an appropriate phase shift as

[0104]

[0105] Using appropriate phase compensation we combine all N spectra to form the stator current spectrum 560 in the frequency domain, and the stator current spectrum 560 is

[0106]

[0107] where the weight β n (ω)550 is optimized by the minimum variance beamforming method (i.e., by minimizing the noise variance of the spectrum at each frequency).

[0108]

[0109] Let the weight β n (ω)550 be the vector

[0110] w = [β1(ω), …, β n (ω), …, β N (ω)] ∈ R N×1 , (21)

[0111]

[0112] and a = [1, …, 1, …, 1] T ∈ R N×1 . (23)

[0113] Then, the optimization problem in (20) is reformulated as a standard minimum variance beamforming problem.

[0114]

[0115] The closed - form solution of (20) is given in terms of 550 as

[0116]

[0117] Note that w is frequency - dependent, which means that for each frequency, we solve (25) to obtain different w. Once the estimated spectrum S(ω)=425 is obtained, the fault features 430 are extracted at different frequencies according to (1)-(8). If the magnitude of the fault feature in 440 is greater than a specific value, for example - 70 dB compared to the operating signal of the early fault, the computer - implemented fault detection method / program generates a fault feature code indicating the detected fault (fault feature). In this case, the fault feature code indicates one of the severity levels of the fault based on the magnitude of the fault feature. For example, the severity level is determined by three levels corresponding to the normal level (< - 70 dB), the warning level (≈ - 60 dB), and the severe (critical) level (> - 40 dB). Then, the corresponding control signal 450 is sent to the controller 220 to control the induction motor 200. If no fault is detected, the fault detection system 100 is configured to keep receiving / acquiring the operating current (stator current 410) of the induction motor 200 from the current sensor 150 via the sensor interface 130 to keep monitoring the stator current 410. While acquiring the operating current 410 from the current sensor 150 via the sensor interface 130, the fault detection system (circuit module) 100 keeps generating the status data of the severity level of the induction motor and sending it to the user interface 140, which enables the display unit to display the message information (status information) of the severity level. In some cases, the operation uses a keyboard connected to the user interface 140 to send a control signal to the controller 220 via the network 250 to control the rotational speed of the induction motor 220 in response to the displayed status information of the severity level of the induction motor 220.

[0118] In addition, when the severity level of the induction motor is equal to or greater than the critical threshold level, the fault detection system sends a control signal to the controller 220 of the induction motor 200 via the network 250 to reduce the rotational speed of the induction motor 200 or stop operating the induction motor 200.

[0119] To verify our proposed method, an MCSA - based analysis is carried out taking the eccentricity fault as an example. An experiment is conducted using a motor with an eccentricity fault in the laboratory. To generate the eccentricity of the rotor, two bearings supporting the rotor are removed and replaced by two external bearings fixed outside the motor, so that the air - gap can be manually adjusted within a certain range. Four gap sensors are placed in the stator component to monitor the horizontal and vertical air - gaps at both ends, thus ensuring the accuracy of the adjustment. A magnetic powder brake is used as the load, and its torque can be adjusted by changing its input operating current. A diagram of our experimental setup is shown in Figure 1A and is shown in. During operation, the three - phase time - domain stator currents are recorded for further analysis.

[0120] To simulate the early eccentric fault, we set the eccentricity level to 3%, where the eccentricity level is defined by the following formula:

[0121]

[0122] where d is the distance between the actual rotor axis and the ideal rotor, and δ0 is the average air-gap length in a normal motor. When d = 0, meaning the rotor and stator are coaxial, the motor is normal, as shown in Figure 2A When d ≠ 0, meaning the rotor and stator are not coaxial, the motor fails, as shown in Figure 2B Experiments were conducted under various load conditions by adjusting the input operating current of the magnetic powder brake.

[0123] The fault detection circuit module 100 is configured to collect the stator current for 60 s at a sampling rate f sa = 10 = kHz for different load conditions. Figure 3 Two embodiments of the time-domain stator current data collected when we have a constant load and a variable load respectively are plotted. We observe that the current amplitude is relatively flat under constant load conditions, while it changes significantly under variable load conditions. Extracting fault features from the stator current with varying amplitudes is very meaningful but challenging. To extract the fault frequency components, we consider various spectral analysis methods. To implement multi-sequence data, we segment the 60-second time series data into t s = 0.02 s time steps of t w = 4 s sequences.

[0124] Figure 6A and Figure 6B and Figure 6C respectively show exemplary graphs of the stator current spectra using the Fourier transform spectrum average spectrum and the minimum variance (MV) spectrum under one load condition; Figure 7A and Figure 7B and Figure 7C respectively show exemplary graphs of the stator current spectra using the Fourier transform (FT) spectrum average spectrum and the minimum variance (MV) spectrum under another load condition; and Figure 8A and Figure 8B and Figure 8C respectively show exemplary graphs of the stator current spectra using the Fourier transform spectrum average spectrum and the minimum variance (MV) spectrum under another load condition.

[0125] We can observe that the FM spectrum fails to detect the fault characteristics under variable load conditions. Although the average spectrum reduces noise to some extent and works well in some cases, it is not consistently so. The MV spectrum implemented by the method we proposed successfully detects the fault characteristic components consistently for different variable loads. In our case, the frequency components of eccentricity are 30 Hz and 90 Hz. In other words, the frequency range of the MV spectrum is set to indicate the fault characteristics at the frequency components of 30 Hz and 90 Hz of eccentricity.

[0126] The above embodiments of the present invention can be implemented in any of a variety of ways. For example, the embodiments can be implemented using hardware, software, or a combination thereof. When implemented in software, the software code can be executed on any suitable processor or set of processors (whether provided in a single computer or distributed among multiple computers). Such a processor can be implemented as an integrated circuit, and one or more processors are among the integrated circuit components. However, the processor can be implemented using any suitable format of circuitry.

[0127] In addition, the embodiments of the present invention can be embodied as a method, and embodiments of the method have been provided. The actions taken as part of the method can be sequenced in any suitable manner. Thus, the embodiments can be configured such that the actions are performed in a different order than shown, which can include performing some actions simultaneously, even if those actions are shown as sequential actions in the illustrative embodiments.

[0128] The use of ordinal terms such as "first", "second", etc. in the claims to modify the claim elements themselves does not mean any priority, precedence, or order of one claim element with respect to another, nor does it mean the order in time of the acts of a method, but is only used as a label to distinguish one claim element having a particular name from another element having the same name (but using an ordinal term) to distinguish the claim elements.

[0129] Although the present invention has been described by way of examples of the preferred embodiments, it should be understood that various other adaptations and modifications can be made within the spirit and scope of the present invention.

[0130] Therefore, the purpose of the appended claims is to cover all such changes and modifications that fall within the true spirit and scope of the present invention.

Claims

1. A fault detection system for extracting fault features of an induction motor operating under variable load or variable speed conditions, comprising: A sensor interface configured to acquire the operating signals of the induction motor; A memory configured to store a computer-implemented fault detection method; A digital signal processor configured to perform the steps of the computer-implemented fault detection method by using the operating signals acquired via the sensor interface, wherein the steps include: Segmenting the operating signals into N segments by means of a time window, wherein the segmentation is performed for each of the N segments to include an overlapping time period between adjacent segmented operating signals; Converting each segmented operating signal into N frequency domains; Shifting the phase of each segmented operating signal by a preset phase; Performing a beamforming method on the converted segmented operating signals, wherein an optimized weighting factor is applied to each of the converted segmented operating signals; and Extracting a fault signal in the frequency domain from the spectrum formed by the minimum variance beamforming method.

2. The fault detection system according to claim 1, wherein, The sensor interface iteratively acquires the operating signals within a predetermined time period.

3. The fault detection system according to claim 1, wherein The eccentricity-related frequency domain is determined to be at approximately 1 / 2 of the operating frequency of the induction motor and at approximately 3 / 2 of the operating frequency.

4. The fault detection system according to claim 1, wherein, The digital signal processor collects the operation signal for 60 seconds at a sampling rate f sa = 10 kHz or at least twice the highest frequency of the signal.

5. The fault detection system according to claim 1, wherein The operating signal is the stator current of the induction motor.

6. The fault detection system according to claim 1, wherein, The extracted spectrum is a minimum variance (MV) spectrum.

7. The fault detection system according to claim 6, wherein, The frequency range of the MV spectrum is set to indicate the fault features at 30 Hz and 90 Hz of the frequency components indicating eccentricity.

8. The fault detection system according to claim 1, wherein The site of the induction motor is separated from the site of the fault detection system, wherein data communication between the site of the induction motor and the site of the fault detection system is carried out via a network, and the network is an optical fiber network, a wireless network, an Internet network or a data communication network composed of at least two combinations among the optical fiber network, the wireless network and the Internet network.

9. The fault detection system according to claim 8, wherein, When the severity level of the induction motor is equal to or greater than the critical threshold level, the fault detection system sends a control signal to the controller of the induction motor via the network to reduce the speed of the induction motor or stop operating the induction motor.

10. A computer-implemented fault detection method for extracting fault features of an induction motor operating under variable load or variable speed conditions, the computer-implemented fault detection method comprising the following steps: Acquiring the operating signals of the induction motor via a sensor interface connected to a current sensor arranged in the induction motor; Segmenting the operating signals into N segments by means of a time window, wherein the segmentation is performed for each of the N segments to include an overlapping time period between adjacent segmented operating signals; Converting each segmented operating signal into N frequency domains; Shifting the phase of each segmented operating signal by a phase related to frequency and sequence time shift; Performing a minimum variance beamforming method on the converted segmented operating signals, wherein an optimized weighting factor is applied to each of the converted segmented operating signals; and Extract the fault signal in the frequency domain from the spectrum formed by the minimum variance beamforming method.

11. The computer-implemented fault detection method according to claim 10, wherein, The sensor interface iteratively acquires the operation signal within a predetermined time period.

12. The computer-implemented fault detection method according to claim 10, wherein, The eccentricity-related frequency domain is determined to be at approximately 1 / 2 of the operating frequency of the induction motor and at approximately 3 / 2 of the operating frequency.

13. The computer-implemented fault detection method according to claim 10, wherein, A method using a digital signal processor collects the operation signal for 60 seconds at a sampling rate f sa = 10 kHz or at least twice the highest frequency of the signal.

14. The computer-implemented fault detection method according to claim 10, wherein, The operation signal is the stator current of the induction motor.

15. The computer-implemented fault detection method according to claim 10, wherein, The extracted spectrum is the minimum variance (MV) beamforming spectrum.

16. The computer-implemented fault detection method according to claim 15, wherein, The frequency range of the MV spectrum is set to the fault characteristics at 30 Hz and 90 Hz of the frequency components indicating eccentricity.

17. The computer-implemented fault detection method according to claim 10, wherein, The site of the induction motor is separated from the site of the fault detection system, wherein data communication between the site of the induction motor and the site of the fault detection system is carried out via a network, and the network is a fiber optic network, a wireless network, an Internet network, or a data communication network composed of at least two combinations among the fiber optic network, the wireless network, and the Internet network.

18. The computer-implemented fault detection method according to claim 17, wherein, When the severity level of the induction motor is equal to or greater than the critical threshold level, the fault detection system sends a control signal to the controller of the induction motor via the network to reduce the rotational speed of the induction motor or stop operating the induction motor.

Images