A method for fault diagnosis and physical positioning of rolling bearings based on time-frequency phase characteristics
By using multi-channel vibration sensors and cross-spectral analysis, the problems of missed and false alarms of early-stage minor bearing faults have been solved, enabling precise fault location and improving the reliability of diagnosis and maintenance guidance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUXI BRACH 703TH RES INST OF CHINA SHIPBUILDING IND CORP
- Filing Date
- 2026-03-11
- Publication Date
- 2026-06-16
Smart Images
Figure CN122220832A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of rotating machinery condition monitoring technology, and relates to a method for fault diagnosis and physical location of rolling bearings based on time-frequency phase characteristics. Specifically, it relates to a method for early and weak fault diagnosis of rolling bearings and precise location of physical faults using multi-channel vibration signals. Background Technology
[0002] Rolling bearings are core supporting components in rotating machinery, but they are also the weakest links most prone to failure. Accurate and timely fault diagnosis of bearings, especially early warning of faults, is of paramount importance for ensuring safe equipment operation, reducing maintenance costs, and preventing major accidents.
[0003] Current mainstream bearing fault diagnosis technologies typically rely on vibration signals collected by a single accelerometer mounted on the bearing housing. The basic principle is that when a local defect occurs in a bearing component (such as the inner or outer rings, or rolling elements), the rolling elements periodically roll over the defect, generating a series of weak impact pulses. The repetition frequency of these impact pulses is the "characteristic frequency" of this type of fault. Therefore, the core of existing technologies is to use various signal processing methods (such as envelope demodulation, Fast Fourier Transform (FFT), and wavelet transform) to find the energy peak corresponding to the theoretical fault characteristic frequency on the spectrum.
[0004] However, these technologies face a common bottleneck in diagnosing early, subtle faults:
[0005] Extremely low signal-to-noise ratio leads to missed detections: The impact energy at the incipient stage of a fault is extremely weak, completely submerged in the strong background noise from gear meshing, fluid impact, and structural resonance. In the frequency spectrum, its energy peak may be far below the noise level, making it ineffective for diagnostic methods and resulting in missed fault detections.
[0006] Phase information loss leads to false alarms: Traditional power spectrum analysis methods (such as FFT) only focus on the amplitude information (energy magnitude) of the signal, while completely ignoring the phase information. This makes it impossible to fundamentally distinguish between weak real fault signals and random noise with similar energy, resulting in a high false alarm rate.
[0007] The ambiguity of the fault source leads to diagnostic uncertainty: this is an inherent physical defect of single-point measurement methods. When a sensor detects an impact signal, it cannot physically determine with 100% certainty that the impact originated inside the bearing being tested. Random knocks from the outside, vibrations transmitted from other components (such as hydraulic systems), can all generate interference peaks in the frequency spectrum similar to the fault's characteristic frequency. This "uncertainty about the signal source" is the core pain point leading to frequent false alarms and poor reliability in existing diagnostic systems. Furthermore, even if the fault is confirmed, the specific physical location of the fault cannot be determined, causing significant inconvenience to maintenance work. Summary of the Invention
[0008] This invention aims to solve the following problems existing in current bearing fault diagnosis technology: 1) Due to the low signal-to-noise ratio and neglect of phase information, it is difficult to detect early and weak faults, resulting in a high false alarm rate. 2) Because it cannot distinguish between real fault impacts and external interference impacts, and cannot confirm the uniqueness of the fault source, it results in a high false alarm rate. 3) It can only determine the fault type, but cannot provide the specific physical location of the fault point on the bearing raceway, resulting in poor maintenance guidance.
[0009] To address the aforementioned problems and technical requirements, the inventors have proposed a method for rolling bearing fault diagnosis and physical location based on time-frequency phase characteristics. This method deploys at least two vibration sensors on the same bearing housing and creatively utilizes cross-spectral analysis to extract key spatial phase features, thus constructing a complete technical process for bearing fault diagnosis and physical location. The technical solution of this invention is as follows: A method for fault diagnosis and physical location of rolling bearings based on time-frequency phase characteristics includes the following steps: Step 1: Multi-channel synchronous data acquisition and signal enhancement processing Vibration time-series signals at at least two different circumferential positions on the monitored bearing housing are acquired and subjected to time-domain impact feature enhancement processing to obtain two enhanced impact signals.
[0010] Step 2: Perform cross-spectral analysis on the two impact signals, calculate and extract the spatial phase characteristic spectrum, which includes: The coherence function spectrum is used to quantify the degree of linear correlation between two impulse signals at each frequency point in the frequency domain.
[0011] Cross-phase spectrum is used to provide the phase difference between one impulse signal and another impulse signal at various frequency points in the frequency domain.
[0012] Step 3: Fault diagnosis decision based on the "three-in-one" physical evidence chain Based on the peak energy of the impact signal, the coherence function spectrum, and the cross-position spectrum, fault diagnosis conditions are set, the true fault characteristic frequency that meets the conditions is found, and compared with the theoretical values of the fault characteristic frequencies of various rolling bearings. The rolling bearing fault type corresponding to the closest theoretical value is output as the fault type corresponding to the true fault characteristic frequency.
[0013] Step 4: Precise circumferential fault location based on spatial phase difference Based on the geometric relationship between the fault point and the two vibration time sequence signal acquisition points on the monitored bearing housing, a circumferential angle calculation model is established. The actual fault characteristic frequency, the stable phase difference extracted at the actual fault characteristic frequency, and the circumferential position of the vibration time sequence signal acquisition point are substituted into the model to calculate the circumferential position of the fault point on the monitored bearing housing.
[0014] The further technical solution is that step one also includes: Install a speed sensor on the bearing housing being monitored and align it with the mark on the rotor; The absolute phase reference of the rotor provided by the speed sensor is used to calibrate and verify the fault location results.
[0015] Preferably, the circumferential positions of the two vibration timing signal acquisition points are distributed along the orthogonal directions of the monitored bearing housing, such as horizontal and vertical.
[0016] The further technical solution is that the time-domain impact feature enhancement processing in step one includes, for each vibration time-series signal: Adaptive signal decomposition technology is used to decompose the vibration time series signal into a series of intrinsic mode functions (IMF) components; Based on the statistical indicators characterizing the impact of the signal, the IMF component with the largest statistical indicator is selected from each component. This component best characterizes the impact component and is used as the enhanced impact signal for subsequent cross-spectral analysis. Among them, the statistical indicators are kurtosis value, energy entropy, or correlation coefficient.
[0017] The further technical solution is as follows: Step two specifically includes: After performing Fourier transform on the two impact signals, their respective self-power spectra and cross-power spectra are calculated. The coherence function spectrum is calculated based on the self-power spectrum and the cross-power spectrum. The spectral value close to 1 indicates that the two impact signals are highly correlated at the corresponding frequency points and originate from the same fault point; The mutual potential spectrum is calculated based on the mutual power spectrum. When the spectral value is stable, it indicates that the phase difference of the periodic shock wave originating from the same fault point at the corresponding frequency point is stable. The stable phase difference directly reflects the time difference of the shock wave propagating from the fault point to two different acquisition points.
[0018] The further technical solution is that, in step three, fault diagnosis conditions are set based on the peak energy of the impact signal, the coherence function spectrum, and the cross-position spectrum, and the true fault characteristic frequencies that meet the conditions are identified. Specifically, this includes: For any suspected fault characteristic frequency A frequency is identified as a true fault characteristic frequency when all of the following fault diagnosis conditions are met simultaneously. : Energy significance: in At this point, the peak energy values of both impact signals were higher than the preset decibel value of the background noise; Uniqueness of source: At this point, the spectral values of the coherence function spectra of the two impact signals are not less than a preset threshold; Position stability: At this point, the spectral values of the mutual position spectra of the two impact signals fluctuate within a range less than the preset angle over a period of time, while... The harmonics also exhibit a stable phase difference relationship corresponding to the fundamental frequency.
[0019] The further technical solution is that, in step four, a circumferential angle calculation model is established based on the geometric relationship between the fault point and the two vibration time-series signal acquisition points on the monitored bearing housing, including: Based on the fault point and two vibration timing signal acquisition points At the circumferential position on the monitored bearing housing, the fault point is calculated to... The difference in arc length between points is expressed as: ; Based on the propagation of the shock wave from the fault point to... Time difference of points and the propagation speed of the shock wave in the monitored bearing housing material Calculate the arrival time of the shock wave The difference in path length between the points; Solve for the case where the path length difference equals the arc length difference. The value is used as the circumferential position of the fault point on the monitored bearing housing.
[0020] The beneficial technical effects of this invention are: 1) It enables physical tracing of faults, fundamentally solving the problem of false alarms. This invention innovatively uses the "coherence function" as key physical evidence for determining the homology of signals and incorporates it into a rigorous "three-in-one" fault diagnosis decision logic. This enables the method to distinguish between genuine internal bearing faults and external random interference from a physical perspective, fundamentally solving the problem of false alarms caused by unclear signal sources in traditional methods, and greatly improving the reliability and accuracy of diagnosis.
[0021] 2) Significantly improves the detection sensitivity of early-stage minor faults. The fault diagnosis decision-making of this invention no longer relies solely on weak and unreliable energy peaks. Even in the early stages of a fault, when the energy of the original vibration signal is far lower than the background noise, it can be accurately captured after the time-domain impact characteristics are enhanced, as it possesses two deterministic spatial phase characteristics: "significant energy" and "unique source" and "stable location." This significantly lowers the detectable threshold of the fault, achieving true early warning.
[0022] 3) A breakthrough was achieved from "fault diagnosis" to "precise physical location". This invention creatively utilizes "mutual position spectrum" information (stable phase difference) and combines it with the physical layout of sensors to successfully transform temporal phase information into spatial physical location information. It can output precise location reports such as "peeling exists in the outer 315° direction," providing unprecedented high-value decision-making basis for condition-based maintenance and root cause analysis of equipment, and has significant engineering application value.
[0023] 4) The method is robust and has excellent anti-interference ability. Decision-making through a "three-in-one" chain of evidence makes diagnostic results insensitive to fluctuations in a single feature. Even when energy features are not obvious or are subject to strong noise interference, as long as the spatial phase features are stable, accurate judgments can still be made, demonstrating extremely strong anti-interference capabilities and environmental adaptability. Attached Figure Description
[0024] Figure 1 This is a flowchart of the rolling bearing fault diagnosis and physical location method based on time-frequency phase characteristics provided in this application; Figure 2 This is a diagram showing the arrangement of the vibration sensor provided in this application on the monitored bearing housing; Wherein: 1-the housing of the monitored bearing housing, (0)-the bearing spindle, A-vibration sensor A (0°), B-vibration sensor B (90°), C-the fault point. Detailed Implementation
[0025] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.
[0026] Please refer to Figure 1 As shown, one embodiment of this application provides a method for fault diagnosis and physical location of rolling bearings based on time-frequency phase characteristics, specifically including the following steps: Step 1: Obtain vibration time-series signals at at least two different circumferential positions on the monitored bearing housing, and perform time-domain impact feature enhancement processing to obtain two enhanced impact signals.
[0027] In this step, at least two different circumferential positions are selected on the bearing housing housing being monitored as signal acquisition points, and vibration sensors are installed at each point to synchronously acquire vibration timing signals. The precise physical installation angle (circumferential angle) of each sensor is recorded. In this embodiment, the circumferential positions of the two signal acquisition points are distributed along orthogonal directions of the bearing housing being monitored, such as the horizontal and vertical directions.
[0028] Optionally, three or more signal acquisition points can be set, and the layout can be designed as a ring to improve the accuracy and robustness of physical positioning, and even three-dimensional positioning.
[0029] Step 2: Perform cross-spectral analysis on the two impact signals to calculate and extract the key spatial phase characteristic spectra.
[0030] This step is the core technology of this invention, aiming to extract key evidence capable of physical tracing and localization. Cross-spectral analysis is performed on the impact signals (enhanced) from two acquisition channels. The spatial phase characteristic spectrum to be extracted includes the coherence function spectrum (…). ) and mutual position spectrum ( ).in, The function quantifies the degree of linear correlation between two impulse signals at each frequency point in the frequency domain. Spectral lines provide the phase difference between one impulse signal and another at various frequency points in the frequency domain.
[0031] Step 3: Set fault diagnosis conditions based on the peak energy of the impact signal, the coherence function spectrum, and the cross-position spectrum. Find the true fault characteristic frequency that meets the conditions and compare it with the theoretical value of the fault characteristic frequency of various rolling bearings. Output the rolling bearing fault type corresponding to the closest theoretical value as the fault type corresponding to the true fault characteristic frequency.
[0032] Among them, rolling bearing failure types include outer ring failure, inner ring failure, and rolling element failure. The theoretical values of the failure characteristic frequencies of various rolling bearings can be calculated based on bearing parameters (such as the number of rolling elements, rolling element diameter, and bearing pitch circle diameter) and real-time rotational speed (to obtain the shaft's rotational frequency). The calculation process will not be described in detail here.
[0033] Step 4: Based on the geometric relationship between the fault point and the two vibration time sequence signal acquisition points on the monitored bearing housing, establish a circumferential angle calculation model. Substitute the actual fault characteristic frequency, the stable phase difference extracted at the actual fault characteristic frequency, and the circumferential position of the vibration time sequence signal acquisition point into the model to calculate the circumferential position of the fault point on the monitored bearing housing.
[0034] The solution model is based on the physical principle that "phase difference corresponds to propagation time difference", which transforms the phase information in the time dimension into the physical location information in the spatial dimension.
[0035] In step one, the acquired raw vibration signal typically contains a large amount of background noise and non-impact components, which severely affects the identification of early weak impact characteristics. To improve the sensitivity to these weak impacts and effectively suppress background noise, this embodiment modifies each vibration timing signal channel (e.g., signal...) and Independent time-domain impact feature enhancement processing is performed. In one possible implementation, for each vibration time-series signal, the enhancement process includes: using adaptive signal decomposition techniques such as Variational Mode Decomposition (VMD) to decompose the vibration time-series signal into a series of Intrinsic Mode Function (IMF) components. Then, based on statistical indicators that characterize the signal's impact, such as kurtosis, energy entropy, or correlation coefficient, the IMF component with the largest statistical indicator is selected from each component. This component best characterizes the impact component and is used as the enhanced impact signal for subsequent cross-spectral analysis. The time-domain impact feature enhancement processing aims to "purify" the original signal, filtering out most steady-state noise and non-impact interference, significantly amplifying the potential periodic impact features, and laying the foundation for subsequent spatial phase feature extraction.
[0036] In this embodiment, step one further includes: installing a speed sensor (such as a keyway sensor) on the monitored bearing housing, with the sensor probe aligned with a mark on the rotor (such as a keyway or reflective mark) to synchronously acquire the rotor's real-time speed and absolute phase reference. The absolute phase reference refers to a fixed reference point used to measure the absolute phase of rotor vibration. The speed sensor detects a specific mark on the rotor, generating a pulse signal each time the mark passes, marking the zero-degree position of the rotor rotation. The absolute phase reference provided by the speed sensor is used to calibrate and verify the fault location results calculated in step four, ensuring the accuracy of the location.
[0037] The reason for extracting the spatial phase feature spectrum in step two is to consider that a real periodic impact originating from a fixed physical point has a highly coupled and stable phase relationship between its fundamental frequency and its harmonics, while the phase of background noise is random and uncorrelated. In one possible implementation, step two specifically includes the following: After performing Fourier transforms on the two impact signals, their individual power spectra and cross-power spectra are calculated. The coherence function spectrum is then calculated based on the individual and cross-power spectra. Its range is , A value close to 1 (preferably 1, but can also be greater than or equal to a preset threshold) indicates that the two impact signals are highly correlated at the corresponding frequency points and originate from the same fault point (physical source), which is strong physical evidence that the two signals originate from the same stable physical source.
[0038] The mutual potential spectrum is calculated based on the mutual power spectrum. When the spectral value is stable, it indicates that the phase difference of the periodic shock wave originating from the same fault point at the corresponding frequency point is stable and definite. This stable phase difference directly reflects the time difference of the shock wave propagating from the fault point to two different acquisition points, which is the direct basis for physical positioning.
[0039] Step three establishes a "three-in-one" fault diagnosis decision logic to confirm the authenticity of the fault from a physical perspective. In one possible implementation, fault diagnosis conditions are set based on the peak energy of the impact signal, the coherence function spectrum, and the cross-spectrum, to identify the true fault characteristic frequencies that meet these conditions. Specifically, this includes: For any suspected fault characteristic frequency A frequency is identified as a true fault characteristic frequency when all of the following fault diagnosis conditions are met simultaneously. : (1) Energy significance: in When the energy peaks of both impact signals are higher than the preset decibel value of the background noise, they are judged to have significant energy.
[0040] (2) Uniqueness of source: In At this point, the coherence function spectrum of the two impact signals ( The spectral value of the signal is not less than a preset threshold (preferably, greater than or equal to 0.85). This condition is used to exclude all interference signals from different sources that are transmitted from a distance by vibrations or non-periodic impacts, thereby enabling physical tracing of the fault signal and fundamentally solving the problem of false alarms.
[0041] (3) Position stability: In At this point, the cross-spectral density of the two impact signals ( The spectral value of ) fluctuates within a range smaller than a preset angle (e.g., ±5°) over a period of time, while The harmonics also exhibit a stable phase difference relationship corresponding to the fundamental frequency. This condition is used to confirm that the physical source is fixed and periodic, rather than random or mobile.
[0042] Through the aforementioned "three-in-one" chain of evidence, this method can effectively distinguish, from physical principles, real periodic impacts originating from internal fixed defects in bearings from all pseudo-impact signals such as external random knocks, remote vibration transmission, and non-periodic noise, greatly improving the reliability of diagnosis and the sensitivity of early fault detection.
[0043] Once the fault is confirmed by the above steps, step four, fault location calculation, is initiated. In one possible implementation, a circumferential angle calculation model is established based on the geometric relationship between the fault point and the two vibration time-series signal acquisition points on the monitored bearing housing. Specifically, this includes: based on the fault point and the geometric relationship between the two vibration time-series signal acquisition points... At the circumferential position on the monitored bearing housing, the fault point is calculated to... The difference in arc length between points is expressed as: According to the propagation of the shock wave from the fault point to... Time difference of points and the propagation speed of the shock wave in the monitored bearing housing material (This can be approximated as a constant), calculate the arrival time of the shock wave. The difference in path length between points. Finally, solve for the condition where the difference in path length equals the difference in arc length. The value is used as the circumferential position of the fault point on the monitored bearing housing, that is, to determine the angle of the fault point relative to sensor A or B.
[0044] The path length difference is expressed as:
[0045] in, These are the actual fault characteristic frequencies. To extract from mutual position spectra Stable phase difference at the fundamental frequency.
[0046] The above method will be further explained below with reference to specific embodiments. Taking the diagnosis and location of early spalling failure of the outer ring of a main bearing of a certain test bench as an example, this method specifically includes the following: 1. Signal Acquisition Vibration sensor deployment: Two high-sensitivity accelerometers (A and B) are selected and securely mounted on the housing 1 of the main bearing housing using magnetic mounts. With the bearing housing horizontally to the right as the 0° reference, sensor A is installed at the 3 o'clock position (0°), and sensor B is installed at the 12 o'clock position (90°). Figure 2 As shown.
[0047] Speed sensor deployment: Install a photoelectric speed sensor with its probe aligned with a keyway on the spindle 0 to generate a key phase pulse signal.
[0048] Data Acquisition: A multi-channel dynamic signal acquisition instrument was used to synchronously acquire the vibration time-series signal of sensor A at a sampling frequency of 25.6kHz. Vibration timing signal of sensor B And the bond phase pulse signal.
[0049] 2. Signal Processing The three signals mentioned above are acquired simultaneously and formed into a data frame. The acquired signals are then... and The signals are subjected to time-domain impulse enhancement units. These units utilize the VMD algorithm to decompose the signals. Calculations show that... The 4th IMF component and The 4th IMF component The kurtosis values are the largest, indicating that they best reflect the impact components in the equipment. Therefore, these two IMF components are selected as the enhanced impact signals for subsequent cross-spectral analysis.
[0050] 3. Spatial phase feature extraction The enhanced two IMF signals are fed into the phase spectrum extraction unit. This unit performs cross-spectral analysis, in which the two impulse signals... and The cross-power spectrum calculated after Fourier transform Represented as:
[0051] in, and They are and Fourier transform, yes conjugate, It expresses expectation.
[0052] based on Two key feature spectra were calculated and generated, where the coherence function spectrum is represented as follows:
[0053] in, These are two impact signals. and The self-power spectrum is calculated after Fourier transform. The range of values is .
[0054] when When the frequency is such that the two signals are highly correlated at that frequency point, it strongly suggests that they originate from the same physical source. This provides strong evidence for tracing the physical source of fault signals and can effectively rule out external random interference and remotely transmitted vibrations. When the frequency is 0, it indicates that the two signals are uncorrelated at that frequency point, which is usually random noise or signals from different sources.
[0055] The mutual position spectrum is represented as:
[0056] in, This represents the argument of a complex number. For periodic impacts originating from the same fault point, its value at the true fault characteristic frequency... Phase difference at It should be stable and deterministic. This phase difference directly reflects the time difference of the shock wave propagating from the fault point to the two different sensor acquisition points, and is the direct basis for accurate physical location of the fault.
[0057] 4. Fault diagnosis and decision making At a rotation speed of Under the condition of rpm, the theoretical value of the outer ring fault characteristic frequency (BPFO) is calculated to be... Hz.
[0058] Energy significance: Examining the autopower spectrum of the enhanced signal revealed that... Hz and If the energy peak at a frequency Hz is higher than the preset decibel value of the background noise, it is determined that the energy at that frequency is significant.
[0059] Source uniqueness: Check the coherence function spectrum , found in Hz and The coherence function value at Hz reaches a preset threshold (≥0.85), indicating a unique source. This strongly proves that these two frequency components are highly correlated on both sensors, meaning they originate from the same fixed physical source.
[0060] Positional stability: Check mutual position spectrum agreed phase difference (That is, the phase difference between sensor B and sensor A). It was found in... The phase difference at Hz is stable at around -45° (with fluctuations within ±5°). The phase difference at Hz is stable at around -90° (with fluctuations within ±10°), which conforms to the harmonic phase doubling relationship. This proves that the position of the physical source is stable and unchanging.
[0061] Conclusion: Since the three physical conditions of energy, source, and location are simultaneously satisfied at the fundamental frequency and harmonics of the fault characteristic frequency, the diagnostic module ultimately determines that the main bearing has an outer ring fault.
[0062] 5. Precise circumferential fault location After the fault was confirmed, the circumferential fault location calculation began. First, the stable phase difference was extracted: the fault fundamental frequency was extracted from the cross-phase spectrum. Stable phase difference at Hz This means that the signal phase of sensor B lags behind the signal phase of sensor A by -45°. Therefore, the shock wave generated by fault point C arrives at sensor B (at the 90° position) later than at sensor A (at the 0° position), indicating that the fault point is closer to sensor A and farther from sensor B. Based on Calculate the propagation of the shock wave from the fault point to Time difference of points :
[0063] Calculate the difference in path length between the two sensor acquisition points for the shock wave. :
[0064] in, The propagation speed of the shock wave in the bearing housing material needs to be determined through experimental calibration or empirical values. Because The value is positive, confirming once again that fault point C is closer to sensor A.
[0065] Assuming the average radius of the bearing outer ring is It is known that sensor A is located at... Location, sensor B is located Location. Fault point C is located at an unknown angle on the outer raceway. (by (Using this as a reference, counterclockwise is positive). Assuming the shock wave propagates along the outer raceway surface with the shortest arc length, then: Arc length from fault point C to sensor A (Angles need to be converted to radians).
[0066] Arc length from fault point C to sensor B (Angles need to be converted to radians).
[0067] The built-in solution model solves equations To determine the circumferential angle of fault point C. .
[0068] In this embodiment, it is assumed that the effective propagation speed calibrated on-site is... , Hz, then .set up The built-in solution module will traverse or iterate to solve the problem. .for (Right now ):
[0069]
[0070]
[0071] This solution result The results obtained from the phase difference and the calibrated wave velocity The results are highly consistent. Ultimately, it can be concluded that the fault point C is located at a circumferential position of 315° (or -45°) on the outer ring raceway of the bearing.
[0072] 6. Final Result Output The output module (such as the monitoring software interface) displays the diagnostic report: "Alarm: Early spalling fault exists in the outer ring of the main bearing. Physical location of the fault: 315° direction of the outer ring raceway (near the lower right). It is recommended to perform an endoscopic inspection or replace the fault during the next maintenance window." Through the above embodiments, the method provided by the present invention successfully detected early and weak faults under strong background noise, and through the analysis of spatial phase characteristics, it achieved physical tracing and precise location of the fault source, providing unprecedented precise guidance for equipment maintenance.
[0073] The above descriptions are merely preferred embodiments of this application, and the present invention is not limited to the above embodiments. It is understood that other improvements and variations directly derived or conceived by those skilled in the art without departing from the spirit and concept of the present invention should be considered to be included within the protection scope of the present invention.
Claims
1. A method for fault diagnosis and physical location of rolling bearings based on time-frequency phase characteristics, characterized in that, The method includes: The vibration time-series signals at at least two different circumferential positions on the monitored bearing housing are acquired and subjected to time-domain impact feature enhancement processing to obtain two enhanced impact signals. Cross-spectral analysis was performed on the two impact signals to calculate and extract the spatial phase characteristic spectrum, which includes: The coherence function spectrum is used to quantify the degree of linear correlation between two impulse signals at each frequency point in the frequency domain. Cross-phase spectrum is used to provide the phase difference between one impulse signal and another impulse signal at various frequency points in the frequency domain. Based on the peak energy of the impact signal, the coherence function spectrum, and the cross-position spectrum, fault diagnosis conditions are set, the true fault characteristic frequency that meets the conditions is found, and compared with the theoretical value of the fault characteristic frequency of various rolling bearings, the rolling bearing fault type corresponding to the closest theoretical value is output as the fault type corresponding to the true fault characteristic frequency. Based on the geometric relationship between the fault point and the two vibration time sequence signal acquisition points on the monitored bearing housing, a circumferential angle calculation model is established. The true fault characteristic frequency, the stable phase difference extracted at the true fault characteristic frequency, and the circumferential position of the vibration time sequence signal acquisition point are substituted into the model to calculate the circumferential position of the fault point on the monitored bearing housing.
2. The method for diagnosing and physically locating rolling bearing faults according to claim 1, characterized in that, The cross-spectral analysis of the two impact signals, and the calculation and extraction of the spatial phase characteristic spectrum, includes: After performing Fourier transform on the two impact signals, their respective self-power spectra and cross-power spectra are calculated. The coherence function spectrum is calculated based on the self-power spectrum and the cross-power spectrum. A spectrum value close to 1 indicates that the two impact signals are highly correlated at the corresponding frequency points and originate from the same fault point. The mutual potential spectrum is calculated based on the mutual power spectrum. When the spectrum value is stable, it indicates that the phase difference of the periodic shock wave originating from the same fault point at the corresponding frequency point is stable. The stable phase difference directly reflects the time difference of the shock wave propagating from the fault point to two different acquisition points.
3. The method for diagnosing and physically locating rolling bearing faults according to claim 2, characterized in that, The spectrum of the coherence function is expressed as: in, The cross-power spectrum is calculated after Fourier transforming the two impact signals. These are the self-power spectra calculated from the Fourier transforms of the two impact signals. The range of values is .
4. The method for diagnosing and physically locating rolling bearing faults according to claim 2, characterized in that, The mutual position spectrum is represented as follows: in, The cross-power spectrum is calculated after Fourier transforming the two impact signals. It represents the argument of a complex number.
5. The method for diagnosing and physically locating rolling bearing faults according to claim 1, characterized in that, The step of setting fault diagnosis conditions based on the peak energy of the impact signal, the coherence function spectrum, and the cross-position spectrum, and finding the true fault characteristic frequencies that meet the conditions, includes: For any suspected fault characteristic frequency A frequency is identified as a true fault characteristic frequency when all of the following fault diagnosis conditions are met simultaneously. : In the At this point, the peak energy values of both impact signals were higher than the preset decibel value of the background noise; In the At this point, the spectral values of the coherence function spectra of the two impact signals are not less than a preset threshold; In the At this point, the spectral values of the mutual position spectra of the two impact signals fluctuate within a range less than a preset angle over a period of time, while the... The harmonics also exhibit a stable phase difference relationship corresponding to the fundamental frequency.
6. The method for diagnosing and physically locating rolling bearing faults according to claim 1, characterized in that, The step of establishing a circumferential angle calculation model based on the geometric relationship between the fault point and two vibration time-series signal acquisition points on the monitored bearing housing includes: Based on the fault point and two vibration timing signal acquisition points The fault point is calculated to the circumferential position on the monitored bearing housing. The difference in arc length between points is expressed as: ; According to the propagation of the shock wave from the fault point to Time difference of points and the transmission speed of the shock wave in the monitored bearing housing material Calculate the arrival time of the shock wave The difference in path length between the points; Solve for the case where the path length difference is equal to the arc length difference. The value is used as the circumferential position of the fault point on the monitored bearing housing.
7. The method for diagnosing and physically locating rolling bearing faults according to claim 6, characterized in that, The path length difference is expressed as: in, The actual fault characteristic frequency, To extract from the mutual position spectrum Stable phase difference at the fundamental frequency.
8. The method for diagnosing and physically locating rolling bearing faults according to claim 1, characterized in that, The time-domain impact feature enhancement processing includes, for each vibration time-series signal: The vibration time-series signal is decomposed into a series of intrinsic mode functions (IMF) components using adaptive signal decomposition technology. Based on the statistical index characterizing the impact of the signal, the IMF component with the largest statistical index is selected from each component. This component best characterizes the impact component and is used as the enhanced impact signal for subsequent cross-spectral analysis. The statistical indicators mentioned above are kurtosis, energy entropy, or correlation coefficient.
9. The method for diagnosing and physically locating rolling bearing faults according to claim 1, characterized in that, The method further includes: Install a speed sensor on the bearing housing being monitored and align it with the mark on the rotor; The absolute phase reference of the rotor provided by the speed sensor is used to calibrate and verify the fault location results.
10. The method for diagnosing and physically locating rolling bearing faults according to claim 1, characterized in that, The circumferential positions of the two vibration timing signal acquisition points are distributed along the orthogonal direction of the monitored bearing housing.