Rail transit electromechanical equipment defect fault vibration detection analysis method and system

Abstract

The application discloses a kind of rail transit electromechanical equipment defect fault vibration detection analysis method and system, the method includes the operation data of the health state of rail transit electromechanical equipment high-low frequency signal respectively using improved ensemble empirical mode decomposition method for signal decomposition, screening inherent mode component number;Respectively for high, low frequency signal calculation relevant control limit parameter and control limit, utilize the weight of the instantaneous energy density of signal and signal-to-noise ratio to dynamically adjust high frequency and low frequency signal, calculate the comprehensive monitoring index of high frequency signal and low frequency signal, as the determination threshold of rail transit electromechanical equipment defect fault vibration detection analysis;The comprehensive monitoring index of the data to be measured is calculated, and compared with determination threshold to determine defect fault vibration.This application can effectively improve the efficiency and accuracy of signal decomposition, and comprehensively consider high-low frequency signal to improve fault detection accuracy.

Description

Technical Field

[0001] This invention relates to fault detection technology, and more particularly to a method and system for detecting and analyzing vibration faults in rail transit electromechanical equipment. Background Technology

[0002] With the acceleration of urbanization, rail transit systems are under unprecedented pressure, especially during peak hours, when the safety of rail transit electromechanical equipment is directly related to passenger safety. Traditional maintenance methods are no longer adequate to meet the demands of high frequency and high load.

[0003] When special rail transit equipment has defects or malfunctions, it will generate significant vibrations during operation. Therefore, one of the mainstream methods for defect detection is to detect the frequency and amplitude of the vibration to obtain the time-frequency characteristics of the vibration signal, and then combine this characteristic analysis to determine whether the equipment has a fault. Existing methods for analyzing vibration signals include Short-Time Fourier Transform (STFT), S-Transform, Wavelet Transform, and Ensemble Empirical Mode Decomposition (EEMD). However, these methods still have some shortcomings. For example, the STFT uses a fixed-size window for Fourier transform, which means that the time and frequency resolutions are fixed and it is not suitable for the time-varying characteristics of non-stationary signals. The S-Transform usually involves complex mathematical operations, which may result in higher computational complexity compared to the STFT, especially when dealing with large-scale datasets. The performance of the Wavelet Transform is highly dependent on the properties of the selected mother wavelet; different mother wavelets are suitable for different types of signal analysis, resulting in poor adaptability. Ensemble Empirical Mode Decomposition (EEMD) is an improved method of Empirical Mode Decomposition (EMD). It reduces mode aliasing and improves the stability of decomposition by adding white noise to the signal. However, the resulting IMF components require manual selection, lacking effective information filtering for different scenarios. This increases the subjectivity and workload of the analysis, reducing the efficiency and accuracy of signal decomposition. Furthermore, when equipment components have defects, the contact between rotating parts and the defect will generate impact. The high-frequency band of the impact vibration signal contains most of the characteristic information, with some contained in the low-frequency band. However, the low-frequency band often has high environmental noise energy, which can easily mask energy changes at characteristic frequencies. Most current research methods select to analyze the high-frequency features of the impact signal, discarding some features contained in the low-frequency band. This results in insufficient analysis of the characteristic information of the vibration signal, easily leading to missed fault detection and reducing fault detection accuracy. Summary of the Invention

[0004] Purpose of the invention: The present invention aims to provide a vibration detection and analysis method for defects and faults in rail transit electromechanical equipment that can effectively improve the efficiency and accuracy of signal decomposition and comprehensively consider high and low frequency signals to improve the accuracy of fault detection. Another purpose of the present invention is to provide a vibration detection and analysis system for defects and faults in rail transit electromechanical equipment.

[0005] Technical solution: The present invention provides a vibration detection and analysis method for defects and faults in rail transit electromechanical equipment, comprising the following steps:

[0006] (1) Use the operational data of rail transit electromechanical equipment under healthy conditions as training data, separate high and low frequency signals in the training data, and perform preprocessing;

[0007] (2) The preprocessed high and low frequency signals are decomposed by the improved ensemble empirical mode decomposition method. The number of intrinsic mode components is selected based on the correlation coefficient, energy density and average period between each intrinsic mode component and the original signal.

[0008] (3) Calculate the relevant control limit parameters and control limits for high-frequency and low-frequency signals respectively, and construct a comprehensive control limit index; use the instantaneous energy density and signal-to-noise ratio of the signal to dynamically adjust the weight of high-frequency and low-frequency signals, calculate the comprehensive monitoring index of high-frequency and low-frequency signals, and use it as the judgment threshold for vibration detection and analysis of defects and faults in rail transit electromechanical equipment.

[0009] (4) Collect the current vibration data of rail transit electromechanical equipment as the data to be measured. After processing and calculation in the above steps (1)-(3), obtain the comprehensive monitoring index of the data to be measured, and compare it with the judgment threshold to make a defect and fault vibration judgment.

[0010] Preferably, the selection principle for the intrinsic modal components is: when If the value is greater than a given threshold, the intrinsic mode component is deemed valid and retained.

[0011]

[0012] in, This indicates the effectiveness of the IMF of the i-th intrinsic mode component compared to the original signal. Indicates the weighting coefficient. This represents the correlation coefficient between the i-th IMF and the original signal. , Let represent the maximum and minimum values ​​of the correlation coefficients between all IMFs and the original signal, respectively; E represents the energy density; and T represents the average period. This represents the energy coefficient of the i-th IMF. , These represent the maximum and minimum values ​​of all IMF energy coefficients, respectively.

[0013] Preferably, the correlation coefficient between the i-th IMF and the original signal for:

[0014]

[0015] in, It is the original signal and covariance It is the original signal Standard deviation of the original signal It is a function of time t. yes The standard deviation; n is the number of IMFs; The IMFs representing the vibration signals obtained through screening are as follows: .

[0016] Preferably, the energy coefficient It is the product of energy density E and average period T, where energy density for:

[0017]

[0018] In the formula, The sequence length of the Intrinsic Modal Components (IMFs) is given. yes The corresponding amplitude value; The IMFs representing the vibration signals obtained through screening are as follows: ;

[0019] Average period for:

[0020]

[0021] In the formula for The number of extreme points within a component.

[0022] Preferably, in step (3), the construction of the comprehensive control limit index includes:

[0023] PCA was used to extract principal components from the IMF components obtained in step (2), and principal component score matrices and principal component loading matrices were constructed. Statistical values ​​were then calculated using the principal component loading matrices and principal component score matrices. Sum of squared prediction errors ;

[0024] Then, through the chi-square distribution... Calculate and construct comprehensive control limit indicators.

[0025] Preferably, in step (3), the step of dynamically adjusting the weights of high-frequency and low-frequency signals using the instantaneous energy density and signal-to-noise ratio of the signal, and calculating the comprehensive monitoring index of the high-frequency and low-frequency signals, includes:

[0026] Calculate the weight of the high-frequency signal based on the instantaneous energy density and signal-to-noise ratio of the high- and low-frequency signals. :

[0027]

[0028] The weight of low-frequency signals for:

[0029] in, Indicates the signal-to-noise ratio of high-frequency signals. This represents the instantaneous energy density of a high-frequency signal. Indicates the signal-to-noise ratio of low-frequency signals. This represents the instantaneous energy density of a low-frequency signal.

[0030] The comprehensive monitoring index for high-frequency and low-frequency signals is as follows:

[0031]

[0032] In the formula, fai represents the comprehensive monitoring index. Indicates the comprehensive control limit index for high-frequency signals. This indicates the comprehensive control limit index for low-frequency signals.

[0033] Preferably, step (4) includes: calculating the comprehensive monitoring index of the data to be tested, determining whether the comprehensive monitoring index of the data to be tested is greater than the comprehensive monitoring index of the training data, if so, saving the index of the data sample to be tested, otherwise continuing to calculate the comprehensive monitoring index of the next data sample to be tested and comparing it.

[0034] If n out of m consecutive test data samples are stored in the sample index, then these m consecutive test data samples are marked as abnormal and an alarm is issued. At the same time, the sampling time of the first sample among the m consecutive test data samples is recorded as the sampling time of the abnormal sample; n is not less than m / 2, and m and n are integers.

[0035] The present invention discloses a vibration detection and analysis system for defects and faults in rail transit electromechanical equipment, comprising:

[0036] The training data preprocessing module is used to separate high- and low-frequency signals from the training data and perform preprocessing.

[0037] The signal decomposition module is used to decompose the preprocessed high- and low-frequency signals using an improved ensemble empirical mode decomposition method. The number of intrinsic mode components is selected based on the correlation coefficient and energy coefficient between each intrinsic mode component (IMF) and the original signal. The energy coefficient is the product of energy density and average period.

[0038] The judgment threshold calculation module for defect and fault vibration detection analysis is used to calculate relevant control limit parameters and control limits for high and low frequency signals respectively, and construct a comprehensive control limit index. Then, the instantaneous energy density and signal-to-noise ratio of the signal are used to dynamically adjust the weight of high frequency and low frequency signals, calculate the comprehensive monitoring index of high frequency and low frequency signals, and use it as the judgment threshold for defect and fault vibration detection analysis of rail transit electromechanical equipment.

[0039] The defect and fault vibration detection and analysis module is used to collect vibration data of rail transit electromechanical equipment as test data, calculate the comprehensive monitoring index of the test data, and compare it with the judgment threshold to make defect and fault vibration judgment.

[0040] The present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements the aforementioned method for detecting and analyzing vibrations and defects in rail transit electromechanical equipment.

[0041] The present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for detecting and analyzing vibrations and defects in rail transit electromechanical equipment.

[0042] Beneficial Effects: Compared with the prior art, the present invention has the following significant advantages: 1. When using Ensemble Empirical Mode Decomposition (EEMD) for signal decomposition, an IMF signal screening principle is introduced. The screening formula is constructed based on the correlation coefficient ρ between each IMF and the original signal and the product of energy density and average period ET, which can effectively improve the efficiency and accuracy of signal decomposition; 2. The high and low frequency signals of the original signal are separated, and then, based on the characteristics of the high and low frequency band signals, their comprehensive scores are calculated based on the instantaneous energy density and signal-to-noise ratio of each frequency band to dynamically adjust the weight of the feature information contained in the high and low frequency bands, making full use of the information contained in the high and low frequency signals, so as to make the fault detection results more accurate. Attached Figure Description

[0043] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0044] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, the vibration detection and analysis method for defects and faults in rail transit electromechanical equipment according to the present invention includes the following steps:

[0046] (1) Use the operating data of rail transit electromechanical equipment during a period of time marked as healthy as training data. For example, the vibration data of the left wire rope of a certain elevator during operation, the vibration data of the left side of the bearing in the lower pit of a certain escalator during operation, or the vibration data of key mechanical components (such as drive device, guide rail, door lock mechanism) during platform screen door operation. Use filters to separate high and low frequency signals from the training data, and use noise reduction methods such as wavelet thresholding to process the low frequency signals.

[0047] Although most of the characteristic information of vibration signals resides in the high-frequency band, a small amount of useful information remains in the low-frequency band. To ensure the accuracy of the input data, it is still necessary to process the low-frequency signals to extract the useful information. Low-frequency signals typically suffer from significant noise interference, therefore, noise reduction processing is required for them.

[0048] (2) The ensemble empirical mode decomposition (EEMD) method is used to decompose the high-frequency signal and the low-frequency signal after noise reduction, and effective IMFs are selected.

[0049] The EEMD method is essentially an improvement on the EMD algorithm. Based on the characteristic that white noise has zero mean, white noise is added to the signal before EMD decomposition. The decomposition results are then averaged; the more times the averaging is performed, the smaller the impact of noise on the decomposition result. The signal after EEMD decomposition... for

[0050]

[0051] in, These are intrinsic modal components. For the remaining items, The number of IMFs when the EEMD decomposition termination condition is met.

[0052] Typically, the number of Intrinsic Modal Components (IMFs) selected is determined by human experience, often choosing the first n IMFs to form the reconstructed signal, without filtering out effective information for different scenarios.

[0053] This invention introduces a signal filtering principle, namely, calculating the correlation coefficient between each IMF and the original signal. and energy coefficient Using this method to select the number of intrinsic mode components can effectively improve the efficiency and accuracy of signal decomposition.

[0054] The IMFs of the vibration signal obtained by EEMD decomposition are represented as follows: Its correlation coefficient for:

[0055]

[0056] in, It is a signal and The correlation coefficient, It is the original signal and covariance It is the original signal Standard deviation of the original signal It is a function of time t. yes The standard deviation of the correlation coefficient. The closer the value is to 1, the closer the correlation between the separated intrinsic mode component and the signal, and the more feature information it contains.

[0057] Energy coefficient It is the product of energy density E and average period T, where energy density for:

[0058]

[0059] In the formula, The sequence length of the Intrinsic Modal Components (IMFs) is given. yes The corresponding amplitude value.

[0060] Average period for:

[0061]

[0062] In the formula for The number of extreme points within a component.

[0063] according to and Constructing effective IMF screening principles: When the i-th IMF component is as effective as the original signal... If the value is greater than a given threshold, the component is considered valid and retained.

[0064]

[0065] in The effectiveness of the i-th IMF component compared to the original signal. This represents the weighting coefficient; when the correlation coefficient and energy coefficient are of equal importance... The possible value is 0.5.

[0066] (3) Calculate the relevant control limit parameters and control limits for high-frequency and low-frequency signals respectively, and construct a comprehensive control limit index.

[0067] After obtaining all valid IMF components in step (2), they are centered and standardized. Then, PCA is performed on the IMF components to extract principal components (PCs), constructing the principal component score matrix and principal component loading matrix. The principal component loading matrix and principal component score matrix are then used to calculate... Statistic and Squared Prediction Error .

[0068]

[0069]

[0070] in These are the original observations. It is the principal component loading matrix. It is a diagonal matrix of principal component variances (eigenvalues). The statistic represents the degree of anomaly of a sample in the principal component space. A higher value indicates a higher degree of abnormality in the sample, and samples exceeding the control limit may have problems.

[0071] These are the original observations. These are the observations reconstructed using PCA. The value reflects the degree to which the sample deviates from the principal component model. The higher the value, the greater the deviation of the sample from the model. If it exceeds the control limit, it may be an outlier.

[0072] Determined using the F-distribution Control Limits :

[0073]

[0074] in It is the sample size. It is the number of principal components. This indicates that, at a significance level of 0.95, the degrees of freedom are... and The critical value of the F-distribution.

[0075] Calculate SPE control limits :

[0076]

[0077]

[0078] in It is the confidence limit of the standard normal distribution, satisfying the formula: , The value is 0.01. It is a constant related to the data distribution. , , These represent the sum of eigenvalues, the sum of squares, and the sum of cubes within the residual space, respectively.

[0079] Get SPE and T 2 After controlling for the limits, the chi-square distribution is used. Calculate and construct the comprehensive control limit index fai':

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] in It is based on the covariance matrix The correction factor calculated from the trace is used to adjust the chi-square distribution. It is a correction factor based on the sum of squared traces of the covariance matrix. It is a phi matrix, temp is a temporary value, and is the sum of Z and φ. The product of It is an identity matrix.

[0086] The comprehensive control limit index (fai) integrates... and The statistic is used to comprehensively assess the degree of anomaly in a sample. A value exceeding the control limit indicates that the sample has a significant anomaly.

[0087] (4) Use the instantaneous energy density and signal-to-noise ratio of the signal to dynamically adjust the weight of high-frequency and low-frequency signals, and calculate the comprehensive monitoring index of high-frequency and low-frequency signals.

[0088] High-frequency signals typically contain faster fluctuations and impulses, therefore their instantaneous energy density usually exhibits a higher instantaneous peak value. Low-frequency signals generally reflect long-term, slow changes, with relatively smaller changes in instantaneous energy density. Therefore, instantaneous energy density is used to measure the local energy of a signal at a given moment, while the signal-to-noise ratio (SNR) measures the overall strength of the signal relative to noise. Instantaneous energy density and SNR are calculated separately for high-frequency and low-frequency vibration signals. Based on the instantaneous energy density and SNR of each frequency band, a comprehensive score is calculated to dynamically adjust the weights.

[0089] The instantaneous energy density of a signal refers to the average power of the signal over an extremely short time interval. For the input signal... Its instantaneous energy density for:

[0090]

[0091] Signal-to-noise ratio (SNR) is the ratio of signal power to noise power, used to measure signal quality, and is calculated as follows:

[0092]

[0093] in It is a signal power, It is the power of the noise;

[0094] In decibels (dB), the signal-to-noise ratio (SNR) can be calculated using the following formula:

[0095]

[0096] After calculating the instantaneous energy density and signal-to-noise ratio for the high and low frequency bands respectively, the weight of the high-frequency band signal is calculated. :

[0097]

[0098] The weight of low-frequency signals for: ,in, Indicates the signal-to-noise ratio of high-frequency signals. This represents the instantaneous energy density of a high-frequency signal. Indicates the signal-to-noise ratio of low-frequency signals. This represents the instantaneous energy density of a low-frequency signal.

[0099] The comprehensive monitoring index for high-frequency and low-frequency signals is calculated as follows:

[0100]

[0101] In the formula, fai represents the comprehensive monitoring index. Indicates the comprehensive control limit index for high-frequency signals. This indicates the comprehensive control limit index for low-frequency signals.

[0102] (5) Collect the current vibration data of the rail transit electromechanical equipment as the test data. After processing and calculation in the above steps (1)-(4), obtain the comprehensive monitoring index of the test data. Determine whether the comprehensive monitoring index of the test data is greater than the comprehensive monitoring index of the training data. If so, save the index of the test data sample. Otherwise, continue to calculate the comprehensive monitoring index of the next test data sample and compare them.

[0103] If n out of m consecutive test data samples (n is not less than m / 2, and m and n are integers) are stored in the sample index, then these m consecutive test data samples are marked as abnormal, an alarm is issued, and the sampling time of the first sample among the m consecutive test data samples is recorded as the sampling time of the abnormal sample.

[0104] Based on the same inventive concept, the present invention provides a vibration detection and analysis system for defects and faults in rail transit electromechanical equipment, comprising:

[0105] The training data preprocessing module is used to separate high- and low-frequency signals from the training data and perform preprocessing.

[0106] The signal decomposition module is used to decompose the preprocessed high- and low-frequency signals using an improved ensemble empirical mode decomposition method. The number of intrinsic mode components is selected based on the correlation coefficient and energy coefficient between each intrinsic mode component (IMF) and the original signal. The energy coefficient is the product of energy density and average period.

[0107] The judgment threshold calculation module for defect and fault vibration detection analysis is used to calculate relevant control limit parameters and control limits for high and low frequency signals respectively, and construct a comprehensive control limit index. Then, the instantaneous energy density and signal-to-noise ratio of the signal are used to dynamically adjust the weight of high frequency and low frequency signals, calculate the comprehensive monitoring index of high frequency and low frequency signals, and use it as the judgment threshold for defect and fault vibration detection analysis of rail transit electromechanical equipment.

[0108] The defect and fault vibration detection and analysis module is used to collect vibration data of rail transit electromechanical equipment as test data, calculate the comprehensive monitoring index of the test data, and compare it with the judgment threshold to make defect and fault vibration judgment.

[0109] Based on the same inventive concept, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements the above-mentioned method for detecting and analyzing vibrations and defects in rail transit electromechanical equipment.

[0110] Based on the same inventive concept, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for detecting and analyzing vibrations and defects in rail transit electromechanical equipment.

[0111] The invention will now be further illustrated with three specific embodiments.

[0112] Example 1: Vibration Detection and Analysis of Straight Ladder Steel Wire Rope

[0113] This embodiment uses a subway station escalator as an example to detect the vibration data of its wire rope. The vibration detection and analysis method for defects and faults in the wire rope of an escalator, as described in this invention, includes the following steps:

[0114] (1) Vibration signals of the left steel wire rope of the straight ladder under healthy operating conditions were collected as training data. The sampling frequency was 256 kHz, and the duration of a single acquisition was 180 seconds. High-pass and low-pass filters were used to separate the high and low frequency signals of the training data, with a cutoff frequency of 256 kHz. Low-frequency signals (≤25.6 kHz) were denoised using wavelet thresholding, with the wavelet basis being Daubechies 4 and the decomposition level being 6 levels; high-frequency signals were preserved.

[0115] (2) Perform EEMD decomposition on the high-frequency signal and the denoised low-frequency signal respectively. Calculate the correlation coefficient and energy coefficient of each IMF with the original signal, and set the weight coefficient to 0.5 to filter effective IMF components. Set a threshold of 0.3 to retain effective IMF components.

[0116] (3) For high-frequency and low-frequency signals, calculate the principal components of PCA respectively, obtain the statistic T² and the squared prediction error SPE, calculate the control limits, and then construct the comprehensive control limit index fai' through the chi-square distribution. Comprehensive control limit index for high-frequency signals The comprehensive control limit index for low-frequency signals is 0.715. The result was 0.592.

[0117] (4) Calculate the instantaneous energy density and signal-to-noise ratio of high-frequency and low-frequency signals, dynamically adjust the weights, and calculate the comprehensive monitoring index of high-frequency and low-frequency signals.

[0118] The instantaneous energy density of the high-frequency signal was calculated. The signal-to-noise ratio is 0.85. 25 dB; instantaneous energy density of low-frequency signals The signal-to-noise ratio is 0.15. The value is 20dB. This allows us to obtain the weighting of the high-frequency signal. The weight of the low-frequency signal is 0.876. The result was 0.124. Ultimately, the comprehensive monitoring index (fai) for the training data was calculated to be 0.6997, which was used as the judgment threshold.

[0119] (5) Collect the vibration signal of the left steel wire rope of the same ladder after it has been running for a period of time as the data to be measured. The duration of a single acquisition is 90 seconds. After processing and calculation in the above steps (1)-(4), the comprehensive monitoring index fai of the data to be measured is 1.2848.

[0120] Since the comprehensive monitoring index of the data to be tested exceeds the judgment threshold, the record is saved. Further, the comprehensive monitoring index values ​​of 6 out of 10 consecutive samples to be tested exceed the judgment threshold. Therefore, these 10 samples are marked as abnormal, an alarm is issued, and the sampling time of the first sample is recorded.

[0121] On-site inspection confirmed slight wear on the left steel wire rope of the straight ladder, consistent with the vibration test results, thus verifying the effectiveness of this method.

[0122] Example 2: Vibration Detection and Analysis of the Lower Sink Bearing of an Escalator

[0123] This embodiment uses the bearing of the escalator's lower stairwell in a subway station as an example to detect vibration data at the bearing. The present invention provides a method for vibration detection and analysis of defects in escalator lower stairwell bearings, comprising the following steps:

[0124] (1) Vibration signals of the left bearing of the escalator pit were collected as training data when the escalator was running in a healthy state. The sampling frequency was 51.2 kHz and the duration of a single acquisition was 2.56 seconds. High-pass and low-pass filters were used to separate the high and low frequency signals of the training data, with a cutoff frequency of 10 kHz. The low-frequency signal was denoised using the wavelet thresholding method, with the wavelet basis being Daubechies 4 and the decomposition level being 5 levels; the high-frequency signal was preserved.

[0125] (2) Perform EEMD decomposition on the high-frequency signal and the noise-reduced low-frequency signal respectively. Filter the effective IMF components by calculating the correlation coefficient and energy coefficient of each IMF with the original signal. Set a threshold of 0.25 to retain the effective IMF components.

[0126] (3) For high-frequency and low-frequency signals, calculate the principal components of PCA respectively, obtain the statistic T² and the squared prediction error SPE, calculate the control limits, and then construct the comprehensive control limit index fai' through the chi-square distribution. Comprehensive control limit index for high-frequency signals The comprehensive control limit index for low-frequency signals is 0.648. The result was 0.465.

[0127] (4) Calculate the instantaneous energy density of the high-frequency signal. The signal-to-noise ratio is 0.78. 22dB; instantaneous energy density of low-frequency signals The signal-to-noise ratio is 0.22. The value is 18 dB. This gives the weight of the high-frequency signal. The value is 0.81, indicating a low-frequency weight. The result was 0.19. Finally, the comprehensive monitoring index fai of the training data was calculated to be 0.6132, which was used as the judgment threshold.

[0128] (5) When the same escalator makes an abnormal noise during operation, the vibration signal of the left bearing of the lower step pit is collected as the data to be measured. The duration of a single collection is 2.56 seconds. After processing and calculation in the above steps (1)-(4), the comprehensive monitoring index fai of the data to be measured is 0.7828.

[0129] Since the comprehensive monitoring index of the data to be tested exceeds the judgment threshold, the record is saved. Further, the comprehensive monitoring index values ​​of 7 out of 10 consecutive samples to be tested exceed the judgment threshold. Therefore, these 10 samples are marked as abnormal, an alarm is issued, and the sampling time of the first sample is recorded.

[0130] On-site inspection confirmed that the left bearing of the escalator's lower pit had slight wear, consistent with the vibration test results. This method is applicable to the detection of defects in the bearing of the escalator's lower pit.

[0131] Example 3: Guide Rail Vibration Detection and Analysis

[0132] This embodiment uses the guide rail of a rail transit platform door as an example to detect vibration data at its bearing. The vibration detection and analysis method for defects and faults in rail transit platform door guide rails according to this invention includes the following steps:

[0133] (1) Vibration signals of the guide rails of the platform screen door under healthy operating conditions were collected as training data. The sampling frequency was 2kHz, and the duration of a single acquisition was 3 seconds. High-pass and low-pass filters were used to separate the high and low frequency signals of the training data, with a cutoff frequency of 500Hz. The low-frequency signal was denoised using the wavelet thresholding method, with the wavelet basis being Daubechies 4 and the decomposition level being 4 levels; the high-frequency signal was retained.

[0134] (2) Perform EEMD decomposition on the high-frequency signal and the noise-reduced low-frequency signal respectively. Filter the effective IMF components by calculating the correlation coefficient and energy coefficient of each IMF with the original signal. Set a threshold of 0.2 to retain the effective IMF components.

[0135] (3) For high-frequency and low-frequency signals, calculate the principal components of PCA respectively, obtain the statistic T² and the squared prediction error SPE, calculate the control limits, and then construct the comprehensive control limit index fai' through the chi-square distribution. Comprehensive control limit index for high-frequency signals The comprehensive control limit index for low-frequency signals is 0.6. The result was 0.39.

[0136] (4) Calculate the instantaneous energy density of the high-frequency signal. The signal-to-noise ratio is 0.7. 20dB; instantaneous energy density of low-frequency signals The signal-to-noise ratio is 0.3. The value is 15 dB. This gives the weight of the high-frequency signal. The value is 0.757, indicating a low-frequency weight. The result was 0.243. Ultimately, the comprehensive monitoring index (fai) for the training data was calculated to be 0.54897, which was used as the judgment threshold.

[0137] (5) When the same guide rail experiences slight jamming during operation, the vibration signal of the left bearing of the lower step pit is collected as the data to be measured. The duration of each acquisition is 3 seconds. After processing and calculation in the above steps (1)-(4), the comprehensive monitoring index fai of the data to be measured is 0.7043.

[0138] Since the comprehensive monitoring index of the data to be tested exceeds the judgment threshold, the record is saved. Further, the comprehensive monitoring index values ​​of 5 out of 10 consecutive samples to be tested are checked, so these 10 samples are marked as abnormal, an alarm is issued, and the sampling time of the first sample is recorded.

[0139] On-site inspection confirmed that the guide rail had slight wear, which was consistent with the vibration test results. This method is applicable to the detection of guide rail defects.

[0140] To further verify the effectiveness of this invention, the following methods were used to analyze the vibration of escalator lower stairwell bearings based on multiple samples: the method of this invention, a method that only uses high-frequency features of vibration signals (ignoring low-frequency signal features while maintaining the same process as the method of this invention), and a method in the ensemble empirical mode decomposition method that uses manual experience to select IMF components (while maintaining the same process as the method of this invention). The results showed that the accuracy rate of defect detection using only high-frequency features of vibration signals was 82.5%, the accuracy rate of the traditional manual experience-based IMF classification method was 86.25%, and the accuracy rate of the method of this invention was 93.75%, all higher than the two methods mentioned above. Therefore, the method of this invention can improve the accuracy of vibration detection for defects in rail transit electromechanical equipment.

Claims

1. A method for vibration detection and analysis of defects and faults in rail transit electromechanical equipment, characterized in that, Includes the following steps: (1) Use the operational data of rail transit electromechanical equipment under healthy conditions as training data, separate high and low frequency signals in the training data, and perform preprocessing; (2) The preprocessed high and low frequency signals are decomposed by the improved ensemble empirical mode decomposition method. The number of intrinsic mode components is selected based on the correlation coefficient, energy density and average period between each intrinsic mode component and the original signal. In step (2), the selection principle for the intrinsic modal components is: when If the value is greater than a given threshold, the intrinsic mode component is deemed valid and retained. in, This indicates the effectiveness of the IMF of the i-th intrinsic mode component compared to the original signal. Indicates the weighting coefficient. This represents the correlation coefficient between the i-th IMF and the original signal. , Let represent the maximum and minimum values ​​of the correlation coefficients between all IMFs and the original signal, respectively; E represents the energy density; and T represents the average period. This represents the energy coefficient of the i-th IMF. , These represent the maximum and minimum values ​​of all IMF energy coefficients, respectively. (3) Calculate the relevant control limit parameters and control limits for high-frequency and low-frequency signals respectively, and construct a comprehensive control limit index; use the instantaneous energy density and signal-to-noise ratio of the signal to dynamically adjust the weight of high-frequency and low-frequency signals, calculate the comprehensive monitoring index of high-frequency and low-frequency signals, and use it as the judgment threshold for vibration detection and analysis of defects and faults in rail transit electromechanical equipment. In step (3), the construction of the comprehensive control limit index includes: PCA was used to extract principal components from the IMF components obtained in step (2), and principal component score matrices and principal component loading matrices were constructed. Statistical values ​​were then calculated using the principal component loading matrices and principal component score matrices. Sum of squared prediction errors Then through chi-square distribution Calculate and construct the comprehensive control limit index fai'; In step (3), the step of dynamically adjusting the weights of high-frequency and low-frequency signals using the instantaneous energy density and signal-to-noise ratio of the signals, and calculating the comprehensive monitoring index of high-frequency and low-frequency signals, includes: calculating the weights of high-frequency signals based on the instantaneous energy density and signal-to-noise ratio of the high and low-frequency signals. : The weight of low-frequency signals for: in, Indicates the signal-to-noise ratio of high-frequency signals. This represents the instantaneous energy density of a high-frequency signal. Indicates the signal-to-noise ratio of low-frequency signals. This represents the instantaneous energy density of the low-frequency signal; therefore, the comprehensive monitoring index for both high-frequency and low-frequency signals is: In the formula, fai represents the comprehensive monitoring index, f represents the comprehensive control limit index for high-frequency signals, and f represents the comprehensive control limit index for low-frequency signals. (4) Collect the current vibration data of rail transit electromechanical equipment as the data to be measured. After processing and calculation in the above steps (1)-(3), obtain the comprehensive monitoring index of the data to be measured, and compare it with the judgment threshold to make a defect and fault vibration judgment.

2. The vibration detection and analysis method for defects and faults in rail transit electromechanical equipment according to claim 1, characterized in that, The correlation coefficient between the i-th IMF and the original signal for: in, It is the original signal and covariance It is a signal Standard deviation of the original signal It is a function of time t; yes The standard deviation; n is the number of IMFs; The IMFs representing the vibration signals obtained through screening are as follows: .

3. The method for vibration detection and analysis of defects and faults in rail transit electromechanical equipment according to claim 1, characterized in that, Energy coefficient It is the product of energy density E and average period T, where energy density for: In the formula, The sequence length of the Intrinsic Modal Components (IMFs) is given. yes The corresponding amplitude value; The IMFs representing the vibration signals obtained through screening are as follows: ; Average period for: In the formula for The number of extreme points within a component.

4. The vibration detection and analysis method for defects and faults in rail transit electromechanical equipment according to claim 1, characterized in that, Step (4) includes: calculating the comprehensive monitoring index of the data to be tested, determining whether the comprehensive monitoring index of the data to be tested is greater than the comprehensive monitoring index of the training data, if so, saving the index of the sample, otherwise continuing to calculate the comprehensive monitoring index of the next sample to be tested and comparing them. If n out of m consecutive test data samples are stored in the sample index, then these m consecutive samples are marked as abnormal and an alarm is issued. At the same time, the sampling time of the first sample among the m consecutive samples is recorded as the sampling time of the abnormal sample; n is not less than m / 2, and m and n are integers.

5. A vibration detection and analysis system for rail transit electromechanical equipment defects and faults, used to implement the vibration detection and analysis method for rail transit electromechanical equipment defects and faults according to any one of claims 1-4, characterized in that, include: The training data preprocessing module is used to separate high- and low-frequency signals from the training data and perform preprocessing. The signal decomposition module is used to decompose the preprocessed high- and low-frequency signals using an improved ensemble empirical mode decomposition method. The number of intrinsic mode components is selected based on the correlation coefficient and energy coefficient between each intrinsic mode component (IMF) and the original signal. The energy coefficient is the product of energy density and average period. The judgment threshold calculation module for defect and fault vibration detection analysis is used to calculate relevant control limit parameters and control limits for high and low frequency signals respectively, and construct a comprehensive control limit index. Then, the instantaneous energy density and signal-to-noise ratio of the signal are used to dynamically adjust the weight of high frequency and low frequency signals, calculate the comprehensive monitoring index of high frequency and low frequency signals, and use it as the judgment threshold for defect and fault vibration detection analysis of rail transit electromechanical equipment. The defect and fault vibration detection and analysis module is used to collect vibration data of rail transit electromechanical equipment as test data, calculate the comprehensive monitoring index of the test data, and compare it with the judgment threshold to make defect and fault vibration judgment.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the vibration detection and analysis method for defects and faults in rail transit electromechanical equipment according to any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the vibration detection and analysis method for defects and faults in rail transit electromechanical equipment according to any one of claims 1-4.

Images