EMD-based vibration signal rotation speed identification method
Through the vibration signal speed recognition method based on EMD, high-pass filtering and EMD decomposition technology are used to eliminate the influence of high frequency and select related modal components, accurate speed recognition within the range of 0RPM-30000RPM is achieved, and the large error sum in the existing technology is solved The problem of poor versatility improves recognition accuracy and range.
Patent Information
- Application Number
- CN202410920924.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-07-04
- Filing Date
- 2024-07-10
- Publication Date
- 2024-11-08
AI Technical Summary
In the prior art, the vibration signal speed recognition method has large errors, high cost, high construction difficulty, experience dependence, low recognition accuracy in noise environment, poor method versatility, and low or high speeds, especially in high frequency bands. The correct speed cannot be extracted.
Using the vibration signal speed recognition method based on EMD, high-pass filtering is performed by obtaining the vibration acceleration signal of the device, high-frequency structure resonance frequency signal is eliminated, EMD decomposition is performed, and relevant eigenmodal function components are selected for signal reconstruction, and the sum of the FFT, obtain the velocity spectrum, and obtain the difference velocity spectrum by smooth filtering and subtracting the velocity average energy spectrum, and take its maximum value as the rotation frequency output.
It realizes speed identification without the specified speed range, covers the speed range of 0RPM-30000RPM, eliminates the impact of high frequency on speed frequency calculation, improves the fundamental frequency proportion of rotation frequency, clears low frequency noise, and accurately recognizes high-speed and low-speed speeds. .
Smart Images

Figure CN118916669A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of signal recognition, and in particular relates to a vibration signal rotation speed recognition method based on EMD. Background Art
[0002] In order to accurately diagnose the health status of rotating equipment, the rotational speed is an indispensable condition.
[0003] At present, there are three main methods for speed identification:
[0004] The first method is to use the rated speed of the equipment as the speed of the equipment. Since the equipment may be variable frequency and its actual speed will change with the change of equipment load, the speed of the equipment is usually not equal to the rated speed, so this method has a large error;
[0005] The second method is to install a speed sensor to directly collect speed data, but it is difficult to implement in engineering due to cost constraints, construction difficulties, and difficulty in synchronous speed measurement in wireless systems.
[0006] The third method is to directly identify the rotation speed through the collected vibration data. The invention patent: A method for identifying the rotation speed of equipment based on vibration signal data drive, authorization announcement number: CN 110633686 B, summarizes the current methods of obtaining the rotation speed using vibration signals. Engineers determine the rotation speed of the equipment by analyzing the spectrum or envelope spectrum corresponding to the vibration signal. In these methods, the accuracy of speed identification depends on the experience of engineers, which is highly subjective and subject to the complexity of the vibration signal. It requires accurate prior parameters, has low identification accuracy in a strong noise environment, cannot be extracted when the high frequency band has high energy, and has poor method versatility. The method proposed in this invention patent has certain application value, but its algorithm specifies the effective frequency band [4Hz-100Hz], which is actually equivalent to estimating the rotation speed range. It can only identify the rotation speed range of 240RPM-6000RPM, and cannot identify lower and higher rotation speeds. In addition, this method will extract the wrong rotation speed when the double frequency or higher frequency occupies the main component.
[0007] Empirical Mode Decomposition (EMD) is a new adaptive signal time-frequency processing method. This method decomposes the signal based on the time scale characteristics of the data itself without presetting any basis functions. It is particularly suitable for the analysis and processing of nonlinear and non-stationary signals. Summary of the invention
[0008] Technical problem: In view of the above problems existing in the prior art, the technical problem to be solved by the present invention is to propose a vibration signal speed identification method based on EMD, which does not require a specified speed range and can effectively eliminate the influence of high frequency on the speed frequency calculation.
[0009] Technical solution: In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:
[0010] A vibration signal rotation speed identification method based on EMD, comprising the following steps:
[0011] S1. Obtaining the original acceleration signal of device vibration;
[0012] S2, pre-processing the acceleration signal, high-pass filtering, and removing the high-frequency structural resonance frequency signal;
[0013] S3, performing EMD decomposition on the preprocessed signal;
[0014] S4, selecting an intrinsic mode function component related to the original signal, and reconstructing the signal of the selected relevant intrinsic mode function component;
[0015] S5, integrating the reconstructed signal, and performing FFT after integration to obtain a reconstructed velocity spectrum;
[0016] S6, performing smoothing filtering on the reconstructed velocity spectrum to obtain a velocity average energy spectrum;
[0017] S7. Subtract the velocity average energy spectrum from the reconstructed velocity spectrum to obtain a difference velocity spectrum, and take the maximum value of the difference velocity spectrum as the rotation frequency output.
[0018] Furthermore, in step S2, the high-pass filter frequency is 500 Hz.
[0019] Further, the step S4 includes the following steps:
[0020] S4.1. Calculate the correlation coefficient ρ between the original signal and each layer’s intrinsic mode function component according to the following formula:
[0021]
[0022] Where: x is the original signal, y is the reconstructed signal, ρ is the correlation coefficient, Co(x, y) is the covariance, σ is the standard deviation, and n is the total number of calculated data;
[0023] S4.2. Select relevant intrinsic mode function components through the correlation coefficient ρ, and select the intrinsic mode function components with a correlation coefficient of 0.3≤ρ<0.5 for signal reconstruction.
[0024] Beneficial effects: Compared with the prior art, the present invention has the following advantages:
[0025] 1. The high-pass filter frequency is 500Hz. Compared with the existing algorithm effective frequency band [4Hz-100Hz], the speed range covers 0RPM-30000RPM (frequency*60 is the speed), covering rotating equipment below 30000RPM, and can be used for speed identification of most equipment;
[0026] 2. The influence of high frequency on the speed frequency calculation can be effectively eliminated through high-pass filtering, and the intrinsic mode function components with a correlation coefficient of 0.3≤ρ<0.5 are selected for signal reconstruction to further weaken the influence of non-rotation frequency, retain the rotation frequency, and increase the proportion of the rotation frequency base frequency;
[0027] 3. Subtract the average energy spectrum of the velocity from the reconstructed velocity spectrum to obtain the difference velocity spectrum and remove the influence of low-frequency noise;
[0028] 4. It can identify both high speed and low speed very well. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 is a flow chart of the method of the present invention;
[0030] Figure 2 It is the vibration acceleration time domain diagram and vibration acceleration frequency domain diagram of working condition 1;
[0031] Figure 3 This is the vibration velocity spectrum after the working condition 1 in Example 1 is processed according to the existing identification method;
[0032] Figure 4 It is the time domain diagram and frequency domain diagram of the signal after preprocessing in Example 1;
[0033] Figure 5 This is the spectrum after the working condition 1 in Example 1 is processed according to the method of the present invention;
[0034] Figure 6 This is the spectrum after the working condition 2 in Example 2 is processed by the existing identification method;
[0035] Figure 7 1 is a vibration acceleration time domain diagram and a vibration acceleration frequency domain diagram of the signal of working condition 2 in Example 2;
[0036] Figure 8 It is the time domain diagram and frequency domain diagram of the signal after preprocessing under working condition 2 in Example 2;
[0037] Fig. 9 This is the spectrum after the working condition 2 in Example 2 is processed according to the method of the present invention;
[0038] Fig.10 yes Fig. 9 Enlarged view of the low-frequency band (1Hz-50Hz) of the spectrum. DETAILED DESCRIPTION
[0039] The present invention is further illustrated below in conjunction with specific examples. The examples are implemented based on the technical solutions of the present invention. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention.
[0040] Comparative Example 1
[0041] Working condition 1:
[0042] Motor drive end data of the vulcanizing blower;
[0043] Measurement parameters: vibration acceleration data;
[0044] Sampling rate: 5120;
[0045] Number of sampling points: 16384;
[0046] Actual speed: about 2983rpm;
[0047] Fault: Loose;
[0048] like Figure 2 As shown in the figure, the motor of the vulcanizing blower cannot obtain the accurate speed through the ordinary FFT peak recognition method. The method introduced in the already disclosed invention patent: A method for identifying the speed of equipment based on vibration signal data drive, authorization announcement number: CN 110633686B is used for processing. The processed graph is as follows Figure 3 As shown, although the proportion of rotational speed is increased by integration according to this method, because the double frequency is dominant, the target frequency in the vibration velocity spectrum is output as the current rotational frequency of the equipment. In the figure, X represents the value of the horizontal axis, and Y represents the value of the vertical axis. The target frequency is: 99.375, so the rotational speed will be incorrectly diagnosed as 99.375*60=5962.5rpm.
[0049] Example 1
[0050] Working condition 1:
[0051] Motor drive data of vulcanizing blower;
[0052] Measurement parameters: vibration acceleration data;
[0053] Sampling rate: 5120;
[0054] Number of sampling points: 16384;
[0055] Actual speed: about 2983rpm;
[0056] Fault: Loose;
[0057] For the speed identification of the vulcanizing blower motor under the same working condition as in Comparative Example 1, this embodiment proposes a vibration signal speed identification method based on EMD, comprising the following steps:
[0058] S1. Obtain the original acceleration signal of the equipment vibration. The vibration acceleration time domain diagram and vibration acceleration frequency domain diagram of the signal are as follows: Figure 2 shown.
[0059] S2. Preprocess the acceleration signal, high-pass filter, filter frequency is 500Hz, remove the high-frequency structural resonance frequency signal, improve the signal-to-noise ratio of the frequency conversion signal; the time domain and frequency domain diagrams of the preprocessed signal are as follows Figure 4 shown.
[0060] S3. Use the adaptive time-frequency data analysis method empirical mode decomposition to perform EMD decomposition on the signal data.
[0061] S4, selecting the intrinsic mode function components related to the original signal, and reconstructing the selected relevant intrinsic mode function components, including the following steps:
[0062] S4.1. Calculate the correlation coefficient ρ between the original signal and each layer’s intrinsic mode function component according to the following formula:
[0063]
[0064] Where: x is the original signal, y is the reconstructed signal, ρ is the correlation coefficient, Cov(x, y) is the covariance, and σ is the standard deviation. i is the x value in the ith data, is the mean of x, y i is the y value in the i-th data, is the average value of y, and n is the total number of data calculated;
[0065] S4.2. Select the relevant intrinsic mode function components through the correlation coefficient ρ, and select the intrinsic mode function components with the correlation coefficient 0.3≤ρ<0.5 for signal reconstruction; for acceleration signals, high frequency generally dominates, and the purpose of reconstruction is to further weaken the influence of non-rotation frequency, retain the rotation frequency, and increase the proportion of the rotation frequency base frequency;
[0066] When the correlation coefficient ρ<0.3, the original signal and the intrinsic mode function component are weakly correlated or extremely weakly correlated; when 0.3≤ρ<0.5, the original signal and the intrinsic mode function component are moderately correlated; when 0.6≤ρ<0.8, the original signal and the intrinsic mode function component are strongly correlated; when 0.8≤ρ<1, the original signal and the intrinsic mode function component are extremely strongly correlated; the moderate correlation of 0.3≤ρ<0.5 is selected because the rotation frequency contributes to the vibration acceleration signal, but the frequency higher than the rotation frequency often contributes more, and the purpose is to eliminate the influence of this part of the frequency on the rotation frequency identification.
[0067] S5. Integrate the reconstructed signal in the frequency domain, and obtain the reconstructed velocity spectrum by FFT (Fast Fourier Transform) after integration, such as Figure 5 shown.
[0068] S6. Smooth filter the reconstructed velocity spectrum to obtain the velocity average energy spectrum, such as Figure 5 shown.
[0069] S7. Subtract the velocity average energy spectrum from the reconstructed velocity spectrum to obtain the difference velocity spectrum, such as Figure 5 As shown, the maximum value of the difference speed spectrum is taken as the frequency output; the purpose of this step is mainly to eliminate the influence of low-frequency noise. When integrating, the low-frequency band of the non-rotation frequency is amplified as a whole, and the difference is just taken to remove the influence of the low-frequency integral noise; X represents the value of the horizontal axis, and Y represents the value of the vertical axis. The maximum value on the difference speed spectrum is 49.6875Hz, and the identified speed is 49.6875*60=2981.25rpm, which is basically consistent with the actual speed.
[0070] Comparative Example 2
[0071] Working condition 2:
[0072] Axial bearing seat of the slow drive side of the SAG mill;
[0073] Measurement parameters: vibration acceleration data;
[0074] Sampling rate: 51200;
[0075] Number of sampling points: 262144;
[0076] Actual speed: about 200rp m
[0077] Fault: bearing failure;
[0078] The invention patent: A method for identifying the rotation speed of equipment based on vibration signal data drive, authorization announcement number: CN110633686 B, is used for processing. The processed graphics are as follows Figure 6As shown, it can be clearly seen from the vibration acceleration time domain diagram that there is an impact signal modulated by the rotation frequency, which is also reflected in the vibration velocity spectrum. However, because the frequency of 53.32 is dominant in the vibration velocity spectrum, the speed will be mistakenly diagnosed as 53.32*60=3199.2rpm / min. Therefore, the speed identified by this method is very different from the actual speed.
[0079] Example 2
[0080] Working condition 2:
[0081] SAG mill slow drive side bearing seat axial
[0082] Measurement parameters: vibration acceleration data
[0083] Sampling rate: 51200;
[0084] Number of sampling points: 262144;
[0085] Actual speed: about 200rpm;
[0086] Fault: bearing failure;
[0087] For the semi-autogenous mill under the same working condition as in Example 2, the present embodiment proposes a vibration signal speed identification method based on EMD, comprising the following steps:
[0088] S1. Obtain the original acceleration signal of the equipment vibration. The vibration acceleration time domain diagram and vibration acceleration frequency domain diagram of the signal are as follows: Figure 7 shown.
[0089] S2. Preprocess the acceleration signal, high-pass filter, filter frequency is 500Hz, remove the high-frequency structural resonance frequency signal, improve the signal-to-noise ratio of the frequency conversion signal; the time domain and frequency domain diagrams of the preprocessed signal are as follows Figure 8 shown.
[0090] S3. Use the adaptive time-frequency data analysis method empirical mode decomposition to perform EMD decomposition on the signal data.
[0091] S4, selecting the intrinsic mode function components related to the original signal, and reconstructing the selected relevant intrinsic mode function components, including the following steps:
[0092] S4.1. Calculate the correlation coefficient ρ between the original signal and each layer’s intrinsic mode function component according to the following formula:
[0093]
[0094] Where: x is the original signal, y is the reconstructed signal, ρ is the correlation coefficient, Cov(x, y) is the covariance, and σ is the standard deviation. iis the x value in the ith data, is the mean of x, y i is the y value in the i-th data, is the average value of y, and n is the total number of data calculated;
[0095] S4.2. Select the relevant intrinsic mode function components through the correlation coefficient ρ, and select the intrinsic mode function components with a correlation coefficient of 0.3≤ρ<0.5 for signal reconstruction. For acceleration signals, high frequencies generally dominate, and the purpose of reconstruction is to further weaken the influence of non-rotational frequencies, retain rotational frequencies, and increase the proportion of rotational frequency fundamental frequencies:
[0096] When the correlation coefficient ρ<0.3, the original signal and the intrinsic mode function component are weakly correlated or extremely weakly correlated; when 0.3≤ρ<0.5, the original signal and the intrinsic mode function component are moderately correlated; when 0.6≤ρ<0.8, the original signal and the intrinsic mode function component are strongly correlated; when 0.8≤ρ<1, the original signal and the intrinsic mode function component are extremely strongly correlated.
[0097] S5. Integrate the reconstructed signal in the frequency domain and obtain the reconstructed velocity spectrum by FFT after integration, such as Fig. 9 and Fig.10 shown.
[0098] S6. Smooth filter the reconstructed velocity spectrum to obtain the velocity average energy spectrum, such as Fig. 9 and Fig.10 shown.
[0099] S7. Subtract the velocity average energy spectrum from the reconstructed velocity spectrum to obtain the difference velocity spectrum, such as Fig. 9 and Fig.10 As shown, the maximum value of the difference speed spectrum is taken as the frequency output; the purpose of this step is mainly to eliminate the influence of low-frequency noise. When integrating, the low-frequency band of the non-rotating frequency is amplified as a whole, and the difference is just taken to remove the influence of the low-frequency integral noise; in the figure, X represents the value of the horizontal axis, and Y represents the value of the vertical axis. The maximum value on the difference speed spectrum is 3.32, and the identified speed is 3.32*60=199.2rpm / min, which is basically consistent with the actual speed.
[0100] Through Example 1 and Example 2, it can be found that the output spectrum after being processed by the method of the present invention can clearly identify the frequency through the peak of the difference spectrum (difference speed spectrum) and the output speed frequency from the difference spectrum, without the misidentification of the existing method in comparison. The identified speed is basically consistent with the actual speed, and both high and low speeds can be well identified.
[0101] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A vibration signal rotation speed identification method based on EMD, characterized in that: The following steps are involved: S1. Obtaining the original acceleration signal of device vibration; S2, pre-processing the acceleration signal, high-pass filtering, and removing the high-frequency structural resonance frequency signal; S3, performing EMD decomposition on the preprocessed signal; S4, selecting an intrinsic mode function component related to the original signal, and reconstructing the signal of the selected relevant intrinsic mode function component; S5, integrating the reconstructed signal, and performing FFT after integration to obtain a reconstructed velocity spectrum; S6, performing smoothing filtering on the reconstructed velocity spectrum to obtain a velocity average energy spectrum; S7. Subtract the velocity average energy spectrum from the reconstructed velocity spectrum to obtain a difference velocity spectrum, and take the maximum value of the difference velocity spectrum as the rotation frequency output.
2. The vibration signal rotation speed identification method based on EMD according to claim 1 is characterized in that: The high-pass filter frequency in step S2 is 500 Hz.
3. The vibration signal rotation speed identification method based on EMD according to claim 1 is characterized in that: The step S4 comprises the following steps: S4.
1. Calculate the correlation coefficient ρ between the original signal and each layer’s intrinsic mode function component according to the following formula: Where: x is the original signal, y is the reconstructed signal, ρ is the correlation coefficient, Cov(x, y) is the covariance, σ is the standard deviation, and n is the total number of calculated data; S4.
2. Select relevant intrinsic mode function components through the correlation coefficient ρ, and select the intrinsic mode function components with a correlation coefficient of 0.3≤ρ<0.5 for signal reconstruction.
Citation Information
Patent Citations
Revolving speed curve estimation method based on combination of EEMD-HHT (Ensemble Empirical Mode Decomposition-Hilbert-Huang Transform) and time-frequency reassignment
CN107679013A
Equipment rotating speed identification method based on vibration signal data driving
CN110633686A
Washing machine, method for controlling drum rotating speed, and program
JP2009165682A
Speed and position sensing device for EMD two-cycle diesel engines
US20110197441A1
Cited By
Full-band beat vibration identification method for rotating equipment
CN120721213A
