A method for detecting newly added line spectra in vibration signals of dry vacuum pumps

Abstract

The present invention belongs to the field of dry vacuum pumps, and specifically is a method for detecting newly added line spectra in the vibration signal of a dry vacuum pump, comprising the following steps: in the initial stage of operation of the dry vacuum pump, collecting the vibration signal of the dry vacuum pump and sending it to a processor to obtain the line spectrum position of the vibration signal power spectrum; mapping the line spectrum position characteristics of the vibration signal power spectrum into a feature vector, and saving it as a reference feature vector in the processor; during the operation of the dry vacuum pump during a set period of time, collecting the vibration signal of the dry vacuum pump in real time, and forming a feature vector under real-time working conditions into the processor; the processor obtains a fault evaluation index based on the reference feature vector and the feature vector under real-time working conditions; and judges the working state of the dry vacuum pump. The present invention avoids the computational complexity of fast Fourier transform in line spectrum detection, and only uses the finite impulse response filter method. On a lightweight processor, it can realize the detection of newly added spectrum lines in the vibration signal.

Description

Technical Field

[0001] The invention belongs to the field of dry vacuum pumps, and in particular relates to a method for detecting newly added line spectra in a vibration signal of a dry vacuum pump. Background Art

[0002] The extraction and identification of vibration signal features plays a crucial role in dry vacuum pump fault prediction. As a typical complex electromechanical system primarily composed of rotating machinery and bearings, the dry vacuum pump's operating vibration signal is characterized by a broad spectrum signal dominated by line spectrum groups. Currently, line spectrum detection is primarily performed on a computer through spectrum detection methods. Determining the health of the pump based on line spectrum mainly relies on manual observation, which is inefficient. Manual identification methods are also inaccurate and can lead to errors. Therefore, designing a new line spectrum detection method for vibration signals is particularly important. Summary of the Invention

[0003] In the fault diagnosis of dry vacuum pump vibration signals, changes in the line spectrum structure are often closely related to vacuum pump bearing failures. The purpose of the present invention is to provide a method for detecting newly added line spectra in dry vacuum pump vibration signals for lightweight processors. This method avoids the computational complexity of fast Fourier transforms in line spectrum detection and only uses the finite impulse response filter method. On lightweight processors such as the STM32F103, it can detect newly added spectral lines in the vibration signal, providing a new means for detecting vibration signals for faults such as rotor misalignment in dry vacuum pumps.

[0004] The technical solution adopted by the present invention to achieve the above-mentioned object is: a method for detecting newly added line spectra in a dry vacuum pump vibration signal, comprising the following steps:

[0005] 1) starting the dry vacuum pump, collecting a vibration signal of the dry vacuum pump in the initial stage of operation of the dry vacuum pump, and sending the collected vibration signal of the dry vacuum pump to a processor, and obtaining the line spectrum position of the power spectrum of the vibration signal according to a line spectrum extraction method;

[0006] 2) Mapping the line spectrum position characteristics of the vibration signal power spectrum into a feature vector to form a feature vector in the initial operation of the dry vacuum pump, and saving it as a reference feature vector in the processor;

[0007] 3) During the operation of the dry vacuum pump during the set period, the vibration signal of the dry vacuum pump is collected in real time, and a feature vector under the real-time working condition is generated and sent to the processor;

[0008] 4) The processor obtains a fault evaluation index based on the reference feature vector and the feature vector under the real-time working condition;

[0009] 5) Determine the working status of the dry vacuum pump based on the fault evaluation indicators.

[0010] The method of obtaining the line spectrum position of the vibration signal power spectrum according to the line spectrum extraction method includes the following steps:

[0011] 1-1) performing a fast Fourier transform calculation on the N points of the collected vibration signal;

[0012] Y(n)=FFT(S(n),N)

[0013] Wherein, FFT is the fast Fourier transform operation, n is the sample number, n=1, 2, 3, ..., N, S(n) is the input signal acquisition amount, and Y(n) is the Fourier spectrum value;

[0014] 1-2) Obtain the power spectrum X(n) by taking the square root of the amplitude of the Fast Fourier Transform result and dividing it by the number of N points multiplied by 2, and retain the first N / 2 points;

[0015] X(n)=2*abs(Y(n)) / N

[0016] Wherein, n is the sample number, n=1, 2, 3, ..., N / 2; abs is the modulo operation.

[0017] 1-3) Calculate the mean of the N / 2 points before 1 in the power spectrum X(n), and obtain the position number n of the N / 2 points before the power spectrum X(n) that is greater than the mean as the position number of the line spectrum.

[0018] The line spectrum position characteristics of the vibration signal power spectrum are mapped into a characteristic vector to form a characteristic vector in the initial operation of the dry vacuum pump, specifically:

[0019] 2-1) Define a feature vector as an array of M long integers, where each long integer contains N / 2M bits, resulting in a total of N / 2 bits in the array.

[0020] 2-2) According to the position serial number of the line spectrum obtained in the line spectrum extraction method, mark the N / 2 bits of this array respectively. The bit corresponding to the position where the line spectrum exists is 1, and the bit corresponding to the position where the line spectrum does not exist is 0. The complete M long integer data are obtained, that is, the characteristic vector of the dry vacuum pump vibration signal.

[0021] The step 4) is specifically as follows:

[0022] The processor performs an XOR operation on the currently collected feature vector and the reference feature vector according to the position number, and sums the results of the XOR operation according to the position number. The summation process starts from the highest bit of the position number. If one bit of the feature vector obtained after the XOR operation is 1, the evaluation index is increased by 1, and the final integer is used as the fault evaluation index.

[0023] The processor performs an XOR operation on the currently collected feature vector and the reference feature vector according to the position sequence number, that is:

[0024] R=SUM(A⊕B)

[0025] Among them, A is the currently collected eigenvector, B is the reference eigenvector, ⊕ is the XOR symbol, and R is the final result, reflecting the status of the newly added spectral line.

[0026] The working state of the dry vacuum pump is judged according to the fault evaluation index, specifically: the final result R is judged to reflect the number of newly added spectral lines. When R is greater than the set threshold, it reflects that the number of newly added spectral lines exceeds the set safety number, that is, the state of the pump has deteriorated.

[0027] In step 1), the collecting of the dry vacuum pump vibration signal is specifically: collecting the dry vacuum pump vibration signal by a sensor provided on the dry vacuum pump, wherein the sensor is a piezoelectric vibration sensor.

[0028] A device for detecting newly added line spectra in a dry vacuum pump vibration signal, comprising a memory and a processor;

[0029] The memory is used to store the computer program and the collected feature vectors or reference feature vectors;

[0030] The processor is configured to implement a method for detecting a newly added line spectrum in a vibration signal of a dry vacuum pump when executing the computer program.

[0031] A computer-readable storage medium stores a computer program. When a collected reference feature vector and a current real-time feature vector are stored simultaneously, the computer program, when executed by a processor, implements a method for detecting newly added line spectra in a dry vacuum pump vibration signal.

[0032] The present invention has the following beneficial effects and advantages:

[0033] 1. The present invention can provide real-time status assessment for bearing faults and motor faults that often occur in dry vacuum pumps, and can be used for fault prediction of dry vacuum pumps.

[0034] 2. This method has less computational complexity, fast calculation, and requires less computing resources, and can solve the problem of fault prediction on embedded systems.

[0035] 3. The present invention avoids the computational complexity of fast Fourier transform in line spectrum detection and only uses the finite impulse response filter method to realize the detection of newly added spectral lines in the vibration signal on a lightweight processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flowchart of the method of the present invention;

[0037] Figure 2 Schematic diagram of the XOR operation principle between feature vectors of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0039] like Figure 1 FIG. 1 is a flowchart of the method of the present invention. The present invention can provide real-time status assessment for bearing faults and motor faults that often occur in dry vacuum pumps, and can be used for fault prediction of dry vacuum pumps. Specifically, the method includes the following steps:

[0040] 1) Turning on the dry vacuum pump. In the initial stage of operation of the dry vacuum pump, a piezoelectric vibration sensor provided on the dry vacuum pump collects a vibration signal of the dry vacuum pump, and sends the collected vibration signal of the dry vacuum pump to a processor. Using a line spectrum extraction method, the line spectrum position of the vibration signal power spectrum is obtained;

[0041] In this embodiment, taking a 4096-point signal as an example, the vibration extraction method mainly performs a 4096-point FFT calculation on the 4096-point vibration acquisition signal, namely:

[0042] Performing N-point fast Fourier transform calculation on the collected vibration signal;

[0043] Y(n)=FFT(S(n),N)

[0044] Wherein, FFT is a fast Fourier transform operation, n is the sample number, n = 1, 2, 3, ..., 4096, S(n) is the input signal acquisition amount, and Y(n) is the Fourier spectrum value; the FFT calculation result is obtained by taking the square of the amplitude and dividing it by 4096 to obtain the power spectrum X(n), where n is the sample number.

[0045] X(n)=2*abs(Y(n)) / N

[0046] Where n is the sample number, n = 1, 2, 3, ..., 2048; abs is the modulo operation

[0047] Since the vibration signal is real-valued, the first 2048 points of the power spectrum X(n) are symmetrical with the last 2048 points. Therefore, only the first 2048 points are needed, i.e., X(n), n = 1:2048. Calculate the mean of the first 2048 points of the power spectrum X(n), and take the position number n of the first 2048 points of the power spectrum X(n) that is greater than the mean as the position number of the line spectrum.

[0048] 2) Mapping the line spectrum position characteristics of the vibration signal power spectrum into a feature vector to form a feature vector in the initial operation of the dry vacuum pump, and saving it as a reference feature vector in the processor;

[0049] The method for mapping position features to feature vectors is as follows: This embodiment uses the example of defining a feature vector as an array of 32 long integers. First, the feature vector is defined as an array of 32 long integers, each containing 64 bits, for a total of 2048 bits. The 2048 bits of this array are labeled according to the position numbers of the line spectra calculated in the line spectrum extraction method. Bits corresponding to positions with line spectra are set to 1, and bits corresponding to positions without line spectra are set to 0. These 32 long integers are the feature vectors of the vibration signal.

[0050] 3) During the operation of the dry vacuum pump during the set period, the vibration signal of the dry vacuum pump is collected in real time, and a feature vector under the real-time working condition is generated and sent to the processor;

[0051] 4) The processor obtains a fault evaluation index based on the reference feature vector and the feature vector under the real-time working condition;

[0052] like Figure 2 The figure shows the principle diagram of the XOR operation between feature vectors of the present invention. In step 4), the method for estimating the abnormal signal fault severity evaluation index is as follows: First, during the initial operation of the dry vacuum pump, a vibration signal is collected, the line spectrum positions are extracted using a line spectrum extraction method, and then the line spectrum position features are mapped into a feature vector to form a feature vector under normal operating conditions. This feature vector is 2048 bits, or 32 long integers. This feature vector is saved as a reference feature vector. Then, during the later operation of the dry vacuum pump, vibration signals are regularly collected, the line spectrum positions are extracted using a line spectrum extraction method, and then the line spectrum position features are mapped into a feature vector to form a feature vector under the current operating conditions. The feature vector under the current operating conditions is compared with the reference feature vector to form a fault evaluation index, thereby evaluating the operating status of the dry vacuum pump. The specific method for forming the fault evaluation index is to perform a bitwise XOR operation on the current feature vector and the reference feature vector, and then sum the results bit by bit. Starting from the most significant bit, if a bit in the feature vector obtained after the XOR operation is 1, the evaluation index is increased by 1. The resulting integer is used as the fault evaluation index.

[0053] The processor performs an XOR operation on the currently collected feature vector and the reference feature vector according to the position sequence number, that is:

[0054] R=SUM(A⊕B)

[0055] Among them, A is the currently collected eigenvector, B is the reference eigenvector, ⊕ is the XOR symbol, and R is the final result, reflecting the status of the newly added spectral line.

[0056] 5) Determine the working status of the dry vacuum pump based on the fault evaluation index. Specifically, the final result R reflects the number of newly added spectral lines. When R is greater than the set threshold, it indicates that the number of newly added spectral lines exceeds the set safety number, indicating that the pump status has deteriorated.

[0057] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for detecting newly added line spectra in a dry vacuum pump vibration signal, characterized in that: The following steps are involved: 1) starting the dry vacuum pump, collecting a vibration signal of the dry vacuum pump in the initial stage of operation of the dry vacuum pump, and sending the collected vibration signal of the dry vacuum pump to a processor, and obtaining the line spectrum position of the power spectrum of the vibration signal according to a line spectrum extraction method; The method of obtaining the line spectrum position of the vibration signal power spectrum according to the line spectrum extraction method includes the following steps: 1-1) performing a fast Fourier transform calculation on N points of the collected vibration signal; Y(n)=FFT(S(n),N) Wherein, FFT is the fast Fourier transform operation, n is the sample number, n = 1, 2, 3, ..., N, S(n) is the input signal, and Y(n) is the Fourier spectrum value; 1-2) Obtain the power spectrum X(n) by taking the square root of the amplitude of the Fast Fourier Transform result and dividing it by the number of N points multiplied by 2, and retain the first N / 2 points; X(n)=2*abs(Y(n)) / N Where n is the sample number, n = 1, 2, 3, ..., N / 2; abs is the modulo operation; 1-3) Calculate the mean of the first N / 2 points in the power spectrum X(n), and obtain the position number n of the first N / 2 points in the power spectrum X(n) that is greater than the mean as the position number of the line spectrum; 2) Mapping the line spectrum position characteristics of the vibration signal power spectrum into a feature vector to form a feature vector in the initial operation of the dry vacuum pump, and saving it as a reference feature vector in the processor; The line spectrum position characteristics of the vibration signal power spectrum are mapped into a characteristic vector to form a characteristic vector in the initial operation of the dry vacuum pump, specifically: 2-1) Define a feature vector as an array of M long integers, where each long integer contains N / 2M bits, resulting in a total of N / 2 bits in the array. 2-2) According to the position serial number of the line spectrum obtained in the line spectrum extraction method, mark the N / 2 bits of this array respectively. The bit corresponding to the position where the line spectrum exists is 1, and the bit corresponding to the position where the line spectrum does not exist is 0. The complete M long integer data are obtained, that is, the feature vector of the dry vacuum pump vibration signal; 3) During the operation of the dry vacuum pump during the set period, the vibration signal of the dry vacuum pump is collected in real time, and a feature vector under the real-time working condition is generated and sent to the processor; 4) The processor obtains a fault evaluation index based on the reference feature vector and the feature vector under the real-time working condition; 5) Determine the working status of the dry vacuum pump based on the fault evaluation indicators.

2. A method for detecting newly added line spectra in a dry vacuum pump vibration signal according to claim 1, characterized in that: Step 4) is as follows: The processor performs an XOR operation on the currently collected feature vector and the reference feature vector according to the position number, and sums the results of the XOR operation according to the position number. The summation process starts from the highest bit of the position number. If one bit of the feature vector obtained after the XOR operation is 1, the evaluation index is increased by 1, and the final integer is used as the fault evaluation index.

3. A method for detecting newly added line spectrum in a dry vacuum pump vibration signal according to claim 2, characterized in that: The processor performs an XOR operation on the currently collected feature vector and the reference feature vector according to the position sequence number, that is: R=SUM(A⊕B) Among them, A is the currently collected eigenvector, B is the reference eigenvector, ⊕ is the XOR symbol, and R is the final result, reflecting the status of the newly added spectral line.

4. A method for detecting newly added line spectra in a dry vacuum pump vibration signal according to claim 1, characterized in that: The working state of the dry vacuum pump is judged according to the fault evaluation index, specifically: the final result R is judged to reflect the number of newly added spectral lines. When R is greater than the set threshold, it reflects that the number of newly added spectral lines exceeds the set safety number, that is, the state of the pump has deteriorated.

5. The method for detecting newly added line spectrum in a dry vacuum pump vibration signal according to claim 1, characterized in that: In step 1), the collecting of the dry vacuum pump vibration signal is specifically: collecting the dry vacuum pump vibration signal by a sensor provided on the dry vacuum pump, wherein the sensor is a piezoelectric vibration sensor.

6. A device for detecting newly added line spectra in a dry vacuum pump vibration signal, characterized in that: including memory and processor; The memory is used to store the computer program and the collected feature vectors or reference feature vectors; The processor is configured to implement a method for detecting newly added line spectra in a vibration signal of a dry vacuum pump as described in any one of claims 1 to 5 when executing the computer program.

7. A computer-readable storage medium, characterized in that A computer program is stored on the storage medium. When the collected reference feature vector and the current real-time feature vector are stored at the same time, the computer program is executed by the processor to implement a method for detecting a newly added line spectrum in a dry vacuum pump vibration signal as described in any one of claims 1 to 5.

Images