Petrochemical pipeline corrosion vibration signal feature extraction method

Abstract

The invention discloses a petrochemical pipeline corrosion vibration signal feature extraction method, and particularly relates to a signal processing method for vibration signal change caused by structure damage due to corrosion of a petrochemical pipeline. The system mainly comprises a low-reflectivity fiber grating vibration sensor, a high-speed vibration signal demodulator, an upper computer and a pipeline corrosion vibration signal processing program. According to the system, a low-reflectivity fiber grating sensor is specially packaged into a cable, and the sensor is wound and arranged along a petrochemical device pipeline and is fixed through high-temperature-resistant glue, so that vibration signals generated in the working process of the pipeline are fully transmitted. And identifying the working state of the pipeline by using the modal identification method provided by the invention. The system has the advantages of simple structure, high real-time performance, accurate identification and the like. Real-time online monitoring in a petrochemical complex working condition environment can be realized, and the production efficiency and the working safety are greatly improved.

Description

Technical Field

[0001] The invention relates to a vibration signal characteristic processing method for online monitoring of corrosion conditions of petrochemical pipelines. Background Art

[0002] As an economical and environmentally friendly means of transportation, pipeline transportation is responsible for transporting gas, liquid, slurry and other means of production in the petrochemical industry. However, due to the special substances it transports, chemical corrosion often occurs during its operation. In order to avoid disasters caused by long-term corrosion of petrochemical pipelines, it is necessary to monitor the corrosion of pipelines.

[0003] At present, there are mainly the following methods for monitoring corrosion of petrochemical pipelines:

[0004] The pipeline corrosion detection method introduced in the Chinese patent (patent number CN202221015972.4) "A Corrosion Monitoring Device for Oil and Gas Pipelines" designs two arc blocks, clamps them on the oil and gas pipelines to be tested, and obtains pipe wall data by rolling the rolling parts on the inner wall on the surface of the oil and gas pipelines. This method can more accurately obtain the changes in the pipe wall caused by corrosion and deformation of the oil and gas pipeline, but the measurement is time-consuming and labor-intensive. The monitoring device needs to be installed at the corresponding position for each measurement point. In addition, in petrochemical production, corrosion often occurs at pipe bends and changes in pipe diameter, so the application of this device is very limited.

[0005] The Chinese patent (patent number CN202110243852.3) "An optical fiber monitoring device and method for underground metal pipeline corrosion in subways" introduces that an excitation current is applied to the metal pipeline through an excitation unit, and the generated magnetic field will propagate along the sensing unit to cause the magnetostrictive composite material to deform, and finally the fiber grating sensor attached to the magnetostrictive material is used to monitor the deformation. This method determines the corrosion condition of the metal pipeline by the difference between adjacent sensing units, but for the petrochemical industry, it is often not allowed to apply current excitation to the working pipeline, and this method is also difficult to adapt to industry requirements.

[0006] At present, the domestic petrochemical industry pipeline corrosion monitoring means mostly use various mechanical instruments and electromagnetic sensors for monitoring. Mechanical instruments rely on regular manual inspections to read data, and the impact on normal production work must also be considered during the monitoring process; while electrical sensors often require a certain amplification circuit to work properly during use, which brings certain hidden dangers to safe construction. Therefore, a new pipeline corrosion monitoring technology is urgently needed. Summary of the invention

[0007] To solve the above problems, the present invention proposes a method for extracting features of petrochemical pipeline corrosion vibration signals. The system can realize online monitoring of the corrosion status of petrochemical pipelines, solve the current problems of untimely and inaccurate manual inspections, and improve production safety.

[0008] To realize the above invention, the present invention adopts the following technical solutions:

[0009] A system and method for extracting characteristics of vibration signals of petrochemical pipeline corrosion mainly include a low-reflectivity fiber Bragg grating vibration sensor, a high-speed vibration signal demodulator, a host computer and a pipeline corrosion vibration signal processing program. The system collects vibration signals generated during the operation of petrochemical pipelines through fiber Bragg grating vibration sensors, that is, low-reflectivity fiber Bragg grating vibration sensors are arranged along the petrochemical pipelines, and then the collected optical signals are converted into electrical signals through a high-speed vibration signal demodulator and transmitted to the host computer, and the characteristic information in the signal is extracted using a modal recognition method to determine the corrosion damage of the pipeline.

[0010] The sensor of the present invention adopts a low-reflectivity optical fiber grating vibration sensor.

[0011] The original vibration signal of the present invention is collected and denoised by wavelet transform to filter out complex environmental noise, as follows:

[0012] The original signal is:

[0013] s(n)=f(n)+σe(n)

[0014] Where e(n) is the noise, σ is the noise intensity, and the Birge-Massart model is used to determine the threshold required for denoising the coefficients of each layer

[0015] S1. Given a specified decomposition level j, all coefficients of j+1 and higher levels are retained;

[0016] S2. For the i-th layer (1≤i≤j), keep the n with the largest absolute value i coefficients, n i Determined by the following formula:

[0017] n i =M(j+2-i) α

[0018] Where M and α are empirical coefficients. By default, M = L(1), which is the length of the sparseness after the first layer of decomposition; in general, M satisfies L(1)≤M≤2L(1); the value of α is generally 3.

[0019] The denoised signal of the present invention uses a modal decomposition method to extract features, and the specific method is as follows:

[0020] S1. Obtaining vibration data of the petrochemical pipeline under working excitation according to the vibration signal collected by the sensor;

[0021] S2. Calculate the power spectrum and the power spectrum matrix according to the obtained vibration data;

[0022] S3, performing singular value decomposition on the obtained power spectrum matrix to obtain an observable matrix and a controllable inversion matrix;

[0023] S4, calculating the structural state matrix through the first two order observable matrices;

[0024] S5, performing eigendecomposition on the obtained structural state matrix, and calculating modal parameters according to the obtained eigenvalues ​​and eigenvectors;

[0025] S6. Perform fuzzy C-means clustering analysis on the obtained modal frequencies to obtain the cluster center of each modal cluster;

[0026] S7. Use the mean phase shift MPD to determine the order of the mode and obtain the final modal parameter identification result.

[0027] The beneficial effects produced by the present invention are:

[0028] A system and method for extracting vibration signal characteristics of petrochemical pipeline corrosion is proposed. Since optical fiber sensing has small size and high sensitivity, it is both a sensing unit and a transmission medium, and has the characteristics of anti-electromagnetic interference intrinsic explosion-proof, it is suitable as a sensor for petrochemical pipelines in complex environmental conditions after packaging. By winding and arranging low-reflectivity fiber Bragg grating vibration sensors along the petrochemical pipeline, the vibration signal changes caused by corrosion deformation of the petrochemical pipeline are collected to infer the corrosion of the pipeline, and the real-time online monitoring of the corrosion of the petrochemical pipeline is realized. The location and severity of corrosion can be accurately determined. It has the advantages of small size, easy installation, anti-electromagnetic interference, and real-time monitoring. It solves the shortcomings of traditional monitoring methods, such as the need for regular manual inspections, which is time-consuming and labor-intensive, and poor ability to resist sudden risks, and improves production safety.

[0029] Unlike the corrosion of oil and gas pipelines, which is mainly pinhole corrosion, the corrosion of petrochemical plants is the thinning of the inner wall of the pipeline at a certain position inside the corrosion zone. The vibration frequency becomes lower after disturbance in the transported gas and liquid medium. The present invention can measure the position by detecting the changes in this part. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of a system for extracting characteristics of vibration signals of petrochemical pipeline corrosion according to the present invention;

[0031] Numbers in the figure: 1. Low reflectivity fiber Bragg grating vibration sensor, 2. High-speed vibration signal demodulator, 3. Host computer;

[0032] Figure 2 It is a schematic diagram of the signal processing flow adopted by the present invention. DETAILED DESCRIPTION

[0033] The following is a further detailed description of a system and method for extracting vibration signal features of a petrochemical pipeline corrosion according to the present invention in conjunction with specific embodiments and drawings.

[0034] like Figure 1 As shown, a petrochemical pipeline corrosion vibration signal feature extraction system includes a low reflectivity fiber Bragg grating vibration sensor 1, a high-speed vibration signal demodulator 2, and a host computer 3. The system collects the vibration signal generated during the operation of the petrochemical pipeline through the fiber Bragg grating vibration sensor, and then converts the collected optical signal into an electrical signal through the high-speed vibration signal demodulator and transmits it to the host computer, and uses the modal recognition method to extract the feature information in the signal to determine the corrosion damage of the pipeline.

[0035] like Figure 2 As shown in the figure, after the original vibration signal is collected, the wavelet transform method is used to denoise and filter out the complex environmental noise, as follows:

[0036] The original signal is:

[0037] s(n)=f(n)+σe(n)

[0038] Where e(n) is the noise, σ is the noise intensity, and the Birge-Massart model is used to determine the threshold required for denoising the coefficients of each layer

[0039] S1. Given a specified decomposition level j, all coefficients of j+1 and higher levels are retained;

[0040] S2. For the i-th layer (1≤i≤j), keep the n with the largest absolute value i coefficients, n i Determined by the following formula:

[0041] n i =M(j+2-i) α

[0042] Where M and α are empirical coefficients. By default, M=L(1), which is the length of the sparseness after the first layer of decomposition. In general, M satisfies L(1)≤M≤2L(1). The value of α is 3.

[0043] For the original vibration signal collected by the sensor, the wavelet transform method is first used to remove noise, and the threshold required for noise reduction of each layer coefficient is determined by the Birge-Massart model:

[0044] S1. Given a specified decomposition level j, all coefficients of j+1 and higher levels are retained;

[0045] S2. For the i-th layer (1≤i≤j), keep the n with the largest absolute value i coefficients, n i Determined by the following formula:

[0046] n i =M(j+2-i) α

[0047] Where M and α are empirical coefficients. By default, M = L(1), which is the length of the sparse layer after the first layer of decomposition; in general, M satisfies L(1)≤M≤2L(1); the value of α is generally 3.

[0048] Then the feature extraction is performed through the modal decomposition method:

[0049] S1. Obtaining vibration data of the petrochemical pipeline under working excitation according to the vibration signal collected by the sensor;

[0050] S2. Calculate the power spectrum and the power spectrum matrix according to the obtained vibration data;

[0051] S3, performing singular value decomposition on the obtained power spectrum matrix to obtain an observable matrix and a controllable inversion matrix;

[0052] S4, calculating the structural state matrix through the first two order observable matrices;

[0053] S5, performing eigendecomposition on the obtained structural state matrix, and calculating modal parameters according to the obtained eigenvalues ​​and eigenvectors;

[0054] S6. Perform fuzzy C-means clustering analysis on the obtained modal frequencies to obtain the cluster center of each modal cluster;

[0055] S7. Use the mean phase shift MPD to determine the order of the mode and obtain the final modal parameter identification result.

[0056] By comparing the historical modal data currently monitored, the corrosion degree of the pipeline at the current location can be inferred.

Claims

1. A method for extracting characteristics of vibration signals of petrochemical pipeline corrosion, comprising a low-reflectivity fiber Bragg grating vibration sensor, a high-speed vibration signal demodulator and a host computer; characterized in that: The original vibration signal generated during the operation of the petrochemical pipeline is collected using a fiber grating sensor with intrinsic explosion-proof characteristics. That is, a low-reflectivity fiber grating vibration sensor is arranged along the petrochemical pipeline, and the environmental noise under most complex petrochemical working conditions is removed by the wavelet transform method; then the collected optical signal is converted into an electrical signal by a high-speed vibration signal demodulator and transmitted to the host computer; the denoised signal is modally decomposed to extract the characteristic signals under different modes; the modal recognition method is used to extract the characteristic information in the signal to determine the corrosion damage of the pipeline.

2. The method for extracting characteristics of a petrochemical pipeline corrosion vibration signal according to claim 1 is characterized in that: The modal decomposition is performed by calculating the power spectrum matrix and performing singular value decomposition on it to obtain an observable matrix, and performing eigenvector calculation on the obtained first two-order observable matrices and the structural state matrix to obtain modal parameters; then the fuzzy C-means clustering analysis and the mean phase shift MPD are used to determine the order of the mode; the corrosion damage of the pipeline at the same sensor monitoring position is analyzed by comparing and analyzing the same modal signals at the same sensor monitoring position; the upper computer software is used to set the pipeline corrosion alarm threshold, and once the collected signal exceeds the threshold, an automatic alarm will be implemented to avoid disasters.

3. The method for extracting characteristics of petrochemical pipeline corrosion vibration signals according to claim 1 is characterized in that: After the original vibration signal is collected, the wavelet transform method is used to denoise and filter out the complex environmental noise, as follows: The original signal is: s(n)=f(n)+σe(n) Where e(n) is the noise, σ is the noise intensity, and the Birge-Massart model is used to determine the threshold required for denoising the coefficients of each layer S1. Given a specified decomposition level j, all coefficients of j+1 and higher levels are retained; S2. For the i-th layer (1≤i≤j), keep the n with the largest absolute value i coefficients, n i Determined by the following formula: n i =M(j+2-i) α Where M and α are empirical coefficients; by default, M = L(1), which is the length of the sparseness after the first layer of decomposition; In general, M satisfies L(1)≤M≤2L(1); the value of α is 3.

4. The method for extracting characteristics of petrochemical pipeline corrosion vibration signals according to claim 1 is characterized in that: After denoising, the characteristic signal is obtained after modal decomposition. The specific method is as follows: S1. Obtaining vibration data of the petrochemical pipeline under working excitation according to the vibration signal collected by the sensor; S2. Calculate the power spectrum and the power spectrum matrix according to the obtained vibration data; S3, performing singular value decomposition on the obtained power spectrum matrix to obtain an observable matrix and a controllable inversion matrix; S4, calculating the structural state matrix through the first two order observable matrices; S5, performing eigendecomposition on the obtained structural state matrix, and calculating modal parameters according to the obtained eigenvalues ​​and eigenvectors; S6. Perform fuzzy C-means clustering analysis on the obtained modal frequencies to obtain the cluster center of each modal cluster; S7. Use the mean phase shift MPD to determine the order of the mode and obtain the final modal parameter identification result.

Images