Pipeline leakage identification method based on combination of OVMD and GBDT algorithm

Abstract

The invention relates to the field of offshore platform pipeline leakage detection, and provides a pipeline leakage identification method combining OVMD with a GBDT algorithm. The method comprises the following steps: firstly, establishing a pipeline leakage model, collecting leakage signals under different working conditions, collecting real background noise on an offshore platform, and injecting the real background noise into laboratory leakage signals to simulate an actual working environment; and decomposing the constructed original sample data set by using an OVMD method to obtain a plurality of signal component functions, extracting feature indexes, and classifying and identifying the extracted feature indexes by using a GBDT algorithm to form a confusion matrix and a precision-recall rate curve. According to the method, the identification accuracy of the leakage signal is effectively improved, and the problems that the acoustic emission signal acquisition difficulty is large, and an existing method is low in identification accuracy, poor in robustness and complex to implement are solved.

Description

Technical Field

[0001] The present invention relates to the field of offshore platform pipeline leakage detection, and in particular to a pipeline leakage identification method combining OVMD with GBDT algorithm. Background Art

[0002] As offshore oil and gas platform pipeline systems age, their safety issues become increasingly prominent. In complex marine environments, both natural gas pipelines and crude oil processing facilities are subject to vibration from various sources. The accumulated effects of this vibration over time can lead to pipeline fatigue and even damage. If a leak is not detected and prevented promptly, it poses a significant safety threat to the offshore platform and could even cause catastrophic consequences such as explosions, resulting in immeasurable economic losses and social impact. Therefore, leak prevention in offshore oil and gas pipelines has become a key industry concern. However, traditional theoretical analysis methods are difficult to implement in many practical situations and cannot fully guarantee the safety and reliability of the equipment. To address this, the industry has gradually introduced and widely used acoustic emission technology for real-time monitoring and diagnosis. Acoustic emission testing is a dynamic, real-time, and non-destructive assessment method. By capturing the acoustic wave signals spontaneously generated by the material's internal structure in response to stress or environmental changes, it accurately detects the overall structural health and provides early warning of potential damage. This technology, which requires no external excitation sources, offers high defect detection capabilities, comprehensive assessment performance, and broad applicability.

[0003] To identify pipeline leaks using acoustic emission technology, the first step is to acquire pipeline leakage signals. However, due to practical reasons, acquiring high-quality pipeline leakage signals is difficult. Furthermore, offshore platform pipelines operate in an extremely complex environment and are subject to strong noise interference, such as mechanical vibration noise generated by high pipeline vibration intensity, and wave and wind noise in the marine environment. These interferences further complicate the acquisition of high-quality acoustic signals.

[0004] Machine learning and deep learning are two approaches to identifying collected acoustic emission signals. However, traditional methods suffer from low recognition accuracy and robustness under strong noise interference. Furthermore, traditional machine learning methods require expanding the signal's information dimension to achieve high accuracy, which complicates the processing.

[0005] In view of this, it is necessary to provide a new technical solution to solve the above problems. To this end, the present invention improves the traditional variational mode decomposition (VMD) and proposes a pipeline leakage identification method that combines optimal variational mode decomposition (OVMD) with the gradient boosting decision tree (GBDT) algorithm. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to propose a method of injecting actual background noise into the laboratory leakage signal to solve the problem of difficulty in obtaining high-quality acoustic emission signals, and to propose a new method to solve the problems of insufficient recognition accuracy, low robustness and complex implementation process of existing methods.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a pipeline leakage identification method (OVMD-GBDT) combining optimal variational mode decomposition (OVMD) with gradient boosting tree (GBDT) algorithm, comprising the following steps: S1. Build a pipeline leakage model, set up an acoustic emission sensor on the pipeline to be tested, and collect multiple sets of pipeline leakage signals; S2. Install the acoustic emission sensors on at least two pipelines on the offshore platform to collect actual background noise signals; S3, injecting the actual background noise signal into the pipeline leakage signal to construct an original sample data set; S4, decomposing the original sample data set constructed in step S3 using the OVMD method to obtain multiple signal component functions, the OVMD method includes: introducing the Pearson correlation coefficient to select the optimal modal number , using the Peacock optimization algorithm to adjust the penalty factor and fidelity coefficient Conduct optimization; S5, extracting characteristic indicators from the signal component function obtained in step S4 to form a characteristic vector set; S6. Use the GBDT algorithm to classify and identify the feature vector set extracted above to form a confusion matrix and a precision-recall curve.

[0008] Preferably, the pipeline leakage model in step S1 includes: a pipeline, a pressure device and the acoustic emission sensor, the acoustic emission sensor is connected to a preamplifier, the preamplifier is connected to an acoustic emission detection device, and the acoustic emission detection device is connected to an industrial computer.

[0009] Preferably, the pipeline leakage model in step S1 includes two areas: one is an experimental section, in which multiple control valves are evenly distributed for simulating and controlling leakage conditions; the other is a buffer section, which is connected to a pressure device through a flexible rubber pipe, and the pressure device adopts an electric pressure self-controlled pressure test pump.

[0010] Preferably, the acoustic emission sensor installed for collecting the actual background noise signal in step S2 is the same as that in step S1, and the sampling mode is a long-term timing sampling mode.

[0011] Preferably, in step S3, the pipeline leakage signal in step S1 is first evenly divided according to the length of the sampling point to obtain multiple leakage signal samples, and then the actual background noise signal collected in S2 is also divided into sampling point segments with the same number as the pipeline leakage signal, and noise signal samples with the same number as the leakage signal samples are constructed. Finally, by injecting the noise signal samples into the leakage signal samples, the original sample data set containing the leakage signal samples and the noise signal samples is formed.

[0012] Preferably, the Pearson correlation coefficient is introduced in step S4 to select the optimal modal number. , mode selection is performed according to the following model: ; in, is the Pearson correlation coefficient function, which is calculated as follows: ; is the original signal; is the first one obtained by VMD decomposition modal components; is the residual signal after decomposition; It is The correlation coefficient between the modal components and the original signal; is the correlation coefficient between the residual signal and the original signal; is the residual correlation coefficient attenuation factor. According to the model setting, the optimal modal number can be automatically selected. , the specific process is as follows: (1) Set the initial mode number , perform preliminary VMD decomposition on the signal to be analyzed; (2) Calculate the Pearson correlation coefficient between each modal component and the original signal for the current modal number , and further calculate the correlation coefficient between the residual signal and the original signal ; (3) If the correlation of the residual signal Greater than the smallest correlation coefficient among all modal components, that is, , indicating that there is still under-decomposition, then K needs to be increased by 1 and return to the second step to continue iteration; (4) If satisfied ,and When , the current mode number is a reasonable upper limit and can be saved as the optimal value.

[0013] (5) After the stop condition is met, the current The optimal number of modes is taken as the final estimate and the program is terminated.

[0014] Preferably, the Peacock optimization algorithm is introduced in step S4 to adjust the penalty factor and fidelity coefficient Optimization is performed, including: according to the optimal modal number obtained in step S4 Generate multiple candidate parameter combinations , and perform VMD decomposition on it as an individual, calculate the corresponding residual signal and obtain the evaluation index, select the position with the best fitness, and then generate a new position based on the peacock optimization strategy and perform residual calculation and index evaluation again.

[0015] Preferably, in the step S5, the characteristic indicators of the signal component function are extracted by extracting time domain features and frequency domain features from the characteristic indicators of the signal component function obtained in the step S4, and constructing a characteristic vector set of the original sample data set; wherein, the time domain features include: mean, standard deviation, minimum value, maximum value, root mean square value, skewness, kurtosis, peak value, margin, and pulse index; the frequency domain features include spectral mean, spectral standard deviation, spectral energy, spectral center, spectral flatness, spectral roll-off point, spectral skewness, frequency kurtosis, spectral drop, and spectral coefficient of variation.

[0016] Preferably, the GBDT method is used for classification in step S6, 70% of the feature vector set is used as a training set and 30% is used as a test set to form a confusion matrix and a precision-recall curve as the recognition result.

[0017] The pipeline leakage identification method (OVMD-GBDT) provided by the present invention combines the optimal variational mode decomposition (OVMD) with the gradient boosted tree (GBDT) algorithm. It has the following beneficial effects: 1. This invention addresses, to a certain extent, the difficulty of acquiring high-quality pipeline leakage signals. By constructing a pipeline leakage model in a controlled laboratory environment, the method collects pipeline leakage signals under various leakage conditions. The method then collects actual background noise signals from an offshore platform. This background noise signal is injected into the laboratory-collected leakage signals to simulate pipeline leakage signals in a real-world environment. This creates a dataset that closely resembles real-world operating conditions, addressing the difficulty of acquiring high-quality leakage signals.

[0018] 2. This paper proposes an improved VMD method, namely the OVMD method, which solves the problem of the traditional VMD method on the number of modalities. , penalty factor and fidelity coefficient A very sensitive problem. By introducing the Pearson correlation coefficient to analyze the modal number The peacock optimization algorithm is introduced to optimize the penalty factor. and fidelity coefficient Optimization is performed to improve the accuracy of signal decomposition and the effect of feature extraction, and to minimize the residual signal.

[0019] 3. The present invention solves the problem that after the traditional VMD method obtains multiple signal component functions, the recognition efficiency is reduced due to too many information sequences. The present invention improves the recognition efficiency by extracting characteristic indicators from the signal component functions. Specifically, the extracted characteristic indicators are divided into two parts: time domain and frequency domain. The time domain features include mean, standard deviation, maximum value, minimum value, root mean square value, skewness, kurtosis, peak value, margin, and pulse index. The frequency domain features include spectrum mean, spectrum standard deviation, spectrum energy, spectrum center, spectrum flatness, spectrum roll-off point, spectrum skewness, frequency kurtosis, spectrum drop, and spectrum variation coefficient. By selecting these characteristic values ​​in the signal, the information sequence is reduced, the recognition efficiency is improved, and the problem of low recognition efficiency of the traditional VMD method is solved.

[0020] 4. This invention addresses the low recognition accuracy of traditional machine learning methods in low signal-to-noise ratio conditions. This invention creates noise interference by injecting actual background noise into the pipeline leakage signal. It then uses an improved VMD method to decompose and extract features from the injected signal. Finally, it uses the GBDT method to classify and identify the fused signal, achieving a high recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the overall process of the present invention; Figure 2 Schematic diagram of the pipeline leakage model of the present invention; Figure 3 This is a spectrum diagram after the noise signal is injected into the present invention; Figure 4 This is the time-frequency diagram after the noise signal is injected into the present invention; Figure 5 Flowchart for calibrating the modal number K based on the correlation coefficient of the present invention; Figure 6This is a flowchart of optimizing the penalty factor α and the fidelity coefficient τ based on the Peacock optimization algorithm of the present invention; Figure 7 It is the confusion matrix diagram of the present invention; Figure 8 This is the precision-recall curve of the present invention.

[0022] The accompanying drawings are marked as follows: 1. pipeline; 2. control valve; 3. acoustic emission sensor; 4. preamplifier; 5. acoustic emission detection device; 6. industrial computer. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] Please refer to the attached Figure 1 , Figure 1 The figure is a schematic diagram of the overall process of a pipeline leakage identification method according to the present invention using an OVMD combined with a GBDT algorithm. The method provided by the present invention may include the following steps: S1. Build a pipeline leakage model. Install an acoustic emission sensor on the pipeline to collect multiple sets of pipeline leakage signals. The pipeline leakage model mainly includes pipeline 1, a pressure device, a pressure gauge, and a series of control valves 2. Pipeline 1 simulates the working pipeline of an offshore platform. The pressure device provides pressure to enable fluid flow within pipeline 1. The pressure gauge reflects the current pressure within the pipeline. Control valves 2 control the size of the leak, simulating different leakage conditions.

[0025] S2. Install the acoustic emission sensors on at least two pipelines on the offshore platform to collect actual background noise signals. The "Liwan 3-1" offshore platform was selected for actual investigation, and actual background noise signals were collected on the 19-meter platform and the 29-meter platform.

[0026] S3. Inject the actual background noise signal into the pipeline leakage signal to construct an original sample data set. Because both the actual background noise signal and the pipeline leakage signal are continuous signals, the noise percentage is determined by signal power. Then, an appropriate number of samples are selected for injection to obtain the original sample data set.

[0027] S4. Decompose the original sample data set constructed in step S3 using the OVMD method to obtain multiple signal component functions. The traditional variational mode decomposition (VMD) method is improved to obtain the OVMD method, the core of which is the number of modes based on the Pearson correlation coefficient. The Peacock Optimization Algorithm (POA) is introduced to adjust the penalty factor and fidelity coefficient A modal number estimation model based on the Pearson correlation coefficient is proposed, the goal of which is to minimize the effective modal number. to prevent The pseudo-modal phenomenon caused by excessively large values. and fidelity coefficient Optimization is performed to minimize the residual signal as much as possible, thereby improving the accuracy of subsequent analysis or prediction tasks.

[0028] S5. Extract characteristic indicators of the signal component functions obtained by the OVMD method in step S4 to form a feature vector set to improve recognition efficiency.

[0029] S6. Use the GBDT algorithm to classify and identify the extracted feature vector set, generating a confusion matrix and a precision-recall curve to reflect the accuracy and performance of the method. All data is input into the GBDT algorithm for classification and identification. The resulting confusion matrix reflects the model's data recognition performance, and the precision-recall curve reflects the model's recognition accuracy.

[0030] The overall concept of this method is to first obtain leakage signals under controlled laboratory conditions under various leakage conditions. The actual background noise signal of the pipeline is then collected and injected into the leakage signal to construct a dataset that simulates real-world operating conditions. The OVMD method, an improvement on the traditional variational mode decomposition (VMD) method, is then used to decompose the constructed dataset into signal component functions. Feature indices are selected to extract features from the signal component functions, which are then input into the gradient boosted tree algorithm (GBDT) for classification and recognition. The resulting confusion matrix and precision-recall curve are used to reflect recognition accuracy and method performance.

[0031] Please refer to the attached Figure 2 , Figure 2 It is a schematic diagram of the pipeline leakage model in the present invention, showing the overall structure of the pipeline leakage model.

[0032] The pipeline leakage model primarily consists of a pipeline 1, a pressure device, a pressure gauge, and a series of control valves 2. The selected pipeline model strictly adheres to the GB / T 8163-2018 standard and is made of seamless steel pipe, suitable for high-pressure fluid transportation. Pipeline 1 has a diameter of 105 mm, a wall thickness of 5 mm, and a total length of 6.2 m. The system is divided into two key areas: the experimental section, which is 5 meters long and contains multiple high-pressure welded needle control valves 2 evenly distributed throughout to simulate and control leakage conditions; and the buffer section, which is 1.2 meters long and connected to the pressure device via a flexible rubber pipe. The pressure device utilizes the high-performance Z4DSY electric pressure automatic test pump. The acoustic emission sensor 3 comprises two R15 sensors, mounted at either end of pipeline 1, 10 cm from the boundary of the experimental section. Industrial-grade white vaseline is used as the coupling medium between the pipeline surface and the sensor, which is then secured with tape.

[0033] The data acquisition device uses an acoustic emission detection device 5 of model PCI-2 produced by the American Physical Acoustics Corporation (PAC). The device has a wide and accurate sampling frequency range (1kHz to 3MHz) and has a built-in efficient data storage function. It can store acoustic emission waveform information on the hard disk at a speed of 10 million sampling points per second. The data acquisition stage is operated and controlled using the AEwin software that is compatible with the PCI-2 device. The acoustic emission sensor 3 used in the experiment is R15 type, and the center frequency is set to 150kHz. In order to further optimize the signal quality and enhance the contrast between the signal and the noise, this experiment is equipped with a 2 / 4 / 6-channel preamplifier 4, whose gain can be adjusted to 20 / 40 / 60dB respectively. The acquisition parameters are pre-set and optimized on the AEwin software platform of the industrial computer 6. Then a broken lead experiment was carried out to verify the coupling state of the sensor and the pipeline and the stability of the signal transmission, and to confirm whether the storage function of the data acquisition device is normal. The adjustable final gain is set to 40db and the sampling frequency is set to 1MHz. After debugging, according to Figure 1 The pipeline leakage acoustic emission experiment is carried out in the above manner to collect multiple groups of pipeline leakage signals.

[0034] In order to collect the actual background noise signal of the pipeline, we went to the "Liwan 3-1" offshore platform for a detailed field survey, and finally selected the pipelines of the 19-meter platform and the 29-meter platform as the collection objects. The 19-meter platform has typical mechanical vibration noise characteristics due to the high vibration intensity of the pipeline, and is prone to fatigue damage; while the 29-meter platform is in the cooling process after the compressor is pressurized. The pipeline at this location is under great pressure, especially the elbows due to stress concentration problems, which are prone to damage. Therefore, these two locations with typical noise characteristics and potential risks were selected as the collection points for on-site noise signals. The model and parameter settings of the installed acoustic emission sensor are consistent with those in the laboratory. Industrial white vaseline is also used as a coupling agent, and the sensor is fixed with tape; the sampling mode is a long-term timed sampling mode.

[0035] Please refer to Figure 3 and Figure 4 , which are respectively the spectrum diagram and time-frequency diagram after the actual background noise signal is injected into the pipeline leakage signal in the present invention.

[0036] Since the noise level in the laboratory is relatively low and cannot reflect the actual noise situation of the offshore platform, the measured field noise is injected into the leakage signal collected in the laboratory to verify the acoustic emission-based pipeline leak detection method. Since both are continuous signals, the signal power is used to determine the noise percentage, that is: ; in is the noise power, is the signal power, 、 are the RMS values ​​of noise and leakage signal respectively, is the number of samples. Therefore, the leakage signal of the fused noise It can be obtained by the following formula: ; in is the collected leakage signal; is the measured noise signal.

[0037] To generate noisy leakage signal samples, this study first evenly segmented the pipeline leakage signals collected in the laboratory into segments of 1024 sampling points, resulting in 900 leakage signal samples. The actual noise signals collected from the offshore platform were similarly divided into segments of 1024 sampling points, creating another 900 noise signal samples. Finally, by injecting the noise signal samples into the leakage signal samples, a raw sample dataset containing 1800 samples was formed.

[0038] Please refer to Figure 5 , Figure 5This is a flow chart of the modal quantity calibration of the traditional VMD method in the present invention.

[0039] The calibration of the mode number based on the correlation coefficient is performed as follows: During the VMD decomposition, the mode number The selection of has a significant impact on the final decomposition effect. If the value is too large, it will easily cause the signal to be over-decomposed and produce pseudo-modes; If the setting is too small, insufficient decomposition may occur. Therefore, to address this problem, an adaptive mode number determination method based on correlation coefficient is proposed. This method uses the Pearson correlation coefficient to determine the modal number. Weighing two signals and The linear correlation between them is calculated as follows: ; This paper proposes a modal number estimation model based on the Pearson correlation coefficient, the goal of which is to minimize the effective modal number. , mode selection is performed according to the following model: ; in, is the Pearson correlation coefficient function; is the original signal; is the first one obtained by VMD decomposition modal components; is the residual signal after decomposition; It is The correlation coefficient between the modal components and the original signal; is the correlation coefficient between the residual signal and the original signal; is the residual correlation coefficient attenuation factor. According to the model setting, the optimal modal number can be automatically selected. , the specific process is as follows: (1) Set the initial mode number , perform preliminary VMD decomposition on the signal to be analyzed; (2) Calculate the Pearson correlation coefficient between each modal component and the original signal for the current modal number , and further calculate the correlation coefficient between the residual signal and the original signal ; (3) If the correlation of the residual signal Greater than the smallest correlation coefficient among all modal components, that is, , indicating that there is still under-decomposition, then K needs to be increased by 1 and return to the second step to continue iteration; (4) If satisfied ,and When , the current mode number is a reasonable upper limit and can be saved as the optimal value.

[0040] (5) After the stop condition is met, the current The optimal number of modes is taken as the final estimate and the program is terminated.

[0041] Please refer to Figure 6 , Figure 6 The present invention uses the Peacock Optimization Algorithm (POA) to adjust the penalty factor and fidelity coefficient Flowchart for optimization.

[0042] The Peacock optimization algorithm is introduced to optimize the fidelity coefficient τ and penalty factor α in the following way: On the basis of what has been determined, two key parameters still need to be further determined in VMD decomposition: penalty factor and fidelity coefficient These two parameters have a great influence on the decomposition quality and are directly related to the modal components obtained in the end. and the residual signal To improve the accuracy of signal decomposition and feature extraction, it is necessary to and Optimize to minimize the residual signal, thereby improving the accuracy of subsequent analysis or prediction tasks. To this end, a parameter estimation method based on the Peacock Optimization Algorithm (POA) is introduced. This method aims to minimize the energy of the residual signal, and its optimization index is the VMD residual evaluation function. : ; in, is the length of the original signal; is the original input signal; is a parameter The decomposition obtained The Peacock Optimization Algorithm (POA) achieves optimized search by simulating the process of male peacocks displaying their tail feathers to attract female peacocks. The POA algorithm includes two key links: display behavior and update mechanism. It has the advantages of fast convergence speed and strong global search ability. Assume that in D-dimensional space, the population consists of The position of each peacock Represents a set of parameter combinations to be optimized (penalty factor and fidelity coefficient), each position corresponds to a fitness value (i.e. ). The peacock individual attracts the position of the dominant individual by displaying behavior, gradually updating the position and optimizing parameters. The specific update formula is as follows: ; Where: For the In the iteration The location of each peacock, is the position of the best individual in the current population, is a random number in [0,1], is a standard normal random number, is the attenuation factor. In the VMD algorithm, the parameter combination It will significantly affect the decomposition results. In order to achieve efficient parameter optimization, each group As an "individual" in the POA algorithm, and the residual error index it produces As a fitness function, the POA algorithm is used to perform a global search and update of the parameter space to obtain a better solution. In this embodiment, in order to achieve efficient parameter optimization, each group As an "individual" in the POA algorithm, and the residual error index it produces As a fitness function, the POA algorithm is used to perform a global search and update of the parameter space to obtain a better solution. The basic steps are: (1) Using the method of modal number calibration based on correlation coefficient in step S4, first determine the optimal modal number of the current signal. ; (2) According to Value generates multiple candidate parameter combinations , and perform VMD decomposition on it as an individual, calculate the corresponding residual signal and obtain the evaluation index ; (3) Set the position with the best fitness as ; (4) Generate new positions based on the Peacock optimization strategy , and perform residual calculation and indicator evaluation again; (5) If a better individual appears, update ; (6) If the stopping criteria are not met, return to step (4) and continue; if the termination criteria are met, take the current parameters As the optimal parameters of OVMD.

[0043] Please refer to Figure 7 and Figure 8 , Figure 7 and Figure 8 The confusion matrix and precision-recall curve after classification and recognition using the gradient boosted tree algorithm (GBDT) are shown.

[0044] Using the improved VMD method to decompose the signal can obtain multiple signal component functions. However, if these signal component functions are directly identified, the recognition efficiency will be reduced due to too many information sequences. In order to improve the recognition efficiency, this embodiment proposes a method for extracting feature indicators from the signal component function for identification. The extracted features include 10 time domain features and 10 frequency domain features. The 10 time domain features are: mean, standard deviation, minimum value, maximum value, root mean square value, skewness, kurtosis, peak value, margin, and pulse index. The 10 frequency domain features are spectral mean, spectral standard deviation, spectral energy, spectral center, spectral flatness, spectral roll-off point, spectral skewness, frequency kurtosis, spectral drop, and spectral coefficient of variation. The above 20 indicators are selected from the signal component function of the improved VMD for feature extraction, and 80 feature vectors are extracted for each sample to form a feature vector set of the original sample data set. The feature vector set is then input into the GBDT algorithm for classification and identification: The core idea of ​​GBDT is to optimize the model by gradually building decision trees, so that each new tree added at a time is based on reducing the error of the previous model. The step model is as follows: ; in, It is the previous round model, Indicates the decision trees; the final GBDT model can be expressed as a CART additive model: ; in, Indicates the A decision tree, Representative The parameters of the decision tree, Represents the number of decision trees. In order to optimize the general loss function, GBDT uses the gradient value of the loss function to build a model. superior, is the eigenvector, is the corresponding category. The algorithm flow of GBDT classification problem is as follows: (1) Initialize model parameters: ; (2) For , calculate the probability that each sample belongs to each category: ; (3) For , calculate the probability residual: ; And Fit a regression tree and get the mth round The leaf node area of ​​the regression tree of the class ,in .

[0045] (4) For calculate The value on: ; Finally, the calculated Wheel models of various types: ; The GBDT algorithm is used to train and classify the feature vectors extracted from the signal component function. After feature extraction, 1800 sets of feature vectors are classified using the GBDT algorithm. 70% of each category is used as a training set and 30% as a test set, that is, 1260 sets of training set samples and 540 sets of test set samples. Figure 7 It can be concluded that the method proposed in this embodiment correctly identified 268 noisy leakage signals and 269 measured noise signals, identified 2 measured noise signals as noisy leakage signals, and identified 1 noisy leakage signal as a measured noise signal, with an identification accuracy rate of 99.44%. Figure 8 It can be concluded that the precision and recall of the prediction results are both very close to 1. The above proves that the method described in this embodiment has a high recognition accuracy.

[0046] Through the method described in this embodiment, a dataset simulating leakage signals of real offshore platform pipelines was successfully constructed, and the problems of low recognition accuracy and insufficient robustness of traditional machine learning methods were solved, providing an important basis for the structural analysis and optimization of offshore platform pipelines.

[0047] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A pipeline leakage identification method combining OVMD and GBDT algorithm, characterized in that: The following steps are involved: S1. Build a pipeline leakage model, set up an acoustic emission sensor on the pipeline to be tested, and collect multiple sets of pipeline leakage signals; S2. Install the acoustic emission sensors on at least two pipelines on the offshore platform to collect actual background noise signals; S3, injecting the actual background noise signal into the pipeline leakage signal to construct an original sample data set; S4, decomposing the original sample data set constructed in step S3 using the OVMD method to obtain multiple signal component functions, the OVMD method includes: introducing the Pearson correlation coefficient to select the optimal modal number , using the Peacock optimization algorithm to adjust the penalty factor and fidelity coefficient Conduct optimization; S5, extracting characteristic indicators from the signal component function obtained in step S4 to form a characteristic vector set; S6. Use the GBDT algorithm to classify and identify the feature vector set extracted above to form a confusion matrix and a precision-recall curve.

2. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1 is characterized in that: The pipeline leakage model in step S1 includes: a pipeline, a pressure device and the acoustic emission sensor, the acoustic emission sensor is connected to a preamplifier, the preamplifier is connected to an acoustic emission detection device, and the acoustic emission detection device is connected to an industrial computer.

3. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 2 is characterized in that: The pipeline leakage model in step S1 includes two areas: one is the experimental section, in which multiple control valves are evenly distributed for simulating and controlling leakage conditions; the other is the buffer section, which is connected to the pressure device through a flexible rubber pipe, and the pressure device uses an electric pressure self-controlled pressure test pump.

4. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1 is characterized in that: The acoustic emission sensor installed for collecting the actual background noise signal in step S2 is the same as that in step S1, and the sampling mode is a long-term timing sampling mode.

5. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1 is characterized in that: In step S3, the pipeline leakage signal in step S1 is first evenly divided according to the length of the sampling points to obtain multiple leakage signal samples. Then, the actual background noise signal collected in step S2 is also divided into sampling point segments with the same number as the pipeline leakage signal, and noise signal samples with the same number as the leakage signal samples are constructed. Finally, the noise signal samples are injected into the leakage signal samples to form the original sample data set including the leakage signal samples and the noise signal samples.

6. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1 is characterized in that: The Pearson correlation coefficient is introduced in step S4 to select the optimal number of modes , mode selection is performed according to the following model: ； in, is the Pearson correlation coefficient function, which is calculated as follows: ; is the original signal; is the first one obtained by VMD decomposition modal components; is the residual signal after decomposition; It is The correlation coefficient between the modal components and the original signal; is the correlation coefficient between the residual signal and the original signal; is the residual correlation coefficient attenuation factor; according to the model setting, the optimal modal number can be automatically selected , the specific process is as follows: (1) Set the initial mode number , perform preliminary VMD decomposition on the signal to be analyzed; (2) Calculate the Pearson correlation coefficient between each modal component and the original signal for the current modal number , and further calculate the correlation coefficient between the residual signal and the original signal ; (3) If the correlation of the residual signal Greater than the smallest correlation coefficient among all modal components, that is, , indicating that there is still under-decomposition, then K needs to be increased by 1 and return to the second step to continue iteration; (4) If satisfied ,and When , the current mode number is a reasonable upper limit and can be saved as the optimal value; (5) After the stop condition is met, the current The optimal number of modes is taken as the final estimate and the program is terminated.

7. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1 or 6, characterized in that: The Peacock optimization algorithm is introduced in step S4 to adjust the penalty factor and fidelity coefficient Perform optimization, including: according to the optimal modal number obtained in step S4 Generate multiple candidate parameter combinations , and perform VMD decomposition on it as an individual, calculate the corresponding residual signal and obtain the evaluation index, select the position with the best fitness, and then generate a new position based on the peacock optimization strategy and perform residual calculation and index evaluation again.

8. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1 is characterized in that: In the step S5, the characteristic indicators of the signal component function are extracted by extracting the time domain characteristics and frequency domain characteristics of the characteristic indicators of the signal component function obtained in the step S4, and constructing the characteristic vector set of the original sample data set; wherein the time domain characteristics include: mean, standard deviation, minimum value, maximum value, root mean square value, skewness, kurtosis, peak value, margin, and pulse index; the frequency domain characteristics include spectrum mean, spectrum standard deviation, spectrum energy, spectrum center, spectrum flatness, spectrum roll-off point, spectrum skewness, frequency kurtosis, spectrum drop, and spectrum coefficient of variation.

9. The pipeline leakage identification method combining OVMD and GBDT algorithm according to claim 1, characterized in that: In the S6 step, the GBDT method is used for classification, 70% of the feature vector set is used as a training set, and 30% is used as a test set to form a confusion matrix and a precision-recall curve as the recognition result.

Images