Submarine pipeline health assessment method and system based on multi-state quantity data fusion

Abstract

The invention discloses a technical scheme of a submarine pipeline health assessment method and system based on multi-state quantity data fusion, and the method comprises the steps: improving the defect classification precision while reducing the dimensionality through dynamic balance global variance retention and local discrimination enhancement; and a solution with high efficiency and robustness is provided for feature fusion and accurate evaluation of the multi-source high-dimensional pipeline data. The submarine pipeline health assessment method comprises the following steps: collecting monitoring data of a sensor in a submarine pipeline, the monitoring data including but not limited to stress strain, acceleration, vibration and temperature in the submarine pipeline; carrying out fusion and dimension reduction processing on the obtained multi-source heterogeneous data, and finding out a mutual relation of changes among parameters of the submarine pipeline in the same time period; calculating a pipeline damage condition coefficient and a repair index according to an analysis result of the multi-source heterogeneous data; and dividing repair grades according to the repair indexes, comprehensively judging the health state of the submarine pipeline, and giving a repair suggestion.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of subsea pipeline health assessment, and particularly relates to a subsea pipeline health assessment method and system based on multi-state quantity data fusion. BACKGROUND

[0002] The subsea pipeline is long-term served in the complex marine environment, is affected by seawater corrosion, water pressure fluctuation, ocean current impact, geological disasters and the like, and is prone to cause multiple defects such as cracking, deformation, misplacement, leakage and corrosion. The cumulative effect and interaction of these defects seriously threaten the marine ecological safety. For example, pipeline cracking or leakage can cause oil and gas leakage and cause large-area marine ecological pollution. However, frequent maintenance of the subsea pipeline faces real problems such as high cost and long cycle, and therefore, in order to balance the contradiction between “high maintenance cost” and “high ecological risk”, a method system for subsea pipeline health assessment needs to be constructed to quantize the pipeline state and repair priority and solve the dilemma of “over-maintenance” and “passive repair”.

[0003] The monitoring data generated by the subsea pipeline running in the complex marine environment has typical high-dimensional and multi-source characteristics, and a traditional single dimension reduction method cannot effectively extract key features. Although principal component analysis can separate data redundancy and compress dimensions to the principal component space through covariance matrix decomposition, the unsupervised characteristics lead to the loss of discriminative information of the defect category. Although the maximum margin criterion optimizes the projection direction based on the inter-class scatter and the intra-class scatter to enhance the separability, but when it directly acts on the original high-dimensional data, the calculation complexity is high and is easily disturbed by noise. Therefore, a new clustering method is needed to reduce and fuse the monitoring data of the subsea pipeline. SUMMARY

[0004] In order to make up for the deficiencies of the prior art, the application aims to provide a technical scheme of a subsea pipeline health assessment method and system based on multi-state quantity data fusion, which balances the global variance reservation and local discrimination enhancement dynamically, reduces the dimension while improving the defect classification accuracy, and provides a solution with high efficiency and robustness for feature clustering and accurate evaluation of multi-source high-dimensional pipeline data.

[0005] The first object of the application is to provide a subsea pipeline health assessment method based on multi-state quantity data fusion, comprising the following steps:

[0006] S1. Collecting monitoring data of sensors in the subsea pipeline, the monitoring data including but not limited to stress and strain, acceleration, vibration and temperature in the subsea pipeline;

[0007] S2. Clustering and dimensionally reducing the obtained multi-source heterogeneous data to find the mutual relationship of the changes of the subsea pipeline parameters in the same time period;

[0008] S3. Calculate the pipeline damage condition coefficient and repair index according to the analysis result of the multi-source heterogeneous data;

[0009] S4. Divide the repair level according to the repair index, comprehensively judge the health state of the submarine pipeline, and give repair suggestions.

[0010] Further, the specific operation of the step S2 is as follows:

[0011] S21. Each sample in the multi-source heterogeneous data is represented by a vector, and the samples are combined to form a sample matrix, and the sample matrix is centrally processed to obtain a centralized sample matrix;

[0012] S22. The covariance of the centralized sample matrix is solved to obtain a covariance matrix C;

[0013] S23. The sample mean vector of the multi-source heterogeneous data sample is solved, and the between-class scatter S b and the within-class scatter S w are solved by using the maximum margin criterion algorithm, and the formulas of the between-class scatter S b and the within-class scatter S w are as follows:

[0014]

[0015] Wherein, c is the number of classes, n i is the number of samples in the i-th class, μ i is the mean vector of the i-th sample, represents the j-th sample of the i-th class, and μ is the total sample mean vector;

[0016] S24. The sample covariance matrix C, the between-class scatter S b and the within-class scatter S w are substituted into the following objective function formula to obtain the projection matrix W * :

[0017]

[0018] Wherein, the constraint term s.t. W T W=I stipulates that W is an orthogonal projection matrix; W T is the transpose of the matrix W; I is the unit matrix; tr(W T (S b -S w )W) is the trace of the discriminant metric matrix after projection, which is equal to the sum of the difference between the between-class distance and the within-class distance in the projection space; tr(W TCW) is the trace of the projected data covariance matrix, which equals to the sum of eigenvalues in all projection directions; a is a trade-off parameter to adjust the weight between PCA and MMC, when a = 1, it is the standard PCA algorithm, and when a = 0, it is the MMC algorithm.

[0019] PCA is an unsupervised dimensionality reduction method, and the core goal is to maximize the variance of the projected data to capture the main variation direction and structure of the overall data. MMC is a supervised discriminant method, which uses data labels, and the core goal is to maximize the difference between "inter-class dispersion" and "intra-class dispersion", aiming to make data of different classes more separated (large inter-class distance) and data within the same class more aggregated (small inter-class distance) in the projection space.

[0020] S25. Solve the objective function and convert it into a generalized eigenvalue problem to obtain the following formula:

[0021] (C + aS)W = AW,

[0022] wherein, AW is a diagonal matrix containing eigenvalues; in the m*m diagonal matrix, m is the dimension of the projected subspace.

[0023] S26. Substitute the projection matrix W * into the following formula to obtain the reduced dimensionality feature matrix Z:

[0024] Z = XW * ,

[0025] wherein, X is a matrix containing the original data as the input of the dimensionality reduction algorithm; the first k principal components of the feature matrix are taken according to the number of features retained in the feature matrix Z to achieve the purpose of dimensionality reduction.

[0026] Further, the pipeline health condition and repair evaluation index system comprises a target layer, a condition layer and an index layer, the target layer is a generalization of the pipeline health evaluation index system, representing the overall comprehensive level of the pipeline health condition; the condition layer can represent the detection condition of the system, including five detection conditions of rupture, deformation, misplacement, leakage and corrosion; the index layer is a different element expressing each classification index, reflecting the pipeline health condition.

[0027] Further, the specific operation of the step S3 is as follows:

[0028] S31. Introduce expert opinions through the analytic hierarchy process to screen and study each influence factor of the pipeline, and construct an evaluation index system;

[0029] S32. Comb the influence factors of each index on the health of the pipeline system, conduct single factor analysis based on mathematical statistics, decompose the comprehensive evaluation results of multi-factor influence into single factor influence and conduct correlation analysis, obtain the health degree score standard corresponding to the pipeline health condition and repair evaluation index system;

[0030] S33. According to the established evaluation system, establish and train the pipeline health condition and repair evaluation model;

[0031] S34. Optimize the model parameters by the control variable method;

[0032] S35. Substitute the constructed model into the sample data for training and self-validation;

[0033] S36. Divide the evaluation health degree and evaluation degree level, and adopt the fuzzy comprehensive evaluation method to make comprehensive evaluation on the input index in the model.

[0034] Further, in step 3, the damage condition coefficient S is calculated according to the multi-source detection results, then the defect parameter is calculated according to the following formula (1) or (2), and then the pipeline repair index RI is calculated by combining the region importance parameter, the pipeline importance parameter and the soil quality influence parameter, which is used to analyze the pipeline health condition; let the defect parameter F, the calculation formula is as follows:

[0035] When S<40, F=0.25S (1)

[0036] When S≥40, F=10 (2)

[0037]

[0038] RI=0.7F+0.1K+0.05E+0.15T (4)

[0039] In the formula, L is the total length of the pipeline to be evaluated, L i is the longitudinal length of the i-th defect, P i is the weight of the i-th defect, n is the number of defects, K is the region importance parameter, E is the pipeline importance parameter, and T is the soil quality influence parameter.

[0040] Further, in step 4, the repair level is divided according to the repair index, and the repair suggestion is given, which specifically includes the following levels:

[0041] Level 1: Repair index RI<3, indicating that the pipeline structure condition is basically perfect and does not need to be repaired;

[0042] Level 2: 3≤RI<5, indicating that the pipeline structure will not be damaged in the short term, but repair plan should be made and the corresponding repair method should be adopted;

[0043] Grade 3: 5≤RI<7, indicating that the pipeline structure may be damaged in the short term, and the corresponding repair method should be repaired as soon as possible;

[0044] Grade 4: RI≥7, indicating that the pipeline structure has been damaged or will be damaged soon, and the corresponding repair method should be repaired immediately.

[0045] Another object of the present application is to provide a submarine pipeline health assessment system based on multi-state quantity data fusion, for executing a submarine pipeline health assessment method based on multi-state quantity data fusion, comprising a data acquisition module, a data fusion module, an evaluation index system construction module, wherein,

[0046] The data acquisition module comprises a submarine pipeline sensing monitoring network formed by a combination of a fiber grating strain sensor, a fiber grating temperature sensor, a fiber grating vibration sensor and a MEMS accelerometer, and the measurement data of each measurement node monitored by the sensor monitoring network is organized into a multi-dimensional vector;

[0047] The data fusion module uses a PCA-MMC fusion method based on principal component analysis PCA and maximum margin criterion MMC, to fuse and process long-term different types of data obtained by the sensor monitoring network in the data acquisition module, and to find out the mutual relationship between the vibration, displacement deformation, temperature and stress-strain changes of the submarine pipeline under the same time period and the same environmental conditions;

[0048] The evaluation index system construction module is used to analyze the correlation of the data changes obtained in the data fusion module, to construct a submarine pipeline health state and repair evaluation model, to calculate a pipeline damage condition coefficient and a repair index according to the analysis results of the multi-source data, to divide the repair grade according to the repair index, to comprehensively judge the health state of the submarine pipeline, and to give repair suggestions.

[0049] Compared with the prior art, the present application has the following advantages:

[0050] This invention effectively addresses the problems of traditional methods, such as limited data dimensions, insufficient defect coupling analysis, and delayed maintenance decisions, by constructing a multi-state data fusion-based subsea pipeline health assessment system. The data acquisition module integrates fiber optic strain, temperature, and vibration sensors, as well as MEMS accelerometers, forming a multi-dimensional vector monitoring network. This network can simultaneously acquire multiple physical quantities such as stress-strain, vibration spectrum, temperature distribution, and deformation, overcoming the limitation of a single sensor reflecting only local static parameters. The data fusion module enhances the ability to identify multi-defect coupling risks by correlating and analyzing the relationships between vibration, displacement, temperature, and strain in long-term monitoring data. The assessment index system construction module establishes a health status and repair assessment model based on the correlation of multi-source data. Through the quantitative calculation of pipeline damage coefficients and repair indices, it prioritizes repairs and generates targeted recommendations. This avoids resource waste caused by over-maintenance and enables timely response to high-risk defects. It significantly reduces the risk of ecological damage due to missed detections or delayed repairs while lowering operation and maintenance costs. This provides reliable technical support for balancing subsea pipeline safety and economic benefits and offers a standardized solution for the intelligent operation and maintenance of deep-sea energy infrastructure. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the experimental system composition of the present invention;

[0052] Figure 2 This is a flowchart illustrating the subsea pipeline health assessment method of the present invention;

[0053] Figure 3 This is a flowchart of the multi-source heterogeneous data fusion processing of the present invention;

[0054] Figure 4 This is a technical roadmap for the subsea pipeline health status and repair assessment model of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0056] Example 1

[0057] In this embodiment, the fiber optic grating monitoring system for submarine pipelines consists of the following components: Figure 1As shown, a total of 10 temperature sensors are arranged every 1 meter, and every 5 sensors are individually accessed to the demodulator; a total of two strings of 13 sensors are arranged every 1 meter, and the monitoring distance is 13 meters; the FBG strain, temperature and vibration sensors are connected to the fiber grating demodulator through the FC interface for wavelength data demodulation, the demodulator is connected to the industrial computer, and the host computer software on the industrial computer can obtain the wavelength data of each sensor in real time, and save the daily data, and finally connect the 4G wireless transmission module to realize remote transmission of data. The multi-source heterogeneous data during September 3, 2022 to September 6, 2022 is collected by using the system, the data acquisition frequency of the demodulator is 100Hz, and the data saving frequency is 1Hz.

[0058] As shown in Figure 2 and Figure 3 , a submarine pipeline health assessment method based on multi-state quantity data fusion includes the following steps:

[0059] (1) Collecting monitoring data of sensors in the submarine pipeline, the monitoring data including but not limited to stress and strain, acceleration, vibration, and temperature in the submarine pipeline.

[0060] Collecting real-time data monitored by the submarine pipeline sensing monitoring network as input signals of the system.

[0061] (2) Aggregating and dimensionally reducing the obtained multi-source heterogeneous data to find the mutual relationship between the changes of submarine pipeline parameters in the same time period.

[0062] The multi-state quantity data fusion processing technology used in the present application is based on the aggregation method combining principal component analysis and maximum margin criterion. The aggregation method combining principal component analysis and maximum margin criterion has the following specific steps:

[0063] ①Each sample in the multi-source heterogeneous data is represented by a vector, and the samples are combined to form a sample matrix. The sample matrix is centrally processed to obtain a centralized sample matrix.

[0064] ②The covariance of the centralized sample matrix is calculated to obtain the covariance matrix C.

[0065] ③The sample mean vector of the multi-source heterogeneous data sample is calculated, and the maximum margin criterion algorithm is used to obtain the between-class scatter S b and the within-class scatter S w , the formula of the between-class scatter S b and the within-class scatter S w is as follows:

[0066]

[0067]

[0068] where c is the number of classes, n i is the number of samples in the i-th class, μ i is the mean vector of the i-th class, represents the j-th sample in the i-th class, and μ is the mean vector of all samples.

[0069] (4) Substitute the sample covariance matrix C, the between-class scatter S b and the within-class scatter S w into the following objective function to obtain the projection matrix W * :

[0070]

[0071] where the constraint s.t. W T W = I specifies that W is an orthogonal projection matrix; W T is the transpose of the matrix W; I is the identity matrix; tr(W T (S b -S w )W) is the trace of the discriminant metric matrix after projection, which is equal to the sum of the differences between the between-class distances and the within-class distances in the projection space; tr(W T CW) is the trace of the data covariance matrix after projection, which is equal to the sum of the eigenvalues in all projection directions; and a is a weighting parameter used to adjust the weight between PCA and MMC, with a = 1 being the standard PCA algorithm and a = 0 being the MMC algorithm.

[0072] PCA is an unsupervised dimensionality reduction method, and the core objective is to maximize the variance of the projected data to capture the main variation direction and structure of the overall data. MMC is a supervised discriminant method that uses data labels, and the core objective is to maximize the difference between the between-class scatter and the within-class scatter, aiming to make the data of different classes more separated (large between-class distance) and the data within the same class more clustered (small between-class distance) in the projection space.

[0073] (5) Solve the objective function by converting it into a generalized eigenvalue problem to obtain the following formula:

[0074] (C + λS)W = ΛW,

[0075] where Λ is a diagonal matrix containing eigenvalues; and m is the dimension of the subspace after projection.

[0076] (6) Substitute the projection matrix W * into the following formula to obtain the feature matrix Z after dimensionality reduction:

[0077] Z = XW * ,

[0078] Wherein, X is a matrix containing original data as the input of the dimension reduction algorithm; the first k principal components of the feature matrix are taken according to the number of features retained according to the feature matrix Z, so as to achieve the purpose of dimension reduction.

[0079] (3) Calculate the pipeline damage condition coefficient and the repair index according to the analysis result of the multi-source heterogeneous data.

[0080] The pipeline health condition and repair evaluation index system includes a target layer, a condition layer and an index layer, as shown in Table 1. The target layer is a generalization of the pipeline health evaluation index system, representing the overall comprehensive level of the pipeline health condition; the condition layer can represent the detection condition of the system, including five detection conditions of rupture, deformation, misplacement, leakage and corrosion; the index layer is a different element expressing each classification index, reflecting the pipeline health condition.

[0081] Table 1: Submarine pipeline health condition evaluation index system

[0082]

[0083] In the table, L is the total length of the pipeline to be evaluated, L i is the longitudinal length of the i-th defect, P i is the weight of the i-th defect, n is the number of defects, K is the regional importance parameter, E is the pipeline importance parameter, and T is the soil influence parameter.

[0084] According to the damage condition coefficient S calculated from the deformation, stress and strain, bending angle and other multi-source detection results, the defect parameters are calculated according to the formula (1) or (2) in the table, and then the pipeline repair index is calculated by combining the regional importance parameter, the pipeline importance parameter and the soil influence parameter, which is used to analyze the pipeline health condition.

[0085] As Figure 4 shown, the specific process of this step is as follows:

[0086] ① Build the evaluation index system, introduce expert opinions through the analytic hierarchy process, and screen and study each influencing factor of the pipeline to build the evaluation index system.

[0087] Through the analytic hierarchy process, expert opinions are introduced, a cross-disciplinary expert team is formed to carry out three rounds of Delphi consultation, each influencing factor of the pipeline is screened and studied, and a five-level index structure is optimized: the target layer focuses on the comprehensive evaluation of pipeline health, the condition layer is divided into five failure modes of rupture, deformation, misplacement, leakage and corrosion, and the index layer quantifies the defect parameter F and the repair index RI.

[0088] ②Sort out the influence factors of each index on the health of the pipeline system, conduct single-factor analysis based on mathematical statistics, decompose the comprehensive evaluation results of multi-factor influence into single-factor influence and conduct correlation analysis, and obtain the health degree scoring standard corresponding to the pipeline health condition and repair evaluation index system.

[0089] ③According to the established evaluation system, establish and train the pipeline health condition and repair evaluation model.

[0090] Based on the established evaluation index system (including target layer, condition layer, and index layer), a weighted linear regression model is used, with defect parameters F and repair index RI as dependent variables, and quantitative indexes of five failure modes such as rupture, deformation, misplacement, leakage, and corrosion as independent variables. The model parameters are fitted by least squares method to obtain the contribution weight of each index to the pipeline health condition, thereby establishing a comprehensive evaluation model for calculating the health degree score and repair index RI of the pipeline.

[0091] ④Optimize the selection of model parameters by controlling variable method.

[0092] Specifically, fix part of the variables, adjust the other variables one by one, observe the changes of the model output, and select the parameter combination that makes the model prediction accuracy and stability optimal.

[0093] ⑤The constructed model is substituted into sample data for training and self-validation.

[0094] Specifically, the collected multi-source heterogeneous data is divided into training set and validation set. The training set is used to fit the model parameters by least squares method to obtain the optimal weight, and then the validation set is used for self-validation. The mean square error (MSE) and determination coefficient (R 2 ) of the model on the validation set are calculated to ensure the accuracy and reliability of the model.

[0095] ⑥Divide the evaluation health degree and evaluation degree into grades, and use fuzzy comprehensive evaluation method to make comprehensive evaluation on the input indexes in the model.

[0096] (4) According to the repair index, divide the repair level, comprehensively judge the health state of the submarine pipeline, and give repair suggestions.

[0097] According to the repair index, divide the repair level, give repair suggestions, which include the following levels:

[0098] Level 1: Repair index RI < 3, indicating that the pipeline structure condition is basically sound and does not need repair;

[0099] Level 2: 3≤RI<5, indicating that the pipeline structure will not fail in the short term, but repair plan should be made and the corresponding repair method should be adopted;

[0100] Grade 3: 5≤RI<7, indicating that the pipeline structure may be damaged in the short term, and the corresponding repair method should be repaired as soon as possible;

[0101] Grade 4: RI≥7, indicating that the pipeline structure has been damaged or will soon be damaged, and the corresponding repair method should be repaired immediately.

[0102] Embodiment 2

[0103] The embodiment provides a submarine pipeline health assessment system based on multi-state quantity data fusion, which is used to execute the submarine pipeline health assessment method based on multi-state quantity data fusion in embodiment 1. The submarine pipeline health assessment system comprises a data acquisition module, a data fusion module and an evaluation index system construction module.

[0104] Specifically, the data acquisition module comprises a submarine pipeline sensing and monitoring network formed by a combination of a fiber grating strain sensor, a fiber grating temperature sensor, a fiber grating vibration sensor and a MEMS accelerometer. The measurement data of each measurement node monitored by the sensing and monitoring network is organized into a multi-dimensional vector. The data fusion module uses a fusion method PCA-MMC based on principal component analysis PCA and maximum margin criterion MMC, which is used to fuse and process long-term different types of data obtained by the sensing and monitoring network in the data acquisition module, and find the mutual relationship between the vibration, displacement deformation, temperature and stress-strain change of the submarine pipeline under the same time period and the same environmental condition. The evaluation index system construction module is used to analyze the correlation of the mutual relationship of the data change obtained in the data fusion module, construct a submarine pipeline health state and repair evaluation model, calculate a pipeline damage condition coefficient and a repair index according to the analysis result of the multi-source data, divide a repair grade according to the repair index, comprehensively judge the health state of the submarine pipeline, and give a repair suggestion.

[0105] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for health assessment of subsea pipelines based on multi-state quantity data fusion, characterized in that, Comprise the following steps: S1. Collect monitoring data of sensors in the submarine pipeline, the monitoring data including but not limited to stress and strain, acceleration, vibration, temperature in the submarine pipeline; S2. The obtained multi-source heterogeneous data is fused and dimensionally reduced, and the mutual relationship between the changes of the submarine pipeline parameters in the same time period is found out; S3. According to the analysis result of the multi-source heterogeneous data, the pipeline damage condition coefficient and the repair index are calculated; S4. According to the repair index, the repair grade is divided, the health status of the submarine pipeline is comprehensively judged, and the repair suggestion is given.

2. The method for health assessment of subsea pipeline based on multi-state data fusion according to claim 1, characterized in that, The specific operation of the step S2 is as follows: S21. Each sample in the multi-source heterogeneous data is represented by a vector, and the samples are combined to form a sample matrix. The sample matrix is centrally processed to obtain a centralized sample matrix; S22. The covariance of the centralized sample matrix is solved to obtain the covariance matrix C; S23. Calculate the sample mean vector of the multi-source heterogeneous data samples, and calculate the between-class scatter S using the maximum margin criterion algorithm b and the within-class scatter S w The formulas of the between-class scatter S b and the within-class scatter S w are as follows, respectively: where c is the number of classes, n i is the number of samples in the i-th class, μ i is the mean vector of the i-th class, denotes the j-th sample of the i-th class, and μ is the mean vector of the total samples. S24. The sample covariance matrix C, the between-class scatter S b and the within-class scatter S w are substituted into the following objective function formula to obtain the projection matrix W * : where the constraint term s.t. W T W = I specifies that W is an orthogonal projection matrix; W T is the transpose of matrix W; I is the identity matrix; tr(W T (S b -S w )W) is the trace of the discriminant metric matrix after projection, which is equal to the sum of the difference between the inter-class distance and the intra-class distance in the projection space; tr(W T CW) is the trace of the data covariance matrix after projection, which is equal to the sum of the eigenvalues in all projection directions; and a is a trade-off parameter to adjust the weight between PCA and MMC, when a = 1, it is the standard PCA algorithm, and when a = 0, it is the MMC algorithm. S25. The objective function is solved, which is converted into a generalized eigenvalue problem, and the following formula is obtained: (C+λS)W=ΛW, wherein Λ is a diagonal matrix containing eigenvalues; S26. Substitute the projection matrix W into the following equation to obtain the reduced dimension feature matrix Z: * Z = W X Z = XW * , Wherein, X is a matrix containing original data, which is used as the input of the dimension reduction algorithm; According to the number of characteristics reserved by the characteristic matrix Z, the first k principal components of the characteristic matrix are taken out to realize dimension reduction.

3. The method for health assessment of subsea pipeline based on multi-state data fusion according to claim 1, characterized in that, The pipeline health condition and repair evaluation index system comprises a target layer, a condition layer and an index layer. The target layer is a generalization of the pipeline health evaluation index system, representing the overall comprehensive level of the pipeline health condition. The condition layer can represent the detection condition of the system, including five detection conditions of rupture, deformation, misplacement, leakage and corrosion. The index layer is a different element for expressing each classification index, reflecting the pipeline health condition.

4. The method for health assessment of subsea pipeline based on multi-state data fusion according to claim 3, characterized in that, The specific operation of the step S3 is as follows: S31. The expert opinions are introduced by the analytic hierarchy process, the influencing factors of the pipeline are researched, and the evaluation index system is constructed; S32. The influencing factors of each index on the health of the pipeline system are sorted out, the single factor analysis based on mathematical statistics is carried out, the comprehensive evaluation result of the multiple factor influence is decomposed into single factor influence and correlation analysis is carried out, and the health degree scoring standard corresponding to the pipeline health condition and repair evaluation index system is obtained; S33. According to the established evaluation system, the pipeline health condition and repair evaluation model is established and trained; S34. The model parameters are selected by the control variable method; S35. The constructed model is substituted into the sample data for training and self verification; S36. The health degree and the evaluation degree grade are divided, and the fuzzy comprehensive evaluation method is adopted to make comprehensive evaluation on the input index in the model.

5. The method for health assessment of subsea pipelines based on multi-state data fusion according to claim 4, characterized in that, In the step 3, the damage condition coefficient S is calculated according to the multi-source detection result, then the defect parameter is calculated according to the following formula (1) or (2), and then the pipeline repair index RI is calculated by combining the region importance parameter, the pipeline importance parameter and the soil quality influence parameter, which is used for analyzing the pipeline health condition; Let the defect parameter F be calculated as follows: When S<40, F=0.25S (1) When S≥40, F=10 (2) RI=0.7F+0.1K+0.05E+0.15T (4) where L is the total length of the pipeline being evaluated, L i is the longitudinal length of the ith defect, P i is the weight of the ith defect, n is the number of defects, K is a regional importance parameter, E is a pipeline importance parameter, and T is a soil impact parameter.

6. The method for health assessment of subsea pipelines based on multi-state data fusion of claim 4, wherein, In the step 4, the repair index is used to divide the repair grade and give the repair suggestion, which specifically includes the following grades: Grade 1: repair index RI < 3, indicating that the pipeline structure condition is basically sound, no repair; Grade 2: 3≤RI < 5, indicating that the pipeline structure will not be damaged in the short term, but repair plan should be made, and the corresponding repair method; Grade 3: 5≤RI < 7, indicating that the pipeline structure may be damaged in the short term, and should be repaired as soon as possible, and the corresponding repair method; Grade 4: RI≥7, indicating that the pipeline structure has been damaged or will be damaged soon, and should be repaired immediately, and the corresponding repair method.

7. A subsea pipeline health assessment system based on multi-state variable data fusion, for performing a method of subsea pipeline health assessment based on multi-state variable data fusion according to any one of claims 1-6, characterized in that, The data acquisition module, the data fusion module and the evaluation index system construction module are included, wherein, The data acquisition module includes a submarine pipeline sensing monitoring network formed by a combination of fiber grating strain sensors, fiber grating temperature sensors, fiber grating vibration sensors and MEMS accelerometers, and the measurement data of each measurement node monitored by the sensor monitoring network is organized into a multi-dimensional vector; The data fusion module uses a fusion method PCA-MMC based on principal component analysis PCA and maximum margin criterion MMC, to fuse and process long-term different types of data obtained by the sensor monitoring network in the data acquisition module, and find out the mutual relationship between the vibration, displacement deformation, temperature and stress strain change of the submarine pipeline under the same time period and the same environmental conditions; The evaluation index system construction module is used to analyze the correlation of the data change relationship obtained in the data fusion module, construct a submarine pipeline health state and repair evaluation model, calculate the pipeline damage condition coefficient and repair index according to the analysis results of the multi-source data, divide the repair grade according to the repair index, comprehensively judge the health state of the submarine pipeline, and give repair suggestions.

Images