Method and system for early warning of the state of a submarine pipeline based on scour-vibration coupling perception

Abstract

The application discloses a kind of based on scour-vibration coupling perception's submarine pipeline state early warning method and system, comprising: obtaining the historical marine environment information of target submarine pipeline area, according to this layout monitoring sensor array, multi-source monitoring data are collected and form multi-source monitoring data set after being preprocessed.The multi-physical field coupling analysis model is constructed, the data set is input into model analysis current submarine environment under the response condition of pipeline, and the pipeline response simulation information is obtained.The resonance influence caused by different sections due to environmental scour is analyzed, resonance derived features are generated, and pipeline multi-modal fusion features are constructed in combination with monitoring data.Analysis model of submarine pipeline state is used to predict the future response state of pipeline, and risk early warning prompt is carried out;At the same time, based on resonance derived features, the fatigue life of each section is analyzed, the fatigue-prone section is identified, and maintenance suggestions are generated and pushed.Further improve the accuracy and timeliness of submarine pipeline state perception, effectively improve the early warning intelligence of submarine pipeline.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ocean engineering, and in particular to a seabed pipeline state early warning method and system based on scour-vibration coupling sensing. BACKGROUND

[0002] Currently, the monitoring of the state of the seabed pipeline usually adopts a separate and uncoupled analysis method. It is usually composed of two independent modules of a vibration monitoring system and a seabed topography monitoring system. The vibration monitoring system is composed of an acceleration sensor, a data collector and signal processing software installed on the pipeline. Then the pipeline vibration signals (such as acceleration and frequency) are collected, and the vibration intensity is judged by spectrum analysis algorithm to determine whether it exceeds the standard, which is mainly used for early warning of fatigue damage caused by vortex-induced vibration (VIV). The seabed topography monitoring system realizes topography monitoring through a multi-beam sounding sonar and a profile instrument carried by a remotely operated vehicle (ROV) for periodic cruising. It periodically (such as once a year) acquires the topography data of the seabed around the pipeline, compares the historical data, and manually judges the development degree of the scour pit.

[0003] The prior art is two independent modules, and only simple manual data comparison is performed at the final management layer. The vibration monitoring system provides near real-time vibration data, but cannot explain whether the vibration change is caused by scour; the topography monitoring system can provide accurate topography change, but is seriously lagging and cannot be used for real-time early warning. Disadvantages and deficiencies: uncoupling: the biggest defect is that vibration and scour are regarded as two independent processes, and the strong coupling feedback mechanism (vibration aggravates scour, and scour changes the vibration characteristics) therebetween is ignored. This leads to the inability to accurately early warn the chain failure caused by the coupling effect. Serious lag: the topography monitoring is periodic (months or years), and cannot capture the rapid development of the scour pit in extreme events such as storms, missing the best intervention opportunity. Rely on manual experience: the early warning relies on the subjective judgment of engineers on two independent data, lacks automatic and intelligent decision support, is low in efficiency and prone to errors. Lack of prediction ability: only the occurred conditions can be monitored and alarmed, and the development trend of the scour-vibration coupling system in a future period of time cannot be predicted based on the current state, and the method does not have foresight.

[0004] To solve the above problems, the present application provides a seabed pipeline state early warning method and system based on scour-vibration coupling sensing, which integrates real-time monitoring, coupling simulation, intelligent prediction and active early warning into one method and system, and realizes real-time sensing and foresight early warning of the scour-vibration coupling state of the seabed pipeline. SUMMARY

[0005] The present application overcomes the defects of the prior art and provides a seabed pipeline state early warning method and system based on scour-vibration coupling sensing.

[0006] To achieve the above object, the first aspect of the present application provides a seabed pipeline state early warning method based on scour-vibration coupling perception, comprising:

[0007] Obtain historical marine environment information of a target seabed pipeline layout area, set a monitoring sensor array based on the historical marine environment information, monitor the target seabed pipeline area to obtain multi-source monitoring data and perform preprocessing to generate a multi-source monitoring data set;

[0008] Construct a multi-physical field coupling analysis model, input the multi-source monitoring data set into the multi-physical field coupling analysis model to analyze the response state of the target seabed pipeline under the seabed environment in the current monitoring period, and obtain pipeline response simulation information;

[0009] Based on the pipeline response simulation information, analyze the segment resonance influence between different seabed pipeline segments caused by seabed environment scour and generate resonance derived features, and construct pipeline multi-modal fusion features in combination with the multi-source monitoring data set;

[0010] Construct a seabed pipeline state analysis model, import the pipeline multi-modal fusion features to analyze the response state of the target seabed pipeline in the future time period, obtain response state analysis information, judge whether there is a pipeline operation risk and perform risk early warning prompt;

[0011] According to the pipeline modal fusion features, extract resonance derived features, analyze the fatigue life of each segment of the target seabed pipeline through the resonance derived features, and identify the easy fatigue segment nodes to generate maintenance suggestions for pushing.

[0012] In the present scheme, the historical marine environment information of the target seabed pipeline layout area is obtained, the monitoring sensor array is set based on the historical marine environment information, the target seabed pipeline area is monitored to obtain multi-source monitoring data and perform preprocessing to generate a multi-source monitoring data set, and specifically includes:

[0013] The historical marine environment information of the target seabed pipeline layout area is obtained through big data retrieval technology, the historical marine environment information is feature extracted to obtain flow velocity and direction time series data, wave statistical feature parameters and seabed geological exploration data, and historical marine environment feature information is obtained;

[0014] According to the historical marine environment feature information, spatial gridding processing is performed, the K-means clustering algorithm is used to divide the pipeline along the line into different environmental units, each unit is classified according to the bottom type and flow field characteristics, and an environmental zoning atlas with different environmental characteristics is formed;

[0015] The principal component analysis method is introduced to evaluate the region type of each sub-region in the environmental distribution map, the principal component score of the corresponding environmental characteristics of each sub-region is calculated, and the environmental characteristics within the preset principal component score range are selected as the key environmental characteristics;

[0016] The region type of each sub-region is evaluated by the key environmental characteristics in each sub-region, the monitoring sensor setting scheme of different sub-regions is set through the region type evaluation result, and the monitoring sensor array is set in the target submarine pipeline laying area through the set sensor setting scheme;

[0017] The target submarine pipeline region is monitored by the set monitoring sensor array to obtain multi-source monitoring data, which is input into the adaptive Kalman filter to eliminate noise interference caused by wave disturbance;

[0018] The Db4 wavelet base function is used for 6-layer wavelet packet decomposition to extract the vortex-induced vibration energy eigenvalue, the high-frequency noise component is eliminated through empirical mode decomposition processing, the instantaneous change rate of the pipe bottom gap and the standard deviation of the flow velocity gradient are calculated, and each type of monitoring data is further spatio-temporally aligned to finally generate a multi-source monitoring data set.

[0019] In the scheme, the multi-physical field coupling analysis model is constructed, the multi-source monitoring data set is input into the multi-physical field coupling analysis model to analyze the response of the target submarine pipeline in the submarine environment in the current monitoring period, and pipeline response simulation information is obtained, specifically including:

[0020] The pipeline structure information and submarine geographic information of the target submarine pipeline are obtained, and a multi-physical field coupling analysis model is constructed based on three-dimensional finite element analysis method, which is constructed by cascading and coupling fluid dynamics sub-model, pipeline structure mechanics sub-model and sediment dynamics model;

[0021] The boundary conditions and initial calculation domain are set according to the pipeline structure information and submarine geographic information, and the material properties and constitutive relationship are set; the multi-source monitoring data set is obtained, and the multi-source monitoring data set is input into the multi-physical field coupling analysis model to analyze the response of the target submarine pipeline in the submarine environment in the current monitoring period;

[0022] In the fluid dynamics sub-model, the Navier-Stokes equation is solved by using large eddy simulation method to simulate the complex three-dimensional unsteady turbulent flow field around the pipeline, calculate the fluctuating pressure distribution of each point on the pipeline surface, vortex shedding frequency and vortex strength, and finally output the fluid response simulation information with space-time attributes;

[0023] The fluid load field is extracted from the fluid response simulation information and used as the input to the pipeline structure mechanical sub-model. The Newmark-β integral method is used to solve the pipeline dynamic equations for transient dynamic analysis. The vibration displacement, acceleration and equivalent stress distribution of the target subsea pipeline at each time step are calculated, and finally the pipeline structure response simulation information is output.

[0024] The near-bottom velocity field and bed shear stress are extracted from the fluid response simulation information and input into the sediment dynamics model. The changes in seabed scouring and deposition around the pipeline are analyzed by the modified sediment initiation probability formula and sediment transport flux formula. The continuous evolution process of the development depth, width and morphology of scour pits is dynamically simulated, and finally the sediment response simulation information is output.

[0025] Subsequently, the simulation information of pipeline structure response and sediment response is imported into the fluid dynamics sub-model to update the fluid computation domain. The simulation process is repeated iteratively until the response status of the subsea pipeline at all time steps of the current monitoring period is simulated. Then, time-series alignment and correlation are performed to generate pipeline response simulation information containing spatiotemporal evolution characteristics.

[0026] In this scheme, the step of analyzing the segment resonance effects caused by seabed scouring between different subsea pipeline sections based on the pipeline response simulation information and generating resonance-derived features specifically includes:

[0027] The pipeline response simulation information is obtained, and the target subsea pipeline is divided into several sections according to the deployment scheme of the monitoring sensor array on the target subsea pipeline. Based on the section division results, section labels are generated and associated with the pipeline response simulation information.

[0028] Based on the associated pipeline response simulation information, the vortex shedding frequency and flow direction time history of each sub-section are calculated. The vortex shedding frequency is calculated from the three-dimensional unsteady turbulent field around the pipeline. Vibration mode parameters are extracted from the vibration response data of each pipeline section by the empirical mode decomposition method to generate a vibration mode feature set and perform two-level resonance coupling analysis.

[0029] The natural frequency sequence and local vortex shedding frequency sequence corresponding to each sub-segment are extracted through the vibration mode feature set. The natural frequency sequence and local vortex shedding frequency sequence of each sub-segment are compared, and the instantaneous frequency ratio and locking coefficient of the corresponding sub-segment are calculated.

[0030] Based on the instantaneous frequency ratio and locking coefficient, the critical section of the target subsea pipeline with isolated vortex-induced vibration is identified. The span length and vibration amplitude of the critical section are extracted by pipeline response simulation information, and weighted and fused with the locking coefficient to generate the section resonance confidence, and finally the first resonance derived feature is generated.

[0031] Based on the vibration mode feature set, the vibration transfer function matrix between all sections is constructed using the transfer path analysis method. The coherence function and phase spectrum between the vibration responses of adjacent and spaced sections are calculated to quantify the transmission path and directional characteristics of vibration energy in the subsea pipeline system.

[0032] Subsequently, the fluid field response simulation characteristics of each sub-section are obtained through pipeline response simulation information. The heterogeneous excitation caused by fluid loads with opposite flow directions or velocity differences greater than a preset threshold in adjacent sections is analyzed, and a heterogeneous excitation distribution feature vector is generated.

[0033] By calculating the vibration transfer function matrix to quantify the cancellation or enhancement effect of out-of-phase excitation on the overall vibration mode of the pipeline, the segment coupling coefficient characterizing the intensity of dynamic interaction between segments, the attenuation coefficient characterizing the vibration energy transfer efficiency, and the modal participation factor reflecting the contribution of each segment in the overall pipeline system vibration mode are generated, thus obtaining the second resonance derived characteristics.

[0034] In this solution, the construction of pipeline multimodal fusion features by combining the multi-source monitoring dataset specifically includes:

[0035] Acquire the first resonance-derived features, the second resonance-derived features, pipeline response simulation information, and multi-source monitoring datasets;

[0036] The simulated vibration data sequence is extracted from the pipeline response simulation information, and the measured vibration data sequence is extracted from the multi-source monitoring dataset. The dynamic time warping algorithm is introduced to align the simulated vibration data sequence and the measured vibration data sequence and mark the difference sequence segments to generate a fused vibration data sequence.

[0037] The fused vibration data sequence, the first resonance derived feature, the second resonance derived feature, the pipeline response simulation information, and the multi-source monitoring dataset are input into a multimodal feature fusion network based on a multilayer perceptron architecture for multimodal feature fusion.

[0038] Feature interaction and compression are performed through a multilayer perceptron. The temporal dependency and spatial correlation between features are captured by a gated recurrent unit. An attention mechanism is introduced to dynamically calculate attention weights through the gated recurrent unit network for adaptive fusion, ultimately generating pipeline multimodal fusion features.

[0039] In this solution, the construction of a subsea pipeline status analysis model, importing the pipeline's multimodal fusion feature analysis to determine the target subsea pipeline's response status within a future time period, obtaining response status analysis information, determining whether there are pipeline operational risks, and issuing risk warnings specifically includes:

[0040] A state analysis model for subsea pipelines is constructed based on a Bayesian inference network. A bidirectional long short-term memory network is used to extract the temporal dependencies of multimodal features as the observation layer. The pipeline vibration state, stress state and geometric state form a multidimensional hidden state layer.

[0041] Based on the pre-set training dataset, the expectation-maximization algorithm is used to learn the state transition probability matrix and the observation probability distribution. The state transition probability matrix is ​​estimated by the Metropolis-Hastings algorithm in Markov chain Monte Carlo sampling to capture the nonlinear and uncertain characteristics of pipeline state evolution, and finally obtain a subsea pipeline state analysis model that meets the expectations.

[0042] The pipeline multimodal fusion features are obtained, and the pipeline multimodal fusion features are input into the subsea pipeline state analysis model to obtain the transient prior distribution and probability likelihood function. The posterior probability distribution is calculated by a stochastic variational inference algorithm.

[0043] Monte Carlo sampling is performed in a Bayesian inference network based on the posterior probability distribution to infer the response state of the target subsea pipeline in the future time period, obtain the probability distribution of the response state of each pipeline in the future time step, and generate response state analysis information.

[0044] Based on the response state analysis information, a number of Markov chains are generated using a Markov algorithm. The Markov chains represent the state transition paths of the target subsea pipeline in the future time period. A risk assessment space is generated through the generated Markov chains.

[0045] In the risk assessment space, the confidence level of each Markov chain is calculated using the Viterbi algorithm to generate a confidence ranking table. The Markov chain with the highest confidence level is selected as the target state transition sequence based on the confidence ranking table. The sequence is then compared with preset risk assessment thresholds to generate a risk assessment report and provide risk warning prompts.

[0046] In this solution, the step of extracting resonance-derived features based on the pipeline modal fusion features, analyzing the fatigue life of each section of the target subsea pipeline through the resonance-derived features, and identifying fatigue-prone section nodes to generate and push maintenance suggestions specifically includes:

[0047] The pipeline multimodal fusion features are obtained, and the first resonance derived features and the second resonance derived features are obtained through the pipeline multimodal fusion features. The dynamic stress amplitude time series, resonance duration, segment coupling coefficient, vibration energy transfer efficiency and modal participation factor are extracted to obtain the resonance derived feature set.

[0048] Based on the resonance-derived feature set, the dynamic stress time history of each sub-segment is cyclically counted using the rainflow counting algorithm to obtain the initial stress cyclic statistical matrix of each sub-segment.

[0049] A vibration transmission weight matrix is ​​constructed based on the segment coupling coefficient, and matrix operations are performed with the initial stress cycle statistics matrix to correct the vibration transmission influence between adjacent segments and generate a regularized stress cycle distribution.

[0050] The SN curve characteristics of the target subsea pipeline material are obtained, and the fatigue damage degree of each sub-segment in the current monitoring cycle is calculated using the Miner linear cumulative damage theory in combination with the regularized stress cycle distribution, generating a cumulative damage degree distribution map.

[0051] The historical operating time and load of the pipeline are obtained, and the cumulative damage distribution map is input into the pre-trained remaining life prediction model to analyze the remaining fatigue life of each sub-section and obtain the remaining life probability distribution map.

[0052] The remaining life probability distribution map is used to identify fatigue-prone segment nodes with lifespans below a preset threshold using spatial clustering analysis. Maintenance priorities are then set based on the corresponding remaining lifespans, and a maintenance priority ranking table is generated.

[0053] Meteorological data of the target subsea pipeline deployment area is obtained, and the subsea pipeline sections that need maintenance are extracted through the maintenance priority ranking table. The maintenance time window is generated in combination with the meteorological data to generate a pipeline maintenance auxiliary report and push it out.

[0054] The second aspect of the present invention provides a submarine pipeline status early warning system based on scour-vibration coupling sensing. The submarine pipeline status early warning system based on scour-vibration coupling sensing is used to implement the submarine pipeline status early warning method based on scour-vibration coupling sensing as described in any of the above claims. The system includes a multi-source intelligent sensing module, a pipeline response multi-physics field coupling analysis module, and an intelligent early warning and decision-making module.

[0055] The multi-source intelligent sensing module consists of a vibration sensor array, an acoustic Doppler current profiler, an integrated ranging unit, and a data preprocessing and edge computing unit, and is used to monitor the target subsea pipeline.

[0056] The pipeline response multiphysics coupling analysis module is composed of a multiphysics coupling analysis model, which performs scour-vibration coupling sensing on the target subsea pipeline based on real-time monitoring data processed by data preprocessing and edge computing units.

[0057] The intelligent early warning and decision-making module consists of a coupled state assessment unit, an adaptive early warning unit, and a decision support unit. It analyzes the response state of the target subsea pipeline through the scour-vibration coupled sensing results, determines whether there is a risk to pipeline operation, and provides risk warning prompts and intervention measures suggestions.

[0058] A third aspect of the present invention provides a computer-readable storage medium comprising a subsea pipeline status early warning method program based on scour-vibration coupling sensing. When the subsea pipeline status early warning method program based on scour-vibration coupling sensing is executed by a processor, it implements the steps of the subsea pipeline status early warning method based on scour-vibration coupling sensing as described in any of the preceding claims. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments or examples of the present invention, the drawings used in the embodiments or examples will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained according to these drawings without creative effort.

[0060] Figure 1 A flowchart of the first method of a subsea pipeline status early warning method based on scour-vibration coupling sensing provided in an embodiment of the present invention;

[0061] Figure 2 This is a flowchart of a second method for an early warning method for the condition of a subsea pipeline based on scour-vibration coupling sensing, provided in an embodiment of the present invention.

[0062] Figure 3 This is a block diagram of a subsea pipeline status early warning system based on scour-vibration coupling sensing, provided in an embodiment of the present invention.

[0063] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0065] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0066] Figure 1 A flowchart of the first method of a subsea pipeline status early warning method based on scour-vibration coupling sensing provided in an embodiment of the present invention;

[0067] like Figure 1As shown, the present invention provides a first method flowchart for an early warning method for the condition of a subsea pipeline based on scour-vibration coupling sensing, comprising:

[0068] S102, acquire historical marine environmental information of the target subsea pipeline deployment area, set up a monitoring sensor array based on the historical marine environmental information, monitor the target subsea pipeline area to acquire multi-source monitoring data and preprocess it to generate a multi-source monitoring dataset.

[0069] S104, Construct a multiphysics coupling analysis model, input the multi-source monitoring dataset into the multiphysics coupling analysis model to analyze the response status of the target subsea pipeline in the current monitoring period, and obtain pipeline response simulation information;

[0070] S106, Based on the pipeline response simulation information, analyze the impact of segment resonance caused by seabed scouring between different subsea pipeline sections and generate resonance-derived features, and construct pipeline multimodal fusion features by combining the multi-source monitoring dataset;

[0071] S108, Construct a subsea pipeline status analysis model, import the multimodal fusion feature analysis of the pipeline to analyze the response status of the target subsea pipeline in the future time period, obtain response status analysis information, determine whether there is a pipeline operation risk and issue a risk warning.

[0072] S110, extract resonance-derived features based on the pipeline modal fusion features, analyze the fatigue life of each section of the target subsea pipeline through the resonance-derived features, and identify fatigue-prone section nodes to generate maintenance suggestions for push.

[0073] Furthermore, in a preferred embodiment of this scheme, the step of acquiring historical marine environmental information of the target subsea pipeline deployment area, setting up a monitoring sensor array based on the historical marine environmental information, monitoring the target subsea pipeline area to acquire multi-source monitoring data, and preprocessing the data to generate a multi-source monitoring dataset specifically includes:

[0074] Historical marine environmental information of the target submarine pipeline layout area is obtained by big data retrieval technology. Feature extraction is performed on the historical marine environmental information to obtain time series data of flow velocity and direction, wave statistical characteristic parameters and seabed geological exploration data, so as to obtain historical marine environmental characteristic information.

[0075] Based on historical marine environmental characteristics, spatial gridding is performed, and the pipeline is divided into different environmental units using the K-means clustering algorithm. Each unit is combined and classified according to the substrate type and flow field characteristics to form an environmental zoning map with different environmental characteristics.

[0076] Principal component analysis is introduced to evaluate the regional type of each sub-region in the environmental distribution map. By calculating the principal component scores of the corresponding environmental features of each sub-region, environmental features within the preset principal component score range are selected as key environmental features.

[0077] The region type of each sub-region is assessed by identifying the key environmental characteristics within each sub-region. Based on the results of the region type assessment, a monitoring sensor setting scheme is set for different sub-regions. Based on the set sensor setting scheme, a monitoring sensor array is set in the target subsea pipeline deployment area.

[0078] The target subsea pipeline area is monitored by a set monitoring sensor array to obtain multi-source monitoring data, which is then imported into an adaptive Kalman filter to eliminate noise interference caused by wave disturbances.

[0079] The energy features of vortex-induced vibration were extracted by 6-level wavelet packet decomposition using the Db4 wavelet basis function. High-frequency noise components were eliminated by empirical mode decomposition. At the same time, the instantaneous rate of change of the pipe bottom gap and the standard deviation of the velocity gradient were calculated. The monitoring data of various types were further spatiotemporally aligned to finally generate a multi-source monitoring dataset.

[0080] It should be noted that in most subsea pipeline monitoring systems, monitoring arrays are typically deployed at fixed intervals. However, due to differences in environmental conditions across different sections, fixed-distance deployments may suffer from insufficient monitoring accuracy or redundancy. Therefore, this study utilizes big data retrieval technology to obtain historical marine environmental information of the target subsea pipeline deployment area. Deep feature extraction is then performed on this raw data, primarily acquiring three types of core data: time-series data on current velocity and direction (reflecting the spatiotemporal variation of ocean currents), wave statistical parameters (including wave height, wave period, and wave direction), and seabed geological exploration data (containing information on seabed type, soil structure, and seabed topography). This forms historical marine environmental characteristic information. Based on this characteristic information, spatial gridding is first performed to divide the pipeline area into regular grid cells. Then, the K-means clustering algorithm is used to classify and group cells based on seabed type and flow field characteristics, aggregating adjacent cells with similar physical environments into the same category. Finally, an environmental zoning map that visually displays spatial heterogeneity is generated. To further quantify the importance of environmental features in each sub-region, principal component analysis (PCA) was introduced to reduce the dimensionality of the multidimensional environmental features in each sub-region of the map. By calculating the principal component scores of the environmental features in each sub-region and setting threshold ranges, the key environmental feature combinations with the greatest impact on the regional environment were selected. Based on these key features, each sub-region was refined into a type assessment, and differentiated monitoring sensor configuration schemes were developed based on the assessment results (e.g., increasing the density of flow velocity sensors in areas with complex flow fields and deploying geological monitoring equipment in areas with unstable sediment). Finally, a scientifically deployed monitoring sensor array was completed in the target pipeline area. After real-time acquisition of multi-source monitoring data from the target pipeline area by this sensor array, the raw monitoring data was first imported into an adaptive Kalman filter. The real-time adjustment of filter parameters effectively eliminated random noise interference caused by wave disturbances. Subsequently, a 6-level wavelet packet decomposition was performed using the Db4 wavelet basis function to extract eddy-induced vibration energy feature values ​​of different frequency bands from the vibration signal. Furthermore, empirical mode decomposition was used to eliminate residual high-frequency noise components. The instantaneous rate of change of the pipe bottom gap (reflecting the dynamic process of scouring) and the standard deviation of the velocity gradient (characterizing the stability of the flow field) are calculated simultaneously. Finally, a spatiotemporal registration algorithm is used to accurately align and fuse all types of monitoring data in terms of timestamps and spatial locations, ultimately generating a high-quality multi-source monitoring dataset, providing a reliable data foundation for subsequent analysis.

[0081] Furthermore, in a preferred embodiment of this scheme, the step of constructing a multiphysics coupling analysis model, inputting the multi-source monitoring dataset into the multiphysics coupling analysis model to analyze the response status of the target subsea pipeline in the current monitoring period, and obtaining pipeline response simulation information, specifically includes:

[0082] The pipeline structure information and seabed geographic information of the target subsea pipeline are obtained, and a multiphysics coupling analysis model is constructed based on the three-dimensional finite element analysis method. The multiphysics coupling analysis model is constructed by cascading coupling of a fluid dynamics sub-model, a pipeline structural mechanics sub-model, and a sediment dynamics model.

[0083] Based on the pipeline structure information and seabed geographic information, boundary conditions and initial computational domain are set, and material properties and constitutive relations are set; multi-source monitoring dataset is obtained, and the multi-source monitoring dataset is input into the multi-physics coupling analysis model to analyze the response status of the target seabed pipeline in the seabed environment during the current monitoring period;

[0084] In the fluid dynamics sub-model, the Navier-Stokes equations are solved using the large eddy simulation method to simulate the complex three-dimensional unsteady turbulent flow field around the pipe, calculate the pulsating pressure distribution, vortex shedding frequency and vortex intensity at various points on the pipe surface, and finally output fluid response simulation information with spatiotemporal attributes.

[0085] The fluid load field is extracted from the fluid response simulation information and used as the input to the pipeline structure mechanical sub-model. The Newmark-β integral method is used to solve the pipeline dynamic equations for transient dynamic analysis. The vibration displacement, acceleration and equivalent stress distribution of the target subsea pipeline at each time step are calculated, and finally the pipeline structure response simulation information is output.

[0086] The near-bottom velocity field and bed shear stress are extracted from the fluid response simulation information and input into the sediment dynamics model. The changes in seabed scouring and deposition around the pipeline are analyzed by the modified sediment initiation probability formula and sediment transport flux formula. The continuous evolution process of the development depth, width and morphology of scour pits is dynamically simulated, and finally the sediment response simulation information is output.

[0087] Subsequently, the simulation information of pipeline structure response and sediment response is imported into the fluid dynamics sub-model to update the fluid computation domain. The simulation process is repeated iteratively until the response status of the subsea pipeline at all time steps of the current monitoring period is simulated. Then, time-series alignment and correlation are performed to generate pipeline response simulation information containing spatiotemporal evolution characteristics.

[0088] It should be noted that current subsea pipeline condition monitoring mostly employs vibration monitoring systems and seabed topography monitoring systems as independent modules, failing to consider corresponding coupling effects, resulting in insufficient sensing accuracy. Therefore, this study obtains detailed pipeline structural information (including pipe diameter, wall thickness, material grade, etc.) and seabed geographic information (such as seabed topography and sediment type) of the target subsea pipeline. Based on this information, a comprehensive simulation model is constructed using the three-dimensional finite element method, consisting of a cascaded coupling of a fluid dynamics sub-model, a pipeline structural mechanics sub-model, and a sediment dynamics model. First, the boundary conditions of the model are set according to the pipeline structure and seabed geographic information, and the computational domain is initialized. Simultaneously, material properties, including the constitutive relation of steel and seabed soil parameters, are accurately set. Then, the preprocessed multi-source monitoring dataset is used as input to drive this coupled model to perform a high-precision simulation of the pipeline response under the seabed environment during the current monitoring period. In the fluid dynamics sub-model, the Navier-Stokes equations are solved using the large eddy simulation method to finely simulate the complex three-dimensional unsteady turbulent flow field around the pipeline. Key parameters such as the fluctuating pressure distribution, vortex shedding frequency, and vortex intensity at various points on the pipeline surface are calculated, outputting fluid response simulation information with spatiotemporal attributes. The fluid load field extracted from the fluid response simulation information is input to the pipeline structural mechanics sub-model. The Newmark-β integral method is used to solve the pipeline's dynamic equations for transient dynamic analysis, calculating the detailed distribution of vibration displacement, acceleration, and equivalent stress at each time step, outputting pipeline structural response simulation information. Simultaneously, the near-bottom velocity field and bed shear stress are extracted from the fluid response information and input to the sediment dynamics model. The modified sediment initiation probability formula and sediment transport flux formula are used to analyze the scouring and deposition processes of the seabed around the pipeline, dynamically simulating the continuous evolution of the development depth, width, and morphology of scour pits, outputting sediment response simulation information. Finally, the simulation information of pipeline structure response and sediment response is fed back into the fluid dynamics sub-model to update the fluid computation domain in real time. Through this bidirectional coupled iterative simulation process, the simulation is continuously calculated until all time steps of the current monitoring cycle are completed. Finally, all output data are time-series aligned and spatially correlated to generate pipeline response simulation information that fully reflects the spatiotemporal evolution characteristics of the pipeline, overcoming the shortcomings of independent data monitoring in traditional monitoring methods.

[0089] Furthermore, in a preferred embodiment of this scheme, the step of analyzing the segment resonance effects caused by seabed scouring between different subsea pipeline sections based on the pipeline response simulation information and generating resonance-derived features specifically includes:

[0090] The pipeline response simulation information is obtained, and the target subsea pipeline is divided into several sections according to the deployment scheme of the monitoring sensor array on the target subsea pipeline. Based on the section division results, section labels are generated and associated with the pipeline response simulation information.

[0091] Based on the associated pipeline response simulation information, the vortex shedding frequency and flow direction time history of each sub-section are calculated. The vortex shedding frequency is calculated from the three-dimensional unsteady turbulent field around the pipeline. Vibration mode parameters are extracted from the vibration response data of each pipeline section by the empirical mode decomposition method to generate a vibration mode feature set and perform two-level resonance coupling analysis.

[0092] The natural frequency sequence and local vortex shedding frequency sequence corresponding to each sub-segment are extracted through the vibration mode feature set. The natural frequency sequence and local vortex shedding frequency sequence of each sub-segment are compared, and the instantaneous frequency ratio and locking coefficient of the corresponding sub-segment are calculated.

[0093] Based on the instantaneous frequency ratio and locking coefficient, the critical section of the target subsea pipeline with isolated vortex-induced vibration is identified. The span length and vibration amplitude of the critical section are extracted by pipeline response simulation information, and weighted and fused with the locking coefficient to generate the section resonance confidence, and finally the first resonance derived feature is generated.

[0094] Based on the vibration mode feature set, the vibration transfer function matrix between all sections is constructed using the transfer path analysis method. The coherence function and phase spectrum between the vibration responses of adjacent and spaced sections are calculated to quantify the transmission path and directional characteristics of vibration energy in the subsea pipeline system.

[0095] Subsequently, the fluid field response simulation characteristics of each sub-section are obtained through pipeline response simulation information. The heterogeneous excitation caused by fluid loads with opposite flow directions or velocity differences greater than a preset threshold in adjacent sections is analyzed, and a heterogeneous excitation distribution feature vector is generated.

[0096] By calculating the vibration transfer function matrix to quantify the cancellation or enhancement effect of out-of-phase excitation on the overall vibration mode of the pipeline, the segment coupling coefficient characterizing the intensity of dynamic interaction between segments, the attenuation coefficient characterizing the vibration energy transfer efficiency, and the modal participation factor reflecting the contribution of each segment in the overall pipeline system vibration mode are generated, thus obtaining the second resonance derived characteristics.

[0097] It should be noted that traditional monitoring methods acquire vibration data from multiple vibration monitoring sensors at various locations, and then use spatial interpolation algorithms to summarize the vibration distribution of the entire pipeline. However, due to the complexity of the seabed environment, there may be undersea currents or different geological conditions in different directions, resulting in differences in vibration conditions. This means that vibrations between adjacent sections may resonate, leading to either enhanced or weakened vibrations. Point monitoring and spatial interpolation alone cannot account for this situation, thus reducing the monitoring accuracy of the seabed pipeline. Therefore, based on a pre-designed monitoring sensor array deployment scheme, the target seabed pipeline is numerically divided into several continuous segments, and a unique identifier is generated for each segment to achieve precise correlation with the spatial coordinates in the simulation information. Based on this correlation, the time series of vortex shedding frequencies and flow direction time histories for each sub-segment are extracted from the fluid response simulation information. The vortex shedding frequencies are obtained by analyzing the vortex motion characteristics in the three-dimensional unsteady turbulent field surrounding the pipe. Simultaneously, the vibration acceleration response data of each pipe segment is processed using the empirical mode decomposition method to extract its natural frequencies, damping ratios, and mode shapes, forming a complete vibration mode feature set. By comparing and analyzing the natural frequency sequences of each segment extracted from the vibration mode feature set with the local vortex shedding frequency sequences, the instantaneous frequency ratio and its statistical lock-in coefficient at each time step are calculated, thereby identifying critical segments that may experience isolated vortex-induced vibration. Furthermore, the suspension length and vibration amplitude parameters of these critical segments are obtained from the simulation information, and a weighted fusion algorithm is used with the lock-in coefficient to generate a quantified segment resonance confidence index, ultimately forming the first resonance derived feature, which directly reflects the resonance risk level of the local pipe segment.

[0098] Subsequently, based on the full pipeline vibration modal feature set, a vibration transfer function matrix between all sections was constructed using the transfer path analysis method. By calculating the coherence function and phase spectrum characteristics between vibration response signals of adjacent and inter-section sections, the transmission path and directional characteristics of vibration energy in the pipeline system were quantified. Simultaneously, the anisotropic excitation phenomenon caused by opposite flow directions or velocity differences exceeding a preset threshold in adjacent sections was analyzed from fluid field response simulation information, generating spatially distributed anisotropic excitation feature vectors. The modulation effect of anisotropic excitation on the overall pipeline vibration mode was calculated using the vibration transfer function matrix, generating section coupling coefficients characterizing the intensity of dynamic interaction between sections, attenuation coefficients characterizing vibration energy transfer efficiency, and system modal participation factors obtained through modal decomposition, ultimately forming the second resonance-derived feature. The core role of resonance feature analysis lies in overcoming the limitations of traditional isolated analysis of single sections. By establishing the correlation between local vibration and system response, it can not only identify high-risk sections of local resonance but also reveal the transmission law and coupling mechanism of vibration energy within the pipeline system. Its significance lies in providing two key dimensions of information for pipeline condition assessment: the first resonance-derived features directly indicate local structural safety risks, providing a basis for key monitoring and maintenance; the second resonance-derived features reveal the system-level dynamic interaction mechanism, providing a scientific basis for understanding complex vibration phenomena, predicting long-term evolution trends, and optimizing overall safety strategies, ultimately providing core technical support for achieving accurate safety early warning and intelligent maintenance decisions throughout the entire life cycle of pipeline systems.

[0099] Furthermore, in a preferred embodiment of this scheme, the step of constructing pipeline multimodal fusion features by combining the multi-source monitoring dataset specifically includes:

[0100] Acquire the first resonance-derived features, the second resonance-derived features, pipeline response simulation information, and multi-source monitoring datasets;

[0101] The simulated vibration data sequence is extracted from the pipeline response simulation information, and the measured vibration data sequence is extracted from the multi-source monitoring dataset. The dynamic time warping algorithm is introduced to align the simulated vibration data sequence and the measured vibration data sequence and mark the difference sequence segments to generate a fused vibration data sequence.

[0102] The fused vibration data sequence, the first resonance derived feature, the second resonance derived feature, the pipeline response simulation information, and the multi-source monitoring dataset are input into a multimodal feature fusion network based on a multilayer perceptron architecture for multimodal feature fusion.

[0103] Feature interaction and compression are performed through a multilayer perceptron. The temporal dependency and spatial correlation between features are captured by a gated recurrent unit. An attention mechanism is introduced to dynamically calculate attention weights through the gated recurrent unit network for adaptive fusion, ultimately generating pipeline multimodal fusion features.

[0104] It should be noted that multimodal feature fusion technology combines numerical simulation and measured data to construct a comprehensive feature vector that fully reflects the pipeline's condition. First, data alignment and preprocessing are performed: simulated vibration data sequences are extracted from the pipeline response simulation information, while measured vibration data sequences are extracted from the multi-source monitoring dataset. Considering the time scale and phase differences between the numerical simulation and measured data, a dynamic time warping algorithm is introduced to perform nonlinear alignment of the two sequences. This algorithm matches the two sequences in the time domain by finding the optimal bending path and automatically marks significantly different sequence segments. Based on the alignment results, a fused vibration data sequence is generated, which retains the integrity of the simulation data while incorporating the authenticity of the measured data. Subsequently, all features are input into a specially designed multimodal feature fusion network for processing. This network uses a multilayer perceptron as its basic architecture to perform feature interaction and dimensionality compression on the multi-source inputs, including the fused vibration data sequence, two types of resonance-derived features, simulation information, and monitoring data. The network uses gated recurrent unit layers to capture the temporal dependencies between multimodal features and extracts spatial correlation features through convolutional operations. To further enhance the fusion effect, an attention mechanism is introduced to dynamically calculate the attention weights of each feature through a gated recurrent unit network, achieving adaptive evaluation and weighted fusion of the importance of different features. The resulting multimodal fusion feature of the pipeline combines the advantages of mechanistic analysis from numerical simulation with the authenticity of measured data, while also including local resonance characteristics and system coupled vibration characteristics, providing a comprehensive and reliable data foundation for subsequent pipeline condition assessment and risk warning.

[0105] Furthermore, in a preferred embodiment of this scheme, the step of constructing a subsea pipeline status analysis model, importing the multimodal fusion feature analysis of the pipeline to analyze the response status of the target subsea pipeline within a future time period, obtaining response status analysis information, determining whether there is a pipeline operation risk, and issuing a risk warning specifically includes:

[0106] A state analysis model for subsea pipelines is constructed based on a Bayesian inference network. A bidirectional long short-term memory network is used to extract the temporal dependencies of multimodal features as the observation layer. The pipeline vibration state, stress state and geometric state form a multidimensional hidden state layer.

[0107] Based on the pre-set training dataset, the expectation-maximization algorithm is used to learn the state transition probability matrix and the observation probability distribution. The state transition probability matrix is ​​estimated by the Metropolis-Hastings algorithm in Markov chain Monte Carlo sampling to capture the nonlinear and uncertain characteristics of pipeline state evolution, and finally obtain a subsea pipeline state analysis model that meets the expectations.

[0108] The pipeline multimodal fusion features are obtained, and the pipeline multimodal fusion features are input into the subsea pipeline state analysis model to obtain the transient prior distribution and probability likelihood function. The posterior probability distribution is calculated by a stochastic variational inference algorithm.

[0109] Monte Carlo sampling is performed in a Bayesian inference network based on the posterior probability distribution to infer the response state of the target subsea pipeline in the future time period, obtain the probability distribution of the response state of each pipeline in the future time step, and generate response state analysis information.

[0110] Based on the response state analysis information, a number of Markov chains are generated using a Markov algorithm. The Markov chains represent the state transition paths of the target subsea pipeline in the future time period. A risk assessment space is generated through the generated Markov chains.

[0111] In the risk assessment space, the confidence level of each Markov chain is calculated using the Viterbi algorithm to generate a confidence ranking table. The Markov chain with the highest confidence level is selected as the target state transition sequence based on the confidence ranking table. The sequence is then compared with preset risk assessment thresholds to generate a risk assessment report and provide risk warning prompts.

[0112] It should be noted that the subsea pipeline state analysis model built based on a Bayesian inference network employs a bidirectional long short-term memory network (Bi-LSTM) as the observation layer to extract complex temporal dependencies in the pipeline's multimodal fusion features. Simultaneously, a multidimensional hidden state layer is constructed using pipeline vibration state, stress state, and geometric morphology state, forming a complete dynamic Bayesian network architecture. During model training, the expectation-maximization algorithm is used to process the pre-set training dataset. The state transition probability matrix is ​​estimated using the Metropolis-Hastings algorithm in Markov chain Monte Carlo sampling. This algorithm effectively captures the nonlinear and uncertain characteristics of the pipeline state evolution process through an acceptance-rejection sampling strategy. Simultaneously, it learns the observation probability distribution using observation data, ultimately obtaining a fully trained state analysis model. Subsequently, the real-time acquired multimodal fusion features of the pipeline are input into the trained model. The probability likelihood function is calculated through the observation layer, and combined with the transient prior distribution obtained from historical state inference, the posterior probability distribution is calculated using a stochastic variational inference algorithm. Based on the obtained posterior probability distribution, sequential Monte Carlo sampling is performed in the Bayesian inference network. Through importance sampling and resampling strategies, the response state of the target subsea pipeline within future time periods is estimated, outputting the probability distribution and confidence interval of the pipeline's response state at multiple future time steps, forming response state analysis information containing uncertainty information. Subsequently, based on this information, multiple Markov chain states transitioning paths are generated using the Markov chain Monte Carlo method. These chains represent the possible future state evolution trajectories of the pipeline, and a multi-dimensional risk assessment space is constructed by aggregating these paths. In this risk assessment space, the maximum likelihood probability of each Markov chain is calculated using the Viterbi algorithm, generating a Markov chain sequence list sorted by confidence level. The chain with the highest confidence level is selected as the most probable state transition path, and it is compared and analyzed with preset risk assessment indicators such as stress threshold, vibration threshold, and displacement threshold. Finally, a quantitative risk assessment report is generated, triggering corresponding risk warning prompts. The significance of this Bayesian state analysis model lies in its ability to effectively handle the uncertainty and nonlinear characteristics in pipeline state assessment, providing confidence interval estimates of future states through probabilistic inference methods, rather than a single deterministic result. Its core value lies in three aspects: first, effectively capturing the temporal dynamic characteristics of multimodal features through bidirectional long short-term memory networks; second, accurately characterizing the uncertainty of state transitions through Markov chain methods; and third, providing quantitative risk warning levels through probabilistic risk assessment. This state analysis method based on probabilistic reasoning significantly improves the reliability and accuracy of pipeline risk warnings, providing important technical support for the safe operation and intelligent maintenance of subsea pipelines.

[0113] Figure 2 This is a flowchart of a second method for an early warning method for the condition of a subsea pipeline based on scour-vibration coupling sensing, provided in an embodiment of the present invention.

[0114] like Figure 2 As shown, the present invention provides a second method flowchart for an early warning method for the condition of subsea pipelines based on scour-vibration coupling sensing, comprising:

[0115] S202, obtain the multimodal fusion features of the pipeline, obtain the first resonance derived features and the second resonance derived features through the multimodal fusion features of the pipeline, extract the dynamic stress amplitude time series, resonance duration, segment coupling coefficient, vibration energy transfer efficiency and modal participation factor, and obtain the resonance derived feature set;

[0116] S204, Based on the resonance-derived feature set, the dynamic stress time history of each sub-segment is cyclically counted using the rainflow counting algorithm to obtain the initial stress cyclic statistical matrix of each sub-segment;

[0117] S206, construct a vibration transmission weight matrix based on the segment coupling coefficient, and perform matrix operations with the initial stress cycle statistics matrix to correct the vibration transmission influence between adjacent segments and generate a regularized stress cycle distribution;

[0118] S208, Obtain the SN curve characteristics of the target subsea pipeline material, and calculate the fatigue damage degree of each sub-segment in the current monitoring cycle using the Miner linear cumulative damage theory in combination with the regularized stress cycle distribution, and generate a cumulative damage degree distribution map.

[0119] S210, Obtain the historical operating time and load of the pipeline, and input the cumulative damage distribution map into the pre-trained remaining life prediction model to analyze the remaining fatigue life of each sub-section and obtain the remaining life probability distribution map.

[0120] S212, using the remaining life probability distribution map, identify fatigue-prone segment nodes with lifespans below a preset threshold using spatial clustering analysis, and set maintenance priorities based on the corresponding remaining lifespans to generate a maintenance priority ranking table.

[0121] S214. Obtain meteorological data of the target submarine pipeline deployment area, extract the submarine pipeline sections that need maintenance through the maintenance priority ranking table, and generate a pipeline maintenance auxiliary report by combining the meteorological data with the maintenance time window and push it.

[0122] It should be noted that pipeline fatigue life assessment and maintenance decisions are based on resonance-derived features from multimodal fusion features. First resonance-derived features (including dynamic stress amplitude time series and resonance duration) and second resonance-derived features (including segment coupling coefficient, vibration energy transfer efficiency, and modal participation factor) are extracted from the pipeline's multimodal fusion features and integrated to form a resonance-derived feature set. Based on this feature set, a rainflow counting algorithm is used to perform cyclic counting of the dynamic stress time history of each sub-segment, counting the number of cycles under different stress levels to generate an initial stress cycle statistical matrix. Subsequently, a vibration transmission weight matrix is ​​constructed based on the segment coupling coefficient, which quantifies the vibration energy transfer efficiency between adjacent segments. The vibration transmission weight matrix is ​​combined with the initial stress cycle statistical matrix through matrix operations to correct the additional stress influence caused by vibration transmission from adjacent segments, generating a regularized stress cycle distribution that better reflects actual working conditions. Based on the SN curve characteristics of the target subsea pipeline material and combined with the regularized stress cycle distribution, Miner's linear cumulative damage theory is used to calculate the fatigue damage degree of each sub-segment within the current monitoring cycle, forming a cumulative damage degree distribution map reflecting spatial distribution characteristics. The cumulative damage distribution map, along with historical pipeline operating time and load data, is input into a pre-trained remaining life prediction model. This model calculates the crack propagation rate using the Paris formula and considers corrosion fatigue effects, outputting a probability distribution map of the remaining fatigue life for each sub-segment. Based on this distribution map, spatial clustering analysis is used to identify fatigue-prone segments with lifespans below a preset threshold. Maintenance priorities are then set according to the remaining lifespan, generating a maintenance priority ranking table. Finally, combined with meteorological data from the target pipeline area, the optimal operating time window is calculated for the sections requiring maintenance, generating a pipeline maintenance auxiliary report containing specific maintenance recommendations and schedules, which is then sent to relevant personnel. The significant advantage lies in providing accurate fatigue life assessment and maintenance decision support through resonance characteristic analysis. Its core value lies in considering the interaction between pipe segments through a vibration transmission weight matrix, overcoming the limitations of traditional isolated analysis, and combining meteorological conditions with maintenance decisions, thus improving the safety and effectiveness of maintenance operations. This provides a scientific basis and technical support for preventative maintenance and full life-cycle management of subsea pipelines, significantly improving the controllability and economy of pipeline safe operation.

[0123] Figure 3 An embodiment of the present invention provides a submarine pipeline status early warning system 3 based on scour-vibration coupling sensing, used to implement the submarine pipeline status early warning method based on scour-vibration coupling sensing as described in any of the above claims. The system includes a multi-source intelligent sensing module, a pipeline response multi-physics field coupling analysis module, and an intelligent early warning and decision-making module.

[0124] The multi-source intelligent sensing module consists of a vibration sensor array, an acoustic Doppler current profiler, an integrated ranging unit, and a data preprocessing and edge computing unit, and is used to monitor the target subsea pipeline.

[0125] The pipeline response multiphysics coupling analysis module is composed of a multiphysics coupling analysis model, which performs scour-vibration coupling sensing on the target subsea pipeline based on real-time monitoring data processed by data preprocessing and edge computing units.

[0126] The intelligent early warning and decision-making module consists of a coupled state assessment unit, an adaptive early warning unit, and a decision support unit. It analyzes the response state of the target subsea pipeline through the scour-vibration coupled sensing results, determines whether there is a risk to pipeline operation, and provides risk warning prompts and intervention measures suggestions.

[0127] In another aspect, the present invention provides a computer-readable storage medium including a subsea pipeline status early warning method program based on scour-vibration coupling sensing. When the subsea pipeline status early warning method program based on scour-vibration coupling sensing is executed by a processor, it implements the steps of the subsea pipeline status early warning method based on scour-vibration coupling sensing as described in any of the preceding claims.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0129] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0130] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0131] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0132] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0133] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for early warning of the condition of subsea pipelines based on scour-vibration coupled sensing, characterized in that, include: Historical marine environmental information of the target subsea pipeline deployment area is obtained, and a monitoring sensor array is set up based on the historical marine environmental information to monitor the target subsea pipeline area, obtain multi-source monitoring data, and perform preprocessing to generate a multi-source monitoring dataset. A multiphysics coupling analysis model is constructed, and the multi-source monitoring dataset is input into the multiphysics coupling analysis model to analyze the response status of the target subsea pipeline in the current monitoring period, so as to obtain pipeline response simulation information. Based on the pipeline response simulation information, the influence of segment resonance caused by seabed scouring between different subsea pipeline sections is analyzed and resonance-derived features are generated. Combined with the multi-source monitoring dataset, pipeline multimodal fusion features are constructed. A subsea pipeline status analysis model is constructed, and the multimodal fusion feature analysis of the pipeline is imported to analyze the response status of the target subsea pipeline in the future time period. The response status analysis information is obtained, and it is determined whether there is any pipeline operation risk and a risk warning is issued. Based on the multimodal fusion features of the pipeline, resonance-derived features are extracted. The fatigue life of each section of the target subsea pipeline is analyzed through the resonance-derived features, and maintenance suggestions are generated and pushed out for easily fatigued section nodes. Specifically, the step of analyzing the segment resonance effects caused by seabed scouring between different subsea pipeline sections based on the pipeline response simulation information and generating resonance-derived features includes: The pipeline response simulation information is obtained, and the target subsea pipeline is divided into several sections according to the deployment scheme of the monitoring sensor array on the target subsea pipeline. Based on the section division results, section labels are generated and associated with the pipeline response simulation information. Based on the associated pipeline response simulation information, the vortex shedding frequency and flow direction time history of each sub-section are calculated. The vortex shedding frequency is calculated from the three-dimensional unsteady turbulent field around the pipeline. Vibration mode parameters are extracted from the vibration response data of each pipeline section by the empirical mode decomposition method to generate a vibration mode feature set and perform two-level resonance coupling analysis. The natural frequency sequence and local vortex shedding frequency sequence corresponding to each sub-segment are extracted through the vibration mode feature set. The natural frequency sequence and local vortex shedding frequency sequence of each sub-segment are compared, and the instantaneous frequency ratio and locking coefficient of the corresponding sub-segment are calculated. Based on the instantaneous frequency ratio and locking coefficient, the critical section of the target subsea pipeline with isolated vortex-induced vibration is identified. The span length and vibration amplitude of the critical section are extracted by pipeline response simulation information, and weighted and fused with the locking coefficient to generate the section resonance confidence, and finally the first resonance derived feature is generated. Based on the vibration mode feature set, the vibration transfer function matrix between all sections is constructed using the transfer path analysis method. The coherence function and phase spectrum between the vibration responses of adjacent and spaced sections are calculated to quantify the transmission path and directional characteristics of vibration energy in the subsea pipeline system. Subsequently, the fluid field response simulation characteristics of each sub-section are obtained through pipeline response simulation information. The heterogeneous excitation caused by fluid loads with opposite flow directions or velocity differences greater than a preset threshold in adjacent sections is analyzed, and a heterogeneous excitation distribution feature vector is generated. By calculating the vibration transfer function matrix to quantify the cancellation or enhancement effect of out-of-phase excitation on the overall vibration mode of the pipeline, the segment coupling coefficient characterizing the intensity of dynamic interaction between segments, the attenuation coefficient characterizing the vibration energy transfer efficiency, and the modal participation factor reflecting the contribution of each segment in the overall pipeline system vibration mode are generated, thus obtaining the second resonance derived characteristics.

2. The method for early warning of subsea pipeline status based on scour-vibration coupling sensing according to claim 1, characterized in that, The process of acquiring historical marine environmental information of the target subsea pipeline deployment area, setting up a monitoring sensor array based on the historical marine environmental information, monitoring the target subsea pipeline area to acquire multi-source monitoring data, and preprocessing the data to generate a multi-source monitoring dataset specifically includes: Historical marine environmental information of the target submarine pipeline layout area is obtained by big data retrieval technology. Feature extraction is performed on the historical marine environmental information to obtain time series data of flow velocity and direction, wave statistical characteristic parameters and seabed geological exploration data, so as to obtain historical marine environmental characteristic information. Based on historical marine environmental characteristics, spatial gridding is performed, and the pipeline is divided into different environmental units using the K-means clustering algorithm. Each unit is combined and classified according to the substrate type and flow field characteristics to form an environmental zoning map with different environmental characteristics. Principal component analysis is introduced to evaluate the region type of each sub-region in the environmental zoning map. By calculating the principal component scores of the corresponding environmental features of each sub-region, environmental features within the preset principal component score range are selected as key environmental features. The region type of each sub-region is assessed by identifying the key environmental characteristics within each sub-region. Based on the results of the region type assessment, a monitoring sensor setting scheme is set for different sub-regions. Based on the set sensor setting scheme, a monitoring sensor array is set in the target subsea pipeline deployment area. The target subsea pipeline area is monitored by a set monitoring sensor array to obtain multi-source monitoring data, which is then imported into an adaptive Kalman filter to eliminate noise interference caused by wave disturbances. The energy features of vortex-induced vibration were extracted by 6-level wavelet packet decomposition using the Db4 wavelet basis function. High-frequency noise components were eliminated by empirical mode decomposition. At the same time, the instantaneous rate of change of the pipe bottom gap and the standard deviation of the velocity gradient were calculated. The monitoring data of various types were further spatiotemporally aligned to finally generate a multi-source monitoring dataset.

3. The method for early warning of subsea pipeline status based on scour-vibration coupling sensing according to claim 1, characterized in that, The construction of a multiphysics coupling analysis model involves inputting the multi-source monitoring dataset into the model to analyze the response status of the target subsea pipeline in the current monitoring period, thereby obtaining pipeline response simulation information. Specifically, this includes: The pipeline structure information and seabed geographic information of the target subsea pipeline are obtained, and a multiphysics coupling analysis model is constructed based on the three-dimensional finite element analysis method. The multiphysics coupling analysis model is constructed by cascading coupling of a fluid dynamics sub-model, a pipeline structural mechanics sub-model, and a sediment dynamics model. Based on the pipeline structure information and seabed geographic information, boundary conditions and initial computational domain are set, and material properties and constitutive relations are set; multi-source monitoring dataset is obtained, and the multi-source monitoring dataset is input into the multi-physics coupling analysis model to analyze the response status of the target seabed pipeline in the seabed environment during the current monitoring period; In the fluid dynamics sub-model, the Navier-Stokes equations are solved using the large eddy simulation method to simulate the complex three-dimensional unsteady turbulent flow field around the pipe, calculate the pulsating pressure distribution, vortex shedding frequency and vortex intensity at various points on the pipe surface, and finally output fluid response simulation information with spatiotemporal attributes. The fluid load field is extracted from the fluid response simulation information and used as the input to the pipeline structure mechanical sub-model. The Newmark-β integral method is used to solve the pipeline dynamic equations for transient dynamic analysis. The vibration displacement, acceleration and equivalent stress distribution of the target subsea pipeline at each time step are calculated, and finally the pipeline structure response simulation information is output. The near-bottom velocity field and bed shear stress are extracted from the fluid response simulation information and input into the sediment dynamics model. The changes in seabed scouring and deposition around the pipeline are analyzed by the modified sediment initiation probability formula and sediment transport flux formula. The continuous evolution process of the development depth, width and morphology of scour pits is dynamically simulated, and finally the sediment response simulation information is output. Subsequently, the simulation information of pipeline structure response and sediment response is imported into the fluid dynamics sub-model to update the fluid computation domain. The simulation process is repeated iteratively until the response status of the subsea pipeline at all time steps of the current monitoring period is simulated. Then, time-series alignment and correlation are performed to generate pipeline response simulation information containing spatiotemporal evolution characteristics.

4. The method for early warning of subsea pipeline status based on scour-vibration coupling sensing according to claim 1, characterized in that, The construction of pipeline multimodal fusion features by combining the multi-source monitoring dataset specifically includes: Acquire the first resonance-derived features, the second resonance-derived features, pipeline response simulation information, and multi-source monitoring datasets; The simulated vibration data sequence is extracted from the pipeline response simulation information, and the measured vibration data sequence is extracted from the multi-source monitoring dataset. The dynamic time warping algorithm is introduced to align the simulated vibration data sequence and the measured vibration data sequence and mark the difference sequence segments to generate a fused vibration data sequence. The fused vibration data sequence, the first resonance derived feature, the second resonance derived feature, the pipeline response simulation information, and the multi-source monitoring dataset are input into a multimodal feature fusion network based on a multilayer perceptron architecture for multimodal feature fusion. Feature interaction and compression are performed through a multilayer perceptron. The temporal dependency and spatial correlation between features are captured by a gated recurrent unit. An attention mechanism is introduced to dynamically calculate attention weights through the gated recurrent unit network for adaptive fusion, ultimately generating pipeline multimodal fusion features.

5. The method for early warning of subsea pipeline status based on scour-vibration coupling sensing according to claim 1, characterized in that, The process of constructing a subsea pipeline status analysis model, importing the pipeline's multimodal fusion feature analysis to analyze the target subsea pipeline's response status within a future time period, obtaining response status analysis information, determining whether there are pipeline operational risks, and issuing risk warnings specifically includes: A state analysis model for subsea pipelines is constructed based on a Bayesian inference network. A bidirectional long short-term memory network is used to extract the temporal dependencies of multimodal features as the observation layer. The pipeline vibration state, stress state and geometric state form a multidimensional hidden state layer. Based on the pre-set training dataset, the expectation-maximization algorithm is used to learn the state transition probability matrix and the observation probability distribution. The state transition probability matrix is ​​estimated by the Metropolis-Hastings algorithm in Markov chain Monte Carlo sampling to capture the nonlinear and uncertain characteristics of pipeline state evolution, and finally obtain a subsea pipeline state analysis model that meets the expectations. The pipeline multimodal fusion features are obtained, and the pipeline multimodal fusion features are input into the subsea pipeline state analysis model to obtain the transient prior distribution and probability likelihood function. The posterior probability distribution is calculated by a stochastic variational inference algorithm. Monte Carlo sampling is performed in a Bayesian inference network based on the posterior probability distribution to infer the response state of the target subsea pipeline in the future time period, obtain the probability distribution of the response state of each pipeline in the future time step, and generate response state analysis information. Based on the response state analysis information, a number of Markov chains are generated using a Markov algorithm. The Markov chains represent the state transition paths of the target subsea pipeline in the future time period. A risk assessment space is generated through the generated Markov chains. In the risk assessment space, the confidence level of each Markov chain is calculated using the Viterbi algorithm to generate a confidence ranking table. The Markov chain with the highest confidence level is selected as the target state transition sequence based on the confidence ranking table. The sequence is then compared with preset risk assessment thresholds to generate a risk assessment report and provide risk warning prompts.

6. The method for early warning of subsea pipeline status based on scour-vibration coupling sensing according to claim 1, characterized in that, The process of extracting resonance-derived features based on the multimodal fusion features of the pipeline, analyzing the fatigue life of each section of the target subsea pipeline using these features, and identifying fatigue-prone sections and nodes to generate and push maintenance recommendations specifically includes: The pipeline multimodal fusion features are obtained, and the first resonance derived features and the second resonance derived features are obtained through the pipeline multimodal fusion features. The dynamic stress amplitude time series, resonance duration, segment coupling coefficient, vibration energy transfer efficiency and modal participation factor are extracted to obtain the resonance derived feature set. Based on the resonance-derived feature set, the dynamic stress time history of each sub-segment is cyclically counted using the rainflow counting algorithm to obtain the initial stress cyclic statistical matrix of each sub-segment. A vibration transmission weight matrix is ​​constructed based on the segment coupling coefficient, and matrix operations are performed with the initial stress cycle statistics matrix to correct the vibration transmission influence between adjacent segments and generate a regularized stress cycle distribution. The SN curve characteristics of the target subsea pipeline material are obtained, and the fatigue damage degree of each sub-segment in the current monitoring cycle is calculated using the Miner linear cumulative damage theory in combination with the regularized stress cycle distribution, generating a cumulative damage degree distribution map. The historical operating time and load of the pipeline are obtained, and the cumulative damage distribution map is input into the pre-trained remaining life prediction model to analyze the remaining fatigue life of each sub-section and obtain the remaining life probability distribution map. The remaining life probability distribution map is used to identify fatigue-prone segment nodes with lifespans below a preset threshold using spatial clustering analysis. Maintenance priorities are then set based on the corresponding remaining lifespans, and a maintenance priority ranking table is generated. Meteorological data of the target subsea pipeline deployment area is obtained, and the subsea pipeline sections that need maintenance are extracted through the maintenance priority ranking table. The maintenance time window is generated in combination with the meteorological data to generate a pipeline maintenance auxiliary report and push it out.

7. A subsea pipeline condition early warning system based on scour-vibration coupling sensing, characterized in that, To implement the subsea pipeline status early warning method based on scour-vibration coupling sensing as described in any one of claims 1-6, the system includes a multi-source intelligent sensing module, a pipeline response multi-physics field coupling analysis module, and an intelligent early warning and decision-making module; The multi-source intelligent sensing module consists of a vibration sensor array, an acoustic Doppler current profiler, an integrated ranging unit, and a data preprocessing and edge computing unit, and is used to monitor the target subsea pipeline. The pipeline response multiphysics coupling analysis module is composed of a multiphysics coupling analysis model, which performs scour-vibration coupling sensing on the target subsea pipeline based on real-time monitoring data processed by data preprocessing and edge computing units. The intelligent early warning and decision-making module consists of a coupled state assessment unit, an adaptive early warning unit, and a decision support unit. It analyzes the response state of the target subsea pipeline through the scour-vibration coupled sensing results, determines whether there is a risk to pipeline operation, and provides risk warning prompts and intervention measures suggestions.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a subsea pipeline status early warning method program based on scour-vibration coupling sensing. When the subsea pipeline status early warning method program based on scour-vibration coupling sensing is executed by a processor, it implements the steps of the subsea pipeline status early warning method based on scour-vibration coupling sensing as described in any one of claims 1 to 6.

Images