A multi-source sensing integrated development monitoring method for underwater immersed tube sinking

Abstract

The application discloses a kind of multi-source sensing integrated development monitoring methods for underwater immersed tube sinking, belong to multi-source sensing monitoring technical field.The method obtains the targeted data of the whole process of immersed tube sinking, carries out feature extraction by first fusion model data, obtains initial parameter set of sinking state, inputs initial parameter set of sinking state into second self-adaptive decoupling model, obtains sinking control adaptation parameter set, inputs sinking control adaptation parameter set into third intelligent early warning model, generates collaborative control instruction set based on the dynamic control instruction matrix, and issues collaborative control instruction set to executing mechanism through underwater internet of things edge gateway.The application realizes multi-dimensional data accurate collection of immersed tube sinking, fully captures the dynamic change of attitude and structure, achieves efficient analysis of multi-field coupling parameters, supports precise adaptation of control instruction to actual working conditions, guarantees monitoring and control closed-loop response in underwater complex environment, and improves sinking operation safety and accuracy.

Description

Technical Field

[0001] This invention relates to the field of multi-source sensing monitoring technology, specifically to a multi-source sensing integrated development and monitoring method for underwater immersed tube placement. Background Technology

[0002] With the rapid development of cross-sea passage projects, underwater immersed tunnels are becoming increasingly widely used due to their advantages such as high traffic capacity and minimal impact on the surrounding environment. Immersion of the tunnel section is a core step in tunnel construction, and the working environment is complex and variable. Precise monitoring of multiple dimensions, including the tunnel section's attitude, structural deformation, and environmental loads, is necessary to ensure immersion accuracy and structural safety. However, existing immersed tunnel monitoring technologies still have many shortcomings and cannot meet the precise control requirements of complex underwater environments.

[0003] Chinese patent (publication number: CN114737991A) discloses a safety monitoring method for the final joint structure of an underwater immersed tunnel. This method uses sensors such as fiber optic grating arrays and vibration pickups arranged at the final joint to monitor aspects such as water leakage, dynamic load, and structural health, effectively filling the gap in the identification of local risks during the operation and maintenance phase of the immersed tunnel. However, this patent focuses on the operation and maintenance monitoring after the immersed tube docking is completed, and does not involve core requirements such as dynamic attitude adjustment and multi-physics field coupling parameter analysis during the immersion process. It cannot provide support for real-time control of the immersion process, and it does not build a closed-loop system for data fusion and control command generation, making it difficult to adapt to the dynamic requirements of immersion operations.

[0004] To address the dynamic needs of immersed tunnel monitoring and the structural modeling deficiencies in existing patents, Chinese patent (publication number: CN119577901A) proposes an automated design and layout method for monitoring in tunnel sections of rail transit shield tunnels. This method, based on AutoCAD secondary development, achieves automated layout of monitoring points and monitoring sections, improving monitoring design efficiency. However, this technology is applicable to surface and surrounding environment monitoring in tunnel sections, without considering the special characteristics of the underwater environment. Furthermore, it only focuses on optimizing the layout of monitoring points and lacks a coupled analysis and intelligent control model for multi-source monitoring data. It cannot solve complex problems such as fluid-structure interaction and structural deformation coordination during the underwater immersed tunnel immersion process, making it difficult to directly apply to the accurate monitoring and dynamic control of immersed tunnel immersion.

[0005] Therefore, there is an urgent need for a monitoring method for immersed tube placement that can adapt to complex underwater environments, realize multi-source data coupling and analysis, dynamic control command generation, and closed-loop control, in order to solve the core problems of existing technologies such as insufficient monitoring targeting during the placement process and poor connection between data processing and control, and ensure the safety and accuracy of immersed tube placement operations. Summary of the Invention

[0006] Based on the above-mentioned technical problems, this application discloses a multi-source sensor integrated development and monitoring method for underwater immersed tube placement, specifically including: acquiring targeted data of the entire process of immersed tube placement, wherein the targeted data includes first core data and second core data;

[0007] The first fusion model is used to extract features from the first core data and the second core data to obtain the initial parameter set of the immersion state. The first fusion model is constructed based on the synergistic mechanism of the fluid-structure coupling dynamic response and structural deformation of the immersed tube.

[0008] The initial parameter set of the sinking state is input into the second adaptive decoupling model to obtain the sinking control adaptation parameter set. The second adaptive decoupling model is constructed based on the adaptive orthogonal stripping algorithm of the multi-physics coupling parameters of the sinking process.

[0009] The set of sinking control adaptation parameters is input into the third intelligent early warning model. Combined with the preset sinking accuracy and structural safety threshold, a dynamic control command matrix is ​​generated. The third intelligent early warning model is constructed based on the real-time fusion assessment of sinking environmental risk factors and structural health status.

[0010] Based on the aforementioned dynamic control instruction matrix, a collaborative control instruction set is generated;

[0011] The collaborative control command set is sent to the actuators of the immersion equipment via an underwater IoT edge gateway. The actuators include ballast water tank regulating valves, attitude adjustment thrusters, and docking hydraulic fine-tuning devices.

[0012] Preferably, the targeted data includes first core data and second core data. The first core data is multi-dimensional vibration spectrum data of the immersed tube attitude. The acquisition process of the multi-dimensional vibration spectrum data of the immersed tube attitude includes: symmetrically arranging N sets of triaxial velocity sensors at the top, bottom and sides of the immersed tube, with the sampling time synchronized with the entire process of immersed tube placement; collecting vibration signals of the immersed tube in the three attitude directions of roll, pitch and sway through the sensors, denoising the vibration signals using a wavelet threshold denoising algorithm, and performing Fourier transform on the denoised vibration signals to obtain the multi-dimensional vibration spectrum data of the immersed tube attitude. The vibration spectrum data includes the vibration frequency, amplitude and phase information of each attitude direction.

[0013] Preferably, the second core data is the three-dimensional deformation field distribution data of the pipe section joint, and the process of obtaining the three-dimensional deformation field distribution data of the pipe section joint includes: uniformly arranging the data around the circumference of the immersed tunnel joint. A laser displacement sensor and a strain gauge are used. The laser displacement sensor is used to collect the normal displacement data of the joint, and the strain gauge is used to collect the tangential strain data of the joint. Based on the data collected by the laser displacement sensor and the strain gauge, a three-dimensional reconstruction algorithm is used to construct the three-dimensional deformation field of the joint. The constructed three-dimensional deformation field is smoothed to remove abnormal deformation points, and the distribution data of the three-dimensional deformation field of the pipe section joint is obtained.

[0014] Preferably, the construction process of the first fusion model includes: determining the coupling correlation factor between the first core data and the second core data based on the synergistic mechanism of the fluid-structure interaction dynamic response and structural deformation of the immersed tube. Establish the coupled dynamic equations:

[0015]

[0016] in, For the density of the immersed tube structure, Let be the displacement vector of the immersed tube structure. For time parameters, This is the fluid viscosity damping coefficient. Here is the stiffness matrix of the immersed tube structure. Let be the force vector of fluid-structure interaction, where Water pressure, It is the synergistic force of structural deformation.

[0017] Preferably, based on the coupled dynamic equations, a feature extraction network is constructed. This network includes convolutional layers, an attention mechanism layer, and fully connected layers. The convolutional layers extract the vibration spectrum features of the first core data and the deformation field features of the second core data. The attention mechanism layer assigns weights to the two types of features, with the weight assignment coefficients determined by the coupling correlation factor. Adaptive adjustment: The fully connected layer is used to map the weighted fused features to the initial parameter set of the sinking state.

[0018] Preferably, the construction process of the first fusion model includes: determining the coupling correlation factor between the first core data and the second core data based on the synergistic mechanism of the fluid-structure interaction dynamic response and structural deformation of the immersed tube. The coupled dynamic equations are established as shown in the following equation:

[0019]

[0020] in, For the density of the immersed tube structure, Let be the displacement vector of the immersed tube structure. For time parameters, This is the fluid viscosity damping coefficient. Here is the stiffness matrix of the immersed tube structure. Let be the force vector of fluid-structure interaction, where Water pressure, The synergistic force of structural deformation;

[0021] Based on the coupled dynamic equations, a feature extraction network is constructed. This network includes convolutional layers, an attention mechanism layer, and a fully connected layer. The convolutional layers extract the vibration spectrum features of the first core data and the deformation field features of the second core data. The attention mechanism layer assigns weights to the two types of features, with the weight assignment coefficients determined by the coupling correlation factor. Adaptive adjustment: The fully connected layer is used to map the weighted fused features to the initial parameter set of the sinking state.

[0022] Preferably, the construction process of the second adaptive decoupling model includes: determining the orthogonal basis vector set of the multi-physics coupling parameters based on the adaptive orthogonal stripping algorithm of the multi-physics coupling parameters during the sinking process. ,in Let the dimensions of the coupling parameters be defined; construct an adaptive decoupling objective function:

[0023]

[0024] in, No. A vector of coupling parameters for each physical field. For the first The projection matrix of each physical field. They are orthogonal basis vectors. The regularization coefficient is used to avoid model overfitting. The adaptive decoupling objective function is solved by gradient descent algorithm to obtain the orthogonal stripping coefficients of each physical field coupling parameter. Based on the orthogonal stripping coefficients, a second adaptive decoupling model is constructed. The model is used to strip the coupling parameters in the initial parameter set of the immersion state into independent single physical field parameters to obtain the immersion control adaptation parameter set.

[0025] Preferably, the set of sinking control adaptation parameters includes ballast adjustment parameters, attitude adjustment parameters, and docking fine-tuning parameters. The ballast adjustment parameters are quantitative parameters that characterize the dynamic balance between buoyancy and gravity of the immersed tube. The attitude adjustment parameters are quantitative parameters that characterize the correction of the roll, pitch, and heave angles of the immersed tube. The docking fine-tuning parameters are quantitative parameters that characterize the three-dimensional deformation compensation of the docking joint. Each parameter is obtained by mapping the independent single-physics field parameters output by the second adaptive decoupling model.

[0026] Preferably, the construction process of the third intelligent early warning model includes: screening environmental risk factors and structural health status assessment indicators, wherein the environmental risk factors include water flow velocity, wave height, and water density, and the structural health status assessment indicators include equivalent stress of the immersed tube structure and maximum deformation of the joint; and establishing a real-time fusion assessment function for the environmental risk factors and structural health status assessment indicators, as shown in the following formula:

[0027]

[0028] in, for The comprehensive evaluation value of the sinking status at any given moment. for The comprehensive quantitative value of environmental risk factors at any given time. for The comprehensive quantitative value of the structural health status at any given time. , These are the weighting coefficients for environmental risk factors and structural health status assessment indicators, respectively, and they satisfy the following conditions: + =1, , The system dynamically adjusts according to the sinking conditions; based on the comprehensive evaluation function and combined with the preset sinking accuracy threshold and structural safety threshold, it constructs an early warning judgment logic and generates control instructions to form a third intelligent early warning model. The model is used to generate a dynamic control instruction matrix based on the comparison results between the comprehensive evaluation value and the threshold.

[0029] Preferably, the dynamic control command matrix is ​​a 3×3 matrix, with the row dimension corresponding to the three types of actuators and the column dimension corresponding to the core control parameters of each actuator. Each element in the matrix is ​​a specific quantified command value for a single control parameter of a single actuator. The first row of the dynamic control command matrix M corresponds to the ballast water tank regulating valve. To adjust the valve opening, For regulating valve adjustment rate, The first row represents the feedback threshold for the regulating valve action; the second row corresponds to the attitude adjustment thruster. To output thrust to the thruster For the working direction of the thruster, The third row corresponds to the intermittent adjustment frequency of the thruster; it also corresponds to the docking hydraulic fine-tuning device. For fine-tuning of hydraulic devices, For fine-tuning the speed of the hydraulic device, This refers to the operating pressure threshold of the hydraulic device.

[0030] Preferably, the underwater IoT edge gateway serves as a communication relay unit between the slumbering monitoring system and the actuator. It employs an IP68-level waterproof sealing design, supports multi-protocol compatibility and conversion, and enables encrypted transmission of collaborative control command sets and real-time feedback of actuator operating data. The edge gateway has a built-in local data cache. When the underwater communication link is interrupted, the data cache automatically stores the collaborative control command sets to be sent, and automatically resends them in the command generation order after the communication link is restored. After receiving the collaborative control command sets and executing actions, the actuator collects its own operating status parameters in real time, including the actuator's workload, action execution accuracy, and equipment operating temperature. These parameters are then fed back to the slumbering monitoring system through the underwater IoT edge gateway, forming a closed-loop control link of sensor acquisition, model calculation, command issuance, and status feedback.

[0031] Compared with the prior art, the technical solution of this application has the following technical effects:

[0032] This invention constructs a first fusion model based on the synergistic mechanism of fluid-structure interaction dynamic response and structural deformation of immersed tubes. By adaptively allocating feature weights for vibration spectrum data and three-dimensional deformation field data through an attention mechanism layer and coupling correlation factors, the model achieves deep fusion of the two types of core data. This model effectively overcomes the limitations of isolated processing of multi-source data, accurately extracts core parameters of the immersion state, comprehensively reflects the dynamic correlation between the attitude of the immersed tube and structural deformation, provides reliable data support for subsequent control, and significantly improves the comprehensiveness and accuracy of state perception during the immersion process.

[0033] This invention constructs a second adaptive decoupling model based on an adaptive orthogonal stripping algorithm for multi-physics coupling parameters during the immersion process. By constructing orthogonal basis vector sets and solving the objective function with regularization constraints, the model achieves precise stripping of multi-physics coupling parameters. This model effectively solves the technical challenge of cross-interference between multiple field parameters such as water flow, waves, and structural internal forces, transforming complex coupling parameters into independent single-physics parameters, which are then mapped to targeted control parameters such as ballast adjustment, attitude adjustment, and docking fine-tuning. This ensures that the control commands are highly adapted to the actual needs of immersion, improving the accuracy and effectiveness of control.

[0034] This invention dynamically adjusts the weight coefficients of two types of evaluation indicators through a third intelligent early warning model, adapting to the core requirements of different operational stages such as water entry, uniform sinking, and docking. Combined with the early warning judgment logic constructed by preset thresholds, it can generate a dynamic control command matrix to achieve real-time response and stable control during the sinking process. This mechanism effectively avoids the limitations of traditional fixed threshold early warning, preventing attitude deviation caused by environmental interference and avoiding safety risks caused by excessive structural stress, thus comprehensively ensuring the operational safety and docking accuracy of immersed tube sinking.

[0035] This invention constructs a closed-loop control link through an underwater IoT edge gateway, encompassing sensor acquisition, model computation, command issuance, and status feedback. The gateway's multi-protocol compatibility, encrypted transmission, and breakpoint resume functionality ensure stable transmission of control commands in complex underwater environments. The actuators provide real-time feedback of operating status parameters, enabling dynamic correction of the control effect. This effectively solves engineering pain points such as unstable underwater communication and command loss. The entire system forms a complete closed loop from data acquisition to command execution, improving the stability and anti-interference capabilities of the technical solution and providing reliable assurance for practical engineering applications.

[0036] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0037] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0039] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0040] Figure 1 : Overall flowchart of the multi-source sensor integrated development and monitoring method for underwater immersed tube placement;

[0041] Figure 2 : Schematic diagram of the first fusion model architecture;

[0042] Figure 3 : Schematic diagram of the second adaptive decoupling model architecture;

[0043] Figure 4 : Schematic diagram of the third intelligent early warning model architecture;

[0044] Figure 5 : Multi-dimensional vibration spectrum of underwater immersed tube attitude;

[0045] Figure 6Iterative curves of the second adaptive decoupling model;

[0046] Figure 7 : A graph showing the change in the comprehensive evaluation value of the third intelligent early warning model;

[0047] Figure 8 Comparison chart of the execution accuracy of actuator actions. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0049] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0050] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0051] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0052] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0053] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0054] Example 1

[0055] This embodiment mainly describes a multi-source sensor integration development and monitoring method for underwater immersed tube placement, such as... Figure 1 As shown, this specifically includes: acquiring target data for the entire process of immersed tube placement, wherein the target data includes first core data and second core data;

[0056] The first fusion model is used to extract features from the first core data and the second core data to obtain the initial parameter set of the immersion state. The first fusion model is constructed based on the synergistic mechanism of the fluid-structure coupling dynamic response and structural deformation of the immersed tube.

[0057] The initial parameter set of the sinking state is input into the second adaptive decoupling model to obtain the sinking control adaptation parameter set. The second adaptive decoupling model is constructed based on the adaptive orthogonal stripping algorithm of the multi-physics coupling parameters of the sinking process.

[0058] The set of sinking control adaptation parameters is input into the third intelligent early warning model. Combined with the preset sinking accuracy and structural safety threshold, a dynamic control command matrix is ​​generated. The third intelligent early warning model is constructed based on the real-time fusion assessment of sinking environmental risk factors and structural health status.

[0059] Based on the aforementioned dynamic control instruction matrix, a collaborative control instruction set is generated;

[0060] The collaborative control command set is sent to the actuators of the immersion equipment via an underwater IoT edge gateway. The actuators include ballast water tank regulating valves, attitude adjustment thrusters, and docking hydraulic fine-tuning devices.

[0061] Furthermore, the first core data acquisition of the immersed tube attitude multi-dimensional vibration spectrum data includes sensor arrangement, data acquisition parameter setting and data preprocessing;

[0062] Sensor arrangement: A total of 8 sets (N=8) of triaxial accelerometers (range ±5g, accuracy 0.001g) and velocity sensors (range 0-5m / s, accuracy 0.001m / s) are arranged at the top center, bottom center and symmetrical positions on both sides (1.5m apart on each side). The sensors are fixed to the outer wall of the immersed tube by waterproof brackets, and the cables are introduced into the data acquisition box inside the immersed tube through sealed sleeves.

[0063] Data acquisition parameter settings: The sampling frequency is set to 300Hz, and the sampling duration is completely synchronized with the sinking operation duration (from the sinking of the tube into the water to the completion of docking). During the acquisition process, the temperature compensation module built into the sensor eliminates the influence of underwater temperature (0-25℃) on the acquisition accuracy.

[0064] Data preprocessing: Denoising is performed using a wavelet thresholding algorithm. The denoising formula is:

[0065]

[0066]

[0067]

[0068] in, These are the original coefficients after wavelet transform. These are the denoised wavelet coefficients. The standard deviation of noise. The number of sampling points for the vibration signal is given; first, the collected vibration signal is decomposed using db4 wavelet decomposition (with 5 decomposition levels) to obtain the wavelet coefficients. Threshold calculation is based on the formula Calculate the standard deviation of noise Then determine the threshold. ,in The number of sampling points per sampling is set to 3000. The coefficients are reconstructed by processing the decomposed wavelet coefficients according to the wavelet threshold denoising formula, retaining the effective signal coefficients and suppressing the noise coefficients. The denoised vibration signal is then reconstructed by inverse wavelet transform. The spectrum conversion performs a fast Fourier transform (FFT) on the denoised vibration signal to obtain the vibration spectrum data in the three attitude directions of roll, pitch, and heave. The characteristic frequencies, peak amplitudes, and phase information in the range of 1-100Hz in each direction are extracted to form the first core dataset.

[0069] Furthermore, the acquisition of three-dimensional deformation field distribution data of the second core data pipe section joint includes sensor arrangement and three-dimensional deformation construction;

[0070] Sensor arrangement: 12 sets (M=12) of laser displacement sensors (range 0-50mm, accuracy 0.005mm) and strain gauges (range ±2000mm) are evenly arranged around the circumference of the immersed tube joint. Precision 1 The laser displacement sensor is arranged perpendicular to the surface of the joint to collect normal displacement data, and the strain gauge is attached along the tangential direction of the joint to collect tangential strain data.

[0071] Three-dimensional deformation field construction: Before data acquisition and calibration, the laser displacement sensor was calibrated using a standard displacement stage (accuracy 0.001mm), and the strain gauge was zero-point calibrated using a strain calibrator to eliminate initial errors. Deformation vector calculation: Based on the normal displacement data acquired by the laser displacement sensor and the tangential strain data acquired by the strain gauge, combined with the geometric dimensions (diameter, wall thickness) of the immersed tube joint, the expression for the three-dimensional deformation field is shown in the following equation:

[0072]

[0073] in, for Always attach coordinates to the seam The three-dimensional deformation vector at that location. These are the coordinates along... Deformation in three directions; a three-dimensional deformation field of the joint is constructed using the 3D reconstruction module of ANSYS software. Data optimization employs a Gaussian filtering algorithm to smooth the constructed three-dimensional deformation field. The filtering window size is set to 3×3 to remove abnormal deformation points caused by sensor errors (points with deformation exceeding ±5mm are judged as abnormal points). Finally, the three-dimensional deformation field distribution data of the pipe section joint is obtained, forming the second core dataset.

[0074] Furthermore, such as Figure 2 As shown, the construction and operation of the first fusion model includes determining the basic parameters of the model and building the feature extraction network;

[0075] Determining the basic parameters of the model: Coupling correlation factor Calculation: Based on the synergistic mechanism of fluid-structure interaction dynamics and structural deformation of immersed tubes, numerical simulation was used to determine... The initial value is 0.65. The dynamic adjustment range is 0.4-0.8, and the adjustment is based on the fluid-structure interaction force. Cooperative forces with structural deformation ratio, ; Coupled dynamic equation parameter settings: immersed tube structure density Fluid viscosity damping coefficient Stiffness matrix of immersed tube structure The water pressure was obtained through ANSYS finite element simulation (a finite element model was constructed based on the concrete material parameters and structural dimensions of the immersed tube, with a mesh generation accuracy of 0.1m). The pressure data is collected in real time by pressure sensors installed on the outer wall of the immersed tube (range 0-10MPa, accuracy 0.01MPa).

[0076] Network structure: The convolutional layer contains two convolutional blocks (first convolutional block: 3×3 kernel size, 64 kernels, stride 1, ReLU activation function; second convolutional block: 3×3 kernel size, 128 kernels, stride 1, ReLU activation function), used to extract vibrational spectrum features (frequency, amplitude, phase) and deformation field features (3D deformation, deformation gradient); Attention mechanism layer: based on coupling correlation factor Weights are assigned to the two types of features, with the vibration spectrum feature having the following weight: The characteristic weight of the deformation field is 1- It achieves adaptive fusion of two types of core data; the fully connected layer contains two hidden layers (with 256 and 128 neurons respectively) and one output layer (with 32 neurons), which maps the fused features to an initial parameter set for the submerged state. The parameter set includes 32 dimensions of initial parameters such as the submerged tube attitude angle, vibration intensity, docking seam deformation, and fluid-structure interaction force.

[0077] Furthermore, such as Figure 3 As shown, the construction and operation of the second adaptive decoupling model includes determining the orthogonal basis vector set, solving the adaptive decoupling objective function, and generating the adjustment and adaptation parameter set;

[0078] The orthogonal basis vector set is determined by an adaptive orthogonal stripping algorithm based on the coupling parameters of multiple physics fields (flow field, structural field, force field) during the sinking process. The orthogonal basis vector set is constructed through the Gram-Schmidt orthogonalization process. (m=6, i.e., 6 coupled physical field dimensions: water flow force, wave force, immersed tube gravity, buoyancy, structural internal forces, and docking pressure).

[0079] Solving the adaptive decoupling objective function: The objective function is constructed by taking the coupling parameter vector from the initial parameter set of the sinking state. With projection matrix (Constructed using orthogonal basis vector sets, with dimensions 32×6) Substitute into the objective function The regularization coefficient The learning rate is set to 0.01 to avoid model overfitting; the algorithm uses gradient descent to solve the objective function, with a learning rate of 0.001 and 1000 iterations. The difference between the objective function values ​​of two adjacent iterations is used as the learning rate. The iteration stops when the time is right, and the orthogonal stripping coefficients of each physical field coupling parameter are obtained;

[0080] Generation of the set of control and adaptation parameters: The coupling parameters are separated into independent single physical field parameters by the stripping coefficient, and then converted into ballast adjustment parameters (buoyancy and gravity balance deviation value, ballast water injection or discharge rate), attitude adjustment parameters (roll, pitch, and heave angle correction amount, correction rate) and docking fine-tuning parameters (three-dimensional deformation compensation amount of docking joint, compensation speed) through the mapping function (based on empirical data of immersed tube immersion project), forming the set of immersion control and adaptation parameters.

[0081] Furthermore, such as Figure 4 As shown, the construction and operation of the third intelligent early warning model includes the selection and quantification of evaluation indicators, the operation of integrated evaluation functions, and the generation of dynamic control instruction matrices;

[0082] Assessment Indicator Screening and Quantification: Quantification of Environmental Risk Factors: Environmental risk factors include water flow velocity, wave height, and water density. The weights of each factor are set as follows: =0.4、 =0.3、 =0.3; After collecting the raw data of each factor through the corresponding sensor, it is quantified according to the following formula: ,in, This is a real-time water flow velocity measurement value. =2m / s is the baseline value for water flow velocity; This is the real-time wave height data. =1.5m is the baseline value for wave height; This represents the real-time water density data. =1025kg / This serves as a baseline value for water density. , , These are the weighting coefficients for water flow velocity, wave height, and water density, respectively, satisfying... Quantification of structural health status: The structural health status assessment indicators include the equivalent stress of the immersed tunnel structure and the maximum deformation of the joint. The weights of each indicator are set as follows: , After acquiring raw data using stress and deformation sensors, the data is quantified according to the following formula: Where σ is the real-time equivalent stress acquired by the immersed tunnel structure. =20MPa is the allowable stress benchmark value for immersed tube structures; This is the real-time maximum deformation value of the butt joint. =3mm is the reference value for the maximum allowable deformation of the butt joint; , These are the weighting coefficients for the equivalent stress of the immersed tunnel structure and the maximum deformation of the joint, respectively, satisfying... .

[0083] Fusion evaluation function calculation: Weight coefficients are dynamically adjusted according to the sinking conditions. and Submerged tube entry stage , (Focusing on environmental risks), docking phase , (Focusing on structural safety); Comprehensive assessment value calculation: according to the formula Calculate the real-time comprehensive evaluation value, where and The range of values ​​for all values ​​is normalized to [0,1], therefore The value range is [0,1].

[0084] Dynamic control instruction matrix generation: Threshold setting: Preset sinking accuracy threshold =0.3 (normal state), structural safety threshold =0.7 (warning status), when < Maintain the current control parameters when Make fine adjustments as needed. Emergency adjustments are made as needed; Command matrix construction: Based on the comparison results of the comprehensive evaluation value and the threshold, a 3×3 dynamic control command matrix is ​​generated. For example, the matrix in the fine-tuning state during the docking phase is: The first row corresponds to the ballast water tank regulating valve (opening degree 80%, regulating rate 0.2m³ / min, feedback threshold 0.1MPa), the second row corresponds to the attitude adjustment thruster (propulsion force 5kN, working direction 15°, regulating frequency 2Hz), and the third row corresponds to the docking hydraulic fine-tuning device (fine-tuning amount 0.5mm, fine-tuning speed 0.1mm / s, pressure threshold 10MPa).

[0085] Furthermore, the actuators include the ballast water tank regulating valve, the attitude adjustment thruster, and the docking hydraulic fine-tuning device;

[0086] Ballast water tank regulating valve: After receiving instructions, the valve opening is adjusted via an electric actuator (adjustment accuracy ±1%), and the valve opening feedback signal and tank pressure signal are collected in real time and fed back to the edge gateway; Attitude adjustment thruster: According to the thrust and working direction instructions, the output power and jet direction of the thruster are adjusted via the hydraulic system, with an adjustment response time ≤50ms, and the status parameters such as the thruster working current and output thrust are collected in real time; Docking hydraulic fine-tuning device: Based on the fine-tuning amount and fine-tuning speed instructions, the fine-tuning mechanism is driven by the servo hydraulic system, with a fine-tuning accuracy of ±0.01mm, and the hydraulic system pressure, fine-tuning displacement and other data are collected in real time.

[0087] The workflow of an underwater IoT edge gateway includes communication protocol conversion, encrypted data transmission, caching and retransmission, and status feedback forwarding.

[0088] Communication Protocol Conversion: Supports converting the TCP / IP protocol of the upper-level monitoring terminal to the CAN and Modbus protocols compatible with the actuator, achieving instruction format adaptation; Encrypted Data Transmission: Employs the AES256 encryption algorithm to encrypt transmitted control commands and status data, preventing data leakage or tampering caused by underwater electromagnetic interference; Caching and Retransmission: Features a built-in 16GB local cache module, supporting the storage of over 1000 control commands. When the communication link is interrupted (e.g., due to water flow interference causing wireless signal loss), it automatically caches pending commands. Once the link is restored (signal strength ≥ -85dBm), the commands are retransmitted sequentially according to their generation order, ensuring no command loss; Status Feedback Forwarding: Forwards the actuator's operating status parameters (workload, action execution accuracy, equipment operating temperature) to the upper-level monitoring terminal in real time, with a forwarding cycle of 100ms.

[0089] Based on Example 1, to verify the practicality, accuracy, and stability of the multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to the present invention, a standard immersed tube section from a cross-sea immersed tunnel project was selected for on-site testing. The test object was a reinforced concrete immersed tube with a length of 120m, a diameter of 15m, and a wall thickness of 1.2m. The water depth for the placement operation was 60m, the water flow velocity range in the operation area was 0.5-1.8m / s, the wave height was 0.3-1.2m, the water density was 1020-1028kg / m³, the design docking accuracy requirement for the immersed tube was ±20mm, the allowable structural stress was 20MPa, and the maximum allowable deformation of the docking joint was 3mm. Throughout the test, the hardware system and software model were deployed according to the method scheme of the present invention. Data was collected through multi-source sensors, and control commands were generated through calculations by a three-layer core model. Finally, the dynamic control of the immersed tube placement was realized through the actuator, and key data of each stage were recorded simultaneously to complete the effect verification.

[0090] The experiment first verified the accuracy of the core data acquisition module. Multi-dimensional vibration spectrum data acquisition of the immersed tube attitude was performed using eight sets of triaxial accelerometers and velocity sensors. Three-dimensional deformation field distribution data acquisition of the tube section joint was performed using twelve sets of laser displacement sensors and strain gauges. A dataset of 180 minutes was continuously collected throughout the entire operation. Typical data from different operational stages (water entry stage, uniform sinking stage, and docking stage) were selected for analysis. Table 1 shows the verification results of the core data acquisition accuracy, where VSD is vibration spectrum data, DFD is deformation field data, FRE is frequency, AMP is amplitude, DEF is deformation, ME is measurement error, and RE is relative error. The data in the table shows that the frequency measurement errors in the three attitude directions (roll, pitch, and heave) of the vibration spectrum data are all ≤0.05Hz, with a relative error of ≤1.8%; the amplitude measurement error is ≤0.01mm, with a relative error of ≤2.5%; and the deformation field distribution data shows that the deformation measurement errors in the three directions (x, y, and z) are all ≤0.008mm, with a relative error of ≤0.67%. The acquisition accuracy of all data meets the requirements of engineering monitoring, providing a reliable data foundation for subsequent model calculations.

[0091] Table 1. Verification Results of Core Data Acquisition Accuracy

[0092] VSD Roll FRE (Hz) 12.5 12.46 0.04 1.6% VSD Pitch FRE (Hz) 8.3 8.27 0.03 1.2% VSD sag FRE (Hz) 5.7 5.68 0.02 0.88% VSD Roll AMP (mm) 0.8 0.78 0.02 2.5% VSD Pitch AMP (mm) 0.6 0.59 0.01 1.67% VSD AMP (mm) for heave 1.2 1.18 0.02 1.67% DFD x-axis DEF (mm) 1.2 1.208 0.008 0.67% DFD Dy to DEF (mm) 0.9 0.905 0.005 0.56% DFD z-axis DEF (mm) 1.5 1.506 0.006 0.4%

[0093] The raw vibration signals collected during the water entry stage were preprocessed using wavelet threshold denoising and then subjected to spectral analysis to obtain... Figure 5 The multi-dimensional vibration spectrum diagram of the immersed tube attitude contains three sub-plots, corresponding to the vibration spectrum curves in the roll, pitch, and heave directions, respectively. The spectrum curves in each direction are smooth and free of obvious noise interference, clearly presenting the core characteristic frequencies and corresponding peak characteristics of 10-15Hz (roll), 6-10Hz (pitch), and 4-8Hz (heave). This provides high-quality, low-noise basic data support for the subsequent feature extraction and multi-source data coupling analysis of the first fusion model.

[0094] Next, the computational performance and control effect of the three-layer core model were verified. The first fusion model achieved adaptive fusion of two types of core data by coupling dynamic equations and feature extraction networks. The second adaptive decoupling model completed the accurate decoupling of multi-physics coupling parameters based on the adaptive orthogonal stripping algorithm. The third intelligent warning model generated control instructions through a fusion evaluation function with dynamic weight adjustment. Table 2 shows the verification results of the model's computational performance and control effect, where FM is the first fusion model, SDM is the second adaptive decoupling model, EWM is the third intelligent warning model, OT is the operation time, DT is the delay time, AAR is the attitude adjustment accuracy, DPA is the docking precision, and SSR is the structural safety rate. According to the data, the first fusion model has a single operation time of ≤80ms, the second adaptive decoupling model has a single operation time of ≤60ms, the third intelligent early warning model has a single operation time of ≤40ms, the total operation time of the three-layer model is ≤180ms, and the control delay is ≤100ms, which meets the real-time control requirements. The attitude adjustment accuracy of the immersed tube in roll, pitch, and heave is ≤0.3°, and the final docking accuracy reaches ±12mm, which is better than the design requirement of ±20mm. The maximum equivalent stress of the structure during the test is 15.8MPa, which does not exceed the allowable stress of 20MPa. The maximum deformation of the dock joint is 1.2mm, which is less than the allowable value of 3mm, and the structural safety rate reaches 100%.

[0095] Table 2. Verification results of model computational performance and control effect

[0096] FM Single OT (ms) 75 SDM Single OT (ms) 52 EWM Single OT (ms) 38 Overall model system Total OT (ms) 165 Overall model system Regulating DT (ms) 85 Regulation effect Roll AAR (°) 0.25 Regulation effect Pitch AAR (°) 0.22 Regulation effect Drooping AAR (°) 0.18 Regulation effect DPA (mm) ±12 Structural safety Maximum equivalent stress (MPa) 15.8 Structural safety Maximum DEF (mm) 1.2 Structural safety SSR (%) 100

[0097] like Figure 6 As shown, the iterative curve of the second adaptive decoupling model contains two curves, representing the curves without a regularization term and the curve with a regularization coefficient. The iterative curves of the objective function with a value of 0.01 show that the curve without regularization tends to stabilize after 200 iterations, but the objective function value fluctuates significantly. The curve with regularization converges smoothly after 300 iterations, with no significant fluctuation in the objective function value. This fully demonstrates the efficient convergence characteristics of the second adaptive decoupling model based on the adaptive orthogonal stripping algorithm, as well as its stability in the precise stripping of multi-physics coupling parameters, providing algorithmic support for the accurate output of sinking and deployment control adaptation parameters.

[0098] The curve showing the change in the comprehensive evaluation value of the third intelligent early warning model, as shown in the figure. Figure 7As shown in the diagram, the process includes three stages: the submerged tube immersion stage (0-30 min), the uniform immersion stage (30-150 min), and the docking stage (150-180 min). The trend of change, and the preset sinking accuracy threshold. =0.3 and structural safety threshold =0.7, the water entry stage is significantly affected by environmental risk factors. Fluctuating between 0.4 and 0.6, the model focuses on environmental risk regulation; the environment is stable during the uniform settling phase. The value should remain stable between 0.2 and 0.3, maintaining the current control parameters; the weighting of structural health status should be increased during the docking phase. The value is varied between 0.3 and 0.5, and fine-tuning is used to ensure docking accuracy throughout the process. Not exceeding the safety threshold =0.7, no warning trigger, achieving stable control of the sinking process.

[0099] Finally, the stability of the closed-loop control link was verified. The underwater IoT edge gateway adopts an IP68 waterproof design, supports multi-protocol compatibility conversion and AES-256 encrypted transmission. During the test, three communication link interruptions were simulated (each interruption lasting 30 seconds). The edge gateway successfully cached the commands to be sent in each instance, and completed the command resending within 10 seconds after the link was restored. There were no command losses. After the actuator received the command, the opening accuracy of the ballast water tank regulating valve was ±0.5%, the response time of the attitude adjustment thruster was ≤40ms, and the fine-tuning accuracy of the docking hydraulic fine-tuning device was ±0.005mm. Figure 8 The figure shows a comparison of the execution accuracy of the actuators. The three curves in the figure correspond to the execution error changes of the ballast water tank regulating valve, attitude adjustment thruster, and docking hydraulic fine-tuning device, respectively. As can be seen from the figure, the execution error of all actuators is controlled within ±0.02mm, and the error does not accumulate significantly with the increase of the number of adjustments, and remains at a low level, which verifies the stability and reliability of the closed-loop control link.

[0100] Based on the above experimental data and analysis results, the core data acquisition accuracy, model calculation performance, control effect, and closed-loop control stability of the technical solution of this invention all meet the actual engineering requirements. It can effectively realize accurate monitoring and dynamic control of the entire process of underwater immersed tube placement, providing a strong guarantee for the safety and reliability of immersed tube placement operations.

[0101] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A multi-source sensor integrated development and monitoring method for underwater immersed tube placement, characterized in that, include: Acquire targeted data for the entire process of immersed tube placement, the targeted data including first core data and second core data; The first fusion model is used to extract features from the first core data and the second core data to obtain the initial parameter set of the immersion state. The first fusion model is constructed based on the synergistic mechanism of the fluid-structure coupling dynamic response and structural deformation of the immersed tube. The initial parameter set of the sinking state is input into the second adaptive decoupling model to obtain the sinking control adaptation parameter set. The second adaptive decoupling model is constructed based on the adaptive orthogonal stripping algorithm of the multi-physics coupling parameters of the sinking process. The set of sinking control adaptation parameters is input into the third intelligent early warning model. Combined with the preset sinking accuracy and structural safety threshold, a dynamic control command matrix is ​​generated. The third intelligent early warning model is constructed based on the real-time fusion assessment of sinking environmental risk factors and structural health status. Based on the aforementioned dynamic control instruction matrix, a collaborative control instruction set is generated; The collaborative control command set is sent to the actuators of the immersion equipment via an underwater IoT edge gateway. The actuators include ballast water tank regulating valves, attitude adjustment thrusters, and docking hydraulic fine-tuning devices.

2. The multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 1, characterized in that, The target data includes first core data and second core data. The first core data is multi-dimensional vibration spectrum data of the immersed tube attitude. The process of acquiring the multi-dimensional vibration spectrum data of the immersed tube attitude includes: symmetrically arranging N sets of triaxial velocity sensors at the top, bottom and sides of the immersed tube, and synchronizing the sampling time with the entire process of the immersed tube sinking. Vibration signals of the immersed tube in three attitude directions (roll, pitch, and sway) are collected by sensors. The vibration signals are then denoised using a wavelet threshold denoising algorithm. The denoised vibration signals are then subjected to Fourier transform to obtain multi-dimensional vibration spectrum data of the immersed tube attitude. The vibration spectrum data includes the vibration frequency, amplitude, and phase information of each attitude direction.

3. The multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 2, characterized in that, The second core data is the three-dimensional deformation field distribution data of the pipe section joint. The process of obtaining the three-dimensional deformation field distribution data of the pipe section joint includes: uniformly arranging the data around the circumference of the immersed tunnel joint. A laser displacement sensor and a strain gauge are used. The laser displacement sensor is used to collect the normal displacement data of the joint, and the strain gauge is used to collect the tangential strain data of the joint. Based on the data collected by the laser displacement sensor and the strain gauge, a three-dimensional reconstruction algorithm is used to construct the three-dimensional deformation field of the joint. The constructed three-dimensional deformation field is smoothed to remove abnormal deformation points, and the distribution data of the three-dimensional deformation field of the pipe section joint is obtained.

4. The multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 1, characterized in that, The construction process of the first fusion model includes: determining the coupling correlation factor between the first core data and the second core data based on the synergistic mechanism of the fluid-structure interaction dynamic response and structural deformation of the immersed tube. Establish the coupled dynamic equations: in, For the density of the immersed tube structure, Let be the displacement vector of the immersed tube structure. For time parameters, This is the fluid viscosity damping coefficient. Here is the stiffness matrix of the immersed tube structure. Let be the force vector of fluid-structure interaction, where Water pressure, It is the synergistic force of structural deformation.

5. A multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 4, characterized in that, Based on the coupled dynamic equations, a feature extraction network is constructed. This network includes convolutional layers, an attention mechanism layer, and a fully connected layer. The convolutional layers extract the vibration spectrum features of the first core data and the deformation field features of the second core data. The attention mechanism layer assigns weights to the two types of features, with the weight assignment coefficients determined by the coupling correlation factor. Adaptive adjustment: The fully connected layer is used to map the weighted fused features to the initial parameter set of the sinking state.

6. The multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 1, characterized in that, The construction process of the second adaptive decoupling model includes: determining the orthogonal basis vector set of the multi-physics coupling parameters based on the adaptive orthogonal stripping algorithm of the multi-physics coupling parameters during the sinking process. ,in Let the dimensions of the coupling parameters be defined; construct an adaptive decoupling objective function: in, No. A vector of coupling parameters for each physical field. For the first The projection matrix of each physical field. They are orthogonal basis vectors. The regularization coefficient is used to avoid model overfitting. The adaptive decoupling objective function is solved by gradient descent algorithm to obtain the orthogonal stripping coefficients of each physical field coupling parameter. Based on the orthogonal stripping coefficients, a second adaptive decoupling model is constructed. The model is used to strip the coupling parameters in the initial parameter set of the immersion state into independent single physical field parameters to obtain the immersion control adaptation parameter set.

7. A multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 6, characterized in that, The set of sinking control adaptation parameters includes ballast adjustment parameters, attitude adjustment parameters, and docking fine-tuning parameters. The ballast adjustment parameters are quantitative parameters that characterize the dynamic balance between buoyancy and gravity of the immersed tube. The attitude adjustment parameters are quantitative parameters that characterize the correction of the roll, pitch, and heave angles of the immersed tube. The docking fine-tuning parameters are quantitative parameters that characterize the three-dimensional deformation compensation of the docking joint. Each parameter is obtained by mapping the independent single-physics field parameters output by the second adaptive decoupling model.

8. The multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 1, characterized in that, The construction process of the third intelligent early warning model includes: screening environmental risk factors and structural health status assessment indicators. Environmental risk factors include water flow velocity, wave height, and water density. Structural health status assessment indicators include equivalent stress of the immersed tube structure and maximum deformation of the joint. A real-time fusion assessment function for environmental risk factors and structural health status assessment indicators is established, as shown in the following formula: in, for The comprehensive evaluation value of the sinking status at any given moment. for The comprehensive quantitative value of environmental risk factors at any given time. for The comprehensive quantitative value of the structural health status at any given time. , These are the weighting coefficients for environmental risk factors and structural health status assessment indicators, respectively. , The system dynamically adjusts according to the sinking conditions; based on the comprehensive evaluation function and combined with the preset sinking accuracy threshold and structural safety threshold, it constructs an early warning judgment logic and generates control instructions to form a third intelligent early warning model. The model is used to generate a dynamic control instruction matrix based on the comparison results between the comprehensive evaluation value and the threshold.

9. A multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 1, characterized in that, The dynamic control command matrix has rows corresponding to three types of actuators and columns corresponding to the core control parameters of each actuator. Each element in the matrix is ​​a specific quantified command value for a single control parameter of a single actuator. The first row of the dynamic control command matrix M corresponds to the ballast water tank regulating valve. To adjust the valve opening, For regulating valve adjustment rate, The first row represents the feedback threshold for the regulating valve action; the second row corresponds to the attitude adjustment thruster. To output thrust to the thruster For the working direction of the thruster, The third row corresponds to the intermittent adjustment frequency of the thruster; it also corresponds to the docking hydraulic fine-tuning device. For fine-tuning of hydraulic devices, For fine-tuning the speed of the hydraulic device, This refers to the operating pressure threshold of the hydraulic device.

10. A multi-source sensor integrated development and monitoring method for underwater immersed tube placement according to claim 1, characterized in that, The underwater IoT edge gateway serves as a communication relay unit between the slumbering monitoring system and the actuator. It features an IP68-level waterproof and sealed design, supports multi-protocol compatibility and conversion, and enables encrypted transmission of collaborative control command sets and real-time feedback of actuator operating data. The edge gateway has a built-in local data cache. When the underwater communication link is interrupted, the data cache automatically stores the collaborative control command sets to be sent, and automatically resends them in the command generation order after the communication link is restored. After receiving the collaborative control command sets and executing actions, the actuator collects its own operating status parameters in real time, including the actuator's workload, action execution accuracy, and equipment operating temperature. These parameters are then fed back to the slumbering monitoring system through the underwater IoT edge gateway, forming a closed-loop control link of sensor acquisition, model calculation, command issuance, and status feedback.

Images