Deep water revetment unit row construction error reverse modeling correction method and device

Abstract

The application discloses a deep-water revetment unit row construction error reverse modeling correction method and device, and relates to the field of data processing. Water profile data of a target construction area where a deep-water revetment unit row is located is acquired; a hydrological state tensor is constructed according to the water profile data; the hydrological state tensor is input into a prediction correction model, and a layered water resistance field and a non-uniform sinking path dynamic response equation of the unit row body are constructed; based on the layered water resistance field and the non-uniform sinking path dynamic response equation, a prediction result is determined; a posture response vector flow of the unit row body after entering water is acquired through a sensor device; the posture response vector flow and the prediction result are dynamically matched, and a paving path correction value and a pitch attitude adjustment suggestion amount are output; and based on the paving path correction value and the pitch attitude adjustment suggestion amount, the posture and the sinking path of the unit row body are dynamically corrected. The application facilitates improving the accuracy of deep-water revetment unit row construction operation in a complex hydrological scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a deep water revetment unit row construction error reverse modeling correction method and device. BACKGROUND

[0002] In the deep water revetment construction process, especially for the accurate laying operation of unit row body in the deep water section of large rivers such as the Yangtze River, the sinking process of the row body faces significant stratified water disturbance characteristics.

[0003] At present, under such complex hydrodynamic environment, it is often accompanied by vertical salinity gradient, water temperature stratification structure and high concentration of suspended sediment particles, thereby causing the buoyancy and resistance of the row body during sinking to present significant depth dependence and nonlinear distribution. This stratified water resistance distribution differentiation phenomenon causes the row body sinking process to fail to meet the ideal uniform rate assumption, and is prone to nonlinear response behaviors such as acceleration mutation, path deviation and attitude disturbance in the local water layer, thereby affecting the laying accuracy and structural stability. However, the error model processing method relying on traditional shallow water empirical parameters is difficult to capture its true characteristics, and cannot accurately describe the dynamic response behavior of the row body during sinking, resulting in low accuracy of deep water revetment unit row construction operation in complex hydrological scenarios.

[0004] Therefore, there is an urgent need for a deep water revetment unit row construction error reverse modeling correction method and device. SUMMARY

[0005] The present application provides a deep water revetment unit row construction error reverse modeling correction method and device, which facilitates to improve the accuracy of deep water revetment unit row construction operation in complex hydrological scenarios.

[0006] In a first aspect of the present application, a deep water revetment unit row construction error reverse modeling correction is provided, the method comprising: obtaining water profile data of a target construction area where a deep water revetment unit row is located, the water profile data including water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient; constructing a hydrological state tensor according to the water profile data; inputting the hydrological state tensor into a prediction correction model to construct a stratified water resistance field and a non-uniform sinking path dynamic response equation of the unit row body; determining a prediction result based on the stratified water resistance field and the non-uniform sinking path dynamic response equation; obtaining an attitude response vector flow of the unit row body after entering the water sent by a sensor device; dynamically matching the attitude response vector flow with the prediction result to output a laying path correction value and a pitch attitude adjustment suggestion amount; and dynamically correcting the attitude and sinking path of the unit row body based on the laying path correction value and the pitch attitude adjustment suggestion amount.

[0007] By adopting the technical scheme, complete water body profile data including water depth, water temperature, water body density, water body salinity, water body turbidity and flow velocity gradient are acquired, a high-dimensional hydrological state tensor is constructed and input into a prediction correction model, nonlinear changes of buoyancy and resistance suffered by the body under layered disturbance conditions are effectively described, and a dynamic response equation conforming to actual working conditions is established. Combined with the attitude response vector flow acquired by the sensor, dynamic matching is performed, path deviation and attitude deviation during sinking are identified in real time, and correction amounts are output based on the same, so that the control system is driven to perform accurate adjustment, and a closed-loop correction mechanism based on measured feedback is formed. The method breaks through the adaptability bottleneck of traditional empirical models to complex hydrological disturbance, significantly improves the sinking accuracy, attitude stability and response robustness of the unit body in a deep water environment with different water resistance, and is suitable for high-quality bank protection construction in a complex hydrological scene, and facilitates to improve the accuracy of deep water bank protection unit row construction in a complex hydrological scene.

[0008] Optionally, the constructing a hydrological state tensor according to the water body profile data specifically comprises: reconstructing continuity of the water body profile data in a profile layer by using a spatial interpolation method, and performing smoothing processing in a depth direction by using a Gaussian kernel regression and spline fitting method to obtain continuous tensor functions of various hydrological properties in a three-dimensional space; mapping the continuous tensor functions corresponding to the water depth, the water temperature, the water body density, the water body salinity, the water body turbidity and the flow velocity gradient into six tensor channels respectively, and constructing a four-dimensional state tensor with spatial transverse coordinates, vertical depth coordinates and time coordinates as indexes; and performing time sequence filtering on the four-dimensional state tensor by using a time moving window processing and a convolution smoothing algorithm to obtain the hydrological state tensor.

[0009] By adopting the technical scheme, the spatial interpolation, Gaussian kernel regression and spline fitting in the depth direction are performed on the water body profile data, high-precision continuity reconstruction of multiple hydrological properties in a three-dimensional space is realized, and a four-dimensional state tensor is further constructed by introducing a time dimension, so as to comprehensively reflect spatial heterogeneity and time sequence change characteristics of hydrological disturbance. Various hydrological parameters are uniformly mapped into a multi-channel tensor structure, and filtering processing is performed by using a time moving window and a convolution smoothing algorithm, so as to significantly improve continuity and robustness of the data. The hydrological state tensor provides a complete structure, high resolution and physically consistent input basis for subsequent dynamic response modeling and prediction correction, effectively solves the problems of non-uniform information dimension and discontinuous disturbance scale under a complex water body profile, and guarantees adaptability and accuracy of the model in a deep water dynamic environment.

[0010] Optionally, the hydrological state tensor is input into a prediction correction model to construct a layered water body resistance field and a non-uniform sinking path dynamic response equation of the unit array body, specifically including: inputting the water body density tensor and the flow velocity gradient tensor in the hydrological state tensor into a resistance field construction unit in the prediction correction model, calculating the buoyancy and resistance of the unit array body in each profile layer based on the water body density and local flow velocity change rate of each depth layer, and constructing corresponding layered water body buoyancy and resistance functions; based on the layered water body buoyancy and resistance functions, the layered water body resistance field is constructed; the layered water body buoyancy and resistance functions are input as input variables into a dynamic response modeling unit in the prediction correction model, combined with the mass parameters, gravity constant and constant component load parameters of the unit array body, to construct a non-uniform sinking path dynamic response equation of the unit array body in the vertical sinking process, which is a layered cumulative nonlinear variable coefficient second-order differential equation set, including the sinking acceleration function, sinking velocity function and sinking displacement function of the unit array body at each depth position and time node in the vertical space.

[0011] By using the above technical scheme, the water body density tensor and the flow velocity gradient tensor in the hydrological state tensor are input into the prediction correction model to construct the layered water body buoyancy and resistance functions, accurately reflecting the dynamic changes of the mechanical action on the array body in different depth layers, breaking through the simplified assumption of uniform water resistance of the traditional model. Further combined with the mass parameters, gravity constant and constant component load parameters of the unit array body, a layered cumulative nonlinear variable coefficient second-order differential equation set is established to systematically depict the acceleration, velocity and displacement evolution relationship of the array body in the non-uniform sinking state. The dynamic response modeling mechanism has high physical consistency and depth resolution, and can realize fine prediction of the dynamic behavior of the array body in the layered disturbance environment, providing accurate mechanical benchmarks for attitude response matching and path correction, effectively improving the control accuracy and modeling credibility in deep water revetment operation.

[0012] Optionally, the prediction result is determined based on the layered water body resistance field and the non-uniform sinking path dynamic response equation, specifically including: using a numerical solution method to perform time step solution on the non-uniform sinking path dynamic response equation to generate the acceleration function, velocity function and displacement function of the unit array body in the continuous time domain; mapping the acceleration function, velocity function and displacement function to the depth dimension to form the acceleration profile, velocity profile and displacement profile, and combining to construct a state prediction matrix as the prediction result.

[0013] By adopting the technical scheme, the acceleration function, the speed function and the displacement function of the body in the continuous time domain are generated by using the time stepping numerical solution method for the non-uniform speed depositing path dynamic response equation, so that the time sequence evolution characteristics of the dynamic response in the depositing process can be accurately described. Further, the above-mentioned dynamic functions are mapped to the depth dimension to construct the acceleration profile, the speed profile and the displacement profile, and the depositing behavior difference of the body under different water layer structures is systematically shown, thereby breaking through the neglect of the profile response of the traditional single time sequence model. Finally, the state prediction matrix constructed integrates the time and space information, which is used as a high-dimensional reference for subsequent error matching and path correction, thereby significantly improving the adaptability and error identification accuracy of the model to the nonlinear disturbance path, and laying a stable numerical foundation for accurate control in the deep water revetment construction process.

[0014] Optionally, the posture response vector flow and the prediction result are dynamically matched to output a laying path correction value and a pitch attitude adjustment suggestion, specifically including: the posture response vector flow is constructed into a time-synchronized posture response data matrix, the posture response vector flow includes a pitch angle, a roll angle, an attitude angular velocity, a settling acceleration and a vertical displacement of the unit body in the depositing process; a posture prediction reference sequence corresponding to a time-synchronized time period is extracted from the prediction result to form a posture prediction sequence corresponding to the posture response vector flow; the posture response data matrix and the posture prediction sequence are input into a dynamic matching unit in the prediction correction model, a dynamic time warping algorithm is used for nonlinear alignment matching to generate a dynamic error function; based on the dynamic error function, the laying path correction value and the pitch attitude adjustment suggestion are calculated, the laying path correction value is a cumulative offset distance of the vertical path of the unit body, and the pitch attitude adjustment suggestion is a correction direction and amplitude of the attitude angle of the unit body.

[0015] By adopting the technical scheme, the posture response vector flow is constructed into a time-synchronized data matrix, and the posture prediction sequence extracted from the prediction result is dynamically time-warped to effectively identify the nonlinear deviation between the actual response and the prediction state in the depositing process. By constructing a dynamic error function, the offset trend of the body in the vertical path and the attitude angle is quantified, and then the laying path correction value and the pitch attitude adjustment suggestion are accurately calculated to realize the two-dimensional correction control of the spatial position and the attitude behavior of the body. This mechanism breaks through the limitations of the static error estimation method in complex disturbance environment, has high time sequence adaptability and response sensitivity, provides accurate and efficient feedback basis for subsequent dynamic correction, and greatly improves the attitude stability and path control accuracy of the body in the deep water revetment construction.

[0016] Optionally, based on the laying path correction value and the pitch attitude adjustment suggestion, the attitude and laying path of the unit laying body are dynamically corrected, specifically including: generating a speed control instruction for adjusting the release rate of the main winch and a rhythm control instruction for correcting the laying rhythm according to the laying path correction value; generating a tension control instruction for adjusting the tension of the side anchor cable and an attitude control instruction for driving the tail attitude adjusting mechanism of the laying body according to the pitch attitude adjustment suggestion; the speed control instruction, the rhythm control instruction, the tension control instruction and the attitude control instruction are synchronously input to the laying control terminal, and the main winch control device, the anchor cable control device and the laying body attitude executing device of the laying control terminal are used to perform the laying path fine adjustment and attitude stability control of the unit laying body.

[0017] By adopting the above technical solution, the laying path correction value and the pitch attitude adjustment suggestion are converted into specific speed control instructions, rhythm control instructions, tension control instructions and attitude control instructions, which are respectively used for the main winch control device, the anchor cable control device and the laying body attitude executing device, so as to realize the precise linkage control of the laying path and attitude of the unit laying body. The dynamic correction mechanism generates multi-dimensional control actions according to real-time feedback results, has high responsiveness and strong execution closed loop capability, can effectively eliminate the deviation accumulation and attitude drift in the non-uniform laying process, and ensures the precise alignment and stable landing of the laying body in the complex hydrological disturbance environment. This technology breaks through the traditional manual experience adjustment mode, realizes the automatic, high-precision and multi-channel collaborative laying behavior closed loop management, and significantly improves the construction quality, safety level and system robustness of deep water revetment operation.

[0018] Optionally, the historical water depth data, the historical water temperature data, the historical water density data, the historical water salinity data, the historical water turbidity data and the historical flow gradient data of the target construction area are spatially registered and hierarchically reconstructed to construct a historical hydrological state tensor; the unit laying body attitude response vector flow and control parameter sequence in the corresponding historical construction task of the target construction area are collected to construct a training sample set; based on the historical hydrological state tensor and the training sample set, a supervised learning mechanism is adopted to take the laying control feedback as the target output, and the variational Bayesian optimization method is used to train the resistance function parameters, the dynamic equation coefficients and the control mapping weights in the initial model, so as to complete the training of the prediction correction model.

[0019] By adopting the technical scheme, the historical hydrological data in multiple dimensions such as historical water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient are spatially registered and reconstructed in layers to construct a high-fidelity historical hydrological state tensor, and a training sample set is systematically constructed by combining the attitude response vector flow and the control parameter sequence in the corresponding historical construction task, so as to comprehensively capture the dynamic behavior law of the unit row under deep water disturbance conditions. A supervised learning mechanism is adopted to take the sinking control feedback as the target output, and a variational Bayesian optimization method is used to jointly train the resistance function parameters, the dynamic response structure and the control mapping weight of the prediction correction model to realize global convergence of the parameters and improvement of the model generalization ability. The training strategy not only strengthens the coupling expression capability of the model to complex hydrological disturbance and nonlinear response, but also provides a highly accurate and robust model basis for subsequent real-time prediction and dynamic correction, thereby significantly improving the intelligent level of deep water revetment construction.

[0020] In a second aspect of the present application, a deep water revetment unit row construction error reverse modeling correction device is provided, the device comprising an acquisition module and a processing module, wherein the acquisition module is configured to acquire water profile data of a target construction area where a deep water revetment unit row is located, the water profile data comprising water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient; the processing module is configured to construct a hydrological state tensor according to the water profile data; the processing module is further configured to input the hydrological state tensor into a prediction correction model to construct a layered water resistance field and a non-uniform sinking path dynamic response equation of the unit row body; the processing module is further configured to determine a prediction result based on the layered water resistance field and the non-uniform sinking path dynamic response equation; the acquisition module is further configured to acquire an attitude response vector flow sent by a sensor device after the unit row body enters the water; the processing module is further configured to dynamically match the attitude response vector flow and the prediction result to output a laying path correction value and a pitch attitude adjustment suggestion; and the processing module is further configured to dynamically correct the attitude and sinking path of the unit row body based on the laying path correction value and the pitch attitude adjustment suggestion.

[0021] In a third aspect of the present application, an electronic device is provided, which comprises a processor, a memory, a user interface and a network interface, the memory is configured to store instructions, the user interface and the network interface are configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method described above.

[0022] In a fourth aspect of the present application, a computer readable storage medium is provided, which stores instructions, when the instructions are executed, the method described above is performed.

[0023] To sum up, the one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0024] By acquiring complete water profile data including water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient, constructing a high-dimensional hydrological state tensor and inputting the prediction correction model, the nonlinear changes of the buoyancy and resistance suffered by the body under the condition of layered disturbance are effectively described, and a dynamic response equation conforming to the actual working condition is established. Combined with the attitude response vector flow obtained by the sensor, the path deviation and attitude deviation during sinking can be identified in real time, and the correction amount is output accordingly to drive the control system to implement accurate adjustment, forming a closed-loop correction mechanism based on measured feedback. This method breaks through the adaptability bottleneck of traditional empirical models to complex hydrological disturbances, significantly improves the sinking accuracy, attitude stability and response robustness of the unit row body in the deep water environment with different water resistance, and is suitable for high-quality revetment construction in complex hydrological scenarios, and is convenient for improving the accuracy of deep water revetment unit row construction in complex hydrological scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 A flowchart of a deep water revetment unit row construction error reverse modeling correction method provided by an embodiment of the application is shown in the figure.

[0026] Figure 2 A module schematic diagram of a deep water revetment unit row construction error reverse modeling correction device provided by an embodiment of the application is shown in the figure.

[0027] Figure 3 A structural schematic diagram of an electronic device provided by an embodiment of the application is shown in the figure.

[0028] Explanation of reference signs: 21, acquisition module; 22, processing module; 31, processor; 32, communication bus; 33, user interface; 34, network interface; 35, memory. DETAILED DESCRIPTION

[0029] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be described clearly and completely below in conjunction with the drawings in the embodiments of the specification. Obviously, the described embodiments are only a part of the embodiments of the application, not all the embodiments.

[0030] In the description of the embodiments of the application, the words such as "for example" or "for instance" are used to represent an example, illustration or description. Any embodiment or design scheme described as "for example" or "for instance" in the embodiments of the application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words such as "for example" or "for instance" are intended to present the relevant concept in a specific manner.

[0031] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first", "second", "third", etc. are used only for descriptive purposes and should not be construed as indicating or implying relative importance or implying the indicated technical features. Therefore, the features defined with "first", "second", etc. can be explicitly or implicitly included one or more of the features. The terms "include", "contain", "have" and their variants mean "include but not limited to", unless otherwise specifically emphasized.

[0032] In deep water revetment construction operations, especially for the implementation of high-precision laying tasks of unit row bodies in deep water areas of large rivers such as the Yangtze River, the row body sinking process is generally affected by significant stratified water disturbance.

[0033] At present, in such a complex hydrodynamic environment, there are obvious vertical salinity gradient, water temperature stratification and high concentration of suspended sediment particles. These factors together cause the buoyancy and resistance experienced by the row body during sinking to exhibit strong depth dependence and nonlinear variation characteristics, forming a typical stratified water resistance distribution differentiation effect. This effect causes the row body to produce non-uniform acceleration changes, path drift and attitude deflection in a specific depth region, causing the sinking behavior to exhibit significant nonlinear response, severely restricting the laying precision and structural stability. The existing empirical error model based on shallow water conditions generally ignores the deep water disturbance mechanism and stratified fluid dynamic characteristics, and is difficult to accurately characterize the dynamic evolution behavior of the row body sinking process, resulting in insufficient construction precision in complex hydrological environments, limiting the refinement and intelligent development of unit row body laying construction in deep water revetment projects.

[0034] To solve the above technical problems, the present application provides a deep water revetment unit row construction error reverse modeling correction method, referring to Figure 1 , Figure 1 A flowchart of a deep water revetment unit row construction error reverse modeling correction method provided by an embodiment of the present application. The method is applied to a server and includes steps S110 to S170, which are as follows:

[0035] S110, obtain water profile data of a target construction area where the deep water revetment unit row is located, the water profile data including water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient.

[0036] Specifically, the server first collects raw water profile data in the construction area in real time by accessing multiple types of sensor devices deployed in the target construction area, including multibeam echo sounders, CTD profilers, laser turbidity meters, and three-dimensional Doppler current meters, etc. The data collected by each sensor corresponds to six types of hydrological indicators, including water depth, water temperature, water density, water salinity, water turbidity, and flow gradient. The sampled data is transmitted back to the server through a communication link and is synchronized, normalized, and timestamped by a unified space-time registration module to ensure consistency and interactivity of all water profile data in space and time dimensions.

[0037] For example, in a deep water section of the Yangtze River, a multibeam echo sounder installed under the hull of a deployment ship continuously measures the water depth profile at different latitude and longitude coordinates; a CTD profiler mounted on a vertical profile deployment platform records the continuous changes of water temperature and water salinity from the water surface to the bottom; and a laser turbidity meter and a three-dimensional Doppler current meter operating synchronously measure the turbidity changes and flow velocity vectors in different depths, respectively. After receiving the multi-source data returned by the above devices, the server assembles various indicators into a standardized multi-layer water profile data set based on the profile coordinate system of the construction area, which serves as the basis for subsequent construction of the hydrological state tensor.

[0038] The server here refers to a central processing device or an edge computing node deployed in the construction management system, which has comprehensive functions such as data collection access, space-time data fusion, model inference calculation, and control instruction issuance. In the deep water revetment unit sinking control system, the server plays the following roles: first, as a data center, it receives and analyzes various hydrological data streams sent by water profile data collection devices in real time; second, as a computing core, it performs high-density computing tasks such as hydrological state tensor construction, prediction correction model inference, dynamic error analysis, and control strategy generation; third, as an instruction dispatcher, it distributes attitude and path correction control instructions to the laying and executing system to realize closed-loop control.

[0039] The server can be an industrial-grade edge server deployed on a construction control ship, or a high-performance data processing cluster in a remote control center, depending on project requirements. For example, in a certain construction task, an edge server is mounted on the main control platform, which integrates multi-thread task management and deep learning inference modules, and can complete state tensor construction and dynamic prediction response calculation within seconds of receiving water profile data, and immediately output control instructions to the winch system and anchor cable control device to realize real-time and accurate adjustment of the unit row body attitude and sinking path.

[0040] S120, constructing a hydrological state tensor according to the water profile data.

[0041] Specifically, taking the revetment operation of a deep water section of the Yangtze River as an example, after the server receives the water temperature data and water density data collected in layers within a range of 30 meters in water depth, the data is first aligned with the construction profile transverse grid coordinates, and then the bidirectional spline function is used to interpolate and expand the 10 sampling layers to construct a continuous profile tensor of water temperature and density at a depth interval of 0.5 meters; the above data from different time nodes are then integrated in time through a sliding time window to form a four-dimensional state tensor containing six channels of water temperature, water density, water salinity, etc., for subsequent calling by the prediction correction model to construct a layered water resistance function and a dynamic response equation.

[0042] In one possible implementation, according to the water profile data, a hydrological state tensor is constructed, specifically including: using a spatial interpolation method to reconstruct the continuity of the water profile data within the profile layer, and using a Gaussian kernel regression and spline fitting method to smooth in the depth direction to obtain continuous tensor functions of various hydrological properties in three-dimensional space; mapping the continuous tensor functions of water depth, water temperature, water density, water salinity, water turbidity, and flow velocity gradient respectively into six tensor channels, and indexing with spatial transverse coordinates, vertical depth coordinates, and time coordinates to construct a four-dimensional state tensor; performing time series filtering on the four-dimensional state tensor through time moving window processing and convolution smoothing algorithm to obtain a hydrological state tensor.

[0043] Specifically, the process is sequentially executed by the server in three stages, including spatial reconstruction, tensor mapping, and time series filtering, based on obtaining complete water profile data, to generate a hydrological state tensor with spatial continuity and time series stability. In the first stage, the server first performs interpolation reconstruction of the collected water depth data, water temperature data, water density data, water salinity data, water turbidity data, and flow velocity gradient data within the profile layer. Taking a hydrological property f(x, y, z) as an example, x and y represent the plane coordinates of the construction area, and z represents the depth. The spatial interpolation uses a weighted inverse distance interpolation method to construct as follows:

[0044]

[0045] where f i represents the property value at sampling point i, (x i , y i ) is the plane coordinate of the sampling point, is a smoothing term, is a weight decay parameter, usually taking a value between 1.5 and 2.5. This interpolation method forms a two-dimensional tensor layer with continuous distribution for each hydrological property within the spatial cross section. Then, in the depth direction, the server uses a Gaussian kernel regression smoothing function to further optimize the interpolation result, which is constructed as follows:

[0046]

[0047] where z j is the depth value corresponding to the depth sampling point j, is the fitting coefficient, and σ is the bandwidth of the Gaussian kernel, reflecting the fitting smoothness. If it is necessary to ensure the consistency of the fitting strength of deep and shallow layers, a cubic spline function can be used to further perform boundary preservation and curvature constraint on the direction.

[0048] In the second stage, the server forms a six-channel tensor structure T(x, y, z, c) by taking the processed hydrological attribute tensors of various types as channels, where c corresponds to water depth, water temperature, water density, water salinity, water turbidity, and flow velocity gradient in turn, and the tensor dimension is a three-dimensional coordinate structure. To introduce the time variation feature, in each time window t k The construction process is repeated and stacked in time sequence, and finally a four-dimensional tensor T(x, y, z, t, c) is constructed, which completely describes the hydrological disturbance dynamics of the target construction area in space and time.

[0049] In the third stage, the server performs time sequence filtering on the four-dimensional tensor to improve data robustness. The time moving window algorithm is used to perform sliding average on the tensor in the t dimension, assuming that the window width is w, the step size is s, and the mass of the unit array is m. Then, the filtered tensor at each time point t k is given by the following formula:

[0050]

[0051] To further suppress high-frequency noise and abnormal disturbance of the profile, while maintaining the clear boundary of the hydrological structure, the server further applies a one-dimensional convolution smoothing operation to the t-dimensional tensor sequence. The kernel function is constructed as follows:

[0052]

[0053] where, is a Gaussian kernel function, is a time dimension smoothing parameter. Through this process, the server finally outputs a four-dimensional hydrological state tensor with time continuity, depth layer clarity, and spatial structure consistency, which serves as the core input of the prediction correction model.

[0054] For example, in a revetment section with a water depth of 25 meters in a certain construction area, the server processes the real-time water density and flow velocity gradient data collected from 12 profile points according to the above process, generates a hydrological state tensor T(x, y, z, t, c) with a resolution of every 5 minutes as a time frame, and a total of six channels, which is used for subsequent modeling of the layered water resistance field and the non-uniform velocity response behavior of the array.

[0055] S130, input the hydrological state tensor into the prediction correction model to construct a layered water body resistance field and a non-uniform sinking path dynamic response equation of the unit array.

[0056] Specifically, when the server implements this step, first, the water body density tensor and the flow velocity gradient tensor are extracted from the constructed hydrological state tensor and input into the resistance field construction unit of the prediction correction model. Based on the water body density change rate and the local flow velocity gradient of each depth layer, the module calculates the buoyancy and resistance of the unit array in different depth profiles, and constructs the layered buoyancy function and the layered resistance function accordingly. The server uses the layer-by-layer integration method to splice and accumulate the buoyancy and resistance functions along the vertical space to obtain a continuously distributed layered water body resistance field under the spatial coordinates, which is used to describe the coupling relationship between the array and the layered disturbed water body in the entire sinking path.

[0057] In one possible implementation, the hydrological state tensor is input into the prediction correction model to construct a layered water body resistance field and a non-uniform sinking path dynamic response equation of the unit array, specifically including: inputting the water body density tensor and the flow velocity gradient tensor in the hydrological state tensor into the resistance field construction unit of the prediction correction model, calculating the buoyancy and resistance of the unit array in each profile layer based on the water body density and the local flow velocity change rate of each depth layer, and constructing the corresponding layered water body buoyancy and resistance functions; based on the layered water body buoyancy and resistance functions, a layered water body resistance field is constructed; the layered water body buoyancy and resistance functions are input as input variables, combined with the mass parameters, the gravitational constant and the constant component load parameters of the unit array, into the dynamic response modeling unit of the prediction correction model, to construct a non-uniform sinking path dynamic response equation of the unit array in the vertical sinking process. The non-uniform sinking path dynamic response equation is a layered cumulative nonlinear variable coefficient second-order differential equation group, which includes the sinking acceleration function, the sinking velocity function and the sinking displacement function of the unit array at each depth position and time node in the vertical space.

[0058] Specifically, this process is implemented by the server in stages after the construction of the hydrological state tensor. First, the water body density tensor and the flow velocity gradient tensor are extracted and input into the resistance field construction unit of the prediction correction model. In each depth profile z k Based on the water body density and the local flow velocity change rate, the buoyancy B(z k ) and the resistance D(z k ) of the unit array are calculated. The buoyancy function is constructed based on the Archimedes principle, and the resistance function introduces a velocity coupling term and a density gradient term, which are specifically expressed as follows:

[0059]

[0060]

[0061]

[0062] Among them, B(z) k ) represents the depth z k Buoyancy, Where is the density of the water, g is the gravitational acceleration constant, and V is the volume of water discharged. For depth z k At the resistance, v k Let A be the current velocity of the pump body and A be the projected area of ​​the force. C0 is the drag coefficient, and C0 is the static drag constant. and These are the weighting adjustment coefficients for the density gradient term and the velocity gradient term, respectively. For density gradient, This represents the magnitude of the local velocity gradient.

[0063] The server maps the aforementioned buoyancy and drag functions into three-dimensional space, forming a continuously distributed, layered water drag field to dynamically reflect the impact of hydrological disturbances on the stress state of the discharge vessel. Subsequently, the server calls the dynamic response modeling unit in the prediction correction model to convert B(z) into a dynamic response modeling function. k ) and D(z k ,v k As input variables, together with the mass m, gravity constant g, and component dead load parameter G of the unit array, a nonlinear variable coefficient second-order differential equation is constructed, as follows:

[0064]

[0065] Where s(z,t) is the displacement function of the slab at depth z and time t. Let be the sinking acceleration function, v(z) be the sinking velocity function, and G be the gravitational force caused by the constant load on the raft. This equation is applied at different depths z. k Sub-equations are established on the above, and a hierarchical accumulation strategy is used to solve the system, forming a complete dynamic response structure for the non-uniform sinking path.

[0066] For example, in a real deep-water construction project, the server identified a halocline layer within the range of z=10 meters to z=14 meters, causing a sudden change in water density. Combined with the increase in local high-velocity shear zones, the prediction correction model calculated the extent of this change. Rising rapidly, and then The significant enhancement caused changes in the acceleration of the sprue. Based on this, the server constructed an updated dynamic response equation for the non-uniform sinking path, enabling the system to anticipate and correct attitude disturbances in advance.

[0067] S140, determining the prediction result based on the layered water body resistance field and the non-uniform sinking path dynamic response equation.

[0068] Specifically, when the server determines the prediction result based on the layered water body resistance field and the non-uniform sinking path dynamic response equation, first, on the basis of the constructed nonlinear variable coefficient second-order differential equation group, a time-stepping numerical solution method is used to solve the dynamic path. In specific implementation, the server selects a high-precision algorithm such as the Runge-Kutta method or an implicit multi-step method, takes the initial speed and initial displacement as boundary conditions, and performs step-by-step integration of the sinking acceleration function at fixed time intervals, to sequentially solve the sinking speed function and the sinking displacement function, and obtain the acceleration, speed, and displacement values of each node of the body in the time dimension. This solving process is simultaneously carried out on each depth profile layer, and a multi-dimensional time sequence body dynamic response structure is constructed.

[0069] Subsequently, the server expands the above-mentioned sinking acceleration function, sinking speed function, and sinking displacement function generated at each time into acceleration profiles, speed profiles, and displacement profiles along the depth direction, and uniformly maps these three types of profile data in the spatial structure to the construction profile grid, to combine and construct a state prediction matrix as the core prediction result output by the prediction correction model. Taking a certain deep water revetment construction as an example, the server identifies that the water body resistance is dramatically enhanced between z=13m and z=15m, and generates a prediction response that the body appears a sinking speed drop and posture change trend in this region through numerical solution, to provide a complete behavior benchmark for the subsequent dynamic matching module. This prediction result not only reflects the vertical motion trend of the body, but also has the ability to capture the abnormal disturbance response behavior in advance, thereby supporting the generation of high-precision control instructions.

[0070] In one possible implementation, the prediction result is determined based on the layered water body resistance field and the non-uniform sinking path dynamic response equation, and specifically includes: using a numerical solution method to perform time-stepping solution on the non-uniform sinking path dynamic response equation, to generate acceleration functions, speed functions, and displacement functions of the unit body in the continuous time domain; mapping the acceleration functions, speed functions, and displacement functions to the depth dimension to form acceleration profiles, speed profiles, and displacement profiles, and combining and constructing a state prediction matrix as the prediction result.

[0071] Specifically, this process is implemented by the server after the non-uniform sinking path dynamic response equation is constructed, first, the non-linear variable coefficient second-order differential equation group is solved by a numerical solution method, to generate acceleration functions a(t), speed functions v(t), and displacement functions s(t) in the continuous time domain. The standard form of the dynamic response equation is:

[0072]

[0073] where m is the mass of the unit row body, s(t) is the vertical displacement function, v(t) is the sinking velocity function, a(t) is the sinking acceleration function, B(z(t)) is the buoyancy function at the current position z(t) of the row body, D(z(t), v(t)) is the resistance function under the position and velocity conditions, and G is the constant gravity term. The server solves it using the Runge-Kutta fourth-order method, discretizes the time domain into a node sequence t0, t1,..., t n , and recursively calculates the values of a(t k ), v(t k ), and s(t k ) at each time point to form a sequence of dynamic response functions in the continuous time domain.

[0074] After the solution is completed, the server remaps the above acceleration function, velocity function, and displacement function to the vertical depth dimension. Since the z(t) of the row body is a monotonically increasing function during sinking, the server establishes the inverse function mapping relationship t(z) of the function value, and then generates the acceleration profile a(z), the velocity profile v(z), and the displacement profile s(z) in the depth domain. This conversion ensures that each depth profile can obtain complete dynamic response information. Finally, the server fuses the three sets of profile data to construct a state prediction matrix:

[0075]

[0076] The state prediction matrix comprehensively expresses the dynamic response characteristics of the unit row body at different depth positions, and is a reference for subsequent attitude response matching and path correction. For example, in a deep water revetment laying task with a water depth of 18 meters, the server identifies the acceleration peak a(14.5)=0.84m / s 2 at z=14.5m caused by density mutation through the prediction process, and accordingly predicts that there is a vertical disturbance risk at this position, and supports the control module to make fine planning in advance. This implementation ensures that the row body sinking state has a time-depth consistent modeling basis, greatly improving the path prediction accuracy under non-uniform speed disturbance.

[0077] S150, obtaining the attitude response vector flow of the unit row body after entering the water sent by the sensor device.

[0078] Specifically, when the server implements the process, first, a data communication link is established with the multiple sets of attitude monitoring sensors deployed on the unit row body, and a real-time attitude response vector stream containing post-entry dynamic response data is received. The attitude response vector stream is collected by the inertial measurement unit, attitude angle sensor, accelerometer, and high-frequency underwater positioning module, and specifically includes five types of core attitude response variables: pitch angle, roll angle, attitude angular velocity, vertical acceleration, and current depth position. The server standardizes each data record according to the timestamp and synchronously fills in the missing data at fixed time intervals to form a time-continuous attitude response data stream, establishing a high-resolution data basis for subsequent dynamic matching and deviation estimation.

[0079] For example, in a Yangtze River deep water section revetment construction task, the six-axis inertial navigation device installed on the unit row body reports pitch angle and acceleration data 20 times per second. The server receives the following vector sequence through the interface: at t = 4.0 seconds, the pitch angle is 3.2°, the roll angle is 1.1°, the attitude angular velocity is 0.18° / s, the vertical acceleration is -0.92 m / s², and the current depth is z = 11.8 meters. The server embeds this set of values into the attitude response vector stream, continuously stacks and updates them in chronological order, ensuring that the constructed attitude response data matrix fully covers the dynamic behavior characteristics of the unit row body during the entire sinking process. This process provides physical response basis and data support for subsequent dynamic error alignment and control parameter correction of path prediction results.

[0080] S160, dynamically match the attitude response vector stream with the prediction results, output the row-laying path correction value and the pitch attitude adjustment suggestion.

[0081] Specifically, when the server implements the process, first, the attitude response vector stream is constructed into a time-synchronized attitude response data matrix, and the matrix dimension is consistent with the state prediction matrix output by the prediction correction model. Each time point in the attitude response data matrix contains five types of response variables: pitch angle, roll angle, attitude angular velocity, vertical acceleration, and sinking depth, which correspond to the acceleration function, velocity function, and displacement function in the state prediction matrix to form a predicted profile. The server calls the dynamic matching unit in the prediction correction model and uses the dynamic time warping algorithm to nonlinearly align the two time series matrices, constructs a dynamic error function to quantify the timing deviation and amplitude error between the actual response and the predicted behavior. The dynamic error function is formed by the weighted superposition of the point-to-point differences of each dimension variable, which is used to extract abnormal paragraphs and identify drift trends.

[0082] For example, in a deep water revetment operation, the server identifies that the pitch rate in the attitude response data matrix deviates from the pitch reference path in the predicted profile significantly in the time period from t = 6.2 seconds to t = 8.0 seconds, and the dynamic error function exceeds the threshold value 0.12 in this interval. The server calculates the laying path correction value as -0.37 meters accordingly, indicating that the vertical path of the unit row has a cumulative sinking offset of 0.37 meters; at the same time, the server further analyzes that the dominant direction of the pitch error is a negative mutation trend, and generates a pitch attitude adjustment suggestion value of +2.6°, indicating that the control system adjusts the current disturbance deviation by lifting the angle of the tail attitude adjusting mechanism. These output values are directly used as input parameters of the control instruction generation module to realize the linkage control of attitude adjustment and path stability.

[0083] In a possible implementation, the attitude response vector flow is dynamically matched with the prediction result to output the laying path correction value and the pitch attitude adjustment suggestion value, and specifically includes: constructing the attitude response vector flow into a time-synchronized attitude response data matrix, the attitude response vector flow including the pitch angle, roll angle, attitude angular velocity, sinking acceleration, and vertical displacement of the unit row in the sinking process; extracting the attitude prediction reference sequence corresponding to the time-synchronized time period from the prediction result to form the attitude prediction sequence corresponding to the attitude response vector flow; inputting the attitude response data matrix and the attitude prediction sequence into the dynamic matching unit in the prediction correction model, performing nonlinear alignment matching by using the dynamic time warping algorithm to generate a dynamic error function; and based on the dynamic error function, calculating the laying path correction value and the pitch attitude adjustment suggestion value, the laying path correction value being the cumulative offset distance of the vertical path of the unit row, and the pitch attitude adjustment suggestion value being the correction direction and amplitude of the attitude angle of the unit row.

[0084] Specifically, the process is jointly implemented by the server after the attitude response vector flow and the prediction result are generated. First, the server constructs the attitude response vector flow into a structured time-synchronized matrix, each row corresponding to a time step t i , and each column storing the pitch angle p(t i ), roll angle r(t i ), attitude angular velocity Ω(t i ), sinking acceleration a(t i ), and vertical displacement s(t i ), respectively. The server extracts the corresponding interval consistent with the time axis from the state prediction matrix, constructs the mapping relationship by using the vertical displacement function s(t), and then obtains the attitude prediction sequence, which contains the predicted pitch angle, attitude angular velocity, and sinking speed in the same time period.

[0085] The server inputs the posture response data matrix and the posture prediction sequence into a dynamic matching unit in the prediction correction model, and performs nonlinear alignment matching using a dynamic time warping (DTW) algorithm. The DTW algorithm finds a set of paths to minimize the global distance defined by the following formula:

[0086]

[0087] wherein, is a Euclidean distance measurement function for calculating the single-point difference between the posture response and the prediction value. Each w k represents the alignment mapping index point between the posture response and the prediction path.

[0088] The dynamic error function is defined as:

[0089]

[0090] wherein, and are the values of the jth variable in the response matrix and the prediction matrix, such as the pitch angle and the vertical acceleration, is the weight coefficient of each variable.

[0091] The server performs integral processing on the error function to extract the cumulative path offset as the laying path correction value, which is calculated as follows:

[0092]

[0093] wherein, is the response settling acceleration, is the prediction settling acceleration, is the weighted correction function, representing the significance of disturbance in different time periods.

[0094] At the same time, the server calculates the positive and negative direction offset trend and the average offset amplitude of the pitch angle p(t) in the maximum slope segment of the error function to generate the pitch attitude adjustment suggestion , whose value is expressed by the formula:

[0095]

[0096] wherein, and are the time averages of the response and the prediction pitch angle, respectively, and the sign function indicates the direction.

[0097] ​For example, taking the Yangtze River revetment body sinking as an example, in the interval of t=5.5 seconds to t=8.0 seconds, the server identifies the error accumulation section through DTW, obtains the cumulative displacement offset of-0.42 meters and the positive yaw angle offset of +2.3°, and generates speed adjustment and tail attitude fine adjustment control instructions accordingly, so as to realize dynamic correction of the sinking path and the body attitude, and significantly improve the stability and precision in the construction process.

[0098] S170, based on the laying path correction value and the pitch attitude adjustment suggestion, dynamically correcting the attitude and sinking path of the unit row body.

[0099] Specifically, by obtaining complete water profile data including water depth, water temperature, water density, water salinity, water turbidity and flow gradient, a high-dimensional hydrological state tensor is constructed and input into the prediction correction model, effectively describing the nonlinear changes of the buoyancy and resistance of the body under layered disturbance conditions, and establishing a dynamic response equation that conforms to the actual working conditions. Combined with the attitude response vector flow obtained by the sensor, the sinking path offset and attitude deviation can be identified in real time, and the correction amount can be output accordingly to drive the control system to implement accurate adjustment, forming a closed-loop correction mechanism based on real-time feedback. This method breaks through the adaptability bottleneck of traditional empirical models to complex hydrological disturbances, significantly improves the sinking accuracy, attitude stability and response robustness of the unit row body in deep water and heterogeneous water resistance environment, and is suitable for high-quality revetment construction in complex hydrological scenarios, and facilitates to improve the accuracy of deep water revetment unit row construction in complex hydrological scenarios.

[0100] In one possible implementation, based on the laying path correction value and the pitch attitude adjustment suggestion, the attitude and sinking path of the unit row body are dynamically corrected, specifically including: generating a speed control instruction for adjusting the release rate of the main winch and a rhythm control instruction for correcting the laying rhythm according to the laying path correction value; generating a tension control instruction for adjusting the tension of the side anchor cable and a posture control instruction for driving the tail attitude adjustment mechanism of the body according to the pitch attitude adjustment suggestion; synchronously inputting the speed control instruction, the rhythm control instruction, the tension control instruction and the posture control instruction into the laying control terminal, and adjusting the sinking path and stabilizing the attitude of the unit row body through the main winch control device, the anchor cable control device and the body attitude execution device of the laying control terminal.

[0101] Specifically, first, the server obtains a laying path correction value, which represents the cumulative vertical offset of the unit row body compared to the predicted laying path during the current laying period. When the correction value is positive, it indicates that the actual laying speed of the row body is too slow, and the cable release rate of the main winch needs to be increased. When the correction value is negative, it indicates that the laying speed of the row body is too fast, and the cable release rate of the main winch needs to be slowed down. The server adjusts the control current or speed setting value of the main winch in proportion to the numerical value and sign direction of the correction value, generates a speed control instruction, and uses the instruction to adjust the cable release speed of the main winch in real time, so that the vertical trajectory of the row body gradually tends to the ideal path after correction. In addition, the server synchronously analyzes the change trend of the correction value in the time dimension, extracts the rhythm disturbance characteristics, such as periodic fluctuations or local abnormal acceleration, and generates a rhythm control instruction based on the existing construction standard rhythm library and disturbance identification module. The instruction is used to coordinate the release rhythm of the main winch during unit length laying, including start and stop time, slow start segment length, rhythm stabilization window, and other parameters, to further suppress system resonance or local over-stress caused by path fluctuations.

[0102] Secondly, the server obtains a pitch adjustment suggestion, which reflects the deviation between the average pitch angle of the row body during laying and the predicted value. When the suggestion is positive, it indicates that the front end of the row body is relatively high, and the tail needs to be adjusted to press down or the front buoyancy needs to be increased. When the suggestion is negative, it indicates that the front end of the row body is relatively low, and the tail needs to be lifted or the front pitch needs to be reduced. The server generates a tension control instruction accordingly, the core of which is to control the tension distribution of the left and right anchor cable systems, and to apply a pitch adjustment torque to the row body by increasing the tension on one side and reducing the tension on the other side, so as to guide it to return to a balanced state. The server calculates the tension change value and adjustment period based on the change amplitude of the suggestion, system inertia parameters, and response lag time, and uses them as input parameters to generate the tension control instruction. At the same time, the server uses the tail integrated attitude adjustment actuator of the row body to correct the angle for asymmetric disturbances that cannot be adjusted by the anchor cable. According to the amplitude of the attitude adjustment suggestion and the current attitude error, the rotation direction and angle of the attitude adjustment actuator are set, and finally the attitude control instruction is generated to accurately adjust the attitude of the row body tail by scheduling the servo motor or hydraulic actuator.

[0103] The above four types of control instructions are packaged and encoded in the server, and are transmitted in real time to the laying control terminal through the control network. The terminal schedules the main winch control device, anchor cable control device, and attitude adjustment control device to respond cooperatively, and realizes the integrated closed-loop dynamic correction of attitude-path. The instruction generation mechanism not only quickly adapts to the nonlinear response under deep water disturbance conditions, but also has good scalability and engineering deployment adaptability.

[0104] In one possible implementation, historical water depth data, historical water temperature data, historical water density data, historical water salinity data, historical water turbidity data, and historical flow velocity gradient data of a target construction area are spatially registered and reconstructed in layers to construct a historical hydrological state tensor; a unit body posture response vector flow and a control parameter sequence in a corresponding historical construction task of the target construction area are collected to construct a training sample set; based on the historical hydrological state tensor and the training sample set, an initial model is trained by using a supervised learning mechanism, taking a sinking control feedback as a target output, and by using a variational Bayesian optimization method to train resistance function parameters, dynamic equation coefficients, and control mapping weights in the initial model, so as to complete training of a prediction correction model.

[0105] Specifically, the server first performs data preprocessing on historical data of a target construction area, specifically including spatially registering historical water depth data, historical water temperature data, historical water density data, historical water salinity data, historical water turbidity data, and historical flow velocity gradient data, that is, aligning multi-source hydrological data to a consistent spatial grid based on a unified spatial reference system. Then, layer interpolation technology is used to reconstruct the data of various hydrological properties in the depth direction, ensuring the continuity and physical consistency of the data in different depth layers. After registration and reconstruction are completed, the server constructs the above data into a three-dimensional multi-channel tensor, and performs time sequence padding and sliding window segmentation processing, finally generating a historical hydrological state tensor with clear spatial, temporal, and attribute dimensions as an environmental expression for training input.

[0106] In the stage of constructing the training sample set, the server synchronously calls the corresponding unit body posture response vector flow and control parameter sequence in the historical construction task, the former including the pitch angle, roll angle, sinking speed, and attitude angular velocity of the body in each historical operation period, and the latter recording control signals such as the main winch speed, anchor cable tension, and attitude adjustment mechanism setting value at each time point. The server pairs each group of historical hydrological state tensors with its corresponding posture response vector flow and control parameter sequence to construct a standard training sample set, and uses a supervised learning mechanism to take the sinking control feedback (such as attitude deviation and path correction amount) as the target output. The server jointly optimizes the resistance function parameters, dynamic equation coefficients, and control mapping weights in the initial model by using a variational Bayesian optimization method, which takes into account parameter uncertainty expression and high-dimensional collaborative modeling capability, to ensure that the prediction correction model generated after training can accurately fit the dynamic behavior of the body under different hydrological disturbance conditions, and to realize response prediction and control strategy recommendation for the sinking process in future construction scenarios. This implementation process has realized an effective transition from an empirical model to a data-driven model in multiple actual projects, significantly improving the scene adaptation capability and prediction accuracy of the deep water body control system.

[0107] The application also provides a deep water revetment unit row construction error reverse modeling correction device, which is referred toFigure 2 , Figure 2 A module schematic diagram of a deep water revetment unit row construction error reverse modeling correction device provided for an embodiment of the present application. The device is a server, and the server comprises an acquisition module 21 and a processing module 22, wherein the acquisition module 21 acquires water profile data of a target construction area where the deep water revetment unit row is located, and the water profile data comprises water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient; the processing module 22 constructs a hydrological state tensor according to the water profile data; the processing module 22 inputs the hydrological state tensor into a prediction correction model to construct a layered water resistance field and a non-uniform speed sinking path dynamic response equation of the unit row body; the processing module 22 determines a prediction result based on the layered water resistance field and the non-uniform speed sinking path dynamic response equation; the acquisition module 21 acquires a posture response vector flow sent by a sensor device after the unit row body enters the water; the processing module 22 dynamically matches the posture response vector flow with the prediction result to output a laying path correction value and a pitch attitude adjustment suggestion amount; and the processing module 22 dynamically corrects the posture and sinking path of the unit row body based on the laying path correction value and the pitch attitude adjustment suggestion amount.

[0108] In a possible implementation, the hydrological state tensor is constructed according to the water profile data, and specifically includes: the processing module 22 reconstructs the continuity of the water profile data in the profile layer by using a spatial interpolation method, and performs smoothing processing in the depth direction by using a Gaussian kernel regression and a spline fitting method to obtain continuous tensor functions of various hydrological properties in a three-dimensional space; the processing module 22 respectively maps the continuous tensor functions of the water depth, the water temperature, the water density, the water salinity, the water turbidity and the flow velocity gradient into six tensor channels, and constructs a four-dimensional state tensor by taking spatial horizontal coordinates, vertical depth coordinates and time coordinates as indexes; and the processing module 22 performs time sequence filtering on the four-dimensional state tensor by using a time moving window processing and a convolution smoothing algorithm to obtain the hydrological state tensor.

[0109] In a possible implementation, the processing module 22 inputs the hydrological state tensor into the prediction correction model to construct a layered water body resistance field and a non-uniform sinking path dynamic response equation of the unit array, specifically including: the processing module 22 inputs the water body density tensor and the flow velocity gradient tensor in the hydrological state tensor into a resistance field construction unit in the prediction correction model, calculates the buoyancy and resistance of the unit array in each profile layer based on the water body density and the local flow velocity change rate of each depth layer, and constructs corresponding layered water body buoyancy and resistance functions; the processing module 22 constructs a layered water body resistance field based on the layered water body buoyancy and resistance functions; the processing module 22 inputs the layered water body buoyancy and resistance functions as input variables, combines the mass parameters, the gravity constant and the constant component load parameters of the unit array, and inputs them into a dynamic response modeling unit in the prediction correction model to construct a non-uniform sinking path dynamic response equation of the unit array in the vertical sinking process, the non-uniform sinking path dynamic response equation being a layered cumulative nonlinear variable coefficient second-order differential equation group, and the nonlinear variable coefficient second-order differential equation group including sinking acceleration functions, sinking velocity functions and sinking displacement functions of the unit array at each depth position and time node in the vertical space.

[0110] In a possible implementation, the processing module 22 determines the prediction result based on the layered water body resistance field and the non-uniform sinking path dynamic response equation, specifically including: the processing module 22 uses a numerical solution method to perform time step solving on the non-uniform sinking path dynamic response equation to generate acceleration functions, velocity functions and displacement functions of the unit array in a continuous time domain; the processing module 22 maps the acceleration functions, the velocity functions and the displacement functions to the depth dimension to form acceleration profiles, velocity profiles and displacement profiles, and combines to construct a state prediction matrix as the prediction result.

[0111] In a possible implementation, the processing module 22 dynamically matches the attitude response vector flow with the prediction result, and outputs the laying path correction value and the pitch attitude adjustment suggestion, specifically including: the processing module 22 constructs the attitude response vector flow into a time-synchronized attitude response data matrix, the attitude response vector flow including the pitch angle, the roll angle, the attitude angular velocity, the sinking acceleration and the vertical displacement of the unit row body during the sinking process; the processing module 22 extracts the attitude prediction reference sequence corresponding to the time-synchronized time period from the prediction result, to form the attitude prediction sequence corresponding to the attitude response vector flow; the processing module 22 inputs the attitude response data matrix and the attitude prediction sequence into the dynamic matching unit in the prediction correction model, and performs nonlinear alignment matching by using the dynamic time warping algorithm to generate a dynamic error function; the processing module 22 calculates the laying path correction value and the pitch attitude adjustment suggestion based on the dynamic error function, the laying path correction value being the cumulative offset distance of the vertical path of the unit row body, and the pitch attitude adjustment suggestion being the correction direction and amplitude of the attitude angle of the unit row body.

[0112] In a possible implementation, the processing module 22 dynamically corrects the attitude and the sinking path of the unit row body based on the laying path correction value and the pitch attitude adjustment suggestion, specifically including: the processing module 22 generates the speed control instruction for adjusting the release rate of the main winch and the rhythm control instruction for correcting the laying rhythm according to the laying path correction value; the processing module 22 generates the tension control instruction for adjusting the side anchor cable tension and the attitude control instruction for driving the tail attitude adjustment mechanism of the row body according to the pitch attitude adjustment suggestion; the processing module 22 synchronously inputs the speed control instruction, the rhythm control instruction, the tension control instruction and the attitude control instruction into the laying control terminal, and performs the sinking path fine adjustment and the attitude stability control of the unit row body through the main winch control device, the anchor cable control device and the row body attitude execution device of the laying control terminal.

[0113] In a possible implementation, the processing module 22 performs spatial registration and layered reconstruction on the historical water depth data, the historical water temperature data, the historical water density data, the historical water salinity data, the historical water turbidity data and the historical flow gradient data of the target construction area, to construct a historical hydrological state tensor; the processing module 22 collects the unit row body attitude response vector flow and the control parameter sequence in the corresponding historical construction task of the target construction area, to construct a training sample set; the processing module 22 trains the initial model by using the supervised learning mechanism, taking the sinking control feedback as the target output, and training the resistance function parameters, the dynamic equation coefficients and the control mapping weights in the initial model by using the variational Bayesian optimization method, to complete the training of the prediction correction model.

[0114] It should be noted that the apparatus provided in the above embodiments is only used as an example for dividing the above functional modules to achieve their functions, and in actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0115] The application also provides an electronic device, referring to Figure 3 , Figure 3 The application provides an electronic device. The structure of the electronic device is shown in the figure. The electronic device can include at least one processor 31, at least one network interface 34, a user interface 33, a memory 35, and at least one communication bus 32.

[0116] The communication bus 32 is used to realize the connection and communication between the components.

[0117] The user interface 33 can include a display screen (Display) and a camera (Camera). Optionally, the user interface 33 can also include a standard wired interface and a wireless interface.

[0118] The network interface 34 can optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface).

[0119] The processor 31 can include one or more processing cores. The processor 31 connects various parts of the server through various interfaces and lines, executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 35, and calling data stored in the memory 35. Optionally, the processor 31 can be realized in at least one of the hardware forms of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 31 can integrate a combination of one or several of central processing units (Central Processing Unit, CPU), graphics processing units (Graphics Processing Unit, GPU), and modems. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 31, but can be realized by a separate chip.

[0120] The memory 35 can include a Random Access Memory (RAM) and a Read-Only Memory (ROM). Optionally, the memory 35 includes a non-transitory computer-readable storage medium. The memory 35 can be configured to store instructions, programs, codes, code sets, or instruction sets. The memory 35 can include a program storage area and a data storage area. The program storage area can be configured to store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the methods described above, etc. The data storage area can be configured to store data related to the methods described above, etc. The memory 35 can optionally be at least one storage device located remotely from the processor 31. As shown, the memory 35, as a computer storage medium, can include an operating system, a network communication module, a user interface module, and an application program of a deepwater revetment unit row construction error reverse modeling correction method. Figure 3

[0121] As shown in the electronic device, the user interface 33 is mainly configured to provide an interface for user input and obtain data input by the user. The processor 31 can be configured to call the application program of the deepwater revetment unit row construction error reverse modeling correction method stored in the memory 35, and when executed by one or more processors, cause the electronic device to perform the method of one or more of the above embodiments. Figure 3

[0122] It should be noted that, for the above-mentioned method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited to the action sequence described, because according to the present application, certain steps can be performed in other sequences or at the same time. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0123] The present application also provides a computer-readable storage medium having instructions stored therein. When executed by one or more processors, the electronic device performs the method described in one or more of the above embodiments.

[0124] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0125] ​​In several embodiments provided in the present application, it should be understood that the disclosed apparatus can be implemented in other manners. For example, the division of the apparatus embodiments is merely illustrative, and the division of units can be changed according to actual conditions, such as a combination or integration of some units, or a deletion of some features, or an addition of some features. In addition, the coupling or direct coupling or communication connection between the shown or discussed units can be indirect coupling or communication connection through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0126] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0127] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0128] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, magnetic disk or optical disk, etc. Various program codes that can store program codes.

[0129] The above is only exemplary embodiments of the present disclosure, which cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon considering the specification and practicing the true principles of the present disclosure. The present application is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or conventional techniques in the art not described in the present disclosure. The specification and embodiments are only considered exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A deep water revetment unit row construction error inverse modeling correction method, characterized in that, The method comprises: obtaining water profile data of a target construction area where the deep water revetment unit row is located, the water profile data comprising water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient; constructing a hydrological state tensor according to the water profile data; inputting the hydrological state tensor into a prediction correction model to construct a layered water resistance field and a non-uniform sinking path dynamic response equation of the unit row body; determining a prediction result based on the layered water resistance field and the non-uniform sinking path dynamic response equation; obtaining an attitude response vector flow sent by a sensor device after the unit row body enters water; dynamically matching the attitude response vector flow with the prediction result to output a laying path correction value and a pitch attitude adjustment suggestion; based on the laying path correction value and the pitch attitude adjustment suggestion, dynamically correcting the attitude and sinking path of the unit row body.

2. The deep water bulkhead unit row construction error inverse modeling correction method according to claim 1, characterized in that, The method further comprises: reconstructing the continuity of the water profile data in the profile layer by using a spatial interpolation method, and performing smoothing processing in the depth direction by using a Gaussian kernel regression and spline fitting method to obtain continuous tensor functions of various hydrological properties in three-dimensional space; mapping the continuous tensor functions of the water depth, the water temperature, the water density, the water salinity, the water turbidity and the flow velocity gradient into six tensor channels respectively, and constructing a four-dimensional state tensor with spatial horizontal coordinates, vertical depth coordinates and time coordinates as indexes; performing time sequence filtering on the four-dimensional state tensor by using a time moving window processing and convolution smoothing algorithm to obtain the hydrological state tensor.

3. The deep water bulkhead unit row construction error inverse modeling correction method of claim 1, wherein, The method further comprises: inputting the water density tensor and the flow velocity gradient tensor in the hydrological state tensor into a resistance field construction unit in the prediction correction model, calculating the buoyancy and resistance of the unit row body in each profile layer based on the water density and local flow velocity change rate of each depth layer, and constructing corresponding layered water buoyancy and resistance functions; based on the layered water buoyancy and resistance functions, constructing the layered water resistance field; inputting the layered water buoyancy and resistance functions as input variables, combining the mass parameters, gravity constant and constant component load parameters of the unit row body, inputting into a dynamic response modeling unit in the prediction correction model, and constructing a non-uniform sinking path dynamic response equation of the unit row body in the vertical sinking process, the non-uniform sinking path dynamic response equation being a layered cumulative nonlinear variable coefficient second-order differential equation set, the nonlinear variable coefficient second-order differential equation set comprising sinking acceleration functions, sinking velocity functions and sinking displacement functions of the unit row body at each depth position and time node in the vertical space.

4. The deep water bulkhead unit row construction error inverse modeling correction method of claim 1, wherein, The method further comprises: The non-uniform speed drop path dynamic response equation is solved by a numerical solution method to perform time step solving, to generate acceleration functions, velocity functions and displacement functions of the unit row body in a continuous time domain; The acceleration functions, the velocity functions and the displacement functions are mapped to a depth dimension to form acceleration profiles, velocity profiles and displacement profiles, and are combined to construct a state prediction matrix as the prediction result.

5. The deep water bulkhead unit row construction error inverse modeling correction method of claim 1, wherein, The posture response vector flow and the prediction result are dynamically matched to output a laying path correction value and a pitch attitude adjustment suggestion, and specifically include: The posture response vector flow is constructed as a time-synchronized posture response data matrix, and the posture response vector flow includes a pitch angle, a roll angle, an attitude angular velocity, a sinking acceleration and a vertical displacement of the unit row body during the drop process; A posture prediction reference sequence corresponding to a time-synchronized time period is extracted from the prediction result to form a posture prediction sequence corresponding to the posture response vector flow; The posture response data matrix and the posture prediction sequence are input into a dynamic matching unit in the prediction correction model, and a dynamic time warping algorithm is used for nonlinear alignment matching to generate a dynamic error function; Based on the dynamic error function, the laying path correction value and the pitch attitude adjustment suggestion are calculated, the laying path correction value is a cumulative offset distance of the vertical path of the unit row body, and the pitch attitude adjustment suggestion is a correction direction and amplitude of the attitude angle of the unit row body.

6. The deep water bulkhead unit row construction error inverse modeling correction method of claim 1, wherein, Based on the laying path correction value and the pitch attitude adjustment suggestion, the posture and drop path of the unit row body are dynamically corrected, and specifically include: According to the laying path correction value, a speed control instruction for adjusting the release rate of the main winch and a rhythm control instruction for correcting the laying rhythm are generated; According to the pitch attitude adjustment suggestion, a tension control instruction for adjusting the tension of the side anchor cable and an attitude control instruction for driving the tail attitude adjusting mechanism of the row body are generated; The speed control instruction, the rhythm control instruction, the tension control instruction and the attitude control instruction are synchronously input into a laying control terminal, and the main winch control device, the anchor cable control device and the row body attitude executing device of the laying control terminal are used to perform drop path fine tuning and attitude stability control on the unit row body.

7. The deep water bulkhead unit row construction error inverse modeling correction method of claim 1, wherein, The method further includes: The historical water depth data, the historical water temperature data, the historical water density data, the historical water salinity data, the historical water turbidity data and the historical flow gradient data of the target construction area are spatially registered and reconstructed to construct a historical hydrological state tensor; The unit row body posture response vector flow and the control parameter sequence in the corresponding historical construction task of the target construction area are collected to construct a training sample set; Based on the historical hydrological state tensor and the training sample set, an initial model is trained by a supervised learning mechanism, with drop control feedback as a target output, by a variational Bayesian optimization method to train resistance function parameters, dynamic equation coefficients and control mapping weights, to complete training of the prediction correction model.

8. A deep water revetment unit row construction error inverse modeling correction device, characterized in that, The device comprises an acquisition module (21) and a processing module (22), wherein, The acquisition module (21) is configured to acquire water profile data of a target construction area where the deep water revetment unit row is located, and the water profile data comprises water depth, water temperature, water density, water salinity, water turbidity and flow velocity gradient; The processing module (22) is configured to construct a hydrological state tensor according to the water profile data; The processing module (22) is further configured to input the hydrological state tensor into a prediction correction model to construct a layered water resistance field and a non-uniform speed sinking path dynamic response equation of the unit row body; The processing module (22) is further configured to determine a prediction result based on the layered water resistance field and the non-uniform speed sinking path dynamic response equation; The acquisition module (21) is further configured to acquire a posture response vector flow sent by a sensor device after the unit row body enters water; The processing module (22) is further configured to dynamically match the posture response vector flow with the prediction result to output a laying path correction value and a pitch attitude adjustment suggestion amount; The processing module (22) is further configured to dynamically correct the posture and sinking path of the unit row body based on the laying path correction value and the pitch attitude adjustment suggestion amount.

9. An electronic device, comprising: The electronic device comprises a processor (31), a memory (35), a user interface (33) and a network interface (34), the memory (35) is configured to store instructions, the user interface (33) and the network interface (34) are configured to communicate with other devices, and the processor (31) is configured to execute the instructions stored in the memory (35) to enable the electronic device to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores instructions, when the instructions are executed, the method of any one of claims 1 to 7 is performed.

Images