A stability and durability analysis method for underwater structure tensor information based on memory characteristics

Abstract

The application discloses a kind of stability sustained analysis method of underwater structure tensor information based on memory characteristics;It relates to underwater measurement technical field, the application is measured to underwater structure in N time points, obtains N three-dimensional height difference tensor matrix sequence;With the dimension of one of three-dimensional height difference tensor matrix as benchmark and as benchmark matrix, matrix alignment operation is carried out to other N-1 three-dimensional height difference tensor matrix, and alignment matrix is obtained;Difference is obtained between alignment matrix and benchmark matrix, and point difference matrix is obtained;For each point difference matrix, select edge irregular area and calculate effective selected area value;Based on the three-dimensional height difference tensor matrix obtained in current stage measurement, current memory length and historical effective selected area value set, current effective selected area value is updated by linear transformation and nonlinear activation processing function, the stability change trend of underwater structure is analyzed and early warning is carried out, and the accuracy of underwater structure tensor information stability analysis is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of underwater measurement technology, and in particular to a method for the continuous stability analysis of tensor information of underwater structures based on memory characteristics. Background Technology

[0002] Underwater structures are subject to complex geological and hydrodynamic coupling effects over long periods, making them prone to scour expansion and structural instability. Once an underwater structure becomes unstable, it directly threatens the safety of the structure and personnel. Therefore, stability analysis is a core element in ensuring its safety, reliability, and durability. Commonly used methods for analyzing the stability of underwater structures include theoretical calculations, numerical simulations, and field measurements. However, theoretical calculations neglect dynamic coupling and rely heavily on empirical parameters; numerical simulations are computationally intensive and suffer from model simplification and distortion; and field measurements in deep water have low accuracy and high cost. Existing methods share common shortcomings in terms of adaptability to dynamic loads and reliability of deep-water data, necessitating optimization through the integration of intelligent algorithms and high-precision sensing technologies.

[0003] This invention proposes a method for the continuous stability analysis of tensor information of underwater structures based on memory characteristics. A multibeam echo sounder and an ROV (Remotely Operated Vehicle) are used in collaboration to acquire high-precision matrix data. By obtaining the dynamic changes of the effective selected area values ​​at each measurement stage, the stable state of the tensor information of underwater structures can be continuously analyzed. Summary of the Invention

[0004] In view of the above-mentioned prior art, the present invention provides a method for continuous stability analysis of underwater structures based on tensor information with memory characteristics, which mainly solves the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution of this invention is implemented as follows:

[0006] In a first aspect, the present invention provides a method for sustained stability analysis of underwater structures based on tensor information with memory characteristics, the method comprising the following steps:

[0007] The underwater structure was measured at N time points to obtain N sequences of three-dimensional elevation difference tensor matrices;

[0008] Using the dimension of one of the three-dimensional elevation difference tensor matrices as a reference and as the reference matrix, matrix alignment is performed on the other N-1 three-dimensional elevation difference tensor matrices to obtain the alignment matrix;

[0009] The difference between the alignment matrix and the reference matrix is ​​used to obtain the point difference matrix;

[0010] For each point difference matrix, select irregular edge regions and calculate the effective selection value;

[0011] Based on the three-dimensional elevation difference tensor matrix obtained from the current stage of measurement, the current memory length, and the set of historical valid selected areas, where the current memory length includes the set of all three-dimensional elevation difference tensors obtained from previous stages of measurement, the current valid selected areas are updated through linear transformation and nonlinear activation processing functions, specifically including:

[0012] A linear transformation is performed on the three-dimensional elevation difference tensor matrix obtained from the current stage measurement and the set of historical valid selected area values, and then input into the nonlinear activation processing function; based on the output value of the nonlinear activation processing function, a new matrix is ​​extracted from the three-dimensional elevation difference tensor matrix obtained from the current stage measurement, and the three-dimensional elevation difference tensor matrix obtained from the current stage measurement is combined with the current memory length to form a new memory length;

[0013] Extract the required data based on the new memory length and the set of historical valid selected areas, and calculate the valid selected area value for the current stage according to the predefined formula and the new matrix;

[0014] Based on the current effective selected area values, analyze the stability change trend of underwater structures and issue early warnings.

[0015] As a preferred embodiment of the present invention, the step of measuring the underwater structure at N time points to obtain a sequence of N three-dimensional elevation difference tensor matrices specifically includes:

[0016] Based on the geometric dimensions of the underwater structure to be monitored and the key monitoring areas, a three-dimensional coordinate system is established and reference benchmarks are set.

[0017] Based on the established coordinate system and reference benchmarks, underwater topographic surveying equipment is used to collect data, scan the target structure and its surrounding area, obtain raw point cloud data with three-dimensional coordinates, and perform preprocessing.

[0018] The preprocessed point cloud data is gridded using an interpolation algorithm to obtain a three-dimensional elevation matrix in the form of a regular grid.

[0019] Using the three-dimensional elevation matrix obtained from the first measurement as the reference surface matrix, the three-dimensional elevation matrix obtained from the current measurement is subtracted element by element from the reference surface matrix to obtain the three-dimensional elevation difference tensor matrix.

[0020] As a preferred embodiment of the present invention, the matrix alignment operation includes spatially aligning the three-dimensional elevation difference tensor matrix at an earlier time point with the three-dimensional elevation difference tensor matrix at a later time point to eliminate positional deviations and ensure that corresponding points are consistent.

[0021] As a preferred embodiment of the present invention, the point difference matrix represents the change in elevation difference between two time points, and is used to quantify the local changes on the surface of underwater structures.

[0022] As a preferred embodiment of the present invention, the selection of irregular edge regions includes identifying regions with significant changes in the point difference matrix, setting the elements outside the region to zero to obtain a processed matrix, and calculating the effective selection region value based on the processed matrix. The effective selection region value is obtained by statistically averaging the elements within the region.

[0023] As a preferred embodiment of the present invention, the step of performing a linear transformation on the three-dimensional elevation difference tensor matrix and the set of historical valid selected area values ​​obtained from the current stage measurement, and then inputting them into the nonlinear activation processing function, specifically includes:

[0024] The current three-dimensional elevation difference tensor matrix is ​​linearly combined using a preset first weight matrix and a first bias vector through the first transformation branch to obtain the current state feature vector.

[0025] The second transformation branch extracts at least one nearest valid selected area value from the set of historical valid selected area values, and performs a linear combination operation through a preset second weight matrix and a second bias vector to obtain the historical state feature vector.

[0026] The current state feature vector is fused with the historical state feature vector to obtain a comprehensive feature vector. The comprehensive feature vector is then input into a nonlinear activation processing function for computation to obtain at least one gating signal, including: an information retention gating signal, used to quantify the importance of historical memory; and an information update gating signal, used to quantify the novelty and importance of the current measurement information.

[0027] As a preferred embodiment of the present invention, the step of extracting a new matrix from the three-dimensional elevation difference tensor matrix obtained by the current stage measurement based on the output value of the nonlinear activation processing function, and combining the three-dimensional elevation difference tensor matrix obtained by the current stage measurement with the current memory length to form a new memory length includes:

[0028] The information update gate signal is used as a spatial attention mask to selectively focus on the current three-dimensional elevation difference tensor matrix and extract a new matrix that focuses on the significantly changing region.

[0029] The information retention gate signal is weighted and calculated with each historical 3D elevation difference tensor matrix in the current memory length. When the information retention gate signal value is lower than the preset threshold, the earliest part of the matrix in the current memory length is discarded. Then, the 3D elevation difference tensor matrix obtained from the current stage measurement is added to the current memory length to obtain a new memory length.

[0030] As a preferred embodiment of the present invention, the step of extracting the required data based on the new memory length and the set of historical valid selection area values, and calculating the valid selection area value at the current stage according to a predefined formula and a new matrix, specifically includes:

[0031] From the new memory length, select the K nearest three-dimensional elevation difference tensors to the current stage;

[0032] From the set of historical valid selected areas, select k historical valid selected areas that correspond in time to the k three-dimensional elevation difference tensor matrices;

[0033] By using linear regression to calculate the historical trend prediction value from the k historical effective selected area values, and by performing statistical calculations based on the area of ​​interest defined by the new matrix, a current instantaneous change value is obtained.

[0034] Calculate the difference norm between the historical trend prediction value and the current instantaneous change, and then weight the difference norm with k historical effective selected area values ​​to obtain the effective selected area value for the current stage.

[0035] As a preferred embodiment of the present invention, the analysis of stability change trend includes solving the overall deviation value, the range and trend information of scour or siltation, and comparing and analyzing with historical deviation values, the range and trend of scour or siltation, to achieve stability analysis and early warning of underwater structure tensor information.

[0036] Secondly, the present invention provides a stability persistence analysis system for underwater structures based on tensor information with memory characteristics, the system comprising:

[0037] The data acquisition and sequence generation module is used to measure underwater structures at N time points and obtain N three-dimensional elevation difference tensor matrix sequences.

[0038] The matrix alignment processing module is used to perform matrix alignment operations on the other N-1 three-dimensional elevation difference tensor matrices, taking the dimension of one of the three-dimensional elevation difference tensor matrices as a reference and using it as a reference matrix, to obtain an alignment matrix.

[0039] The change detection module is used to subtract the alignment matrix from the reference matrix to obtain the point difference matrix;

[0040] The key region extraction module is used to select irregular edge regions and calculate the effective selection value for each point difference matrix;

[0041] The memory feature update module is used to update the current effective selection value based on the three-dimensional elevation difference tensor matrix obtained from the current stage measurement, the current memory length, and the set of historical effective selection values. The current memory length includes the set of all three-dimensional elevation difference tensors obtained from the previous stage measurement. The current effective selection value is then updated through linear transformation and nonlinear activation processing function.

[0042] The stability early warning analysis module is used to analyze the stability change trend of underwater structures and issue early warnings based on the current effective selected area values.

[0043] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any method for continuous stability analysis of underwater structure tensor information based on memory characteristics.

[0044] Fourthly, the present invention also provides an electronic device, including a processor and a memory, the memory storing a plurality of instructions; the processor loads instructions from the memory to execute steps in any method for sustained stability analysis of underwater structure tensor information based on memory characteristics.

[0045] Fifthly, the present invention also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in any method for the continuous stability analysis of underwater structure tensor information based on memory characteristics.

[0046] The beneficial effects of this invention are as follows: Compared with the theoretical calculations and numerical simulations of traditional methods, this invention obtains high-precision data through on-site measurements, constructs a three-dimensional elevation difference tensor matrix sequence, and effectively identifies local deformation regions of underwater structures through matrix alignment and point difference calculation. Simultaneously, it introduces an adaptive update mechanism with memory characteristics, employing linear transformations and nonlinear activation functions to dynamically screen key features, overcoming the shortcomings of traditional methods in adapting to dynamic loads. This effectively improves the accuracy of tensor information stability analysis of underwater structures, making the evaluation results more precise and reliable. Attached Figure Description

[0047] Figure 1 A schematic diagram illustrating the steps of a method for sustained stability analysis of underwater structures based on tensor information with memory properties;

[0048] Figure 2 This is a schematic diagram of matrix Z;

[0049] Figure 3 A schematic diagram illustrating the principle of generating valid selected area values ​​for the current stage;

[0050] Figure 4 Example diagram for generating valid selected area values ​​at the current stage;

[0051] Figure 5 This is a schematic diagram of a system for continuous stability analysis of underwater structures based on tensor information with memory properties. Detailed Implementation

[0052] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. In the following description, the expression "some embodiments" refers to a subset of all possible embodiments; however, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0053] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0054] It should be understood that the present invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Furthermore, the terminology used herein is intended only to describe particular embodiments and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “compose” and / or “comprising,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0055] It should also be noted that when an element is referred to as being "fixed to" another element, it can be directly attached to the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0056] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0057] Firstly, this invention provides a method for the sustained stability analysis of underwater structures based on tensor information with memory characteristics. Please refer to the attached document. Figure 1 The method includes the following steps:

[0058] Step S1: Measure the underwater structure at N time points to obtain N three-dimensional elevation difference tensor matrix sequences;

[0059] As a preferred embodiment of the present invention, the step of measuring the underwater structure at N time points to obtain a sequence of N three-dimensional elevation difference tensor matrices specifically includes:

[0060] Based on the geometric dimensions of the underwater structure to be monitored and the key monitoring areas, a three-dimensional coordinate system is established and reference benchmarks are set.

[0061] Based on the established coordinate system and reference benchmarks, underwater topographic surveying equipment is used to collect data, scan the target structure and its surrounding area, obtain raw point cloud data with three-dimensional coordinates, and perform preprocessing.

[0062] In some embodiments, the underwater topographic surveying equipment includes: a multibeam echo sounder system, capable of efficiently and accurately acquiring large-area underwater topographic point cloud data; or three-dimensional acoustic scanning sonar, such as mechanical scanning sonar or three-dimensional forward-looking sonar, suitable for fine scanning of structural surfaces; or an integrated system combining GNSS (Global Navigation Satellite System) and an inertial measurement unit (IMU), which can be used to accurately determine the position and attitude of measurement vehicles, such as survey vessels and ROVs.

[0063] In this embodiment, preprocessing of the raw point cloud data includes cleaning and correction. This includes removing outliers and filtering out noise caused by debris, bubbles, or signal interference in the water. Attitude and tide level corrections are also performed, using IMU and tide gauge data to eliminate elevation changes caused by the ship's movement and tides, thus normalizing all data to a unified elevation datum.

[0064] The preprocessed point cloud data is meshed using an interpolation algorithm to obtain a three-dimensional elevation matrix in the form of a regular grid. The row and column indices of this matrix correspond to the regular grid coordinates on the horizontal plane, and the value of each element in the matrix corresponds to the elevation value at that grid coordinate.

[0065] Using the three-dimensional elevation matrix obtained from the first measurement as the reference surface matrix, the three-dimensional elevation matrix obtained from the current measurement is subtracted element by element from the reference surface matrix to obtain the three-dimensional elevation difference tensor matrix.

[0066] For example, the first measurement is performed at a certain time (e.g., the first measurement is performed in 2024), and the three-dimensional elevation difference tensor matrix A1 is obtained.

[0067]

[0068] in, Represents the first in the matrix line, number Column elements, The dimension is .

[0069] A second measurement will be performed after a stable period (e.g., in 2025) to obtain the three-dimensional elevation difference tensor matrix A2.

[0070]

[0071] in, Represents the first in the matrix line, number Column elements, The dimension is .

[0072] Step S2: Using the dimension of one of the three-dimensional elevation difference tensor matrices as a reference and as the reference matrix, perform matrix alignment operation on the other N-1 three-dimensional elevation difference tensor matrices to obtain the alignment matrix.

[0073] As a preferred embodiment of the present invention, the matrix alignment operation includes spatially aligning the three-dimensional elevation difference tensor matrix at an earlier time point with the three-dimensional elevation difference tensor matrix at a later time point to eliminate positional deviations and ensure that corresponding points are consistent.

[0074] For example, using the second three-dimensional elevation difference tensor matrix as the reference matrix, assuming that the size of matrix A1 is small, matrix A1 is shifted to a larger size; matrix A1 is interpolated and transformed to the size of matrix A2 to obtain the transformation matrix B1, as shown in the formula:

[0075]

[0076] Step S3: Subtract the alignment matrix from the reference matrix to obtain the point difference matrix;

[0077] For example, the tensor point difference matrix DELTA_A is obtained by subtracting the transformation matrix B1 and matrix A2, as shown in the formula:

[0078]

[0079] As a preferred embodiment of the present invention, the point difference matrix represents the change in elevation difference between two time points, and is used to quantify the local changes on the surface of underwater structures.

[0080] Step S4: For each point difference matrix, select irregular edge regions and calculate the effective selection area value;

[0081] As a preferred embodiment of the present invention, the selection of irregular edge regions includes identifying regions with significant changes in the point difference matrix, setting the elements outside the region to zero to obtain a processed matrix, and calculating the effective selection region value based on the processed matrix. The effective selection region value is obtained by statistically averaging the elements within the region.

[0082] For example, a region with irregular edges is selected in the difference matrix, and the elements outside the selected region are set to zero, resulting in matrix Z.

[0083] like Figure 2 As shown, assuming matrix A2 is a 4×5 matrix, the selected irregular edge region has a region range with an outer zero matrix Z, and the effective selected region value B(1,2) is calculated as follows:

[0084]

[0085] Step S5, see Figure 3 All the three-dimensional elevation difference tensor matrices obtained from the measurements are expanded according to the measurement time and labeled as T1, T2, ..., TN. During the calculation of each stage, based on the three-dimensional elevation difference tensor matrix obtained from the current stage, the current memory length, and the set of historical valid selected area values, where the current memory length includes the set of all three-dimensional elevation difference tensor matrices obtained from measurements before the current stage, the current valid selected area values ​​are updated through linear transformation and nonlinear activation processing functions, specifically including:

[0086] A linear transformation is performed on the three-dimensional elevation difference tensor matrix obtained from the current stage measurement and the set of historical valid selected area values, and then input into the nonlinear activation processing function; based on the output value of the nonlinear activation processing function, a new matrix is ​​extracted from the three-dimensional elevation difference tensor matrix obtained from the current stage measurement, and the three-dimensional elevation difference tensor matrix obtained from the current stage measurement is combined with the current memory length to form a new memory length;

[0087] Extract the required data based on the new memory length and the set of historical valid selected areas, and calculate the valid selected area value for the current stage according to the predefined formula and the new matrix;

[0088] For example, after stabilizing the cycle, a third measurement is performed to obtain the three-dimensional elevation difference tensor matrix A3. Since the size of the obtained matrix is ​​uncertain, matrix alignment is performed, transforming it to the same size as a certain matrix, such as the size of matrix A2, with the command B2, and the effective selection area is denoted as B(2,3).

[0089] Similarly, the Nth measurement is performed to obtain the three-dimensional elevation difference tensor matrix AN. If a matrix transformation is performed to form a new matrix, the command is BN, and the effective selection area is denoted as B(N-1, N).

[0090] As a preferred embodiment of the present invention, the three-dimensional elevation difference tensor matrix obtained from the current stage measurement and the set of historical valid selected area values ​​are linearly transformed and then input into a nonlinear activation processing function. This nonlinear activation processing function includes a Sigmoid function, the output range of which is (0, 1), making it highly suitable as a gating signal. The specific process includes:

[0091] The current three-dimensional elevation difference tensor matrix is ​​linearly combined using a preset first weight matrix and a first bias vector through the first transformation branch to obtain the current state feature vector.

[0092] The second transformation branch extracts at least one nearest valid selected area value from the set of historical valid selected area values, and performs a linear combination operation through a preset second weight matrix and a second bias vector to obtain the historical state feature vector.

[0093] The current state feature vector is fused with the historical state feature vector to obtain a comprehensive feature vector. The comprehensive feature vector is then input into a nonlinear activation processing function for computation to obtain at least one gating signal, including: an information retention gating signal, used to quantify the importance of historical memory; and an information update gating signal, used to quantify the novelty and importance of the current measurement information.

[0094] As a preferred embodiment of the present invention, the step of extracting a new matrix from the three-dimensional elevation difference tensor matrix obtained by the current stage measurement based on the output value of the nonlinear activation processing function, and combining the three-dimensional elevation difference tensor matrix obtained by the current stage measurement with the current memory length to form a new memory length includes:

[0095] The information update gate signal is used as a spatial attention mask to selectively focus on the current three-dimensional elevation difference tensor matrix and extract a new matrix that focuses on the significantly changing region.

[0096] The information retention gate signal is weighted and calculated with each historical 3D elevation difference tensor matrix in the current memory length. When the information retention gate signal value is lower than the preset threshold, the earliest part of the matrix in the current memory length is discarded. Then, the 3D elevation difference tensor matrix obtained from the current stage measurement is added to the current memory length to obtain a new memory length.

[0097] As a preferred embodiment of the present invention, the step of extracting the required data based on the new memory length and the set of historical valid selection area values, and calculating the valid selection area value at the current stage according to a predefined formula and a new matrix, specifically includes:

[0098] From the new memory length, select the K nearest three-dimensional elevation difference tensors to the current stage;

[0099] From the set of historical valid selected areas, select k historical valid selected areas that correspond in time to the k three-dimensional elevation difference tensor matrices;

[0100] By using linear regression to calculate the historical trend prediction value from the k historical effective selected area values, and by performing statistical calculations based on the area of ​​interest defined by the new matrix, a current instantaneous change value is obtained.

[0101] Calculate the difference norm between the historical trend prediction value and the current instantaneous change, and then weight the difference norm with k historical effective selected area values ​​to obtain the effective selected area value for the current stage.

[0102] For example, see Figure 4 The three-dimensional elevation difference tensor matrix obtained from the four measurements is expanded according to time and labeled as T1, T2, T3, and T4, respectively.

[0103] In stage T1, matrix A1 is obtained. Since stage T1 is the initial measurement stage, the memory length and effective selection area value are zero. Therefore, the input of this stage is only matrix A1, and the output is also matrix A1.

[0104] In stage T2, matrix A2 is obtained. Matrix A2, matrix A1 obtained in the previous stage, and manually selected matrix Z are used as inputs. A linear transformation is performed on matrices A2 and A1, and the result is input into a nonlinear activation processing function. The function output value determines whether to retain matrix A1, i.e., the function output value is multiplied by the historical matrix. If the function outputs 1 (retain) or 0 (not retain), the original matrix A2 is compressed by the function to generate a new matrix (which can be named matrix B2). The original matrix is ​​then combined with matrix A1 from the previous stage to form a new memory length. Then, the required matrix is ​​selected from the new memory length and calculated with the manually selected matrix Z and matrix B2 to obtain the effective selection value (B(1,2)) for stage T2.

[0105] Similarly, in stage T3, matrix A3 is obtained. Using A3, A1, A2, and B(1,2) as input, a linear transformation is performed on matrix A3 and the effective selection value B(1,2) from stage T2. After processing by a nonlinear activation function, a new matrix B3 is generated. The original matrix A3 is then combined with matrices A1 and A2 to form a new memory length. Next, the required data is selected from matrices A1, A2, A3, and the effective selection value B(1,2) and calculated with the new matrix B3 to generate the effective selection value B(2,3) for stage T3. Finally, the effective selection value B(2,3) for stage T3 is output. This process is repeated until all calculations in stage T4 are completed, at which point the final effective selection value is output.

[0106] Step S6: Based on the current effective selected area values, analyze the stability change trend of underwater structures and issue an early warning.

[0107] As a preferred embodiment of the present invention, the analysis of stability change trend includes solving the overall deviation value, the range and trend information of scour or siltation, and comparing and analyzing with historical deviation values, the range and trend of scour or siltation, to achieve stability analysis and early warning of underwater structure tensor information.

[0108] In this embodiment, the analysis and determination of the stability of underwater structures includes the following:

[0109] (1) Judgment 1, Overall Deviation Judgment:

[0110] After long-term monitoring, a range of historical data of effective selected area values ​​of an underwater structure during a stable period is selected and its standard deviation and variance are calculated. Based on the standard deviation and variance, a stable interval is set. The effective selected area value B(N-1, N) is compared with the stable interval to obtain the overall deviation.

[0111] For example, if the effective selected region value B(N-1, N) falls within the stable interval, it is determined to be "overall stable".

[0112] If the effective selected area value B(N-1, N) is lower than the lower limit of the stable interval, it is judged as "overall slight deviation".

[0113] If the effective selected region value B(N-1, N) exceeds the upper limit of the stable interval, it is judged as "significant overall deviation".

[0114] (2) Judgment 2, Judgment of the same trend of change:

[0115] The determination of whether there is the same trend is made by calculating the sign difference of the effective selection area value B(N-1, N). For example, if a short-term, continuous effective selection area value sequence is selected, the slope of the sequence is calculated. If multiple consecutive effective selection area values ​​have the same sign (all positive or all negative) and the absolute value shows a non-decreasing trend, it is determined that "there is a unidirectional continuous trend".

[0116] For example, if all values ​​are negative (indicating the direction of scouring) and the absolute values ​​continue to increase, it is determined that "the trend is consistent, and there is a risk of continued accelerated scouring." This is a key warning signal; even if the overall deviation has not yet exceeded the limit, this trend indicates that the situation is deteriorating.

[0117] (3) Judgment 3, judgment of scour or siltation volume:

[0118] The range of scour or siltation can be obtained by calculating the volume of the effective selected area value B(N-1, N).

[0119] For example, within the valid selection area defined by the point difference matrix DELTA-A(N-1,N), the elevation variation value for each grid cell. Multiply by the area dA of the cell. And for all cells... Summing these values ​​yields the total volume change, V.

[0120] like , indicates the volume of siltation; , represents the scouring volume (absolute value).

[0121] The calculated volume V is then compared with the preset engineering safety threshold.

[0122] Therefore, this invention solves the technical problem of insufficient data accuracy in deep-water environments by constructing a three-dimensional elevation difference tensor matrix sequence. Through matrix alignment and point difference calculation, it effectively identifies local deformation regions of underwater structures. By introducing an adaptive update mechanism with memory characteristics and using linear transformation and nonlinear activation functions to dynamically screen key features, it overcomes the shortcomings of traditional methods in adapting to dynamic loads. Finally, by analyzing and judging the stability of underwater structures, it achieves stability early warning, enabling continuous and accurate analysis of the stability state of underwater structures.

[0123] Secondly, this invention provides a stability continuous analysis system for underwater structures based on tensor information with memory characteristics. Please refer to the attached document for details. Figure 5 The system includes:

[0124] The data acquisition and sequence generation module is used to measure underwater structures at N time points and obtain N three-dimensional elevation difference tensor matrix sequences.

[0125] The matrix alignment processing module is used to perform matrix alignment operations on the other N-1 three-dimensional elevation difference tensor matrices, taking the dimension of one of the three-dimensional elevation difference tensor matrices as a reference and using it as a reference matrix, to obtain an alignment matrix.

[0126] The change detection module is used to subtract the alignment matrix from the reference matrix to obtain the point difference matrix;

[0127] The key region extraction module is used to select irregular edge regions and calculate the effective selection value for each point difference matrix;

[0128] The memory feature update module is used to update the current effective selection value based on the three-dimensional elevation difference tensor matrix obtained from the current stage measurement, the current memory length, and the set of historical effective selection values. The current memory length includes the set of all three-dimensional elevation difference tensors obtained from the previous stage measurement. The current effective selection value is then updated through linear transformation and nonlinear activation processing function.

[0129] The stability early warning analysis module is used to analyze the stability change trend of underwater structures and issue early warnings based on the current effective selected area values.

[0130] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any method for continuous stability analysis of underwater structure tensor information based on memory characteristics.

[0131] Fourthly, the present invention also provides an electronic device, including a processor and a memory, the memory storing a plurality of instructions; the processor loads instructions from the memory to execute steps in any method for sustained stability analysis of underwater structure tensor information based on memory characteristics.

[0132] In this embodiment, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0133] Fifthly, the present invention also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in any method for the continuous stability analysis of underwater structure tensor information based on memory characteristics.

[0134] In this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be accomplished by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0135] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the stability persistence analysis methods for underwater structures based on memory characteristics provided in embodiments of this application.

[0136] It should be noted that, through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0137] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for stability and durability analysis of underwater structure tensor information based on memory characteristics, characterized by, The method comprises the following steps: measuring the underwater structure at N time points to obtain a sequence of N three-dimensional elevation difference tensor matrices; performing matrix alignment on the other N-1 three-dimensional elevation difference tensor matrices with the dimensions of one of the three-dimensional elevation difference tensor matrices as a reference and as a reference matrix to obtain aligned matrices; the matrix alignment operation comprises spatially aligning the three-dimensional elevation difference tensor matrix at an earlier time point with the three-dimensional elevation difference tensor matrix at a later time point to eliminate positional deviation and ensure the correspondence of points; differencing the aligned matrices from the reference matrix to obtain a point difference matrix; the point difference matrix represents the change in elevation difference between two time points and is used to quantify local changes in the surface of the underwater structure; for each point difference matrix, selecting an irregular region at the edge thereof and calculating an effective selected area value; based on the three-dimensional elevation difference tensor matrix obtained in the current stage, the current memory length and the set of historical effective selected area values, the current memory length comprising a set of all three-dimensional elevation difference tensor matrices obtained by measurement before the current stage, updating the current stage effective selected area value through linear transformation and nonlinear activation processing function, specifically comprising: performing linear transformation on the three-dimensional elevation difference tensor matrix obtained in the current stage and the set of historical effective selected area values, and then inputting them into the nonlinear activation processing function; according to the output value of the nonlinear activation processing function, extracting a new matrix from the three-dimensional elevation difference tensor matrix obtained in the current stage, and combining the three-dimensional elevation difference tensor matrix obtained in the current stage with the current memory length to form a new memory length; according to the new memory length and the required data extracted from the set of historical effective selected area values, calculating the current stage effective selected area value according to a predefined formula and the new matrix; based on the current stage effective selected area value, analyzing the stability change trend of the underwater structure and giving a warning.

2. The method of claim 1, wherein the method is characterized by: The measurement of the underwater structure at N time points to obtain a sequence of N three-dimensional elevation difference tensor matrices specifically comprises: establishing a three-dimensional coordinate system and setting a reference datum point according to the geometric size and key monitoring area of the underwater structure to be monitored; using an underwater topographic surveying device to collect data and scan the target structure and its surrounding area according to the established coordinate system and reference datum point, to obtain raw point cloud data with three-dimensional coordinates and perform preprocessing; performing gridding processing on the preprocessed point cloud data through an interpolation algorithm to obtain a three-dimensional elevation matrix in a regular grid form; subtracting the three-dimensional elevation matrix obtained in the current measurement from the three-dimensional elevation matrix obtained in the first measurement element by element to obtain a three-dimensional elevation difference tensor matrix.

3. The method of claim 1, wherein the method is characterized by: The selection of the irregular region at the edge thereof comprises identifying a region with significant changes in the point difference matrix, setting the elements outside the region to zero to obtain a processed matrix, and calculating the effective selected area value based on the processed matrix, the effective selected area value being obtained by counting the average value of the elements in the region.

4. The method of claim 1, wherein the method is characterized by: the linear transformation on the three-dimensional elevation difference tensor matrix obtained in the current stage and the set of historical effective selected area values, and then inputting them into the nonlinear activation processing function, specifically comprising: The current three-dimensional elevation difference tensor matrix is obtained through the first transformation branch, linear combination operation is performed through a preset first weight matrix and a first bias vector, and a current state feature vector is obtained; The nearest neighbor at least one effective selected area value is extracted from the historical effective selected area value set through the second transformation branch, linear combination operation is performed through a preset second weight matrix and a second bias vector, and a historical state feature vector is obtained; The current state feature vector and the historical state feature vector are fused to obtain a comprehensive feature vector, and the comprehensive feature vector is input into a nonlinear activation processing function for operation to obtain at least one gate signal, including: an information retention gate signal for quantifying the importance degree of historical memory; An information update gate signal for quantifying the novelty and importance degree of the current measurement information.

5. The method of claim 4, wherein the method is characterized by: According to the output value of the nonlinear activation processing function, a new matrix is extracted from the three-dimensional elevation difference tensor matrix obtained in the current stage measurement, and the three-dimensional elevation difference tensor matrix obtained in the current stage measurement is combined with the current memory length to form a new memory length, including: The information update gate signal is used as a spatial attention mask to selectively focus on the three-dimensional elevation difference tensor matrix in the current stage, and a new matrix focused on the significant change area is extracted; The information retention gate signal is weighted with each historical three-dimensional elevation difference tensor matrix in the current memory length, and when the information retention gate signal value is lower than a preset threshold, the earliest part of the matrix in the current memory length is discarded, and then the three-dimensional elevation difference tensor matrix obtained in the current stage measurement is added to the current memory length to obtain a new memory length.

6. The method of claim 5, wherein the method is characterized by, According to the new memory length and the required data extracted from the historical effective selected area value set, the current stage effective selected area value is calculated according to a predefined formula and the new matrix, specifically including: Selecting the K three-dimensional elevation difference tensor matrices closest to the current stage from the new memory length; Selecting the k historical effective selected area values corresponding to the K three-dimensional elevation difference tensor matrices in time from the historical effective selected area value set; Calculating the k historical effective selected area values by linear regression to obtain a historical trend prediction value, and calculating a current instantaneous change value according to the attention area defined by the new matrix; The difference norm of the historical trend prediction value and the current instantaneous change value is calculated, and the difference norm is weighted with the k historical effective selected area values to obtain the current stage effective selected area value.

7. The method of claim 1, wherein the method is characterized by: The analysis of the stability change trend includes solving the overall deviation value, the scouring or silting range and the change trend information, and comparing and analyzing the historical deviation value, the scouring or silting range and the change trend state to realize the stability analysis and early warning of the underwater structure tensor information.

Images