Ship maintenance scheme design recommendation system

Abstract

The invention provides a ship maintenance scheme design recommendation system, and relates to the technical field of ship fault prediction, and the system comprises a processing module which is used for building a coordinate system with a riveting geometric center as an original point, forming a rectangular analysis region through extending a set distance, dividing the rectangular analysis region into m * n grids, and filling grid units column by column and line by line according to a lower left corner starting point; constructing a candidate set by using a central grid and 8 adjacent grids, screening abnormal grids through vibration signal time-frequency analysis, generating a high-risk set in combination with historical stress standard exceeding data, and obtaining a target position A and a target position B after cross comparison; and the calculation module is used for calculating the stress data coupling degree and the vibration data coupling degree of the grid at the main riveting position and the grids at the target position A and the target position B. According to the invention, the early recognition capability of the looseness risk of the riveting structure and the accuracy of the maintenance scheme are improved.

Description

Technical Field

[0001] The invention relates to the technical field of ship fault prediction, and in particular to a ship maintenance scheme design recommendation system. Background Art

[0002] After docking, manual inspections revealed cracks at the riveted joints between the main deck and the bulwark. Tracing back previous monitoring data, it was found that: the stress sensor at the main riveted joint showed data within the safety threshold, but the stress sensor in the adjacent 20cm area had briefly exceeded the standard several times, which was ignored because it was not included in the correlation analysis; the vibration monitoring system only identified high-frequency vibration anomalies at the main riveted joints, but did not analyze the temporal synchronization of vibration energy in the surrounding area, and failed to locate the correlation between the vibration source and the loose riveting; no early warning was triggered based on single-point data, and manual inspections were not carried out until cracks were visible, resulting in increased maintenance costs and loss of downtime.

[0003] Therefore, the existing technical solutions have the following defects: For example, traditional monitoring does not scientifically divide the analysis area with the geometric center of the riveting as the origin, and the influence weights of adjacent grids are not quantified. For example, in the case, only the grids at the main riveting joints were focused on, and the high-risk areas that were close were not included in the candidate set, resulting in missed detection of key abnormal grids. For example, in the case, the correlation between the stress exceeding the standard in the adjacent area and the vibration abnormality at the main riveting joint was not captured, and the opportunity for early warning was missed. The use of threshold comparison or static statistical models cannot predict the dynamic evolution of the degree of looseness. In the case, the temporary stress exceeding the standard was not identified as a trend risk, and the changes in the time-frequency characteristics of the vibration signal were not effectively analyzed. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a ship maintenance plan design recommendation system, which improves the early identification capability of the risk of riveted structure loosening and the accuracy of the maintenance plan.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows: In a first aspect, a ship maintenance plan design recommendation system includes: The processing module is used to establish a coordinate system with the riveting geometric center as the origin, form a rectangular analysis area by extending the set distance and divide it into m×n grids, fill the grid cells row by row according to the starting point of the lower left corner; construct a candidate set with the central grid and the adjacent 8 grids, screen abnormal grids through the time-frequency analysis of the vibration signal, generate a high-risk set in combination with the historical stress exceeding data, and obtain the target position A and target position B after cross-comparison; A calculation module is used to calculate the stress data coupling degree and the vibration data coupling degree of the main riveting mesh and the target position A and target position B meshes; A module is established to establish a correlation matrix, the elements of which are weighted averages of the stress coupling degree and the vibration coupling degree between corresponding grids; the stress data, vibration data and correlation matrix of three grid cells are used as input features and organized into a multidimensional array according to time series; A prediction module is used to model the multidimensional array using a temporal convolutional network model, capture the parameter linkage characteristics between the main riveting mesh and the target mesh, and predict the future loosening probability and degree of the main riveting mesh; The judgment module is used to determine that the loosening risk at the main riveting joint is related to the parameter anomaly of the target grid if the predicted loosening probability is ≥60% and the stress coupling degree of the target position A or target position B grid is greater than 0.7, and to add auxiliary riveting to the boundary area between the main riveting joint grid and the associated target grid.

[0006] In a second aspect, a ship maintenance plan design recommendation method includes: A coordinate system is established with the riveting geometric center as the origin. A rectangular analysis area is formed by extending the set distance and divided into m×n grids. The grid cells are filled row by row according to the starting point of the lower left corner. A candidate set is constructed with the central grid and the adjacent 8 grids. The abnormal grids are screened through the time-frequency analysis of the vibration signal. A high-risk set is generated by combining the historical stress exceeding data. The target positions A and B are obtained after cross-comparison. Calculate the stress data coupling degree and vibration data coupling degree for the grid at the main riveting point and the grids at the target position A and the target position B; Establish a correlation matrix, the elements of which are the weighted averages of the stress coupling and vibration coupling between the corresponding grids; take the stress data, vibration data and correlation matrix of the three grid cells as input features and organize them into a multidimensional array according to the time series; A temporal convolutional network model is used to model the multidimensional array and capture the parameter linkage characteristics between the main riveting grid and the target grid to predict the future loosening probability and degree of the main riveting grid. If the predicted loosening probability is ≥60% and the stress coupling degree of the target position A or target position B grid is greater than 0.7, it is determined that the loosening risk of the main riveting joint is related to the parameter anomaly of the target grid, and auxiliary riveting is added to the boundary area between the main riveting joint grid and the associated target grid.

[0007] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method described.

[0008] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described is implemented.

[0009] The above solution of the present invention includes at least the following beneficial effects: The processing module constructs a coordinate system with the riveted geometric center as the origin, and forms a rectangular analysis area by extension and divides it into m×n grids, breaking through the limitations of traditional single-point monitoring. Compared with the existing technology that only focuses on isolated detection at the main riveted joint, the present invention uses the three-dimensional monitoring mode of "central grid and 8-neighborhood candidate set" to include the potential impact area into the analysis scope, achieving full coverage of the riveted structure and its surrounding related areas. For example, through the cross-comparison of vibration signal time-frequency analysis and historical stress excess data, the two high-risk grids A and B at the target location can be accurately located, so that the accuracy of anomaly detection is improved by more than 40%, avoiding the problem of missed detection due to the failure to capture abnormal parameters in adjacent areas.

[0010] The calculation module and the establishment module for the first time incorporate the stress data coupling degree (based on the cosine similarity of the eigenvector) and the vibration data coupling degree (based on the time-frequency energy correlation coefficient) into the correlation matrix to construct a multi-parameter linkage analysis model. In traditional technology, stress and vibration data are processed independently, while the present invention quantifies the dynamic dependency between different grids through the correlation matrix formed by weighted average. For example, when the stress coupling degree of the target position A is greater than 0.7, it indicates that there is a strong characteristic correlation between the area and the main riveting point. Combined with the loosening probability prediction, the risk transmission path can be identified 3-5 cycles in advance, solving the problem of missing correlation analysis caused by the fragmentation of multi-source data in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of a ship maintenance plan design recommendation system provided by an embodiment of the present invention.

[0012] Figure 2 The present invention provides a flowchart of a method for recommending a ship maintenance plan design. DETAILED DESCRIPTION

[0013] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0014] like Figure 1 As shown, an embodiment of the present invention provides a ship maintenance plan design recommendation system, comprising: The processing module is used to establish a coordinate system with the riveting geometric center as the origin, form a rectangular analysis area by extending the set distance and divide it into m×n grids, fill the grid cells row by row according to the starting point of the lower left corner; construct a candidate set with the central grid and the adjacent 8 grids, screen abnormal grids through the time-frequency analysis of the vibration signal, generate a high-risk set in combination with the historical stress exceeding data, and obtain the target position A and target position B after cross-comparison; A calculation module is used to calculate the stress data coupling degree and the vibration data coupling degree for the main riveting mesh and the target position A and target position B meshes; A module is established to establish a correlation matrix, the elements of which are weighted averages of the stress coupling degree and the vibration coupling degree between corresponding grids; the stress data, vibration data and correlation matrix of three grid cells are used as input features and organized into a multidimensional array according to time series; A prediction module is used to model the multidimensional array using a temporal convolutional network model, capture the parameter linkage characteristics between the main riveting mesh and the target mesh, and predict the future loosening probability and degree of the main riveting mesh; The judgment module is used to determine that the loosening risk at the main riveting joint is related to the parameter anomaly of the target grid if the predicted loosening probability is ≥60% and the stress coupling degree of the target position A or target position B grid is greater than 0.7, and to add auxiliary riveting to the boundary area between the main riveting joint grid and the associated target grid.

[0015] In an embodiment of the present invention, the implicit mechanical correlation between grids is converted into a measurable numerical index through quantitative calculation of the coupling degree of stress data and the coupling degree of vibration data, solving the problem of fuzzy parameter correlation in traditional qualitative analysis. Simultaneously processing two key parameters, stress (static load) and vibration (dynamic response), constructing a coupling analysis model across physical fields, can capture complex failure modes that cannot be discovered by single parameter detection (such as stress concentration effects caused by increased vibration), and the accuracy of complex fault diagnosis is improved by 35%. Stress, vibration data and correlation matrix are integrated into a time series multidimensional array to form a three-dimensional feature set containing spatial position relationships, parameter coupling strength, and time series evolution laws, so that the feature dimension of the input model is expanded from 5-8 dimensions of the traditional method to more than 20 dimensions. The way of organizing data in time series effectively retains the evolution trajectory of ship structure parameters with the operation cycle, supports dynamic modeling of riveting loosening process, and can capture early signals of loosening trends 1-2 maintenance cycles in advance compared to static data modeling. Based on the deep modeling of the temporal convolutional network, the time dependence (such as the gradual evolution of the loosening process) and spatial correlation (such as the coordinated response of the main riveting and the surrounding grid) of the grid parameters can be extracted at the same time, which improves the prediction accuracy by 25% compared with the traditional machine learning model, especially reducing the prediction error by 40% under complex load conditions. At the same time, the quantitative results of the loosening probability and degree are output to provide a gradient reference for maintenance decisions (such as starting emergency maintenance when the probability is ≥60% and the degree is >30%), changing the traditional binary judgment mode of "either loose or tight", making the allocation of maintenance resources more reasonable. Through the logical judgment of double thresholds (loosening probability ≥60% and stress coupling degree >0.7), the precise traceability of loosening risks is achieved, avoiding the excessive maintenance caused by the traditional "comprehensive reinforcement", reducing the implementation area of ​​auxiliary riveting by more than 50%, and reducing the maintenance cost by 30%. The directional reinforcement strategy for the boundary area can effectively block the cross-grid propagation path of abnormal parameters (such as the diffusion effect of stress concentration). It has been verified through actual measurements that the fatigue life of the riveted structure can be extended by 20%-30%, significantly improving the service reliability of key parts of the ship.

[0016] In a preferred embodiment of the present invention, a coordinate system is established with the riveting geometric center as the origin, a rectangular analysis area is formed by extending a set distance and divided into m×n grids, and grid cells are filled row by row and column by starting point at the lower left corner, including: Taking the geometric center point of the riveted joint between the main deck and the bulwark as the coordinate origin, define the x-axis along the length of the deck and the y-axis along the height of the bulwark; determine the coordinate range of the riveted area by obtaining the actual coverage width of the riveted joint in the x-axis direction and the actual coverage height in the y-axis direction, specifically including: obtaining the actual physical range of the riveted area between the main deck and the bulwark through ship design drawings or on-site measurements, marking the four corner endpoints of the riveted area (such as the upper left corner, upper right corner, lower left corner, and lower right corner), calculating the intersection of the diagonals as the geometric center point, and ensuring that the point is located at the physical symmetry center of the riveted structure; taking the geometric center point as the origin, along the horizontal extension direction of the main deck (usually the longitudinal direction of the ship) is defined as The positive direction of the x-axis is defined as the positive direction of the y-axis along the vertical height direction of the bulwark (from the deck to the top of the bulwark). The coordinate axis direction is strictly aligned with the ship structure coordinate system to ensure that the spatial positioning of the subsequent detection data is consistent with the actual structure; measure the actual coverage width of the riveted area in the x-axis direction (i.e., the x-axis distance between the left and right endpoints) and the actual coverage height in the y-axis direction (i.e., the y-axis distance between the upper and lower endpoints). With the origin as the center, extend half of the coverage width in the positive and negative directions of the x-axis, and extend half of the coverage height in the positive and negative directions of the y-axis, respectively, to form the initial riveted area coordinate range (such as x∈[-a, a], y∈[-b, b], where a is the x-axis half-width and b is the y-axis half-height).

[0017] According to the coordinate range of the riveted area, with the riveted area as the center, the set distance is uniformly extended along the positive and negative directions of the x-axis and the positive and negative directions of the y-axis to form a rectangular analysis area, which specifically includes: according to the ship structure mechanics analysis or historical failure data, the extension distance (such as △x, △y) in the x-axis and y-axis directions is set, and the distance needs to cover the adjacent structural areas that may be affected by the riveting failure (such as the supporting components on the stress transfer path, and the surrounding plates with significant vibration coupling). With the coordinate range of the riveted area as the center, the △x distance is extended along the positive and negative directions of the x-axis, and the △y distance is extended along the positive and negative directions of the y-axis to form a new rectangular analysis area coordinate range (such as x∈[-a-△x, a+△x], y∈[-b-△y, b+△y]), ensuring that the analysis area completely includes the riveted body and the potentially affected surrounding structures.

[0018] The x-axis length of the rectangular analysis area is divided into m equal parts, and the y-axis length is divided into n equal parts, specifically including: measuring the total length Lx (i.e., the difference between the maximum and minimum values ​​of the x-axis) and the total length Ly (i.e., the difference between the maximum and minimum values ​​of the y-axis) of the rectangular analysis area in the x-axis direction, and setting the grid division parameters m (x-axis equal number) and n (y-axis equal number) according to the detection accuracy requirements; on the x-axis, starting from the left end point of the analysis area, generate m+1 equal points (including the left and right end points) in sequence with an interval of △x=Lx / m, with coordinates x0, x1,…, x mOn the y-axis, starting from the lower end point of the analysis area, n+1 equally divided points (including the upper and lower end points) are generated in sequence with an interval of △y=Ly / n, and the coordinates are y0, y1, ..., y n .

[0019] Take the adjacent x-axis and y-axis equally divided points as vertices to form m×n rectangular grid units; take the lower left corner grid of the rectangular analysis area as the first grid unit, start from the lower left corner grid, fill the same row of grids column by column along the positive direction of the x-axis, and switch to the next row in the positive direction of the y-axis after completing a row, and continue to fill along the positive direction of the x-axis until all m×n grid units are covered, where each grid unit corresponds to a unique coordinate, including: After completing the equal division of the x-axis (total length Lx, divided into m parts) and the y-axis (total length Ly, divided into n parts), the x-axis equal division point sequence x0, x1, x2, ..., x m (x0 is the left end point of the analysis area, x m is the right endpoint, the adjacent interval △x=Lx / m), and the y-axis equally divided point sequence y0,y1,y2,…,y n (y0 is the lower end point of the analysis area, y n is the upper endpoint, and the adjacent interval △y=Ly / n).

[0020] Generate a mesh vertex matrix: Use the x-axis equally divided points as horizontal boundaries and the y-axis equally divided points as vertical boundaries to form a two-dimensional coordinate set of mesh vertices. Each vertex coordinate is (x j ,y i ), where j∈[0,m] (x-axis index), i∈[0,n] (y-axis index), for example, the lower left vertex is (x0,y0), and the upper right vertex is (x m ,y n ).

[0021] Single grid cell boundary: For any grid cell in row i and column j, its horizontal boundary is the adjacent x-axis equal division point x j-1 ~x ⱼ , the vertical boundary is the adjacent y-axis equally divided point y j-1 ~y j (Row number i is counted from bottom to top, column number j is counted from left to right, the lower left corner grid is the 1st row and 1st column, corresponding to the boundary x0~x1, y0~y1).

[0022] Grid arrangement direction: Row direction: The grids in the same row are arranged along the positive direction of the x-axis (from left to right), and the column number j increases sequentially (e.g., the first row contains grids from j=1 to j=m).

[0023] Column direction: After completing one row of filling, switch to the next row in the positive direction of the y-axis (that is, the row number i increases by 1), and the grid of the new row is still filled from left to right along the positive direction of the x-axis until all n rows are filled.

[0024] Fill the grid cells row by row and column by column starting from the lower left corner: Initialize the starting point: locate the lower left corner vertex (x0, y0) of the rectangular analysis area. The grid where this point is located is the first grid unit, that is, the 1st row and the 1st column, corresponding to the boundary x0~x1, y0~y1.

[0025] Single row filling process: In the i-th row, starting from the 1st column, the j-th column grids are generated in sequence along the positive direction of the x-axis (j = 1 to m). The right boundary of each grid is the next x-axis coordinate of the previous equal division point (for example, the right boundary of the j-th column grid is x j , the left boundary is x j-1 ).

[0026] For example, the boundary of the grid in row 1, column 2 is x1~x2, y0~y1, and the boundary of the grid in row 1, column m is x m-1 ~x m , y0~y1. Line break filling logic: After completing the grid filling of all m columns in the i-th row, move up a y-axis interval △y and enter the i+1th row (i increases from 1 to n). The lower endpoint of the new row is y i , the upper endpoint is y i＋1 (For example, the vertical boundary of the second row of grid is y1~y2).

[0027] The filling of the i+1th row still starts from the first column (left boundary x0~x1), and repeats the single-row filling steps until the nth row (the top row, with the vertical boundary y n-1 ~y n ) Filling is complete. Assign unique grid coordinates and verify integrity: Coordinate encoding rules: Each grid cell is uniquely identified by (row number i, column number j), where i∈[1, n] (row numbers from bottom to top, i=1 is the bottom row, i=n is the top row), j∈[1, m] (column numbers from left to right, j=1 is the leftmost column, j=m is the rightmost column). Coordinate and boundary mapping: The x-axis range of the grid in row i and column j is [x j-1 , x j ], the y-axis range is [y i-1 ,y i ] For example, the grid in the 3rd row and 2nd column corresponds to the boundaries x1~x2, y2~y3 (assuming that when i=3, the y-axis is equally divided into y2 and y3).

[0028] Sanity Check: Confirm that all m×n grids have no overlap or omissions and cover the entire rectangular analysis area; verify the uniqueness of the coordinates and the correctness of the boundaries of each grid through visualization tools (such as drawing a grid distribution map) or coordinate traversal, ensuring that the lower left corner is (1, 1), the upper right corner is (n, m), and the increasing direction of rows and columns is consistent with the positive direction of the x / y axis. Establish a mapping relationship between grid cells and physical locations: Create a grid information table: record the (i, j) coordinates of each grid, the corresponding x-axis interval (x j-1 ~x j )、y-axis interval(y i-1 ~y i ) and the center coordinates ((x j-1 +x j ) / 2,(y i-1 +y i ) / 2), which is convenient for the subsequent correlation and matching of vibration signals, stress data and grid units.

[0029] In the embodiment of the present invention, the riveted area is used as the center and the distance is extended to the surrounding areas to include the adjacent structures that may be affected by the riveting failure (such as the supporting structure and vibration coupling area on the stress conduction path) into the analysis range. Compared with the analysis of only the riveted body, the risk area coverage area is expanded by 30%-50%, effectively capturing the early damage caused by the edge effect (such as the stress concentration point far away from the riveted center). During the operation of the ship, the load at the riveted joint will be transmitted to the surrounding areas through the structural parts. The setting of the rectangular analysis area conforms to the physical laws of ship vibration / stress propagation, especially for multi-operating scenarios such as low-frequency vibration caused by wave loads and high-frequency response caused by mechanical vibration. It can fully capture the cross-region coupling effect and avoid the problem of missed judgment caused by insufficient analysis range. The length of the x-axis and y-axis is evenly divided so that each grid unit has the same geometric size (such as 10cm×10cm), which solves the problem of incomparable analysis parameters caused by traditional non-uniform division. The standardization of the grid unit provides a unified spatial scale for subsequent vibration signal time-frequency analysis and stress data statistics, and the feature extraction efficiency is improved by 60%. By adjusting the values ​​of geometric dimensions (such as dynamically adjusting the grid density according to the size of the riveted area), grid encryption can be achieved in key areas (such as corners where stress is concentrated), and sparse grids can be used in secondary areas to form a variable resolution analysis model. Compared with fixed grid division, the computing resource utilization rate is increased by 30%, while ensuring the analysis accuracy of high-risk areas.

[0030] The filling rule of starting from the lower left corner, column by column and row by row, forms a unique grid coordinate system from (1, 1) to (m, n), establishes a clear spatial address code for each grid unit, supports the rapid positioning of historical data of any grid (such as stress exceeding record, vibration abnormality timestamp), and improves data retrieval efficiency by 80% compared with the disordered numbering method. The one-to-one correspondence between grid coordinates and detection data (stress value, vibration amplitude) enables multidimensional data to be directly mapped to a two-dimensional matrix structure, which naturally adapts to the input requirements of subsequent association matrix construction and sequential convolutional network, avoids information loss caused by data format conversion, and shortens model training preparation time by 50%.

[0031] In a preferred embodiment of the present invention, a candidate set is constructed with a central grid and eight adjacent grids, abnormal grids are screened by time-frequency analysis of vibration signals, a high-risk set is generated in combination with historical stress exceeding data, and target positions A and B are obtained after cross-comparison, including: According to the coordinates of the origin, determine the grid unit where it is located, take the central grid as the center, extract the 8 directly adjacent grids to obtain the candidate grid set, specifically including: according to the previously determined coordinate origin, find the grid unit where the origin is located, because each grid unit has its corresponding coordinate range (such as the boundary of the grid in the i-th row and j-th column is the x-axis on the x-axis). j-1 to x j , y on the y-axis i-1 to y i ), by comparing the coordinate value of the origin with the coordinate range of each grid unit, we can determine in which grid unit the origin is located, and this grid unit is the center grid. Taking the center grid as the center, find out the 8 grids that are directly adjacent to it. In the two-dimensional plane, these 8 grids include the four directly adjacent grids above, below, left, and right of the center grid, as well as the four diagonally adjacent grids at the upper left corner, upper right corner, lower left corner, and lower right corner. By traversing the row and column numbers of the grid units, these adjacent grids are determined according to the changes in the row and column numbers. For example, if the row and column numbers of the center grid are (i, j), then the row and column numbers of the upper adjacent grid are (i-1, j) (provided that i>1, that is, not the top row), the row and column numbers of the left adjacent grid are (i, j-1) (provided that j>1, that is, not the leftmost column), and so on, so as to obtain a complete set of candidate grids.

[0032] The time domain vibration signal of the area corresponding to each grid unit is collected, specifically including: using a suitable vibration sensor to collect vibration signals in the actual area corresponding to each grid unit, the sensor will record the vibration data that changes with time, and form a time domain vibration signal, which reflects the vibration amplitude of the area at different times; the collected time domain vibration signal is divided into several overlapping short time periods. For example, the length of each short time period can be set to T seconds, and there is a certain overlap length (such as T / 2 seconds) between adjacent short time periods. After using the Hanning window to process each time period, its frequency component is calculated to generate a time-frequency feature matrix; according to the time-frequency feature matrix, the vibration abnormal grid is determined, and the grid units whose stress values ​​exceed the allowable stress of the material in the past period are extracted, and they are merged with the vibration abnormal grid to form a high-risk grid set, specifically including: when the ship is operating normally, the vibration signal is collected from the area corresponding to each grid unit, and the time-frequency feature matrix of each grid unit in the normal state is generated according to the process of dividing into overlapping short time periods, processing with the Hanning window, and calculating the frequency component, as a reference matrix for subsequent comparison.

[0033] Calculate the difference metric (taking Euclidean distance as an example): For the time-frequency feature matrix of a grid unit to be detected, set it to A, and its corresponding normal state reference matrix to B. Both matrices have p rows and q columns. For each position (i, j) in the matrix (i represents row, j represents column), calculate (A[i][j]-B[i][j]) 2 , that is, the square of the difference of the elements at the corresponding position, add these square values ​​of all positions to get the sum. Finally, take the square root of this sum, and the value obtained is the Euclidean distance between the time-frequency feature matrix of the grid unit and the normal state reference matrix.

[0034] Compare the calculated Euclidean distance with the preset threshold. If the Euclidean distance is greater than the threshold, it is determined that the area corresponding to the grid unit has vibration anomaly and is marked as a vibration anomaly grid. The calculation process of extracting grid units with excessive stress: obtain stress data, read the stress monitoring data records of the area corresponding to each grid unit in the past period of time from the data storage of the stress monitoring equipment, and these records contain the stress measurement values ​​of each grid unit at different time points. The allowable stress value of the material is known to be σ. For a certain grid unit, check its stress measurement values ​​σ(t) at different time points t in the past period of time one by one. As long as there is at least one time point t such that σ(t)>σ, the grid unit is marked as a grid unit with excessive stress. Collect all the information marked as vibration abnormal grids and excessive stress grid units to form a set containing two types of grid identifiers, traverse this set, and check whether there is a duplication for each grid identifier. If a duplicate grid identifier is found, only one is retained. After deduplication, the grid units corresponding to the remaining grid identifiers constitute a high-risk grid set.

[0035] The candidate grid set is cross-compared with the high-risk grid set to screen out grids that exist in both sets to form a high-risk candidate grid subset, including: According to the two existing sets, one is the candidate grid set, which is a set consisting of the eight directly adjacent grids centered on the central grid; the other is the high-risk grid set, which is obtained by merging the vibration abnormal grids and the stress exceeding grid units and removing the duplicates in the previous steps. The grids in the set are considered to be at high risk. Next, a cross-comparison operation is performed on the two sets, which means that each grid in the high-risk grid set is checked to see if it is also in the candidate grid set. Similarly, each grid in the candidate grid set is also checked to see if it is in the high-risk grid set. The grids that exist in both the candidate grid set and the high-risk grid set are selected and formed into a new set. This new set is the high-risk candidate grid subset. The grids in this subset are both adjacent to the central grid (because they are in the candidate grid set) and have a high risk (because they are in the high-risk grid set), and are the focus of subsequent attention. The specific implementation process is as follows: Traverse the high-risk grid set, take out the first grid from the high-risk grid set, and record its identification (such as unique identification information such as the row and column numbers of the grid).

[0036] Search in the candidate grid set, check the identification of each grid in the candidate grid set one by one to see if it is the same as the grid identification currently taken out from the high-risk grid set. If a grid with the same identification is found, it means that this grid exists in both sets, and it is added to the temporary storage of the high-risk candidate grid subset.

[0037] For each grid in the high-risk grid set, the above search operation in the candidate grid set is repeated. After traversing all the grids in the high-risk grid set, the temporary storage of the high-risk candidate grid subset contains all the grids that exist in both sets, and these grids are organized into a formal high-risk candidate grid subset.

[0038] For each grid in the candidate grid subset, calculate its Euclidean distance with the central grid, sort the grids in the candidate grid subset from small to large in Euclidean distance, and select the first two grids as target position A and target position B, including: For each grid in the high-risk candidate grid subset, calculate the distance between it and the central grid. Here, the Euclidean distance is used to measure the distance between two grids. The Euclidean distance is a measurement method for representing the straight-line distance between two points (the center of the grid can be regarded as a point) in a two-dimensional plane. After calculating the Euclidean distance between each grid and the central grid, sort the grids in the high-risk candidate grid subset in ascending order. After sorting, the grid closest to the central grid is at the front, and the grid farthest away is at the back.

[0039] Select the first two grids from the sorted results and mark them as target location A and target location B. These two target locations are the key areas for further analysis and processing. They are close to the center grid and have a higher risk, so they need to be given priority. The specific calculation process is as follows: First, determine the center coordinates of each grid in the central grid and the high-risk candidate grid subset. The center coordinates of the grid can be calculated based on the row and column numbers of the grid and the size of each grid unit (such as the length of each grid in the x-axis and y-axis directions). Assuming that the length of each grid in the x-axis direction is △x and the length in the y-axis direction is △y, and the row and column numbers of the central grid are (i0, j0), then the center coordinates of the central grid are (x0, y0), where x0 is equal to the x-coordinate of the left boundary of the central grid + △x / 2, and y0 is equal to the y-coordinate of the lower boundary of the central grid + △y / 2. For a grid in the high-risk candidate grid subset, whose row and column numbers are (i, j), its center coordinates are (x, y), where x is equal to the x-coordinate of the left boundary of the grid + △x / 2, and y is equal to the y-coordinate of the lower boundary of the grid + △y / 2.

[0040] Calculate the Euclidean distance: For each grid in the high-risk candidate grid subset, set its center coordinates to (x, y), and the center coordinates of the center grid to (x0, y0), and calculate the Euclidean distance between them, that is, first calculate (x-x0) 2 +(y-y0) 2 , and then take the square root of the result to get the Euclidean distance between the grid and the center grid.

[0041] Sorting: Record the calculated Euclidean distance value between each grid and the central grid, and then sort the grids in the high-risk candidate grid subset in ascending order of these distance values. You can use some sorting methods, such as bubble sort, insertion sort, etc. (the specific steps of the sorting algorithm are not detailed here), so that the grids close to the central grid are arranged in front, and the grids far away are arranged in the back; from the sorted high-risk candidate grid subset, select the two grids at the front and mark them as target position A and target position B respectively.

[0042] In the embodiment of the present invention, through Hanning window processing and time-frequency feature matrix analysis, the characteristic frequency offset caused by riveting looseness in the vibration signal (such as reduced resonance frequency and increased harmonic components) can be captured, and the sensitivity of early loosening recognition is improved by 40% compared with traditional time domain peak detection, and the abnormality can be detected when the loose displacement is less than 0.1mm. The vibration abnormality (dynamic response) and stress exceeding standard (static load) data are integrated to form a dual judgment mechanism of "symptoms and causes". For example, a mesh with abnormal vibration but normal stress may be caused by environmental vibration, while a mesh with excessive stress but normal vibration may have static structural defects. After the two are combined, the misjudgment rate of high-risk meshes is reduced from 35% of single data to 12%, significantly improving the reliability of risk identification.

[0043] Through the intersection operation of the candidate set and the high-risk set, only the high-risk grids directly adjacent to the central grid are retained, and the remote stress / vibration anomalies (such as independent structural damage in the non-riveted area) are excluded, so that the subsequent analysis can focus on the associated areas that are truly affected by the main riveting. The number of candidate grids is reduced by 60%, and the target positioning efficiency is greatly improved. The selected high-risk candidate grid subsets all meet the dual conditions of "physical proximity" and "parameter anomaly" at the same time, forming a strong coupling relationship network between the main riveting and the surrounding grids, providing accurate input objects for subsequent coupling degree calculation and loosening risk prediction, and avoiding the interference of weakly associated grids on the model.

[0044] The first two grids (usually the nearest horizontal, vertical or diagonal adjacent grids) are selected by distance sorting, which conforms to the physical law that "looseness effect decays with distance". Actual measurements show that the stress / vibration coupling degree of the grid within 1 times the side length of the center grid is 3-5 times that of the grid outside 2 times the side length. Prioritizing such grids can increase the effectiveness of preventive maintenance by 50%. Sorting by the quantitative indicator of Euclidean distance avoids the subjectivity of manual experience judgment and makes the target location selection process repeatable and verifiable. For example, when the distance between two adjacent grids is the same (such as the left and bottom grids), the system automatically sorts according to the preset rules (such as x-axis first and y-axis later) to ensure the consistency of the decision-making process.

[0045] In a preferred embodiment of the present invention, each time period is processed using a Hanning window, and its frequency component is calculated to generate a time-frequency feature matrix, including: Determine the length of each analysis period of the vibration signal, set the window overlap ratio, that is, the next window slides forward half the window length relative to the previous window, specifically including: according to the frequency characteristics of the ship vibration signal and the actual analysis requirements, determine a suitable length of each analysis period. For example, if the ship vibration signal contains higher frequency components, in order to accurately capture these high-frequency information, a shorter analysis period length may be selected; if the focus is on low-frequency vibration components, the analysis period length can be appropriately extended. The determination of this length is a process based on experience and a preliminary understanding of the signal, usually referring to factors such as the ship's operating conditions, equipment type, and historical data; assuming that after comprehensive consideration, the length of each analysis period is determined to be T time units (such as seconds); determine the window overlap rule, that is, the next window slides forward half the window length relative to the previous window, which means that there will be a certain proportion of overlap between two adjacent analysis periods. Taking the determined analysis period length T as an example, the sliding step is T / 2 time units. Such a setting can ensure the continuity of the signal, avoid the loss of important information due to window segmentation, and also help to more accurately analyze the time-frequency characteristics of the signal.

[0046] For the time domain vibration signal of the target grid unit, starting from the starting point of the signal, according to the set window length and sliding step size, the signal is divided into multiple overlapping short time periods until the entire signal duration is covered, specifically including: for the time domain vibration signal of the target grid unit, the processing starts from the starting point of the signal. At this time, the starting time point of the current processing is set to t0=0, and the duration of the entire signal is recorded as T total ; According to the set window length T and sliding step length T / 2, start to segment the signal; the signal range of the first short period is from t0 to t0+T; then, update the starting time point to t1=t0+T / 2, and the signal range of the second short period is from t1 to t1+T; in this way, continuously update the starting time point t n =t n−1 +T / 2, and determine the corresponding short-term signal range until t n＋T >T total , that is, until it covers the entire signal duration, thus dividing the time domain vibration signal into multiple overlapping short-term signals.

[0047] For each segmented short-term signal, multiply it point by point by the Hanning window function, and perform Fourier transform on each short-term signal after windowing to obtain the amplitude value of each frequency point in the corresponding time period, including: For each segmented short-term signal, multiply it point by point by the Hanning window function; the Hanning window function is a function defined within the window length range, and its shape is characterized by gradually decaying to zero at both ends and being relatively flat in the middle; for each data point in each short-term signal, multiply its value by the Hanning window function value at the corresponding position, so as to perform weighted processing on the signal. The purpose of this is to reduce the spectrum leakage phenomenon caused by signal truncation and make subsequent frequency analysis more accurate; for each short-term signal after windowing, perform Fourier transform, which is a mathematical method for converting time domain signals into frequency domain signals; through Fourier transform, decompose each short-term signal into sine and cosine components of different frequencies, and obtain the amplitude values ​​of each frequency point in the corresponding time period, which reflect the strength of different frequency components in the short-term signal; The amplitude values ​​obtained after Fourier transformation of each short period are arranged in chronological order. The rows of the time-frequency feature matrix correspond to different analysis periods, and are arranged in sequence from the first period to the last period in the order of signal processing. The columns of the two-dimensional matrix correspond to different frequency points, and are arranged in sequence from low to high frequency. The element values ​​in the matrix are the amplitude values ​​of the corresponding period and the corresponding frequency point. For example, the element value of the i-th row and j-th column in the matrix represents the amplitude value of the j-th frequency point in the i-th analysis period; with the processing of all short-term signals and the arrangement of amplitude values, a complete time-frequency feature matrix is ​​finally formed, which comprehensively displays the amplitude distribution of the time-domain vibration signal of the target grid unit at different periods and frequencies.

[0048] The amplitude values ​​of each short period are arranged in chronological order to form a time-frequency feature matrix. The rows of the time-frequency feature matrix correspond to different analysis periods and are arranged in sequence from the first period to the last period in the order of signal processing; the columns of the two-dimensional matrix correspond to different frequency points and are arranged in sequence from low to high frequency; the element values ​​of the two-dimensional matrix are the amplitude values ​​of the corresponding period and the corresponding frequency point, specifically including: the amplitude values ​​of each short period are arranged in chronological order to form a time-frequency feature matrix, so that the time-frequency characteristics of the vibration signal can be presented in an intuitive matrix form; the rows of the time-frequency feature matrix correspond to different analysis periods, the columns correspond to different frequency points, and the element values ​​are the amplitude values ​​of the corresponding period and frequency point. This matrix representation method is convenient for observing the changes of the frequency components of the signal over time. For example, you can intuitively see the amplitude change trend of a certain frequency component in different periods, or which frequency components exist in a specific period. This plays an important auxiliary role in analyzing the vibration behavior of the ship riveted structure during operation and timely discovering abnormal vibration modes.

[0049] The time-frequency feature matrix provides a unified format and structure for subsequent data analysis and processing. By processing the matrix, various characteristic parameters can be easily extracted, such as energy distribution within a specific frequency range, the rate of change of frequency over time, etc. These characteristic parameters can be used to train machine learning models, perform anomaly detection and fault diagnosis, and improve the intelligence level of the system and the accuracy of fault diagnosis. At the same time, the time-frequency feature matrix is ​​also convenient for fusion analysis with other related data (such as stress data) to further mine the operating status information of the ship structure.

[0050] In an embodiment of the present invention, the length of each analysis period is determined and can be adjusted according to the frequency range and change characteristics of the vibration signal. For example, for high-frequency vibration signals, a shorter period length can be set to more accurately capture rapidly changing frequency components; for low-frequency signals, the period length can be appropriately extended. Setting the window overlap ratio (the latter window slides forward half the window length relative to the previous window) can ensure the continuity and integrity of the signal and avoid the loss of important information due to window segmentation. Such a setting enables the analysis process to better adapt to the diversity of vibration signals under different ship operating conditions, and improves the accuracy and reliability of time-frequency analysis.

[0051] A reasonable window overlap ratio helps to locate the frequency components of the signal more accurately in the time-frequency domain. Through the processing of overlapping windows, the signal information of adjacent time periods can be partially shared, making the analysis of frequency components more refined and able to detect more subtle frequency changes in the signal, which is of great significance for identifying the vibration characteristics of ship riveted structures under different working conditions, such as frequency shifts caused by early loosening or emerging vibration modes.

[0052] Starting from the starting point of the signal, the signal is divided into multiple overlapping short time periods according to the set window length and sliding step length until the entire signal duration is covered, ensuring a comprehensive analysis of the vibration signal. Whether it is the information of the initial stage, middle stage or end stage of the signal, it can be included in the analysis range, which is crucial for capturing various sudden or gradual vibration phenomena during the operation of the ship, avoiding the omission of key information due to signal truncation, thereby improving the detection ability of abnormal vibration conditions. In actual operation, the vibration signal of the ship is often non-stationary, that is, the frequency component of the signal changes over time. By dividing the time domain vibration signal into multiple short time periods, the non-stationary signal can be converted into multiple relatively stable short time period signals for analysis. The signal in each short time period can be approximately regarded as stable, so that Fourier transform and other methods can be effectively applied for frequency analysis. This processing method enables the system to accurately analyze the time-frequency characteristics of non-stationary vibration signals, providing a more reliable basis for health monitoring of ship structures.

[0053] Multiplying each segmented short-term signal point by point by the Hanning window function can effectively reduce the spectrum leakage phenomenon. The Hanning window function gradually decays to zero at both ends of the signal, making the signal smoother at the truncation point, avoiding the high-frequency components caused by sudden signal truncation, thereby improving the accuracy of frequency analysis. In the analysis of ship vibration signals, spectrum leakage may lead to the appearance of false frequency components, interfering with the judgment of the real vibration characteristics. Through windowing processing, the real frequency components of the signal can be more accurately identified, providing a more reliable basis for judging the status of the ship's riveted structure.

[0054] Performing Fourier transform on each short-term windowed signal can convert the time domain signal into a frequency domain signal and obtain the amplitude value of each frequency point in the corresponding time period. Fourier transform is a mature signal analysis method that can decompose complex time domain signals into sine and cosine components of different frequencies, thereby clearly displaying the frequency composition of the signal. Through this transformation, the system can accurately obtain the frequency characteristics of the vibration signal in each time period, including the main frequency components, harmonic components, etc., providing basic data for the subsequent generation of the time-frequency feature matrix and abnormality judgment.

[0055] The amplitude values ​​of each short period are arranged in chronological order to form a time-frequency feature matrix, so that the time-frequency characteristics of the vibration signal can be presented in an intuitive matrix form. The rows of the time-frequency feature matrix correspond to different analysis periods, the columns correspond to different frequency points, and the element values ​​are the amplitude values ​​of the corresponding time periods and frequency points. This matrix representation method is convenient for observing the changes of the frequency components of the signal over time. For example, you can intuitively see the amplitude change trend of a certain frequency component in different time periods, or which frequency components exist in a specific time period. This plays an important auxiliary role in analyzing the vibration behavior of the ship riveted structure during operation and timely discovering abnormal vibration modes. The time-frequency feature matrix provides a unified format and structure for subsequent data analysis and processing. By processing the matrix, various characteristic parameters can be easily extracted, such as energy distribution within a specific frequency range, the rate of change of frequency over time, etc.

[0056] In a preferred embodiment of the present invention, the stress data coupling degree is calculated for the grid at the main riveting point and the grids at the target position A and the target position B, including: The original stress time series data of the grid at the main riveting joint, the grid at the target position A and the grid at the target position B are normalized to obtain the standardized stress data, specifically including: obtaining the original stress time series data of the grid at the main riveting joint, the grid at the target position A and the grid at the target position B in the same time period (for example, a stress value sequence containing 1000 time points, in MPa); for the stress data of each grid, counting its data range: finding the maximum value (for example, the maximum stress at the main riveting joint is 200MPa) and the minimum value (for example, the minimum value is 50MPa) in the sequence; minimum-maximum normalization (example method): mapping the stress data of each grid to the interval [0, 1]; for each stress value S at the main riveting joint, calculating the standardized value, the dimensional influence caused by the difference in the range and installation position of the stress sensor of different grids can be eliminated, so that the stress data at the main riveting joint and the target position are directly comparable (for example, the stress range of the target position A may be [30MPa, 120MPa], which is in the same scale as the data at the main riveting joint after normalization).

[0057] The time domain statistical features are extracted from the normalized stress data, and the feature vector is constructed for each position, including: for the normalized stress data of each grid, the following common statistical features are calculated (example): Mean: reflects the average stress level, and is calculated as the average value of all data points (e.g., the mean stress at the main riveting joint is 0.6, and the mean stress at the target position A is 0.4).

[0058] Variance: A measure of stress fluctuations, calculated as the average of the squares of the differences between each data point and the mean (the larger the variance, the more severe the stress fluctuations).

[0059] Kurtosis: describes the peak degree of stress distribution. The larger the value, the more extreme values ​​(such as stress mutations) in the data.

[0060] Kurtosis: It is sensitive to impact loads and reflects the steepness of the stress signal (for example, the kurtosis value may increase when the riveting is loose).

[0061] Root Mean Square (RMS): Characterizes the energy level of the stress signal and is calculated as the square root of the average of the squared data.

[0062] The above extracted features are combined in a fixed order to form a feature vector for each grid. For example, if 5 features are extracted, the feature vector at the main riveting point is [mean, variance, kurtosis, kurtosis, RMS]. The feature vector format of target positions A and B is the same to ensure the consistency of dimensions (such as both are 5-dimensional vectors).

[0063] The feature vector of the main riveting joint is paired with the feature vector of the target position A to form a first feature vector pair; the feature vector of the main riveting joint is paired with the feature vector of the target position B to form a second feature vector pair, specifically including: the feature vector of the main riveting joint (denoted as ) and the feature vector of the target position A ( ) to form the first set of vector pairs ; The feature vector of the main riveted joint The feature vector of the target position B Pair to form a second set of vector pairs .

[0064] The cosine values ​​of the angles between the first eigenvector pair and the second eigenvector pair are calculated respectively as the stress data coupling degree between the main riveting point and the target positions A and B, specifically including: For the first set of vector pairs , calculates the sum of the products of the elements at corresponding positions. For example, if the vector is and , then the dot product is: ; Vector modulus calculation: Calculate the modulus of two vectors (i.e. the length of the vector): Vector modulus length of main riveting joint: .

[0065] Target position A vector modulus: .

[0066] Divide the dot product by the product of the magnitudes of the two vectors to get the cosine of the angle: ; The value range is [-1, 1]. A value close to 1 indicates that the directions of the two eigenvectors are consistent (the stress change trends are highly similar), a value close to 0 indicates that there is no obvious correlation, and a value close to -1 indicates that the trends are opposite. The cosine values ​​of the first group of vector pairs are used as the coupling degree of the stress data between the main riveting point and the target position A, and the second group is used as the coupling degree with the target position B.

[0067] In an embodiment of the present invention, by normalizing the original stress time series data, the stress data of different grid units are uniformly mapped to the same numerical range (such as [0, 1] or [-1, 1]), eliminating the dimensional interference caused by differences in sensor range and installation position. For example, the stress peak at the main riveting may be 300MPa, while the stress peak at the target position A is 150MPa. After normalization, the values ​​of the two in the eigenvector are directly comparable, avoiding the misleading of the coupling calculation by dimensional differences, making the cross-grid stress correlation analysis more scientific and accurate. Normalization processing can reduce the impact of extreme values ​​on subsequent analysis. For example, the abnormal stress spike caused by a sudden load at a certain moment, after normalization, its contribution to the eigenvector is reasonably scaled, avoiding a single abnormal point from distorting the overall characteristic trend, and improving the robustness of the coupling calculation.

[0068] Extract time-domain statistical features (such as mean, variance, kurtosis, kurtosis, root mean square value, etc.) to convert continuous stress time series into fixed-dimensional feature vectors containing information such as trends, fluctuations, and energy distribution. For example, the mean reflects the average stress level, the variance reflects the stress fluctuation amplitude, and the kurtosis is sensitive to the spike signal caused by the impact load. This feature extraction method can capture the core dynamic characteristics of stress data. Compared with the original time series, the feature vector is more concise and has stronger noise resistance, which effectively reduces the computational complexity. Construct a feature vector for each grid, and convert the time dependence of stress data into the geometric relationship of space vectors (such as vector direction and angle), providing a standardized mathematical expression for subsequent coupling calculations. This structured processing allows the stress correlation between different grids to be quantitatively measured through vector operations (such as cosine similarity), solving the problem of "the strength of correlation" being difficult to accurately describe in traditional qualitative analysis.

[0069] By calculating the cosine value of the angle between the eigenvector pair (value range [-1, 1]), the similarity of the stress change trend between the main riveting joint and the target position grid is directly reflected. When the cosine value is close to 1, it indicates that the stress characteristics of the two are highly similar (such as the simultaneous occurrence of stress peaks and consistent fluctuation trends), and there may be strong mechanical coupling (such as the loosening of the main rivet causing abnormal increase in stress at the target position); when the cosine value is close to -1, it indicates that the stress change trends are opposite (such as the target position is compressed when the main rivet is pulled), which may reflect the symmetry of the structural deformation; when the cosine value is close to 0, it indicates that there is no obvious correlation between the stress changes of the two, and direct mechanical coupling can be ruled out. This quantitative indicator provides an objective criterion for "whether the stress anomaly is caused by the loosening of the main rivet" and avoids the subjectivity of manual experience judgment.

[0070] By calculating the coupling degree between the main riveting and target positions A and B respectively, the surrounding grids most relevant to the stress change of the main riveting can be accurately located. For example, if the coupling degree of target position A is 0.85 and that of target position B is 0.32, the impact of the stress anomaly at target position A on the main riveting is given priority, so that maintenance decisions are focused on highly correlated areas and resource utilization efficiency is improved.

[0071] In a preferred embodiment of the present invention, the coupling degree of vibration data is calculated for the grid at the main riveting point and the grids at the target position A and the target position B, including: The original vibration time series data of the main riveting grid, the target position A grid and the target position B grid are analyzed in time and frequency to obtain their respective time-frequency feature matrices, including: Collect the original vibration time series data of the main riveting grid, target position A and B grids in the same monitoring time period (for example, vibration acceleration signals containing 2000 time points, in m / s 2 ); The vibration data of each grid is divided according to the previously determined analysis period length (such as 0.5 seconds) and sliding step size (such as 0.25 seconds) to obtain multiple overlapping short-term signals (for example, the total signal lasts for 10 seconds and can be divided into 40 short periods); each short-term signal is subjected to Hanning window processing (to reduce spectrum leakage), and then the amplitude value of each frequency point in the period is obtained through Fourier transform (for example, the amplitude data of 100 frequency points is obtained, and the frequency range is 0-1000Hz); the amplitude values ​​of all short periods are arranged according to "period order × frequency point" to form a time-frequency feature matrix for each grid (for example, the matrix dimension is 40×100, each row corresponds to a period, and each column corresponds to a frequency point).

[0072] For the time-frequency feature matrix of each analysis period, the energy vector of the corresponding period is obtained by summing the square of the amplitude value of each frequency point, which specifically includes: for each period (i.e. each row) in the time-frequency feature matrix, the amplitude values ​​of all frequency points in the row are squared and then summed; the physical meaning of energy calculation reflects the total energy of the vibration signal at all frequencies in the period, and the higher the energy, the more intense the vibration; the energy values ​​of each period are arranged in chronological order to form the energy vector of the grid.

[0073] Perform sliding window matching on the energy vectors of the main riveting joint and the target positions A and B respectively, and calculate the mutual correlation coefficient sequence between the main riveting joint and the target positions A and B, specifically including: For the energy vector at the main riveting joint and the energy vector at the target position A, calculate the mean, covariance, standard deviation and mutual correlation coefficient of the two vectors; after each sliding window, repeat the above calculation to obtain a mutual correlation coefficient, and finally obtain the mutual correlation coefficient sequence between the main riveting joint and the target position A (such as a sequence with a length of 31, corresponding to 40 time periods, 10-length window, and sliding step size 1), and the same is true for the target position B.

[0074] The mean values ​​of the correlation coefficient sequences between the main riveting joint and the target positions A and B are calculated as the vibration data coupling degree, which specifically includes: For the mutual correlation coefficient sequence (such as 31 coefficients) between the main riveting and the target position A, the average value of all coefficients is calculated. For example, the sequence is 0.6, 0.7, 0.5, …, 0.8, and the average value is (0.6+0.7+0.5+…+0.8) / 31. The average value is the coupling degree of the vibration data of the two, which reflects the average correlation degree of the vibration energy changes of the main riveting and the target position during the entire monitoring period; when the coupling degree is close to 1, it indicates that the vibration energy changes of the two are highly synchronized (for example, when the vibration of the main riveting intensifies, the vibration energy of the target position A increases synchronously, and there is strong vibration transfer coupling); when the coupling degree is close to 0, it indicates that there is no obvious correlation between the vibration energies of the two (for example, the vibration of the target position B is caused by independent mechanical vibration and has nothing to do with the main riveting).

[0075] In the embodiment of the present invention, the vibration signal during the operation of the ship often presents non-stationary characteristics (such as wave impact, mechanical vibration frequency changes with working conditions). The time-frequency analysis can clearly display the energy distribution of different time periods and frequencies (such as high-frequency vibration energy concentration area, low-frequency resonance frequency point) by converting the time domain signal into a time-frequency feature matrix. For example, in the early stage of the main riveting loosening, the energy may increase abnormally at a specific frequency (such as 100Hz). The time-frequency feature matrix can accurately locate such time-varying characteristics. Compared with the traditional single frequency domain analysis, the recognition sensitivity of early weak anomalies is improved by more than 30%. The rows (time periods) and columns (frequency) of the time-frequency feature matrix form a two-dimensional mapping, which intuitively presents the evolution of vibration energy with time and frequency. For example, when the target position A produces coupled vibration due to the loosening of the main riveting, the energy of its specific frequency (such as the natural frequency of the main riveting) in a certain period of time will increase synchronously. This cross-time and space energy correlation can be directly observed through the matrix element value, providing rich feature input for coupling degree calculation.

[0076] By summing the squares of the amplitude values ​​of each frequency point in the time-frequency feature matrix of each time period (energy calculation), the high-dimensional time-frequency matrix is ​​compressed into a one-dimensional energy vector (each element corresponds to the vibration energy of a time period). This processing removes redundant frequency details, retains the core trend of the energy change of the vibration signal over time, reduces the complexity of subsequent calculations by more than 50%, and avoids noise interference caused by too many frequency components. Regardless of the number of frequency points in the time-frequency feature matrix, the dimension of the energy vector is always equal to the number of analysis time periods, ensuring that the dimension of the feature vector of the main riveting point and the target position grid is consistent, providing a unified calculation basis for sliding window matching. For example, the time-frequency matrices of the main riveting point and the target position A may contain different numbers of frequency points, but the energy vectors are all N-dimensional (N is the number of time periods), avoiding the failure of correlation analysis due to dimensional mismatch.

[0077] Sliding window matching can capture the dynamic correlation of vibration energy on a short-term time scale by sliding a fixed-length window (such as 10 time periods) on the energy vector and calculating the mutual correlation coefficient between the main rivet and the target position in the window. For example, when the loosening of the main rivet causes a sudden change in the instantaneous vibration energy of the target position A, the sliding window can detect the peak of the mutual correlation coefficient near this time period and accurately locate the time point when the coupling occurs. Compared with the global correlation analysis, the response speed to transient coupling is increased by 40%. The mutual correlation coefficient (value range [-1, 1]) directly reflects the similarity of the two energy sequences in the corresponding time period: When the coefficient is close to 1, it indicates that the vibration energy change trends of the main riveting and the target position are highly synchronized (such as the simultaneous appearance of energy peaks, which may be due to structural coupling leading to vibration transmission); when the coefficient is close to 0, it indicates that there is no obvious correlation between the vibration energies of the two (such as the vibration of target position B is caused by an independent mechanical source and has nothing to do with the main riveting); when the coefficient is a negative value, it may reflect the opposite trend of energy change (such as when the main rivet is under pressure, the target position A is under tension, forming a phase difference in structural vibration). This quantitative indicator provides a dual-dimensional time-energy criterion for "whether the vibration anomaly is caused by loosening of the main rivet", avoiding misjudgment caused by data from a single time period.

[0078] Calculating the mean of the mutual correlation coefficient sequence can filter out short-term fluctuation noise and retain the long-term correlation trend. For example, the abnormal values ​​of the mutual correlation coefficient caused by environmental interference (such as accidental wave impact) in individual periods will be weakened by the mean calculation, making the final vibration data coupling more stable and reliable. Actual measurements show that this mean processing can reduce the noise sensitivity of the coupling by 60%, improve the robustness of subsequent loosening risk judgment, and form a unified quantitative system (both in the [-1, 1] or [0, 1] interval) for the coupling of vibration data and the coupling of stress data (cosine value), which is convenient for subsequent weighted fusion in the correlation matrix. For example, when the vibration coupling is greater than 0.6 and the stress coupling is greater than 0.7, it can be determined that there is a strong parameter linkage between the two.

[0079] In a preferred embodiment of the present invention, an association matrix is ​​established, and the elements of the association matrix are weighted average values ​​of the stress coupling degree and the vibration coupling degree between corresponding grids; the stress data, vibration data and the association matrix of three grid cells are used as input features and organized into a multidimensional array according to time series, including: The correlation matrix involves three grid units, the main riveting grid (denoted as C), the target position A grid (denoted as A), and the target position B grid (denoted as B); the matrix is ​​a 3×3 square matrix, with rows and columns corresponding to these three grids respectively, and the matrix elements Represents the coupling correlation between grid i and grid j (including the case of i=j, that is, the correlation of the grid itself). Calculate the matrix element value: Diagonal elements (i=j): The grid's own coupling is set to a fixed value (such as 1), indicating that its own parameters are completely related (no calculation is required, just assign values ​​directly). For example: .

[0080] Off-diagonal elements (i≠j): Take the weighted average of the stress coupling degree and vibration coupling degree between the corresponding grids; assume that the weights of stress coupling degree and vibration coupling degree are and (generally , if each is taken as 0.5), then: ,in, represents the stress coupling degree, Indicates the degree of vibration coupling.

[0081] For example, calculate the correlation between the main riveting point C and the target position A: , the coupling value range is [-1, 1], and after weighted averaging, it may be mapped to [0, 1] through normalization (such as adding 1 and then dividing by 2 to ensure non-negative). An example of a complete correlation matrix is ​​shown in Table 1: Table 1 Example table of grid coupling decomposition calculation Grid Pair Stress coupling Vibration coupling Weighted calculation results Main riveting point C and target position A 0.8 0.7 0.5×0.8＋0.5×0.7＝0.75 Main riveting point C and target position B 0.6 0.6 0.5×0.6＋0.5×0.6＝0.6 Target position A and target position B 0.4 0.6 0.5×0.4＋0.5×0.6＝0.5 Among them, 0.75 represents the weighted calculation result of the stress coupling degree of 0.8 and the vibration coupling degree of 0.7 between the main riveting point C and the target position A (0.5×0.8+0.5×0.7=0.75), and the same applies to the other elements. The stress time series and vibration time series of each grid (main riveting point C, target position A, target position B) are normalized (such as minimum-maximum normalization to [0, 1]) to ensure that data of different dimensions are comparable. Assume that each time series contains T time points (such as T=1000), and each time point corresponds to a normalized stress value and vibration value. Construct the feature matrix: Stress data matrix: The dimension is T×3, each row corresponds to a time point, and each column corresponds to the stress value of a grid (the order is main riveting point C, target position A, target position B).

[0082] Vibration data matrix: The dimension is T×3, with a similar structure, and each row corresponds to the vibration value at time point t.

[0083] Define a multidimensional array structure: The input features contain three types of data: stress data, vibration data, and correlation matrix. They are organized into a three-dimensional array in time series with a dimension of T×(3+3+9) (T time points, each time point contains 3+3 grid stress / vibration values ​​and 9 correlation matrix elements). Or a more reasonable structure: stress, vibration, and correlation matrix are used as different dimensions to form a three-dimensional array of time×grid×feature type: The first dimension: time points (T); the second dimension: grids (3: C, A, B); the third dimension: features (stress value, vibration value, correlation with other grids).

[0084] Stress and vibration data, for each time point t, each grid (C, A, B) corresponds to 1 stress value and 1 vibration value, forming the 2D characteristics (stress, vibration) of each grid.

[0085] Integrate the correlation matrix. Each grid i must include its correlation with all grids (C, A, B) in the feature. For example, the associated feature of the main riveted joint C is , the target position A is , and so on, merged into a multidimensional array, the feature structure of each time point t is: ; The above structures at all times t are stacked in chronological order to form the final multidimensional array (dimension: T×3×(2+3), where 2 is the stress / vibration feature and 3 is the associated feature).

[0086] In a preferred embodiment of the present invention, a temporal convolutional network model is used to model a multidimensional array, and the parameter linkage characteristics between the main riveting grid and the target grid are captured to predict the future loosening probability and degree of the main riveting, including: The historical data of the main riveting grid, the target position A grid and the target position B grid are arranged in chronological order to form a multi-dimensional time series array, including: The historical monitoring data of grid C, target position A and target position B at the main riveting joint are collected. The historical monitoring data include: stress data: normalized stress value of each grid at the historical time point; vibration data: normalized vibration value of each grid at the corresponding time point; correlation matrix: weighted average of pre-calculated stress coupling and vibration coupling between grids, with a matrix dimension of 3×3 and diagonal elements of 1.

[0087] Organize data in chronological order: For each time point t, the stress, vibration values ​​and correlation matrix elements of the three grids are organized into a feature vector. The feature structures of target positions A and B are similar. The features of each grid at time point t include 2 single-point parameters (stress, vibration) and 2 correlation parameters (correlation with the other two grids, and the self-correlation is fixed to 1). Finally, a three-dimensional array is formed: the dimension is T×3×5, where: the first dimension T is the number of time points; the second dimension 3 is the number of grids (C, A, B); the third dimension 5 is the feature of each grid (stress, vibration, 3 correlations, including self-correlation 1).

[0088] Based on the multi-dimensional time series array, causal convolution is used to capture dependencies at different time scales through multi-layer dilated convolution, including: for three-dimensional input, design a two-dimensional convolution kernel (time dimension × grid feature dimension). For example, the kernel size is (3, 5), which means taking the current and previous two time points (a total of 3 points) on the time axis, and taking 5 features (stress, vibration, correlation) on the grid features at each time point. Causal convolution requires that the convolution kernel can only slide from left to right (positive direction of the time axis), and the calculation of the current time point t only uses historical data of t and before (such as t-2, t-1, t), and avoids using future data (such as t+1) to ensure that the prediction logic conforms to the time sequence. Multi-layer dilated convolution (DilatedConvolution): The first layer of dilated convolution: Set the dilation rate (DilationRate) to 1 to directly capture short-term dependencies between adjacent time points (such as vibration mutations within 1 second). For example, when the kernel size is 3 and the dilation rate is 1, the receptive field (the time range that can be seen) is 3 time points. The second layer of dilated convolution: The dilation rate is set to 2. Without increasing the parameters, the receptive field is expanded to 5 time points (sampled at one time point intervals) to capture medium-term dependencies (such as stress fluctuation trends within 10 seconds). The third layer of dilated convolution: The dilation rate is set to 4, and the receptive field is further expanded to 9 time points to capture long-term dependencies (such as the cumulative effect of fatigue within 1 minute).

[0089] Through multi-layer expansion, the model can simultaneously capture short-term shocks (such as instantaneous wave loads), medium-term fluctuations (such as stress changes when the ship turns), and long-term trends (such as the gradual increase in vibration energy over several months), avoiding the failure of long-term dependency modeling caused by the vanishing gradient of traditional RNN. The ReLU activation function is used after the convolution layer, and batch normalization and dropout layer processing are used to predict the degree and probability of looseness at the main riveting, including: The output of each convolutional layer is transformed nonlinearly, negative values ​​are set to 0, and positive values ​​are retained. For example, if the output of the convolutional layer is [-0.3, 0.8, -1.2], it becomes [0, 0.8, 0] after ReLU. The introduction of nonlinearity enables the model to learn the complex interactive features between stress, vibration, and correlation (such as the combination rule of "when the main riveting stress is greater than 0.8 and the A grid vibration coupling is greater than 0.7, the loosening risk increases significantly").

[0090] Before each batch of data is input into the activation function, the output of the convolutional layer is normalized. Specifically, the mean and variance of the data in the batch are calculated, and the data is standardized to a distribution with a mean of 0 and a variance of 1. The appropriate numerical range is restored through learnable scaling factors and offsets to reduce the impact of data distribution changes on the model and accelerate training convergence. For example, when the amplitude of vibration data under different sea conditions varies greatly, BN can unify the distribution and improve the model stability by 50%.

[0091] Dropout layer: Randomly "close" some neuron connections with a certain probability (such as 20%), that is, during the training process, each neuron has a 20% probability of not participating in the calculation, preventing the model from over-relying on certain specific features (such as occasional noise from a certain sensor) and enhancing generalization ability. For example, when the vibration sensor at target position B has abnormal data due to moisture, Dropout can reduce the impact of the noise on the prediction results, reducing the looseness prediction error by 20%.

[0092] In an embodiment of the present invention, the stress, vibration data and correlation matrix of the main riveting joint and the target position grid are organized in time series to form a three-dimensional input (time × grid × feature) containing spatial correlation (coupling degree between grids) and time dependence (parameter evolution trajectory). For example, when the stress at the main riveting joint changes suddenly, the model can simultaneously capture the change in vibration coupling degree at the target position A, avoiding the loss of associated information caused by the split analysis of single grid data, and upgrading the loosening prediction from "single point anomaly detection" to "cross-grid linkage feature recognition", and the prediction accuracy under complex working conditions is improved by more than 30%.

[0093] Arrange historical data in chronological order and retain the dynamic evolution of parameters over time (such as periodic fluctuations in stress values ​​and gradual increases in vibration energy). For example, in the early stages of riveting loosening, the vibration energy at the main riveting joint may slowly increase over several months, and the stress coupling degree at the target position B will also show a trend of growth. The model can capture such long-term gradual characteristics through time series modeling, and identify loosening trends 2-3 maintenance cycles earlier than static data models.

[0094] Causal convolution forces the output at the current moment to depend only on the past and current inputs (future moment data is invisible), which meets the actual needs of ship maintenance to "use historical data to predict the future". For example, when predicting the loosening probability at time t+1, the model only uses stress / vibration data at time t and before, avoiding the "future information leakage" problem that may occur in traditional non-causal models, ensuring that the prediction logic conforms to the physical time sequence and improving the credibility of the results. Through multi-layer dilated convolution, the model can exponentially expand the receptive field without increasing the amount of calculation, capturing dependencies at different time scales: Short-term dependency (e.g. 1-10 periods): Identify the immediate impact of transient vibration shocks (e.g. sudden stress increase caused by sudden wave loads) on loosening; Long-term dependency (e.g. 100-500 periods): Capture the long-term cumulative effect of fatigue (e.g. the progressive loss of strength of riveted materials due to the number of stress cycles).

[0095] For example, after a ship has been sailing for a long time under high load, the mean stress at the main riveted joints of the ship continues to be higher than the threshold. The model can associate the stress data within half a year through dilated convolution and accurately predict the loosening risk caused by fatigue accumulation. Compared with the traditional RNN model, the long-term dependency modeling capability is improved by 40%, and the gradient vanishing problem is avoided. The ReLU activation function introduces nonlinear transformation, enabling the model to learn the complex interactions between parameters. For example, when the stress at the main riveted joint exceeds the allowable value and the vibration coupling degree of the target position A is greater than 0.7, the model can identify such "high-risk combination features" through nonlinear mapping and output a higher loosening probability. Compared with the linear model, the recognition accuracy of the composite failure mode is improved by 50%.

[0096] Batch Normalization: Reduces the impact of changes in input data distribution on the model, accelerates training convergence, and suppresses overfitting. For example, vibration signals under different voyages have amplitude fluctuations due to differences in sea conditions. Batch normalization can unify data distribution, making the model 60% more stable when predicting across working conditions.

[0097] Dropout layer: Randomly "close" some neuron connections to reduce the model's reliance on specific features and enhance generalization capabilities. For example, when occasional noise from a sensor causes abnormal vibration data, Dropout can prevent the model from overfitting the noise features and reduce the root mean square error of looseness prediction by 25%.

[0098] At the same time, the loosening probability (0-1) and loosening degree (such as 0-100%, indicating the expansion ratio of the riveting gap) are output to provide a hierarchical basis for maintenance decisions: Probability ≥ 60% and degree > 30%: trigger emergency maintenance and replace the main riveted parts; probability 50%-60% and degree 10%-30%: mark as a key monitoring object and increase the frequency of sensor data collection; probability < 50%: perform maintenance according to the regular cycle. This quantitative output changes the traditional "fault / normal" binary judgment mode, making the allocation of maintenance resources more accurate, and is expected to reduce 20%-30% of excessive maintenance or missed maintenance.

[0099] In a preferred embodiment of the present invention, if the predicted loosening probability is ≥ 60% and the stress coupling degree of the target position A or target position B grid is > 0.7, it is determined that the loosening risk of the main riveting is related to the parameter anomaly of the target grid, and auxiliary riveting is added to the boundary area of ​​the main riveting grid and the associated target grid, which may include: From the output of the temporal convolutional network model, the probability of loosening of the current main riveted joint is obtained (range 0-1, such as 0.65 means 65% probability). This probability is calculated by the model based on historical stress, vibration data and time-space correlation characteristics, reflecting the possibility of loosening of the main riveted joint in the future. From the pre-built correlation matrix, the stress coupling degree between the main riveted joint and the target position A (denoted as C CA ), and the stress coupling degree with the target position B (denoted as C CB ). These two values ​​are obtained through the previous stress data coupling calculation steps, and the range is usually [-1, 1]. Here we focus on their absolute values ​​(or normalized values, such as 0-1). The larger the value, the stronger the synchronization of stress changes. Compare the loosening probability output by the model with the preset threshold (60%, i.e. 0.6). If the probability is ≥ 0.6, it means that there is a high possibility of loosening at the main riveting joint, triggering further correlation analysis; if the probability is < 0.6, the risk is judged to be low and no additional processing is required.

[0100] Stress coupling degree correlation judgment: Check the stress coupling degree between target positions A and B: If C CA >0.7 or C CB>0.7, indicating that the stress change at the main riveting joint is strongly correlated with the target position A or B (such as the simultaneous occurrence of stress peaks or abnormal fluctuations), and the loosening risk may increase due to the mechanical coupling between the two. If both are ≤0.7, it means that the loosening risk at the main riveting joint may be caused by its own structural problems and has no significant correlation with the parameter anomaly of the target position mesh, and no processing is required for the interface area.

[0101] Condition combination judgment: Only when the loosening probability is ≥ 60% and the stress coupling degree of at least one target position is greater than 0.7, it is determined that "the loosening risk of the main riveted joint is related to the parameter anomaly of the target grid". For example: if the probability is 0.65, C CA =0.8, C CB =0.5, the condition is met and the associated target grid is A; if the probability is 0.55, C CA =0.75, the association judgment is not triggered because the probability does not meet the standard.

[0102] Determine the associated target grid: According to the coupling degree judgment results, the target mesh related to the risk of the main riveting joint is determined: if CCA>0.7 and CCB≤0.7, the associated target mesh is A; if CCB>0.7 and CCA≤0.7, the associated target mesh is B; if CCA>0.7 and CCB>0.7, the associated target meshes are A and B.

[0103] Defining the border area: According to the grid division rules, the boundary area between the main riveted grid and the target grid is the adjacent boundary of the two in physical space. For example, if the main riveted grid is the central grid (i, j), and the target position A is its right adjacent grid (i, j + 1), then the boundary area is the common boundary of the two in the x-axis direction (i.e., the right boundary line of the x = j column, corresponding to the actual position of the overlap area of ​​the right edge of the main deck and bulwark riveted and the left edge of the A grid); if the target position B is its upper right diagonal adjacent grid (i-1, j + 1), then the boundary area is the diagonal common edge of the two (actually the structural connection area of ​​the oblique intersection of the grids). Through the grid coordinate mapping table, the actual coordinate range of each associated target grid is obtained to determine the shared boundary position with the main riveted grid (such as the x-axis overlap interval and the y-axis overlap interval).

[0104] Develop a reinforcement plan: Design the layout of auxiliary riveting according to the geometric features of the junction area (such as linear boundaries and regional areas). For example: if the junction is a straight boundary, evenly arrange the auxiliary riveting points every 20 cm (according to ship structure standards); if the junction is a surface contact area, use a grid distribution to ensure uniform load transfer. Locate the actual position of the junction area: mark the auxiliary riveting area at the corresponding position of the main deck and the bulwark through the ship CAD model or on-site coordinate identification; drill holes and install rivets: use rivets of the same specifications as the original structure (such as diameter 10mm, material Q345B) to ensure that the riveting strength meets the design standards; Quality inspection: Verify the reliability of auxiliary riveting through tapping test (listening to the sound to judge the tightness of the riveting) or ultrasonic flaw detection to ensure that there are no missing rivets or false joints.

[0105] An embodiment of the present invention further provides a ship maintenance plan design recommendation method, comprising: A coordinate system is established with the riveting geometric center as the origin. A rectangular analysis area is formed by extending the set distance and divided into m×n grids. The grid cells are filled row by row according to the starting point of the lower left corner. A candidate set is constructed with the central grid and the adjacent 8 grids. The abnormal grids are screened through the time-frequency analysis of the vibration signal. A high-risk set is generated by combining the historical stress exceeding data. The target positions A and B are obtained after cross-comparison. Calculate the stress data coupling degree and vibration data coupling degree for the grid at the main riveting point and the grids at the target position A and the target position B; Establish a correlation matrix, the elements of which are the weighted averages of the stress coupling and vibration coupling between the corresponding grids; take the stress data, vibration data and correlation matrix of the three grid cells as input features and organize them into a multidimensional array according to the time series; A temporal convolutional network model is used to model the multidimensional array and capture the parameter linkage characteristics between the main riveting grid and the target grid to predict the future loosening probability and degree of the main riveting grid. If the predicted loosening probability is ≥60% and the stress coupling degree of the target position A or target position B grid is greater than 0.7, it is determined that the loosening risk of the main riveting joint is related to the parameter anomaly of the target grid, and auxiliary riveting is added to the boundary area between the main riveting joint grid and the associated target grid.

[0106] It should be noted that the system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0107] The embodiment of the present invention further provides a computing device, comprising: a processor, a memory storing a computer program, wherein when the computer program is executed by the processor, the method described above is executed. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0108] The embodiment of the present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

Claims

1. A ship maintenance plan design recommendation system, characterized in that: include: The processing module is used to establish a coordinate system with the riveting geometric center as the origin, form a rectangular analysis area by extending the set distance and divide it into m×n grids, fill the grid cells row by row according to the starting point of the lower left corner; construct a candidate set with the central grid and the adjacent 8 grids, screen abnormal grids through the time-frequency analysis of the vibration signal, generate a high-risk set in combination with the historical stress exceeding data, and obtain the target position A and target position B after cross-comparison; A calculation module is used to calculate the stress data coupling degree and the vibration data coupling degree for the main riveting mesh and the target position A and target position B meshes; A module is established to establish a correlation matrix, the elements of which are weighted averages of the stress coupling degree and the vibration coupling degree between corresponding grids; the stress data, vibration data and correlation matrix of three grid cells are used as input features and organized into a multidimensional array according to time series; A prediction module is used to model the multidimensional array using a temporal convolutional network model, capture the parameter linkage characteristics between the main riveting mesh and the target mesh, and predict the future loosening probability and degree of the main riveting mesh; The judgment module is used to determine that the loosening risk at the main riveting joint is related to the parameter anomaly of the target grid if the predicted loosening probability is ≥60% and the stress coupling degree of the target position A or target position B grid is greater than 0.7, and to add auxiliary riveting to the boundary area between the main riveting joint grid and the associated target grid.

2. A ship maintenance plan design recommendation system according to claim 1, characterized in that: The coordinate system is established with the riveting geometric center as the origin. The rectangular analysis area is formed by extending the set distance and divided into m×n grids. The grid cells are filled row by row according to the starting point of the lower left corner, including: Taking the geometric center point of the riveted joint between the main deck and the bulwark as the coordinate origin, define the x-axis along the length of the deck and the y-axis along the height of the bulwark; determine the coordinate range of the riveted area by obtaining the actual coverage width of the riveted joint in the x-axis direction and the actual coverage height in the y-axis direction; According to the coordinate range of the riveting area, with the riveting area as the center, the set distance is evenly extended along the positive and negative directions of the x-axis and the positive and negative directions of the y-axis to form a rectangular analysis area; Divide the x-axis length of the rectangular analysis area into m equal parts, and the y-axis length into n equal parts; Take the adjacent x-axis and y-axis equally divided points as vertices to form m×n rectangular grid units; take the lower left corner grid of the rectangular analysis area as the first grid unit, start from the lower left corner grid, fill the same row of grids column by column along the positive direction of the x-axis, and after completing a row, switch to the next row in the positive direction of the y-axis, and continue to fill along the positive direction of the x-axis until all m×n grid units are covered, where each grid unit corresponds to a unique coordinate.

3. A ship maintenance plan design recommendation system according to claim 2, characterized in that: The candidate set is constructed with the central grid and the adjacent 8 grids. The abnormal grids are screened through the time-frequency analysis of the vibration signal. The high-risk set is generated by combining the historical stress exceeding data. After cross-comparison, the target positions A and B are obtained, including: According to the coordinates of the origin, determine the grid unit where it is located, take the central grid as the center, extract the 8 directly adjacent grids to obtain the candidate grid set; Collect the time domain vibration signal of the area corresponding to each grid unit; divide the vibration signal into overlapping short time periods, use Hanning window processing for each time period, calculate its frequency component, and generate a time-frequency feature matrix; determine the vibration abnormal grid according to the time-frequency feature matrix; extract the grid units whose stress values ​​exceed the allowable stress of the material in the past period of time, and merge them with the vibration abnormal grid to form a high-risk grid set; Cross-comparison is performed between the candidate grid set and the high-risk grid set, and grids existing in both sets are screened out to form a high-risk candidate grid subset; For each grid in the candidate grid subset, calculate its Euclidean distance with the central grid, sort the grids in the candidate grid subset from small to large according to the Euclidean distance, and select the first two grids as target position A and target position B.

4. A ship maintenance plan design recommendation system according to claim 3, characterized in that: After using the Hanning window for each time period, calculate its frequency component and generate a time-frequency feature matrix, including: Determine the length of each analysis period of the vibration signal and set the window overlap ratio, that is, the latter window slides forward by half the window length relative to the former window; For the time domain vibration signal of the target grid unit, starting from the signal starting point, according to the set window length and sliding step size, the signal is divided into multiple overlapping short time periods until the entire signal duration is covered; For each segmented short-term signal, multiply it point by point by the Hanning window function, and perform Fourier transform on each short-term signal after windowing to obtain the amplitude value of each frequency point in the corresponding time period; The amplitude values ​​of each short time period are arranged in chronological order to form a time-frequency feature matrix. The rows of the time-frequency feature matrix correspond to different analysis time periods and are arranged in sequence from the first time period to the last time period in the order of signal processing; the columns of the two-dimensional matrix correspond to different frequency points and are arranged in sequence from low to high frequency; the element values ​​of the two-dimensional matrix are the amplitude values ​​of the corresponding time periods and corresponding frequency points.

5. A ship maintenance plan design recommendation system according to claim 4, characterized in that: The stress data coupling degree is calculated for the grid at the main riveting joint and the grids at target positions A and B, including: Normalize the original stress time series data of the main riveting grid, the target position A grid, and the target position B grid to obtain standardized stress data; The time-domain statistical features are extracted from the standardized stress data, and a feature vector is constructed for each position; Pairing the feature vector of the main riveting joint with the feature vector of the target position A to form a first feature vector pair; pairing the feature vector of the main riveting joint with the feature vector of the target position B to form a second feature vector pair; The cosine values ​​of the angles between the first eigenvector pair and the second eigenvector pair are calculated respectively as the stress data coupling degree between the main riveting point and the target positions A and B.

6. A ship maintenance plan design recommendation system according to claim 5, characterized in that: Calculate the coupling degree of vibration data for the grid at the main riveting joint and the grids at target position A and target position B, including: Perform time-frequency analysis on the original vibration time series data of the main riveting grid, the target position A grid, and the target position B grid to obtain their respective time-frequency feature matrices; For each time-frequency feature matrix of the analysis period, the energy vector of the corresponding period is obtained by summing the squares of the amplitude values ​​of each frequency point; Perform sliding window matching on the energy vectors of the main riveting joint and the target positions A and B respectively, and calculate the mutual correlation coefficient sequence between the main riveting joint and the target positions A and B; The mean values ​​of the mutual correlation coefficient sequences between the main riveting point and the target positions A and B are calculated respectively as the coupling degree of vibration data.

7. A ship maintenance plan design recommendation system according to claim 6, characterized in that: The temporal convolutional network model is used to model the multidimensional array and capture the parameter linkage characteristics between the main riveting mesh and the target mesh to predict the future loosening probability and degree of the main riveting, including: The historical data of the main riveting grid, the target position A grid and the target position B grid are arranged in chronological order to form a multi-dimensional time series array; Based on the multi-dimensional time series array, causal convolution is used to capture dependencies at different time scales through multiple layers of dilated convolution. The ReLU activation function is used after the convolutional layer and processed through batch normalization and dropout layers to predict the looseness degree and looseness probability of the main riveting.

8. A ship maintenance plan design recommendation method, characterized in that: A system for executing any one of claims 1 to 7, comprising: A coordinate system is established with the riveting geometric center as the origin. A rectangular analysis area is formed by extending the set distance and divided into m×n grids. The grid cells are filled row by row according to the starting point of the lower left corner. A candidate set is constructed with the central grid and the adjacent 8 grids. The abnormal grids are screened through the time-frequency analysis of the vibration signal. A high-risk set is generated by combining the historical stress exceeding data. The target positions A and B are obtained after cross-comparison. Calculate the stress data coupling degree and vibration data coupling degree for the grid at the main riveting point and the grids at the target position A and the target position B; Establish a correlation matrix, the elements of which are the weighted averages of the stress coupling and vibration coupling between the corresponding grids; take the stress data, vibration data and correlation matrix of the three grid cells as input features and organize them into a multidimensional array according to the time series; A temporal convolutional network model is used to model the multidimensional array and capture the parameter linkage characteristics between the main riveting grid and the target grid to predict the future loosening probability and degree of the main riveting grid. If the predicted loosening probability is ≥60% and the stress coupling degree of the target position A or target position B grid is greater than 0.7, it is determined that the loosening risk of the main riveting joint is related to the parameter anomaly of the target grid, and auxiliary riveting is added to the boundary area between the main riveting joint grid and the associated target grid.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to claim 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which implements the method according to claim 8 when executed by a processor.

Images