A full-field DIC measured node and FEM calculated node adaptive matching and screening method and system for crane modal analysis

Abstract

This invention discloses an adaptive matching and screening method and system for full-field DIC measured nodes and FEM calculated nodes for crane modal analysis. First, a K-D tree is constructed and ICP coordinate registration is completed. Then, candidate points are screened by adaptively generating a search radius based on the measurement point density and structural dimensions. Gaussian spatial similarity and MAC mode shape similarity are fused, with fusion weights adaptively allocated according to the modal order. The minimum half-wavelength of the mode is calculated through the mode shape gradient, and adjacent node matching penalty constraints are added. The KM algorithm is used to solve the global one-to-one matching. Finally, the validity of measurement point sampling is verified based on the Nyquist criterion, and a compliant matching set or an empty set is output. This invention takes into account both spatial location and vibration dynamics characteristics, avoiding non-physical cross-region matching; built-in sampling verification avoids modal aliasing; it is compatible with various crane two-dimensional shell and three-dimensional solid simulation models, providing reliable measurement point mapping data for finite element model correction and mode shape correlation analysis.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of structural dynamics testing and numerical simulation, specifically involving an adaptive matching and screening method and system for full-field DIC measured nodes and FEM calculation nodes for crane modal analysis. Background Technology

[0002] Digital image correlation methods (DIC) can non-contactly acquire the full-field displacement response of a structure by capturing speckle images of its surface, and then obtain the mode shapes of each node through modal recognition algorithms. This method can identify tens of thousands or even millions of nodal mode shape data, far exceeding the number of measurement points required by traditional accelerometers. Meanwhile, finite element methods (such as commercial solvers like Altair OptiStruct and ANSYS) can calculate the theoretical mode shapes of any node in the structure.

[0003] When conducting subsequent work such as correlation analysis between experimental and simulation results and correction of finite element models, it is necessary to establish a one-to-one correspondence between the measured nodes of DIC and the calculated nodes of FEM. However, the following technical challenges exist: DIC nodes are random, disordered, and unstructured point cloud data, while FEM nodes are structured mesh data with clear element connection relationships; the density of DIC nodes is usually much higher than that of FEM nodes, and their distribution is uneven; they are located in different coordinate systems and may have rigid body displacement deviations; most importantly, existing technologies do not fully utilize the key dynamic feature of mode shape for node matching, resulting in matching results lacking physical meaning.

[0004] In existing technologies, such as the paper "Comparison of Measurement Deviation of Aircraft Casing Deformation Based on Data Mapping Optimization [J]. Optics and Precision Engineering, 2023, 31(20):2930-2942," a mapping method for DIC measurement data and finite element simulation data is proposed, which uses point cloud registration and nearest neighbor search to achieve data mapping. However, this method is only applicable to data mapping of static deformation fields, and its essence is to treat the displacement field as a scalar field or vector field for spatial interpolation. For physical quantities such as mode shapes with specific dynamic characteristics (multiple orders, spatial periodicity, orthogonality, node / anti-node distribution), simple spatial distance measurement cannot reflect the physical essence of the mode shape, resulting in the lack of dynamic physical meaning in the matching results. In addition, existing technologies do not utilize the spatial continuity of the mode shape changes of adjacent nodes to constrain the matching results, which easily leads to errors of local matching inconsistency (i.e., adjacent FEM nodes are matched to DIC nodes that are spatially far apart). Furthermore, the point cloud adaptive neighborhood search mechanism has been used for computer vision tasks such as semantic segmentation, but its application field is completely different from that of this invention.

[0005] Currently, there is no systematic method specifically for matching and screening the nodes of the measured modal vibration modes in DIC and the nodes of the calculated modal vibration modes in FEM for cranes. In particular, there is a lack of technical solutions to optimize the matching results by utilizing the consistency constraints of the spatial gradient of the modal vibration modes. Summary of the Invention

[0006] This invention provides an adaptive matching and screening method and system for full-field DIC measured nodes and FEM calculated nodes for crane modal analysis, aiming to solve the technical problem that existing technologies cannot automatically, accurately, and efficiently perform one-to-one matching and effective node screening for massive DIC measured nodes and FEM calculated nodes of cranes.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following solution:

[0008] An adaptive matching and selection method for full-field DIC measured nodes and FEM calculated nodes for crane modal analysis includes the following steps:

[0009] Step S1, Data Acquisition and Preprocessing: Collect the actual DIC node set and FEM calculated node set of the crane, construct the DIC node spatial index and complete the coarse alignment of the two sets of node coordinates;

[0010] Step S2, Adaptive Neighborhood Candidate Matching: Based on the local density of DIC measurement points around each FEM node, the search radius is adaptively generated, and the candidate DIC node set corresponding to each FEM node is obtained by filtering.

[0011] Step S3: Calculate the weighted similarity of multi-mode vibrations: Combine spatial distance features and multi-mode vibration MAC features, use weights that are adaptively adjusted according to the modal order to calculate the comprehensive similarity, and simplify the candidate matching set;

[0012] Step S4, Global one-to-one matching with mode shape spatial continuity constraints: Construct a bipartite graph using FEM and DIC nodes, determine the minimum half-wavelength of the mode and the adjacent matching distance constraint based on the mode shape spatial gradient, and solve for the global optimal one-to-one matching after correcting the edge weights of the bipartite graph.

[0013] Step S5, Nyquist sampling validity verification: Based on the average spacing of DIC nodes after matching and the minimum half wavelength of the mode, the sampling criterion verification is completed to determine whether the measurement point meets the mode recognition sampling requirements;

[0014] Step S6: Output matching results: If the sampling verification is successful, output a set of valid FEM-DIC matching node pairs; if the verification is unsuccessful, output an empty set.

[0015] Further optimization, step S1 specifically includes the following steps:

[0016] Step S1.1, Data Reading: Obtain the DIC measured node set ,in For spatial coordinates, For the front The first-order DIC mode shape vector, where M is the total number of DIC nodes. Let be the modal order parameter, and ;

[0017] Obtain the FEM numerical computation node set ,in For spatial coordinates, is the vector of the first k modes of the FEM, and N is the total number of nodes in the FEM.

[0018] Step S1.2, Spatial Index Construction: Construct a KD-tree spatial index for the unordered DIC point cloud nodes.

[0019] Step S1.3, Point Cloud Coordinate Registration: The ICP iterative nearest point algorithm is used to solve the rotation and translation transformation matrices of the two sets of node sets, eliminating rigid body displacement and rotation deviations between the experimental and simulation coordinate systems, and completing the coarse alignment of coordinates.

[0020] Further optimization, step S2 specifically includes the following steps:

[0021] Step S2.1, Parameter Preset: Set the maximum search radius ;Statistical analysis of all FEM nodes The number of global reference nodes is obtained by averaging the number of DIC nodes within the radius sphere. .

[0022] Step S2.2, Node-by-Node Adaptive Search Radius Calculation: DIC nodes located within this radius are retrieved from the spatial index as candidate matching nodes; the adaptive search radius of the current FEM node is calculated using the following formula: ;

[0023] in, With this FEM node as the center and a radius of The number of DIC nodes inside the sphere; The dimension index is taken as the dimension index for a three-dimensional volume structure. For two-dimensional shell structures, take .

[0024] Step S2.3, Candidate Node Retrieval: Retrieve all DIC nodes within the r-space range in the KD tree to form the initial candidate set of the FEM node.

[0025] Further optimization, step S3 specifically includes the following steps:

[0026] Step S3.1: For each candidate matching pair Calculate spatial distance similarity Similarity to modal shape ;

[0027] (1) Formula for spatial distance similarity: ;in: The Euclidean distance between FEM and DIC nodes; This is a spatial scale parameter, and its value is 2 to 3 times the average node spacing of all FEM nodes.

[0028] (2) Formula for similarity of MAC mode shapes: ; , These are the first k-th order mode shape vectors of FEM and DIC, respectively.

[0029] Step S3.2: Calculate the modal adaptive weights: ; The time weight increases linearly with the order; when k>6, the weight is fixed at 0.85.

[0030] Step S3.3, Calculation of overall similarity S: .

[0031] Step S3.4: Set the similarity threshold S th =0.6, only retain the DIC nodes with a comprehensive similarity of Nc=10 and greater than the threshold, forming a simplified candidate set.

[0032] Further optimization, step S4 specifically includes the following steps:

[0033] Step S4.1: Identify adjacent node pairs in the FEM mesh: If two FEM nodes share the same cell, or the Euclidean distance between the nodes is less than 1.2 to 1.5 times the average cell size of the mesh, they are determined to be adjacent node pairs, and a set of adjacent node pairs is constructed. .

[0034] Step S4.2: Pre-calculate based on FEM node set The following modal space gradient method was used to calculate the previous values. Minimum wavelength of first mode ,

[0035] Define the maximum allowable spacing. Where t is the modal oversampling coefficient, .

[0036] The modal space gradient method includes:

[0037] Step S4.2.1: For the first First-order mode, for each node in a given set of nodes Through its neighborhood Using the nearest neighbor nodes, the tangent plane of the local mode shape is fitted using the least squares method, and the mode shape gradient vector is calculated. The gradient magnitude is The number of nearest neighbors is 6 ≤ K ≤ 15;

[0038] Step S4.2.2: Traverse all nodes and determine the maximum gradient magnitude of this mode. ;

[0039] Step S4.2.3: Determine the mode shape amplitude of this mode. If it has been normalized, then ;

[0040] Step S4.2.4: Calculate the minimum half wavelength of this mode. ; This indicates the shortest total wavelength of the nth mode;

[0041] Step S4.2.5: For the previous For each of the three modes, the minimum half-wavelength of each mode is taken as the final criterion, i.e. .

[0042] Step S4.3: Determine the adjacent matching distance and calculate the penalty factor:

[0043] 1) For each pair of adjacent nodes and its candidate matching DIC nodes and ,calculate and Among them, nodes belong Candidate set, nodes belong The candidate set;

[0044] 2) If Then, simply delete the edge from the bipartite graph. and ;like Then forced In other cases, the relative deviation penalty factor... Where β is the allowable threshold coefficient for relative distance deviation. .

[0045] Step S4.4: Bipartite graph edge weight correction: adjust the edge weights... and The original comprehensive similarity is multiplied by 1. If the same edge is modified by multiple adjacent constraints, then its original comprehensive similarity is multiplied by the sum of all the factors involving that edge. The product of.

[0046] Step S4.5: The maximum weight matching algorithm is used to solve for the globally optimal matching on the corrected bipartite graph. Specifically, after correcting the weights of all matching edges and removing invalid edges, a standard bipartite graph is constructed. The left vertex is all FEM nodes, and the right vertex is the selected DIC candidate nodes. Each edge is assigned a comprehensive similarity value after being corrected by a penalty factor. If the number of vertices on the left and right sides is inconsistent, virtual nodes with a weight of 0 are added to the side with fewer nodes. An equal-sized square matrix is ​​constructed to adapt to the input of the Kuhn-Munkres Hungarian algorithm. Vertex labels, relaxation variables, and matching labels are initialized. Feasible augmenting paths are found iteratively through depth-first search. The vertex labels and relaxation values ​​are dynamically updated, and the matching set is continuously expanded until no new legal augmenting paths are added, resulting in a one-to-one matching matrix with the largest global total weight. Finally, all pairings associated with virtual nodes are removed, and the real FEM-DIC matching relationships are retained. This achieves a globally optimal mapping where each FEM node matches only one DIC node, and each DIC node corresponds to only one FEM node. Nodes that do not form a valid pair are uniformly marked as invalid nodes.

[0047] Further optimization, step S5, node validity verification based on the Nyquist sampling theorem, includes the following steps:

[0048] Step S5.1: Calculate the average spatial spacing of all DIC nodes after matching. Specifically, the steps are as follows: extract the set of DIC node coordinates corresponding to all valid matches output in step S4; traverse any two DIC nodes in the set and calculate the three-dimensional Euclidean distance between the two points; summarize all the distance values ​​of all pairs of nodes and sum them up to count the total number of distance samples; divide the total distance by the total number of samples to obtain the average spatial spacing between nodes.

[0049] Step S5.2: Obtain the highest valid modal order of the matching DIC set. , The minimum half wavelength of the DIC set is calculated using the modal gradient method in step S4.2. ;

[0050] Step S5.3: Nyquist sampling decision: If This triggers a "DIC sampling does not meet the Nyquist condition" warning and jumps to step S6.

[0051] Step S5.4: Perform grid resolution verification if sampling is qualified: If It outputs a "FEM mesh resolution insufficient" message without interrupting the process.

[0052] Further optimization, step S6, which outputs the final set of matching and valid node pairs, includes the following steps:

[0053] Step S6.1: If step S5 does not trigger the Nyquist sampling warning, output all one-to-one matching node pairs; each matching pair includes FEM number, DIC number, three-dimensional coordinates, mode shape vectors of each order, and comprehensive similarity;

[0054] Step S6.2: If step S5 triggers a Nyquist sampling warning, output an empty set directly.

[0055] An adaptive matching and filtering system for full-field DIC measured nodes and FEM calculated nodes for crane modal analysis is provided to implement any of the methods described above. The system includes a hardware processing unit and six functional modules mounted on the hardware processing unit: a data acquisition and coarse alignment module, an adaptive neighborhood candidate matching module, a multi-order modal weighted similarity calculation module, a mode shape gradient constraint and global bipartite graph matching module, a Nyquist sampling verification module, and a matching result output module.

[0056] The data acquisition and coarse alignment module completes the reading of DIC and FEM node sets, the construction of KD tree spatial index, and the registration of two sets of point cloud ICP coordinates.

[0057] The adaptive neighborhood candidate matching module calculates the search radius for each FEM node according to the density adaptive formula and retrieves and generates a simplified candidate DIC set for each node.

[0058] The multi-modal weighted similarity calculation module solves for spatial similarity, MAC modal similarity, and adaptive order weight, respectively, and fuses them to obtain the comprehensive similarity of each matching pair and filters candidates according to a threshold.

[0059] The modal gradient constraint and global bipartite graph matching module calculates the modal gradient, minimum half-wavelength of the mode, and adjacent matching penalty coefficient, corrects the edge weights of the bipartite graph, and solves the global one-to-one optimal matching through the Kuhn-Munkres algorithm.

[0060] The Nyquist sampling verification module calculates the average node spacing of the DIC after matching, performs Nyquist sampling criterion verification and FEM mesh resolution consistency verification, and outputs corresponding prompt information.

[0061] The matching result output module outputs a complete matching node dataset or an empty set based on the sampling verification results. The output data includes the node number, coordinates, mode shape, and similarity information.

[0062] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the methods described above.

[0063] A computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements any one of the methods described above.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. This invention distinguishes the dimension index for two typical structures: the two-dimensional shell of the crane main beam and the three-dimensional solid of the boom. It adaptively calculates the retrieval radius based on the density of local DIC measurement points. The search range is automatically narrowed in dense measurement point areas, which greatly reduces invalid candidate points and reduces the computational load of graph matching. The retrieval upper limit is expanded in sparse measurement point areas to avoid the loss of effective matching points, thus solving the problems of missed matching and redundant calculation that are common in fixed radius matching.

[0066] 2. This invention designs an adaptive MAC weight calculation formula based on modal order. When the spatial changes of higher-order modes are drastic, the similarity ratio of the modes is automatically increased, while the spatial position matching of lower-order modes is emphasized. Quantitative verification by experiments shows that the average MAC value of this scheme reaches 0.9922, which is much higher than that of pure spatial matching (0.9649) and static deformation mapping scheme (0.9761). The correlation between experimental and simulated modes is greatly enhanced, providing a high-precision mapping basis for finite element model correction.

[0067] 3. The scheme described in this invention is based on the minimum half-wavelength determination logic of the mode shape spatial gradient, derives the maximum allowable spacing for matching adjacent FEM nodes, and designs a linear penalty factor to correct the bipartite graph matching weight; matches with excessive spatial deviation are directly eliminated, and similarity attenuation constraints are applied to matches with slight deviations. From the perspective of structural vibration deformation mechanism, it ensures that the matching space of adjacent finite grids is close to DIC measurement points, and eliminates the local spatial fragmentation distortion caused by traditional point-by-point independent matching.

[0068] 4. An automated judgment standard is established based on the Nyquist sampling theorem. By comparing the average distance between DIC measurement points after matching with the minimum half wavelength of the mode, the undersampling condition of measurement points is automatically identified and invalid matching datasets are intercepted. At the same time, FEM mesh resolution consistency verification is added to give early warning of insufficient mesh division and avoid drawing incorrect engineering conclusions based on unqualified measurement points for damage identification and correlation analysis.

[0069] 5. The solution described in this invention adopts the Kuhn-Munkres maximum weight bipartite graph matching algorithm, adds virtual nodes to adapt to scenarios with inconsistent numbers of DIC and FEM nodes, achieves a strict one-to-one correspondence between FEM and DIC nodes, and eliminates the problem of matching one measurement point with multiple grids or repeated mapping. The matching results are standardized and can be directly connected to subsequent dynamic post-processing software. Attached Figure Description

[0070] Figure 1 This is an overall flowchart of the method described in Embodiment 1 of the present invention;

[0071] Figure 2 This is a schematic diagram of the finite element model of the crane in Example 1;

[0072] Figure 3 This is a schematic diagram of the adaptive neighborhood search radius in Example 1;

[0073] Figure 4 This is a schematic diagram of the MAC distribution of the matching pairs in Example 1;

[0074] Figure 5 The diagram shows the distribution of FEM nodes and DIC nodes in Example 1, and the connection of matching points.

[0075] Figure 6 This is a schematic diagram comparing the probability density distribution of coordinate registration error in Example 1. Detailed Implementation

[0076] To enable those skilled in the art to more clearly understand the technical solution of the present invention, the present invention will be further described in detail below with reference to embodiments and accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.

[0077] Example 1:

[0078] like Figure 1 As shown, an adaptive matching and selection method for full-field DIC measured nodes and FEM calculated nodes for crane modal analysis includes the following steps:

[0079] Step S1: Data acquisition and preprocessing, specifically:

[0080] In this embodiment, the test subject is a long ,Width ,thick The steel free plate is used to simulate a local area of ​​the main beam structure of a crane.

[0081] DIC Actual Measurement: Two high-speed cameras were used (resolution) Pixels, sampling frequency The surface speckle pattern was photographed, and the approximate surface area was calculated using VIC-3D software. The first three modal shapes of each node (input modal order parameters) Each node contains spatial coordinates. and the mode vectors of the corresponding orders .

[0082] FEM calculation: A finite element model of the plate was built using Altair Hyper Mesh, selecting 4-node quadrilateral plate elements, resulting in a total of 50 nodes. ,like Figure 2 As shown. Modal analysis was performed using the Opti Struct solver, and the results were output to a .pch file. The first three bending mode shapes were extracted to obtain the FEM node set. ,in .

[0083] A KD-tree spatial index was constructed for 2000 DIC nodes, with a construction time of approximately 0.01 seconds. The point cloud registration method (ICP algorithm) was used to align the coordinates of the DIC node set and the FEM node set. After registration, the root mean square error decreased from the initial 14.916 mm to 1.919 mm.

[0084] Step S2: Adaptive neighborhood candidate matching, specifically:

[0085] In this embodiment, a maximum search radius is set. First, calculate the number of reference nodes. For each FEM node, statistics are performed with its center and radius... The number of DIC nodes within the sphere is calculated, and then the average value of all FEM nodes is taken. In this example, the global average density is approximately... Multiply by the volume of the sphere to get .

[0086] For each FEM node Calculate the number of DIC nodes within its local sphere. Then follow the formula Determine the search radius. Since this embodiment uses a plate structure (two-dimensional shell), an exponent is used. Adaptive results: Sparse node regions Dense areas at the ends Each FEM node receives an average of approximately 5–15 candidate DIC nodes, such as Figure 3 As shown in the figure, a single FEM node (marked with a pentagram) at coordinates (88.9, 50.0) is used as the observation object to visually demonstrate the effect of the density adaptive radius mechanism. In the figure, × represents the original DIC measurement points across the entire region, solid black dots are candidate DIC points for the current FEM node, the dashed circle represents the adaptively calculated retrieval radius r = 5.13 mm, and the outer dashed rectangle is the bounding box of the node's measurement points. Example 1 uses a two-dimensional shell structure with a dimension exponent α = 1 / 2. In this region, the DIC measurement point density is relatively high, so the algorithm automatically reduces the retrieval radius, retaining only nearby valid measurement points and filtering out redundant DIC points at a distance. If the algorithm switches to a sparse area at the edge of the specimen, the calculated r will increase to approximately 12 mm, ensuring sufficient candidate measurement points. Compared to the fixed radius retrieval scheme, this adaptive mechanism significantly reduces the number of candidate points in dense areas, reducing the subsequent similarity calculation overhead.

[0087] Step S3: Calculation of weighted similarity of multi-mode vibration modes

[0088] In this embodiment, spatial scale parameters are taken. Similarity threshold The top 10 candidate points are retained. The first three modes ( The corresponding weight is For each candidate matching pair Calculate spatial distance similarity Similarity to modal shape Then obtain the overall similarity. Typical matching results are shown in Table 1 below. The frequency distribution of MAC values ​​for each match is as follows: Figure 4 As shown in the figure, the horizontal axis of the histogram represents the MAC value, and the vertical axis represents the frequency of matching pairs. The data comes from all 50 groups of effective FEM-DIC matching nodes in Example 1. It can be seen that the MAC values ​​of the vast majority of matching pairs are concentrated in the range of 0.98 to 1, with the highest peak frequency approaching 0.995. Only a very small number of matches have a MAC value below 0.97, and the overall average reaches 0.9922. Compared with the average MAC value of 0.9649 for pure spatial matching and 0.9761 for the static mapping method, the weighted similarity and gradient constraint scheme of this invention significantly improves the correlation of mode shape matching, proving that simultaneously integrating spatial location and multi-mode features can significantly improve the physical rationality of pairing. The high MAC ratio indicates that the matching modes are highly consistent.

[0089] Table 1 Typical Matching Results

[0090]

[0091] Step S4: Global one-to-one matching and modal shape space continuity constraints, specifically including:

[0092] Step S4.1: Determine the set of FEM neighboring node pairs (sharing the same unit or the Euclidean distance is less than) (node ​​pairs).

[0093] Step S4.2: Use the modal spatial gradient method to pre-calculate based on the FEM node set. Calculate the minimum wavelength of the first three modes. Specifically, for the third (highest) mode, calculations are performed according to S4.2.1 to S4.2.5: through the neighborhood... Fit the gradient to the nearest neighbor nodes to obtain Normalized mode amplitude minimum half wavelength Taking the minimum value of the first three orders, we get... (The first two wavelengths are longer). Taking the oversampling coefficient t=4, the maximum permissible spacing is... .

[0094] Step S4.3: For each pair of adjacent nodes and its candidate DIC node, calculate and .

[0095] Step S4.4: If Delete the corresponding edge; otherwise, calculate the relative deviation penalty factor. In this embodiment, take In this example Therefore, it is not necessary to force .

[0096] Step S4.5: Multiply the original comprehensive similarity by Update edge weights. When the same edge is modified by multiple adjacent constraints, the weight is multiplied by all corresponding constraints. The product of.

[0097] Step S4.6: The Kuhn-Munkres algorithm is used to solve for the maximum weight matching in the bipartite graph. After applying constraints, none of the matching results violate spatial continuity, and the computation time is approximately 2.1 seconds. The matching connections of FEM and DIC nodes are as follows: Figure 5 As shown. Figure 5 The hollow circle represents 50 structured FEM mesh nodes, the densely scattered points represent 2000 unordered DIC measurement points, and the diagonal line segments are the algorithm output matching lines. All lines are short-distance segments, and there are no cross-regional connections between adjacent FEM nodes and distant DIC measurement points. Thanks to the mode gradient continuity penalty constraint in step S4, the spatial positions of the matching DIC measurement points corresponding to adjacent FEM meshes are synchronously adjacent, which conforms to the actual vibration deformation distribution law of the crane's thin plate. This solves the spatial fragmentation and non-physical pairing defects that occur in traditional independent point-by-point matching, and intuitively demonstrates the optimization effect of continuity constraints.

[0098] Step S5: Node validity verification and consistency check based on the Nyquist sampling theorem, specifically:

[0099] Average Spacing of DIC Nodes after Matching Maximum spacing ,like Figure 6 As shown, Figure 6 The curves are divided into two sets of X-direction coordinate error distributions: before registration and after registration. Before registration, the mean value is 14.916 mm, and the peak probability density falls in the range of 14~16 mm. After registration, the mean value is 1.919 mm, and the peak value shrinks to a narrow range of 1~2 mm.

[0100] In Example 1, there was a significant rigid body offset between the two sets of original coordinate systems of the measurement points. After the preprocessing and registration in step S1, the spatial reference height of the two sets of point clouds was unified, which greatly eliminated the mismatch caused by the coordinate deviation and provided a reliable coordinate basis for subsequent adaptive retrieval and accurate similarity calculation. This proves that the preprocessing and registration step is a necessary prerequisite for high-precision matching.

[0101] Based on the matched DIC node set, the highest effective modal order is taken. (The first three modes were reliably identified). The minimum half-wavelength was recalculated using the mode shape spatial gradient method in step S4.2: Nyquist condition test: This is much smaller than the threshold, satisfying the sampling theorem. Further calculations... Relative differences This does not trigger the FEM mesh resolution warning.

[0102] Step S6: Output the final set of matching and valid node pairs, specifically:

[0103] Output 50 one-to-one node pairs. Each matching pair contains the FEM node number, DIC node number, spatial coordinates, and the first three mode shapes.

[0104] Results verification: The MAC matrix of the experiment and simulation was calculated using the matched node pairs. The diagonal values ​​were all greater than 0.99, and the spatial distance between adjacent matched nodes was continuous, indicating that the matching was correct and reliable.

[0105] Quantitative comparison and verification of different matching methods:

[0106] To quantitatively verify the advantages of the method of the present invention compared with the prior art, the same test subject (length) as in Example 1 was used. ,Width ,thick Using the same DIC / FEM data (steel free plate), node matching was performed using the three methods shown in Table 2, and the average MAC value (first three modes) after matching was calculated.

[0107] Table 2 shows the node matching results using three methods.

[0108]

[0109] The comparison results show that the average MAC value (0.9922) calculated by the method of the present invention is significantly better than that of the pure spatial distance matching method (0.9649) and the method of Liu Hui et al. (2023) (0.9761), and the matching results satisfy the Nyquist sampling condition, which verifies the advantages of the present invention in matching accuracy and physical rationality.

[0110] Example 2:

[0111] The core algorithm of this invention is independent of any specific FEM solver. Taking the Altair HyperMesh+Opti Struct platform used in this embodiment as an example, the modal analysis results are output to a .pch file, which contains nodal coordinates and displacement components of each mode shape. The required node set and mode shape vectors can be extracted through the Hyper Mesh API or by directly parsing the ASCII code of the .pch file. Similarly, this invention can also be adapted to other finite element software such as Nastran and ANSYS, requiring only corresponding modifications to the data reading module.

[0112] In the test-simulation correlation analysis of crane main beam structures, this invention reduces the time required for manual matching of a single node from approximately 2 minutes to less than 2 seconds for fully automated matching of all 50 nodes, while maintaining spatial continuity in the matching results. This significantly improves analysis efficiency and reliability. Furthermore, the automatic screening mechanism based on the modal sampling theorem effectively avoids modal aliasing problems caused by insufficient node density.

[0113] Example 3:

[0114] An adaptive matching and filtering system for full-field DIC measured nodes and FEM calculated nodes for crane modal analysis is provided to implement the method described in Embodiment 1 above. The system includes a hardware processing unit and six functional modules mounted on the hardware processing unit: data acquisition and coarse alignment module, adaptive neighborhood candidate matching module, multi-mode weighted similarity calculation module, mode gradient constraint and global bipartite graph matching module, Nyquist sampling verification module, and matching result output module.

[0115] The data acquisition and coarse alignment module completes the reading of DIC and FEM node sets, the construction of KD tree spatial index, and the registration of two sets of point cloud ICP coordinates.

[0116] The adaptive neighborhood candidate matching module calculates the search radius for each FEM node according to the density adaptive formula and retrieves and generates a simplified candidate DIC set for each node.

[0117] The multi-modal weighted similarity calculation module solves for spatial similarity, MAC modal similarity, and adaptive order weight, respectively, and fuses them to obtain the comprehensive similarity of each matching pair and filters candidates according to a threshold.

[0118] The modal gradient constraint and global bipartite graph matching module calculates the modal gradient, minimum half-wavelength of the mode, and adjacent matching penalty coefficient, corrects the edge weights of the bipartite graph, and solves the global one-to-one optimal matching through the Kuhn-Munkres algorithm.

[0119] The Nyquist sampling verification module calculates the average node spacing of the DIC after matching, performs Nyquist sampling criterion verification and FEM mesh resolution consistency verification, and outputs corresponding prompt information.

[0120] The matching result output module outputs a complete matching node dataset or an empty set based on the sampling verification results. The output data carries complete supplementary information such as node number, coordinates, mode shape, and similarity.

[0121] Example 4:

[0122] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in Embodiment 1 above.

[0123] Example 5:

[0124] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1 above.

[0125] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and adjustments without departing from the principle of the present invention, and these improvements and adjustments should also be considered within the scope of protection of the present invention.

Claims

1. A full-field DIC measured node and FEM calculated node adaptive matching and screening method for crane modal analysis, characterized in that, Includes the following steps: Step S1, Data Acquisition and Preprocessing: Collect the actual DIC node set and FEM calculated node set of the crane, construct the DIC node spatial index and complete the coarse alignment of the two sets of node coordinates; Step S2, Adaptive Neighborhood Candidate Matching: Based on the local density of DIC measurement points around each FEM node, the search radius is adaptively generated, and the candidate DIC node set corresponding to each FEM node is obtained by filtering. Step S3: Calculate the weighted similarity of multi-mode vibrations: Combine spatial distance features and multi-mode vibration MAC features, use weights that are adaptively adjusted according to the modal order to calculate the comprehensive similarity, and simplify the candidate matching set; Step S4, Global one-to-one matching with mode shape spatial continuity constraints: Construct a bipartite graph using FEM and DIC nodes, determine the minimum half-wavelength of the mode and the adjacent matching distance constraint based on the mode shape spatial gradient, and solve for the global optimal one-to-one matching after correcting the edge weights of the bipartite graph. Step S5, Nyquist sampling validity verification: Based on the average spacing of DIC nodes after matching and the minimum half wavelength of the mode, the sampling criterion verification is completed to determine whether the measurement point meets the mode recognition sampling requirements; Step S6: Output matching results: If the sampling verification is successful, output a set of valid FEM-DIC matching node pairs; if the verification is unsuccessful, output an empty set.

2. The matching and screening method of claim 1, wherein, Step S1 specifically includes the following steps: Step S1.1, data reading: obtaining the DIC measured node set wherein is the spatial coordinate, is the pre order DIC modal shape vector, and M is the total number of DIC nodes, is the modal order parameter, and ; Obtain the FEM numerical computation node set ,in For spatial coordinates, is the vector of the first k modes of the FEM, and N is the total number of nodes in the FEM; Step S1.2, Spatial Index Construction: Construct a KD-tree spatial index for the unordered DIC point cloud nodes; Step S1.3, Point Cloud Coordinate Registration: The ICP iterative nearest point algorithm is used to solve the rotation and translation transformation matrices of the two sets of node sets, eliminating rigid body displacement and rotation deviations between the experimental and simulation coordinate systems, and completing the coarse alignment of coordinates.

3. The matching and filtering method according to claim 2, characterized in that, Step S2 specifically includes the following steps: Step S2.1, Parameter Preset: Set the maximum search radius ;Statistical analysis of all FEM nodes The number of global reference nodes is obtained by averaging the number of DIC nodes within the radius sphere. ; Step S2.2, Node-by-Node Adaptive Search Radius Calculation: DIC nodes located within this radius are retrieved from the spatial index as candidate matching nodes; the adaptive search radius of the current FEM node is calculated using the following formula: ; in, With this FEM node as the center and a radius of The number of DIC nodes inside the sphere; The dimension index is taken as the dimension index for a three-dimensional volume structure. For two-dimensional shell structures, take ; Step S2.3, Candidate Node Retrieval: Retrieve all DIC nodes within the r-space range in the KD tree to form the initial candidate set of the FEM node.

4. The matching and filtering method according to claim 3, characterized in that, Step S3 specifically includes the following steps: Step S3.1: For each candidate matching pair Calculate spatial distance similarity Similarity with mode shape ; (1) Formula for spatial distance similarity: ;in: The Euclidean distance between FEM and DIC nodes; This is a spatial scale parameter, and its value is 2 to 3 times the average node spacing of all FEM nodes. (2) Formula for similarity of MAC mode shapes: ; , These are the first k-th order mode shape vectors of FEM and DIC, respectively; Step S3.2: Calculate the modal adaptive weights: ; The time weight increases linearly with the order; when k>6, the weight is fixed at 0.

85. Step S3.3, Calculation of overall similarity S: ; Step S3.4, set the similarity threshold S th = 0.6, only keep the top 10 DIC nodes with the highest similarity scores and greater than the similarity threshold, i.e. Nc=10, to form the reduced candidate set.

5. The matching and filtering method according to claim 4, characterized in that, Step S4 specifically includes the following steps: Step S4.1: Identify adjacent node pairs in the FEM mesh: If two FEM nodes share the same cell, or the Euclidean distance between the nodes is less than 1.2 to 1.5 times the average cell size of the mesh, they are determined to be adjacent node pairs, and a set of adjacent node pairs is constructed. ; Step S4.2: Pre-calculate based on FEM node set The following modal space gradient method was used to calculate the previous values. Minimum wavelength of first mode and define the maximum allowable spacing. Where t is the modal oversampling coefficient, ; The modal space gradient method includes: Step S4.2.1: For the first First-order mode, for each node in a given set of nodes Through its neighborhood Using the nearest neighbor nodes, the tangent plane of the local mode shape is fitted using the least squares method, and the mode shape gradient vector is calculated. The gradient magnitude is The number of nearest neighbors is 6 ≤ K ≤ 15; Step S4.2.2: Traverse all nodes and determine the maximum gradient magnitude of this mode. ; Step S4.2.3: Determine the mode shape amplitude of this mode. If it has been normalized, then ; Step S4.2.4: Calculate the minimum half wavelength of this mode. ; This indicates the shortest total wavelength for the nth mode; Step S4.2.5: For the previous For each of the three modes, the minimum half-wavelength of each mode is taken as the final criterion, i.e. ; Step S4.3: Determine the adjacent matching distance and calculate the penalty factor: 1) For each pair of adjacent nodes and its candidate matching DIC nodes and ,calculate and Among them, nodes belong Candidate set, nodes belong The candidate set; 2) If Then, simply delete the edge from the bipartite graph. and ;like Then forced In other cases, the relative deviation penalty factor... Where β is the allowable threshold coefficient for relative distance deviation. ; Step S4.4: Bipartite graph edge weight correction: adjust the edge weights... and The original comprehensive similarity is multiplied by 1. If the same edge is modified by multiple adjacent constraints, then its original comprehensive similarity is multiplied by the sum of all the factors involving that edge. The product; Step S4.5: Use the maximum weight matching algorithm to find the global optimal matching for the corrected bipartite graph.

6. The matching and filtering method according to claim 5, characterized in that, Step S5, node validity verification based on the Nyquist sampling theorem, includes the following steps: Step S5.1: Calculate the average spatial spacing of all DIC nodes after matching. ; Step S5.2: Obtain the highest valid modal order of the matching DIC set. , The minimum half wavelength of the DIC set is calculated using the modal gradient method in step S4.

2. ; Step S5.3: Nyquist sampling decision: If This triggers the "DIC sampling does not meet the Nyquist condition" warning and jumps to step S6. Step S5.4: Perform grid resolution verification if sampling is qualified: If It outputs a "FEM mesh resolution insufficient" message without interrupting the process.

7. The matching and filtering method according to claim 6, characterized in that, Step S6, which outputs the final set of matching and valid node pairs, includes the following sub-steps: Step S6.1: If step S5 does not trigger the Nyquist sampling warning, output all one-to-one matching node pairs; each matching pair includes FEM number, DIC number, three-dimensional coordinates, mode shape vectors of each order, and comprehensive similarity; Step S6.2: If step S5 triggers a Nyquist sampling warning, output an empty set directly.

8. An adaptive matching and selection system for full-field DIC measured nodes and FEM computed nodes for crane modal analysis, characterized in that, To implement the method described in any one of claims 1 to 7, the system includes a hardware processing unit and six functional modules mounted on the hardware processing unit: a data acquisition and coarse alignment module, an adaptive neighborhood candidate matching module, a multi-mode weighted similarity calculation module, a mode gradient constraint and global bipartite graph matching module, a Nyquist sampling verification module, and a matching result output module. The data acquisition and coarse alignment module completes the reading of DIC and FEM node sets, the construction of KD tree spatial index, and the registration of two sets of point cloud ICP coordinates. The adaptive neighborhood candidate matching module calculates the search radius for each FEM node according to the density adaptive formula and retrieves and generates a simplified candidate DIC set for each node. The multi-modal weighted similarity calculation module solves for spatial similarity, MAC modal similarity, and adaptive order weight, respectively, and fuses them to obtain the comprehensive similarity of each matching pair and filters candidates according to a threshold. The modal gradient constraint and global bipartite graph matching module calculates the modal gradient, minimum half-wavelength of the mode, and adjacent matching penalty coefficient, corrects the edge weights of the bipartite graph, and solves the global one-to-one optimal matching through the Kuhn-Munkres algorithm. The Nyquist sampling verification module calculates the average node spacing of the DIC after matching, performs Nyquist sampling criterion verification and FEM mesh resolution consistency verification, and outputs corresponding prompt information. The matching result output module outputs a complete matching node dataset or an empty set based on the sampling verification results. The output data includes the node number, coordinates, mode shape, and similarity information.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images