Crack estimation device and crack estimation method

Abstract

A crack estimation device for estimating cracks inside a structure with high precision comprises: a measuring unit (10) for measuring the deformation of a measuring surface as a measuring surface deformation vector; a model generating unit (30) for setting the deformation of the measuring surface when a crack occurs on the crack occurrence surface as a measuring surface estimation change vector for a plurality of crack candidates; and a crack state analyzing unit (40) for calculating the similarity between the measuring surface deformation vector and the measuring surface estimation change vector, normalizing the similarity, and estimating cracks (4) occurring on the crack occurrence surface based on the result obtained by multiplying a vector representing the state of the crack occurrence surface for each crack candidate by the normalized similarity and adding the result for all crack candidates.

Description

Technical Field

[0001] The present application relates to a crack estimation device and a crack estimation method. Background Art

[0002] Generally speaking, parts of mechanical components cannot be inspected visually. Cracks can grow unnoticed during standard inspections, potentially affecting the lifespan of the mechanical structure. For example, in rotating electrical machines used in turbine generators, cracks within the rotor structure cannot be detected by standard visual inspection. Therefore, cracks can grow unnoticed during standard inspections, potentially affecting the lifespan of the turbine generator including the rotor structure. Therefore, crack size estimation methods are known as non-destructive inspection methods for detecting cracks within structures (see, for example, Patent Document 1).

[0003] Prior art literature

[0004] Patent Document 1: Japanese Patent Application Laid-Open No. 2012-159477 Summary of the Invention

[0005] In previous crack size estimation methods, the location and size of the cracks inside the structure are estimated by inverse analysis based on the changes in the shape of the structure's surface. In order to perform the inverse analysis, it is necessary to solve the inverse problem. In order to solve the inverse problem, the following three requirements must be met: the solution to the inverse problem can be uniquely determined as the uniqueness of the solution, the solution to the inverse problem exists as the existence of the solution, and the stability of the inverse problem can be maintained as the stability of the solution. However, depending on the results of strain measurement, the three requirements of "uniqueness of solution", "existence of solution" and "stability of solution" are sometimes not met. If any one of these three requirements is not met, the inverse problem will become a bad setting problem, that is, an improper problem, and the accuracy of crack estimation will be reduced.

[0006] The present application has been made to solve the above-mentioned problems, and an object of the present application is to provide a crack estimation device capable of estimating cracks inside a structure with high accuracy.

[0007] The crack inference device disclosed in the present application is characterized in that it comprises: a measuring unit, which takes an observation surface on the surface of a structure as a measuring surface and measures the deformation of the measuring surface as a measuring surface deformation vector; a model generating unit, which generates a shape model obtained by modeling the shape of the structure, takes a candidate surface inside the structure as a crack occurrence surface, and sets the deformation of the measuring surface when a crack occurs on the crack occurrence surface as a measuring surface inference change vector for multiple types of crack candidates; and a crack state analyzing unit, which infers cracks based on the output of the measuring unit and the output of the model generating unit, the crack state analyzing unit calculates the similarity between the measuring surface deformation vector and the measuring surface inference change vector, normalizes the similarity, and infers the cracks occurring on the crack occurrence surface based on the result obtained by multiplying a vector of a state quantity representing the state of the crack occurrence surface by the normalized similarity for each crack candidate and adding the result for all crack candidates.

[0008] The crack inference device disclosed in the present application calculates the similarity between the measured surface deformation vector and the measured surface inference change vector, normalizes the similarity, and infers cracks occurring on the crack inference surface inside the structure based on the result obtained by multiplying the vector of the state quantity representing the state of the crack occurrence surface for each crack candidate by the normalized similarity and adding the result for all crack candidates. Therefore, the cracks inside the structure can be inferred with high precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a block diagram showing the configuration of the crack estimation device according to the first embodiment.

[0010] Figure 2 This is a perspective view of the structure in the state where a tensile load is applied in the first embodiment.

[0011] Figure 3 This is a perspective view of the structure in a state where a bending moment is applied in the first embodiment.

[0012] Figure 4 This is a diagram showing reference coordinates set for candidate surfaces of a structure in the first embodiment.

[0013] Figure 5 This is a diagram showing how a candidate surface is divided into elements in the first embodiment.

[0014] Figure 6 This is a diagram showing reference coordinates set for the observation plane of the structure in the first embodiment.

[0015] Figure 7 This is a diagram showing how the observation plane is divided into elements in the first embodiment.

[0016] Figure 8 This is a diagram showing displacement change vectors of candidate surfaces in the first embodiment.

[0017] Figure 9 Graph showing the strain change vector of the observation surface in the first embodiment.

[0018] Figure 10 This is a diagram showing load change vectors of candidate surfaces in the first embodiment.

[0019] Figure 11 This is a diagram showing an example of the sides of candidate surfaces for setting the starting point of crack growth in the first embodiment.

[0020] Figure 12 This is a diagram showing a first example of a node identified as a crack in a candidate surface for which a starting point for crack growth is set in the first embodiment.

[0021] Figure 13 This is a diagram showing a second example of a node identified as a crack in a candidate surface for which a starting point for crack growth is set in the first embodiment.

[0022] Figure 14 This is a diagram showing learning data in the first embodiment.

[0023] Figure 15 This is a flowchart showing the operation of the crack estimation device in the first embodiment.

[0024] Figure 16 This is a flowchart showing details of the learning data creation process in the first embodiment.

[0025] Figure 17 This is a flowchart showing details of the learning data creation process in the first embodiment.

[0026] Figure 18 This is a flowchart showing details of the learning data creation process in the first embodiment.

[0027] Figure 19 This is a flowchart showing details of the estimation process in the first embodiment.

[0028] Figure 20 This is a flowchart showing an example of output processing in the first embodiment.

[0029] Figure 21 This is a flowchart showing an example of output processing in the first embodiment.

[0030] Figure 22 This is a diagram showing reference coordinates set for a cylindrical member as another structure in the first embodiment.

[0031] Figure 23 This is a diagram showing a state in which internal pressure is applied to a cylindrical member as another structure in the first embodiment.

[0032] Figure 24 This is a block diagram showing the configuration of a crack estimation device according to the second embodiment.

[0033] Figure 25 This is a flowchart showing the processing in the learning phase and the inverse analysis phase in the second embodiment.

[0034] Figure 26 This is a block diagram showing the configuration of a crack estimation device according to the third embodiment.

[0035] Figure 27 3 is a diagram showing the displacement change vector of the observation plane in the third embodiment.

[0036] Figure 28 This is a diagram showing the angle change vector of the observation plane in the third embodiment.

[0037] Figure 29 This is a block diagram showing the configuration of a crack estimation device according to the fourth embodiment.

[0038] Figure 30 This is a diagram showing candidate surfaces of a structure in the fourth embodiment.

[0039] Figure 31 This is a flowchart showing the process of determining candidate faces in the fourth embodiment.

[0040] Figure 32 This is a schematic diagram showing an example of hardware of the crack estimation device according to the first, third, and fourth embodiments.

[0041] Figure 33 This is a schematic diagram showing an example of hardware of the crack estimation device according to the second embodiment. DETAILED DESCRIPTION

[0042] Hereinafter, a crack estimation device according to an embodiment of the present application will be described in detail with reference to the accompanying drawings.

[0043] Implementation method 1.

[0044] Figure 1 This is a block diagram showing the configuration of the crack estimation device 100 according to the first embodiment. Figure 2 This is a perspective view showing a state where a tensile load 5 is applied to a structure 1 when the structure 1 as a prediction target of the crack prediction device 100 according to the first embodiment is a flat plate. Figure 3This is a perspective view showing a state in which a bending moment 6 is applied to the structure 1 when the structure 1 as the prediction target of the crack prediction device 100 according to the first embodiment is a flat plate.

[0045] like Figure 1 As shown, the crack estimation device 100 includes a measuring unit 10 and an estimation unit 20. The estimation unit 20 estimates Figure 2 as well as Figure 3 The positions and sizes of the cracks 4 inside the structure 1 are shown.

[0046] A candidate surface 3 is set inside the structure 1, and an observation surface 2 is set on the surface of the structure 1. Figure 2 as well as Figure 3 In FIG, a flat plate serving as structure 1 is shown using an orthogonal coordinate system. The surface on which observation surface 2 is set is defined as the xz plane, and the surface on which candidate surface 3 is set is defined as the xy plane. Candidate surface 3 is set at a location where crack 4 is assumed to occur. Observation surface 2 is set within a range where the surface of structure 1 changes due to changes in candidate surface 3.

[0047] The measuring unit 10 of the crack estimation device 100 uses at least a portion of the surface of the structure 1 as the observation surface 2 to measure the deformation of the surface of the observation surface 2. The measuring unit 10 is, for example, a strain gauge installed on the observation surface 2. The strain gauge is composed of a substrate and a resistor material. The material of the substrate is composed of an electrical insulator. The resistor material is installed on the substrate, and a lead wire is provided at a portion protruding from the substrate. The substrate is installed on the surface of the structure 1 via an adhesive, and when the substrate expands or contracts, the resistor material also expands or contracts, and the resistance of the resistor material changes. The lead wire of the resistor material is connected to the data acquisition unit 41 of the estimation unit 20. For example, when the surface of the structure 1 is strained, the resistor material expands or contracts, and the resistance of the resistor material changes. The change in the resistance of the resistor material is transmitted to the data acquisition unit 41 of the estimation unit 20 via the lead wire. Thus, the strain change on the surface of the structure 1 is measured by the strain gauge, and the measurement result is input to the data acquisition unit 41 of the estimation unit 20. With such a structure, it is possible to obtain the result after applying the strain. Figure 2 Tensile load 5 or Figure 3 In the state of the bending moment 6, the measuring unit 10 measures the strain change of the observation surface 2 on the surface of the structure 1. The measuring unit 10 uses the observation surface 2 as the measurement surface and measures the deformation of the measurement surface as the measurement surface deformation vector.

[0048] Alternatively, the measuring unit 10 may be composed of an optical device such as a digital camera and a device that analyzes image information obtained by the optical device. In this case, the strain of the surface of the observation surface 2 is measured in a non-contact manner by obtaining the correlation of the image information obtained by the optical device.

[0049] The estimation unit 20 estimates the cracks 4 within the structure 1 based on changes in the measurement surface measured by the measurement unit 10. The estimation unit 20 estimates the cracks 4 within the structure 1 by using inverse analysis to determine the relationship between changes in the surface shape of the structure 1 and the cracks 4 within the structure 1. The stages processed by the estimation unit 20 include a learning stage and an inverse analysis stage. The inverse analysis stage is performed after the learning stage. In the learning stage, the relationship between the cracks 4 within the structure 1 and the changes in the surface shape of the structure 1 is prepared in advance as learning data. Furthermore, in the inverse analysis stage, the position and size of the cracks 4 are estimated as information on the cracks 4 within the structure 1 using the learning data prepared in the learning stage.

[0050] Such inferences typically use learning data and the least squares method, requiring the calculation of a pseudo-inverse matrix. Therefore, such inferences boil down to solving an inverse problem. Solving an inverse problem requires satisfying three requirements: "uniqueness of solution," "existence of solution," and "stability of solution." However, depending on the strain measurement results of the measurement surface obtained by the measurement unit 10 and the learning data, these three requirements may not be met. For example, if the number of unknowns is greater than the number of observed quantities, countless solutions exist, so "uniqueness of solution" is not satisfied. For example, if the number of unknowns is less than the number of observed quantities, no solution exists, so "existence of solution" is not satisfied. For example, even if stress in the structure 1 causes strain, if the effect of the strain decays more rapidly away from the strained portion, "stability of solution" is not satisfied. Therefore, the inverse problem may become a poorly specified problem, or an inappropriate problem. Therefore, even if one attempts to estimate the location and size of the crack 4 using learning data, a pseudo-inverse matrix may not exist if the problem is inappropriate.

[0051] Therefore, in the crack estimation device 100 according to the first embodiment, the shape model generation unit 31 of the estimation unit 20 models the shape of the structure 1 into a shape model. Furthermore, the estimation model generation unit 32 generates learning data, serving as the estimation model, based on the shape model. The crack state estimation unit 42 of the estimation unit 20 calculates the similarity between the learning data, serving as the estimation model, and the measured surface deformation vector obtained by the measurement unit 10. The calculated similarities are normalized so that the total value becomes 1, thereby obtaining a coefficient vector. Furthermore, the crack state estimation unit 42 of the estimation unit 20 estimates changes in the crack initiation surface, using the candidate surface 3 as the crack initiation surface, based on the coefficient vector and another portion of the estimation model.

[0052] The analysis result output unit 60 displays the remaining service life of the structure 1 or issues a warning to stop use based on the information of the structure 1 obtained from the storage unit 50 and the information of the load applied to the structure 1 and the estimated result of the crack 4 obtained from the crack analysis unit 44 .

[0053] The model generation unit 30 includes a shape model generation unit 31 and an inference model generation unit 32. The shape model generation unit 31 generates a shape model. The inference model generation unit 32 generates a structural analysis model based on the shape model, and generates an inference model based on the structural analysis model. The generated inference model is different depending on the structural analysis model. The structural analysis model is the model used when performing structural analysis.

[0054] To perform structural analysis, a structural analysis model and boundary conditions for the structural analysis model are required. Boundary conditions consist of load conditions and constraints. Therefore, structural analysis requires three things: a structural analysis model, load conditions, and constraints.

[0055] When performing structural analysis using a structural analysis model, load conditions and constraint conditions are defined. Load conditions define the extent of load applied to the structure, i.e., the force vector information at the load-applying locations in the structural model. Constraint conditions, on the other hand, define how and where the structure is supported, i.e., information that minimizes deformation at the supported locations in the structural analysis model.

[0056] The boundary conditions are different depending on the shape model generated. The shape model is a model of the inspection object generated as a whole or a part of the structure 1 based on the measurement surface and the crack occurrence surface.

[0057] When the entire structure 1 is used as a shape model, a temperature distribution can be added as a boundary condition. When using a temperature distribution, for example, information on a known uniform temperature distribution is initially applied to the structural analysis model at a set initial temperature as a load. Subsequently, at an analysis temperature different from the set initial temperature, the entire structure is expanded or contracted according to the difference between the initial temperature and the analysis temperature, thereby performing structural analysis.

[0058] When a part of the structure 1 is used as the shape model, information on displacement changes or load distribution in a surface cut out as the part of the structure 1 is given as a boundary condition.

[0059] When performing structural analysis based on boundary conditions, a model consisting of the measurement surface and crack initiation surface of the shape model divided into a grid pattern is used as the structural analysis model. By dividing candidate surface 3 into a grid pattern, the crack initiation surface is generated as part of the structural analysis model. Furthermore, by dividing observation surface 2 into a grid pattern, the measurement surface is generated as another part of the structural analysis model.

[0060] Figure 4 It shows Figure 2 as well as Figure 3 An example of reference coordinates set for the candidate surface 3 of the structure 1 in FIG. Figure 5 It shows Figure 4 The candidate face 3 is divided into elements 7. The candidate face 3 is divided into n pieces in the x-axis direction and m pieces in the y-axis direction. The points where the grids intersect are represented by positions (i, j). Positions (i, j) are represented by numbers from (0, 0) to (n, m). When the points where the grids intersect are set as nodes, each node is a point on the line that forms element 7. In addition, Figure 5 In FIG, element 7 is represented by a square, but is not limited thereto, and may be, for example, a trapezoid.

[0061] For each node position in the crack occurrence surface, a structural analysis of the crack occurrence surface is performed. For example, in the case where crack 4 occurs at the node at the position (0, 0) in the crack occurrence surface, a structural analysis is performed on the displacement changes of all nodes in the crack occurrence surface from the position (0, 0) to the position (n, m). In this case, the node at the position (0, 0) corresponds to crack 4, so it becomes a hollow. Therefore, a displacement change occurs at the position (0, 0). On the other hand, since it is assumed that there is no crack 4 at the nodes at positions other than (0, 0), no displacement change in the load direction occurs according to the boundary conditions. In addition, by performing a structural analysis of the displacement change of the crack occurrence surface for each node position in this way, the amount of learning data is limited, and the generation time of the learning data can be limited.

[0062] Next, for example, if crack 4 occurs at the node at position (0, 1) on the crack initiation plane, structural analysis is performed on the displacement changes of all nodes in the crack initiation plane from position (0, 0) to position (n, m). In this case, the node at position (0, 1) corresponds to crack 4 and therefore becomes a void. Therefore, displacement changes occur at position (0, 1). On the other hand, since it is assumed that there are no cracks 4 at nodes other than (0, 1), no displacement changes occur in the load direction due to the boundary conditions.

[0063] Next, a structural analysis is performed similarly on the displacement changes of all nodes in the crack initiation plane, for nodes located outside of (0, 0) and (0, 1). Specifically, assuming that a crack 4 has occurred at each node in the crack initiation plane, the displacement changes of all nodes in the crack initiation plane are calculated. Information on at least the maximum displacement change among the thus calculated displacement changes is stored in the storage unit 50. The order of the positions of the nodes designated as crack 4 is predetermined.

[0064] In other words, the following relationship is set for each node and boundary condition in the crack occurrence surface. First, at the nodes in the crack occurrence surface with constraint conditions, the change in the direction of the constraint is set to zero. As a result, the nodes in the crack occurrence surface with constraint conditions do not move in the direction of the constraint. On the other hand, at the nodes in the crack occurrence surface with load conditions that do not crack 4, the change in load in a specific direction is set to other than zero. In addition, at the nodes in the crack occurrence surface with load conditions that crack 4, the change in load in all directions is set to zero.

[0065] Figure 6 It shows the Figure 2 as well as Figure 3 An example of reference coordinates set for the observation plane 2 of the structure 1 in FIG. Figure 7 It shows Figure 6 The observation plane 2 is divided into elements 8. The observation plane 2 is divided into n pieces in the x-axis direction and p pieces in the z-axis direction. The points where the divided grids intersect are represented by positions (k, l). Positions (k, l) are represented by numbers from (0, 0) to (n, p). When the points where the grids intersect are set as nodes, each node is a point on the line that forms element 8. In addition, Figure 7 The middle element 8 is represented by a square, but is not limited thereto and may be, for example, a trapezoid.

[0066] A structural analysis of the measurement surface is performed for each node position on the crack initiation surface. For example, if crack 4 occurs at the node at position (0, 0) on the crack initiation surface, a structural analysis is performed on the deformation of all nodes on the measurement surface from position (0, 0) to position (n, p). In the crack estimation device 100 of the first embodiment, strain changes are used as the deformation of the nodes on the measurement surface. Next, for example, if crack 4 occurs at the node at position (0, 1) on the crack initiation surface, a structural analysis is performed on the strain changes of all nodes on the measurement surface from position (0, 0) to position (n, p).

[0067] Next, a structural analysis is performed on the strain changes at all nodes on the measurement surface, also at nodes located outside of (0, 0) and (0, 1) on the crack initiation plane. Specifically, assuming that a crack 4 has occurred at each node on the crack initiation plane, the strain changes at all nodes on the measurement surface are calculated. Information on at least the maximum strain change among the calculated strain changes is stored in the storage unit 50.

[0068] In other words, the following relationship is established between each node and the boundary conditions in the measurement surface. First, for nodes in the measurement surface with constraint conditions, the change in the direction of the constraint is set to zero. This prevents the nodes in the measurement surface with constraint conditions from moving in the direction of the constraint. On the other hand, for nodes in the measurement surface with load conditions, the change in the load in a specific direction is set to something other than zero.

[0069] In addition, as the strain when the tensile load 5 on the z-axis or the bending moment 6 on the zx plane is applied, the main strain, the equivalent strain defined by the Tresca yield criterion, or the equivalent strain defined by the Mises yield criterion can also be used.

[0070] To summarize the above, the estimated model generation unit 32 of the model generation unit 30 initially performs structural analysis based on pre-set boundary conditions on the shape model generated by the shape model generation unit 31 based on the measurement surface and the crack initiation surface. Next, the structural analysis generates a plurality of estimated measurement surface change vectors, which are derived by estimating changes in the measurement surface. Furthermore, the structural analysis generates a plurality of estimated crack initiation surface change vectors, which are derived by estimating changes in the displacement of the crack initiation surface as changes in the crack initiation surface. Furthermore, an estimated model is generated, which is composed of the generated estimated measurement surface change vectors and estimated crack initiation surface change vectors.

[0071] Specifically, the estimated model generation unit 32 of the model generation unit 30 applies the boundary condition that no cracks 4 have occurred to each node on the crack initiation surface in the structural analysis model. Next, the displacement change at each node on the crack initiation surface in the structural analysis model is calculated. Furthermore, the strain at each node is calculated as the deformation of each node on the measurement surface in the structural analysis model.

[0072] Furthermore, the estimated model generation unit 32 of the model generation unit 30 applies boundary conditions to each node on the crack initiation surface in the structural analysis model, setting each node on the crack initiation surface as a crack. Subsequently, similarly to the above, the strain at each node is calculated as the displacement change at each node on the crack initiation surface and the deformation at each node on the measurement surface.

[0073] The estimated model generating unit 32 of the model generating unit 30 generates a displacement change vector resulting from a difference in displacement change amounts of nodes on the crack initiation surface in the structural analysis model.

[0074] Figure 8 It shows Figure 5 The displacement change vector at each position of the crack 4 of the candidate surface 3 is caused by the difference in the displacement change of each node of the candidate surface 3. Figure 8 As shown, the displacement data of each node included in the column vector of Δ(-,-) are arranged in the order of moving the crack 4 assumed at each node. Here, "-" represents meaningless indefinite data. In the following description, "-" also represents meaningless indefinite data. δ(i, j) is Figure 5 The displacement change of the node at the position (i, j) in the candidate face 3. Moreover, for example, δ 0，0 (i, j) is the displacement data of the node at position (i, j) when crack 4 occurs at the node at position (0, 0), and Δ(0, 0) is the displacement change vector when crack 4 occurs at the node at position (0, 0).

[0075] The following formula (1) represents Figure 8 The crack surface matrix Δ composed of multiple displacement change vectors crack_diff As Figure 8 The displacement change vectors Δ(0, 0) to Δ(n, m) shown in the figure are column vectors. Arranging these column vectors in the order of moving the crack 4 assumed at each node yields the result of Δ(n, m) shown in equation (1). crack_diff .

[0076] [Mathematical formula 1]

[0077]

[0078] Furthermore, the estimated model generating unit 32 of the model generating unit 30 creates a strain change vector due to the difference in strain at the nodes of the measurement surface in the structural analysis model.

[0079] Figure 9 It shows Figure 5 At each position of the crack 4 of the candidate surface 3 shown, Figure 7 The diagram of the strain change vector caused by the difference in strain at each node of the observation surface 2 is shown. Figure 9 As shown, the strain data of each node included in the column vector of E(-,-) are arranged in the order of moving the crack 4 assumed at each node.

[0080] ε(i, j) is Figure 7 The strain data of the node at the position (i, j) in the observation surface 2. 0，0 (k, l) is the strain data of the node at the position (k, l) of the observation surface 2 when a crack 4 occurs at the node at the position (0, 0) of the candidate surface 3, and E(0, 0) is the strain change vector when a crack 4 occurs at the node at the position (0, 0) of the candidate surface 3.

[0081] The following formula (2) is expressed by Figure 9 The measurement surface matrix E composed of multiple strain change vectors measure As Figure 9 The strain change vectors E(0, 0) to E(n, m) shown in the equation (2) are column vectors. Arranging these column vectors in the order of moving the assumed crack 4 at each node yields the following result: measure .

[0082] [Mathematical formula 2]

[0083]

[0084] Then, the estimation model generating unit 32 of the model generating unit 30 creates a load change vector due to the difference in load change amount at the node of the crack initiation surface in the structural analysis model.

[0085] Figure 10 It shows Figure 5 The load change vector at each position of the crack 4 of the candidate surface 3 is shown as a graph caused by the difference in the load change amount of each node of the candidate surface 3. Figure 10 As shown, the load data of each node included in the column vector of Z(-,-) are arranged in the order of moving the crack 4 assumed at each node. 0，0 (i, j) represents the load data for the node at (i, j) when a crack 4 occurs at the node at (0, 0), and Z(0, 0) represents the load change vector when a crack 4 occurs at the node at (0, 0). Specifically, the force at the node where a crack 4 occurs is zero, and the force at the node where no crack 4 occurs is zero.

[0086] The following formula (3) is expressed by Figure 10 The crack surface load matrix Z composed of multiple load change vectors crack_diff As Figure 10 The load change vectors Z(0, 0) to Z(n, m) shown in FIG. 1 are column vectors. Arranging these column vectors in the order of moving the crack 4 assumed at each node yields the Z shown in equation (3). crack_diff .

[0087] [Mathematical formula 3]

[0088]

[0089] The previously calculated Δ crack_diff 、E measure and Z crack_diff The relationship is expressed by the following equations (4) and (5).

[0090] [Formula 4]

[0091] E measure =HΔ crack_diff …(4)

[0092] [Formula 5]

[0093] Z crack_diff =GΔ crack_diff +Z no_crack …(5)

[0094] In equations (4) and (5), H is the observation matrix and G is the stiffness matrix. no_crack is a vector representing the load at each node of the candidate surface 3 under the boundary condition where no cracks 4 have occurred, obtained by the estimated model generation unit 32 of the model generation unit 30. By transforming Equations (4) and (5), the observation matrix H and stiffness matrix G are obtained using the following Equations (6) and (7).

[0095] [Formula 6]

[0096] H=E measure Δ crack_diff -1 …(6)

[0097] [Formula 7]

[0098] G=(Z crack_diff -Z no_crack )Δ crack_diff -1 …(7)

[0099] Here, the elements of the stiffness matrix G are shown in equation (8).

[0100] [Formula 8]

[0101]

[0102] Using the relationship obtained above, calculations simulating the development of cracks on the crack initiation surface are performed to obtain the crack shape as learning data. Figure 11 This is a diagram showing an example of the side 9 of the crack initiation surface that sets the starting point of crack growth. Figure 12FIG. 1 is a diagram showing a first example of a node identified as a crack 4 in a crack generating surface where a starting point of crack development is set. Figure 12 In the example, the node at the position (0, 0) on the edge 9 is determined as the crack 4. Thus, the crack 4 is set at the position (0, 0).

[0103] Figure 13 FIG. 2 is a diagram showing a second example of a node identified as a crack 4 in a crack generating surface where a starting point of crack development is set. Figure 13 In the example, the node at (3, 1) is identified as Crack 4. In this case, the node at (3, 1) where the load data value is the largest in the load change vector of the crack occurrence surface when Crack 4 is present at (3, 0) on edge 9 is set as the new Crack 4. For example, a method for setting the new Crack 4 may be to determine the elements that expand Crack 4 through morphological calculations. Thus, Crack 4 is set at (3, 1) and (3, 0). Learning data is calculated by performing structural analysis under these boundary conditions.

[0104] Here, the learning data includes a latent variable vector Γ indicating whether a node of the crack occurrence surface is crack 4 or not, a displacement change vector Δ caused by the difference in displacement change of the nodes of the crack occurrence surface, a load change vector Z caused by the difference in load change of the nodes of the crack occurrence surface, and a strain change vector E caused by the difference in strain of the measurement surface. Here, the latent variable vector Γ represents the position and size of the crack candidate in the crack occurrence surface. The latent variable vector Γ, the displacement change vector Δ, and the load change vector Z are vectors composed of state quantities representing the state of each node of the crack occurrence surface for each crack candidate, and the strain change vector E is a vector composed of state quantities representing the state of each node of the measurement surface for each crack candidate. Here, as Figure 12 The point on the edge 9 of the crack initiation surface calculated by the estimation model generation unit 32 is used as the starting point. That is, (i, j) = (i, 0) is set as the starting point of the crack 4. The latent variable vector Γ at this time (0) (i，0) is Γ (i，0) , displacement change vector Δ (0) (i，0) is Δ (i，0) , strain change vector E (0) (i，0) It's E (i，0) , load change vector Z (0) (i，0) It's Z (i，0) .

[0105] Next, the load change vector Z (0) (i，0)The node that becomes the largest is set as the next crack 4, which is a new crack candidate. Here, (i, j) = (i, 1) is set as the next crack 4. The latent variable vector Γ when this second crack candidate is set is (1) (i，0) As shown in equation (9), the displacement change vector Δ (1) (i，0) It becomes as shown in formula (10).

[0106] [Formula 9]

[0107]

[0108] [Formula 10]

[0109]

[0110] The displacement change vector Δ shown in formula (10) (1) (i，0) and load change vector Z (1) (i，0) The relationship becomes as shown in formula (11).

[0111] [Mathematical formula 11]

[0112]

[0113] In the latent variable vector Γ (1) (i，0) The latent variable γ (i，0) At the node where (i, j) becomes 1, that is, the node where the crack 4 is set, the load ζ (i，0) (i, j) is zero, the displacement δ (i，0) Since (i, j) is not zero, the load change vector of these data is extracted from equation (11) and is set as Z (1)’ (i，0) It is expressed as formula (12).

[0114] [Mathematical formula 12]

[0115]

[0116] In formula (12), only the load ζ is extracted (i，0) Since (i, j) is zero data, equation (12) becomes as equation (13).

[0117] [Mathematical formula 13]

[0118]

[0119] According to formula (13), the displacement change vector Δ (1)’ (i，0)It can be obtained as shown in formula (14).

[0120] [Mathematical formula 14]

[0121]

[0122] The displacement change vector Δ obtained by equation (14) (1)’ (i，0) Add the displacement δ of the node where crack 4 is not set (i，0) (i, j) information to find the displacement change vector Δ (1) (i，0) , using the observation matrix H and the stiffness matrix G, the strain change vector E is obtained by equations (15) and (16): (1) (i，0) and load change vector Z (1) (i，0) .

[0123] [Mathematical formula 15]

[0124]

[0125] [Formula 16]

[0126]

[0127] The latent variable vector Γ obtained in this way is (1) (i，0) , the displacement change vector Δ caused by the difference in displacement changes of the nodes on the crack occurrence surface (1) (i，0) , Load change vector Z caused by the difference in load change at the nodes of the crack occurrence surface (1) (i，0) , the strain change vector E caused by the difference in strain at the nodes of the measurement surface (1) (i，0) The data are stored as learning data in groups for each crack candidate.

[0128] Next, the load change vector Z (1) (i，0) The largest node is set as the next crack 4, and the latent variable vector Γ is set (2) (i，0) and the displacement change vector Δ (2) (i，0), perform the processing corresponding to equations (9) to (16). Perform the processing corresponding to equations (9) to (16) a total of q times. Furthermore, the processing corresponding to equations (9) to (16) is also performed a total of q times at the next starting point of edge 9 of the crack occurrence surface, and this is performed q times at all nodes on edge 9 of the crack occurrence surface, thereby obtaining learning data from all starting points. Thus, as the number of instances of crack development performed q times from (n+1) nodes, (n+1)*(q+1) learning data is obtained. Figure 14 is the learning data created in the estimation model generation unit 32. Figure 14 In this example, the number of columns is (n+1)*(q+1), and s represents a number greater than or equal to 0 and less than or equal to q. By presetting multiple starting points and developing cracks from each of the set starting points according to predetermined conditions to determine crack candidates, the number of crack candidates is limited, allowing learning data to be generated with a limited processing volume.

[0129] Here, the learning data in the estimation model generating unit 32 is obtained by structural analysis. However, the learning data may be generated by creating a structure including a plurality of cracks 4 and measuring the strain change on the surface at that time.

[0130] Next, the operation of the measuring unit 10 will be described. The measuring unit 10 measures the strain changes on the surface of the observation surface 2 of the structure 1 for each of the following conditions: a case where there are no cracks 4 within the flat plate of the structure 1; and a case where cracks 4 are present within the flat plate of the structure 1. The measured strain changes are arranged as column vectors in the same order as the assumed movement of the cracks 4 at each node, resulting in the following equation (17).

[0131] [Mathematical formula 17]

[0132]

[0133] The measuring unit 10 measures the column vector shown in equation (17) as the measurement surface deformation vector. In the measurement surface deformation vector, the subscript "0*0" indicates that Figure 7 The measurement surface deformation vector thus measured is output to the data acquisition unit 41 of the crack state analysis unit 40 .

[0134] Next, the operation of the crack state analyzer 40 will be described. The crack state analyzer 40 includes a data acquisition unit 41 and a crack state estimation unit 42. The data acquisition unit 41 acquires the measurement surface deformation vector output from the measurement unit 10 and outputs it to the vector similarity calculation unit 43 of the crack state estimation unit 42.

[0135] Next, the operation of the crack state estimation unit 42 will be described. The crack state estimation unit 42 includes a vector similarity calculation unit 43 and a crack analysis unit 44. The vector similarity calculation unit 43 receives information from the estimation model generation unit 32 about the crack state. Figure 14 The strain change vector E caused by the difference in strain at the nodes of the measurement surface in the learning data shown is (s) (i，0) , and set it as the estimated change vector of the measurement surface.

[0136] In order to determine the similarity between the measurement surface deformation vector and the estimated measurement surface change vector expressed in equation (17), the vector similarity calculation unit 43 calculates the Euclidean distance, which is the L2 norm, as shown in equation (18). By using the Euclidean distance as the similarity, a highly accurate similarity can be determined with limited processing power.

[0137] [Mathematical formula 18]

[0138]

[0139] Here, it is assumed that the Euclidean distance α obtained by formula (18) (s) (i，0) The variance of the measurement surface deformation vector σ expressed by equation (17) 2 The same, according to the Euclidean distance α obtained by formula (18) (s) (i，0) and the variance σ of the measurement surface deformation vector 2 , assuming normal distribution to obtain the likelihood function shown in formula (19).

[0140] [Mathematical formula 19]

[0141]

[0142] Here, β takes a value from 1 to (n+1)*(q+1), corresponding to the value of (i, s), representing an instance of learning data. For example, β = 1 means (i, s) = (0, 0), and β = (n+1)*(q+1) means (i, s) = (n, q).

[0143] In order to normalize the likelihood function shown in equation (19), the value C obtained by adding the likelihood functions is obtained as shown in equation (20).

[0144] [Mathematical formula 20]

[0145]

[0146] Formula (21) shows the result obtained by normalizing the likelihood function obtained by formula (19) using C shown in formula (20).

[0147] [Mathematical formula 21]

[0148]

[0149] The normalized likelihood function shown in equation (21) becomes equal to the likelihood function of the latent variable indicating the presence or absence of the crack 4. The vector similarity calculation unit 43 receives the information from the estimation model generation unit 32. Figure 14 The latent variable vector Γ in the learning data is shown (s) (i，0) , the likelihood function shown in Equation (21) is combined with the corresponding latent variable vector Γ (s) (i，0) The expected value of the latent variable in the likelihood function is obtained by multiplying and adding the values ​​to all crack candidates. The vector similarity calculation unit 43 outputs the value shown in equation (22) to the crack analysis unit 44.

[0150] [Mathematical formula 22]

[0151]

[0152] The crack analysis unit 44 performs threshold processing on the expected value of the latent variable obtained from equation (22) using a predetermined threshold value to determine the estimated position and size of the crack 4 on the crack initiation surface. The obtained result is output from the crack analysis unit 44 to the output processing unit 61 of the analysis result output unit 60.

[0153] In Equation (22), the likelihood function shown in Equation (21) is combined with the corresponding latent variable vector Γ (s) (i，0) Multiply and add all crack candidates to find the expected value of the latent variable in the likelihood function, but it can also be combined with the latent variable vector Γ (s) (i，0) Similarly, the displacement change vector Δ is used, which is a vector consisting of state quantities indicating the state of each node of the crack occurrence surface. (s) (i，0) The expected value is obtained by normalizing the likelihood function shown in equation (21) to be equal to the likelihood function of the displacement change of each node of the candidate face 3. It is also possible to calculate the expected value by normalizing the likelihood function shown in equation (21) to the corresponding displacement change vector Δ (s) (i，0) The expected value of the displacement change in the likelihood function is obtained by multiplying and adding the values for all crack candidates. The obtained expected value is thresholded using a predetermined threshold to obtain the estimated position and size of the crack 4 on the crack occurrence surface.

[0154] Moreover, it can also be combined with the latent variable vector Γ (s)(i，0) Similarly, the load change vector Z is a vector composed of state quantities indicating the state of each node of the crack occurrence surface. (s) (i，0) To find the expected value. The normalized likelihood function shown in formula (21) becomes equal to the likelihood function of the load change of each node of candidate face 3. The likelihood function shown in formula (21) can also be compared with the corresponding load change vector Z (s) (i，0) By multiplying and adding all the crack candidates, an expected value of the load change in the likelihood function is calculated. The obtained expected value is thresholded using a predetermined threshold value to determine the estimated position and size of the crack 4 on the crack initiation surface. The crack estimation device of the present application can be implemented even if the vector quantities used in the previous estimation are treated as a two-dimensional array or image data.

[0155] Information on the estimated location and size of the crack 4 is output from the crack analysis unit 44 to the output processing unit 61 of the analysis result output unit 60. The output processing unit 61 obtains the estimated location and size information of the crack 4 from the crack analysis unit 44 and, from the storage unit 50, obtains information on the load applied to the structure 1, physical property values in the structure 1, information on the size of the crack 4 that renders the structure 1 unusable, and information on the location of the crack 4 that renders the structure 1 unusable. Information on the load applied to the structure 1 may also be obtained from the model generation unit 30 via the crack state analysis unit 40. Here, the physical property value is, for example, the longitudinal elastic modulus. Information on the size and location of the crack 4 that renders the structure 1 unusable is used as a threshold value. The physical property values, information on the size and location of the crack 4 that renders the structure 1 unusable, stored in the storage unit 50, are information determined and stored during the product design stage. The output processing unit 61 calculates the remaining useful life of the structure 1 based on this information. The remaining useful life can also be calculated based on changes in the size and position of the cracks 4 over time. The calculated remaining useful life is displayed on a display device 63. The display device 63 is implemented, for example, by a liquid crystal display. The ability to confirm the remaining useful life of the structure 1 on the display device 63 allows for more specific planning of the operation of the structure 1. For example, the time when the structure 1 should be repaired or updated is known in advance, allowing for planned repair and updating of the structure 1.

[0156] Furthermore, if the estimated position of crack 4 exceeds the threshold value at which the structure 1 becomes unusable, or if the estimated size of crack 4 exceeds the threshold value at which the structure 1 becomes unusable, output processing unit 61 transmits a warning message urging the cessation of use of the structure to alarm device 62 or display device 63, causing alarm device 62 or display device 63 to issue a warning. The warning can be implemented, for example, through sound, text, flashing, or lighting. Alarm device 62 can be implemented using a speaker, a light-emitting device, or the like, while display device 63 can be implemented using a liquid crystal display or the like. For example, if alarm device 62 is a speaker, the warning can be issued by sound, while if it is a light-emitting device, the warning can be issued by flashing or lighting. If the warning is issued on the display of display device 63, the warning can be issued by text. The issuance of a warning by alarm device 62 or display device 63 allows the operator of structure 1 to be promptly notified that the structure 1 should cease use.

[0157] Next, the operation of the crack estimation device 100 will be described using a flowchart. Figure 15 This is a flowchart illustrating the processing performed by the crack estimation device 100. Steps S11 to S23 are performed in the learning phase, and steps S24 to S26 are performed in the inverse analysis phase. Steps S11 to S22 are performed in the shape model generation unit 31 of the estimation unit 20, and step S23 is performed in the estimation model generation unit 32 of the estimation unit 20. Steps S11 to S23 are model generation steps. Step S24 is performed in the measurement unit 10 and the data acquisition unit 41. Step S24 is a data acquisition step. Step S25 is performed in the crack state estimation unit 42 of the estimation unit 20. Step S25 is a crack state estimation step. Step S26 is performed in the analysis result output unit 60 of the estimation unit 20. Step S26 is an analysis result output step.

[0158] In step S11, a determination is made as to whether the conditions of the learning data have been accepted. If the conditions of the learning data have not been accepted, the process of step S11 is repeated. The conditions of the learning data include the estimated starting point of the crack 4 and the shape of the crack 4. If the conditions of the learning data have been accepted in step S11, the process proceeds to step S12.

[0159] In step S12, the starting point of the crack 4 is determined based on the conditions of the received learning data, and the process proceeds to step S13. In step S13, the candidate surface 3 is determined based on the starting point of the crack 4, and the process proceeds to step S14. In step S14, the observation surface 2 measured by the measurement unit 10 is determined, and the process proceeds to step S15. In step S15, a shape model is generated based on the shape of the structure 1. Next, the process proceeds to step S16.

[0160] In step S16, the candidate surface 3 is divided into a plurality of elements 7 in a grid pattern, and the process proceeds to step S17. In step S17, nodes are set at intersections of the grids formed by the division into the plurality of elements 7 in step S16, and the process proceeds to step S18. In step S18, multiple structural analysis modes are determined that change the conditions for the cracks 4 in the candidate surface 3, and the process proceeds to step S19. In step S19, the order of learning the cracks at each node is determined based on the structural analysis modes determined in step S18, and the process proceeds to step S20.

[0161] In step S20, the observation surface 2 is divided into a plurality of elements 8 in a grid pattern, and the process proceeds to step S21. In step S21, nodes are set at points where the grids intersect when the observation surface 2 is divided into the plurality of elements 8 in step S20, and the process proceeds to step S22. In step S22, the order of learning strains at each node of the observation surface 2 is determined in the respective structural analysis modes determined in step S18, and the process proceeds to step S23.

[0162] In step S23 , a learning data creation process is executed. Figure 16 、 Figure 17 as well as Figure 18 Details of the learning data creation process are shown. Next, the process proceeds to step S24.

[0163] In step S24, the measurement unit 10 acquires measurement data and transmits the acquired measurement data to the vector similarity calculation unit 43 of the crack state estimation unit 42 via the data acquisition unit 41. The process then proceeds to step S25. In step S25, the crack state estimation unit 42 of the estimation unit 20 performs the estimation process and transmits the result to the analysis result output unit 60. Figure 19 The details of the estimation process are shown. Next, the process proceeds to step S26. In step S26, the analysis result output unit 60 of the estimation unit 20 performs output processing, and the process ends. Figure 20 as well as Figure 21 Shows the details of the output process.

[0164] Figures 16 to 18 It shows Figure 15 The detailed flowchart of the learning data creation process of step S23 is shown in FIG. Figures 16 to 18 The learning data creation process shown in the figure. Figure 16In step S31, information on the starting point of the crack 4 is obtained from the shape model generator 31, and the process proceeds to step S32. In step S32, information on the shape model is obtained from the shape model generator 31, and the process proceeds to step S33. In step S33, information on the structural analysis pattern is obtained from the shape model generator 31, and the process proceeds to step S34. In step S34, information on the order in which the crack 4 was learned is obtained from the shape model generator 31, and the process proceeds to step S35. In step S35, information on the order in which the strain was learned is obtained from the shape model generator 31, and the process proceeds to step S36.

[0165] In step S36, a structural analysis model is created, and the process proceeds to step S37. In step S37, the crack initiation surface is determined, and the process proceeds to step S38. In step S38, the measurement surface is determined, and the process proceeds to step S39. In step S39, the crack initiation surface is divided into a plurality of grid-like elements 7, and the process proceeds to step S40. In step S40, nodes are set at the points where the grids intersect on the crack initiation surface, and the process proceeds to step S41. In step S41, the measurement surface is divided into a plurality of grid-like elements 8, and the process proceeds to step S42. In step S42, nodes are set at the points where the grids intersect on the measurement surface, and the process proceeds to step S43.

[0166] In step S43, the structural analysis model is given the boundary condition that no cracks 4 have occurred at each node of the crack initiation surface, and the process proceeds to step S44. In step S44, under the boundary condition that no cracks 4 have occurred at each node of the crack initiation surface, the displacement change and load change at each node of the crack initiation surface are calculated, and the process proceeds to step S45. In step S45, under the boundary condition that no cracks 4 have occurred at each node of the crack initiation surface, the strain at each node of the measurement surface is calculated, and the process proceeds to step S46.

[0167] In step S46, the structural analysis model is given a boundary condition that the nodes of the crack initiation surface are set to crack 4, and the process proceeds to step S47. In step S47, under the boundary condition that the nodes of the crack initiation surface are set to crack 4, the displacement change and load change of each node in the crack initiation surface are calculated, and the process proceeds to step S48. In step S48, under the boundary condition that the nodes of the crack initiation surface are set to crack 4, the strain of each node of the measurement surface is calculated, and the process proceeds to step S49. Figure 17 Step S49.

[0168] exist Figure 17In step S49, a displacement change vector resulting from the difference in displacement change at the nodes of the crack initiation surface and a load change vector resulting from the difference in load at the nodes are generated, and the process proceeds to step S50. In step S50, a strain change vector resulting from the difference in strain at the nodes of the measurement surface is generated, and the process proceeds to step S51. In step S51, the load change vector resulting from the displacement change is stored in the storage unit 50, and the process proceeds to step S52. In step S52, the strain change vector is stored in the storage unit 50, and the process proceeds to step S53.

[0169] In step S53, it is determined whether the structural analysis has been performed on all nodes of the crack generation surface. If it is determined that the structural analysis has not been performed on all nodes of the crack generation surface, the process proceeds to step S54. In step S54, the node to be set as crack 4 is changed and the process returns to step S55. Figure 16 On the other hand, in step S53, when it is determined that the structural analysis has been performed on all the nodes of the crack occurrence surface, the process proceeds to step S55.

[0170] In step S55, a crack surface matrix Δ is generated which is composed of the displacement change vectors shown in equation (1). crack_diff , enter step S56. In step S56, create the crack surface load matrix Z composed of the load change vectors shown in formula (3) crack_diff , enter step S57. In step S57, create a measurement surface matrix E composed of the strain change vectors shown in formula (2) measure , enter step S58. In step S58, the matrix Δ representing the crack surface is generated as shown in formula (6) crack_diff and the measurement surface matrix E measure The observation matrix H of the relationship between the crack surface load matrix Z is generated as shown in formula (7) crack_diff and the crack surface matrix Δ crack_diff The stiffness matrix G of the relationship is entered into Figure 18 Step S59.

[0171] In step S59, the crack surface load matrix Z crack_diffThe node on the edge 9 of the crack occurrence surface is extracted to become the load change vector of crack 4, and the process proceeds to step S60. In step S60, a latent variable vector Γ is introduced, which uses "1" or "0" to indicate whether the node on the crack occurrence surface is crack 4 or not, and the process proceeds to step S61. In step S61, the node with the largest load in the load change vector is set to crack 4, and the latent variable of the corresponding node is set to "1", and the process proceeds to step S62. In step S62, the displacement change of the node that is considered to be a crack because the latent variable is set to "1" is set to an unknown number, and the displacement change of the node that is considered to be not a crack because the latent variable is set to "0" is set to "0". As a result, the latent variable vector Γ becomes, for example, as shown in equation (9). Next, the process proceeds to step S63. In step S63, the load of the node that is considered to be a crack because the latent variable is set to "1" is set to "0", and the process proceeds to step S64.

[0172] In step S64, the displacement change vector of the node designated as a crack is calculated based on the load change vector, the displacement change vector, and the stiffness matrix G, for example, as shown in equations (11) to (14), and the process proceeds to step S65. In step S65, the strain change vector is calculated based on the displacement change vector calculated in step S64 and the observation matrix H, for example, as shown in equation (15), and the process proceeds to step S66. In step S66, the load change vector is calculated based on the displacement change vector calculated in step S64 and the stiffness matrix G, for example, as shown in equation (16), and the process proceeds to step S67. In step S67, the latent variable vector, displacement change vector, load change vector, and strain change vector calculated through the processing of steps S59 to S66 are stored in the storage unit 50 as learning data, and the process proceeds to step S68.

[0173] In step S68, the processes shown in steps S59 to S67 are repeated a predetermined number of times, and the learning data is stored.

[0174] In step S69, it is determined whether learning data has been created using all nodes on edge 9 as starting points. If learning data has been created using all nodes on edge 9 as starting points, the learning data creation process ends. If learning data has not been created using all nodes on edge 9 as starting points, the process proceeds to step S70. In step S70, the node set as the starting point on edge 9 is changed, and the process proceeds to step S59.

[0175] Figure 19 It shows Figure 15 The details of the estimation process of step S25 are shown in the flowchart. The crack state estimation unit 42 of the estimation unit 20 executes Figure 19The vector similarity calculation unit 43 of the crack state estimation unit 42 executes steps S80 to S89, and the crack analysis unit 44 of the crack state estimation unit 42 executes step S90.

[0176] In step S80, the learning data information is read from the inference model generation unit 32, and the process proceeds to step S81. In step S81, the measurement surface deformation vector expressed by equation (17) is read from the measurement unit 10 via the data acquisition unit 41, and the process proceeds to step S82. In step S82, the strain change vector is acquired from the learning data read by the inference model generation unit 32, and the process proceeds to step S83. Furthermore, in step S80, the learning data information is read from the inference model generation unit 32, and in step S82, the strain change vector is acquired from the learning data. However, step S80 may be omitted, and the strain change vector in the learning data may be acquired from the inference model generation unit 32 in step S82. In step S83, in order to determine the similarity between the measurement surface deformation vector and the strain change vector acquired in step S82, the Euclidean distance as the L2 norm is determined as expressed in equation (18), and the process proceeds to step S84.

[0177] In step S84, it is determined whether the Euclidean distances to the measurement surface deformation vector have been calculated for all strain change vectors in the learning data. If the Euclidean distances to all strain change vectors have been calculated, the process proceeds to step S86. If the Euclidean distances to all strain change vectors have not been calculated, the process proceeds to step S85, where the strain change vectors are modified and the process proceeds to step S82.

[0178] In step S86, the variance σ of the measurement surface deformation vector is set 2 , and proceed to step S87. In step S87, the likelihood function shown in equation (19) is obtained based on the Euclidean distance obtained in step S83 and the variance set in step S86, and the process proceeds to step S88.

[0179] In step S88, the likelihood function is normalized using equations (20) and (21), and the process proceeds to step S89. In step S89, the normalized likelihood function is set equal to the likelihood function of the latent variable corresponding to the normalized likelihood function, and the expected value vector of the latent variable in the likelihood function is calculated as shown in equation (22). Next, the process proceeds to step S90. In step S90, each value of the expected value vector of the latent variable obtained in step S89 is thresholded and classified into cracks and non-cracks. The location and size of the cracks are thus calculated, and the estimation process ends.

[0180] Figure 20 It shows Figure 15 The flowchart of an example of the output process of step S26 is shown in FIG. Figure 1 The details of the process of displaying the remaining use period of the structure 1 in the analysis result output unit 60 are shown in FIG. Steps S101 to S106 are executed in the output processing unit 61, and step S107 is executed in the display device 63.

[0181] In step S101, information on the location and size of crack 4 is obtained, and the process proceeds to step S102. In step S102, information on the load applied to structure 1 is obtained, and the process proceeds to step S103. In step S103, information on the physical property values of structure 1 is obtained, and the process proceeds to step S104. In step S104, information on the location and size of crack 4 at which structure 1 becomes unusable is obtained as a threshold value, and the process proceeds to step S105. In step S105, the amount of growth of crack 4 on the crack initiation surface is calculated based on the location and size of crack 4, the load applied to structure 1, and the physical property values of structure 1. Next, the process proceeds to step S106. In step S106, the remaining useful life is determined based on the amount of growth of crack 4 and the threshold values for the location and size of crack 4, and the process proceeds to step S107. In step S107, information on the determined remaining useful life is output, and the output process ends.

[0182] Figure 21 It shows Figure 15 The flowchart of an example of the output process of step S26 is shown in FIG. Figure 1 Details of the process of issuing a warning urging the cessation of use of the structure 1 in the analysis result output unit 60 are shown. Steps S111, S113 to S116, S119, and S120 are executed in the output processing unit 61. Steps S117 and S118 are executed in the alarm device 62 or the display device 63. Steps S112 and S121 are executed in the display device 63.

[0183] In step S111, information on the location and size of the crack 4 is obtained, and the process proceeds to step S112. In step S112, information on the location and size of the crack 4 is displayed on the display device 63, and the process proceeds to step S113. In step S113, information on the load applied to the structure 1 is obtained, and the process proceeds to step S114. In step S114, information on the physical property values of the structure 1 is obtained, and the process proceeds to step S115. In step S115, information on the location and size of the crack 4 that renders the structure 1 unusable is obtained as a limit value, and the process proceeds to step S116. In step S116, it is determined whether the location and size of the crack 4 exceeds the limit value. If it is determined that the location and size of the crack 4 exceed the limit value, the process proceeds to step S117. In step S117, a warning urging the cessation of use of the structure is issued on the alarm device 62 or the display device 63, and the output process ends.

[0184] On the other hand, if it is determined in step S116 that the position and size of the crack 4 do not exceed the limit value, the process proceeds to step S118. In step S118, the presence of a crack is reported on the alarm device 62 or the display device 63. Next, the process proceeds to step S119. In step S119, it is determined whether the information on the remaining period of use can be obtained. If it is determined that the information on the remaining period of use cannot be obtained, the output process ends. On the other hand, if it is determined that the information on the remaining period of use can be obtained, the process proceeds to step S120. In step S120, the information on the remaining period of use is obtained, and the process proceeds to step S121. In step S121, the remaining period of use is displayed on the display device 63, and the output process ends.

[0185] In the above description, a flat plate is assumed as the structure 1 to be estimated, and the structure is represented by an orthogonal coordinate system of x-axis, y-axis, and z-axis, but the present invention is not limited to this. Figure 22 The cylindrical component 70 is shown. Figure 22 In the figure, it is represented by a cylindrical coordinate system with an r-axis, a z-axis, and an angle θ. Figure 23 Observed from the z-axis direction Figure 22 The figure is obtained by the cylindrical component 70. Figure 23 As shown, during shrinkage fitting, internal pressure 73 is applied to inner circumferential surface 71 of cylindrical member 70. Consequently, cracks 4 develop within cylindrical member 70, causing the shape of outer circumferential surface 72 to change. Cylindrical member 70 is attached, for example, by shrinkage fitting to a retaining ring of a rotor core protruding from the end of a rotor of a rotating electrical machine.

[0186] As described above, the crack estimation device 100 according to the first embodiment includes: a measuring unit 10, which uses an observation surface 2 on the surface of a structure 1 as a measurement surface and measures the deformation of the measurement surface as a measurement surface deformation vector; a model generating unit 30, which generates a shape model obtained by modeling the shape of the structure 1, uses a candidate surface 3 inside the structure 1 as a crack occurrence surface, and sets the deformation of the measurement surface when a crack 4 occurs on the crack occurrence surface as a measurement surface estimation change vector for multiple types of crack candidates; and a crack state analyzing unit 40, which generates a shape model by modeling the shape of the structure 1, uses a candidate surface 3 inside the structure 1 as a crack occurrence surface, and sets the deformation of the measurement surface when a crack 4 occurs on the crack occurrence surface as a measurement surface estimation change vector based on the crack state of the measuring unit 10. The output of the model generation unit 30 and the output of the model generation unit 30 are used to infer the crack 4. The crack state analysis unit 40 calculates the similarity between the measurement surface deformation vector and the measurement surface inferred change vector, and normalizes the similarity. The vector of the state quantity representing the state of the crack occurrence surface for each crack candidate is multiplied by the normalized similarity and the result is added for all crack candidates. The crack 4 occurring on the crack occurrence surface is inferred, so that the uniqueness, existence and stability of the solution are satisfied, and the crack 4 on the crack occurrence surface inside the structure 1 can be inferred with high precision.

[0187] Implementation method 2.

[0188] Figure 24 This is a diagram showing the structure of a crack estimation device 100a according to the second embodiment. Figure 24 The crack estimation device 100a according to the second embodiment shown in FIG. Figure 1 Compared to the crack estimation device 100 according to the first embodiment, the estimation unit 20 is replaced by the estimation unit 20a, the model generation unit 30 is replaced by the model generation unit 30a, the shape model generation unit 31 is replaced by the shape model generation unit 31a, the estimation model generation unit 32 is replaced by the estimation model generation unit 32a, the crack state analysis unit 40 is replaced by the crack state analysis unit 40a, and the data acquisition unit 41 is replaced by the data acquisition unit 41a. Furthermore, the shape model generation unit 31a includes a load setting unit 33, and the data acquisition unit 41a includes a load indication unit 45. Furthermore, a load application unit 11 is newly provided. The remaining configuration of the crack estimation device 100a according to the second embodiment is the same as that of the crack estimation device 100 according to the first embodiment.

[0189] The load setting unit 33 outputs information regarding the magnitude and position of the load to be applied to the structure 1 to the load indicating unit 45. Based on the information received from the load setting unit 33, the load indicating unit 45 outputs instructions to the load applying unit 11. When the measurement unit 10 performs measurement, the load applying unit 11 applies a load of the magnitude indicated by the load indicating unit 45 to the position of the structure 1 indicated by the load indicating unit 45. This allows the measurement unit 10 to measure changes in the surface of the observation surface 2 while the load is applied to the structure 1.

[0190] Next, the operation of the crack estimation device 100a will be described using a flowchart. The basic process of the processing performed by the crack estimation device 100a is as follows: Figure 15 The process shown is the same. Figure 25 This is a flowchart showing the processes in the learning phase and the inverse analysis phase in the crack estimation device 100a according to Embodiment 2. Steps S131 to S134 are processes executed in the learning phase, and steps S141 to S145 are processes executed in the inverse analysis phase.

[0191] In the learning phase, the shape model generating unit 31a determines the inspection object in step S131. The processing in step S131 is the same as Figure 15 The process of step S11 to step S14 is the same as that of step S132. In step S132, the test load is set in the load setting unit 33, and the process of step S133 is carried out. In step S133, the shape model is generated in the shape model generating unit 31a. The process of step S133 is the same as that of step S133. Figure 15The process of step S15 to step S22 is as follows. Next, the process proceeds to step S134. In step S134, the inference model generating unit 32a generates an inference model, and the learning phase ends. The process of step S134 is the same as Figure 15 Here, in step S134, a prediction model is generated under the condition that the test load set in step S132 is added to the structure 1.

[0192] During the inverse analysis phase, a check load is applied to the structure 1 in step S141. Specifically, the load indicating unit 45 first obtains the check load information set in step S132 from the load setting unit 33. Having obtained the check load information, the load indicating unit 45 outputs an instruction to the load applying unit 11 based on the information received from the load setting unit 33. During measurement by the measuring unit 10, the load applying unit 11 applies the load of the magnitude indicated by the load indicating unit 45 to the position of the structure 1 indicated by the load indicating unit 45. Next, the process proceeds to step S142.

[0193] In step S142, the measurement data obtained by the measuring unit 10 is sent to the vector similarity calculation unit 43 of the crack state estimation unit 42 via the data acquisition unit 41a. At this time, the measuring unit 10 measures the deformation of the surface of the observation surface in a state where the load is applied to the structure 1 by the load application unit 11. The processing of step S142 is the same as Figure 15 The process of step S24 is as follows. Next, the process proceeds to step S143. In step S143, the vector similarity calculation unit 43 reads the information of the learning data from the inference model generation unit 32a. The process of step S143 is as follows. Figure 19 The process of step S80 is as follows. Next, the process proceeds to step S144. In step S144, the vector similarity calculation unit 43 of the crack state estimation unit 42 and the crack analysis unit 44 estimate the crack state. The process of step S144 is the same as Figure 19 The processing of steps S81 to S90 is performed accordingly. At this time, the vector similarity calculation unit 43 and the crack analysis unit 44 use the learning data produced under the condition that the test load is applied to the structure 1 and the data obtained from the measurement unit 10 under the condition that the load is applied to the structure 1 to obtain the position and size of the crack 4. Next, the process proceeds to step S145. In step S145, the output processing is performed in the analysis result output unit 60. The processing of step S145 is the same as Figure 15 The above process ends the inverse analysis stage.

[0194] As described above, the model generation unit 30a sets the deformation of the measuring surface when a load is applied to the structure 1 as the measuring surface estimated change vector, and the measuring unit 10 measures the deformation of the measuring surface when a load is applied to the structure 1 as the measuring surface deformation vector, so that the structure 1 that is not pre-loaded can also be inspected, and more types of structures 1 can be inspected.

[0195] Implementation method 3.

[0196] Figure 26 This is a diagram showing the structure of a crack estimation device 100b according to Embodiment 3. Figure 26 The crack estimation device 100b according to the third embodiment shown in FIG. Figure 1 Compared to the crack estimation device 100 according to the first embodiment, the measurement unit 10 is replaced by the measurement unit 10b, the estimation unit 20 is replaced by the estimation unit 20b, the model generation unit 30 is replaced by the model generation unit 30b, the estimated model generation unit 32 is replaced by the estimated model generation unit 32b, the crack state analysis unit 40 is replaced by the crack state analysis unit 40b, the data acquisition unit 41 is replaced by the data acquisition unit 41b, the crack state estimation unit 42 is replaced by the crack state estimation unit 42b, the vector similarity calculation unit 43 is replaced by the vector similarity calculation unit 43b, and the crack analysis unit 44 is replaced by the crack analysis unit 44b. The remaining configuration of the crack estimation device 100b according to the third embodiment is the same as that of the crack estimation device 100 according to the first embodiment. While the crack estimation device 100 according to the first embodiment uses strain change as the deformation of the nodes on the measurement surface, the crack estimation device 100b according to the third embodiment uses at least one of strain change, displacement change, and angle change as the deformation of the nodes on the measurement surface.

[0197] When the strain change is used as the deformation of the node on the measurement surface, the same operation as that of the crack estimation device 100 according to the first embodiment is performed.

[0198] Next, the case where displacement changes are used as deformation of nodes on the measurement surface will be described. When displacement changes are used as deformation of nodes on the measurement surface, the estimated model generation unit 32b of the model generation unit 30b creates displacement change vectors in the structural analysis model that are caused by differences in displacement of nodes on the measurement surface. Figure 27 This is a diagram showing the third embodiment. Figure 5 At each position of the crack 4 of the candidate surface 3, Figure 7 The displacement change vector of each node of the observation surface 2 is shown in FIG. Figure 27The displacement data of each node included in the column vector Dis(-,-) are arranged in the order of moving the crack 4 assumed at each node. d(i, j) is the displacement change of the node at the position (i, j) in the observation surface 2. 0，0 (k, l) is the displacement data of the node at the position (k, l) of the observation surface 2 when a crack 4 occurs at the node at the position (0, 0) of the candidate surface 3, and Dis(0, 0) is the displacement change vector when a crack 4 occurs at the node at the position (0, 0) of the candidate surface 3.

[0199] The following formula (23) shows that Figure 27 The measurement surface matrix Dis composed of multiple displacement change vectors measure When the displacement change is used as the deformation of the nodes in the measurement surface, Dis is used as the measurement surface matrix. measure As Figure 27 The displacement change vectors Dis(0, 0) to Dis(n, m) shown are column vectors. Arranging these column vectors in the order of moving the assumed crack 4 at each node yields the result Dis shown in equation (23). measure .

[0200] [Mathematical formula 23]

[0201]

[0202] When displacement changes are used as deformation of nodes on the measurement surface, the measurement unit 10b is equipped with a displacement sensor to measure the displacement of each node on the observation surface 2. Examples of displacement sensors include laser displacement sensors, eddy current loss displacement sensors, electrostatic capacitance displacement sensors, contact displacement sensors, wire displacement sensors, and laser micrometers. The measurement unit 10b measures the displacement changes on the surface of the observation surface 2 and outputs them as a measurement surface deformation vector.

[0203] In the crack state analysis unit 40b, instead of the measurement surface matrix E measure And using the measurement surface matrix Dis measure , as the measurement surface deformation vector of the displacement change obtained from the measurement unit 10b and the measurement surface matrix Dis measure The Euclidean distance is calculated based on the similarity of , and the position and size of the crack 4 in the crack occurrence surface are estimated.

[0204] Next, the case where angle changes are used as deformations of nodes on the measurement surface will be described. When angle changes are used as deformations of nodes on the measurement surface, the estimated model generation unit 32b of the model generation unit 30b creates angle change vectors in the structural analysis model that are caused by differences in the angles of the nodes on the measurement surface. Figure 28 This is a diagram showing the third embodiment. Figure 5 At each position of the crack 4 of the candidate surface 3, Figure 7 The angle change vector of the angle change of each node of the observation surface 2. Figure 28 The angle data of each node included in the column vector of A(-,-) is arranged in the order of moving the crack 4 assumed at each node. a(i, j) is the angle change of the node at the position (i, j) in the observation plane 2. 0，0 (k, l) is the angle data of the node at the position (k, l) of the observation surface 2 when a crack 4 occurs at the node at the position (0, 0) of the candidate surface 3, and A(0, 0) is the angle change vector when a crack 4 occurs at the node at the position (0, 0) of the candidate surface 3.

[0205] The following formula (24) shows that Figure 28 The measurement surface matrix A composed of multiple angle change vectors measure When the angle change is used as the deformation of the nodes in the measurement surface, A is used as the measurement surface matrix. measure As Figure 28 The angle change vectors A(0, 0) to A(n, m) shown in the figure are column vectors. Arranging these column vectors in the order of moving the crack 4 assumed at each node yields the result of A shown in equation (24). measure .

[0206] [Mathematical formula 24]

[0207]

[0208] When using the angular change as the deformation of the node in the measurement surface, the measurement unit 10 b includes a tilt sensor for measuring the angle of each node of the observation surface 2 .

[0209] In the crack state analysis unit 40b, instead of the measurement surface matrix E measure And using the measurement surface matrix A measure , as the measurement surface deformation vector of the angle change obtained from the measurement unit 10b and the measurement surface matrix A measure The Euclidean distance is calculated based on the similarity of , and the position and size of the crack 4 in the crack occurrence surface are estimated.

[0210] When any change among strain change, displacement change and angle change is selected and used as the deformation of the node in the measuring surface, all change vectors including the strain change vector, the displacement change vector representing the displacement change amount of each node of the measuring surface, and the angle change vector representing the angle change amount of each node of the measuring surface are saved as learning data, and the measuring unit 10b measures at least one of the strain, displacement and angle of each node of the observation surface 2.

[0211] When using displacement or angle changes rather than strain changes as the deformation of the measurement surface, changes in the observation surface 2 of the structure 1 can be measured with higher accuracy and in a shorter time than with strain measurement. Furthermore, when any of strain changes, displacement changes, and angle changes is selected and used as the deformation of the nodes on the measurement surface, various structures 1 can be accommodated.

[0212] Implementation method 4.

[0213] Figure 29 This is a diagram showing the structure of a crack estimation device 100c according to the fourth embodiment. Figure 29 The crack estimation device 100c according to the fourth embodiment shown in FIG. Figure 1 Compared with the crack estimation device 100 according to the first embodiment, the model generation unit 30 is replaced by the model generation unit 30c, and the shape model generation unit 31 is replaced by the shape model generation unit 31c. The other structures of the crack estimation device 100c according to the fourth embodiment are the same as those of the crack estimation device 100 according to the first embodiment.

[0214] Figure 30 FIG. 3 is a diagram showing the appearance of the candidate surface 3 of the structure 1 in the fourth embodiment. Figure 30 As shown, when the stress is obtained by measurement or structural analysis, the maximum stress σ max In the case of a point at which the stress is maximum, the shape model generation unit 31c of the model generation unit 30c determines the point at which the stress is maximum as the location where the crack 4 occurs. Furthermore, the shape model generation unit 31c of the model generation unit 30c determines a surface that is perpendicular to the direction of the stress in the determined location where the crack 4 occurs and that includes the determined location where the crack 4 occurs as a candidate surface 3.

[0215] Figure 31 It shows the decision Figure 30 Flowchart of the processing of the candidate face 3. The shape model generating unit 31c executes Figure 31The processing shown. In step S151, it is determined whether the distribution of stress occurring in the structure 1 has been obtained. If the distribution of stress occurring in the structure 1 has not been obtained, the processing of step S151 is repeated. On the other hand, if the distribution of stress occurring in the structure 1 has been obtained, the process proceeds to step S152. In step S152, it is determined whether there is a point where the stress is maximum. If there is no point where the stress is maximum, the process of step S152 is repeated. On the other hand, if there is a point where the stress is maximum, the process proceeds to step S153. In step S153, the point where the stress is maximum is determined as the location where crack 4 occurs, and the process proceeds to step S154. In step S154, the surface that is perpendicular to the stress in the determined location where crack 4 occurs and includes the location where crack 4 occurs is determined as candidate surface 3.

[0216] In addition, the maximum stress σ occurs max The location of the maximum stress σ is likely to occur in the place where the boundary conditions are set, so it can also be used to calculate the maximum stress σ max Re-evaluate the boundary conditions after the location is reached.

[0217] By determining the candidate surface 3 using the method described above, a surface where cracks 4 are likely to occur inside the structure 1 can be determined as the candidate surface 3 , and the estimation accuracy of the cracks 4 can be further improved.

[0218] Figure 32 This is a schematic diagram showing an example of hardware of the crack estimation device according to the first, third, and fourth embodiments. Figure 33 This is a schematic diagram showing an example of hardware of the crack estimation device according to Embodiment 2. The storage unit 50 is implemented by a memory 202. The memory 202 is, for example, a nonvolatile or volatile semiconductor memory such as ROM, RAM, flash memory, EPROM, or EEPROM, or a magnetic disk, floppy disk, optical disk, high-density disk, minidisc, or DVD.

[0219] The model generation units 30, 30a, 30b, 30c, the crack state analysis units 40, 40a, 40b, and the output processing unit 61 are implemented by a processor 201, such as a CPU or system LSI, that executes a program stored in the memory 202. Alternatively, multiple processing circuits may collaboratively perform the above functions. Furthermore, the above functions may be implemented by dedicated hardware. When implementing the above functions by dedicated hardware, the dedicated hardware may be, for example, a single circuit, a complex circuit, a programmed processor, an ASIC, an FPGA, or a combination thereof. The above functions may also be implemented by a combination of dedicated hardware and software, or a combination of dedicated hardware and firmware. For example, the model generation units 30, 30a, 30b, 30c may be implemented by dedicated hardware, while the crack state analysis units 40, 40a, 40b, and the output processing unit 61 may be implemented by a processor 201, such as a CPU or system LSI, that executes a program stored in the memory 202.

[0220] Various exemplary embodiments are described in this application, but various features, forms, and functions described in one or more embodiments are not limited to application in specific embodiments and can be applied to the embodiments alone or in various combinations.

[0221] Therefore, numerous variations not shown in the examples are conceivable within the scope of the technology disclosed in this application. For example, these include modifying at least one component, adding at least one component, omitting at least one component, and extracting at least one component and combining it with components from other embodiments.

[0222] Explanation of symbols

[0223] 1: Structure; 2: Observation surface; 3: Candidate surface; 4: Crack; 5: Tensile load; 6: Bending moment; 7, 8: Element; 9: Edge; 10, 10b: Measurement unit; 11: Load application unit; 20, 20a, 20b: Estimation unit; 30, 30a, 30b, 30c: Model generation unit; 31, 31a, 31c: Shape model generation unit; 32, 32a, 32b: Estimation model generation unit; 33: Load setting unit; 40, 40a, 40b: Crack state analysis unit; 41, 41a, 41b: Data acquisition unit; 42, 42b: Crack state estimation unit; 43, 43b: Vector similarity calculation unit; 44, 44b: Crack analysis unit; 45: Load indication unit; 50: Storage unit; 60: Analysis result output unit; 61: Output processing unit; 62: Alarm device; 63: Display device; 70: Cylindrical component; 71: Inner circumferential surface; 72: Outer circumferential surface; 73: Internal pressure; 100, 100a, 100b, 100c: Crack estimation device; 201: Processor; 202: Memory.

Claims

1. A crack estimation device, characterized in that: have: a measuring unit that uses an observation surface on the surface of the structure as a measuring surface and measures a deformation of the measuring surface as a measuring surface deformation vector; a model generating unit that generates a shape model by modeling the shape of the structure, uses a candidate surface located inside the structure as a crack initiation surface, and sets, for a plurality of types of crack candidates, a deformation of the measurement surface when a crack occurs on the crack initiation surface as a measurement surface estimated change vector; as well as a crack state analyzing unit that estimates the crack based on the output of the measuring unit and the output of the model generating unit, The crack state analysis unit calculates the similarity between the measurement surface deformation vector and the measurement surface estimated change vector, normalizes the similarity, and estimates the crack occurring on the crack occurrence surface based on the result obtained by multiplying the vector of the state quantity representing the state of the crack occurrence surface by the standardized similarity for each crack candidate and adding the result for all the crack candidates.

2. The crack estimation device according to claim 1, wherein The model generation unit sets a plurality of starting points in advance, and sets the crack candidates by performing crack growth from each of the starting points according to predetermined conditions.

3. The crack estimation device according to claim 1, wherein The similarity is the Euclidean distance.

4. The crack estimation device according to claim 1, wherein The shape model of the structure is a cylindrical coordinate system.

5. The crack estimation device according to any one of claims 1 to 4, characterized in that: The model generation unit sets the deformation of the measurement surface in a state where a load is applied to the structure as the measurement surface estimated change vector. The measuring unit measures the deformation of the measurement surface as the measurement surface deformation vector in a state where the load is applied to the structure.

6. The crack estimation device according to any one of claims 1 to 4, characterized in that: The state quantity indicating the state of the crack initiation surface is any of a latent variable indicating the presence or absence of the crack in the crack candidate in the crack initiation surface, a displacement change in the crack initiation surface, and a load change in the crack initiation surface.

7. The crack estimation device according to any one of claims 1 to 4, characterized in that: The model generation unit sets the nodes of the crack initiation surface as the boundary conditions of the crack and uses the stiffness matrix to obtain the load change vector, and sets the location where the load change vector becomes the largest as the next crack, and develops the crack to be the new crack candidate. The model generation unit repeatedly performs the following processing: setting the boundary conditions anew while a crack is developing and using the stiffness matrix to determine the load change vector, setting the location where the load change vector is maximum as the next crack, developing the crack, and setting it as the new crack candidate; The model generation unit prepares, as learning data, a latent variable vector indicating whether the crack occurs on the crack occurrence surface, a displacement change vector indicating the displacement change of the crack occurrence surface, a load change vector indicating the load change of the crack occurrence surface, and the measurement surface estimated change vector indicating the deformation of the measurement surface among all the crack candidates.

8. The crack estimation device according to any one of claims 1 to 4, characterized in that: The measuring portion measures at least one of a strain change, a displacement change, and an angle change as the deformation of the measuring surface. When the measuring unit measures the strain change, the strain change vector of the observation surface is used as the estimated change vector of the measurement surface. When the measuring unit measures the displacement change, the displacement change vector of the observation plane is used as the measurement plane estimated change vector. When the measuring unit measures the angle change, the angle change vector of the observation plane is used as the measurement plane estimated change vector.

9. A crack estimation method, characterized in that: include: a data acquisition step of taking an observation surface on the surface of the structure as a measurement surface and measuring a deformation of the measurement surface as a measurement surface deformation vector; a model generation step of generating a shape model by modeling the shape of the structure, taking a candidate surface located inside the structure as a crack initiation surface, and setting, for a plurality of types of crack candidates, a deformation of the measurement surface when a crack occurs on the crack initiation surface as a measurement surface estimated change vector; as well as a crack state estimating step of estimating the crack based on the value obtained in the data acquiring step and the value obtained in the model generating step; In the crack state estimation step, the similarity between the measurement surface deformation vector and the measurement surface estimation change vector is calculated, and the similarity is standardized. The crack occurring on the crack occurrence surface is estimated based on the result obtained by multiplying the vector of the state quantity representing the state of the crack occurrence surface with the standardized similarity for each crack candidate and adding the result for all the crack candidates.

10. The crack estimation method according to claim 9, wherein: In the model generation step, a plurality of starting points are set in advance, and cracks are developed from each of the starting points according to predetermined conditions to set the crack candidates.

11. The crack estimation method according to claim 9, wherein: In the model generation step, The nodes of the crack occurrence surface are set as the boundary conditions of the crack and the load change vector is calculated using the stiffness matrix. The location where the load change vector becomes the largest is set as the next crack and the crack is developed and set as the new crack candidate. The following process is repeatedly performed: the boundary conditions are newly set while the crack is developing, the load change vector is obtained using the stiffness matrix, a location where the load change vector becomes maximum is set as the next crack, the crack is developed, and the location is set as the new crack candidate. A potential variable vector indicating whether the crack occurs on the crack occurrence surface, a displacement change vector indicating the displacement change of the crack occurrence surface, a load change vector indicating the load change of the crack occurrence surface, and a measurement surface estimated change vector indicating the deformation of the measurement surface are prepared as learning data for all the crack candidates.

12. The crack estimation method according to any one of claims 9 to 11, characterized in that: Also includes: The analysis result output step determines the remaining service life of the structure based on the crack growth amount calculated based on the inferred crack information, the load applied to the structure and the physical property values in the structure, and the limit values of the crack position and size.

13. The crack estimation method according to any one of claims 9 to 11, characterized in that: Also includes: The analysis result output step issues a warning urging a stop of use of the structure based on the estimated information on the crack and the limit values of the position and size of the crack.

14. The crack estimation method according to any one of claims 9 to 11, characterized in that: In the model generating step, a point where stress in the structure becomes maximum is identified as a location where the crack occurs.

Images