Crack estimation device, crack estimation method, crack inspection method, and failure diagnosis method

Abstract

Possessing: data decision part, decide the shape model of the object structure to be checked and the crack occurrence surface and the observation surface in the shape model; estimation data calculation part, output estimation model, which is used to estimate the state of the crack occurrence surface from the state of the observation surface according to the matrix relating the state of the crack occurrence surface and the state of the observation surface obtained by numerically analyzing the structure analysis model made from the shape model; and crack estimation part, estimating the state of the crack in the crack occurrence surface according to the measured value of the object structure actually measured in the observation surface and the estimation model.

Description

Technical Field

[0001] This application relates to crack detection devices, crack detection methods, crack inspection methods, and fault diagnosis methods. Background Technology

[0002] Generally, cracks inside equipment and other structures cannot be detected by visual inspection and may go unnoticed during routine checks. These cracks can then widen, affecting the structure's lifespan and potentially leading to equipment failure. Therefore, detecting internal cracks is a crucial aspect of equipment fault diagnosis.

[0003] Methods for non-destructively inspecting cracks inside a structure include surface shape measurement, ultrasonic testing, and X-ray inspection (see, for example, Patent Document 1).

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent Application Publication No. 2012-159477 Summary of the Invention

[0007] In non-destructive testing methods such as ultrasonic testing or X-ray inspection, the equipment is difficult to miniaturize. While the shape measurement of structural surfaces is easy to miniaturize, it is difficult to measure internal cracks.

[0008] This application was completed to solve such a problem, and its purpose is to provide a device for predicting internal cracks using a small device.

[0009] The crack prediction device disclosed in this application includes:

[0010] The data decision-making department determines the shape model of the object structure to be inspected, the crack initiation surface in the shape model, and the observation surface;

[0011] The inference data calculation unit outputs an inference model, which is used to infer the state of the crack initiation surface from the state of the observed surface based on a matrix obtained by numerical analysis of the structural analytical model made from the shape model, which correlates the state of the crack initiation surface with the state of the observed surface; and

[0012] The crack prediction section predicts the state of cracks in the crack occurrence surface based on the measured values ​​of the object structure actually measured in the observation surface and the prediction model.

[0013] According to the crack estimation device disclosed in this application, it is possible to estimate the internal cracks of an object structure using a small device based on information obtained by measuring the shape of the structure surface. Attached Figure Description

[0014] Figure 1 This is an overall structural diagram of the fault diagnosis device with a crack prediction device according to Embodiment 1.

[0015] Figure 2 This is a flowchart illustrating the overall process for estimating the size of a crack according to Embodiment 1.

[0016] Figure 3 This is a hardware structure diagram of the crack prediction device involved in Implementation 1.

[0017] Figure 4 This is a functional structure diagram of the crack prediction device involved in Implementation Method 1.

[0018] Figure 5 It is a three-dimensional view showing the structure of an object under tensile load.

[0019] Figure 6 It is a three-dimensional diagram showing the structure of an object under the applied bending moment.

[0020] Figure 7A This is a diagram illustrating the division of the crack-inducing surface of the object structure involved in Embodiment 1.

[0021] Figure 7B This is another diagram illustrating the segmentation of the crack-inducing surface of the object structure involved in Embodiment 1.

[0022] Figure 7C This is another diagram illustrating the segmentation of the crack-inducing surface of the object structure involved in Embodiment 1.

[0023] Figure 8 This is a diagram illustrating the division of the observation plane of the object structure involved in Implementation Method 1.

[0024] Figure 9 It is used for execution Figure 2 The flowchart for step S02.

[0025] Figure 10 This is a diagram illustrating the memory structure for storing information about the displacement changes of the crack-causing surface according to Embodiment 1.

[0026] Figure 11 This is a diagram illustrating the memory structure for storing information about strain changes on the observation surface according to Embodiment 1.

[0027] Figure 12 It is used for execution Figure 2 The flowchart for step S04.

[0028] Figure 13 It is a 3D model constructed from objects of other shapes.

[0029] Figure 14 It is a diagram illustrating the construction of objects of other shapes.

[0030] Figure 15 This is a diagram illustrating the memory structure for storing information about the displacement changes of the observation surface according to Embodiment 2.

[0031] Figure 16 This is a diagram illustrating the memory structure for storing information about the angle change of the observation plane according to Embodiment 2.

[0032] Figure 17 This is a diagram illustrating the memory structure for storing information about load changes on the crack initiation surface according to Embodiment 3.

[0033] Figure 18 This is a flowchart illustrating the actions involved in Implementation Method 5.

[0034] Figure 19 This is a flowchart illustrating the actions involved in Implementation Method 6.

[0035] (Symbol Explanation)

[0036] 01: Object structure; 02: Crack occurrence surface; 03: Crack; 04: Observation surface; 100: Turbine generator; 200: Rotor; 300: Measuring device; 400: Crack prediction device; 401: Processor; 402: Storage device; 410: Alarm device; 420: Input device; 430: Display device; 4021: Volatile storage device; 4022: Auxiliary storage device. Detailed Implementation

[0037] The present application will now be described with reference to the accompanying drawings, in which the same or equivalent parts or portions will be indicated by the same symbols.

[0038] Implementation method 1.

[0039] Figure 1 This is an overall structural diagram of the fault diagnosis device 500, which includes a crack prediction device 400, according to this embodiment. Based on the learning data from the crack prediction device 400 and the measurement data from the measurement device 300, the size (location and size) of cracks inside the structure (object structure 01) of the rotor 200 of the rotating electric motor constituting the turbine generator 100 is predicted. When the size of the crack is greater than or equal to the size that could cause a fault, a warning is issued by the alarm device 410 using sound or a display. The crack prediction device 400 is connected to the input device 420 and the display device 430. Furthermore, the object structure 01 is measured by the measurement device 300.

[0040] Figure 2This is a flowchart illustrating the overall process of a crack estimation device 400 estimating the size of a crack inside an object structure 01, with steps S01, S02, and S04 performed by the crack estimation device 400. Figure 3 yes Figure 1 Hardware structure diagram of the crack prediction device 400. Figure 4 yes Figure 1 Functional structure diagram of the crack prediction device. Figure 5 This is a perspective view showing the state under which a tensile load is applied to object structure 01. Figure 6 This is a perspective view showing the state of applying a bending moment to object structure 01. Figures 7A to 7C This is a diagram illustrating the segmentation of the crack initiation surface 02 in the object's structure. Figure 8 This is a diagram illustrating the division of the observation surface 04 of the object structure 01.

[0041] Figure 2 The flowchart shown illustrates an overview of the data processing performed in the crack prediction device 400. Figure 3 An example of the hardware of the microcomputer within the crack prediction device 400 that causes the flowchart to operate is shown. It consists of a processor 401 and a storage device 402. The storage device 402 includes a volatile storage device 4021 capable of one-time storage, such as random access memory, and a non-volatile auxiliary storage device 4022, such as read-only memory or flash memory. Alternatively, a hard disk may be provided as an auxiliary storage device instead of flash memory. The processor 401 executes the program input from the storage device 402. Figure 1 , Figure 9 The flowchart shown is illustrated below. In this case, the program is input from the auxiliary storage device 4022 to the processor 401 via the volatile storage device 4021. Furthermore, the processor 401 can either output data such as calculation results to the volatile storage device 4021 of the storage device 402, or save data to the auxiliary storage device 4022 via the volatile storage device 4021. In the following description, the volatile storage device 4021 will be described as a primary storage unit.

[0042] Furthermore, in the process executed by processor 401 Figure 2 When the flowchart shown is displayed, it is possible to achieve the following: Figure 4 As shown, based on the execution content, it is divided into multiple functional blocks such as the data determination unit 20, the inferred data calculation unit 30, or the crack inference unit 40. The following sections, along with the steps in the flowchart, will also explain each function.

[0043] exist Figure 2In this context, the learning phase F01 refers to the phase of creating and learning learning data for inference. The learning phase F01 includes the step of determining the conditions of the learning data (step S01) and the step of creating a model to be used in inference based on the learning data (step S02).

[0044] In addition, Figure 2 In the process, the reverse analysis stage F02 is the stage where the shape and location of the crack are inferred from the measurement data obtained by the measuring device 300 based on the learning data produced in the learning stage F01, and the data is output.

[0045] [Explanation of Learning Phase F01]

[0046] The steps to determine the learning data (in) Figure 4 (The function of the data decision-making department 20)

[0047] In step S01, which determines the learning data, as follows: Figure 5 As shown, for the object structure 01 to be speculated, the location where the imagined crack 03 will occur is speculated, and the crack occurrence surface 02 is determined as the inspection area.

[0048] The crack initiation surface 02 can also be determined as shown in (1) to (3) below. However, it is not limited to this.

[0049] (1) The distribution of stress in the object structure is determined in advance by measurement or structural analysis.

[0050] (2) Select an evaluation stress that is suitable for determining the location of crack occurrence based on the material and stress distribution, and take the point where the stress is the largest as the location of crack occurrence.

[0051] (3) Then, determine the surface that is perpendicular to the direction of the maximum principal stress at the location of the crack and opposite to the location of the crack in the object structure in a continuous manner.

[0052] The observable surface near the crack initiation surface 02 is designated as the strain measurement observation surface 04. At this time, in Figure 5 During inspection, a tensile load of 05 is applied to object structure 01. Additionally, as... Figure 6 As shown, during inspection, a bending moment 06 is sometimes applied to the object structure 01. The entire object structure 01 or a portion thereof is used as the inspection area, and a shape model of the inspection area is created. When modeling the entire object structure 01, the constraints on deformation applied to the object structure 01 other than the load, as well as temperature distribution, are taken as boundary conditions for structural analysis. When modeling a portion of the object structure 01, the displacement of the cut surface or the distribution of the load are reflected as boundary conditions in the structural analysis.

[0053] Next, as Figure 7A As shown, the crack initiation surface 02 within the shape model of object structure 01 is divided into grid shapes 08 that become unit faces (features). Figure 7A In this process, the X-direction of the dividing surface is divided into n parts, and the Y-direction is divided into m parts. The position (i, j) represents the point where the divided grids intersect. Therefore, the position (i, j) is represented by numbers from (0, 0) to (n, m). One of the points where the grids intersect is taken as a crack, and the crack moves sequentially at all the points where the grids intersect. In step S01, the order of this movement is determined.

[0054] For each construction analysis that alters the boundary conditions of the crack at crack initiation surface 02, the displacements are calculated from the lattice points of lattice shape 08 at crack initiation surface 02, in the determined order. Furthermore, the stored displacement components are relative. Figure 5 or Figure 6 The load or moment shown is the component with the largest displacement of the crack portion at the crack initiation surface 02.

[0055] As a method for moving cracks, when illustrating the case of the finite element method, such as Figure 7B As shown, there is a method to remove node 602 (connection) that exists between element 600 and element 601. That is, assume that node 602 is not shared, and divide the element into two nodes: the node of element 600 and the node of element 601, with no displacement or force transmission between these two nodes.

[0056] Alternatively, the displacement of the crack initiation surface can be altered to have the same shape or boundary conditions as the conditions that caused the crack. For example, such as Figure 7C As shown, in the analytical model of half the finite element method with the crack initiation surface 02 as the plane of symmetry, the boundary conditions for node 603, which will become the crack, are set to no load and no displacement constraints. For nodes other than those on the crack initiation surface 02, no load and zero displacement in the Z direction are provided as boundary conditions.

[0057] Next, as Figure 8 As shown, the range for obtaining deformation information of the surface used to infer the crack is determined. This range is set as observation surface 04. In this embodiment, strain is used as the deformation of the surface. Observation surface 04 is also divided into a grid shape 09, similar to the crack initiation surface 02. Figure 8In the process, the X-direction of the dividing surface is divided into n parts, and the Z-direction is divided into p parts. The point where the dividing grid shape 09 intersects is represented by the position (k, l). Therefore, the position (k, l) is represented by numbers from (0, 0) to (n, p). For each structural analysis that changes the boundary conditions of the crack on the crack initiation surface 02, the strain calculated at the grid points of the observation surface 04 is stored in the determined order. In step S01, the storage order is determined. In addition, the stored strain components are relative to... Figure 5 or Figure 6 The component of the load or moment strain shown is the component with the largest strain. In addition, for strain when a load or moment is applied to two axes (e.g., the Z-axis and the Y-axis), the principal strain, Tresca's equivalent strain, and Mises' equivalent strain, which are parameters for evaluating the strain of multiaxial loads or multiaxial stress fields caused by structural effects, can also be used.

[0058] The observation surface 04 is not limited to Figure 8 The planar or grid-shaped intersecting point groups (grid point groups) shown can also be discrete points or point groups within a portion of the observation surface.

[0059] The steps to create a model for inference based on the learning data (in...) Figure 4 (The function of the inference data calculation unit 30)

[0060] Next, step S02 will be described in detail. In step S02, based on the learning data determined in step S01, a structural analytical model is created for use in predicting the shape and location of the crack.

[0061] That is, the assumed crack shape and position are changed in the order determined in step S01, and the structural analytical model made from the shape model is numerically analyzed. The displacement of the crack initiation surface 02 and the deformation of the observation surface 04 are stored as vectors in the storage device 402. Then, the analytical results of all the assumed crack shapes stored in the storage device 402 are represented as a matrix.

[0062] Furthermore, by using the linear relationship between the displacement of the crack initiation surface 02 and the deformation of the observation surface 04, the inverse matrix of the positive coefficient matrix of the crack initiation surface matrix and the observation surface matrix is ​​obtained.

[0063] Figure 9 The detailed process of step S02 is shown. Additionally, Figure 10 as well as Figure 11 The memory structure for creating a matrix based on the vector stored in the storage device 402 in step S02 is shown. These memory structures are stored in the storage device 402.

[0064] <Function of the numerical analysis unit 31>

[0065] (1) In Figure 9 In step S0201, the shape model of the crack occurrence location (crack occurrence surface 02) and the strain measurement surface (observation surface 04) determined in step S01, the assumed crack shape and location for learning, and the learning sequence are input into processor 401. Processor 401 executes the following process.

[0066] (2) In step S0202, based on the shape model, a structural analytical model is created through numerical calculations such as the finite element method.

[0067] <Function of the numerical analysis control unit 32>

[0068] (3) In step S0203, the crack initiation surface 02 and observation surface 04 of the analytical model are divided into multiple grid shapes 08 as described above, providing boundary conditions for no cracks. Through analytical construction, the displacement of the crack initiation surface 02 and the deformation of the observation surface 04 are calculated.

[0069] (4) In step S0204, the crack occurrence surface 02 of the analytical model is divided into multiple grid shapes 08 as described above, and each node contained in the grid shape 08 is provided as the boundary condition of the crack. The deformation of the observation surface 04 is calculated by constructing an analytical model.

[0070] (5) In step S0205, for each condition of the nodes that are cracks, the differences in displacement of all nodes on the crack initiation surface 02 before and after crack initiation are arranged in the learning order to create a vector Δ(0, 0) of the displacement change of the crack initiation surface 02. Additionally, the differences in deformation of all nodes on the observation surface 04 before and after crack initiation are arranged in the learning order to create a deformation vector E(0, 0) of the strain change of the observation surface 04 (see below). Figure 10 ).

[0071] (6) In step S0206, the vector is saved in storage device 402.

[0072] (7) In step S0207, it is determined whether a construction analysis was performed that treats all nodes of the crack initiation surface 02 as cracks. Since all nodes of the crack initiation surface 02 are treated as cracks, if the construction analysis was not performed that treats all nodes as cracks, in step S0208, the nodes that are treated as cracks are changed, and the process returns to step S0204 to perform construction analysis. In step S0206, the vector is saved in the storage device 402.

[0073] (8) After analyzing the structure of all nodes of the crack initiation surface 02 as cracks, in step S0209, the vector Δ(0,0) of the displacement change of the crack initiation surface 02 stored in the storage device 402 is arranged in the order of learning, and a crack initiation surface matrix Δ is created as the matrix of the displacement change of the crack initiation surface 02. crack_diff .

[0074] (8-1) Specifically, such as Figure 10 As shown in the memory structure, the column vector of the displacement change vector Δ(0,0) of the crack initiation surface 02 is arranged in the order determined in step S01, representing the displacement data of the nodes of the crack initiation surface 02. In the column vector of Δ(0,0), δ(i,j) represents the displacement of the node at position (i,j) of the crack initiation surface 02.

[0075] (8-2) Furthermore, regarding the information about the location where the crack occurs, as the location (i, j) of the crack occurrence surface 02, a column vector Δ(i, j) is created, and the elements within this column vector are represented by δi_j(i, j). Δ(i, j) represents the displacement of the nodes at the location (i, j) of the crack occurrence surface 02 as a result of the crack construction analysis. This column vector is arranged into rows according to the order of the crack occurrence locations determined in step S01, and a crack occurrence surface matrix Δ representing the displacement changes of the crack occurrence surface 02 is created. crack_diff .

[0076] (9) In addition, in step S0209, based on the deformation vector E(0,0) of the strain change of all nodes of the observation surface 04 stored in the storage device 402, an observation surface matrix E is created as the deformation matrix of the strain change of the observation surface. measure .

[0077] (9-1) Specifically, such as Figure 11 As shown in the memory structure, the column vector of the deformation vector E(0,0) representing the strain change of the observation surface is arranged in the order determined in step S01, containing the strain data of the nodes of the observation surface 04. In the column vector of E(0,0), ε(k,l) represents the strain of the node at position (k,l) of the observation surface 04.

[0078] (9-2) Furthermore, regarding the information about the location where the crack occurs, as the location (i, j) of the crack initiation surface 02, a column vector E(i, j) is created, and the elements within this column vector are represented by εi_j(k, l). E(i, j) represents the strain of the nodes at the location (i, j) of the crack initiation surface 02, which is the result of the crack construction analysis. This column vector is arranged into rows according to the order of the crack initiation locations determined in step S01, and an observation surface matrix E of the strain change of the observation surface is created. measure .

[0079] <Function of the speculative data output unit 33>

[0080] (10) In Figure 9 In step S0210, the linear relationship between the displacement of the crack initiation surface 02 and the deformation of all nodes of the observation surface 04 is used, as shown in equation (1), to define the relationship between the crack initiation surface matrix Δ. crack_diff To the observation surface matrix E measure The coefficient matrix D of the mapping. Then, as shown in equation (2), multiply both sides of equation (1) from the left by the crack initiation surface matrix Δ. crack_diff inverse matrix, Δ crack_diff -1 Thus, as shown in equation (3), based on the crack initiation surface matrix Δ crack_diff and observation surface matrix E measure Construct the coefficient matrix D.

[0081] [Mathematical Expression 1]

[0082] Mathematical Formula 1

[0083]

[0084] [Mathematical Expression 2]

[0085] Mathematical formula 2

[0086]

[0087] [Mathematical Expression 3]

[0088] Mathematical Formula 3

[0089]

[0090] (11) In step S0211, calculate the inverse matrix D of the coefficient matrix D produced in step S0210. -1 .

[0091] (12) In step S0212, as the prediction model, the inverse matrix D is output. -1In this embodiment, an inference model is used as an example, where displacement is used as the state of the crack initiation surface 02 and strain is used as the state of the observation surface 04, and the relationship between them is represented by an inverse matrix. However, the inference model is not limited to the aforementioned inverse matrix. That is, the inference model is a matrix obtained by numerically analyzing the structural analytical model, which associates the state of the crack initiation surface 02 and the state of the observation surface 04, and the model can be used to infer the state of the crack initiation surface 02 from the state of the observation surface 04.

[0092] [Based on the stage F02 of reverse analysis of the learning data (in Figure 4 [The function of the crack prediction section 40]

[0093] exist Figure 2 In step S03, where measurement data is acquired, the deformation of the observation surface 04 of the object structure 01 is measured by the measuring device 300. Here, strain is shown as an example. The measuring device 300 uses a strain gauge, digital image correlation, etc. Regarding the strain here, the strain is measured under two conditions: no internal cracks and cracks occurring. The difference between these is input to the crack prediction device 400. Based on the input measurement values, the crack prediction device 400 performs... Figure 2 Step S04 is shown. Specifically, the processor 401 executes the data stored in the storage device 402 as described below. Figure 12 The flowchart.

[0094] (1) In Figure 12 In step S0401, a deformation vector is created for the observation surface 04, which arranges the measured strain data in the order determined in step S01 as column vectors of strain identical to E(i,j). The measured strain data is set to the same position as the strain data during learning.

[0095] (2) Next, in step S0402, prepare as Figure 1 The output of step S02 is the inference model (inverse matrix D) calculated in the learning phase F01. -1 ).

[0096] (3) In step S0403, based on the deformation vector of the observation surface 04 obtained by measurement in step S0401 and the prediction model (inverse matrix D) calculated in the learning phase F01 in step S0402. -1 ), calculate the displacement vector of the crack initiation surface 02.

[0097] (4) In step S0404, the displacement vector of the crack initiation surface O2 is transformed into the displacement distribution of the crack initiation surface using the same order as the learning data determined in step S01. Then, the nodes where the displacement occurs are taken as cracks, and their positions and sizes are determined. The results are output as the crack position and size shown in step S0405.

[0098] (5) In step 0405 (and Figure 2 In step S05, the location and size of the crack output in step S0404 are displayed on the display device 430 as the inspection result.

[0099] As described above, in this embodiment, it is possible to infer internal cracks in the object structure based on information obtained by measuring the shape of the structure surface using a small device equipped with an input device, a display device, a storage device, and a processor.

[0100] In the previous explanation, regarding object construction 01, the flat plate was taken as the object and represented using an orthogonal coordinate system of X, Y, and Z axes, but it can also be like... Figure 13 The diagram shows an application to a cylindrical coordinate system where object 01 is a cylinder 10 with coordinates R-axis, Z-axis, and angle θ. In this case, Figures 7A to 7C as well as Figure 8 The X-axis corresponds to the R-axis, the Y-axis corresponds to angle θ, and the Z-axis corresponds to Z-axis. The cylindrical structure of the object is as follows... Figure 14 The structure of the heat-fitted part with internal pressure 11 shown is illustrated by an example of surface shape change due to internal cracks.

[0101] As an example of an object structure that applies the above-mentioned cylindrical coordinate system, there is a retaining ring that is heat-fitted to the end of the rotor of a rotating electric machine and a heat-fitted part to the rotor core.

[0102] Additionally, the inverse matrix D -1 Alternatively, it can be obtained through matrix operations on a portion of the rigid matrix representing the displacement of the crack initiation surface 02 and the deformation of the observation surface 04, which are obtained analytically from the construction in step S0203.

[0103] Implementation method 2.

[0104] Figure 15 This is a diagram illustrating the memory structure for storing information about the displacement changes of the observation surface 04 according to Embodiment 2. Figure 16 This is a diagram illustrating the memory structure for storing information about the angle change of the observation plane according to Embodiment 2.

[0105] Regarding the inspection results as described in Embodiment 1, the presence or absence of cracks alone cannot determine the period during which the device can be stopped and used. Furthermore, it is impossible to learn all crack shapes that are intended to be detected. To address this issue, the aim is to efficiently learn from a limited amount of crack data and, based on changes in the observation surface 04, infer the location and size of any internal cracks.

[0106] Here, only the changes in the method of constructing the deformation vector and the observation surface matrix of the observation surface 04 are shown when the deformation of the observation surface 04 is not achieved by strain but by displacement or angular change.

[0107] When using displacement changes, Figure 9 In, replacing step S0205 or Figure 11 The strain change is represented by the deformation vector E(0, 0), and as shown... Figure 15 The column vector shown uses the displacement change vector Dis(0, 0). Regarding the column vector Dis(0, 0), the displacement change data of the nodes of observation surface 04 are arranged in the order determined in step S01. Figure 15 In this context, d(k, l) represents the displacement change of the node at position (k, l) of the observation surface 04. Furthermore, regarding the learned information about the location where the crack occurs, a column vector Dis(i, j) is created as the position (i, j) of the crack occurrence surface 02, and the elements within this column vector Dis(i, j) are represented by di_j(k, l). di_j(k, l) represents the positional change of the node at position (k, l) of the crack occurrence surface 02 as a result of the crack construction analysis using the node at position (i, j) of the crack occurrence surface 02. This column vector is arranged in rows according to the order of the crack occurrence positions determined in step S01, creating the observation surface matrix Dis, which is the matrix representing the deformation change of the observation surface. measure .

[0108] When using angle changes, Figure 9 In, replacing step S0205 or Figure 11 The strain change is represented by the deformation vector E(0, 0), and as shown... Figure 16 The column vector shown uses the angle-varying vector A(0, 0). Regarding column vector A(0, 0), the angle-varying data of the nodes of observation surface 04 are arranged in the order determined in step S01. Figure 16In this context, a(k, l) represents the angular change of the node at position (k, l) of the observation surface 04. Furthermore, based on the learned information about the location where the crack occurs, a column vector A(i, j) is created as the position (i, j) of the crack occurrence surface 02, and ai_j(k, l) represents the element within this column vector A(i, j). ai_j(k, l) represents the angular change of the node at position (k, l) of the crack surface, resulting from the analysis of the crack construction at position (i, j) of the crack occurrence surface 02. This column vector is arranged in rows according to the order of the crack occurrence positions determined in step S01, creating the observation surface matrix A, which is the matrix representing the angular change of the observation surface 04. measure .

[0109] As described above, by using this method, the production of learning data corresponding to all shapes of cracks generated in the surface where cracks occur can be partially automated. This allows for efficient inference of the location and size of arbitrary internal cracks based on changes in the observed surface, using less crack data for learning. Furthermore, by using not only strain changes but also displacement and angular changes as the deformation of the observed surface, the variety of measurement methods can be expanded, enabling measurements with higher accuracy and shorter time compared to strain measurements.

[0110] Implementation method 3.

[0111] Figure 17 This is a diagram of the memory structure for storing vector information on the load change of the crack initiation surface 02, as described in Embodiment 3.

[0112] Here, only the changes in the deformation vector of the observation surface 04 and the method of constructing the observation surface matrix are shown when the parameters used to represent the analytical results of the crack initiation surface 02 in matrix form are changed.

[0113] exist Figure 9 In, replacing step S0205 or Figure 10 The vector Δ(0, 0) represents the displacement change of the crack initiation surface 02, while... Figure 17 The diagram shows the column vector of the load variation vector Z(0, 0) at the crack initiation surface. With respect to the column vector Z(0, 0), the load data of the nodes at crack initiation surface 02 are arranged in a predetermined order. Figure 17In this context, ζ(i,j) represents the load of the node at position (i,j) of the crack initiation surface 02. Furthermore, based on the learned information about the location of the crack initiation, a column vector Z(i,j) is created as the position (i,j) of the crack initiation surface 02, and the elements within this column vector Z(i,j) are represented by ζi_j(i,j). ζi_j(i,j) represents the load of the node at position (i,j) of the crack surface, as a result of analyzing the crack structure at position (i,j) of the crack initiation surface 02. This column vector is arranged in rows according to the order of the crack initiation locations determined in step S01, creating the crack initiation surface matrix Z, which represents the matrix of load changes at the crack initiation surface 02. crack_diff .

[0114] Even when the parameters used in the matrix representation of the analytical results of the crack initiation surface 02 are not only displacement changes but also force changes, the deformation of the observation surface 04 can be achieved by using not only strain changes but also displacement changes and angle changes, thus obtaining the same effect as in Implementation Method 2.

[0115] As described above, by using this method, the production of learning data corresponding to all shapes of cracks generated in the crack initiation surface can be partially automated. This allows for efficient inference of the location and size of arbitrary internal cracks based on changes in the observed surface, using less crack data for learning. Furthermore, the parameters for matrix representation of the analytical results of the crack initiation surface can utilize not only displacement changes but also force changes. This is because the force at the node where the crack occurs is zero, while forces are generated outside of it.

[0116] Implementation method 4.

[0117] In Implementation 1, deformation needs to occur through internal cracks on the observation surface during inspection, so the object structure 01 is limited to structures with pre-applied force, such as heat-fitting parts. However, even without pre-applied force to the object structure 01, it is possible to measure it in the same way by applying a certain load to the object structure 01 under crack-free conditions, such as when determining learning data, and during inspection.

[0118] Specifically, in Figure 2 In step S01, when determining the learning data, the load to be applied to the object structure 01 and the location of the load application are determined and added to the boundary conditions during structure analysis. Then, in step S03, when acquiring the measurement data, the load determined during the learning data determination is applied at the location of the load application, and measurement is performed. Thus, even without pre-applying force to the object structure 01 being measured, the shape and location of the crack can be inferred.

[0119] Implementation method 5.

[0120] In addition to inverse analysis to infer the shape and location of the cracks, it is also possible to use... Figure 12 Step S0404 involves examining the object structure 01 by determining the predicted location and size of the crack, the external force applied to the rotor structure during the operation of equipment such as the turbine generator 100, and the material properties of the structure where the crack occurred, to calculate the crack propagation life and the remaining service life of the equipment. This allows for the determination of the remaining usable time of the equipment and enables planned repairs and upgrades. Furthermore, while Embodiment 1 has been described using a turbine generator as an example, it is not limited to this method.

[0121] Specifically, such as Figure 18 As shown, for in Figure 12 The information regarding the location and size of the crack, as predicted in step S0404, is supplemented by the information shown in step S22. As supplementary information, (1) the external force applied to the object structure 01, (2) the physical properties of the material used in the object structure 01, and (3) the size and location of the crack that renders the object structure 01 unusable. This information can be obtained during the product design phase, for example, from... Figure 3 The input device 420 is shown. Based on the information predicted in step S0404 and the two input information input in step S22, in step S23, the amount of crack development in the operating conditions of the turbine generator and other equipment using the object structure 01 is calculated according to destruction mechanics. Regarding the calculation of the development amount, it is not only based on destruction mechanics, but also based on the prediction results of the size and location of cracks in the time sequence. Furthermore, in step S24, the period of use that will result in the size of the cracks that make the object structure 01 unusable and the location of the unusable cracks is calculated. Then, in step S25, the remaining period of use is calculated. The calculated period of use is output by the display device 430, which can be used for equipment fault diagnosis.

[0122] Implementation method 6.

[0123] It can also be based on Figure 12 In step S0404, the predicted location and size of the crack, along with predetermined limits on the size and location of cracks within the structure, generate an alarm to prompt the equipment to cease operation, thus enabling fault diagnosis. This allows for rapid determination of when the equipment should be taken out of service. For example, through... Figure 3 The alarm device 410 shown will sound an alarm.

[0124] Specifically, such as Figure 19 As shown, for in Figure 12The information obtained in step S0404 regarding the predicted location and size of the crack is supplemented by the information shown in step S22. As supplementary information, (1) the external force applied to the object structure, (2) the physical properties of the materials used in the object structure, and (3) information regarding the size and location of the crack that renders the object structure unusable. This information can be obtained during the product design phase, for example from... Figure 3 The input device 420 is shown. Based on the two input pieces of information—the information inferred in step S0404 and the information obtained in step S22—in step S26, it is determined whether the size of the crack exceeds a predetermined threshold that renders the device unusable, or whether it exceeds this threshold within the determined period. If it does, as shown in step S27, an alarm urging the device to stop using is generated. For example, by... Figure 3 The alarm device 410, as shown, issues an alarm. If the time limit has not been exceeded, as shown in step S28, only the presence of a crack is displayed on the display device 430. Furthermore, in this case, the remaining usage period shown in Embodiment 5 can also be displayed simultaneously. By generating an alarm in this way, it can be used for equipment fault diagnosis.

[0125] This application describes various exemplary implementation methods and embodiments, but the various features, methods, and functions described in one or more implementation methods are not limited to the application of a specific implementation method, and can be applied to the implementation method alone or in various combinations.

[0126] Therefore, within the scope of the technology disclosed in this application, numerous variations not illustrated are contemplated. For example, these include variations of at least one constituent element, additions, omissions, and the extraction of at least one constituent element and its combination with constituent elements of other embodiments.

Claims

1. A crack prediction device, characterized in that, have: The data determination unit determines the shape model of the object structure to be inspected, the crack initiation surface in the shape model, and the observation surface; The speculative data calculation unit outputs the inverse matrix of the matrix obtained by numerically analyzing the structural analytical model made from the shape model, which correlates the state of the crack initiation surface determined by the data determination unit with the state of the observation surface. as well as The crack estimation unit, based on calculations of a column vector and its inverse matrix derived from measurements of the object structure actually measured in the observation surface determined by the data determination unit, estimates the state of the cracks in the crack occurrence surface determined by the data determination unit. The matrix of the inferred data calculation unit is a matrix obtained by numerically analyzing a structural analytical model made from the shape model, assuming the shape and position of the crack in the crack initiation surface determined by the data determination unit. It is a matrix relating the state of the crack initiation surface and the state of the observation surface to multiple assumed crack representations. It is a matrix obtained by arranging the states of the crack initiation surface in a predetermined order for each crack shape and a matrix obtained by arranging the states of the observation surface in a predetermined order for each crack shape.

2. The crack prediction device according to claim 1, characterized in that, The crack estimation unit calculates the displacement vector in the crack occurrence surface based on the deformation vector of the observation surface and the inverse matrix, which are generated from the deformation of the object structure actually measured in the observation surface, and infers the location and size of the crack in the crack occurrence surface based on the displacement vector.

3. The crack prediction device according to claim 1, characterized in that, The object structure is a retaining ring heat-fitted to the end of the rotor of a rotating electric motor, and the object structure is modeled in a cylindrical coordinate system.

4. A crack prediction device, characterized in that, have: The data determination unit determines the shape model of the object structure to be inspected, the crack initiation surface in the shape model, and the observation surface; The speculative data calculation unit outputs the inverse matrix of a matrix obtained by numerically analyzing the structural analytical model made from the shape model, which correlates the state of the crack initiation surface determined by the data determination unit with the state of the observation surface; and The crack estimation unit, based on calculations of a column vector and its inverse matrix derived from measurements of the object structure actually measured in the observation surface determined by the data determination unit, estimates the state of the cracks in the crack occurrence surface determined by the data determination unit. The inferred data calculation unit includes: The numerical analysis unit divides the crack initiation surface and the observation surface into unit surfaces, and performs numerical analysis on the structural analytical model based on the boundary conditions of the divided unit surfaces. The numerical analysis control unit generates the structural analytical model from the shape model, sequentially sets the boundary conditions of the structural analytical model for cracks occurring at the crack initiation surface, and sequentially performs analysis under the sequentially set boundary conditions in the numerical analysis unit, storing the analysis results of the crack initiation surface and the analysis results of the observation surface in a storage device; and The inference data output unit calculates a positive coefficient matrix that maps from the crack occurrence surface matrix obtained by representing the analytical results of the crack occurrence surface stored in the storage device to the observation surface matrix obtained by representing the analytical results of the observation surface stored in the storage device. The inverse of the positive coefficient matrix is ​​then output as the inference model. The matrix of the inferred data calculation unit is a matrix obtained by numerically analyzing a structural analytical model made from the shape model, assuming the shape and position of the crack in the crack initiation surface determined by the data determination unit. It is a matrix relating the state of the crack initiation surface and the state of the observation surface to multiple assumed crack representations. It is a matrix obtained by arranging the states of the crack initiation surface in a predetermined order for each crack shape and a matrix obtained by arranging the states of the observation surface in a predetermined order for each crack shape.

5. The crack prediction device according to claim 4, characterized in that, The input boundary conditions for the constructed analytical model are: in the crack initiation surface, the connection between the unit surfaces obtained by dividing the crack initiation surface is broken, or the displacement of the crack initiation surface is changed to a shape or boundary condition that is equal to the situation that produces the crack.

6. The crack prediction device according to claim 4, characterized in that, The analytical results of the observation surface are represented as a vector based on any changes in displacement, strain, and angle of the observation surface.

7. The crack detection device according to claim 4, characterized in that, The analytical results of the crack initiation surface are represented as a vector based on the displacement or load change of the crack initiation surface.

8. The crack detection device according to claim 4, characterized in that, The crack estimation unit calculates the displacement vector in the crack occurrence surface based on the deformation vector of the observation surface and the inverse matrix, which are generated from the deformation of the object structure actually measured in the observation surface, and infers the location and size of the crack in the crack occurrence surface based on the displacement vector.

9. The crack prediction device according to any one of claims 4 to 8, characterized in that, The object structure is a retaining ring heat-fitted to the end of the rotor of a rotating electric motor, and the object structure is modeled in a cylindrical coordinate system.

10. A method for predicting cracks, characterized in that, have: The steps are: input the shape model of the object structure to be inspected, the crack initiation surface in the shape model, and the observation surface; The step of outputting the inverse matrix of the matrix that associates the state of the crack initiation surface and the state of the observation surface in the input step of the construction analysis model made from the shape model; as well as The step of inferring the state of the cracks in the crack initiation surface input in the input step is based on the calculation of the column vector and the inverse matrix made from the measured values ​​of the object structure actually measured in the observation surface input in the input step. The matrix is ​​a matrix obtained by numerically analyzing the structural analytical model made from the shape model, assuming the shape and position of the cracks in the crack initiation surface input in the input step. It is a matrix relating the state of the crack initiation surface and the state of the observation surface to multiple assumed cracks. It is a matrix obtained by arranging the states of the crack initiation surface in a predetermined order for each crack shape and a matrix obtained by arranging the states of the observation surface in a predetermined order for each crack shape.

11. A method for inspecting cracks, characterized in that, Based on the location and size of the crack in the object structure as inferred by the crack estimation method of claim 10, the external force applied to the object structure, and the physical property values ​​of the material used in the object structure, the development life of the crack is determined, and the remaining period up to the development life is calculated.

12. A fault diagnosis method, characterized in that, An alarm is generated if, based on the location and size of a crack in an object structure estimated using the crack estimation method of claim 10, the external force applied to the object structure, and the physical property values ​​of the material used in the object structure, it is determined that the size of the crack exceeds a predetermined threshold or exceeds the predetermined threshold within a predetermined period.

13. A method for predicting cracks, characterized in that, have: The steps are: input the shape model of the object structure to be inspected, the crack initiation surface in the shape model, and the observation surface; The step of outputting the inverse matrix of the matrix that correlates the state of the crack initiation surface and the state of the observation surface in the input step from the construction analytical model made from the shape model; as well as The step of inferring the state of the cracks in the crack initiation surface input in the input step is based on the calculation of the column vector and the inverse matrix made from the measured values ​​of the object structure actually measured in the observation surface input in the input step. The steps of the output prediction model include: The steps include: sequentially setting boundary conditions for the analytical model of the structure where cracks occur at all nodes of the unit surface obtained by dividing the crack initiation surface, performing numerical analysis, and storing the analytical results of the crack initiation surface and the analytical results of the observation surface into a storage device. as well as Based on the parsing results stored in the storage device, the steps of calculating the crack occurrence surface matrix obtained by representing the crack occurrence surface as a matrix and the observation surface matrix obtained by representing the observation surface as a matrix, determining the positive coefficient matrix that maps the crack occurrence surface matrix to the observation surface matrix, and outputting the inverse matrix of the positive coefficient matrix as the inference model are as follows: The matrix is ​​a matrix obtained by numerically analyzing the structural analytical model made from the shape model, assuming the shape and position of the cracks in the crack initiation surface input in the input step. It is a matrix relating the state of the crack initiation surface and the state of the observation surface to multiple assumed cracks. It is a matrix obtained by arranging the states of the crack initiation surface in a predetermined order for each crack shape and a matrix obtained by arranging the states of the observation surface in a predetermined order for each crack shape.

14. A method for inspecting cracks, characterized in that, Based on the location and size of the crack in the object structure as predicted by the crack prediction method of claim 13, the external force applied to the object structure, and the physical property values ​​of the material used in the object structure, the development life of the crack is determined, and the remaining period up to the development life is calculated.

15. A fault diagnosis method, characterized in that, An alarm is generated if, based on the location and size of a crack in an object structure estimated using the crack estimation method of claim 13, the external force applied to the object structure, and the physical property values ​​of the material used in the object structure, it is determined that the size of the crack exceeds a predetermined threshold or exceeds the predetermined threshold within a predetermined period.

Images