Component dynamic state identification method and system, medium and equipment

Abstract

The invention discloses a component dynamics state recognition method and system based on a multilayer sensing network, a medium and equipment, and the method comprises the steps: constructing a component geometric model, carrying out the first grid cell division of the component geometric model, numbering a first node, and recording the coordinate of the first node; dividing a second grid unit, numbering second nodes, recording coordinates of the second nodes, collecting first strain data under the coordinates of the first nodes and second strain data under the coordinates of the second nodes in the movement process of the component under the action of external force, and constructing three multi-layer sensor networks; training a multilayer perceptron network by using the training set data, taking the first node coordinate as input, taking the first strain data as input data of a loss function, testing the trained multilayer perceptron network by using a test set, and solving an equivalent load of a second node coordinate of the force application surface of the component according to a node load expression; and calculating a component deformation state cloud picture.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical structure component state recognition, and in particular to a component dynamic state recognition method, system, medium and equipment based on a multi-layer perception network. Background Art

[0002] In the field of mechanical engineering, identifying the state of key mechanical component structures under dynamic models is a crucial step in design, analysis, and optimization. With the advancement of science and technology and the rapid development of the manufacturing industry, the complexity and precision requirements of mechanical structural components are increasing, and the operating environment is becoming increasingly harsh. Traditional dynamic analysis methods face numerous challenges. Traditional methods use physical models and complex numerical calculation methods to perform calculations, relying on sensor measurements to obtain a large amount of real-world component data. These methods often have limitations when dealing with complex structures and component structural requirements under complex operating conditions. For example, there are problems with a large gap between calculated and actual loads, sparse measurement points due to a small number of sensor measurement points, and excessive noise in sensor acquisition signals.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0004] The present invention provides a component dynamic state identification method, system, medium, and device based on a multi-layer perception network. These methods are used to calculate the displacement and strain at any point on a component, as well as the magnitude of its external load, while significantly improving identification accuracy. This method can be applied to cantilever beams, thick plates, blocks, and other structures. By constructing component geometric models for these cantilever beams, thick plates, and blocks and identifying their dynamic states, it is currently being used for vibration transfer path analysis and testing of aircraft wings, gas turbine casings, and feedwater pump bodies.

[0005] A component dynamic state recognition method based on a multi-layer perception network is used to construct component geometric models and identify component dynamic states for cantilever beams, thick plates, and blocks;

[0006] The method comprises,

[0007] Step 1: constructing a component geometric model and dividing the component geometric model into first grid units, numbering first nodes and recording first node coordinates, and dividing the component geometric model into second grid units, numbering second nodes and recording second node coordinates, wherein the first grid size is larger than the second grid size;

[0008] Step 2: Collect first strain data at the first node coordinate and second strain data at the second node coordinate during the movement of the component under the action of external force, wherein a strain gauge is installed at the first node coordinate, and a strain gauge is installed at the second node coordinate. The first node coordinate and its first strain data constitute a training set, and the second node coordinate and its second strain data constitute a test set;

[0009] Step 3: Build three multilayer perceptron networks, with the input being the normalized x-axis coordinates of the nodes , y-axis coordinate , z-axis coordinate and time nodes , the outputs are x-axis displacement , displacement in the y-axis direction , z-axis displacement , combining the three multilayer perceptron network outputs into a node displacement vector ;

[0010] Step 4: Using the training set data to train a multilayer perceptron network, the first node coordinate is used as input, and the first strain data is used as input data of a loss function, wherein the loss function is composed of four parts: observation loss, initial value loss, boundary loss, and virtual work loss; wherein the loss function calculation process adopts a finite element model function method;

[0011] Step 5: The trained multilayer perceptron network is tested using a test set. The second node coordinates are used as input for the multilayer perceptron network to calculate and output the target second strain data. The target second strain data is compared with the second strain data. If the difference is less than a threshold, step 6 is executed, and the process returns to step 1 to adjust the first grid size and the second grid size, and returns to step 3 to adjust the network parameters of the multilayer perceptron network, and / or returns to step 4 to adjust the training parameters of the multilayer perceptron network. Furthermore, the threshold is 80-90%.

[0012] Step 6: Solve the equivalent load of the second node coordinate of the force-applying surface of the component according to the node load expression. The equivalent load of the second node coordinate is the stress magnitude of the second node on the outer surface of the component in a direction perpendicular to the outer surface.

[0013] Step 7: Calculate the component deformation state cloud map, calculate the second node displacement size, second node strain size, second node stress size, second node deformed coordinates, strain energy size, boundary support reaction size of the component according to the displacement at the second node coordinate and draw a cloud map.

[0014] In the component dynamic state recognition method based on the multilayer perceptron network, the multilayer perceptron network model is divided into four parts: input layer, subnetwork, connection layer, and output layer. The input layer is connected to three independent and identical subnetworks respectively. The output data of the three subnetworks are combined into a three-channel vector through the connection layer and output through the output layer. The input layer input data is a 4-row n-column matrix, where n is equal to the total number of nodes, and each column is sequentially represented as the x-axis coordinate of the node. , y-axis coordinate , z-axis coordinate and time coordinates , where i is the grid node number, j is the sampling time point number, and each sub-network consists of a normalization layer, multiple fully connected layers, and an activation layer connected in sequence. The number of neurons in the last fully connected layer is 1, and the output layer outputs a 3-row n-column matrix, where each row represents the node displacement in the x-axis direction. , y-axis displacement , z-axis displacement .

[0015] In the component dynamic state recognition method based on the multi-layer perception network, the activation layer is a Sigmoid activation layer or a Relu activation layer.

[0016] In the component dynamic state recognition method based on the multi-layer perception network, the loss function is:

[0017] ,

[0018] in, are the loss weight ratios, is the observation loss, is the boundary loss, is the loss of virtual work, is the initial value loss.

[0019] In the component dynamic state recognition method based on the multi-layer perception network, the loss function calculation process includes:

[0020] Step 1: Define the element type function to calculate the element displacement. The type function expression is:

[0021] ,

[0022] in, , 、 、 is the shape function The natural coordinate system coordinates of the type function are expressed as:

[0023]

[0024] The relationship between node displacement and element displacement:

[0025]

[0026] in, is the value of the type function at the kth node, is the type function matrix, is the displacement of any node of the element, is the element node displacement vector,

[0027] Step 2: Calculate the unit Jacobian matrix And calculate its inverse matrix :

[0028]

[0029] in, is the unit Jacobian matrix, is the inverse matrix of the Jacobian matrix,

[0030] Step 3: Calculate the displacement partial derivative. The displacement partial derivative calculation formula is:

[0031]

[0032] Among them, B is the partial differential operator, which is used to calculate the partial differential of any node displacement in the unit through the node displacement.

[0033] Step 4: Calculate the unit deformation, the expression is:

[0034]

[0035] in, is the positive strain in the x-axis direction, is the positive strain in the y-axis direction, is the positive strain in the z-axis direction, is the shear strain in the xy direction, is the shear strain in the yz direction, is the shear strain in the xz direction,

[0036] Step 5: Calculate the element stress, the expression is:

[0037]

[0038] Where E is the elastic modulus of the material, is the Poisson's ratio of the material, is the normal stress in the x-axis direction, is the normal stress in the y-axis direction, is the normal stress in the z-axis direction, is the shear stress in the xy direction, is the shear stress in the yz direction, is the shear stress in the xz direction,

[0039] Step 6: Calculate the node load. The node load expression is:

[0040]

[0041] in, is the equivalent load of the node along the x-axis, is the equivalent load of the node along the y-axis direction, is the equivalent load of the node along the z-axis, is the material density, is the node equivalent damping, according to the elements in the damping matrix in the finite element method,

[0042] Step 7: Calculate the observation loss. The observation loss expression is:

[0043]

[0044] in, is the node number. Here, only the loss calculation is performed on the grid nodes selected in step 2. No. The normal strain measured in the x-axis direction of each node, No. The normal strain measured in the y-axis direction of each node, No. The normal strain measured in the z-axis direction of each node is No. The normal strain calculated in the x-axis direction of each node, No. The normal strain calculated in the y-axis direction of each node, No. The normal strain calculated in the z-axis direction of each node is:

[0045] Step 8: Calculate the boundary loss. The boundary loss expression is:

[0046]

[0047] in, Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the yoz plane , Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the xoz plane , Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the xoy plane ,

[0048] Step 9: Calculate the virtual work loss. The expression for virtual work loss is:

[0049]

[0050] in, is the node number, is the strain vector of the ith node, is the strain vector of the ith node, is the equivalent load of the node along the x-axis at the i-th node, is the equivalent load of the node along the y-axis at the i-th node, is the equivalent load of the node along the z-axis at the i-th node, is the displacement of the i-th node in the x-axis direction, is the displacement of the i-th node in the y-axis direction, is the displacement of the i-th node in the z-axis direction,

[0051] Step 10: Calculate the initial value loss. The initial value loss expression is:

[0052]

[0053] in, is the displacement in the x-axis direction when the component’s first known sampling point is is the displacement in the y-axis direction when the component’s first known sampling point is is the displacement in the z-axis direction when the component’s first known sampling point is The displacement in the x-axis direction at the first sampling point calculated by the network, The displacement in the y-axis direction at the first sampling point calculated by the network, The displacement in the z-axis direction at the first sampling point calculated by the network.

[0054] In the component dynamic state identification method based on the multi-layer perception network, the external force acts on the component by squeezing, stretching, and twisting.

[0055] In the component dynamic state identification method based on the multi-layer perception network, the component includes a mechanical structure component.

[0056] A component dynamic state recognition system based on a multi-layer perception network, which is used to construct component geometric models and identify component dynamic states for cantilever beams, thick plates, and blocks;

[0057] The system comprises,

[0058] a component geometric model construction unit, constructing the component geometric model and dividing the component geometric model into first grid units, numbering first nodes and recording first node coordinates, and dividing the component geometric model into second grid units, numbering second nodes and recording second node coordinates, wherein the first grid size is larger than the second grid size;

[0059] an acquisition unit, configured to acquire first strain data at a first node coordinate and second strain data at a second node coordinate during movement of a component under an external force, wherein a strain gauge is installed at the first node coordinate and a strain gauge is installed at the second node coordinate, the first node coordinate and its first strain data constitute a training set, and the second node coordinate and its second strain data constitute a test set;

[0060] Multilayer perceptron network unit, which is used to build three multilayer perceptron networks, with the input being the normalized x-axis coordinates of the nodes , y-axis coordinate , z-axis coordinate and time nodes , the outputs are x-axis displacement , y-axis displacement , z-axis displacement , combining the three multilayer perceptron network outputs into a node displacement vector ;

[0061] a training unit, configured to train a multilayer perceptron network using the training set data, taking the first node coordinate as input and the first strain data as input data of a loss function, wherein the loss function is composed of four parts: observation loss, initial value loss, boundary loss, and virtual work loss; wherein the loss function is calculated using a finite element model function method;

[0062] a testing unit configured to test the trained multilayer perceptron network using a test set, use the second node coordinates as input to cause the multilayer perceptron network to calculate and output target second strain data, compare the target second strain data with the second strain data, and if the difference is less than a threshold, execute step six, return to step one to adjust the first grid size and the second grid size, return to step three to adjust network parameters of the multilayer perceptron network, and / or return to step four to adjust training parameters of the multilayer perceptron network;

[0063] a calculation unit, configured to solve an equivalent load of a second node coordinate of the force-applying surface of the component according to a node load expression, wherein the equivalent load of the second node coordinate is a stress magnitude of the second node on the outer surface of the component in a direction perpendicular to the outer surface;

[0064] A cloud map unit is used to calculate the deformation state cloud map of the component, calculate the displacement size of the second node of the component, the strain size of the second node, the stress size of the second node, the coordinates of the second node after deformation, the strain energy size, and the boundary support reaction force size according to the displacement at the second node coordinate and draw a cloud map.

[0065] A computer storage medium includes computer instructions, which, when executed on a computer, cause the computer to execute the method described above.

[0066] An electronic device, comprising:

[0067] A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein:

[0068] When the processor executes the program, the method described is implemented.

[0069] Compared with existing technologies, this invention offers the following advantages: It combines experimentally measured strain data to identify the current displacement changes and load states of component structures, while conforming to finite element constitutive relations and reducing the impact of sensor noise on the model. This provides a real-time analysis method for accurate status detection of key components, early prevention of component failures, and the creation of digital twins of component structures. It also addresses the issue of complex equipment components with limited measurement points, making it difficult to obtain comprehensive sensor measurement information. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are intended only to illustrate preferred embodiments and are not to be construed as limiting the present invention. It should be understood that the drawings described below are merely examples of the present invention, and that those skilled in the art will be able to derive other drawings from these drawings without inventive effort. Throughout the drawings, identical reference numerals are used to denote identical components.

[0071] In the attached figure:

[0072] Figure 1 This is an architecture diagram of the execution steps of the component dynamic state recognition method based on the multi-layer perception network;

[0073] Figure 2 This is a calculation process diagram of a component dynamic state recognition method based on a multi-layer perception network;

[0074] Figure 3 This is a three-dimensional schematic diagram of the cantilever beam mesh model;

[0075] Figure 4This is a schematic diagram of the measurement node locations of the cantilever beam mesh model;

[0076] Figure 5 This is the framework diagram of the multi-layer perception network structure in Matlab;

[0077] Figure 6 is the change in loss function during the training process of the case model;

[0078] Figure 7 This is the stress-strain cloud diagram of the case calculation results.

[0079] The present invention will be further explained below with reference to the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0080] Specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although specific embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0081] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present invention, but the description is based on the general principles of the specification and is not intended to limit the scope of the invention. The scope of protection of the present invention shall be as defined in the attached claims.

[0082] To facilitate understanding of the embodiments of the present invention, further explanation will be given below using specific embodiments as examples in conjunction with the accompanying drawings, and the accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0083] like Figures 1 to 7 As shown in FIG, the component dynamic state recognition method based on the multi-layer perception network includes the following steps:

[0084] Step 1: constructing a component geometric model and dividing the component geometric model into first grid units, numbering first nodes and recording first node coordinates, and dividing the component geometric model into second grid units, numbering second nodes and recording second node coordinates, wherein the first grid size is larger than the second grid size;

[0085] Step 2: Collect first strain data at the first node coordinate and second strain data at the second node coordinate during the movement of the component under the action of external force, wherein a strain gauge is installed at the first node coordinate, and a strain gauge is installed at the second node coordinate. The first node coordinate and its first strain data constitute a training set, and the second node coordinate and its second strain data constitute a test set;

[0086] Step 3: Build three multilayer perceptron networks, with the input being the normalized x-axis coordinates of the nodes , y-axis coordinate , z-axis coordinate and time nodes , the outputs are x-axis displacement , y-axis displacement , z-axis displacement , combining the three multilayer perceptron network outputs into a node displacement vector ;

[0087] Step 4: Using the training set data to train a multilayer perceptron network, the first node coordinate is used as input, and the first strain data is used as input data of a loss function, wherein the loss function is composed of four parts: observation loss, initial value loss, boundary loss, and virtual work loss; wherein the loss function calculation process adopts a finite element model function method;

[0088] Step 5: The trained multilayer perceptron network is tested using a test set. The second node coordinates are used as input for the multilayer perceptron network to calculate and output the target second strain data. The target second strain data is compared with the second strain data. If the difference is less than a threshold, step 6 is executed, and the process returns to step 1 to adjust the first grid size and the second grid size, and returns to step 3 to adjust the network parameters of the multilayer perceptron network, and / or returns to step 4 to adjust the training parameters of the multilayer perceptron network. Furthermore, the threshold is 80-90%.

[0089] Step 6: Solve the equivalent load of the second node coordinate of the force-applying surface of the component according to the node load expression. The equivalent load of the second node coordinate is the stress magnitude of the second node on the outer surface of the component in a direction perpendicular to the outer surface.

[0090] Step 7: Calculate the component deformation state cloud map, calculate the second node displacement size, second node strain size, second node stress size, second node deformed coordinates, strain energy size, boundary support reaction size of the component according to the displacement at the second node coordinate and draw a cloud map.

[0091] In a preferred embodiment of the component dynamic state identification method based on the multilayer perceptron network, the multilayer perceptron network model is divided into four parts: input layer, subnetwork, connection layer, and output layer. The input layer is connected to three independent and identical subnetworks respectively. The output data of the three subnetworks are combined into a three-channel vector through the connection layer and output through the output layer. The input data of the input layer is a 4-row n-column matrix, where n is equal to the total number of nodes, and each column is sequentially represented as the x-axis coordinate of the node. , y-axis coordinate , z-axis coordinate and time coordinates , where i is the grid node number, j is the sampling time point number, and each sub-network consists of a normalization layer, multiple fully connected layers, and an activation layer connected in sequence. The number of neurons in the last fully connected layer is 1, and the output layer outputs a 3-row n-column matrix, where each row represents the node displacement in the x-axis direction. , y-axis displacement , z-axis displacement .

[0092] In a preferred embodiment of the component dynamic state recognition method based on a multi-layer perception network, the activation layer is a Sigmoid activation layer or a Relu activation layer.

[0093] In a preferred embodiment of the component dynamic state identification method based on the multi-layer perception network, the loss function is:

[0094] ,

[0095] in, are the loss weight ratios, is the observation loss, is the boundary loss, is the virtual work loss, is the initial value loss.

[0096] In a preferred embodiment of the component dynamic state identification method based on the multi-layer perception network, the loss function calculation process includes:

[0097] Step 1: Define the element type function to calculate the element displacement. The type function expression is:

[0098] ,

[0099] in, , 、 、 is the shape function The natural coordinate system coordinates of the type function are expressed as:

[0100]

[0101] The relationship between node displacement and element displacement:

[0102]

[0103] in, is the value of the type function at the kth node, is the type function matrix, is the displacement of any node of the element, is the element node displacement vector,

[0104] Step 2: Calculate the unit Jacobian matrix And calculate its inverse matrix :

[0105]

[0106] in, is the unit Jacobian matrix, is the inverse matrix of the Jacobian matrix,

[0107] Step 3: Calculate the displacement partial derivative. The displacement partial derivative calculation formula is:

[0108]

[0109] Among them, B is the partial differential operator, which is used to calculate the partial differential of any node displacement in the unit through the node displacement.

[0110] Step 4: Calculate the unit deformation, the expression is:

[0111]

[0112] in, is the positive strain in the x-axis direction, is the positive strain in the y-axis direction, is the positive strain in the z-axis direction, is the shear strain in the xy direction, is the shear strain in the yz direction, is the shear strain in the xz direction,

[0113] Step 5: Calculate the element stress, the expression is:

[0114]

[0115] Where E is the elastic modulus of the material, is the Poisson's ratio of the material, is the normal stress in the x-axis direction, is the normal stress in the y-axis direction, is the normal stress in the z-axis direction, is the shear stress in the xy direction, is the shear stress in the yz direction, is the shear stress in the xz direction,

[0116] Step 6: Calculate the node load. The node load expression is:

[0117]

[0118] in, is the equivalent load of the node along the x-axis, is the equivalent load of the node along the y-axis direction, is the equivalent load of the node along the z-axis, is the material density, is the node equivalent damping, according to the elements in the damping matrix in the finite element method,

[0119] Step 7: Calculate the observation loss. The observation loss expression is:

[0120]

[0121] in, is the node number. Here, only the loss calculation is performed on the grid nodes selected in step 2. No. The normal strain measured in the x-axis direction of each node, No. The normal strain measured in the y-axis direction of each node, No. The normal strain measured in the z-axis direction of each node is No. The normal strain calculated in the x-axis direction of each node, No. The normal strain calculated in the y-axis direction of each node, No. The normal strain calculated in the z-axis direction of each node is:

[0122] Step 8: Calculate the boundary loss. The boundary loss expression is:

[0123]

[0124] in, Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the yoz plane , Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the xoz plane , Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the xoy plane ,

[0125] Step 9: Calculate the virtual work loss. The expression for virtual work loss is:

[0126]

[0127] in, is the node number, is the strain vector of the ith node, is the strain vector of the ith node, is the equivalent load of the node along the x-axis at the i-th node, is the equivalent load of the node along the y-axis at the i-th node, is the equivalent load of the node along the z-axis at the i-th node, is the displacement of the i-th node in the x-axis direction, is the displacement of the i-th node in the y-axis direction, is the displacement of the i-th node in the z-axis direction,

[0128] Step 10: Calculate the initial value loss. The initial value loss expression is:

[0129]

[0130] in, is the displacement in the x-axis direction when the component’s first known sampling point is is the displacement in the y-axis direction when the component’s first known sampling point is is the displacement in the z-axis direction when the component’s first known sampling point is The displacement in the x-axis direction at the first sampling point calculated by the network, The displacement in the y-axis direction at the first sampling point calculated by the network, The displacement in the z-axis direction at the first sampling point calculated by the network.

[0131] In a preferred embodiment of the component dynamic state identification method based on a multi-layer perception network, the external force is the component being squeezed, stretched, or twisted.

[0132] In a preferred embodiment of the component dynamic state identification method based on a multi-layer perception network, the component includes a mechanical structure component.

[0133] A component dynamic state recognition system based on a multi-layer perception network includes:

[0134] a component geometric model construction unit, constructing the component geometric model and dividing the component geometric model into first grid units, numbering first nodes and recording first node coordinates, and dividing the component geometric model into second grid units, numbering second nodes and recording second node coordinates, wherein the first grid size is larger than the second grid size;

[0135] an acquisition unit, configured to acquire first strain data at a first node coordinate and second strain data at a second node coordinate during movement of a component under an external force, wherein a strain gauge is installed at the first node coordinate and a strain gauge is installed at the second node coordinate, the first node coordinate and its first strain data constitute a training set, and the second node coordinate and its second strain data constitute a test set;

[0136] Multilayer perceptron network unit, which is used to build three multilayer perceptron networks, with the input being the normalized x-axis coordinates of the nodes , y-axis coordinate , z-axis coordinate and time nodes , the outputs are x-axis displacement , displacement in the y-axis direction , z-axis displacement , combining the three multilayer perceptron network outputs into a node displacement vector ;

[0137] a training unit, configured to train a multilayer perceptron network using the training set data, taking the first node coordinate as input and the first strain data as input data of a loss function, wherein the loss function is composed of four parts: observation loss, initial value loss, boundary loss, and virtual work loss; wherein the loss function is calculated using a finite element model function method;

[0138] a testing unit configured to test the trained multilayer perceptron network using a test set, use the second node coordinates as input to cause the multilayer perceptron network to calculate and output target second strain data, compare the target second strain data with the second strain data, and if the difference is less than a threshold, execute step six, return to step one to adjust the first grid size and the second grid size, return to step three to adjust network parameters of the multilayer perceptron network, and / or return to step four to adjust training parameters of the multilayer perceptron network;

[0139] a calculation unit, configured to solve an equivalent load of a second node coordinate of the force-applying surface of the component according to a node load expression, wherein the equivalent load of the second node coordinate is a stress magnitude of the second node on the outer surface of the component in a direction perpendicular to the outer surface;

[0140] A cloud map unit is used to calculate the deformation state cloud map of the component, calculate the displacement size of the second node of the component, the strain size of the second node, the stress size of the second node, the coordinates of the second node after deformation, the strain energy size, and the boundary support reaction force size according to the displacement at the second node coordinate and draw a cloud map.

[0141] A computer storage medium includes computer instructions, which, when executed on a computer, cause the computer to execute the method described above.

[0142] An electronic device, characterized in that the electronic device comprises:

[0143] A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein:

[0144] When the processor executes the program, the method described is implemented.

[0145] In one embodiment, Figure 1 The figure shows the execution step architecture of a component dynamic state identification method based on a multi-layer perception network of the present invention. The steps in the figure are explained step by step based on the case. Figure 2 This is a calculation process diagram of a component dynamic state identification method based on a multi-layer perception network in the present invention.

[0146] For step one, Figure 3 The rod structure shown is a cuboid with a length of 200 mm, a width of 50 mm, and a height of 20 mm. The material is aluminum alloy 6061, and the material parameters are elastic modulus of 68.9 GPa, Poisson's ratio of 0.33, and density of The component geometry model was meshed using C3D8 elements. The coarse mesh used a 2×4×16 grid, resulting in 128 elements and 255 nodes. The fine mesh used a 4×8×32 grid, resulting in 1024 elements and 1485 nodes. The coarse and fine mesh model element nodes were numbered, and their element node matrices and element coordinate matrices were recorded.

[0147] For step 2, use a marker to mark the node positions of the outer surface of the coarse mesh model and the outer surface of the fine mesh model on the outer surface of the cantilever beam. Figure 4As shown, 4 nodes on the front and back, 8 nodes on the top and bottom, and 2 nodes on the right side were evenly selected from the coarse mesh outer surface nodes. The left side was fixed and could not be selected, totaling 26 coarse mesh measurement points. 2 nodes on the front and back, 4 nodes on the top and bottom, and 2 nodes on the right side were evenly selected from the fine mesh outer surface nodes. The left side was fixed and could not be selected, totaling 14 fine mesh measurement points. In the cantilever beam case, a larger number of measurement points was used for demonstration purposes. In practice, a smaller number of measurement points can be used as the measurement point selection scheme. The cantilever beam was placed under working conditions, that is, a concentrated load was applied to the upper surface of the free end of the cantilever beam. After deformation, the cantilever beam gradually came to rest. The strain at each measurement point during the motion was measured. It should be noted that the bidirectional strain gauge used to measure the outer surface strain can only measure the strain in the two perpendicular directions parallel to the measured surface. The strain perpendicular to the side of the belt is defaulted to 0. The three-dimensional strain data of the node in the orthogonal directions of the spatial coordinate system are recorded.

[0148] For step three, Figure 5 The following figure shows a multilayer perceptron network model constructed. This model consists of four parts: input layer, subnetwork, connection layer, and output layer. The input layer is connected to three independent and identical subnetworks. The output data of the three subnetworks is combined into a three-channel vector through the connection layer and output through the output layer. The input layer input data is a four-channel vector, sequentially represented as the node's x-axis coordinate, y-axis coordinate, z-axis coordinate, and time node t, where i is the grid node number. In this case, each subnetwork consists of four fully connected layers and a Relu activation layer connected in sequence. The first three fully connected layers each have 20 neurons, and the last fully connected layer has 1 neuron. The output layer outputs three-channel vectors representing the node's x-axis displacement, y-axis displacement, and z-axis displacement, respectively, where i is the grid node number.

[0149] For step 4, the node coordinate matrices of the coarse grid model and the fine grid model obtained in step 1 and the three-dimensional strain data of the local node space coordinate system orthogonal to the coarse grid and the fine grid obtained in step 2 are combined into the training set (coarse grid model coordinate data and local node strain data) and the test set (fine grid model coordinate data and local node strain data) of the multilayer perceptron network model. Set the training parameters. The training parameters used in this case include the use of the adam solver as the training solver, the maximum number of iterations of 300 times, and the learning rate of 0.01. Figure 2 As shown, according to the loss function calculation step described in step 4 of the invention, the loss function is calculated and the learning parameters are iteratively updated. The loss change curve is shown as Figure 6 shown.

[0150] For step five, the trained network is tested using the test set described in step four. The displacements at the nodes of the fine grid model are calculated using the trained multilayer perceptron model. The observation loss, boundary loss, initial value loss, and virtual work loss are calculated using the loss function calculation method described in step four. The strain magnitude at the selected fine grid coordinate node and the accuracy of the actual measurement results are further calculated. The training effect is evaluated in the above manner. If the gap between the training loss and the test loss is too large or the test node coordinate accuracy is too low, the grid size of step one, the sub-network parameters of the network in step two, and the network training parameters in step three are re-selected. If the training set loss is too large but the loss tends to be stable, the number of fully connected layers or neurons in the sub-network is adjusted. If the training set loss is too large and still has a downward trend, the training parameters such as the solver type, the maximum number of iterations, and the learning rate are adjusted. If the test set loss is much larger than the training set loss, indicating that the network may be overfitting, the grid size can be reduced, the learning rate can be lowered, and the number of sub-network layers and neurons can be adjusted.

[0151] For step six, solve the equivalent load of the external nodes of the force-applying surface. According to the fine mesh model node displacement obtained in step five, the stress at the component node is calculated through the calculation steps in the loss function calculation method in step four. The external node equivalent load is the stress at the node on the outer surface of the component perpendicular to the outer surface.

[0152] For step seven, calculate the component deformation state cloud map. According to the displacement at the test point calculated in step five, calculate the component node displacement, node strain, node stress, node coordinates after deformation, strain energy, boundary support reaction force and draw a cloud map.

[0153] The present invention utilizes the multi-layer perceptron (MLP) deep learning model, leveraging its powerful nonlinear mapping and adaptive learning capabilities to construct a high-precision data-driven model. The present invention can automatically train the network model from experimentally measured sparse strain data, and combined with the constitutive relations of finite element analysis (FEA), it can identify the displacement changes, stress-strain distribution, and external load state of key mechanical components under dynamic conditions. Furthermore, the present invention provides a real-time, efficient analytical tool for accurate state detection of key components, early fault prevention, and the development of digital twin technology, laying a solid foundation for the intelligent and precise development of mechanical engineering. In practice, the present invention uses the finite element model function approach to calculate the loss function, achieving a goodness of fit of 0.97 for strain and 0.96 for stress.

[0154] Although the embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-mentioned specific embodiments and application fields. The above-mentioned specific embodiments are merely illustrative and instructive, and are not restrictive. A person skilled in the art, guided by this specification and without departing from the scope of protection of the claims of the present invention, may also devise various forms, all of which fall within the scope of protection of the present invention.

Claims

1. A component dynamic state recognition method based on a multi-layer perception network, characterized in that: The method is used to construct component geometric models and identify component dynamic states for cantilever beams, thick plates, and blocks; The method comprises the following steps: Step 1: constructing a component geometric model and dividing the component geometric model into first grid units, numbering first nodes and recording first node coordinates, and dividing the component geometric model into second grid units, numbering second nodes and recording second node coordinates, wherein the first grid size is larger than the second grid size; Step 2: Collect first strain data at the first node coordinate and second strain data at the second node coordinate during the movement of the component under the action of external force, wherein a strain gauge is installed at the first node coordinate, and a strain gauge is installed at the second node coordinate. The first node coordinate and its first strain data constitute a training set, and the second node coordinate and its second strain data constitute a test set; Step 3: Build three multilayer perceptron networks, with the input being the normalized x-axis coordinates of the nodes , y-axis coordinate , z-axis coordinate and time nodes , the outputs are x-axis displacement , y-axis displacement , z-axis displacement , combining the three multilayer perceptron network outputs into a node displacement vector ; Step 4: Using the training set data to train a multilayer perceptron network, the first node coordinate is used as input, and the first strain data is used as input data of a loss function, wherein the loss function is composed of four parts: observation loss, initial value loss, boundary loss, and virtual work loss; wherein the loss function calculation process adopts a finite element model function method; Step 5: The trained multilayer perceptron network is tested using a test set. The second node coordinates are used as input for the multilayer perceptron network to calculate and output the target second strain data. The target second strain data is compared with the second strain data. If the difference is less than a threshold, step 6 is executed, and the process returns to step 1 to adjust the first grid size and the second grid size, and returns to step 3 to adjust the network parameters of the multilayer perceptron network, and / or returns to step 4 to adjust the training parameters of the multilayer perceptron network. Furthermore, the threshold is 80-90%. Step 6: Solve the equivalent load of the second node coordinate of the force-applying surface of the component according to the node load expression. The equivalent load of the second node coordinate is the stress magnitude of the second node on the outer surface of the component in a direction perpendicular to the outer surface. Step 7: Calculate the component deformation state cloud map, calculate the second node displacement size, second node strain size, second node stress size, second node deformed coordinates, strain energy size, boundary support reaction size of the component according to the displacement at the second node coordinate and draw a cloud map.

2. The component dynamic state recognition method based on a multi-layer perception network according to claim 1, characterized in that: Preferably, the multilayer perceptron network model is divided into four parts: input layer, subnetwork, connection layer, and output layer. The input layer is connected to three independent and identical subnetworks respectively. The output data of the three subnetworks are combined into a three-channel vector through the connection layer and output through the output layer. The input layer input data is a 4-row n-column matrix, where n is equal to the total number of nodes, and each column is sequentially represented as the x-axis coordinate of the node. , y-axis coordinate , z-axis coordinate and time coordinates , where i is the grid node number, j is the sampling time point number, and each sub-network consists of a normalization layer, multiple fully connected layers, and an activation layer connected in sequence. The number of neurons in the last fully connected layer is 1, and the output layer outputs a 3-row n-column matrix, where each row represents the node displacement in the x-axis direction. , y-axis displacement , z-axis displacement .

3. The component dynamic state recognition method based on multi-layer perception network according to claim 2, characterized in that: The activation layer is a Sigmoid activation layer or a Relu activation layer.

4. The component dynamic state recognition method based on a multi-layer perception network according to claim 1, characterized in that: The loss function is: ， in, are the loss weight ratios, is the observation loss, is the boundary loss, is the virtual work loss, is the initial value loss.

5. The component dynamic state recognition method based on multi-layer perception network according to claim 4 is characterized in that: The loss function calculation process includes: Step 1: Define the element type function to calculate the element displacement. The type function expression is: ， in, , 、 、 is the shape function The natural coordinate system coordinates of the type function are expressed as: ， The relationship between node displacement and element displacement: ， in, is the value of the type function at the kth node, is the type function matrix, is the displacement of any node of the element, is the element node displacement vector, Step 2: Calculate the unit Jacobian matrix And calculate its inverse matrix : ， in, is the unit Jacobian matrix, is the inverse matrix of the Jacobian matrix, Step 3: Calculate the displacement partial derivative. The displacement partial derivative calculation formula is: ， Among them, B is the partial differential operator, which is used to calculate the partial differential of any node displacement in the unit through the node displacement. Step 4: Calculate the unit deformation, the expression is: ， in, is the positive strain in the x-axis direction, is the positive strain in the y-axis direction, is the positive strain in the z-axis direction, is the shear strain in the xy direction, is the shear strain in the yz direction, is the shear strain in the xz direction, Step 5: Calculate the element stress, the expression is: ， Where E is the elastic modulus of the material, is the Poisson's ratio of the material, is the normal stress in the x-axis direction, is the normal stress in the y-axis direction, is the normal stress in the z-axis direction, is the shear stress in the xy direction, is the shear stress in the yz direction, is the shear stress in the xz direction, Step 6: Calculate the node load. The node load expression is: ， in, is the equivalent load of the node along the x-axis, is the equivalent load of the node along the y-axis direction, is the equivalent load of the node along the z-axis, is the material density, is the node equivalent damping, according to the elements in the damping matrix in the finite element method, Step 7: Calculate the observation loss. The observation loss expression is: ， in, is the node number. Here, only the loss calculation is performed on the grid nodes selected in step 2. No. The normal strain measured in the x-axis direction of each node, No. The normal strain measured in the y-axis direction of each node, No. The normal strain measured in the z-axis direction of each node is No. The normal strain calculated in the x-axis direction of each node, No. The normal strain calculated in the y-axis direction of each node, No. The normal strain calculated in the z-axis direction of each node is: Step 8: Calculate the boundary loss. The boundary loss expression is: ， in, Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the yoz plane , Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the xoz plane , Represented as the network output corresponding to the nodes on the outer surface of the component parallel to the xoy plane , Step 9: Calculate the virtual work loss. The expression for virtual work loss is: ， in, is the node number, is the strain vector of the ith node, is the strain vector of the ith node, is the equivalent load of the node along the x-axis at the i-th node, is the equivalent load of the node along the y-axis at the i-th node, is the equivalent load of the node along the z-axis at the i-th node, is the displacement of the i-th node in the x-axis direction, is the displacement of the i-th node in the y-axis direction, is the displacement of the i-th node in the z-axis direction, Step 10: Calculate the initial value loss. The initial value loss expression is: ， in, is the displacement in the x-axis direction when the component’s first known sampling point is is the displacement in the y-axis direction when the component’s first known sampling point is is the displacement in the z-axis direction when the component’s first known sampling point is The displacement in the x-axis direction at the first sampling point calculated by the network, The displacement in the y-axis direction at the first sampling point calculated by the network, The displacement in the z-axis direction at the first sampling point calculated by the network.

6. The component dynamic state recognition method based on a multi-layer perception network according to claim 1, characterized in that: The external force acts as compression, stretching and twisting on the components.

7. The component dynamic state recognition method based on a multi-layer perception network according to claim 1, characterized in that: Components include mechanical structural components.

8. A component dynamic state recognition system based on a multi-layer perception network, characterized in that: The system is used to construct component geometric models and identify component dynamic states for cantilever beams, thick plates, and blocks; The system comprises, a component geometric model construction unit, constructing the component geometric model and dividing the component geometric model into first grid units, numbering first nodes and recording first node coordinates, and dividing the component geometric model into second grid units, numbering second nodes and recording second node coordinates, wherein the first grid size is larger than the second grid size; an acquisition unit, configured to acquire first strain data at a first node coordinate and second strain data at a second node coordinate during movement of a component under an external force, wherein a strain gauge is installed at the first node coordinate and a strain gauge is installed at the second node coordinate, the first node coordinate and its first strain data constitute a training set, and the second node coordinate and its second strain data constitute a test set; Multilayer perceptron network unit, which is used to build three multilayer perceptron networks, with the input being the normalized x-axis coordinates of the nodes , y-axis coordinate , z-axis coordinate and time nodes , the outputs are x-axis displacement , displacement in the y-axis direction , z-axis displacement , combining the three multilayer perceptron network outputs into a node displacement vector ; a training unit, configured to train a multilayer perceptron network using the training set data, taking the first node coordinate as input and the first strain data as input data of a loss function, wherein the loss function is composed of four parts: observation loss, initial value loss, boundary loss, and virtual work loss; wherein the loss function is calculated using a finite element model function method; a testing unit configured to test the trained multilayer perceptron network using a test set, use the second node coordinates as input to cause the multilayer perceptron network to calculate and output target second strain data, compare the target second strain data with the second strain data, and if the difference is less than a threshold, execute step six, return to step one to adjust the first grid size and the second grid size, return to step three to adjust network parameters of the multilayer perceptron network, and / or return to step four to adjust training parameters of the multilayer perceptron network; a calculation unit, configured to solve an equivalent load of a second node coordinate of the force-applying surface of the component according to a node load expression, wherein the equivalent load of the second node coordinate is a stress magnitude of the second node on the outer surface of the component in a direction perpendicular to the outer surface; A cloud map unit is used to calculate the deformation state cloud map of the component, calculate the displacement size of the second node of the component, the strain size of the second node, the stress size of the second node, the coordinates of the second node after deformation, the strain energy size, and the boundary support reaction force size according to the displacement at the second node coordinate and draw a cloud map.

9. A computer storage medium, characterized in that The storage medium includes computer instructions, which, when executed on a computer, enable the computer to execute the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The electronic device comprises: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

Images