Hydro-generator magnetic pole cooling fin tool configuration method and system based on graph neural network

Abstract

The invention discloses a hydro-generator magnetic pole cooling fin tool configuration method and system based on a graph neural network, and belongs to the technical field of hydro-generator assembly process intelligence. The problem of intelligent upgrading of hydro-generator magnetic pole cooling fin tool configuration is solved. The method comprises the following steps: data acquisition: acquiring surface temperature field distribution data, deformation parameter data and three-dimensional vibration spectrum data of the magnetic pole cooling fin, and microscopic morphology image data of a contact area of the magnetic pole cooling fin and a clamp; constructing a node feature vector of a topological graph structure comprising a magnetic pole cooling fin node, a clamp node and an environmental parameter node, and generating a topological graph comprising node features and edge weights by using a graph convolutional neural network; calculating fixture positioning probability distribution, and outputting an optimal tool combination scheme; and executing the physical assembly instruction to complete the configuration of the hydro-generator magnetic pole cooling fin tool. The configuration efficiency of the hydro-generator magnetic pole cooling fin tool is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent technology for hydro-generator assembly processes, specifically relating to a method and system for configuring tooling for hydro-generator magnetic pole heat sinks based on graph neural networks. Background Technology

[0002] Currently, the configuration of tooling fixtures for the rotor pole heat sinks of hydro generators mainly relies on a linear matching method between manual experience and two-dimensional CAD drawings. Traditional solutions use a rule base based on geometric dimensions for fixture positioning point selection, which has three main drawbacks: First, the three-dimensional curved surface structure and asymmetrical layout of the heat sink are difficult to accurately match using manual experience, leading to uneven stress distribution during assembly. Second, existing computer-aided tooling systems fail to effectively integrate multiple physical field parameters such as the material's thermal expansion coefficient and dynamic load distribution, causing the service life prediction error of the tooling fixture to exceed the allowable range. Finally, traditional methods lack the ability to deeply model the topological relationships of irregularly shaped heat sinks, requiring repeated physical trial assembly verification when dealing with novel composite heat dissipation structures.

[0003] Existing intelligent configuration technologies mostly employ machine learning methods based on feature engineering, which have limitations in three dimensions: First, the discretization feature extraction process disrupts the topological correlation between the heat sink surface channels and the supporting ribs, resulting in a large deviation in the calculation of the stress concentration factor of the tooling contact surface; Second, the static parameter model cannot integrate the vibration spectrum data during unit operation in real time, causing the adjustment of the clamping preload to lag behind the actual operating conditions; Third, traditional neural networks have difficulty analyzing the multi-level connection relationships inside the heat sink assembly, increasing the probability of clamping interference when assembling non-standard heat sinks. Summary of the Invention

[0004] The problem to be solved by this invention is to realize the intelligent upgrade of the tooling configuration of the magnetic pole heat sink of the hydro generator, and to propose a method and system for the tooling configuration of the magnetic pole heat sink of the hydro generator based on graph neural network.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for configuring a tooling for a magnetic pole heat sink of a hydro-generator based on a graph neural network includes the following steps:

[0007] S1. Arrange a strain sensor array and a vibration accelerometer array at the root of the ribs of the magnetic pole heat sink, and arrange an infrared thermal imager and a high-definition image acquisition device on the surface of the magnetic pole heat sink;

[0008] S2. The data acquisition module acquires data based on the arrangement scheme in step S1, and collects surface temperature field distribution data, deformation parameter data, three-dimensional vibration spectrum data, and microscopic morphology image data of the contact area between the magnetic pole heat sink and the fixture.

[0009] S3. Based on the data collected in step S2, the graph neural network processing module constructs node feature vectors of a topological graph structure including magnetic pole heat sink nodes, fixture nodes and environmental parameter nodes, and uses a graph convolutional neural network to generate a topological graph including node features and edge weights.

[0010] S4. For the topology map obtained in step S3, configure the optimization module to calculate the fixture positioning probability distribution and output the optimal tooling combination scheme;

[0011] S5. For the optimal tooling combination scheme obtained in step S4, the control execution module executes the physical assembly command to complete the tooling configuration of the hydro-generator magnetic pole heat sink.

[0012] Furthermore, the specific implementation method of step S3 includes the following steps:

[0013] S3.1. Construct node feature vectors for a topology graph structure including magnetic pole heat sink nodes, fixture nodes, and environmental parameter nodes, and define the magnetic pole heat sink graph structure. , where vertex set , Let n represent the vector of the i-th feature point on the surface of the magnetic pole heat sink, where n is the total number of feature points and the edge set is the vector vector of the i-th feature point on the surface of the magnetic pole heat sink. The distance between adjacent feature points The topological connections of mm constitute the structure;

[0014] S3.2. Constructing the vertex feature matrix Each vertex feature vector includes coordinates. Temperature value strain value and surface roughness ;

[0015] S3.3. Establish an adjacency matrix based on the geometry and topology of the magnetic pole heat sink. The node connection conditions of the adjacency matrix satisfy... If and only if ,in, For nodes and The adjacency relationship between elements, where 1 indicates that they are connected and 0 indicates that they are not connected. Let be the node vector of the j-th feature point on the surface of the magnetic pole heat sink. The preset connection threshold;

[0016] S3.4. Perform graph convolution operation based on the adjacency matrix, introduce an attention mechanism in the output layer to calculate edge weight coefficients, and generate a topological graph including node features and edge weights.

[0017] Furthermore, the specific implementation method of step S4 includes the following steps:

[0018] S4.1. Calculate the fixture positioning probability distribution, the expression is:

[0019]

[0020] in, This is the probability distribution matrix for fixture positioning. To normalize a vector into a probability distribution, Let L be the node feature embedding matrix of the Lth layer of the graph neural network. The weight matrix for fixture positioning prediction;

[0021] S4.2. Establish a fixture layout constraint model and set the mechanical equilibrium equations as follows:

[0022]

[0023] in, The force exerted by the clamps is m, and the number of clamps is m. The force is the weight of the magnetic pole heat sink;

[0024] The deformation constraints for constructing the magnetic pole heat sink are as follows:

[0025] mm

[0026] in, Let be the displacement change of the i-th feature point on the surface of the magnetic pole heat sink in the Z direction. It is a function for maximizing the value;

[0027] S4.3. Apply the sequential quadratic programming method for local optimization to obtain the optimal tooling combination scheme. The expression of the sequential quadratic programming method is:

[0028]

[0029] in, The weighting coefficient for the i-th feature point on the surface of the magnetic pole heat sink is given by [the weighting coefficient]. The force balance regularization parameter.

[0030] Furthermore, the specific implementation method for executing the physical assembly instruction in step S5 includes the following steps:

[0031] S5.1. Establish the kinematic model of the robotic arm and use the Denavit-Hartenberg parametric method to determine the transformation matrix of the coordinate system of each joint of the robotic arm;

[0032] S5.2. Perform the five-point calibration method to solve for the transformation matrix from the camera coordinate system to the robot arm base coordinate system. ;

[0033] S5.3. Verify positioning accuracy, requiring repeatability error. mm.

[0034] A tooling configuration system for a hydro-generator magnetic pole heat sink based on a graph neural network is implemented based on the aforementioned tooling configuration method for a hydro-generator magnetic pole heat sink based on a graph neural network. The system includes a data acquisition module, a graph neural network processing module, a configuration optimization module, and a control execution module, which are connected sequentially.

[0035] The beneficial effects of this invention are:

[0036] The present invention discloses a tooling configuration method for magnetic pole heat sinks of hydro generators based on graph neural networks, which improves configuration efficiency by significantly shortening the traditional configuration cycle through graph structure modeling; optimizes accuracy by improving the control accuracy of fixture positioning error, which is significantly better than manual methods; enhances adaptability by supporting automatic adaptation to different types of magnetic pole heat sinks; and reduces maintenance costs by reducing the system maintenance frequency through the potential vector update mechanism of data. Attached Figure Description

[0037] Figure 1 This is a flowchart of a tooling configuration method for a hydro generator magnetic pole heat sink based on a graph neural network, as described in this invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0039] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0040] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 Detailed explanation is as follows:

[0041] Example 1:

[0042] A method for configuring a tooling for a magnetic pole heat sink of a hydro-generator based on a graph neural network includes the following steps:

[0043] S1. Arrange a strain sensor array and a vibration accelerometer array at the root of the ribs of the magnetic pole heat sink, and arrange an infrared thermal imager and a high-definition image acquisition device on the surface of the magnetic pole heat sink;

[0044] S2. The data acquisition module acquires data based on the arrangement scheme in step S1, and collects surface temperature field distribution data, deformation parameter data, three-dimensional vibration spectrum data, and microscopic morphology image data of the contact area between the magnetic pole heat sink and the fixture.

[0045] Furthermore, the temperature field distribution on the surface of the magnetic pole heat sink was collected as follows: The deformation parameters measured at the root of the magnetic pole heat sink fins are as follows: The collected three-dimensional vibration spectrum is The collected microscopic images of the contact area of ​​the fixture are as follows: .

[0046] S3. Based on the data collected in step S2, the graph neural network processing module constructs node feature vectors of a topological graph structure including magnetic pole heat sink nodes, fixture nodes and environmental parameter nodes, and uses a graph convolutional neural network to generate a topological graph including node features and edge weights.

[0047] Furthermore, the specific implementation method of step S3 includes the following steps:

[0048] S3.1. Construct node feature vectors for a topology graph structure including magnetic pole heat sink nodes, fixture nodes, and environmental parameter nodes, and define the magnetic pole heat sink graph structure. , where vertex set , Let n represent the vector of the i-th feature point on the surface of the magnetic pole heat sink, where n is the total number of feature points and the edge set is the vector vector of the i-th feature point on the surface of the magnetic pole heat sink. The distance between adjacent feature points The topological connections of mm constitute the structure;

[0049] S3.2. Constructing the vertex feature matrix Each vertex feature vector includes coordinates. Temperature value strain value and surface roughness ;

[0050] S3.3. Establish an adjacency matrix based on the geometry and topology of the magnetic pole heat sink. The node connection conditions of the adjacency matrix satisfy... If and only if ,in, For nodes and The adjacency relationship between elements, where 1 indicates that they are connected and 0 indicates that they are not connected. Let be the node vector of the j-th feature point on the surface of the magnetic pole heat sink. The preset connection threshold;

[0051] S3.4. Perform graph convolution operation based on the adjacency matrix, introduce an attention mechanism in the output layer to calculate edge weight coefficients, and generate a topological graph including node features and edge weights.

[0052] Furthermore, the expression for performing graph convolution is:

[0053]

[0054] in, It is an adjacency matrix with self-connections. For degree matrix, For trainable weight matrix, It is the ReLU activation function;

[0055] Then, feature propagation is achieved through cascaded graph neural network layers, with the node update formula being:

[0056]

[0057] in, Indicates the first Layered networks at nodes The hidden state, It is the set of adjacent nodes;

[0058] The expression for calculating the edge weight coefficients by introducing an attention mechanism in the output layer is as follows:

[0059]

[0060] in, For attention parameter vectors, Represents vertices The hidden state;

[0061] Define loss function for:

[0062]

[0063] in, , , These are respectively the loss in positioning accuracy, the loss in stress distribution, and the loss in energy consumption. , and These are the weighting coefficients for positioning accuracy loss, stress distribution loss, and energy consumption loss, respectively.

[0064] S4. For the topology map obtained in step S3, configure the optimization module to calculate the fixture positioning probability distribution and output the optimal tooling combination scheme;

[0065] Furthermore, the specific implementation method of step S4 includes the following steps:

[0066] S4.1. Calculate the fixture positioning probability distribution, the expression is:

[0067]

[0068] in, This is the probability distribution matrix for fixture positioning. To normalize a vector into a probability distribution, Let L be the node feature embedding matrix of the Lth layer of the graph neural network. The weight matrix for fixture positioning prediction;

[0069] S4.2. Establish a fixture layout constraint model and set the mechanical equilibrium equations as follows:

[0070]

[0071] in, The force exerted by the clamps is m, and the number of clamps is m. The force is the weight of the magnetic pole heat sink;

[0072] The deformation constraints for constructing the magnetic pole heat sink are as follows:

[0073] mm

[0074] in, Let be the displacement change of the i-th feature point on the surface of the magnetic pole heat sink in the Z direction. It is a function for maximizing the value;

[0075] S4.3. Apply the sequential quadratic programming method for local optimization to obtain the optimal tooling combination scheme. The expression of the sequential quadratic programming method is:

[0076]

[0077] in, The weighting coefficient for the i-th feature point on the surface of the magnetic pole heat sink is given by [the weighting coefficient]. The force balance regularization parameter.

[0078] Furthermore, an initial solution set is generated based on the prediction results of the graph neural network. A simulated annealing algorithm is used for global search, and the temperature decay function is: ;

[0079] Furthermore, the loss function for optimizing prediction is:

[0080]

[0081] in, For the graph Laplace matrix, The KL divergence term constrains the consistency between the predicted distribution and the historical experience distribution, using adjustable weighting coefficients.

[0082] S5. For the optimal tooling combination scheme obtained in step S4, the control execution module executes the physical assembly command to complete the tooling configuration of the hydro-generator magnetic pole heat sink.

[0083] Furthermore, the specific implementation method for executing the physical assembly instruction in step S5 includes the following steps:

[0084] S5.1. Establish the kinematic model of the robotic arm and use the Denavit-Hartenberg parametric method to determine the transformation matrix of the coordinate system of each joint of the robotic arm;

[0085] S5.2. Perform the five-point calibration method to solve for the transformation matrix from the camera coordinate system to the robot arm base coordinate system. ;

[0086] S5.3. Verify positioning accuracy, requiring repeatability error. mm.

[0087] Furthermore, it is equipped with a six-degree-of-freedom robotic arm and an adaptive gripping end effector; as well as a human-machine interaction terminal that integrates a touch screen and a voice command input device;

[0088] Furthermore, the real-time monitoring method involves periodically collecting sensor data and calculating a health index. ,in For strain normalization value, Temperature deviation; when A three-level alarm mechanism is triggered in a timely manner; the hyperparameters of the graph neural network are dynamically adjusted based on historical operation and maintenance data. ,in This represents the number of valid fault samples.

[0089] Furthermore, the human-computer interaction terminal can dynamically adjust the interface, receive real-time monitoring data streams, and update graph node features; a multi-objective optimizer can solve for the Pareto front solution set of fixture configuration parameters. The visualization unit generates a 3D thermal cloud map to display the optimal clamping force distribution. ;

[0090] Furthermore, when the human-computer interaction terminal detects the dominant vibration frequency... Offset exceeds threshold When the temperature gradient is reached, the graph structure reconstruction mechanism is triggered and the node connection relationships are updated; when the temperature gradient is reached... When the safety threshold is exceeded, a fixture position reconfiguration instruction is generated;

[0091] Furthermore, the human-computer interaction terminal can predict the remaining service life of fixture components by training a time-series graph network model based on long-term operating data. ;when It generates a component replacement warning signal in real time.

[0092] Example 2:

[0093] A tooling configuration system for a hydro-generator magnetic pole heat sink based on a graph neural network is implemented based on the tooling configuration method for a hydro-generator magnetic pole heat sink based on a graph neural network described in Example 1. The system includes a data acquisition module, a graph neural network processing module, a configuration optimization module, and a control execution module, which are connected in sequence.

[0094] Furthermore, real-time data interaction and connectivity are achieved through industrial Ethernet.

[0095] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0096] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for assembling a tooling for a magnetic pole heat sink of a hydro-generator based on a graph neural network, characterized in that, Includes the following steps: S1. Arrange a strain sensor array and a vibration accelerometer array at the root of the ribs of the magnetic pole heat sink, and arrange an infrared thermal imager and a high-definition image acquisition device on the surface of the magnetic pole heat sink; S2. The data acquisition module acquires data based on the arrangement scheme in step S1, and collects surface temperature field distribution data, deformation parameter data, three-dimensional vibration spectrum data, and microscopic morphology image data of the contact area between the magnetic pole heat sink and the fixture. S3. Based on the data collected in step S2, the graph neural network processing module constructs node feature vectors of a topological graph structure including magnetic pole heat sink nodes, fixture nodes and environmental parameter nodes, and uses a graph convolutional neural network to generate a topological graph including node features and edge weights. S4. For the topology map obtained in step S3, configure the optimization module to calculate the fixture positioning probability distribution and output the optimal tooling combination scheme; S5. For the optimal tooling combination scheme obtained in step S4, the control execution module executes the physical assembly command to complete the tooling configuration of the hydro-generator magnetic pole heat sink.

2. The method for configuring a tooling for a hydro-generator magnetic pole heat sink based on a graph neural network according to claim 1, characterized in that, The specific implementation method of step S3 includes the following steps: S3.

1. Construct node feature vectors for a topology graph structure including magnetic pole heat sink nodes, fixture nodes, and environmental parameter nodes, and define the magnetic pole heat sink graph structure. , where vertex set , Let n represent the vector of the i-th feature point on the surface of the magnetic pole heat sink, where n is the total number of feature points and the edge set is the vector vector of the i-th feature point on the surface of the magnetic pole heat sink. The distance between adjacent feature points The topological connections of mm constitute the structure; S3.

2. Constructing the vertex feature matrix Each vertex feature vector includes coordinates. Temperature value strain value and surface roughness ; S3.

3. Establish an adjacency matrix based on the geometry and topology of the magnetic pole heat sink. The node connection conditions of the adjacency matrix satisfy... If and only if ,in, For nodes and The adjacency relationship between elements, where 1 indicates that they are connected and 0 indicates that they are not connected. Let be the node vector of the j-th feature point on the surface of the magnetic pole heat sink. The preset connection threshold; S3.

4. Perform graph convolution operation based on the adjacency matrix, introduce an attention mechanism in the output layer to calculate edge weight coefficients, and generate a topological graph including node features and edge weights.

3. The method for configuring a tooling for a hydro-generator magnetic pole heat sink based on a graph neural network according to claim 2, characterized in that, The specific implementation method of step S4 includes the following steps: S4.

1. Calculate the fixture positioning probability distribution, the expression is: in, This is the probability distribution matrix for fixture positioning. To normalize a vector into a probability distribution, Let L be the node feature embedding matrix of the Lth layer of the graph neural network. The weight matrix for fixture positioning prediction; S4.

2. Establish a fixture layout constraint model and set the mechanical equilibrium equations as follows: in, The force exerted by the clamps is m, and the number of clamps is m. The force is the weight of the magnetic pole heat sink; The deformation constraints for constructing the magnetic pole heat sink are as follows: mm in, Let be the displacement change of the i-th feature point on the surface of the magnetic pole heat sink in the Z direction. It is a function for maximizing the value; S4.

3. Apply the sequential quadratic programming method for local optimization to obtain the optimal tooling combination scheme. The expression of the sequential quadratic programming method is: in, The weighting coefficient for the i-th feature point on the surface of the magnetic pole heat sink is given by [the weighting coefficient]. The force balance regularization parameter.

4. The method for configuring a tooling for a hydro-generator magnetic pole heat sink based on a graph neural network according to claim 3, characterized in that, The specific implementation method for executing the physical assembly command in step S5 includes the following steps: S5.

1. Establish the kinematic model of the robotic arm and use the Denavit-Hartenberg parametric method to determine the transformation matrix of the coordinate system of each joint of the robotic arm; S5.

2. Perform the five-point calibration method to solve for the transformation matrix from the camera coordinate system to the robot arm base coordinate system. ; S5.

3. Verify positioning accuracy, requiring repeatability error. mm.

5. A tooling configuration system for a hydro-generator magnetic pole heat sink based on a graph neural network, implemented using the tooling configuration method for a hydro-generator magnetic pole heat sink based on a graph neural network as described in any one of claims 1-4, characterized in that, It includes a data acquisition module, a graph neural network processing module, a configuration optimization module, and a control execution module, which are connected in sequence.

Images