Rolling mill stand analysis method and device based on machine learning

Abstract

The invention provides a rolling mill stand analysis method and device based on machine learning, and relates to the technical field of rolling mill stands, and the method comprises the steps: obtaining a three-dimensional model of a to-be-analyzed rolling mill stand, processing the three-dimensional model of the to-be-analyzed rolling mill stand according to a preset processing rule, and obtaining a parameterized model corresponding to the to-be-analyzed rolling mill stand, a rolling mill stand simulation database is generated based on the parameterized model, a knowledge base is obtained through deep learning and model training, design parameters corresponding to the to-be-analyzed rolling mill stand are input into a pre-constructed deep learning model so as to obtain a prediction analysis result output by the deep learning model, and for different to-be-analyzed rolling mill stands, the design parameters of the to-be-analyzed rolling mill stand simulation database can be obtained. And the data can be input into the deep learning model for finite element simulation analysis, and different finite element simulation models do not need to be constructed for each kind of data, so that the finite element simulation efficiency is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rolling mill frames, and in particular to a rolling mill frame analysis method and device based on machine learning. Background Art

[0002] Finite element analysis (FEA), as a core technology in the field of computer-aided engineering (CAE), has been successfully applied to the structural design of heavy equipment such as rolling mills, providing theoretical support for stress distribution prediction, fatigue life assessment and lightweight design under complex working conditions.

[0003] Although the topological configuration of heavy equipment such as rolling mills is relatively stable (such as the closed frame structure of the rolling mill frame), its design parameters need to dynamically adapt to diverse requirements. For example, the expansion of the load spectrum (such as the upgrade of rolling force); another example is material iteration (from traditional cast steel to high-strength composite steel plates); another example is the derivation of geometric variants (parameter optimization, window size adjustment, etc.). As a result, engineers need to rebuild the finite element model for each parameter combination and perform full-process simulation, making the finite element simulation efficiency low. Summary of the Invention

[0004] The problem solved by the present invention is how to improve the finite element simulation efficiency of a rolling mill stand.

[0005] To solve the above problems, the present invention provides a rolling mill stand analysis method and device based on machine learning.

[0006] In a first aspect, the present invention provides a rolling mill stand analysis method based on machine learning, comprising: Obtain a three-dimensional model of the rolling mill stand to be analyzed; Processing the three-dimensional model of the rolling mill stand to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill stand to be analyzed; A rolling mill stand simulation database is generated based on a parametric model, and a knowledge base is obtained through deep learning and model training. The design parameters corresponding to the rolling mill stand to be analyzed are input into a pre-built deep learning model to obtain the predictive analysis results output by the deep learning model. The predictive analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

[0007] Optionally, the preset processing rules include at least: geometric model processing rules, mesh division rules and boundary condition rules.

[0008] Optionally, the geometric model processing rule includes the following manner: Deleting features that meet a first predetermined geometric condition in the three-dimensional model; Replacing the curved surface in the three-dimensional model that meets the second predetermined geometric condition with a regular geometric surface; retaining the load application region characteristics, constraint region characteristics, and potential high stress region characteristics in the three-dimensional model; retaining the size and chamfer of the predetermined bolt hole in the three-dimensional model, and adding a chamfer feature to the hole edge of the predetermined bolt hole; retaining the original fillet radius of the stress concentration area in the three-dimensional model, and replacing the right angles in the predetermined area with fillets; Preserving the geometric contours of the contact surface and the force application surface in the three-dimensional model; The symmetry planes in the three-dimensional model are marked.

[0009] Optionally, the grid division rule includes the following methods: Using second-order tetrahedral elements to divide the frame body in the three-dimensional model; Dividing the rack columns in the three-dimensional model using at least a first number of thickness-direction units; Dividing the chamfer inside the rack window in the three-dimensional model using a hexahedral swept grid, wherein the arc length of the chamfer is divided into at least a second number of layers of units; Bolt holes with diameters smaller than a predetermined diameter in the three-dimensional model are processed by hole filling, and bolt holes with diameters larger than a predetermined right angle in the three-dimensional model retain their geometric shapes and are divided by radial units of a third layer; The three-dimensional model is subjected to grid encryption verification through the basic grid, the first multiple encrypted grid and the second multiple encrypted grid.

[0010] Optionally, the boundary condition rule includes the following manner: Applying full degree of freedom constraints on the connection surface between the frame base and the foundation of the three-dimensional model; Applying a rolling force in the form of surface pressure to the inner side of the archway window of the three-dimensional model; Symmetry constraints are set on the frame geometric symmetry plane of the three-dimensional model.

[0011] Optionally, it also includes: Constructing a finite element simulation analysis database for a rolling mill stand with known parameters, the finite element simulation analysis database comprising: input parameters, mesh parameters, and output parameters, the input parameters comprising geometric parameters, material parameters, and operating condition parameters of a three-dimensional model of the rolling mill stand; the mesh parameters comprising a cell type, mesh density, and quality index of the mesh; and the output parameters comprising a stress field, a strain field, a displacement field, a vibration frequency, and a vibration mode of the rolling mill stand; wherein the same set of input parameters, mesh parameters, and output parameters are associated in the finite element simulation analysis database; The deep learning model is trained using a constructed finite element simulation analysis database of a rolling mill stand with known parameters.

[0012] Optionally, it also includes: Obtain reference results for the rolling mill stand to be analyzed; Comparing the prediction analysis result output by the deep learning model with the reference result to obtain a comparison result; If the comparison result is greater than the preset error, the deep learning model is adjusted.

[0013] In a second aspect, the present invention provides a rolling mill stand analysis device based on machine learning, comprising: A model acquisition module is used to obtain a three-dimensional model of the rolling mill stand to be analyzed; a data processing module, configured to process the three-dimensional model of the rolling mill stand to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill stand to be analyzed; The prediction module relies on the parameterized model to generate a rolling mill stand simulation database, obtains a knowledge base through deep learning and model training, and inputs the design parameters corresponding to the rolling mill stand to be analyzed into a pre-built deep learning model to obtain the prediction analysis results output by the deep learning model. The prediction analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

[0014] In a third aspect, the present invention provides an electronic device comprising a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the rolling mill stand analysis method based on machine learning as described in the first aspect when executing the computer program.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the machine learning-based rolling mill stand analysis method as described in the first aspect is implemented.

[0016] The beneficial effects of the rolling mill frame analysis method and device based on machine learning of the present invention are: obtaining a three-dimensional model of the rolling mill frame to be analyzed for subsequent finite element analysis of the rolling mill frame. The three-dimensional model of the rolling mill frame to be analyzed is processed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill frame to be analyzed, and a unified data processing specification is established to ensure that the parameterized model input into the deep learning model has a unified form, thereby improving the prediction accuracy of the deep learning model. A rolling mill frame simulation database is generated based on the parameterized model, and a knowledge base is obtained through deep learning and model training. The design parameters corresponding to the rolling mill frame to be analyzed are input into a pre-built deep learning model to obtain the prediction analysis results output by the deep learning model. Different rolling mill frames to be analyzed can be input into the deep learning model for finite element simulation analysis. There is no need to construct different finite element simulation models for each type of data, which effectively improves the efficiency of finite element simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Flowchart of a rolling mill stand analysis method based on machine learning according to an embodiment of the present invention; Figure 2 A flowchart of a method for analyzing a rolling mill stand based on machine learning according to an embodiment; Figure 3 A flowchart of a rolling mill stand analysis method based on machine learning according to another embodiment; Figure 4 Schematic diagram of the structure of a rolling mill stand analysis device based on machine learning according to an embodiment of the present invention; Figure 5 Schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Although certain 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 construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0019] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0020] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0021] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0022] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0023] In related technologies, the design parameters of heavy equipment such as rolling mills must dynamically adapt to diverse needs. This requires engineers to rebuild finite element models and perform full-process simulations for each parameter combination, resulting in the following bottlenecks: (1) Modeling efficiency bottleneck Constructing a finite element model once is time-consuming and requires repeated mesh sensitivity analysis to ensure calculation accuracy, accounting for more than 40% of the overall design cycle.

[0024] (2) Bottlenecks in knowledge inheritance Finite element simulation result data (stress cloud maps, deformation vectors, etc.) are stored in an unstructured and decentralized manner, lacking effective feature extraction and association rule mining. This results in a historical data asset reuse rate of less than 10%, creating a "data island" effect.

[0025] In response to the problems existing in the above-mentioned related technologies, this embodiment provides a rolling mill stand analysis method, device, electronic equipment and storage medium.

[0026] like Figure 1 As shown, an embodiment of the present invention provides a rolling mill stand analysis method based on machine learning, comprising: S100: Obtain a three-dimensional model of a rolling mill stand to be analyzed.

[0027] Specifically, the three-dimensional model of the rolling mill stand to be analyzed may come from any three-dimensional modeling software, for example, SolidEdge software.

[0028] Specifically, the three-dimensional model of the rolling mill frame to be analyzed includes some characteristic parameters of the rolling mill frame to be analyzed, for example, frame thickness, column width, window height, upper beam thickness, bolt diameter and other parameters.

[0029] S200: Processing the three-dimensional model of the rolling mill frame to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill frame to be analyzed.

[0030] Specifically, the preset processing rules include at least: geometric model processing rules, meshing rules and boundary condition rules. The geometric model processing rules are used to convert the three-dimensional model of the rolling mill stand to be analyzed into a continuous geometric model suitable for finite element analysis. This process directly affects the quality of subsequent meshing and the accuracy of the final analysis.

[0031] Meshing rules are the process of discretizing the continuous geometric model obtained after processing the geometric model processing rules into a finite number of interconnected units and nodes. It is the core foundation for subsequent finite element simulation analysis. The quality of meshing directly affects the accuracy, convergence and computational efficiency of the solution.

[0032] Boundary condition rules are used to define the interaction between the three-dimensional model and the outside world. They can simulate the mechanical behavior of the actual physical system by imposing displacement constraints (degree of freedom restrictions) and load conditions to define the interaction between the three-dimensional model and the outside world. The reasonable setting of boundary conditions directly affects the authenticity, convergence and accuracy of the solution.

[0033] S300: Generate a rolling mill stand simulation database based on the parametric model, obtain a knowledge base through deep learning and model training, input the design parameters corresponding to the rolling mill stand to be analyzed into the pre-built deep learning model to obtain the predictive analysis results output by the deep learning model, wherein the predictive analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

[0034] Specifically, the pre-built deep learning model can be any model based on deep learning, for example, a deep neural network (DNN) model or a convolutional neural network (CNN) model.

[0035] Specifically, the pre-built deep learning model is a model trained with training data of known input parameters and finite element simulation analysis results, and has a high prediction accuracy.

[0036] In this embodiment, a three-dimensional model of the rolling mill frame to be analyzed is obtained for subsequent finite element analysis of the rolling mill frame. The three-dimensional model of the rolling mill frame to be analyzed is processed according to preset processing rules to obtain a parametric model corresponding to the rolling mill frame to be analyzed, and a unified data processing specification is established to ensure that the parametric model input into the deep learning model has a unified form, thereby improving the prediction accuracy of the deep learning model. A rolling mill frame simulation database is generated based on the parametric model, and a knowledge base is obtained through deep learning and model training. The design parameters corresponding to the rolling mill frame to be analyzed are input into the pre-built deep learning model to obtain the prediction analysis results output by the deep learning model. For different three-dimensional models of rolling mill frames to be analyzed, they can all be input into the deep learning model for finite element simulation analysis. There is no need to build a different finite element simulation model for each type of data, which effectively improves the efficiency of finite element simulation.

[0037] Optionally, the geometric model processing rules include the following: The features that meet the first predetermined geometric condition are deleted from the 3D model. Specifically, the first predetermined geometric condition may be some minor features that have no effect on the mechanical properties, such as holes, slots, threads, etc., with a diameter less than 5% of the frame thickness.

[0038] The curved surfaces in the 3D model that meet the second predetermined geometric condition are replaced with regular geometric surfaces. Specifically, the second predetermined geometric condition is a regularly shaped curved surface, where a regularly shaped curved surface is a surface that has a clear mathematical definition, geometric symmetry, and can be described by simple equations, such as a sphere, cylinder, or cone.

[0039] Preserve the characteristics of load application areas, constraint areas, and potential high stress areas in the 3D model.

[0040] Specifically, the load application area refers to the geometric location where the external load (force, pressure, moment, etc.) acts in the three-dimensional model, such as a point, line, surface, or body.

[0041] Specifically, the constraint area is a key setting area used to limit the displacement of the model rigid body and simulate the actual support conditions.

[0042] Specifically, the potential high stress area refers to the key part of the three-dimensional model where stress concentration is most likely to occur or the material strength is exceeded under load.

[0043] The size and chamfer of the predetermined bolt holes in the 3D model are retained, and chamfer features are added to the hole edges of the predetermined bolt holes.

[0044] Specifically, the predetermined bolt hole is a key bolt hole, which is a bolt hole that bears complex loads in a structural connection.

[0045] Retain the original fillet radius of stress concentration areas in 3D models and replace right angles with fillets in predetermined areas.

[0046] Specifically, the stress concentration area refers to the area of ​​the three-dimensional model where the local stress increases significantly at the geometric mutation, load mutation or material discontinuity.

[0047] Specifically, the predetermined area is a non-critical area, which refers to an area that has little impact on the overall performance of the structure, has a low stress level, or has an extremely low failure risk.

[0048] Preserve the geometric contours of contact and force surfaces in the 3D model.

[0049] Marks symmetric faces in 3D models.

[0050] In this optional embodiment, the number of unnecessary meshes is reduced by simplifying small features that do not affect the mechanical properties (such as small holes, chamfers, threads, etc.). Irregular curved surfaces are replaced with regular geometric surfaces, so that high-quality structured meshes can be easily generated during subsequent mesh division. The features of key areas such as load application area features, constraint area features, and potential high stress area features are retained so as not to lose the important features of the three-dimensional model. The parameters and features of key bolt holes are retained so as not to lose the important features of key bolt holes. The radius of the original fillet of the stress concentration area is retained, and the right angles of the non-critical area are replaced with fillets. The radius of the important area can be retained, and the right angles of the non-critical area can be replaced with fillets, eliminating details that have little effect on the overall mechanical behavior. The geometric contours of the contact surface and the force action surface in the three-dimensional model are retained to ensure the geometric integrity of the contact surface and the force action surface. The symmetric surfaces in the three-dimensional model are marked and assigned identifiers to facilitate the subsequent application of boundary conditions.

[0051] Optionally, the grid division rules include the following methods: Second-order tetrahedron elements are used to divide the frame body in the three-dimensional model.

[0052] Specifically, the second-order tetrahedron element is a high-precision three-dimensional finite element with curved edges.

[0053] The rack columns in the three-dimensional model are divided using at least a first number of thickness direction units.

[0054] Specifically, the first number of layers is 6, that is, the grid shape is divided into 6 layers along the thickness dimension of the rack column.

[0055] A hexahedral swept grid is used to divide the chamfer inside the rack window in the three-dimensional model, wherein the arc length of the chamfer is divided into at least a second number of units.

[0056] Specifically, hexahedral swept mesh is the core method for generating high-quality structured meshes in finite element analysis. It forms a three-dimensional hexahedral mesh by stretching the two-dimensional quadrilateral mesh of the source surface along the sweeping path.

[0057] Specifically, the second number of layers is 6, which refers to 6 layers of grid units divided in the thickness direction of the structure, which are used to accurately capture the stress / strain gradient in the thickness direction.

[0058] The bolt holes with diameters smaller than the predetermined diameter in the 3D model are processed by hole filling, and the bolt holes with diameters larger than the predetermined right angle in the 3D model retain their geometric shapes and are divided by radial elements of the third layer.

[0059] In particular, radial elements are often used for axisymmetric or rotational structures that require high-precision capture of radial stress / strain gradients, such as pressure vessels, rotating shafts, and pipes.

[0060] Specifically, the third number of layers is 8-12 layers.

[0061] The three-dimensional model is mesh encrypted and verified through the basic mesh, the first multiple encrypted mesh and the second multiple encrypted mesh.

[0062] Specifically, the first multiple may be 1.5 times, and the second multiple may be 2 times.

[0063] In this optional embodiment, the main frame is divided by second-order tetrahedral elements to accommodate complex geometries. The frame columns in the three-dimensional model are divided by at least six layers of thickness-direction elements. Dividing the elements along the thickness can more accurately describe the distribution of stress gradients (such as bending stress and interlaminar stress) or material nonlinearity (such as plastic deformation and damage evolution). Hexahedral swept meshes are used to divide the chamfers within the frame windows in the three-dimensional model. Hexahedral elements can more accurately simulate stress gradient changes in geometrically regular areas (such as chamfers), avoiding false stress oscillations that may occur with tetrahedral elements. The swept mesh can generate uniform layered elements along the curvature of the chamfers to better resolve local high-stress areas (such as fatigue risk points). Three-layer mesh encryption of specific areas of the three-dimensional model (such as the chamfers within the frame windows) can significantly improve the balance between simulation accuracy and computational efficiency.

[0064] Optionally, the boundary condition rules include the following: Apply full degree of freedom constraints to the connection surface between the frame base and the foundation in the three-dimensional model.

[0065] Specifically, the foundation connection surface refers to the geometric surface in the three-dimensional model that directly contacts the external supporting structure (such as the foundation, mounting platform or other fixed components) and transfers loads.

[0066] In the three-dimensional model, a rolling force is applied to the inner side of the arch window of the three-dimensional model in the form of surface pressure.

[0067] Specifically, the arch window refers to a rectangular or special-shaped through-hole structure on the frame used to install the roll bearing seat (or roll assembly). It is located between the columns of the rolling mill frame (arch) and is usually two symmetrically arranged windows (upper and lower or left and right) for installing the bearing seat of the working roll / support roll.

[0068] Set symmetry constraints on the geometric symmetry plane of the frame in the 3D model.

[0069] In this optional embodiment, full-degree-of-freedom constraints are applied to the connection surface between the frame base and the foundation in the 3D model. Full-degree-of-freedom constraints are an efficient, conservative, and physically explicit way to set boundary conditions in frame simulation, and are particularly suitable for preliminary evaluations of rigid connection conditions. In the 3D model, a rolling force is applied in the form of surface pressure to the inside of the arch window of the 3D model. The rolling force is transmitted through the contact surface between the roller bearing seat and the inside of the window. The surface pressure accurately reproduces the physical nature of the distributed load. Symmetry constraints are set on the geometric symmetry surfaces of the frame in the 3D model. This symmetry constraint achieves a balance between efficiency and accuracy.

[0070] Alternatively, as Figure 2 As shown, the rolling mill stand analysis method based on machine learning in this embodiment may further include: S210: Constructing a finite element simulation analysis database for a rolling mill stand with known parameters. The finite element simulation analysis database includes input parameters, mesh parameters, and output parameters. The input parameters include geometric parameters, material parameters, and operating condition parameters of the three-dimensional model of the rolling mill stand. The mesh parameters include the element type, mesh density, and quality index of the mesh. The output parameters include the stress field, strain field, displacement field, vibration frequency, and mode shape of the rolling mill stand. The same set of input parameters, mesh parameters, and output parameters are associated in the finite element simulation analysis database.

[0071] Specifically, building a finite element simulation analysis database for a rolling mill stand with known parameters includes: S210-1: Perform sensitivity analysis on the geometric parameters of the rolling mill frame to determine the main geometric parameters that affect the finite element results, including: frame thickness, column width, window height, upper beam thickness, etc.

[0072] S210-2: Latin hypercube sampling is used to ensure uniform coverage of the parameter space, including frame geometry parameters, material parameters, and operating condition parameters. Adaptive encryption sampling is performed on parameters in key geometric areas (such as stress concentration areas). The number of basic samples is ≥1000 groups, and special case samples are supplemented to ensure the accuracy of subsequent predictions.

[0073] S210-3: Determine the category and scope of the data stored in the database.

[0074] 1) Input parameters include geometric parameters, material parameters and working condition parameters.

[0075] 2) Mesh parameters include element type, mesh density, and quality indicators.

[0076] 3) Output parameters include stress field, strain field, displacement field, vibration frequency and vibration mode.

[0077] Specifically, input and output data are collected and organized, standardized, and stored in the database in the form of key-value pairs. A tiered storage strategy is adopted to divide data into hot data, warm data, and cold data according to the access frequency and importance of the data, and store them in different databases respectively. Corresponding indexes are established to optimize query efficiency.

[0078] S220: Utilize the constructed finite element simulation analysis database of the rolling mill stand with known parameters to train the deep learning model.

[0079] Specifically, for new input, real-time standardization processing is first performed, and then the output results are quickly predicted by combining similar case retrieval and model reasoning.

[0080] In this optional embodiment, a finite element simulation analysis database is constructed based on the association of a large amount of known data to train the deep learning model, so that the prediction results of the deep learning model are more accurate.

[0081] Alternatively, as Figure 3 As shown, the rolling mill stand analysis method based on machine learning in this embodiment may further include: S310: Obtain reference results of the rolling mill stand to be analyzed.

[0082] Specifically, according to the standards and existing processes of the finite element analysis of the frame, a conventional finite element analysis is performed on the newly designed rolling mill frame structure under the design rolling force conditions, and the three-dimensional model of the newly designed rolling mill frame structure is processed strictly in accordance with the preset processing rules in step S200 to ensure the accuracy and consistency of the subsequent analysis process, and the obtained analysis results are organized into reference results corresponding to the predictive analysis results output by the deep learning model in step S300.

[0083] S320: Compare the prediction analysis results output by the deep learning model with the reference results to obtain a comparison result.

[0084] Specifically, the predictive analysis results output by the deep learning model are stored according to unified data formats and standards to ensure the integrity and traceability of the data, providing an accurate basis for subsequent comparative analysis.

[0085] S330: If the comparison result is greater than the preset error, the deep learning model is adjusted.

[0086] Specifically, the preset error should comprehensively consider engineering safety standards, industry practices and the allowable error range in actual production as the basis for judging whether the prediction accuracy of the deep learning model is qualified.

[0087] Specifically, the prediction analysis results output by the deep learning model are compared with the reference results. The comparison indicators include key mechanical performance indicators such as stress field, strain field, displacement field, vibration frequency and mode shape. A combination of quantitative and qualitative methods is used to comprehensively evaluate the error. If the comparison result is less than or equal to the preset error δ, the deep learning model is considered to have passed the verification and has the reliability for engineering application. If the comparison result is greater than the preset error δ, the cause of the error is further analyzed, which may involve insufficient training data for the deep learning model, limitations of the model structure, irrational feature selection, etc. Appropriate optimization measures are taken to address specific causes, such as adjusting the prediction method, including optimizing the deep learning model structure, improving the training algorithm, adjusting hyperparameters, etc., or adding database sample points, especially supplementing sample data related to prediction results with large errors, retraining and optimizing the model until the comparison results meet the requirements, ensuring that the prediction accuracy and reliability of the deep learning model meet the standards for engineering application.

[0088] like Figure 4 As shown, an embodiment of the present invention provides a rolling mill stand analysis device 400 based on machine learning, comprising: The model acquisition module 410 is used to acquire a three-dimensional model of the rolling mill stand to be analyzed.

[0089] The data processing module 420 is used to process the three-dimensional model of the rolling mill frame to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill frame to be analyzed.

[0090] The prediction module 430 is used to generate a rolling mill stand simulation database based on the parameterized model, obtain a knowledge base through deep learning and model training, and input the design parameters corresponding to the rolling mill stand to be analyzed into a pre-built deep learning model to obtain the prediction analysis results output by the deep learning model, wherein the prediction analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

[0091] Optionally, the preset processing rules include at least: geometric model processing rules, mesh division rules and boundary condition rules.

[0092] Optionally, the geometric model processing rule includes the following manner: Deleting features that meet a first predetermined geometric condition in the three-dimensional model; Replacing the curved surface in the three-dimensional model that meets the second predetermined geometric condition with a regular geometric surface; retaining the load application region characteristics, constraint region characteristics, and potential high stress region characteristics in the three-dimensional model; retaining the size and chamfer of the predetermined bolt hole in the three-dimensional model, and adding a chamfer feature to the hole edge of the predetermined bolt hole; retaining the original fillet radius of the stress concentration area in the three-dimensional model, and replacing the right angles in the predetermined area with fillets; Preserving the geometric contours of the contact surface and the force application surface in the three-dimensional model; The symmetry planes in the three-dimensional model are marked.

[0093] Optionally, the grid division rule includes the following methods: Using second-order tetrahedral elements to divide the frame body in the three-dimensional model; Dividing the rack columns in the three-dimensional model using at least a first number of thickness-direction units; Dividing the chamfer inside the rack window in the three-dimensional model using a hexahedral swept grid, wherein the arc length of the chamfer is divided into at least a second number of layers of units; Bolt holes with diameters smaller than a predetermined diameter in the three-dimensional model are processed by hole filling, and bolt holes with diameters larger than a predetermined right angle in the three-dimensional model retain their geometric shapes and are divided by radial units of a third layer; The three-dimensional model is mesh-encrypted by using the basic mesh, the first multiple encryption mesh and the second multiple encryption mesh.

[0094] Optionally, the boundary condition rule includes the following manner: Applying full degree of freedom constraints on the connection surface between the frame base and the foundation of the three-dimensional model; Applying a rolling force in the form of surface pressure to the inner side of the archway window of the three-dimensional model; Symmetry constraints are set on the frame geometric symmetry plane of the three-dimensional model.

[0095] Optionally, it also includes: Constructing a finite element simulation analysis database for a rolling mill stand with known parameters, the finite element simulation analysis database comprising: input parameters, mesh parameters, and output parameters, the input parameters comprising geometric parameters, material parameters, and operating condition parameters of a three-dimensional model of the rolling mill stand; the mesh parameters comprising a cell type, mesh density, and quality index of the mesh; and the output parameters comprising a stress field, a strain field, a displacement field, a vibration frequency, and a vibration mode of the rolling mill stand; wherein the same set of input parameters, mesh parameters, and output parameters are associated in the finite element simulation analysis database; The deep learning model is trained using a constructed finite element simulation analysis database of a rolling mill stand with known parameters.

[0096] Optionally, it also includes: Obtain reference results for the rolling mill stand to be analyzed; Comparing the prediction analysis result output by the deep learning model with the reference result to obtain a comparison result; If the comparison result is greater than the preset error, the deep learning model is adjusted.

[0097] like Figure 5 As shown, an electronic device 500 provided by an embodiment of the present invention includes a memory 510 and a processor 520; the memory 510 is used to store computer programs; the processor 520 is used to implement the rolling mill stand analysis method as described above when executing the computer program.

[0098] In other words, an electronic device 500 includes a memory 510 and a processor 520 coupled to the memory 510; the memory 510 is configured to store a computer program; and the processor 520 is configured to perform the following operations when executing the computer program: Obtain a three-dimensional model of the rolling mill stand to be analyzed; Processing the three-dimensional model of the rolling mill stand to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill stand to be analyzed; A rolling mill stand simulation database is generated based on a parametric model, and a knowledge base is obtained through deep learning and model training. The design parameters corresponding to the rolling mill stand to be analyzed are input into a pre-built deep learning model to obtain the predictive analysis results output by the deep learning model. The predictive analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

[0099] An embodiment of the present invention provides a computer-readable storage medium, wherein a computer program is stored on the storage medium. When the computer program is executed by a processor, the rolling mill stand analysis method described above is implemented.

[0100] In other words, a non-volatile computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the following operations: Obtain a three-dimensional model of the rolling mill stand to be analyzed; Processing the three-dimensional model of the rolling mill stand to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill stand to be analyzed; A rolling mill stand simulation database is generated based on a parametric model, and a knowledge base is obtained through deep learning and model training. The design parameters corresponding to the rolling mill stand to be analyzed are input into a pre-built deep learning model to obtain the predictive analysis results output by the deep learning model. The predictive analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

[0101] An electronic device 500 that can serve as a server or client of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device 500 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 500 can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0102] Electronic device 500 includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). The RAM can also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. An input / output (I / O) interface is also connected to the bus.

[0103] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM). In this application, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network elements. Some or all of these units can be selected based on actual needs to achieve the objectives of the embodiments of the present invention. Furthermore, the functional units in the various embodiments of the present invention can be integrated into a single processing unit, each unit can exist physically separately, or two or more units can be integrated into a single unit. These integrated units can be implemented in either hardware or software functional units.

[0104] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A rolling mill stand analysis method based on machine learning, characterized in that: include: Obtaining a three-dimensional model of a rolling mill stand to be analyzed, and obtaining design parameters corresponding to the rolling mill stand to be analyzed; Processing the three-dimensional model of the rolling mill stand to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill stand to be analyzed; A rolling mill stand simulation database is generated based on a parametric model, and a knowledge base is obtained through deep learning and model training. The design parameters corresponding to the rolling mill stand to be analyzed are input into a pre-built deep learning model to obtain the predictive analysis results output by the deep learning model. The predictive analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

2. The rolling mill stand analysis method based on machine learning according to claim 1, characterized in that: The preset processing rules include at least: geometric model processing rules, grid division rules and boundary condition rules.

3. The rolling mill stand analysis method based on machine learning according to claim 2, characterized in that: The geometric model processing rules include the following methods: Deleting features that meet a first predetermined geometric condition in the three-dimensional model; Replacing the curved surface in the three-dimensional model that meets the second predetermined geometric condition with a regular geometric surface; retaining the load application region characteristics, constraint region characteristics, and potential high stress region characteristics in the three-dimensional model; retaining the size and chamfer of the predetermined bolt hole in the three-dimensional model, and adding a chamfer feature to the hole edge of the predetermined bolt hole; retaining the original fillet radius of the stress concentration area in the three-dimensional model, and replacing the right angles in the predetermined area with fillets; Preserving the geometric contours of the contact surface and the force application surface in the three-dimensional model; The symmetry planes in the three-dimensional model are marked.

4. The rolling mill stand analysis method based on machine learning according to claim 2, characterized in that: The grid division rules include the following methods: Using second-order tetrahedral elements to divide the frame body in the three-dimensional model; Dividing the rack columns in the three-dimensional model using at least a first number of thickness-direction units; Dividing the chamfer inside the rack window in the three-dimensional model using a hexahedral swept grid, wherein the arc length of the chamfer is divided into at least a second number of layers of units; Bolt holes with diameters smaller than a predetermined diameter in the three-dimensional model are processed by hole filling, and bolt holes with diameters larger than a predetermined right angle in the three-dimensional model retain their geometric shapes and are divided by radial units of a third layer; The three-dimensional model is subjected to grid encryption verification through the basic grid, the first multiple encrypted grid and the second multiple encrypted grid.

5. The rolling mill stand analysis method based on machine learning according to claim 2, characterized in that: The boundary condition rules include the following: Applying full degree of freedom constraints on the connection surface between the frame base and the foundation of the three-dimensional model; Applying a rolling force in the form of surface pressure to the inner side of the archway window of the three-dimensional model; Symmetry constraints are set on the frame geometric symmetry plane of the three-dimensional model.

6. The rolling mill stand analysis method based on machine learning according to any one of claims 1 to 5, characterized in that: Also includes: Constructing a finite element simulation analysis database for a rolling mill stand with known parameters, the finite element simulation analysis database comprising: input parameters, mesh parameters, and output parameters, the input parameters comprising geometric parameters, material parameters, and operating condition parameters of a three-dimensional model of the rolling mill stand; the mesh parameters comprising a cell type, mesh density, and quality index of the mesh; and the output parameters comprising a stress field, a strain field, a displacement field, a vibration frequency, and a vibration mode of the rolling mill stand; wherein the same set of input parameters, mesh parameters, and output parameters are associated in the finite element simulation analysis database; The deep learning model is trained using a constructed finite element simulation analysis database of a rolling mill stand with known parameters.

7. The rolling mill stand analysis method based on machine learning according to any one of claims 1 to 5, characterized in that: Also includes: Obtain reference results for the rolling mill stand to be analyzed; Comparing the prediction analysis result output by the deep learning model with the reference result to obtain a comparison result; If the comparison result is greater than the preset error, the deep learning model is adjusted.

8. A rolling mill stand analysis device based on machine learning, characterized in that: include: A model acquisition module is used to obtain a three-dimensional model of the rolling mill stand to be analyzed; a data processing module, configured to process the three-dimensional model of the rolling mill stand to be analyzed according to preset processing rules to obtain a parameterized model corresponding to the rolling mill stand to be analyzed; The prediction module relies on the parameterized model to generate a rolling mill stand simulation database, obtains a knowledge base through deep learning and model training, and inputs the design parameters corresponding to the rolling mill stand to be analyzed into a pre-built deep learning model to obtain the prediction analysis results output by the deep learning model. The prediction analysis results are the finite element simulation analysis results of the rolling mill stand to be analyzed.

9. An electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the rolling mill stand analysis method based on machine learning as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the rolling mill stand analysis method based on machine learning as described in any one of claims 1 to 7 is implemented.

Images