CN117195440A - Main shaft system structure optimization method, device, equipment and storage medium - Google Patents
Main shaft system structure optimization method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN117195440A CN117195440A CN202311478533.6A CN202311478533A CN117195440A CN 117195440 A CN117195440 A CN 117195440A CN 202311478533 A CN202311478533 A CN 202311478533A CN 117195440 A CN117195440 A CN 117195440A
- Authority
- CN
- China
- Prior art keywords
- main shaft
- sample
- neural network
- shaft system
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000005457 optimization Methods 0.000 title claims abstract description 37
- 238000003860 storage Methods 0.000 title claims abstract description 15
- 238000003062 neural network model Methods 0.000 claims abstract description 71
- 238000013461 design Methods 0.000 claims abstract description 64
- 238000005096 rolling process Methods 0.000 claims abstract description 56
- 238000010586 diagram Methods 0.000 claims abstract description 44
- 238000012549 training Methods 0.000 claims abstract description 39
- 230000002787 reinforcement Effects 0.000 claims abstract description 9
- 238000000605 extraction Methods 0.000 claims description 26
- 238000004458 analytical method Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 8
- 238000004088 simulation Methods 0.000 claims description 4
- 238000013526 transfer learning Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000000547 structure data Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000008719 thickening Effects 0.000 description 1
Landscapes
- Numerical Control (AREA)
Abstract
The invention provides a main shaft system structure optimization method, a main shaft system structure optimization device, main shaft system structure optimization equipment and a storage medium, and relates to the technical field of artificial intelligence, wherein the main shaft system structure optimization method comprises the following steps: generating a main shaft system structure diagram based on main shaft system lightweight design data of a target machine tool; inputting the main shaft system structure diagram into each trained graph rolling neural network model in the model group respectively, obtaining a main shaft performance prediction result output by each graph rolling neural network model, wherein each graph rolling neural network model in the model group corresponds to a performance index respectively, and the graph rolling neural network model is obtained based on reinforcement learning training; and optimizing the lightweight design data of the spindle system based on the performance prediction results of each spindle. The invention can improve the lightweight design efficiency of the spindle system.
Description
Technical Field
The present invention relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, a device, and a storage medium for optimizing a spindle system structure.
Background
In order to achieve weight saving of a spindle system, in the prior art, lightweight design data of the spindle system needs to be verified in a test manner, design data of the spindle system is optimized based on test results, and when the lightweight design data of the spindle system is changed, test settings need to be changed, which is time-consuming, resulting in low structural lightweight design efficiency of the spindle system of the machine tool.
Disclosure of Invention
The invention provides a main shaft system structure optimization method, a main shaft system structure optimization device, main shaft system structure optimization equipment and a storage medium, which are used for solving the defect of low design efficiency of a main shaft system structure of a machine tool in the prior art and improving the light design efficiency of the main shaft system structure of the machine tool.
The invention provides a main shaft system structure optimization method, which comprises the following steps:
generating a main shaft system structure diagram based on main shaft system light design data of a target machine tool, wherein the main shaft system structure diagram comprises a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes are connected with the associated second nodes through second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to light parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relations among the components in the main shaft system of the target machine tool;
inputting the main shaft system structure diagram into each trained graph rolling neural network model in a model group respectively, and obtaining a main shaft performance prediction result output by each graph rolling neural network model, wherein the model group comprises at least one graph rolling neural network model, each graph rolling neural network model in the model group corresponds to one performance index respectively, and the graph rolling neural network model is obtained based on reinforcement learning training;
and optimizing the lightweight design data of the spindle system based on each spindle performance prediction result.
According to the main shaft system structure optimization method provided by the invention, the main shaft system lightweight design data of the target machine tool comprises a component drawing, a structural parameter value of a component and a connection relation between the components; the main shaft system light design data generation main shaft system structure diagram based on the target machine tool comprises the following components:
inputting the part drawing corresponding to the first node into a trained image feature extraction model, and obtaining drawing features output by the image feature extraction model;
associating the first node with the drawing feature corresponding to the first node;
and associating the second node with the structural parameter value of the component corresponding to the second node.
According to the main shaft system structure optimization method provided by the invention, the training process of the image feature extraction model comprises the following steps:
acquiring a training batch, wherein the training batch comprises a plurality of sample part drawings;
inputting the sample part drawing into the image feature extraction model to obtain sample drawing features output by the image feature extraction model;
determining a training loss based on the first similarity and the second similarity, updating the image feature extraction model based on the training loss;
the first similarity is the similarity between the sample part drawings, and the second similarity is the similarity between the sample drawing features.
According to the main shaft system structure optimization method provided by the invention, the first similarity between the first sample part drawing and the second sample part drawing is determined based on the following steps:
when the first sample component drawing and the second sample component drawing correspond to components which are not of the same type, determining that the first similarity between the first sample component drawing and the second sample component drawing is a first value, wherein the first value is not more than 0;
when the first sample component drawing and the second sample component drawing correspond to the same type of component, respectively extracting a first line characteristic of the first sample component drawing and a second line characteristic of the second sample component drawing, and acquiring the similarity between the first line characteristic and the second line characteristic as the first similarity.
According to the main shaft system structure optimization method provided by the invention, the training process of the graph convolution neural network model corresponding to the first performance index comprises the following steps:
acquiring structural data of a sample spindle system;
generating a sample spindle system configuration map based on the sample spindle system configuration data;
inputting the sample main shaft system structure diagram into the graph rolling neural network model corresponding to the first performance index, and obtaining a sample prediction result of the first performance index output by the graph rolling neural network model corresponding to the first performance index;
obtaining a reward value based on a sample prediction result of the first performance index;
and updating the graph roll-up neural network model corresponding to the first performance index based on the reward value.
According to the method for optimizing the structure of the spindle system provided by the invention, the obtaining of the reward value based on the sample prediction result of the first performance index comprises the following steps:
generating a sample three-dimensional model of a spindle system of a sample machine tool based on the sample spindle system structural data;
performing finite element simulation analysis on the sample three-dimensional model to obtain a performance analysis result;
and obtaining the rewarding value based on the performance analysis result and the sample prediction result of the first performance index.
According to the main shaft system structure optimization method provided by the invention, the training process of the graph convolution neural network model corresponding to the second performance index comprises the following steps:
and training by adopting a transfer learning algorithm based on the trained graph rolling neural network model corresponding to the first performance index to obtain the graph rolling neural network model corresponding to the second performance index.
The invention also provides a main shaft system structure optimizing device, which comprises:
a diagram generating module, configured to generate a main shaft system structural diagram based on main shaft system lightweight design data of a target machine tool, where the main shaft system structural diagram includes a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes and the associated second nodes are connected by second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to lightweight parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relationships between the components in the main shaft system of the target machine tool;
the map processing module is used for respectively inputting the main shaft system structure map into each trained map rolling neural network model in a model group to obtain a main shaft performance prediction result output by each map rolling neural network model, wherein the model group comprises at least one map rolling neural network model, each map rolling neural network model in the model group corresponds to one performance index, and the map rolling neural network model is obtained based on reinforcement learning training;
and the optimization module is used for optimizing the lightweight design data of the spindle system based on the spindle performance prediction results.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the structure optimization method of the spindle system when executing the computer program.
The invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a spindle system configuration optimization method as described in any of the above.
According to the main shaft system structure optimization method, the device, the equipment and the storage medium, the main shaft system light design data of the machine tool are converted into the graph data (main shaft system structural diagram), the components in the main shaft system, the component light parameter types and the relation among the components are expressed in a graph form, the graph data are processed based on the graph convolution neural network, the main shaft performance prediction result is output, the main shaft system light design data are optimized based on the main shaft performance prediction result, the main shaft system light design data and the main shaft performance internal relation are learned by the graph convolution neural network, the main shaft system is predicted, the influence of the change on the main shaft system performance can be quickly obtained when the main shaft system light design data is changed, the test setting is not required to be modified, the test is performed again, and the main shaft system light design efficiency is improved.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a method for optimizing a main shaft system structure;
FIG. 2 is a schematic structural view of a main shaft system structural optimization device provided by the invention;
fig. 3 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the prior art, in the process of designing a spindle system of a machine tool, what performance is required to be tested on the current design data, and then the design data is modified based on the current performance data to optimize the spindle system structure, but a large number of parameters are required to be set for each test (for example, a test based on ANSYS and other traditional finite element analysis), and a long calculation time is required, when the spindle system structure is changed, the parameters are required to be re-recorded and re-calculated, so that the efficiency of designing the spindle system of the machine tool is low. In order to solve the problem, the invention provides a main shaft system structure optimization method, a main shaft system structure optimization device, main shaft system structure optimization equipment and a storage medium, and aims to improve the design efficiency of a machine tool main shaft system.
The spindle system configuration optimization method, apparatus, device and storage medium of the present invention are described below with reference to fig. 1 to 3.
As shown in fig. 1, the method for optimizing the structure of the spindle system provided by the invention comprises the following steps:
s100, generating a main shaft system structural diagram based on main shaft system lightweight design data of a target machine tool, wherein the main shaft system structural diagram comprises a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes are connected with the associated second nodes through the second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to lightweight parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relations among the components in the main shaft system of the target machine tool;
s200, respectively inputting the main shaft system structure diagram into each trained graph rolling neural network model in a model group, and obtaining a main shaft performance prediction result output by each graph rolling neural network model, wherein the model group comprises at least one graph rolling neural network model, each graph rolling neural network model in the model group corresponds to one performance index, and the graph rolling neural network model is obtained based on reinforcement learning training;
and S300, optimizing the lightweight design data of the spindle system based on the spindle performance prediction results.
According to the main shaft system structure optimization method provided by the invention, the main shaft system light design data of the machine tool is converted into the graph data (main shaft system structural diagram), the components in the main shaft system, the component parameter types and the relation among the components are expressed in the form of the graph, the graph data is processed based on the graph convolution neural network, the main shaft performance prediction result is output, and the main shaft system light design data is optimized based on the main shaft performance prediction result.
In particular, the spindle system includes a plurality of components, such as drive shafts, bearings, supports, etc., that are mechanically coupled to each other, and for each component, the structural parameters thereof affect the performance of the overall spindle system, such as the diameter, length, etc., of the drive shafts. The light design data of the spindle system comprises a component drawing of a component in the spindle system, a structural parameter value of the component and a connection relation among the components. Based on the spindle system lightweight design data, a graph including a plurality of nodes and a plurality of edges may be generated as the spindle system structural diagram.
The main shaft system structure diagram comprises a plurality of first nodes, each first node corresponds to one component, each component is provided with a structural parameter affecting performance, the type of the lightweight parameter is taken as a second node, and the types of the structural parameters affecting performance are different according to different types of the components. For example, the first node is a spindle, and the associated second nodes may respectively correspond to a diameter, a length, whether hollow, a wall thickness, and other lightweight parameter types. Each first node is connected with the associated second node through a second edge, and the first nodes with mechanical connection relationship are connected through the first edge. When the main shaft system structure diagram is generated, the weight values corresponding to the first side and the second side are different, the weight values between the second sides corresponding to different connection relations are also different, and the use of the weight values of the different sides can enable the main shaft system structure diagram to reflect the connection relation of components in the main shaft system and the lightweight parameter types of the components influencing performance.
Each first node association reflects a value of a structure of a component corresponding to the first node, and each second node association reflects a parameter value of a lightweight parameter type corresponding to the second node. That is, the spindle system light design data generating spindle system structure diagram based on the target machine tool includes:
inputting the part drawing corresponding to the first node into a trained image feature extraction model, and obtaining drawing features output by the image feature extraction model;
associating the first node with the drawing feature corresponding to the first node;
and associating the second node with the structural parameter value of the component corresponding to the second node.
In the generated main shaft system structure diagram, each node is associated with a corresponding value, and each side is associated with a corresponding weight value. The primary shaft system architecture diagram may be processed by a graph convolutional neural network.
In order to improve the weight of the influence of the change of the lightweight design data of the main shaft on the performance, the lightweight design data in the component drawing can be specially marked, for example, the lines of the lightweight design data (geometric shape parameter information such as the length and the shape of the main shaft, the diameter and the depth of the aperture, the design and the size of the bearing seat and the like, the wall thickness and the structural layout of the main shaft, the internal cavity, the supporting structure, the reinforcing ribs and whether the structural design information such as the hollow structure and the hollow structure is adopted or not) related to the drawing can be processed in a color, a thickening mode and the like, so that the extracted drawing characteristics can reflect the lightweight design data of the main shaft system more.
Specifically, the training process of the image feature extraction model includes:
acquiring a training batch, wherein the training batch comprises a plurality of sample part drawings;
inputting the sample part drawing into the image feature extraction model to obtain sample drawing features output by the image feature extraction model;
determining a training loss based on the first similarity and the second similarity, updating the image feature extraction model based on the training loss;
the first similarity is the similarity between the sample part drawings, and the second similarity is the similarity between the sample drawing features.
If the performance corresponding to the lightweight design data of the spindle system is required to be accurately predicted, the image feature extraction model is required to accurately extract features related to the performance in the component drawing. However, there is no label data for the performance feature in the component drawing, in the method provided by the invention, training loss is determined based on the first similarity and the second similarity to update the image feature extraction model, specifically, the more similar the first similarity and the second similarity, the smaller the training loss, and the more dissimilar the first similarity and the second similarity, the greater the training loss.
The second similarity may be implemented by calculating cosine similarity between features of the sample drawing. The first degree of similarity between the first sample part drawing and the second sample part drawing may be determined based on the steps of:
when the first sample component drawing and the second sample component drawing correspond to components which are not of the same type, determining that the first similarity between the first sample component drawing and the second sample component drawing is a first value, wherein the first value is not more than 0;
when the first sample component drawing and the second sample component drawing correspond to the same type of component, respectively extracting a first line characteristic of the first sample component drawing and a second line characteristic of the second sample component drawing, and acquiring the similarity between the first line characteristic and the second line characteristic as the first similarity.
The similarity between the first line feature and the second line feature is a second value, the second value being greater than 0.
When two parts of the same type are not corresponding to the two parts, the image feature extraction model should have the capability of outputting two dissimilar drawing features, so that the drawing features output by the image feature extraction model can carry information about the type of the parts. When two component drawings correspond to the same type of component, the image feature extraction model needs to output a feature capable of reflecting information such as size, shape and the like of the component, and in order to achieve the purpose, the first similarity is determined based on the similarity of line features between the two component drawings, and an existing line feature extraction mode can be adopted for extracting the line features. The main component of the drawing is a line, the line characteristics of the image can reflect the position, the shape and the like of the line included in the image, the drawing of the sample part is regarded as the image, and the extracted line characteristics can reflect the difference of the drawing.
And determining training loss based on the first similarity and the second similarity, updating the image feature extraction model, and improving the accuracy of drawing features extracted by the image feature extraction model, thereby being beneficial to improving the accuracy of the main shaft performance prediction result.
The performance of the spindle system is not reflected based on only one performance index, but a plurality, e.g., stiffness, strain, etc. In order to realize comprehensive evaluation of the light design data of the spindle system, the method provided by the invention adopts a model group comprising a plurality of graph convolution neural network models to obtain performance prediction results of a plurality of performance indexes. Each of the graph roll-up neural network models corresponds to one performance index, that is, each of the graph roll-up neural network models is used for outputting a performance prediction result corresponding to one performance index. Specifically, the training process of the graph roll-up neural network model corresponding to the first performance index includes:
acquiring structural data of a sample spindle system;
generating a sample spindle system configuration map based on the sample spindle system configuration data;
inputting the sample main shaft system structure diagram into the graph rolling neural network model corresponding to the first performance index, and obtaining a sample prediction result of the first performance index output by the graph rolling neural network model corresponding to the first performance index;
obtaining a reward value based on a sample prediction result of the first performance index;
and updating the graph roll-up neural network model corresponding to the first performance index based on the reward value.
Specifically, the obtaining the prize value based on the sample prediction result of the first performance index includes:
generating a sample three-dimensional model of a spindle system of a sample machine tool based on the sample spindle system structural data;
performing finite element simulation analysis on the sample three-dimensional model to obtain a performance analysis result;
and obtaining the rewarding value based on the performance analysis result and the sample prediction result of the first performance index.
Based on the parameters of finite element simulation analysis of the sample spindle system structure data, a performance analysis result can be obtained, wherein the performance analysis result is an analysis result of the first performance index, the reward value reflects the difference between the performance analysis result and a sample prediction result of the first performance index, and the larger the difference is, the smaller the reward value is.
In the method provided by the invention, in order to reduce the occupation of calculation data, the efficiency of generating the graph convolution neural network model is improved, the cost of implementing the method is reduced, and the training process of training the convolution neural network model corresponding to a plurality of performance indexes by adopting a transfer learning mode, namely, the training process of the graph convolution neural network model corresponding to a second performance index comprises the following steps:
and training by adopting a transfer learning algorithm based on the trained graph rolling neural network model corresponding to the first performance index to obtain the graph rolling neural network model corresponding to the second performance index.
After each spindle performance prediction result is obtained, the spindle performance prediction result can give reference information to a designer, and the designer refers to the spindle performance prediction result to optimize the lightweight design data of the spindle system.
Further, the optimizing the spindle system lightweight design data based on each of the spindle performance predictors includes:
acquiring first difference data between each spindle performance prediction result and corresponding target performance data, wherein each target performance data corresponds to a performance index;
acquiring reference design data, and acquiring each reference performance prediction result based on the reference design data, wherein each reference performance prediction result corresponds to one performance index;
acquiring second difference data between each reference performance prediction result and the corresponding spindle performance prediction result;
and determining optimization direction data based on the first difference data and the second difference data, and optimizing the lightweight design data of the spindle system based on the optimization direction data.
Specifically, the reference design data is generated by adding noise to the spindle system lightweight design data.
After the spindle performance prediction result is obtained, the first difference data between the spindle performance prediction result and target performance data is obtained, wherein the target performance data is an optimization target, i.e. the target performance data reflects the performance which a designer wants to achieve by a spindle system. According to the difference between the performance prediction result of each performance index corresponding to the current lightweight design data of the spindle system and the target data of each performance index, the performance required to be optimized can be determined.
And adding noise to the light-weight design data of the spindle system to obtain reference design data, wherein the noise is obtained by sampling from normal distribution, and obtaining a reference performance prediction result based on the reference design data, namely, generating a new spindle system structure diagram based on the reference design data and inputting the new spindle system structure diagram into each graph convolution neural network model in the model group to obtain the reference performance prediction result corresponding to each performance index. And determining the optimized direction data based on the second difference data between the reference performance prediction result and the spindle performance prediction result, wherein the optimized direction data reflects the influence of noise added when the reference design data is generated on the spindle system performance. And optimizing the light design data of the spindle system based on the optimization direction data.
The spindle system structure optimizing apparatus provided by the invention is described below, and the spindle system structure optimizing apparatus described below and the spindle system structure optimizing method described above can be referred to correspondingly.
As shown in fig. 2, the main shaft system structure optimizing device provided by the present invention includes:
a diagram generating module 210, configured to generate a spindle system structure diagram based on the spindle system lightweight design data of the target machine tool, where the spindle system structure diagram includes a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes and the associated second nodes are connected by a second edge, the first nodes correspond to components in the spindle system of the target machine tool, the second nodes correspond to lightweight parameter types of components in the spindle system of the target machine tool, and the first edges reflect connection relationships between components in the spindle system of the target machine tool;
the graph processing module 220 is configured to input the main shaft system structure graph to each trained graph rolling neural network model in a model group, and obtain a main shaft performance prediction result output by each graph rolling neural network model, where the model group includes at least one graph rolling neural network model, each graph rolling neural network model in the model group corresponds to a performance index, and the graph rolling neural network model is obtained based on reinforcement learning training;
and an optimization module 230, configured to optimize the spindle system lightweight design data based on each of the spindle performance prediction results.
Fig. 3 illustrates a physical schematic diagram of an electronic device, as shown in fig. 3, where the electronic device may include: processor 310, communication interface (Communications Interface) 320, memory 330 and communication bus 340, wherein processor 310, communication interface 320, memory 330 accomplish communication with each other through communication bus 340. Processor 310 may invoke logic instructions in memory 330 to perform a spindle system configuration optimization method comprising: generating a main shaft system structure diagram based on main shaft system light design data of a target machine tool, wherein the main shaft system structure diagram comprises a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes are connected with the associated second nodes through second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to light parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relations among the components in the main shaft system of the target machine tool;
inputting the main shaft system structure diagram into each trained graph rolling neural network model in a model group respectively, and obtaining a main shaft performance prediction result output by each graph rolling neural network model, wherein the model group comprises at least one graph rolling neural network model, each graph rolling neural network model in the model group corresponds to one performance index respectively, and the graph rolling neural network model is obtained based on reinforcement learning training;
and optimizing the lightweight design data of the spindle system based on each spindle performance prediction result.
Further, the logic instructions in the memory 330 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the spindle system configuration optimization method provided by the above methods, the method comprising: generating a main shaft system structure diagram based on main shaft system light design data of a target machine tool, wherein the main shaft system structure diagram comprises a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes are connected with the associated second nodes through second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to light parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relations among the components in the main shaft system of the target machine tool;
inputting the main shaft system structure diagram into each trained graph rolling neural network model in a model group respectively, and obtaining a main shaft performance prediction result output by each graph rolling neural network model, wherein the model group comprises at least one graph rolling neural network model, each graph rolling neural network model in the model group corresponds to one performance index respectively, and the graph rolling neural network model is obtained based on reinforcement learning training;
and optimizing the lightweight design data of the spindle system based on each spindle performance prediction result.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for optimizing the structure of a spindle system, comprising:
generating a main shaft system structure diagram based on main shaft system light design data of a target machine tool, wherein the main shaft system structure diagram comprises a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes are connected with the associated second nodes through second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to light parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relations among the components in the main shaft system of the target machine tool;
inputting the main shaft system structure diagram into each trained graph rolling neural network model in a model group respectively, and obtaining a main shaft performance prediction result output by each graph rolling neural network model, wherein the model group comprises at least one graph rolling neural network model, each graph rolling neural network model in the model group corresponds to one performance index respectively, and the graph rolling neural network model is obtained based on reinforcement learning training;
and optimizing the lightweight design data of the spindle system based on each spindle performance prediction result.
2. The spindle system structural optimization method according to claim 1, wherein the spindle system lightweight design data of the target machine tool includes a component drawing, a structural parameter value of a component, and a connection relationship between components; the main shaft system light design data generation main shaft system structure diagram based on the target machine tool comprises the following components:
inputting the part drawing corresponding to the first node into a trained image feature extraction model, and obtaining drawing features output by the image feature extraction model;
associating the first node with the drawing feature corresponding to the first node;
and associating the second node with the structural parameter value of the component corresponding to the second node.
3. The method of claim 2, wherein the training process of the image feature extraction model comprises:
acquiring a training batch, wherein the training batch comprises a plurality of sample part drawings;
inputting the sample part drawing into the image feature extraction model to obtain sample drawing features output by the image feature extraction model;
determining a training loss based on the first similarity and the second similarity, updating the image feature extraction model based on the training loss;
the first similarity is the similarity between the sample part drawings, and the second similarity is the similarity between the sample drawing features.
4. The spindle system configuration optimization method of claim 3 wherein the first similarity between the first sample part drawing and the second sample part drawing is determined based on the steps of:
when the first sample component drawing and the second sample component drawing correspond to components which are not of the same type, determining that the first similarity between the first sample component drawing and the second sample component drawing is a first value, wherein the first value is not more than 0;
when the first sample component drawing and the second sample component drawing correspond to the same type of component, respectively extracting a first line characteristic of the first sample component drawing and a second line characteristic of the second sample component drawing, and acquiring the similarity between the first line characteristic and the second line characteristic as the first similarity.
5. The method of claim 2, wherein the training process of the graph roll-up neural network model corresponding to the first performance index comprises:
acquiring structural data of a sample spindle system;
generating a sample spindle system configuration map based on the sample spindle system configuration data;
inputting the sample main shaft system structure diagram into the graph rolling neural network model corresponding to the first performance index, and obtaining a sample prediction result of the first performance index output by the graph rolling neural network model corresponding to the first performance index;
obtaining a reward value based on a sample prediction result of the first performance index;
and updating the graph roll-up neural network model corresponding to the first performance index based on the reward value.
6. The method of claim 5, wherein deriving a prize value based on the sample predictions of the first performance metrics comprises:
generating a sample three-dimensional model of a spindle system of a sample machine tool based on the sample spindle system structural data;
performing finite element simulation analysis on the sample three-dimensional model to obtain a performance analysis result;
and obtaining the rewarding value based on the performance analysis result and the sample prediction result of the first performance index.
7. The method of claim 5, wherein the training process of the graph roll-up neural network model corresponding to the second performance index comprises:
and training by adopting a transfer learning algorithm based on the trained graph rolling neural network model corresponding to the first performance index to obtain the graph rolling neural network model corresponding to the second performance index.
8. A spindle system configuration optimizing apparatus, comprising:
a diagram generating module, configured to generate a main shaft system structural diagram based on main shaft system lightweight design data of a target machine tool, where the main shaft system structural diagram includes a plurality of first nodes and a plurality of first edges, each first node is associated with at least one second node, the first nodes and the associated second nodes are connected by second edges, the first nodes correspond to components in the main shaft system of the target machine tool, the second nodes correspond to lightweight parameter types of the components in the main shaft system of the target machine tool, and the first edges reflect connection relationships between the components in the main shaft system of the target machine tool;
the map processing module is used for respectively inputting the main shaft system structure map into each trained map rolling neural network model in a model group to obtain a main shaft performance prediction result output by each map rolling neural network model, wherein the model group comprises at least one map rolling neural network model, each map rolling neural network model in the model group corresponds to one performance index, and the map rolling neural network model is obtained based on reinforcement learning training;
and the optimization module is used for optimizing the lightweight design data of the spindle system based on the spindle performance prediction results.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the spindle system configuration optimization method according to any one of claims 1 to 7 when executing the computer program.
10. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the spindle system configuration optimization method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311478533.6A CN117195440B (en) | 2023-11-08 | 2023-11-08 | Main shaft system structure optimization method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311478533.6A CN117195440B (en) | 2023-11-08 | 2023-11-08 | Main shaft system structure optimization method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117195440A true CN117195440A (en) | 2023-12-08 |
| CN117195440B CN117195440B (en) | 2024-01-30 |
Family
ID=89003851
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311478533.6A Active CN117195440B (en) | 2023-11-08 | 2023-11-08 | Main shaft system structure optimization method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117195440B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111523267A (en) * | 2020-04-21 | 2020-08-11 | 重庆邮电大学 | Fan main shaft structure optimization method based on parameterized finite element model |
| CN111911539A (en) * | 2020-07-30 | 2020-11-10 | 广州汽车集团股份有限公司 | Engine main bearing cover and matching method thereof and engine |
| CN112949004A (en) * | 2021-04-09 | 2021-06-11 | 中国船舶重工集团海装风电股份有限公司 | Lightweight design method of wind generating set bearing seat and bearing seat thereof |
| CN114378653A (en) * | 2022-01-27 | 2022-04-22 | 上海机床厂有限公司 | Cylindrical grinding chatter online identification and monitoring method based on BP neural network |
| CN114549589A (en) * | 2022-03-01 | 2022-05-27 | 昆明理工大学 | Rotating body vibration displacement measurement method and system based on lightweight neural network |
| CN116021339A (en) * | 2023-03-24 | 2023-04-28 | 中科航迈数控软件(深圳)有限公司 | Method and related device for monitoring cutting force of main shaft of numerical control machine tool |
| WO2023142333A1 (en) * | 2022-01-25 | 2023-08-03 | 大连理工大学 | Correction method and system for thin-walled cylindrical shell model |
| US20230308465A1 (en) * | 2023-04-12 | 2023-09-28 | Roobaea Alroobaea | System and method for dnn-based cyber-security using federated learning-based generative adversarial network |
-
2023
- 2023-11-08 CN CN202311478533.6A patent/CN117195440B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111523267A (en) * | 2020-04-21 | 2020-08-11 | 重庆邮电大学 | Fan main shaft structure optimization method based on parameterized finite element model |
| CN111911539A (en) * | 2020-07-30 | 2020-11-10 | 广州汽车集团股份有限公司 | Engine main bearing cover and matching method thereof and engine |
| CN112949004A (en) * | 2021-04-09 | 2021-06-11 | 中国船舶重工集团海装风电股份有限公司 | Lightweight design method of wind generating set bearing seat and bearing seat thereof |
| WO2023142333A1 (en) * | 2022-01-25 | 2023-08-03 | 大连理工大学 | Correction method and system for thin-walled cylindrical shell model |
| CN114378653A (en) * | 2022-01-27 | 2022-04-22 | 上海机床厂有限公司 | Cylindrical grinding chatter online identification and monitoring method based on BP neural network |
| CN114549589A (en) * | 2022-03-01 | 2022-05-27 | 昆明理工大学 | Rotating body vibration displacement measurement method and system based on lightweight neural network |
| CN116021339A (en) * | 2023-03-24 | 2023-04-28 | 中科航迈数控软件(深圳)有限公司 | Method and related device for monitoring cutting force of main shaft of numerical control machine tool |
| US20230308465A1 (en) * | 2023-04-12 | 2023-09-28 | Roobaea Alroobaea | System and method for dnn-based cyber-security using federated learning-based generative adversarial network |
Non-Patent Citations (1)
| Title |
|---|
| 冯鹏升;薄瑞峰;鲁岩;沈兴全;: "基于拓扑优化的TBT深孔钻床主轴箱轻量化设计", 组合机床与自动化加工技术, no. 11, pages 13 - 16 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117195440B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11423311B2 (en) | Automatic tuning of artificial neural networks | |
| CN112487168B (en) | Semantic question-answering method and device of knowledge graph, computer equipment and storage medium | |
| CN110832509B (en) | Black box optimization using neural networks | |
| CN112287670A (en) | Text error correction method, system, computer device and readable storage medium | |
| CN110245364B (en) | Zero-parallel corpus multi-modal neural machine translation method | |
| WO2020140073A1 (en) | Neural architecture search through a graph search space | |
| CN111027292B (en) | Method and system for generating limited sampling text sequence | |
| CN111047563A (en) | Neural network construction method applied to medical ultrasonic image | |
| CN108961460B (en) | Fault prediction method and device based on sparse ESGP (Enterprise service gateway) and multi-objective optimization | |
| CN111241408B (en) | Recommendation model construction system and method | |
| CN114792126A (en) | Convolutional neural network design method based on genetic algorithm | |
| US20240005129A1 (en) | Neural architecture and hardware accelerator search | |
| CN111177135A (en) | Landmark-based data filling method and device | |
| CN117195440B (en) | Main shaft system structure optimization method, device, equipment and storage medium | |
| CN114398949A (en) | Training method of impulse neural network model, storage medium and computing device | |
| CN117933073A (en) | Method and device for exploring design space of CPU micro-architecture | |
| CN117786082A (en) | Generation type online evaluation method and system based on fine tuning large model | |
| CN113408692B (en) | Network structure searching method, device, equipment and storage medium | |
| CN117058079A (en) | Thyroid imaging image automatic diagnosis method based on improved ResNet model | |
| CN115326397B (en) | Method and related device for establishing crankshaft bearing wear degree prediction model and prediction method | |
| CN115035304A (en) | Image description generation method and system based on course learning | |
| CN115587111A (en) | Radix estimation method and device for database query optimization | |
| US11410036B2 (en) | Arithmetic processing apparatus, control method, and non-transitory computer-readable recording medium having stored therein control program | |
| CN112288077B (en) | Diagnostic adjustment method, system, device and medium for convolutional neural network | |
| CN111767980B (en) | Model optimization method, device and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |