A method and system for automatically generating design key points of manufacturing equipment

Abstract

The present invention discloses a method and system for automatically generating design key points for manufacturing equipment. The method includes: collecting and processing historical design requirements and design key point data; establishing a design requirement-design requirement network diagram; establishing a second-order graph convolutional neural network model based on the design requirement-design requirement network diagram, and training it; importing new manufacturing equipment design requirement data, calculating and generating design key points for the design requirements; and outputting a design key point flow chart. The present invention utilizes an intelligent algorithm model to learn the mapping relationship between historical design requirement data and historical design key point data, automatically generates corresponding design key points and their sequence based on the design requirement data, and guides the high-quality completion of mechanical design tasks; through a guided design key point planning scheme, the design difficulty is greatly reduced, and a new method is provided for intelligent auxiliary design in the field of mechanical design.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing equipment design, and in particular to a method and system for automatically generating design key points of manufacturing equipment. Background Art

[0002] In the process of intelligent manufacturing equipment design, design engineers are often limited by their personal experience and intuition when understanding design requirements. There are often problems of incomplete consideration or deviation in understanding, which leads to repeated modifications of the design products. This process not only consumes a lot of time and manpower costs, but also causes problems such as low design efficiency and poor design quality. Therefore, how to automatically guide mechanical design to complete the design project according to the design points involved in the design process and the design requirements, thereby improving design efficiency and design quality, is an urgent problem to be solved in the field of intelligent manufacturing equipment design. Summary of the invention

[0003] In view of the above problems existing in the prior art, the present invention provides a method and system for automatically generating design key points of manufacturing equipment.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for automatically generating design key points of manufacturing equipment, expressing the design process of manufacturing equipment with design key points, and generating the design key points and sequence of the mechanical structure of manufacturing equipment by constructing and training an intelligent algorithm model, the method comprising:

[0006] S1. Collect and process historical design requirements and design key points data;

[0007] S2. Establish design demand-design requirement network diagram;

[0008] S3. Establish a second-order graph convolutional neural network model based on the design requirements-design requirements network diagram and train it;

[0009] S4. Import the design requirement data of new manufacturing equipment, calculate and generate the design points for the design requirements;

[0010] S5. Output the design key points flow chart.

[0011] Furthermore, in step S1, the design requirement data of the processing history is vectorized into a matrix using word embedding technology for the five parts of the manufacturing equipment, namely, the auxiliary actuator, power and motion transmission mechanism, working actuator, working reference and fixing mechanism, and frame appearance structure, according to the four dimensional information of the historical design requirement data, namely, product information, processing technology information, processing action information, and performance requirement information. The vectorized matrix is ​​a four-column matrix, and the four columns of data respectively represent the four dimensional information of the historical design requirement data.

[0012] Furthermore, in step S2, the construction of the design requirement-design requirement network diagram is carried out by calculating the degree of correlation between the design requirements and the design points and the distance matrix of the nodes in the network diagram, and then using a clustering algorithm to classify the nodes to generate a design requirement-design point directed graph and undirected graph composite network diagram.

[0013] Furthermore, the value range of the association degree is between 0 and 1, and the distance matrix of the network diagram is a digital matrix composed of Euclidean distances between nodes. When the association degree between the design requirements and the design key point nodes is less than 0.5 and the Euclidean distance value between the nodes is less than 1, an undirected graph composed of design requirement-design key point nodes is generated, otherwise a directed graph composed of design key point nodes is generated.

[0014] Furthermore, in step S3, the second-order graph convolutional neural network model structure is composed of an input layer, a first-order graph convolution layer, a first-order pooling layer, a second-order graph convolution layer, a second-order pooling layer, a flattening layer, a multi-layer perceptron layer and an output layer.

[0015] Furthermore, the first-order pooling layer is a maximum pooling layer, which is used to aggregate the feature information between design requirements and design points. The second-order pooling layer is an average pooling layer, which is used to aggregate the feature information between design points. The flattening layer processes the second-order feature data into a one-dimensional matrix for nonlinear transformation between data. The multi-layer perceptron layer and the output layer are both fully connected layers.

[0016] Furthermore, in step S3, the first-order graph convolution layer of the second-order graph convolutional neural network uses the graph convolution layer and the attention layer to convolve the undirected graph of the design requirements-design key points network graph, extracts the correlation features of the design requirements-design key points, obtains the first-order feature data of the network graph, and processes the node weights of the network graph; the second-order graph convolution layer uses the graph convolution layer and the attention layer to convolve the directed graph in the design requirements-design key points network graph on the basis of the first-order features, extracts the sequential features of the design points, and obtains the second-order feature data of the network graph.

[0017] Furthermore, the processing of the node weights of the network graph is to reset the weights of the nodes of the network graph whose attention scores of the attention layer are lower than 0.5 to 0.

[0018] Furthermore, in step S5, the design key points flow chart includes design key points and sequence, and is expressed in the form of a diagram.

[0019] The present invention also provides a system for automatically generating design key points of manufacturing equipment, including the above method, wherein the system includes a historical data collection module, a historical data processing module, an initial network diagram construction module, an intelligent calculation module and a storage module.

[0020] The historical data collection module is used to collect historical design requirement data and historical design key point data;

[0021] The historical data processing module is used to express the historical design requirement data and the historical design process data in vector matrix form;

[0022] The initial network diagram construction module is used to construct a design requirement-design key points network diagram;

[0023] The intelligent calculation module is used to automatically generate the content and sequence of design key points according to the new design requirement data;

[0024] The storage module is used to download and save design key points.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention uses an intelligent algorithm model to learn the mapping relationship between historical design requirement data and historical design key point data, automatically generates corresponding design key points and their sequence according to the design requirement data, and guides the completion of mechanical design tasks with high quality. Through the guided design key point planning scheme, the design difficulty is greatly reduced, providing a new idea for intelligent auxiliary design in the field of mechanical design. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the method of the present invention;

[0028] Figure 2 It is a system configuration diagram of the present invention;

[0029] Figure 3 This is an example diagram of the network structure of a second-order graph convolutional neural network;

[0030] Figure 4 This is an example diagram of the output design points and sequence of the second-order graph convolutional neural network model;

[0031] Figure 5 This is an example diagram of the design key flow chart of the mechanical structure of manufacturing equipment. DETAILED DESCRIPTION

[0032] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0033] The present invention provides a method for automatically generating design key points of manufacturing equipment. Figure 1As shown, the manufacturing equipment design process is expressed by design key points, mainly by constructing and training an intelligent algorithm model, and then using the trained intelligent algorithm model to generate the design key points and sequence of the mechanical structure of the manufacturing equipment. The method includes:

[0034] S1. Collect and process historical design requirements and design key points data;

[0035] The design requirement data of the processing history is vectorized and expressed in matrices according to the four dimensions of product information, processing technology information, processing action information and performance requirement information of the historical design requirement data using word embedding technology, targeting the five parts of the manufacturing equipment, namely the auxiliary actuators, power and motion transmission mechanisms, working actuators, working reference and fixing mechanisms, and frame appearance structures.

[0036] The vectorized matrix is ​​a four-column matrix, and the four columns of data respectively represent four dimensional information of the historical design requirement data.

[0037] S2. Establish design demand-design requirement network diagram;

[0038] The construction of the design demand-design requirement network diagram is carried out by calculating the correlation between the design demand and the design key points and the distance matrix of the nodes in the network diagram, and then using a clustering algorithm to classify the nodes to generate a network diagram in the form of a composite of a directed graph and an undirected graph of the design demand-design key points.

[0039] The value range of the association degree is between 0 and 1, and the distance matrix of the network diagram is a digital matrix composed of the Euclidean distances between nodes. When the association degree between the design requirements and the design key point nodes is less than 0.5 and the Euclidean distance value between the nodes is less than 1, an undirected graph composed of design requirements-design key point nodes is generated, otherwise a directed graph composed of design key point nodes is generated.

[0040] S3. Establish a second-order graph convolutional neural network model based on the design requirements-design requirements network diagram and train it;

[0041] The network architecture of the second-order graph convolutional neural network of the scheme is constructed in the order of input layer, first-order graph convolution layer, first-order pooling layer, second-order graph convolution layer, second-order pooling layer, flattening layer, multi-layer perceptron layer and output layer; among which:

[0042] The first-order graph convolution layer uses the graph convolution layer and the attention layer to convolve the undirected graph of the design requirements-design key points network graph, extracts the associated features of the design requirements-design key points, obtains the first-order feature data of the network graph, and processes the node weights of the network graph; the processing of the node weights of the network graph is to reset the weights of the nodes of the network graph whose attention scores of the attention layer are lower than 0.5 to 0;

[0043] The first-order pooling layer is the maximum pooling layer, which is used to aggregate the feature information of design requirements and design points;

[0044] The second-order graph convolution layer of the second-order graph convolutional neural network uses the graph convolution layer and the attention layer to convolve the directed graph in the design requirements-design key points network graph on the basis of the first-order features, extract the sequential features of the design key points, and obtain the second-order feature data of the network graph;

[0045] The second-order pooling layer is an average pooling layer, which is used to aggregate feature information between design points;

[0046] The flattening layer processes the second-order feature data into a one-dimensional matrix for nonlinear transformation between data;

[0047] Both the multi-layer perceptron layer and the output layer are fully connected layers.

[0048] S4. Import the design requirement data of new manufacturing equipment, calculate and generate the design points for the design requirements;

[0049] S5. Output the design key points flow chart;

[0050] The design key points flow chart includes design key points and sequence, and is expressed in the form of a diagram.

[0051] In order to implement the above method, the present invention also provides a Figure 2 The system for automatically generating design key points of manufacturing equipment shown in the figure includes a historical data collection module, a historical data processing module, an initial network diagram construction module, an intelligent calculation module and a storage module.

[0052] The historical data collection module is used to collect historical design requirement data and historical design key point data;

[0053] The historical data processing module is used to express the historical design requirement data and the historical design process data in vector matrix form;

[0054] The initial network diagram construction module is used to construct a design requirement-design key points network diagram;

[0055] The intelligent calculation module is used to automatically generate the content and sequence of design key points according to the new design requirement data;

[0056] The storage module is used to download and save design key points.

[0057] In one embodiment, a method for automatically generating design key points of manufacturing equipment is provided, and the specific steps include:

[0058] 1. Collect and process historical design requirements and design key points data;

[0059] The design requirement information of historical projects is divided into four dimensions: product information, processing technology information, processing action information and performance requirement information. Product information includes product material, product size, roughness, etc. Processing technology information includes production cycle, processing accuracy, equipment body space constraints, working environment, appearance requirements, assembly module space layout, appearance requirements, etc. Processing action information includes action type, action content, action quantity, action range, etc. Performance requirement information includes brand requirements, standard parts selection requirements and other design performance requirements. Then, the word2vec model of word embedding technology is used to vectorize the design requirement data to obtain the design requirement data matrix. ,matrix It is a four-column data matrix, and the four columns of data represent product information, processing technology information, processing action information and performance requirement information respectively. The specific data format is as follows:

[0060]

[0061] The manufacturing equipment design project is divided into five parts: auxiliary actuator, power and motion transmission mechanism, working actuator, working reference and fixing mechanism, and frame appearance structure. For the design process of each part, important design points and their sequence are established and collected. Among them, the design point data of each part is converted from the product information, processing technology information, processing action information and performance requirement information of the historical design requirement data. The word2vec model is used to vectorize the design point data of each part. Here, taking the power and motion transmission mechanism as an example, the design point data matrix The specific forms are as follows:

[0062]

[0063] Among them, after the design key point data is expressed in the vector matrix, each column of data represents a design key point data. The design point data is expressed as ,in For the matrix No. List.

[0064] 2. Establish design requirements-design requirements network diagram;

[0065] The specific process of constructing the design requirements-design requirements network diagram in this case is as follows:

[0066] First, calculate the design requirements data With multiple design point data , , ,… The correlation between ,use right , , ,… For regression, the formula is as follows:

[0067]

[0068] in, Indicates the number and weight of design points The calculation is as follows:

[0069]

[0070]

[0071] in, is the cosine similarity calculation formula, It is the average value of multiple design points.

[0072] Among them, calculation and , , ,… The multiple correlation coefficient between them is calculated as:

[0073]

[0074] in, for The average value of the complex correlation coefficient is positive and its value range is [0,1]. The Euclidean distance between nodes is calculated as:

[0075]

[0076] Constructing the Euclidean distance matrix ,in represents the number of features of a node, Indicates and The characteristics of each node. Finally, the KNN clustering algorithm is used to cluster the nodes to generate a design requirement-design key point network diagram. When the correlation between the design requirement and the design key point node is less than 0.5 and the Euclidean distance between the nodes is less than 1, an undirected graph consisting of the design requirement-design key point nodes is generated, otherwise a directed graph consisting of the design key point nodes is generated. The nodes in the design requirement-design key point network diagram are composed of design requirements and design key points. The design requirement-design key point network diagram is a composite form of a directed graph and an undirected graph. The undirected graph is used to represent the correlation between the design requirements and the design key points, while the directed graph in the network diagram is used to represent the sequential relationship between the design key points.

[0077] 3. Establish a second-order graph convolutional neural network model based on the design requirements-design requirements network diagram and train it;

[0078] The network structure of the second-order graph convolutional neural network in this case is composed of an input layer, a first-order graph convolution layer, a first-order pooling layer, a second-order graph convolution layer, a second-order pooling layer, a flattening layer, a multi-layer perceptron layer, and an output layer. The specific network architecture is as follows: Figure 3 shown.

[0079] In this model, two graph convolution layers implement high-order graph representation, and use maximum pooling and average pooling operations to aggregate node features. The first-order graph convolution layer uses the graph convolution layer and the attention layer to convolve the initial network graph to extract the relationship between design requirements and design points and obtain the first-order features of the network graph. The attention layer scores each node and resets the weights of nodes with an attention score lower than 0.5 to 0. The second-order graph convolution layer uses the graph convolution layer and the attention layer to convolve the first-order features of the network graph again to extract the sequential features between design points and obtain the second-order features of the network graph. The feature vectors of each node and the sequential vectors between nodes are obtained through two graph convolution layer operations. The flattening layer is used to process this two-dimensional matrix into a one-dimensional matrix. The two fully connected layers perform linear combinations and nonlinear transformations on the one-dimensional feature data to extract deeper features of the design requirement-design point data. Finally, the output layer is used to output the content and order of the design points.

[0080] 4. Import the design requirement data of new manufacturing equipment, calculate and generate the design points for the design requirements; the design requirements of the fixture of the new project are as follows:

[0081] (1) Material: 45 steel

[0082] (2) Fixture size range: 3000*5000*2800

[0083] (3) Weight requirement: total weight not exceeding 40kg

[0084] (4) Foolproofing requirements: compatible product foolproofing and symmetrical structure foolproofing

[0085] (5) Automation level: Fully automated

[0086] (6) Actuator: Both cylinder and electric module are available

[0087] (7) Positioning reference plane: The plane parallel to the product flow direction is used as the reference plane.

[0088] (8) Clamping method: Spring clamping

[0089] (9) Accuracy meets: product 2D accuracy requirements

[0090] (10) Consider the overall stability of the structure

[0091] The above design requirement data is classified and processed according to four dimensions: product information, processing technology information, processing action information and performance requirement information, and vectorized using word2vec technology to obtain the design requirement vectorized matrix :

[0092]

[0093] Vectorize the matrix according to the above design requirements Import the second-order graph convolutional neural network model. The key points and sequence of model output design are as follows: Figure 4 shown.

[0094] The key points of profiling design are:

[0095] 1. Yes, consider reserving the position for gripper pick-up and placement

[0096] 2. Whether to consider anti-collision and wear

[0097] 3. Whether the dimensions involved meet the requirements

[0098] 4. Whether to consider the design of details such as appearance beautification and chamfering

[0099] The key points of fool-proof design are:

[0100] 1. Consider compatible product foolproofing and symmetrical structure foolproofing

[0101] The design key points of reference surface positioning are:

[0102] 1. Is the selection of positioning reference surface of parts and fixtures reasonable?

[0103] 2. Is the positioning point design reasonable?

[0104] The key points of automation design are:

[0105] 1. Whether the selected execution structure meets the automation requirements

[0106] The key points of opening and closing method design are:

[0107] 1. The structure is operable

[0108] 2. Convenience of taking and placing

[0109] 3. No interference in picking and placing parts

[0110] 4. Opening and closing without interference

[0111] The key points of fixed structure design are:

[0112] 1. Is the fixed position of the structure reasonable?

[0113] 2. Whether the hole diameter and hole shape are reasonable

[0114] 5. Output design key points flow chart

[0115] According to the output of the algorithm model, the content and order of the design points are sorted out using a computer program to obtain a design point flow chart for the fixture design. The specific form is as follows: Figure 5 shown.

[0116] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for automatically generating design key points of manufacturing equipment, characterized in that: The manufacturing equipment design process is expressed by design key points, and the design key points and sequence of the mechanical structure of the manufacturing equipment are generated by constructing and training an intelligent algorithm model. The method includes: S1. Collect and process historical design requirements and design key points data; S2. Establish design demand-design requirement network diagram; S3. Establish a second-order graph convolutional neural network model based on the design requirements-design requirements network diagram and train it; S4. Import the design requirement data of new manufacturing equipment, calculate and generate the design points for the design requirements; S5. Output the design key points flow chart.

2. The method for automatically generating design key points of manufacturing equipment according to claim 1, characterized in that: In step S1, the design requirement data of the processing history is vectorized and expressed in a matrix according to the four dimensions of product information, processing technology information, processing action information and performance requirement information of the historical design requirement data using word embedding technology for the five parts of the manufacturing equipment, namely, the auxiliary actuator, power and motion transmission mechanism, working actuator, working reference and fixing mechanism, and frame appearance structure. The vectorized matrix is ​​a four-column matrix, and the four columns of data respectively represent the four dimensions of information of the historical design requirement data.

3. The method for automatically generating design key points of manufacturing equipment according to claim 1, characterized in that: In step S2, the construction of the design demand-design requirement network diagram is carried out by calculating the degree of association between the design demand and the design key points and the distance matrix of the nodes in the network diagram, and then using a clustering algorithm to classify the nodes to generate a network diagram in the form of a composite of a directed graph and an undirected graph of the design demand-design key points.

4. The method for automatically generating design key points of manufacturing equipment according to claim 3, characterized in that: The value range of the association degree is between 0 and 1, and the distance matrix of the network diagram is a digital matrix composed of the Euclidean distances between nodes. When the association degree between the design requirements and the design key point nodes is less than 0.5 and the Euclidean distance value between the nodes is less than 1, an undirected graph composed of design requirements-design key point nodes is generated, otherwise a directed graph composed of design key point nodes is generated.

5. The method for automatically generating design key points of manufacturing equipment according to claim 1, characterized in that: In step S3, the second-order graph convolutional neural network model structure is composed of an input layer, a first-order graph convolution layer, a first-order pooling layer, a second-order graph convolution layer, a second-order pooling layer, a flattening layer, a multi-layer perceptron layer and an output layer.

6. The method for automatically generating design key points of manufacturing equipment according to claim 5, characterized in that: The first-order pooling layer is a maximum pooling layer, which is used to aggregate the feature information between design requirements and design points. The second-order pooling layer is an average pooling layer, which is used to aggregate the feature information between design points. The flattening layer processes the second-order feature data into a one-dimensional matrix for nonlinear transformation between data. The multi-layer perceptron layer and the output layer are both fully connected layers.

7. The method for automatically generating design key points of manufacturing equipment according to claim 5, characterized in that: In step S3, the first-order graph convolution layer of the second-order graph convolutional neural network uses the graph convolution layer and the attention layer to convolve the undirected graph of the design requirements-design key points network graph, extracts the associated features of the design requirements-design key points, obtains the first-order feature data of the network graph, and processes the node weights of the network graph; the second-order graph convolution layer uses the graph convolution layer and the attention layer to convolve the directed graph in the design requirements-design key points network graph on the basis of the first-order features, extracts the sequential features of the design points, and obtains the second-order feature data of the network graph.

8. The method for automatically generating design key points of manufacturing equipment according to claim 7, characterized in that: The processing of the node weights of the network graph is to reset the weights of the nodes of the network graph whose attention scores are lower than 0.5 according to the attention layer to 0.

9. The method for automatically generating design key points of manufacturing equipment according to claim 1, characterized in that: In step S5, the design key points flow chart includes design key points and sequence, and is expressed in the form of a diagram.

10. A system for automatically generating design key points of manufacturing equipment, comprising the method according to any one of claims 1 to 9, characterized in that: The system includes a historical data collection module, a historical data processing module, an initial network diagram construction module, an intelligent calculation module and a storage module; The historical data collection module is used to collect historical design requirement data and historical design key point data; The historical data processing module is used to express the historical design requirement data and the historical design process data in vector matrix form; The initial network diagram construction module is used to construct a design requirement-design key points network diagram; The intelligent calculation module is used to automatically generate the content and sequence of design key points according to the new design requirement data; The storage module is used to download and save design key points.

Images