CN114141320A

CN114141320A - Preparation method and system of magnesium-carbon sliding plate and electronic equipment

Info

Publication number: CN114141320A
Application number: CN202111455647.XA
Authority: CN
Inventors: 曹锋; 宋婉嫕; 毛亚平; 孟凡冰; 王丽娜; 金阳; 黄粟
Original assignee: ANSHAN CHOSUN REFRACTORIES CO LTD
Current assignee: ANSHAN CHOSUN REFRACTORIES CO LTD
Priority date: 2021-12-02
Filing date: 2021-12-02
Publication date: 2022-03-04

Abstract

The application relates to the field of intelligent manufacturing, and particularly discloses a preparation method, a system and electronic equipment of a magnesium-carbon sliding plate, wherein under the condition that other components are not changed, physical performance data and a corresponding microscopic structure diagram of the magnesium-carbon sliding plate collected after a certain component is changed are utilized, the physical performance data and the corresponding microscopic structure diagram are processed on the basis of a context encoder model and a convolutional neural network model to obtain characteristic representation of the component, meanwhile, an adjacency matrix is constructed through logical operation among rules to represent the incidence relation among the components, and the incidence characterization vector of each node containing incidence information is further obtained by using the technology of a neural network to improve the accuracy of decoding regression. Therefore, the optimal proportion of each component in the manufacturing material of the magnesium-carbon sliding plate is determined by an artificial intelligence algorithm, so that the experiment times and cost are reduced, and the finally manufactured magnesium-carbon sliding plate has better physical and chemical properties.

Description

Preparation method and system of magnesium-carbon sliding plate and electronic equipment

Technical Field

The invention relates to the field of intelligent manufacturing design, in particular to a preparation method and system of a magnesium-carbon sliding plate and electronic equipment.

Background

In the modern steel smelting process, particularly with the development of rapid and efficient continuous casting and secondary refining technologies and processes, a sliding gate nozzle system becomes more and more important and becomes an indispensable part in smelting. The sliding gate system is a molten steel control device in the casting process of a continuous casting machine, and can accurately adjust the flow of molten steel from a ladle to a tundish and from the tundish to a crystallizer, so that the inflow and outflow of molten steel are balanced, and the continuous casting operation is easier to control.

The slide plate is the most important component of the sliding gate system, and as the slide plate directly controls the flow of molten steel, the slide plate needs to repeatedly bear the chemical erosion and physical scouring of high-temperature molten steel for a long time and has violent and transient thermal shock and mechanical abrasion effects under the condition of meeting different pouring process conditions, and the use conditions are extremely severe. The improvement of the performance of the sliding plate has been one of the key points of the research on the refractory materials.

In order to remove the nozzle clogging caused by the aluminum deoxidation product, the molten steel needs to be subjected to calcium treatment, but the concentrations of Ca in the molten steel and CaO in the slag water increase, and they react with the alumina component in the sliding plate to generate low-melting-point substances, thereby causing damage to the sliding plate. Due to the excellent high temperature resistance and corrosion resistance of the magnesium oxide, the magnesium oxide can be used for manufacturing a sliding plate for casting calcium-treated steel, but the sliding plate is easy to damage by thermal shock, so that the poor thermal shock stability of the magnesium-carbon sliding plate can be improved to a certain extent by using carbon and carbide with good thermal conductivity or magnesium aluminate spinel as a binding phase.

At present, a large amount of metal aluminum powder is introduced into a magnesium-carbon sliding plate to generate metal combination, so that the thermal shock resistance and the strength of the magnesium-carbon sliding plate are improved. However, the introduction of a large amount of metal aluminum powder may aggravate the aluminothermic reduction reaction of magnesium oxide, so that a large amount of spinel may expand, and therefore, the changes of the structure and the performance of the magnesium-carbon material with high metal aluminum content after being treated at different temperatures need to be researched, and the specific content is the influence of the respective addition amounts of the metal aluminum powder, magnesia, flake graphite, sulfide powder, carbide powder and the like on the structure and the performance of the magnesium-carbon material treated at different temperatures in the air.

The optimal proportion of each component in the preparation formula of the magnesium-carbon skateboard is determined based on an experimental mode, but on one hand, the traditional experiment adopts a controlled variable method, which ignores the relevance among the components. Meanwhile, the number of times of experiments is limited in consideration of the cost of the experiments, and the experiments can only obtain some rules in a fuzzy manner, so that the optimal proportion of each component cannot be accurately deduced. Therefore, in order to determine the optimal ratio of each component in the material for manufacturing the magnesium-carbon sliding plate so as to improve the performance of the magnesium-carbon material, a method for manufacturing the magnesium-carbon sliding plate is desired.

Disclosure of Invention

The present application is proposed to solve the above-mentioned technical problems. The embodiment of the application provides a preparation method, a system and electronic equipment of a magnesium-carbon skateboard, which utilize physical property data of the magnesium-carbon skateboard collected after changing a certain component under the condition that other components are not changed, and a microscopic structure diagram of the magnesium-carbon skateboard, then process the physical property data and the microscopic structure diagram of the magnesium-carbon skateboard based on a context encoder model and a convolutional neural network model to obtain characteristic representation of the components, meanwhile, an adjacency matrix is constructed through logical operation among rules to represent the incidence relation among the components, and the technology of a neural network is further used to obtain the incidence characterization vector of each node containing the incidence information, so that the accuracy of decoding regression is improved. Thus, the optimal proportion of each component in the material for manufacturing the magnesium-carbon sliding plate can be determined, and the performance of the magnesium-carbon material can be improved.

According to one aspect of the present application, there is provided a method for manufacturing a magnesium-carbon skateboard, comprising:

acquiring optimal physical property data and a microscopic structure diagram corresponding to each component in a material for manufacturing the magnesium-carbon sliding plate, wherein the material for manufacturing the magnesium-carbon sliding plate comprises magnesia, scale graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, vulcanized powder, carbonized powder, thermosetting phenolic resin and carbores T60, and the physical property data comprises linear change rate, weight change rate, volume density, apparent porosity and normal-temperature compressive strength;

for the optimal physical performance data corresponding to each component, enabling each physical performance data to pass through a word embedding layer to obtain a plurality of parameter input vectors;

inputting the microscopic structure chart corresponding to each component into a first convolution neural network serving as a filter to obtain a structure input vector;

passing the plurality of parametric input vectors and the structural input vector through a context-based coding encoder to obtain a sequence of coded feature vectors;

all the coding feature vectors in the sequence of the coding feature vectors are cascaded to obtain feature representation vectors corresponding to all the components;

constructing a conjunction adjacency matrix and a disjunction adjacency matrix among feature expression vectors corresponding to each component based on a conjunction and disjunction logical operation rule among the components in the material for manufacturing the magnesium-carbon sliding plate, wherein the conjunction logical operation rule expresses the parallel relation among the rules, and the disjunction logical operation expresses the replacement relation among the rules;

passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix, wherein the first feature matrix and the conjunctive adjacency matrix have the same scale, and the second feature matrix and the disjunctive adjacency matrix have the same scale;

calculating the position points between the first characteristic matrix and the second characteristic matrix to obtain a fused similar adjacency matrix;

performing two-dimensional splicing on the feature expression vectors corresponding to the components to obtain a component feature matrix;

passing the component feature matrix and the class adjacency matrix through a graph neural network to obtain a graph structure feature matrix, each row of the graph structure feature matrix being associated with an associated characterization vector of a corresponding component containing associated information, wherein the graph neural network is adapted to process graph data in an irregular non-Euclidean space to extract associated information of a data sample due to the feature information and the irregular topological structure information; and

the associated characterization vector for each component is passed through a decoder for regression to obtain the optimal ratio for that component.

In the above method for preparing a magnesium-carbon skateboard, inputting the microstructure diagram corresponding to each component into a first convolution neural network as a filter to obtain a structural input vector, including: each layer except the last layer of the first convolution neural network is used for carrying out convolution processing on the microscopic structure diagram to obtain a structural feature diagram; and the last layer of the first convolutional neural network performs feature matrix-based global average pooling on the structural feature map to obtain the structural input vector.

In the above method for manufacturing a magnesium-carbon skateboard, passing the plurality of parameter input vectors and the structure input vector through a context-based coding encoder to obtain a sequence of coded feature vectors, the method includes: passing the plurality of parametric input vectors and the structural input vector through a converter model of the encoder to obtain the sequence of encoded feature vectors.

In the above method for manufacturing a magnesium-carbon sliding plate, constructing a conjunctive adjacency matrix and a disjunctive adjacency matrix between the feature expression vectors corresponding to the components based on a logical operation rule of conjunctive and disjunctive between the components in a material of the magnesium-carbon sliding plate, the method includes: constructing a conjunction adjacency matrix between characteristic expression vectors corresponding to all components based on a conjunction logic operation rule among all components in a manufacturing material of the magnesium-carbon sliding plate according to the following formula;

wherein the formula is:

wherein the content of the first and second substances,

is a conjunctive adjacency matrix, which is used to indicate that the matrix position when a corresponding pair of rules form conjunctive normal forms takes 1, and the matrix position when the non-conjunctive normal forms takes 0.

In the preparation method of the magnesium-carbon sliding plate, an extracted adjacency matrix between characteristic expression vectors corresponding to each component is constructed according to the following formula based on the logic operation rule of extraction between each component in the material for manufacturing the magnesium-carbon sliding plate;

wherein the formula is:

wherein the content of the first and second substances,

is a disjunctive adjacency matrix that indicates that a corresponding pair of rules form a disjunctive normal form with matrix positions taking 1 and non-disjunctive normal form with matrix positions taking 0.

In the above method for manufacturing a magnesium-carbon skateboard, passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix, the method includes: each layer of the second convolutional neural network except the last layer explicitly spatially encodes the conjunctive adjacency matrix and the disjunctive adjacency matrix to obtain a first feature map and a second feature map according to the following formula:

f_i＝active(N_i×f_i-1+B_i)

wherein f is_i-1As input to the ith layer of the second convolutional neural network, f_iIs the output of the ith layer of the second convolutional neural network, N_iA filter of the second convolutional neural network of the ith layer, and B_iActive represents a nonlinear activation function for a bias matrix of the ith layer of the second convolutional neural network; and the last layer of the second convolutional neural network performs global pooling processing on the first feature map and the second feature map along a channel dimension to obtain the first feature matrix and the second feature matrix.

In the above method for manufacturing a magnesium-carbon skateboard, further comprising: setting other label values for indicating upper and lower limits of a ratio range centering on the optimum ratio of the certain component; passing the associated token vector for each component through the decoder to obtain probabilities that the associated token vector belongs to other tag values; and determining whether other tag values are feasible based on the probability.

According to another aspect of the present application, there is provided a system for preparing a magnesium carbon skateboard, comprising:

the system comprises a data acquisition unit, a data acquisition unit and a data processing unit, wherein the data acquisition unit is used for acquiring optimal physical property data and a microscopic structure diagram corresponding to each component in a manufactured material of the magnesium-carbon sliding plate, the manufactured material of the magnesium-carbon sliding plate comprises magnesia, scale graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, sulfide powder, carbide powder, thermosetting phenolic resin and carbores T60, and the physical property data comprises a linear change rate, a weight change rate, a volume density, an apparent porosity and a normal-temperature compressive strength;

the word embedding unit is used for enabling the optimal physical performance data corresponding to the components obtained by the data obtaining unit to pass through a word embedding layer so as to obtain a plurality of parameter input vectors;

a first convolution unit, configured to input the microscopic structure diagram corresponding to each component obtained by the data obtaining unit into a first convolution neural network serving as a filter to obtain a structure input vector;

an encoding unit configured to pass the plurality of parameter input vectors obtained by the word embedding unit and the structure input vector obtained by the first convolution unit through a context-based encoding encoder to obtain a sequence of encoded feature vectors;

the cascade unit is used for cascading all the coding feature vectors in the sequence of the coding feature vectors obtained by the coding unit to obtain feature representation vectors corresponding to all the components;

the adjacency matrix construction unit is used for constructing a conjunction adjacency matrix and a disjunction adjacency matrix between the feature expression vectors corresponding to all the components obtained by the cascade unit based on the conjunction and disjunction logical operation rules among all the components in the material for manufacturing the magnesium-carbon sliding plate, wherein the conjunction logical operation rules express the parallel relation among the rules, and the disjunction logical operation expresses the replacement relation among the rules;

a second convolution unit, configured to pass the conjunctive adjacency matrix obtained by the adjacency matrix construction unit and the disjunctive adjacency matrix obtained by the adjacency matrix construction unit through a second convolution neural network to obtain a first feature matrix and a second feature matrix, where the first feature matrix and the conjunctive adjacency matrix have the same scale, and the second feature matrix and the disjunctive adjacency matrix have the same scale;

a fusion unit, configured to calculate a position-point-based addition between the first feature matrix obtained by the second convolution unit and the second feature matrix obtained by the second convolution unit to obtain a fused similar adjacency matrix;

the two-dimensional splicing unit is used for performing two-dimensional splicing on the feature expression vectors corresponding to the components obtained by the cascading unit to obtain a component feature matrix;

the graph neural network unit is used for enabling the component feature matrix obtained by the two-dimensional splicing unit and the similar adjacency matrix obtained by the fusion unit to pass through a graph neural network to obtain a graph structure feature matrix, each action of the graph structure feature matrix is an associated characterization vector containing associated information of a corresponding component, and the graph neural network is suitable for processing graph data in irregular non-Euclidean air to extract associated information of a data sample due to feature information and irregular topological structure information; and

and the regression unit is used for enabling the associated characterization vector of each component obtained by the graph neural network unit to pass through a decoder for regression so as to obtain the optimal proportion of the component.

In the above system for preparing a magnesium-carbon skateboard, the first rolling unit is further configured to: each layer except the last layer of the first convolution neural network is used for carrying out convolution processing on the microscopic structure diagram to obtain a structural feature diagram; and the last layer of the first convolutional neural network performs feature matrix-based global average pooling on the structural feature map to obtain the structural input vector.

In the above system for preparing a magnesium-carbon skateboard, the encoding unit is further configured to: passing the plurality of parametric input vectors and the structural input vector through a converter model of the encoder to obtain the sequence of encoded feature vectors.

In the above system for manufacturing a magnesium-carbon skateboard, the adjacency matrix construction unit is further configured to: constructing a conjunction adjacency matrix between characteristic expression vectors corresponding to all components based on a conjunction logic operation rule among all components in a manufacturing material of the magnesium-carbon sliding plate according to the following formula;

wherein the formula is:

wherein the content of the first and second substances,

In the preparation system of the magnesium-carbon sliding plate, an extracted adjacency matrix between the characteristic expression vectors corresponding to each component is constructed according to the following formula based on the logic operation rule of extraction between each component in the material for manufacturing the magnesium-carbon sliding plate;

wherein the formula is:

wherein the content of the first and second substances,

In the above system for preparing a magnesium-carbon skateboard, the second convolution unit is further configured to: each layer of the second convolutional neural network except the last layer explicitly spatially encodes the conjunctive adjacency matrix and the disjunctive adjacency matrix to obtain a first feature map and a second feature map according to the following formula:

f_i＝active(N_i×f_i-1+B_i)

In the above preparation system of magnesium carbon skateboard, further comprising: setting other label values for indicating upper and lower limits of a ratio range centering on the optimum ratio of the certain component; passing the associated token vector for each component through the decoder to obtain probabilities that the associated token vector belongs to other tag values; and determining whether other tag values are feasible based on the probability.

According to yet another aspect of the present application, there is provided an electronic device including: a processor; and a memory in which computer program instructions are stored, which, when executed by the processor, cause the processor to perform the method of manufacturing a magnesium-carbon skateboard as described above.

According to yet another aspect of the present application, there is provided a computer readable medium having stored thereon computer program instructions, which, when executed by a processor, cause the processor to perform the method of manufacturing a magnesium-carbon skateboard as described above.

Compared with the prior art, the preparation method, the system and the electronic equipment of the magnesium-carbon skateboard provided by the application have the advantages that the physical property data of the magnesium-carbon skateboard, which are acquired after a certain component is changed under the condition that other components are not changed, and the microscopic structure diagram of the magnesium-carbon skateboard are utilized, the physical property data and the microscopic structure diagram of the magnesium-carbon skateboard are processed on the basis of a context encoder model and a convolutional neural network model to obtain the characteristic representation of the component, meanwhile, an adjacency matrix is constructed through logical operation among rules to represent the association relation among the components, and the association characterization vector of each node containing association information is further obtained by using the technology of a neural network to improve the accuracy of decoding regression. Thus, the optimal proportion of each component in the material for manufacturing the magnesium-carbon sliding plate can be determined, and the performance of the magnesium-carbon material can be improved.

Drawings

The above and other objects, features and advantages of the present application will become more apparent by describing in more detail embodiments of the present application with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings, like reference numbers generally represent like parts or steps.

FIG. 1 is a flow chart of a method of making a magnesium carbon skateboard according to an embodiment of the present application;

FIG. 2 is a schematic diagram of a system architecture of a method for manufacturing a magnesium-carbon skateboard according to an embodiment of the present application;

FIG. 3 is a block diagram of a system for making a magnesium carbon skillet according to an embodiment of the application;

fig. 4 is a block diagram of an electronic device according to an embodiment of the application.

Detailed Description

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and that the present application is not limited by the example embodiments described herein.

Overview of a scene

As described above, in order to remove the nozzle clogging caused by the aluminum deoxidation product, the molten steel needs to be subjected to calcium treatment, but Ca in the molten steel and CaO in the slag water increase in concentration, and they react with the alumina component in the slide plate to generate low melting point substances, thereby causing damage to the slide plate. Due to the excellent high temperature resistance and corrosion resistance of the magnesium oxide, the magnesium oxide can be used for manufacturing a sliding plate for casting calcium-treated steel, but the sliding plate is easy to damage by thermal shock, so that the poor thermal shock stability of the magnesium-carbon sliding plate can be improved to a certain extent by using carbon and carbide with good thermal conductivity or magnesium aluminate spinel as a binding phase.

Accordingly, in the technical solution of the present application, if the optimal ratio is determined by the control variable method or the classification result of the ratio adjustment of a certain component is made, it is obviously impractical, and there is an influence of each component rather than an independent relationship, so the ratio adjustment result or the optimal ratio result can be generated based on the neural network model.

Specifically, in the technical solution of the present application, the experimental data is physical property data of the magnesium-carbon sliding plate collected after changing a certain component under the condition that other components are not changed (including type and weight), that is, a linear change rate, a weight change rate, a volume density, an apparent porosity, and a normal temperature compressive strength, and a microscopic structure diagram of the magnesium-carbon sliding plate. However, since the components are not independent but have a relationship with each other, how to obtain the relationship information between the components becomes critical.

Based on this, the applicant of the present application considers that through the data corresponding to the above-mentioned single components, the feature representation of the component can be obtained through the deep neural network model, so if an adjacency matrix for representing the association relationship between the components can be further obtained, the association characterization vector of each node containing the association information can be obtained by using the technology of the neural network.

Here, it is apparent that there is no node topology information represented by nodes between the respective components, and thus a general adjacency matrix cannot be used, but considering that the respective components can be actually expressed in a regular form, for example, the formulation ratio of the large crystal magnesite 98 of the specification 3 to 1mm is 25% to 32%, it is possible to construct the adjacency matrix by logical operations between the rules.

In particular, the logical operations between rules usually include conjunctions and disjunctions, represented by the symbols Λ and v, respectively, for representing juxtapositions or alternative relationships between the rules, i.e. the meaning of "and" or ". For the components as described above, there is also such a relationship that, for example, the relationship that "the formulation ratio of large crystalline magnesite 98 of 3 to 1mm in specification is 25% to 32%" and the formulation ratio of large crystalline magnesite 98 of 200 mesh in specification is 24% to 28% "is" or ", and the relationship that" the formulation ratio of large crystalline magnesite 98 of 3 to 1mm in specification is 25% to 32% "and the formulation ratio of crystalline flake graphite 98 of 100 mesh in specification is 1% to 1.8" is "and", therefore, discrete logical nodes can be constructed for the rules satisfied for the respective components based on the above method to form an adjacent matrix, which are respectively expressed as:

here, the first and second liquid crystal display panels are,

is a conjunctive adjacency matrix, which is used to indicate that the matrix position when a corresponding pair of rules form conjunctive normal forms takes 1, and the matrix position when the non-conjunctive normal forms takes 0. While

Then, the conjunctive adjacency matrix and the disjunctive adjacency matrix are mapped into a high-dimensional space through a convolutional neural network respectively and are fused, and a similar adjacency matrix corresponding to the feature representation of each node can be obtained.

Therefore, in the technical solution of the present application, first, the optimal physical property data of a single component, that is, the optimal physical property data obtained by changing the proportion of a certain component (including type and weight) under the condition that other components are fixed and unchanged, for example, the optimal physical property data includes linear change rate, weight change rate, volume density, apparent porosity and room temperature compressive strength, and a corresponding microscopic structure diagram. And in order to express the association relation between the parameters, a context encoder model is used for carrying out context encoding, namely, each physical parameter is used for obtaining a parameter input vector in a word embedding mode, a microscopic structure chart is used for obtaining a structure input vector through a first convolution neural network used as a filter, then the parameter input vector and the structure input vector are input into an encoder together to obtain a sequence of encoding characteristic vectors, and each encoding characteristic vector is cascaded, so that a characteristic expression vector of a single component is obtained.

And then, obtaining a first characteristic matrix and a second characteristic matrix by the combined adjacency matrix and the disjointed adjacency matrix through a second convolutional neural network according to the mode, wherein the dimension of the characteristic matrix is the same as that of the original input matrix, and then calculating the points of the first characteristic matrix and the second characteristic matrix to obtain a fused similar adjacency matrix.

And then, performing two-dimensional splicing on the feature representation vectors of each component to obtain a component feature matrix, and obtaining a graph structure feature matrix through a graph neural network together with the class adjacency matrix, wherein each row of the graph structure feature matrix is the associated representation vector containing the associated information of the corresponding component. Here, the graph neural network can be used for processing graph data in an irregular non-euclidean space, so as to extract associated information of a data sample due to feature information and irregular topological structure information, and therefore, compared with an obtained feature representation vector obtained by directly splicing individual components, the obtained feature representation vector can improve the accuracy of decoding regression.

Finally, the associated characterization vector for each component is passed through a decoder for regression to obtain the optimal proportions of the components. In addition, after the optimal proportion of the component is obtained, other label values for representing the upper and lower limits of the proportion range can be set by taking the optimal proportion as the center, for example, 27%, 28%, 29%, 31%, 32% and 33% of the label values with the optimal proportion of 30% can be set, then the probability that the other label values belong is obtained through decoding regression, for example, the label probability of 28%, 29%, 31% and 32% is greater than the preset threshold value, and then the feasible proportion range of 28% -32% is obtained.

Based on this, the application provides a preparation method of a magnesium-carbon sliding plate, which comprises the following steps: acquiring optimal physical property data and a microscopic structure diagram corresponding to each component in a material for manufacturing the magnesium-carbon sliding plate, wherein the material for manufacturing the magnesium-carbon sliding plate comprises magnesia, crystalline flake graphite, carboresp, spherical asphalt, aluminum-silicon alloy powder, vulcanized powder, carbonized powder, thermosetting phenolic resin and carboresp 60, and the physical property data comprises a linear change rate, a weight change rate, a volume density, an apparent porosity and a normal-temperature compressive strength; for the optimal physical performance data corresponding to each component, enabling each physical performance data to pass through a word embedding layer to obtain a plurality of parameter input vectors; inputting the microscopic structure chart corresponding to each component into a first convolution neural network serving as a filter to obtain a structure input vector; passing the plurality of parametric input vectors and the structural input vector through a context-based coding encoder to obtain a sequence of coded feature vectors; all the coding feature vectors in the sequence of the coding feature vectors are cascaded to obtain feature representation vectors corresponding to all the components; constructing a conjunction adjacency matrix and a disjunction adjacency matrix among feature expression vectors corresponding to each component based on a conjunction and disjunction logical operation rule among the components in the material for manufacturing the magnesium-carbon sliding plate, wherein the conjunction logical operation rule expresses the parallel relation among the rules, and the disjunction logical operation expresses the replacement relation among the rules; passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix, wherein the first feature matrix and the conjunctive adjacency matrix have the same scale, and the second feature matrix and the disjunctive adjacency matrix have the same scale; calculating the position points between the first characteristic matrix and the second characteristic matrix to obtain a fused similar adjacency matrix; performing two-dimensional splicing on the feature expression vectors corresponding to the components to obtain a component feature matrix; passing the component feature matrix and the class adjacency matrix through a graph neural network to obtain a graph structure feature matrix, each row of the graph structure feature matrix being associated with an associated characterization vector of a corresponding component containing associated information, wherein the graph neural network is adapted to process graph data in an irregular non-Euclidean space to extract associated information of a data sample due to the feature information and the irregular topological structure information; and passing the associated characterization vector for each component through a decoder for regression to obtain the optimal ratio for that component.

Fig. 1 is a diagram illustrating an application scenario of a preparation method of a magnesium-carbon skateboard according to an embodiment of the application. As shown in fig. 1, in this application scenario, first, optimal physical property data corresponding to each component is obtained from a material (e.g., M as illustrated in fig. 1) of a magnesium-carbon sliding plate, and a microstructure diagram corresponding thereto is obtained by an electron microscope camera (e.g., C as illustrated in fig. 1), wherein the material of the magnesium-carbon sliding plate includes magnesite, flake graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, vulcanized powder, carbonized powder, thermosetting phenol resin, and carbores T60, and the physical property data includes a linear change rate, a weight change rate, a volume density, an apparent porosity, and an ambient pressure strength. The obtained physical property data and the microstructure map are then input into a server (e.g., S as illustrated in fig. 1) that is deployed with a magnesium carbon sled manufacturing algorithm, wherein the server can process the physical property data and the microstructure map with the magnesium carbon sled manufacturing algorithm to obtain the optimal proportions of the components. Furthermore, other label values for representing the upper and lower limits of the scale range may be set around the optimal ratio, for example, 27%, 28%, 29%, 31%, 32% and 33% of the label values with the optimal ratio of 30% may be set, and then the probability that the other label values belong is obtained by decoding regression, for example, the label probability of 28%, 29%, 31% and 32% is greater than a predetermined threshold, so that the feasible scale range of 28% -32% is obtained. Thus, the performance of the magnesium-carbon material can be improved by adding the components in the optimal proportion range.

Having described the general principles of the present application, various non-limiting embodiments of the present application will now be described with reference to the accompanying drawings.

Exemplary method

Fig. 1 illustrates a flow chart of a method of manufacturing a magnesium carbon skateboard. As shown in fig. 1, a method for preparing a magnesium-carbon skateboard according to an embodiment of the present application includes: s110, obtaining optimal physical property data and a microscopic structure diagram corresponding to each component in a manufactured material of the magnesium-carbon sliding plate, wherein the manufactured material of the magnesium-carbon sliding plate comprises magnesia, scale graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, vulcanized powder, carbonized powder, thermosetting phenolic resin and carbores T60, and the physical property data comprises a linear change rate, a weight change rate, a volume density, an apparent porosity and a normal-temperature compressive strength; s120, for the optimal physical performance data corresponding to each component, enabling each physical performance data to pass through a word embedding layer to obtain a plurality of parameter input vectors; s130, inputting the microscopic structure chart corresponding to each component into a first convolution neural network serving as a filter to obtain a structure input vector; s140, passing the plurality of parameter input vectors and the structure input vector through a context-based coding encoder to obtain a sequence of coded feature vectors; s150, all the coding feature vectors in the sequence of the coding feature vectors are cascaded to obtain feature representation vectors corresponding to all the components; s160, constructing a conjunction adjacency matrix and a disjunction adjacency matrix among the feature expression vectors corresponding to all the components based on the conjunction and disjunction logical operation rules among all the components in the material for manufacturing the magnesium-carbon sliding plate, wherein the conjunction logical operation rules express the parallel relation among the rules, and the disjunction logical operation expresses the replacement relation among the rules; s170, passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix, wherein the first feature matrix and the conjunctive adjacency matrix have the same scale, and the second feature matrix and the disjunctive adjacency matrix have the same scale; s180, calculating the position points between the first feature matrix and the second feature matrix to obtain a fused similar adjacency matrix; s190, performing two-dimensional splicing on the feature expression vectors corresponding to the components to obtain a component feature matrix; s200, passing the component feature matrix and the similar adjacency matrix through a graph neural network to obtain a graph structure feature matrix, wherein each row of the graph structure feature matrix is an associated characterization vector containing associated information of a corresponding component, and the graph neural network is suitable for processing graph data in irregular non-Euclidean air to extract associated information of a data sample due to feature information and irregular topological structure information; and S210, passing the associated characterization vector of each component through a decoder for regression to obtain the optimal proportion of the component.

Fig. 2 illustrates an architecture diagram of a preparation method of a magnesium-carbon skateboard according to an embodiment of the present application. As shown in fig. 2, in the network architecture of the preparation method of the magnesium-carbon skateboard, firstly, each obtained physical performance data (e.g., P1 as illustrated in fig. 2) is passed through a word embedding layer (e.g., WEL as illustrated in fig. 2) to obtain a plurality of parameter input vectors (e.g., V1 as illustrated in fig. 2); then, inputting the obtained microscopic structure diagram (e.g., P2 as illustrated in fig. 2) corresponding to the respective components into a first convolutional neural network (e.g., CNN1 as illustrated in fig. 2) as a filter to obtain a structure input vector (e.g., V2 as illustrated in fig. 2); then, passing the plurality of parametric input vectors and the structural input vector through a context-based encoding encoder (e.g., E as illustrated in fig. 2) to obtain a sequence of encoded feature vectors (e.g., VF1 as illustrated in fig. 2); then, all the encoding feature vectors in the sequence of encoding feature vectors are concatenated to obtain feature representation vectors corresponding to the respective components (e.g., VF2 as illustrated in fig. 2); then, constructing a conjunctive adjacency matrix (for example, M1 as illustrated in FIG. 2) and a disjunctive adjacency matrix (for example, M2 as illustrated in FIG. 2) between the feature representation vectors corresponding to the components based on the logical operation rules of conjunction and disjunctive between the components in the material of the magnesium-carbon skateboard; then, passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network (e.g., CNN2 as illustrated in fig. 2) to obtain a first feature matrix (e.g., MF1 as illustrated in fig. 2) and a second feature matrix (e.g., MF2 as illustrated in fig. 2); then, calculating a position-point-wise adjacency-like matrix (e.g., MF as illustrated in fig. 2) between the first feature matrix and the second feature matrix; then, performing two-dimensional splicing on the feature representation vectors corresponding to the components to obtain a component feature matrix (e.g., MFC as illustrated in fig. 2); then, passing the component feature matrix and the adjacency-like matrix through a graph neural network (e.g., GNN as illustrated in fig. 2) to obtain a graph structure feature matrix (e.g., MFF as illustrated in fig. 2), each row of the graph structure feature matrix including associated characterization vectors containing associated information for the corresponding component; and, finally, passing the associated token vector for each component through a decoder for regression (e.g., D as illustrated in fig. 2) to obtain the optimal proportions for that component.

In step S110, optimal physical property data and a microstructure diagram corresponding to each component in a material for manufacturing the magnesium-carbon sliding plate are obtained, wherein the material for manufacturing the magnesium-carbon sliding plate comprises magnesia, flake graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, sulfide powder, carbide powder, thermosetting phenolic resin and carbores T60, and the physical property data comprises a linear change rate, a weight change rate, a volume density, an apparent porosity and a normal temperature compressive strength. As described above, in order to determine the optimal ratio of each component in the material for manufacturing the magnesium-carbon sliding plate, so as to improve the performance of the magnesium-carbon material, in the technical solution of the present application, it is first required to obtain the optimal physical property data and the microstructure diagram corresponding to each component in the material for manufacturing the magnesium-carbon sliding plate, that is, the optimal physical property data obtained by changing the ratio of a certain component (including type and weight) under the condition that other components are fixed and unchanged, for example, the optimal physical property data includes a linear change rate, a weight change rate, a volume density, an apparent porosity, a room temperature compressive strength, and a corresponding microstructure diagram. Here, the magnesia carbon slide board is made of materials including magnesia, flake graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, vulcanized powder, carbonized powder, thermosetting phenol resin, and carbores T60.

In step S120 and step S130, for the optimal physical property data corresponding to each component, each physical property data is passed through a word embedding layer to obtain a plurality of parameter input vectors, and the microstructure diagram corresponding to each component is input into a first convolutional neural network as a filter to obtain a structure input vector. It should be understood that, since the components are not independent but have a relationship with each other, in the technical solution of the present application, in order to obtain the relationship information between the components, a context encoder model is used for context encoding. That is, first, each of the physical property data is passed through a word embedding layer to obtain a plurality of parameter input vectors; then, the microscopic structure diagram corresponding to each component is input into a first convolution neural network serving as a filter to be processed, so as to extract a feature distribution representation of local features of the microscopic structure diagram in a high-dimensional space, and thus a structure input vector is obtained.

Specifically, in this embodiment of the present application, the process of inputting the microstructure diagram corresponding to each component into a first convolutional neural network serving as a filter to obtain a structure input vector includes: firstly, carrying out convolution processing on the microscopic structure diagram by each layer except the last layer of the first convolution neural network to obtain a structural feature diagram; then, the last layer of the first convolutional neural network performs feature matrix-based global average pooling on the structural feature map to obtain the structural input vector. It will be appreciated that by global average pooling, the number of parameters can be reduced and overfitting reduced, thereby improving the accuracy of subsequent calculations.

In steps S140 and S150, the plurality of parameter input vectors and the structure input vector are passed through a context-based coding encoder to obtain a sequence of coded feature vectors, and all the coded feature vectors in the sequence of coded feature vectors are concatenated to obtain feature representation vectors corresponding to the respective components. That is, in the technical solution of the present application, after obtaining the plurality of parameter input vectors and the structure input vector, the plurality of parameter input vectors and the structure input vector are further input to a context-based coding encoder together for coding, so as to obtain a sequence of coded feature vectors. In one particular example, the plurality of parametric input vectors and the structural input vector may be passed through a converter model of the encoder to obtain the sequence of encoded feature vectors. It will be appreciated that since the converter-based encoder model is capable of encoding the input vector based on context, the obtained sequence of encoded feature vectors may obtain information about the association of the various components of the global nature. Then, all the coding feature vectors in the sequence of the coding feature vectors are cascaded, so that feature expression vectors corresponding to all the components are obtained.

In step S160, a conjunctive adjacency matrix and a disjunctive adjacency matrix between the feature expression vectors corresponding to the components are constructed based on the conjunctive and disjunctive logical operation rules among the components in the material of the magnesium-carbon skateboard, wherein the conjunctive logical operation rules represent the parallel relationship among the rules, and the disjunctive logical operation represents the alternative relationship among the rules. It should be understood that, considering that experimental data for the single component can be obtained through a deep neural network model, a feature representation of the component can be obtained, and therefore, in the technical solution of the present application, if an adjacency matrix for representing a correlation between the components can be further obtained, a technology of a graph neural network can be used to obtain an association characterization vector of each node containing correlation information. However, here, it is obvious that the components do not have node topology information represented by nodes, and therefore, a common adjacency matrix cannot be used, but considering that the components can be actually expressed in a regular form, for example, the formulation ratio of the large crystal magnesite 98 with the specification of 3-1mm is 25% -32%, in the technical solution of the present application, the adjacency matrix can be constructed by logical operation between rules.

In particular, it is considered that the logical operations between said rules generally include conjunctions and disjunctions, denoted by the symbols Λ and v, respectively, for representing the juxtaposition or alternative relationships between said rules, i.e. the meaning of "and" or ". For the components as described above, there is also such a relationship that, for example, the relationship that "the formulation ratio of large crystalline magnesite 98 of 3 to 1mm in specification is 25% to 32%" and the formulation ratio of large crystalline magnesite 98 of 200 mesh in specification is 24% to 28% "is" or ", and the relationship that" the formulation ratio of large crystalline magnesite 98 of 3 to 1mm in specification is 25% to 32% "and the formulation ratio of crystalline flake graphite 98 of 100 mesh in specification is 1% to 1.8" is "and", therefore, discrete logical nodes can be constructed for the rules satisfied for the respective components based on the above method, thereby forming an adjacent matrix, respectively expressed as:

here, the first and second liquid crystal display panels are,

Specifically, in the embodiment of the present application, the process of constructing a conjunctive adjacency matrix and a disjunctive adjacency matrix between feature expression vectors corresponding to respective components based on a logical operation rule of conjunctive and disjunctive between the respective components in a material of the magnesium-carbon skateboard includes: constructing a conjunction adjacency matrix between characteristic expression vectors corresponding to all components based on a conjunction logic operation rule among all components in a manufacturing material of the magnesium-carbon sliding plate according to the following formula;

wherein the formula is:

wherein the content of the first and second substances,

Accordingly, in a specific example, an disjunctive adjacency matrix between the characteristic expression vectors corresponding to the components is constructed according to the following formula based on the disjunctive logic operation rule among the components in the manufacturing material of the magnesium-carbon skateboard;

wherein the formula is:

wherein the content of the first and second substances,

In steps S170 and S180, passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix, where the first feature matrix and the conjunctive adjacency matrix have the same scale, and the second feature matrix and the disjunctive adjacency matrix have the same scale, and calculating the point-by-point between the first feature matrix and the second feature matrix to obtain a fused adjacency-like matrix. That is, in the technical solution of the present application, after obtaining the conjunctive adjacency matrix and the disjunctive adjacency matrix, further processing the conjunctive adjacency matrix and the disjunctive adjacency matrix in a second convolutional neural network, respectively, so as to map the adjacency matrix into a high-dimensional space, thereby obtaining a first feature matrix and a second feature matrix, where the dimension of the feature matrix is the same as that of the original input matrix. Then, the points of the first feature matrix and the second feature matrix are calculated and the two feature matrices are fused, so that a class adjacency matrix corresponding to the feature representation of each node can be obtained.

Specifically, in this embodiment of the present application, the process of passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix includes: first, each layer of the second convolutional neural network except the last layer explicitly spatially encodes the conjunctive adjacency matrix and the disjunctive adjacency matrix to obtain a first feature map and a second feature map by the following formula:

f_i＝active(N_i×f_i-1+B_i)

wherein f is_i-1As input to the ith layer of the second convolutional neural network, f_iIs the output of the ith layer of the second convolutional neural network, N_iA filter of the second convolutional neural network of the ith layer, and B_iAnd active represents a nonlinear activation function for the bias matrix of the ith layer of the second convolutional neural network. Then, a last layer of the second convolutional neural network performs global pooling along a channel dimension on the first feature map and the second feature map to obtain the first feature matrix and the second feature matrix. It will be appreciated that by global pooling, the number of parameters can be reduced, thereby reducing overfitting to improve the accuracy of subsequent regression.

In steps S190 and S200, the feature representation vectors corresponding to the components are two-dimensionally concatenated to obtain a component feature matrix, and the component feature matrix and the class adjacency matrix are passed through a graph neural network to obtain a graph structure feature matrix, each row of the graph structure feature matrix is an associated feature vector containing associated information of a corresponding component, wherein the graph neural network is adapted to process graph data in an irregular non-euclidean air to extract associated information of a data sample due to feature information and irregular topological structure information. That is, after the class adjacency matrix is obtained, the feature representation vectors of each component are two-dimensionally spliced to obtain a component feature matrix, and then the component feature matrix and the class adjacency matrix pass through a graph neural network together to obtain a graph structure feature matrix, so that each row of the graph structure feature matrix is the associated representation vector containing the associated information of the corresponding component. It should be understood that, here, the graph neural network can be used to process graph data in an irregular non-euclidean space, so as to extract the associated information of the data sample due to the feature information and the irregular topological structure information, and therefore, the obtained associated characterization vector of the single component can improve the accuracy of decoding regression compared with the feature representation vector obtained by directly splicing.

In step S210, the associated characterization vector for each component is passed through a decoder for regression to obtain the optimal ratio of the component. That is, in the technical solution of the present application, after the graph structure feature matrix is obtained through the graph neural network, the associated token vector of each component is further passed through a decoder for regression, so as to obtain the optimal proportion of the component. In addition, after the optimal proportion of the component is obtained, other label values for representing the upper and lower limits of the proportion range can be set by taking the optimal proportion as the center, for example, 27%, 28%, 29%, 31%, 32% and 33% of label values with the optimal proportion of 30% can be set, then the probability that the other label values belong is obtained through decoding regression, for example, the label probability of 28%, 29%, 31% and 32% is greater than a preset threshold value, and then the feasible proportion range of 28% -32% is obtained.

That is, in one specific example, after the optimal ratio of the component is obtained, other label values for indicating the upper and lower limits of the ratio range are first set centering on the optimal ratio of the certain component; then, the associated feature vector of each component is passed through the decoder to obtain the probability that the associated feature vector belongs to other label values; finally, based on the probability, it is determined whether other tag values are feasible.

In summary, the preparation method of the magnesium-carbon skateboard in the embodiment of the present application is clarified, and the physical property data of the magnesium-carbon skateboard collected after changing a certain component under the condition that other components are not changed and the microscopic structure diagram of the magnesium-carbon skateboard are processed based on the context encoder model and the convolutional neural network model to obtain the feature representation of the component, and meanwhile, the adjacency matrix is constructed through the logical operation between rules to represent the association relationship between the components, and the association characterization vector of each node containing association information is further obtained by using the technology of the graph neural network, so as to improve the accuracy of decoding regression. Thus, the optimal proportion of each component in the material for manufacturing the magnesium-carbon sliding plate can be determined, and the performance of the magnesium-carbon material can be improved.

Exemplary System

Fig. 3 illustrates a block diagram of a system for making a magnesium carbon skateboard according to an embodiment of the present application. As shown in fig. 3, a system 400 for preparing a magnesium-carbon skateboard according to an embodiment of the present application includes: the data acquisition unit 410 is used for acquiring optimal physical property data and a microscopic structure diagram corresponding to each component in a manufactured material of the magnesium-carbon sliding plate, wherein the manufactured material of the magnesium-carbon sliding plate comprises magnesia, scale graphite, carbores p, spherical asphalt, aluminum-silicon alloy powder, sulfide powder, carbide powder, thermosetting phenolic resin and carbores T60, and the physical property data comprises a linear change rate, a weight change rate, a volume density, an apparent porosity and a normal-temperature compressive strength; a word embedding unit 420, configured to, for the optimal physical performance data corresponding to each component obtained by the data obtaining unit 410, pass each physical performance data through a word embedding layer to obtain a plurality of parameter input vectors; a first convolution unit 430, configured to input the microstructure diagram corresponding to each component obtained by the data obtaining unit 410 into a first convolution neural network serving as a filter to obtain a structure input vector; an encoding unit 440, configured to pass the plurality of parameter input vectors obtained by the word embedding unit 420 and the structure input vector obtained by the first convolution unit 430 through a context-based encoding encoder to obtain a sequence of encoded feature vectors; a concatenation unit 450, configured to concatenate all the encoded feature vectors in the sequence of encoded feature vectors obtained by the encoding unit 440 to obtain feature representation vectors corresponding to the respective components; an adjacency matrix construction unit 460, configured to construct a conjunctive adjacency matrix and a disjunctive adjacency matrix between the feature expression vectors corresponding to the components obtained by the cascade unit 450 based on a conjunction and disjunctive logical operation rule between the components in the material of the magnesium-carbon skateboard, where the conjunction logical operation rule represents a parallel relationship between the rules, and the disjunctive logical operation represents an alternative relationship between the rules; a second convolution unit 470, configured to pass the conjunctive adjacency matrix obtained by the adjacency matrix construction unit 460 and the disjunctive adjacency matrix obtained by the adjacency matrix construction unit 460 through a second convolution neural network to obtain a first feature matrix and a second feature matrix, where the first feature matrix and the conjunctive adjacency matrix have the same scale, and the second feature matrix and the disjunctive adjacency matrix have the same scale; a fusion unit 480, configured to calculate a point-by-point addition between the first feature matrix obtained by the second convolution unit 470 and the second feature matrix obtained by the second convolution unit 470 to obtain a fused adjacency-like matrix; a two-dimensional splicing unit 490, configured to perform two-dimensional splicing on the feature representation vectors corresponding to the components obtained by the cascading unit 450 to obtain a component feature matrix; a graph neural network unit 500, configured to pass the component feature matrix obtained by the two-dimensional stitching unit 490 and the class adjacency matrix obtained by the fusion unit 480 through a graph neural network to obtain a graph structure feature matrix, where each row of the graph structure feature matrix is an associated characterization vector containing associated information of a corresponding component, and the graph neural network is adapted to process graph data in an irregular non-euclidean space to extract associated information of a data sample due to feature information and irregular topology information; and a regression unit 510, configured to pass the associated characterization vector of each component obtained by the graph neural network unit 500 through a decoder for regression to obtain an optimal ratio of the component.

In one example, in the above magnesium-carbon skateboard preparation system 400, the first volume unit 430 is further configured to: each layer except the last layer of the first convolution neural network is used for carrying out convolution processing on the microscopic structure diagram to obtain a structural feature diagram; and the last layer of the first convolutional neural network performs feature matrix-based global average pooling on the structural feature map to obtain the structural input vector.

In an example, in the above system 400 for preparing a magnesium-carbon skateboard, the encoding unit 440 is further configured to: passing the plurality of parametric input vectors and the structural input vector through a converter model of the encoder to obtain the sequence of encoded feature vectors.

In one example, in the above magnesium-carbon skateboard preparation system 400, the adjacency matrix building unit 460 is further configured to: constructing a conjunction adjacency matrix between characteristic expression vectors corresponding to all components based on a conjunction logic operation rule among all components in a manufacturing material of the magnesium-carbon sliding plate according to the following formula;

wherein the formula is:

wherein the content of the first and second substances,

In one example, in the above-mentioned system 400 for preparing a magnesium-carbon skateboard, an extracted adjacency matrix between the eigenvectors corresponding to the respective components is constructed in the following formula based on the logical operation rule of extraction between the respective components in the material from which the magnesium-carbon skateboard is made;

wherein the formula is:

wherein the content of the first and second substances,

In one example, in the above system 400 for preparing a magnesium-carbon skateboard, the second convolution unit 470 is further configured to: each layer of the second convolutional neural network except the last layer explicitly spatially encodes the conjunctive adjacency matrix and the disjunctive adjacency matrix to obtain a first feature map and a second feature map according to the following formula:

f_i＝active(N_i×f_i-1+B_i)

In one example, in the above magnesium carbon skateboard preparation system 400, further comprising: setting other label values for indicating upper and lower limits of a ratio range centering on the optimum ratio of the certain component; passing the associated token vector for each component through the decoder to obtain probabilities that the associated token vector belongs to other tag values; and determining whether other tag values are feasible based on the probability.

Here, it can be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the above-described magnesium carbon skateboard preparation system 400 have been described in detail in the above description of the magnesium carbon skateboard preparation method with reference to fig. 1 to 2, and thus, a repetitive description thereof will be omitted.

As described above, the system 400 for manufacturing a magnesium carbon skateboard according to the embodiment of the present application may be implemented in various terminal devices, such as a server of a manufacturing algorithm of a magnesium carbon skateboard, and the like. In one example, the system 400 for preparing a magnesium-carbon skateboard according to an embodiment of the present application may be integrated into a terminal device as a software module and/or a hardware module. For example, the magnesium-carbon sled preparation system 400 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the mg-carbon sled manufacturing system 400 can also be one of many hardware modules of the terminal device.

Alternatively, in another example, the preparation system 400 of the magnesium carbon skateboard and the terminal device may be separate devices, and the preparation system 400 of the magnesium carbon skateboard may be connected to the terminal device through a wired and/or wireless network and transmit the interactive information according to an agreed data format.

Exemplary electronic device

Next, an electronic apparatus according to an embodiment of the present application is described with reference to fig. 4. As shown in fig. 4, the electronic device 10 includes one or more processors 11 and memory 12. The processor 11 may be a Central Processing Unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

Memory 12 may include one or more computer program products that may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, Random Access Memory (RAM), cache memory (cache), and/or the like. The non-volatile memory may include, for example, Read Only Memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer readable storage medium and executed by the processor 11 to implement the functions of the method for preparing a magnesium carbon skateboard of the various embodiments of the present application described above and/or other desired functions. Various content such as feature representation vectors, class adjacency matrices, and the like may also be stored in the computer-readable storage medium.

In one example, the electronic device 10 may further include: an input system 13 and an output system 14, which are interconnected by a bus system and/or other form of connection mechanism (not shown).

The input system 13 may comprise, for example, a keyboard, a mouse, etc.

The output system 14 can output various information including the optimum ratio, regression value, etc. to the outside. The output system 14 may include, for example, a display, speakers, a printer, and a communication network and its connected remote output devices, among others.

Of course, for simplicity, only some of the components of the electronic device 10 relevant to the present application are shown in fig. 4, omitting components such as buses, input/output interfaces, and the like. In addition, the electronic device 10 may include any other suitable components depending on the particular application.

Exemplary computer program product and computer-readable storage Medium

In addition to the above-described methods and apparatus, embodiments of the present application may also be a computer program product comprising computer program instructions that, when executed by a processor, cause the processor to perform the steps in the functions of the method of making a magnesium carbon skateboard according to various embodiments of the present application described in the "exemplary methods" section of this specification, supra.

The computer program product may be written with program code for performing the operations of embodiments of the present application in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server.

Furthermore, embodiments of the present application may also be a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, cause the processor to perform the steps in the method of making a magnesium carbon skateboard described in the "exemplary methods" section of this specification, supra.

The computer-readable storage medium may take any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may include, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The foregoing describes the general principles of the present application in conjunction with specific embodiments, however, it is noted that the advantages, effects, etc. mentioned in the present application are merely examples and are not limiting, and they should not be considered essential to the various embodiments of the present application. Furthermore, the foregoing disclosure of specific details is for the purpose of illustration and description and is not intended to be limiting, since the foregoing disclosure is not intended to be exhaustive or to limit the disclosure to the precise details disclosed.

The block diagrams of devices, apparatuses, systems referred to in this application are only given as illustrative examples and are not intended to require or imply that the connections, arrangements, configurations, etc. must be made in the manner shown in the block diagrams. These devices, apparatuses, devices, systems may be connected, arranged, configured in any manner, as will be appreciated by those skilled in the art. Words such as "including," "comprising," "having," and the like are open-ended words that mean "including, but not limited to," and are used interchangeably therewith. The words "or" and "as used herein mean, and are used interchangeably with, the word" and/or, "unless the context clearly dictates otherwise. The word "such as" is used herein to mean, and is used interchangeably with, the phrase "such as but not limited to".

It should also be noted that in the devices, apparatuses, and methods of the present application, the components or steps may be decomposed and/or recombined. These decompositions and/or recombinations are to be considered as equivalents of the present application.

The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The foregoing description has been presented for purposes of illustration and description. Furthermore, the description is not intended to limit embodiments of the application to the form disclosed herein. While a number of example aspects and embodiments have been discussed above, those of skill in the art will recognize certain variations, modifications, alterations, additions and sub-combinations thereof.

Claims

1. A preparation method of a magnesium-carbon sliding plate is characterized by comprising the following steps:

2. The method for preparing a magnesium-carbon skateboard according to claim 1, wherein inputting the microstructure diagram corresponding to each component into a first convolutional neural network as a filter to obtain a structural input vector comprises:

each layer except the last layer of the first convolution neural network is used for carrying out convolution processing on the microscopic structure diagram to obtain a structural feature diagram; and

the last layer of the first convolutional neural network performs feature matrix-based global average pooling on the structural feature map to obtain the structural input vector.

3. The method of making a magnesium-carbon skateboard of claim 2, wherein passing the plurality of parametric input vectors and the structural input vector through a context-based coding encoder to obtain a sequence of coded feature vectors comprises:

passing the plurality of parametric input vectors and the structural input vector through a converter model of the encoder to obtain the sequence of encoded feature vectors.

4. The preparation method of the magnesium-carbon skateboard of claim 3, wherein constructing a conjunct adjacency matrix and a disjunct adjacency matrix between the feature expression vectors corresponding to the components based on the logical operation rules of conjunct and disjunct among the components in the material of the magnesium-carbon skateboard comprises:

constructing a conjunction adjacency matrix between characteristic expression vectors corresponding to all components based on a conjunction logic operation rule among all components in a manufacturing material of the magnesium-carbon sliding plate according to the following formula;

wherein the formula is:

wherein the content of the first and second substances,

5. The preparation method of the magnesium-carbon sliding plate according to claim 4, wherein an extracted adjacency matrix between the characteristic expression vectors corresponding to the components is constructed according to the following formula based on the logic operation rule of extraction between the components in the material of the magnesium-carbon sliding plate;

wherein the formula is:

wherein the content of the first and second substances,

6. The method for preparing the magnesium-carbon skateboard of claim 5, wherein the passing the conjunctive adjacency matrix and the disjunctive adjacency matrix through a second convolutional neural network to obtain a first feature matrix and a second feature matrix comprises:

each layer of the second convolutional neural network except the last layer explicitly spatially encodes the conjunctive adjacency matrix and the disjunctive adjacency matrix to obtain a first feature map and a second feature map according to the following formula:

f_i＝active(N_i×f_i-1+B_i)

wherein f is_i-1As input to the ith layer of the second convolutional neural network, f_iIs the output of the ith layer of the second convolutional neural network, N_iA filter of the second convolutional neural network of the ith layer, and B_iActive represents a nonlinear activation function for a bias matrix of the ith layer of the second convolutional neural network; and

a last layer of the second convolutional neural network performs global pooling along a channel dimension on the first feature map and the second feature map to obtain the first feature matrix and the second feature matrix.

7. The method of making a magnesium-carbon skateboard of claim 6, further comprising:

setting other label values for indicating upper and lower limits of a ratio range centering on the optimum ratio of the certain component;

passing the associated token vector for each component through the decoder to obtain probabilities that the associated token vector belongs to other tag values; and

based on the probability, it is determined whether other tag values are feasible.

8. A preparation system of magnesium carbon slide, characterized by comprising:

9. The system for preparing a magnesium-carbon skateboard of claim 8, wherein the first volumetric unit is further configured to:

each layer except the last layer of the first convolution neural network is used for carrying out convolution processing on the microscopic structure diagram to obtain a structural feature diagram; and the last layer of the first convolutional neural network performs feature matrix-based global average pooling on the structural feature map to obtain the structural input vector.

10. An electronic device, comprising:

a processor; and

a memory having stored therein computer program instructions that, when executed by the processor, cause the processor to perform the method of making a magnesium-carbon skateboard of any of claims 1-7.