CN117272003A

CN117272003A - Method and device for analyzing bending creep resistance test data of artificial board and related equipment

Info

Publication number: CN117272003A
Application number: CN202311568522.7A
Authority: CN
Inventors: 请求不公布姓名
Original assignee: Guangzhou Oppein Integrated Home Co ltd
Current assignee: Guangzhou Oppein Integrated Home Co ltd
Priority date: 2023-11-23
Filing date: 2023-11-23
Publication date: 2023-12-22
Anticipated expiration: 2043-11-23

Abstract

The application provides a method, a device and related equipment for analyzing bending creep resistance test data of an artificial board, which are used for acquiring a test data characterization vector and a deformation chart characterization vector based on a form data set of a target test board and a data extraction mark indicating a target data extraction dimension, integrating the test data characterization vector and the deformation chart characterization vector, determining a correlation coefficient between each form data set and the target data extraction dimension, determining a board test result included in each form data set, completing automatic targeted extraction of the board test result of the target test board in the target data extraction dimension, and greatly increasing the extraction speed of the board test result. In addition, by changing the target data extraction dimension characterized by the data extraction indicia, sheet test results for the target test sheet under different attributes can be extracted.

Description

Method and device for analyzing bending creep resistance test data of artificial board and related equipment

Technical Field

The application relates to the technical fields of data processing and artificial intelligence, in particular to a method and a device for analyzing bending creep resistance test data of an artificial board and related equipment.

Background

In engineering applications, the sheet material is often used for load bearing structures, and therefore its bending creep resistance is very important. The bending creep test of the artificial board is a test method for evaluating the mechanical properties of the board. In testing, it is often necessary to prepare a sample of the artificial board material and secure it to the test equipment and then simulate the forces experienced by the artificial board material during actual use by applying a constant force or applying different forces within a certain range. Over time, the sheet material may creep and also bend. Through recording and analyzing the deformation and stress data of the plate, the performance index of the plate in bending and creep can be obtained. After the bending creep test of the artificial board is completed, test results such as bending resistance (such as bending rigidity and bending strength), creep performance (such as creep rate and creep strain) and failure analysis results (such as failure conditions of plastic deformation, yield, damage and the like) need to be identified from the recorded test data, currently, manual statistics is generally performed on the test results by adopting a manual mode, the statistical mode is single, and corresponding numerical values are obtained only according to experience. For large-scale tests, the accuracy of test result acquisition cannot be guaranteed.

Disclosure of Invention

In view of this, the embodiments of the present application at least provide a method, an apparatus and a related device for analyzing bending creep test data of an artificial board.

The technical scheme of the embodiment of the application is realized as follows: in one aspect, an embodiment of the present application provides a method for analyzing bending creep test data of an artificial board, which is applied to computer equipment, and the method includes: acquiring a data extraction mark and a morphological data set of a target test board, wherein the data extraction mark is used for indicating a target data extraction dimension; the form data set comprises bending creep resistance test data and a plate deformation graph; performing characterization vector mining on the bending creep resistance test data and the data extraction marks through a first machine learning model to obtain test data characterization vectors corresponding to each form data set; performing characterization vector mining on the plate deformation graph through a second machine learning model to obtain deformation graph characterization vectors corresponding to each form data set; integrating the test data characterization vector and the deformation map characterization vector through a feature integration model to obtain an integrated characterization vector; outputting association coefficients between each morphological data set and a target data extraction dimension based on the integrated characterization vector through a third machine learning model; determining a board test result included in each morphological data set based on the integrated characterization vector through a fourth machine learning model; and determining a board test result of the target test board in the target data extraction dimension from board test results included in the morphological data set based on the association coefficient.

In some embodiments, the performing, by using a first machine learning model, feature vector mining on the bending creep test data and the data extraction mark to obtain test data feature vectors corresponding to each morphological data set includes: combining the bending creep resistance test data in the form data set with the data extraction mark to obtain target test data corresponding to the form data set; inputting the target test data into the first machine learning model; outputting test data characterization vectors corresponding to the corresponding morphological data sets based on the target test data through the first machine learning model; the feature integration model comprises a first feature integration model and a second feature integration model; the feature integration model integrates the test data characterization vector and the deformation map characterization vector to obtain an integrated characterization vector, and the feature integration model comprises the following steps: integrating the test data characterization vector and the deformation graph characterization vector through a first feature integration model to obtain a first integration characterization vector; integrating the test data characterization vector and the deformation graph characterization vector through a second feature integration model to obtain a second integration characterization vector; and combining the first integration characterization vector and the second integration characterization vector to obtain the integration characterization vector.

In some embodiments, the first feature integration model includes X sequentially connected first interactive learning modules, and the second feature integration model includes X sequentially connected second interactive learning modules, where X is a positive integer greater than 1; the integrating is performed by a first feature integration model based on the test data characterization vector and the deformation map characterization vector to obtain a first integration characterization vector, which comprises the following steps: obtaining i first intermediate characterization vectors and i first input array representations output by an i first interactive learning module, and obtaining i second query array representations and i second output weighting array representations output by an i second interactive learning module; wherein i is a positive integer less than X, where j=i+1; determining, by a j-th first interactive learning module, j first intermediate token vectors based on the i first input array representations, the i first intermediate token vectors, the i second query array representations, and i second output weighting array representations, and j first input array representations, j first query array representations, and j first output weighting array representations based on the i first intermediate token vectors; if j is less than X, loading the j first intermediate characterization vectors and j first input array representations to a subsequent first interactive learning module; loading j first query array representations and j first output weighting array representations to a subsequent second interactive learning module; if j=x, determining the j first intermediate token vectors as the first integrated token vectors; wherein if i is 1, the first interactive learning module determines a first intermediate token vector and a first input array representation based on the test data token vector; the first second interactive learning module determines a second query array representation and a second output weighting array representation based on the deformation map characterization vector.

In some embodiments, the method further comprises: acquiring a first debugging sample, wherein the first debugging sample comprises a plurality of first form data set learning examples and annotation information of the first form data set learning examples, and the annotation information is used for indicating a test board corresponding to the first form data set learning examples; the first morphological data set learning example is obtained after a partial shielding operation is adopted; debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of first morphological data set learning examples and annotation information of the first morphological data set learning examples.

In some embodiments, the debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of first morphology data set learning examples and annotation information of the first morphology data set learning examples includes: performing characterization vector mining on the test data in the first morphological data set learning example through the first machine learning model to obtain a test data characterization vector of the first morphological data set learning example; performing characterization vector mining on the plate deformation graph in the first morphological data set learning example through the second machine learning model to obtain a deformation graph characterization vector of the first morphological data set learning example; integrating the deformation map representation vector of the first morphological data set learning example and the test data representation vector of the first morphological data set learning example through the first feature integration model to obtain a first integration representation vector of the first morphological data set learning example; integrating the deformation map representation vector of the first morphological data set learning example and the test data representation vector of the first morphological data set learning example through the second feature integration model to obtain a second integration representation vector of the first morphological data set learning example; combining a first integrated characterization vector of the first morphological data set learning example with a second integrated characterization vector of the first morphological data set learning example to obtain an integrated characterization vector of the first morphological data set learning example; determining a prediction test board classification corresponding to the first morphological data set learning example based on the integrated characterization vector of the first morphological data set learning example; determining a first test board classification distribution based on the predicted test board classifications corresponding to all the first morphology data set learning examples; determining a first cost based on the first test board classification distribution and the second test board classification distribution; the second test board classification distribution is determined based on annotation information of all the first morphological data set learning examples; and optimizing model-learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model based on the first price until a first debugging cut-off requirement is met.

In some embodiments, the method further comprises: obtaining a second debugging sample, wherein the second debugging sample comprises a plurality of second shape data set learning examples and a plurality of third shape data set learning examples, test data in the same second shape data set learning examples are identical to test boards corresponding to the board deformation graph, and the test data in the same third shape data set learning examples are different from the test boards corresponding to the board deformation graph; and debugging the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model based on the plurality of second morphology data set learning examples and the plurality of third morphology data set learning examples.

In some embodiments, the debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of second morphology data set learning examples and the plurality of third morphology data set learning examples comprises: performing characterization vector mining on the test data in the loaded learning example through the first machine learning model to obtain a test data characterization vector of the loaded learning example; the loaded learning example is a second morphological data set learning example or a third morphological data set learning example; performing characterization vector mining on the plate deformation graph in the loaded learning example through the second machine learning model to obtain a deformation graph characterization vector of the loaded learning example; integrating the test data characterization vector of the loaded learning example and the deformation graph characterization vector of the loaded learning example through the first feature integration model to obtain a first integrated characterization vector of the loaded learning example; integrating the test data characterization vector of the loaded learning example and the deformation graph characterization vector of the loaded learning example through the second feature integration model to obtain a second integrated characterization vector of the loaded learning example; predicting a first prediction test board corresponding to the loaded learning example based on the first integrated characterization vector of the loaded learning example; predicting a second prediction test board corresponding to the loaded learning example based on the second integrated characterization vector of the loaded learning example; if the loaded learning example is a second state data set learning example, and meanwhile, a first prediction test board corresponding to the second state data set learning example is different from the corresponding second prediction test board, optimizing model learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model, and enabling the first prediction test board redetermined for the second state data set learning example to be identical to the redetermined second prediction test board after optimizing the learnable variables; if the loaded learning example is a third morphology data set learning example, and meanwhile, the first prediction test board corresponding to the third morphology data set learning example is the same as the second prediction test board corresponding to the third morphology data set learning example, optimizing model variable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model, and enabling the first prediction test board redetermined for the third morphology data set learning example to be different from the second prediction test board redetermined after optimizing the variable variables.

In some embodiments, the method further comprises: obtaining a third debugging sample, wherein the third debugging sample comprises a plurality of sample morphological data sets and associated coefficient annotation information of the sample morphological data sets, the sample morphological data sets comprise a fourth morphological data set learning example and a sample data extraction mark, and the associated coefficient annotation information is used for indicating a marked associated coefficient between the fourth morphological data set learning example and a sample target data extraction dimension indicated by the sample data extraction mark; performing characterization vector mining on the test data in the sample form data set through the first machine learning model to obtain a test data characterization vector of the sample form data set; performing characterization vector mining on the plate deformation graph in the sample form data set through the second machine learning model to obtain a deformation graph characterization vector of the sample form data set; integrating the test data characterization vector of the sample form data set and the deformation map characterization vector of the sample form data set through the first feature integration model to obtain a first integrated characterization vector of the sample form data set; integrating the test data characterization vector of the sample form data set and the deformation map characterization vector of the sample form data set through the second feature integration model to obtain a second integrated characterization vector of the sample form data set; combining the first integrated characterization vector of the sample morphology data set with the second integrated characterization vector of the sample morphology data set to obtain an integrated characterization vector of the sample morphology data set; outputting, by the third machine learning model, a predictive correlation coefficient between the fourth morphology data set learning example and the sample target data extraction dimension based on the integrated characterization vector of the sample morphology data set; determining a second cost based on the predicted association coefficient and the labeling association coefficient; and optimizing model-learnable variables of the first machine learning model, the second machine learning model, the first feature integration model, the second feature integration model and the third machine learning model based on the second cost until a second debugging stop requirement is met.

In a second aspect, the present application provides an artificial board bending creep test data analysis device, including: the board data acquisition module is used for acquiring a data extraction mark and a morphological data set of the target test board, wherein the data extraction mark is used for indicating a target data extraction dimension; the form data set comprises bending creep resistance test data and a plate deformation graph; the test data mining module is used for carrying out characterization vector mining on the bending creep resistance test data and the data extraction marks through a first machine learning model to obtain test data characterization vectors corresponding to each form data set; the image information mining module is used for carrying out characterization vector mining on the plate deformation graph through a second machine learning model to obtain deformation graph characterization vectors corresponding to each form data set; the characterization vector integration module is used for integrating the test data characterization vector and the deformation map characterization vector through a feature integration model to obtain an integrated characterization vector; the association coefficient determining module is used for outputting association coefficients between each morphological data set and the target data extraction dimension based on the integrated characterization vector through a third machine learning model; the test result determining module is used for determining board test results included in each form data set based on the integrated characterization vector through a fourth machine learning model; and the target result determining module is used for determining a board test result of the target test board in the target data extraction dimension from board test results included in the morphological data set based on the association coefficient.

In a third aspect, the present application provides a computer device comprising a memory and a processor, the memory storing a computer program executable on the processor, the processor implementing the steps of the method described above when the program is executed.

The application has at least the beneficial effects that include: according to the artificial board bending creep test data analysis method, the test data characterization vector and the deformation chart characterization vector are obtained based on the form data set of the target test board and the data extraction mark indicating the target data extraction dimension, then the test data characterization vector and the deformation chart characterization vector are integrated to obtain the integrated characterization vector, then the association coefficient between each form data set and the target data extraction dimension is determined based on the integrated characterization vector, the board test result included in each form data set is determined based on the integrated characterization vector, and further the board test result of the target test board in the target data extraction dimension is determined from the board test result included in the form data set based on the association coefficient, so that the board test result of the target test board in the target data extraction dimension is automatically and purposefully extracted, and the board test result extraction speed is greatly increased. In addition, by changing the target data extraction dimension characterized by the data extraction indicia, sheet test results for the target test sheet under different attributes can be extracted.

In addition, compared with the data of the target test board under one type (mode), such as bending creep resistance test data or a board deformation graph, the data information contained in the form data group of the target test board is more complete, and then compared with the integrated characterization vector obtained by integrating the test data characterization vector and the deformation graph characterization vector corresponding to one type of test data characterization vector or the deformation graph characterization vector, the vector characterization data information characterized by the integrated characterization vector is more complete, and then compared with extracting the board test result from only one type of test data characterization vector or a single type of deformation graph characterization vector, the accuracy of the extracted board test result can be improved by extracting the board test result based on the integrated characterization vector. In addition, because the test data characterization vector and the deformation chart characterization vector are integrated, vector characterization data information under different types are mutually fused and promoted, and therefore the accuracy of extracting the plate test result is improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the aspects of the present application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the technical aspects of the application.

Fig. 1 is a schematic implementation flow chart of an artificial board bending creep test data analysis method according to an embodiment of the present application;

fig. 2 is a schematic diagram of a composition structure of an artificial board bending creep test data analysis device according to an embodiment of the present application;

fig. 3 is a schematic hardware entity diagram of a computer device according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application are further elaborated below in conjunction with the accompanying drawings and examples, which should not be construed as limiting the present application, and all other embodiments obtained by those skilled in the art without making inventive efforts are within the scope of protection of the present application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. The term "first/second/third" is merely to distinguish similar objects and does not represent a specific ordering of objects, it being understood that the "first/second/third" may be interchanged with a specific order or sequence, as permitted, to enable embodiments of the present application described herein to be practiced otherwise than as illustrated or described herein.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing the present application only and is not intended to be limiting of the present application.

The embodiment of the application provides a method for analyzing bending creep resistance test data of an artificial board, which is applied to computer equipment 104, wherein the computer equipment 104 can be equipment with data processing capability, such as a server, a notebook computer, a tablet computer, a desktop computer, an intelligent television, mobile equipment (for example, a mobile phone, a portable video player, a personal digital assistant, special message equipment and portable game equipment). The computer device 104 provides a user interface and the tester may enter a data extraction flag (i.e., a flag indicating which data needs to be extracted, which may be represented by a numerical value) or select a panel test result extraction flag, and enter a test panel identification, such as a test panel name, of the target test panel for which panel test result extraction is to be performed, then the computer device 104 initiates a panel test result extraction task based on the data extraction flag and the test panel identification of the target test panel.

Referring to fig. 1, the method for analyzing bending creep test data of an artificial board according to the embodiment of the present application may specifically include: step S101, acquiring a data extraction mark and a morphology data set of the target test board, wherein the data extraction mark is used for indicating a target data extraction dimension.

The target test panel is a panel subjected to a bending creep test. When the bending creep test is carried out, the load effect is kept under different test conditions (such as temperature, humidity, loading rate and the like), so that the target test board is subjected to creep deformation in a certain time, bending creep test data such as deflection, time, load, creep rate, stress, strain data and the like of the target test board are recorded, and evaluation of bending resistance, creep performance, material failure (deformation, yield, damage and the like) and the like of the target test board can be carried out through the data. In addition, a deformation map of the target test board is recorded, and the deformation map can be obtained through shooting by the shooting equipment. In the embodiment of the application, the data extraction mark is used to indicate the target data extraction dimension (that is, the attribute information of the target data and the type of the test result), then, the board test result required to be extracted for the target test board is the board test result of the target test board under the target data extraction dimension, for example, if the board test result of the dimension of the creep performance is required to be extracted, the creep performance can be indicated by the data extraction mark. As an embodiment, the data extraction marks are in the form of numerical labels, such as creep performance marks 11, bending resistance marks 12, deformation yield marks 13, irreversible damage marks 14, etc.

From the above, the morphology data set of the target test sheet includes bending creep test data of the target test sheet and a sheet deformation map of the target test sheet. The bending creep test data are data items under different test environments and conditions (such as one element in a multi-dimensional array, the dimension division can be performed based on the dimensions of test conditions, record items, specific numerical values and the like, the bending creep test data can be represented as a two-dimensional matrix or a tensor with more dimensions), and the plate deformation graph is a corresponding image in the deformation process of the target test plate under the corresponding test conditions as the name implies.

The test process of the target test board can be reflected by the test data combined with the form data set obtained by the board deformation graph, so that the test data can be extracted from the form data set of the target test board to obtain bending creep resistance test data. In the form data set of the test board, the test data capacity of the test board may be relatively large, and at this time, the test data of the test board may be grouped to obtain a plurality of bending creep resistant test data sets.

Optionally, based on the set of morphology data sets and the data extraction markers of the target test sheet, morphology data sets may be constructed for the target test sheet, each morphology data set including bending creep resistance test data, a sheet deformation map, and the data extraction markers, to extract test data characterization vectors and deformation map characterization vectors based on the morphology data sets later.

Step S102, performing characterization vector mining on the bending creep resistance test data and the data extraction mark through a first machine learning model to obtain test data characterization vectors corresponding to each form data set.

And step S103, performing characterization vector mining on the plate deformation graph through a second machine learning model to obtain deformation graph characterization vectors corresponding to each form data set.

In the embodiment of the application, for distinguishing, a machine learning model for performing test data characterization vector mining is regarded as a first machine learning model, and a machine learning model for performing sheet deformation map characterization vector mining is regarded as a second machine learning model. The first machine learning model and the second machine learning model may be formed by a Recurrent Neural Network (RNN), a Convolutional Neural Network (CNN), and a fully-connected network (FCN). For example, the first machine learning model may be a Transformers and the second machine learning model may be a CNN. As some embodiments, the bending creep test data in the form data set can be subjected to characterization vector mining through a first machine learning model to obtain vector characterization data (i.e. characterization vectors used for characterizing the characteristics of corresponding data through the vectors) of the bending creep test data; carrying out characterization vector mining on the plate deformation graph in the morphological data set through a second machine learning model to obtain vector characterization data of the plate deformation graph, and determining the vector characterization data of the plate deformation graph as deformation graph characterization vectors in the application; and carrying out characterization vector mining on the data extraction marks through a first machine learning model to obtain vector characterization data of the data extraction marks, combining the vector characterization data of the bending creep resistance test data with the vector characterization data of the data extraction marks, and taking the vector characterization data obtained by combination as a test data characterization vector corresponding to the morphological data group. The test data characterization vector is vector information for characterizing the characteristics of the test data, and can be a feature vector, and similarly, the deformation chart characterization vector is vector information for characterizing the characteristics of the deformation chart, and can also be a feature vector.

As some embodiments, step S102 includes: combining the bending creep resistance test data in the form data set with the data extraction mark to obtain target test data corresponding to the form data set; inputting the target test data into the first machine learning model; and outputting test data characterization vectors corresponding to the corresponding morphological data sets based on the target test data through the first machine learning model. In this embodiment of the present application, after the bending creep resistance test data and the data extraction mark are combined (for example, the data extraction mark may be directly spliced), the first machine learning model is used to perform the feature vector mining, so that the machine learning model that performs the feature vector mining on the data extraction mark is not required to be separately and accurately performed.

Step S104, integrating the test data characterization vector and the deformation map characterization vector through a feature integration model to obtain an integrated characterization vector.

And integrating (i.e. vector integration, such as addition, splicing, connection and the like) the test data characterization vector and the deformation graph characterization vector to complete the integration of the vector characterization data under the two types of the test data type and the plate deformation graph type, wherein the integrated characterization vector can express not only the characteristic information of the bending creep resistance test data but also the characteristic semantic information of the plate deformation graph. In addition, because the test data characterization vector also expresses the characteristic information of the data extraction mark, the integration characterization vector also expresses the characteristic information of the data extraction mark.

In addition, the characteristic expressed by the bending creep test data and the characteristic expressed by the sheet deformation graph can be mutually supported, and the part, consistent with the characteristic of the sheet deformation graph, of the bending creep test data can be reinforced, so that the accuracy of a sheet test result extracted by the integrated characteristic vector can be ensured.

Optionally, the test data characterization vector and the deformation map characterization vector may be combined to obtain a combined characterization vector; and then loading the combined characterization vector into a feature integration model to perform depth characterization vector mining, and taking vector characterization data output by the feature integration model based on the combined characterization vector as an integrated characterization vector. The feature integration model may be formed by a convolutional neural network, a cyclic neural network, and an affine network, which is not particularly limited.

As some implementations, the feature integration model includes a first feature integration model and a second feature integration model, and step S104 may specifically include: integrating the test data characterization vector and the deformation map characterization vector through a first feature integration model to obtain a first integration characterization vector; integrating the test data characterization vector and the deformation map characterization vector through a second feature integration model to obtain a second integration characterization vector; and combining (i.e. stitching) the first integrated token vector and the second integrated token vector to obtain the integrated token vector. In the embodiment of the present application, since the test data characterization vector not only characterizes the feature information of the bending creep resistance test data, but also characterizes the feature information of the data extraction mark, the feature information of the data extraction mark can guide the first feature integration model and the second feature integration model to focus the bending creep resistance test data and the vector characterization data related to the target data extraction dimension in the plate deformation map in the integration process, thereby ensuring the precision of the plate test result extracted to the target data extraction dimension based on the integration characterization vector.

Step S105, determining, by a third machine learning model, a correlation coefficient between each of the morphological data sets and the target data extraction dimension based on the integrated token vector.

Optionally, because the integrated token vector is obtained by combining the morphological data set and the data extraction label, the association coefficient analysis is directly performed through the integrated token vector by the third machine learning model, so as to obtain the association coefficient (i.e. the numerical value for representing the degree of correlation) between the morphological data set and the target data extraction dimension. The third machine learning model may be built up from a convolutional neural network, an affine network, a cyclic neural network, or the like. For example, the third machine learning model may include one or more fully connected layers.

Step S106, determining the board test result included in each form data set based on the integrated characterization vector through a fourth machine learning model.

It will be appreciated that the integrated token vector expresses the characteristic information of the bending creep resistance test data and the sheet deformation map in the morphology data set, and then the sheet test result included in the morphology data set may be determined based on the analysis (i.e., decoding) performed by the integrated token vector. Because each morphological data set may express information of various dimensions of the target test board, the parsed board test result may be at least one based on the integrated characterization vector.

Optionally, the board test result included in the morphological data set may be obtained by analyzing based on the integrated characterization vector through a fourth machine learning model. The fourth machine learning model may be built by a convolutional neural network, an affine network, a cyclic neural network, or the like, and for example, the fourth machine learning model may be a transducer. Because the integrated token vector integrates the features of the data extraction marks, in the process of analyzing the fourth machine learning model, the part of the integrated token vector, which represents the feature information of the data extraction marks, can teach the fourth machine learning model to analyze the board test results related to the target data extraction dimension, so that high correlation between the analyzed board test results and the target data extraction dimension is ensured, and the operation cost is reduced.

Step S107, determining a board test result of the target test board in the target data extraction dimension from board test results included in the morphology data set based on the correlation coefficient.

As some embodiments, step S107 may specifically include: taking the association coefficient as a supporting weighting coefficient (can be understood as a voting coefficient), and performing supporting accumulation (namely accumulated voting) on each board test result included in the form data set to obtain a supporting accumulated value of each board test result; determining a board test result with the largest supporting accumulated value based on the supporting accumulated value of each board test result; and determining the board test result with the largest supporting accumulated value as the board test result of the target test board in the target data extraction dimension.

For example, if the set of morphology data sets of the target test board includes a morphology data set a, a morphology data set B, and a morphology data set C, determining in step S105 that the association coefficient between the morphology data set a and the target data extraction dimension is x, the association coefficient between the morphology data set B and the target data extraction dimension is y, and the association coefficient between the morphology data set C and the target data extraction dimension is z; if it is determined in step S106 that the sheet test result included in the shape data set a includes a maximum deflection of 2.5mm and a maximum load of 1200N, it is determined that the sheet test result included in the shape data set B includes a maximum deflection of 2.5mm and a creep rate of 0.02%, it is determined that the sheet test result included in the shape data set C includes a maximum deflection of 2.5mm, for the sheet test result of 2.5mm of the maximum deflection, the corresponding support accumulation value is x+y+z, for the sheet test result of 1200N of the maximum load, the corresponding support accumulation value is x, for the sheet test result of 0.02% of the creep rate, the corresponding support accumulation value is y, and obviously, the support accumulation value of the sheet test result of 2.5mm of the maximum deflection is the sheet test result of the target test sheet in the target data extraction dimension.

In the embodiment of the application, the test data characterization vector and the deformation chart characterization vector are obtained based on the form data set of the target test board and the data extraction mark indicating the target data extraction dimension, then the test data characterization vector and the deformation chart characterization vector are integrated to obtain the integrated characterization vector, then the association coefficient between each form data set and the target data extraction dimension is determined based on the integrated characterization vector, the board test result included in each form data set is determined based on the integrated characterization vector, and then the board test result of the target test board under the target data extraction dimension is determined based on the association coefficient from the board test result included in the form data set, so that the board test result of the target test board under the target data extraction dimension is automatically and purposefully extracted, and the board test result extraction speed is greatly increased. In addition, by changing the target data extraction dimension characterized by the data extraction indicia, sheet test results for the target test sheet at different dimensions can be extracted.

The prior art generally only extracts sheet test results for test sheets from the bending creep test data for the test sheets. The test data of the test board may cause errors due to the sensing angle of the sensor, the fitting degree and the like, and the board test result of the test board under a certain dimension extracted from the bending creep test data of the test board may be inaccurate. Based on the embodiment of the application, the board test result is extracted by combining the test data and the board deformation graph, and the image auxiliary test data which can be jointly shot is verified, so that the test result is more accurate, and it can be understood that the test data and the board deformation graph correspond to each other in time, and the test data and the board deformation graph in the same time correspond to each other. In this embodiment, the shape data set of the target test board includes two data types (a test data type and a board deformation graph type), compared with the data of the target test board under one type, such as bending creep test data or a board deformation graph, the shape data set of the target test board includes more complete data information, and then compared with the test data characterization vector or the deformation graph characterization vector corresponding to a single type, the vector characterization data information represented by the integration characterization vector obtained by integrating the test data characterization vector and the deformation graph characterization vector is more complete, and then compared with extracting the board test result from only one type of test data characterization vector or the deformation graph characterization vector of a single type, extracting the board test result based on the integration characterization vector can improve the accuracy of the extracted board test result. In addition, because the test data characterization vector and the deformation chart characterization vector are integrated, vector characterization data information under different types are mutually fused and promoted, and the accuracy of extracting the plate test result is improved.

The following describes an analysis method for bending creep test data of an artificial board provided in an embodiment of the present application, including: step S210, a data-tag group is generated.

In this step, a data-tag group is generated based on the bending creep test data and the target data extraction dimension to be extracted, and the data extraction tag is correspondingly determined based on the specified target data extraction dimension. The data-mark group comprises bending creep resistance test data, a sheet deformation graph and a data extraction mark of the target test sheet.

Step S220, multi-type integration.

When the method is realized, the test data characterization vector and the deformation graph characterization vector are extracted from the data-mark groups, and then the test data characterization vector and the deformation graph characterization vector are integrated to obtain the integrated characterization vector of each data-mark group.

Step S230, the association coefficient is resolved.

Correlation coefficients between the data in each data-tag group and the target data extraction dimension are determined based on the integrated token vector analysis for each data-tag group.

And S240, analyzing the plate test result.

Based on the integrated characterization vector for each data-tag group, a panel test result included in the data in each data-tag group is determined.

Step S250, selecting a board test result.

And analyzing the integrated characterization vector of each data-mark group to determine the association coefficient between the data in each data-mark group and the extraction dimension of the target data as a support weighting coefficient, carrying out support accumulation on the determined plate test result, and determining the maximum support accumulated value of the plate test result.

It can be understood that the test board may include multiple dimension test results, and then the test results of the target test board in multiple dimensions need to be extracted, and the above steps S210 to S240 may be repeated by replacing the data extraction mark, so as to obtain the test results of the target test board in different dimensions.

The application also provides a board test result extraction model, which comprises a first machine learning model, a second machine learning model, a first feature integration model, a second feature integration model, a third machine learning model, a fourth machine learning model, a fully connected network and an output network. In this embodiment of the present application, the first machine learning model is a transducer model, and the second machine learning model is a CNN.

Inputting target test data obtained by combining the bending creep resistance test data of the target test board with the data extraction mark into a first machine learning model, and outputting a test data characterization vector through the first machine learning model; loading a plate deformation graph of the target test plate to a second machine learning model to output deformation graph characterization vectors; and integrating the first feature integration model and the second feature integration model based on the test data characterization vector and the deformation graph characterization vector, wherein the first feature integration model outputs a first integration characterization vector, the second feature integration model outputs a second integration characterization vector, and then combining the first integration characterization vector and the second integration characterization vector to obtain an integration characterization vector. The first feature integration model, the second feature integration model and the fourth machine learning model are all a transducer model, the first feature integration model comprises X first interactive learning modules (converters) which are sequentially connected, the second feature integration model comprises X second interactive learning modules which are sequentially connected, and X is a positive integer greater than 1. The transducer comprises a plurality of interactive learning modules which are connected in sequence. The interactive learning module comprises a multi-head attention network, a first standardized network, a perceptron network and a second standardized network which are sequentially connected, and further comprises a cross-layer identity connection structure (ResNET) with directivity.

In this embodiment of the present application, the first feature integration model and the second feature integration model are integrated based on Cross-attention (Cross-attention), that is, the interactive learning module in the first feature integration model loads the obtained output weighted array representation and the query array representation into the interactive learning module in the same layer in the second feature integration model, so that the interactive learning module in the same layer in the second feature integration model determines an attention array according to the input array representation in the second feature integration model and the output weighted array representation and the query array representation derived from the interactive learning module in the first feature integration model.

In the embodiment of the application, for convenience of distinguishing, the interactive learning module in the first feature integration model is regarded as a first interactive learning module, and the interactive learning module in the second feature integration model is regarded as a second interactive learning module; the input array representation represents the information that the model is currently focusing on, can be regarded as a question or a content to be judged, i first query array representations (i.e. key arrays, which are also called vectors or matrices, for calculating similarity between the input array representation and each element in all other input sequences), i first output weighting array representations (i.e. value arrays, which are also called vectors or matrices, corresponding to each query array representation, in the attention mechanism, of Values in the attention mechanism, which are also called vectors or matrices, corresponding to each query array representation, containing the importance of the input sequence element represented by the query array representation in the importance of the input sequence element in the attention mechanism, can be regarded as a question or a content to be judged), i first query array representations (i.e. key arrays, which are also called vectors or matrices, for calculating similarity between the input array representation and each element in all other input sequences), i first output weighting array representations (i.e. value arrays, which are also called vectors or matrices, corresponding to each query array representation, containing the importance of the input sequence element represented by the query array representation in the attention mechanism, can be regarded as a specific content, but can be similarly encoded or a specific answer is given to the first query array representation; and respectively treating the input array representation, the query array representation, the output weighted array representation and the output intermediate characterization vector determined by the i second interactive learning modules in the second feature integration model as i second input array representations, i second query array representations, i second output weighted array representations and i second intermediate characterization vectors.

In a specific implementation, the step of integrating, by the first feature integration model, the test data token vector and the deformation map token vector to obtain a first integrated token vector may include: obtaining i first intermediate token vectors, i first input array representations and i second query array representations and i second output weighted array representations output by an ith first interactive learning module, wherein i is a positive integer smaller than X, j=i+1, determining j first intermediate token vectors by the jth first interactive learning module based on the i first input array representations, the i first intermediate token vectors, the i second query array representations and the i second output weighted array representations, and determining j first input array representations, j first query array representations and j first output weighted array representations based on the i first intermediate token vectors, and loading the j first intermediate token vectors, j first input array representations and j first output weighted array representations to a subsequent first interactive learning module if j < X, and loading the j first query array representations and j first output weighted array representations to the subsequent second interactive learning module. If j=x, then the j first intermediate token vectors are determined to be the first integrated token vectors. Wherein if i is 1, the first interactive learning module determines a first intermediate token vector and a first input array representation based on the test data token vector, and the first second interactive learning module determines a second query array representation and a second output weighting array representation based on the deformation token vector.

And integrating the test data characterization vector and the deformation graph characterization vector through a second feature integration model to obtain a second integration characterization vector in a consistent mode.

After the integration characterization vector is obtained, the integration characterization vector is loaded into a third machine learning model, and the integration characterization vector is analyzed through the third machine learning model, so that the association coefficient between each form data set and the target data extraction dimension is obtained. Wherein the third machine learning model may include at least one fully connected network.

In addition, the integrated characterization vector is loaded to a fourth machine learning model, the fourth machine learning model is used for analyzing according to the integrated characterization vector, and the board test result included in the morphological data set is determined. And the fourth machine learning model outputs each data in the plate test result according to the sequence positions, and combines the data determined by the sequence positions according to the sequence of the sequence positions to obtain the plate test result included in the morphological data set.

Optionally, in order to increase output efficiency of the board test result, when determining the board test result included in the morphological data set, the embodiment of the application may include: analyzing the integrated characterization vector through a fourth machine learning model, and determining P alternative output results with the maximum triggering probability at each sequence position according to a Top-k sampling mechanism; selecting a target alternative output result comprising a stop mark from P alternative output results with the maximum triggering probability determined by each sequence position; and determining a target output result based on the target alternative output result, wherein the target output result is used for indicating the board test result included in each form data set. After the fourth machine learning model outputs the board test result included in each form data set and the third machine learning model outputs the association coefficient of each form data set and the target data extraction dimension, board test result selection is performed through a full-connection network, and then the output network outputs the board test result with the largest accumulated value as the board test result of the target test board in the target data extraction dimension. The classifier of the output network is, for example, softmax.

In order to ensure the accuracy of the extracted board test results, the board test result extraction model needs to be debugged in advance. Optionally, for the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model, two pre-trained branch tasks, namely multi-type shielding debugging and multi-type unified debugging, are deployed, so that the convergence efficiency of the cost function is improved, and a board test result extraction model with high robustness and precision is obtained.

As some implementations, the process of multi-type mask debugging may include: step S310, a first debugging sample is obtained, wherein the first debugging sample comprises a plurality of first form data set learning examples and annotation information of the first form data set learning examples, and the annotation information is used for indicating a test board corresponding to the first form data set learning examples; the first morphological data set learning example is obtained after a partial occlusion operation is taken.

The partial shielding operation is to perform data shielding on the test data or the plate deformation graph in the first form data set learning example, for example, shielding or replacing the data item in the test data based on a mask, and the shielding range is not limited.

Step S320, debugging the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model based on the plurality of first morphological data set learning examples and the annotation information of the first morphological data set learning examples.

The step S320 may specifically include:

step S3201, performing characterization vector mining on the test data in the first morphological data set learning example through the first machine learning model to obtain a test data characterization vector of the first morphological data set learning example;

step S3202, performing characterization vector mining on the plate deformation graph in the first morphological data set learning example through the second machine learning model to obtain deformation graph characterization vectors of the first morphological data set learning example;

step S3203, integrating the deformation graph characterization vector of the first morphological data set learning example and the test data characterization vector of the first morphological data set learning example through the first feature integration model to obtain a first integrated characterization vector of the first morphological data set learning example;

step S3204, integrating the deformation graph characterization vector of the first morphological data set learning example and the test data characterization vector of the first morphological data set learning example through the second feature integration model to obtain a second integrated characterization vector of the first morphological data set learning example;

Step S3205, combining the first integrated characterization vector of the first morphology data set learning example and the second integrated characterization vector of the first morphology data set learning example to obtain an integrated characterization vector of the first morphology data set learning example;

step S3206, determining a prediction test board classification corresponding to the first morphology data set learning example based on the integrated characterization vector of the first morphology data set learning example.

The test board classification is based on the integrated characterization vector reasoning of the first form data set learning example to obtain the test board classification of the test board corresponding to the first form data set learning example. The test board classification may be a board classification set, such as a composite board, a plywood, a chipboard, a fiberboard, and the like. As some embodiments, the integrated characterization vector of each first morphology data set learning example may be loaded to an affine network (i.e. a fully connected network), and test board classification prediction is performed based on the integrated characterization vector of the first morphology data set learning example through the affine network, so as to obtain a predicted test board classification corresponding to each first morphology data set learning example.

Step S3207, determining a first test board classification distribution based on all the predicted test board classifications corresponding to the first morphology data set learning examples.

And calculating the quantity of the predicted test board classifications corresponding to all the predicted first form data set learning examples, and determining the first test board classification distribution.

In step S3208, a first cost is determined based on the first test board classification distribution and the second test board classification distribution. The second test panel classification profile is determined based on annotation information for all of the first morphology data set learning examples.

Since the comment information of the first-morphology data set learning example indicates the test board corresponding to the first-morphology data set learning example, it is possible to correspondingly determine the test board classification corresponding to the test board based on the comment information, and thus determine the second test board classification distribution. Optionally, a first cost function in the debugging process is determined in advance, and then the first test board classification distribution and the second test board classification distribution are calculated through the first cost function, so as to obtain a first cost. Optionally, the first cost function is a relative entropy cost function.

Step S3209 optimizes the model learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model based on the first price until a first debug cutoff requirement is satisfied.

The first debug cut-off requirement can be that the cost is smaller than a preset value, the number of times of debugging reaches a number threshold value, and the like, and the first debug cut-off requirement is specifically set according to actual needs.

The first cost is determined based on the first test board classification distribution and the second test board classification distribution in the above debugging process, which is different from the cost between the predicted test board classification and the test board classification determined directly by predicting the test board classification and the test board classification represented by the annotation information of each first form data set learning example, in other words, the debugging link does not limit that the test board classification represented by each predicted test board classification and the annotation information needs to be the same, so that the low standard (soft label) of the annotation information can be completed, and the debugging process is simple and easy. Because the subsequent debugging links of other machine learning models also comprise model-learnable variables (i.e. various parameters such as weights and biases) for optimizing the first machine learning model, the second machine learning model and the first feature integration model and the second feature integration model, the debugging of the former link becomes easy, and the accuracy of each machine learning model can be ensured after the detail optimization of the latter link.

Alternatively, the process of multi-type unified debugging may include: step S3210, obtaining a second debug sample, where the second debug sample includes a plurality of second shape data set learning examples and a plurality of third shape data set learning examples, the test data in the same second shape data set learning examples are the same as the test boards corresponding to the board deformation map, and the test data in the same third shape data set learning examples are different from the test boards corresponding to the board deformation map.

Optionally, a small portion of negative examples are randomly selected from a plurality of second shape data set learning examples, one piece of test data is selected from one of the second shape data set learning examples in two second shape data set learning examples of different test boards, one board deformation graph is selected from the other second shape data set learning examples, and the selected test data and board deformation graph form a third shape data set learning example. Thus, the balance of the positive sample and the negative sample in the second debugging sample can be ensured.

Step S3211, debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of second modality data set learning examples and the plurality of third modality data set learning examples.

The step S3211 may specifically include:

step S32111, performing token vector mining on the test data in the loaded learning example through the first machine learning model to obtain a test data token vector of the loaded learning example. The loaded learning example is a second morphological data set learning example or a third morphological data set learning example;

step S32112, performing characterization vector mining on the plate deformation graph in the loaded learning example through the second machine learning model to obtain a deformation graph characterization vector of the loaded learning example;

step S32113, integrating, by the first feature integration model, based on the test data token vector of the loaded learning example and the deformation map token vector of the loaded learning example, to obtain a first integrated token vector of the loaded learning example;

step S32114, integrating, by the second feature integration model, based on the test data token vector of the loaded learning example and the deformation map token vector of the loaded learning example, to obtain a second integrated token vector of the loaded learning example;

step S32115, predicting a first prediction test board corresponding to the loaded learning example based on the first integrated characterization vector of the loaded learning example.

The first predictive test board is a test board corresponding to the loaded learning example, which is predicted by the first integrated characterization vector of the loaded learning example. Optionally, the first predictive test board corresponding to the loaded learning example is predicted based on the first integrated characterization vector of the loaded learning example through the fully connected network.

Step S32116 predicts a second predicted test board corresponding to the loaded learning example based on the second integrated token vector of the loaded learning example.

The second predictive test board is a test board corresponding to the loaded learning example predicted by the second integrated characterization vector of the loaded learning example. Optionally, predicting, through the fully-connected network, a second predictive test board corresponding to the loaded learning example based on the second integrated characterization vector of the loaded learning example.

Step S32117, if the loaded learning example is a second shape data set learning example, and at the same time, the first predicted test board corresponding to the second shape data set learning example is different from the corresponding second predicted test board, optimizing model learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model, and making the first predicted test board redetermined for the second shape data set learning example be the same as the redetermined second predicted test board after optimizing the learnable variables;

In step S32118, if the loaded learning example is a third morphology data set learning example, and at the same time, the first predicted test board corresponding to the third morphology data set learning example is the same as the corresponding second predicted test board, the model learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model are optimized, so that after the learnable variables are optimized, the first predicted test board redetermined for the third morphology data set learning example is different from the redetermined second predicted test board.

Through the debugging task, the method and the device can ensure that the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model accurately perform feature vector mining and feature integration under two types of test data and plate deformation graphs, ensure that the extracted test data feature vectors and deformation graph feature vectors and the precision of the integrated feature vectors are obtained, and further ensure that the precision of association coefficient analysis and plate test result analysis are performed based on the integrated feature vectors.

As some implementations, the method provided in the embodiments of the present application further includes:

Step S410, a third debugging sample is obtained, the third debugging sample comprises a plurality of sample form data sets and associated coefficient annotation information of the sample form data sets, the sample form data sets comprise a fourth form data set learning example and a sample data extraction mark, the associated coefficient annotation information is used for indicating a marked associated coefficient between the fourth form data set learning example and a sample data extraction mark indicating a sample target data extraction dimension;

step S420, performing characterization vector mining on the test data in the sample form data set through the first machine learning model to obtain a test data characterization vector of the sample form data set. Wherein the test data in the morphology data set learning example description pair includes test data and sample data extraction markers in the fourth morphology data set learning example;

and step S430, performing characterization vector mining on the plate deformation graph in the sample form data set through the second machine learning model to obtain a deformation graph characterization vector of the sample form data set. The sheet deformation map in the sample morphology data set can also be understood as the sheet deformation map in the fourth morphology data set learning example;

Step S440, integrating the test data characterization vector of the sample form data set and the deformation map characterization vector of the sample form data set through the first feature integration model to obtain a first integrated characterization vector of the sample form data set;

step S450, integrating the test data characterization vector of the sample form data set and the deformation map characterization vector of the sample form data set through the second feature integration model to obtain a second integrated characterization vector of the sample form data set;

step S460, combining the first integrated characterization vector of the sample morphology data set with the second integrated characterization vector of the sample morphology data set to obtain an integrated characterization vector of the sample morphology data set;

step S470, outputting, by the third machine learning model, a prediction correlation coefficient between the fourth morphology data set learning example and the sample target data extraction dimension based on the integrated characterization vector of the sample morphology data set. Predicting a correlation coefficient refers to the correlation coefficient between a fourth morphological data set learning example predicted by the third machine learning model based on the integrated characterization vector of the morphological data set and the sample target data extraction dimension;

Step S480, determining a second cost based on the prediction association coefficient and the labeling association coefficient.

Optionally, a second cost function may be set for the training process, and the predicted association coefficient and the labeled association coefficient are substituted into the second cost function to determine a second cost. Alternatively, the second cost function may be a log-likelihood function. Based on the second cost function, each machine learning model is enabled to adjust the learnable variable in a targeted manner, and each machine learning model focuses on the fuzzy uncertain sample form data set, and the through samples are not adjusted for the learnable variable.

Step S490 optimizes model learnable variables of the first machine learning model, the second machine learning model, the first feature integration model, the second feature integration model, and the third machine learning model based on the second cost until a second debug cutoff requirement is satisfied. The second debug cutoff requirement may be that the cost of the second cost function is less than a preset cost, or the number of times of debugging is less than a preset number of times, in other words, the second debug cutoff requirement is similar to the first debug cutoff requirement, only specific values are changed or unchanged, and the specific values are set according to needs.

Based on the debugging, the third machine learning model accurately obtains the association coefficient between each morphological data set and the target data extraction dimension based on the first integrated characterization vector output by the first feature integration model and the second integrated characterization vector output by the second feature integration model.

Based on the foregoing embodiments, the embodiments of the present application provide an artificial board bending creep resistance test data analysis device, where each unit included in the device and each module included in each unit may be implemented by a processor in a computer device; of course, the method can also be realized by a specific logic circuit; in practice, the processor may be a central processing unit (Central Processing Unit, CPU), microprocessor (Microprocessor Unit, MPU), digital signal processor (Digital Signal Processor, DSP) or field programmable gate array (Field Programmable Gate Array, FPGA), etc.

Fig. 2 is a schematic structural diagram of an artificial board bending creep test data analysis device according to an embodiment of the present application, and as shown in fig. 2, an artificial board bending creep test data analysis device 200 includes: a panel data acquisition module 210 for acquiring a set of morphological data sets of a target test panel and data extraction indicia for indicating a target data extraction dimension; the form data set comprises bending creep resistance test data and a plate deformation graph; the test data mining module 220 is configured to perform feature vector mining on the bending creep resistance test data and the data extraction mark through a first machine learning model to obtain test data feature vectors corresponding to each form data set; the image information mining module 230 is configured to perform feature vector mining on the sheet deformation map through a second machine learning model, so as to obtain deformation map feature vectors corresponding to each form data set; the token vector integration module 240 is configured to integrate the test data token vector and the deformation map token vector through a feature integration model to obtain an integrated token vector; a correlation coefficient determining module 250, configured to output, based on the integrated token vector through a third machine learning model, a correlation coefficient between each of the morphological data sets and a target data extraction dimension; a test result determining module 260, configured to determine, based on the integrated characterization vector, a board test result included in each of the morphological data sets through a fourth machine learning model; and a target result determining module 270, configured to determine, based on the association coefficient, a board test result of the target test board in the target data extraction dimension from board test results included in the morphological data set.

The description of the apparatus embodiments above is similar to that of the method embodiments above, with similar advantageous effects as the method embodiments. In some embodiments, functions or modules included in the apparatus provided in the embodiments of the present application may be used to perform the methods described in the embodiments of the methods, and for technical details that are not disclosed in the embodiments of the apparatus of the present application, please refer to the description of the embodiments of the methods of the present application for understanding.

It should be noted that, in the embodiment of the present application, if the above-mentioned method for analyzing bending creep test data of an artificial board is implemented in the form of a software functional module, and sold or used as a separate product, the method may also be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application may be essentially or portions contributing to the related art, and the software product may be stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read Only Memory (ROM), a magnetic disk, an optical disk, or other various media capable of storing program codes. Thus, embodiments of the present application are not limited to any specific hardware, software, or firmware, or to any combination of hardware, software, and firmware.

The embodiment of the application provides a computer device, which comprises a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor executes the program to realize part or all of the steps of the method.

Embodiments of the present application provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs some or all of the steps of the above-described method. The computer readable storage medium may be transitory or non-transitory.

Embodiments of the present application provide a computer program comprising computer readable code which, when run in a computer device, performs some or all of the steps for implementing the above method.

Embodiments of the present application provide a computer program product comprising a non-transitory computer-readable storage medium storing a computer program which, when read and executed by a computer, performs some or all of the steps of the above-described method. The computer program product may be realized in particular by means of hardware, software or a combination thereof. In some embodiments, the computer program product is embodied as a computer storage medium, in other embodiments the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.

It should be noted here that: the above description of various embodiments is intended to emphasize the differences between the various embodiments, the same or similar features being referred to each other. The above description of apparatus, storage medium, computer program and computer program product embodiments is similar to that of method embodiments described above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the apparatus, storage medium, computer program and computer program product of the present application, please refer to the description of the method embodiments of the present application.

Fig. 3 is a schematic diagram of a hardware entity of a computer device according to an embodiment of the present application, as shown in fig. 3, the hardware entity of the computer device 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program executable on the processor 1001, the processor 1001 implementing the steps in the method of any of the embodiments described above when the program is executed.

The memory 1002 stores a computer program executable on a processor, and the memory 1002 is configured to store instructions and applications executable by the processor 1001, and may also cache data (e.g., image data, audio data, voice communication data, and video communication data) to be processed or already processed by each module in the processor 1001 and the computer device 1000, which may be implemented by a FLASH memory (FLASH) or a random access memory (Random Access Memory, RAM).

The processor 1001 performs the steps of the method for analyzing bending creep test data of an artificial board according to any one of the above. The processor 1001 generally controls the overall operation of the computer device 1000.

Embodiments of the present application provide a computer storage medium storing one or more programs executable by one or more processors to implement the steps of the method for analyzing bending creep test data of an artificial board according to any of the embodiments above.

It should be noted here that: the description of the storage medium and apparatus embodiments above is similar to that of the method embodiments described above, with similar benefits as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and the apparatus of the present application, please refer to the description of the method embodiments of the present application for understanding. The processor may be at least one of a target application integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD), a programmable logic device (Programmable Logic Device, PLD), a field programmable gate array (Field Programmable Gate Array, FPGA), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, and a microprocessor. It will be appreciated that the electronic device implementing the above-mentioned processor function may be other, and embodiments of the present application are not specifically limited.

The computer storage medium/Memory may be a Read Only Memory (ROM), a programmable Read Only Memory (Programmable Read-Only Memory, PROM), an erasable programmable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable programmable Read Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic random access Memory (Ferromagnetic Random Access Memory, FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Read Only optical disk (Compact Disc Read-Only Memory, CD-ROM); but may also be various terminals such as mobile phones, computers, tablet devices, personal digital assistants, etc., that include one or any combination of the above-mentioned memories.

It should be appreciated that reference throughout this specification to "one embodiment" or "an implementation" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present application, the sequence number of each step/process described above does not mean that the execution sequence of each step/process should be determined by the function and the internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.

Alternatively, the integrated units described above may be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the related art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk.

The foregoing is merely an embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes and substitutions are intended to be covered in the protection scope of the present application.

Claims

1. A method for analyzing bending creep test data of an artificial board, which is applied to computer equipment, the method comprising: acquiring a data extraction mark and a morphological data set of a target test board, wherein the data extraction mark is used for indicating a target data extraction dimension; the form data set comprises bending creep resistance test data and a plate deformation graph; performing characterization vector mining on the bending creep resistance test data and the data extraction marks through a first machine learning model to obtain test data characterization vectors corresponding to each form data set; performing characterization vector mining on the plate deformation graph through a second machine learning model to obtain deformation graph characterization vectors corresponding to each form data set; integrating the test data characterization vector and the deformation map characterization vector through a feature integration model to obtain an integrated characterization vector; outputting association coefficients between each morphological data set and a target data extraction dimension based on the integrated characterization vector through a third machine learning model; determining a board test result included in each morphological data set based on the integrated characterization vector through a fourth machine learning model; and determining a board test result of the target test board in the target data extraction dimension from board test results included in the morphological data set based on the association coefficient.

2. The method of claim 1, wherein the performing, by the first machine learning model, the token vector mining on the bending creep test data and the data extraction markers to obtain test data token vectors corresponding to each morphological data set, comprises: combining the bending creep resistance test data in the form data set with the data extraction mark to obtain target test data corresponding to the form data set; inputting the target test data into the first machine learning model; outputting test data characterization vectors corresponding to the corresponding morphological data sets based on the target test data through the first machine learning model; the feature integration model comprises a first feature integration model and a second feature integration model; the feature integration model integrates the test data characterization vector and the deformation map characterization vector to obtain an integrated characterization vector, and the feature integration model comprises the following steps: integrating the test data characterization vector and the deformation graph characterization vector through a first feature integration model to obtain a first integration characterization vector; integrating the test data characterization vector and the deformation graph characterization vector through a second feature integration model to obtain a second integration characterization vector; and combining the first integration characterization vector and the second integration characterization vector to obtain the integration characterization vector.

3. The method of claim 2, wherein the first feature integration model comprises X sequentially connected first interactive learning modules and the second feature integration model comprises X sequentially connected second interactive learning modules, wherein X is a positive integer greater than 1; the integrating is performed by a first feature integration model based on the test data characterization vector and the deformation map characterization vector to obtain a first integration characterization vector, which comprises the following steps: obtaining i first intermediate characterization vectors and i first input array representations output by an i first interactive learning module, and obtaining i second query array representations and i second output weighting array representations output by an i second interactive learning module; wherein i is a positive integer less than X, where j=i+1; determining, by a j-th first interactive learning module, j first intermediate token vectors based on the i first input array representations, the i first intermediate token vectors, the i second query array representations, and i second output weighting array representations, and j first input array representations, j first query array representations, and j first output weighting array representations based on the i first intermediate token vectors; if j is less than X, loading the j first intermediate characterization vectors and j first input array representations to a subsequent first interactive learning module; loading j first query array representations and j first output weighting array representations to a subsequent second interactive learning module; if j=x, determining the j first intermediate token vectors as the first integrated token vectors; wherein if i is 1, the first interactive learning module determines a first intermediate token vector and a first input array representation based on the test data token vector; the first second interactive learning module determines a second query array representation and a second output weighting array representation based on the deformation map characterization vector.

4. The method according to claim 2, wherein the method further comprises: acquiring a first debugging sample, wherein the first debugging sample comprises a plurality of first form data set learning examples and annotation information of the first form data set learning examples, and the annotation information is used for indicating a test board corresponding to the first form data set learning examples; the first morphological data set learning example is obtained after a partial shielding operation is adopted; debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of first morphological data set learning examples and annotation information of the first morphological data set learning examples.

5. The method of claim 4, wherein the debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of first morphology data set learning examples and annotation information for the first morphology data set learning examples comprises: performing characterization vector mining on the test data in the first morphological data set learning example through the first machine learning model to obtain a test data characterization vector of the first morphological data set learning example; performing characterization vector mining on the plate deformation graph in the first morphological data set learning example through the second machine learning model to obtain a deformation graph characterization vector of the first morphological data set learning example; integrating the deformation map representation vector of the first morphological data set learning example and the test data representation vector of the first morphological data set learning example through the first feature integration model to obtain a first integration representation vector of the first morphological data set learning example; integrating the deformation map representation vector of the first morphological data set learning example and the test data representation vector of the first morphological data set learning example through the second feature integration model to obtain a second integration representation vector of the first morphological data set learning example; combining a first integrated characterization vector of the first morphological data set learning example with a second integrated characterization vector of the first morphological data set learning example to obtain an integrated characterization vector of the first morphological data set learning example; determining a prediction test board classification corresponding to the first morphological data set learning example based on the integrated characterization vector of the first morphological data set learning example; determining a first test board classification distribution based on the predicted test board classifications corresponding to all the first morphology data set learning examples; determining a first cost based on the first test board classification distribution and the second test board classification distribution; the second test board classification distribution is determined based on annotation information of all the first morphological data set learning examples; and optimizing model-learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model based on the first price until a first debugging cut-off requirement is met.

6. The method according to claim 2, wherein the method further comprises: obtaining a second debugging sample, wherein the second debugging sample comprises a plurality of second shape data set learning examples and a plurality of third shape data set learning examples, test data in the same second shape data set learning examples are identical to test boards corresponding to the board deformation graph, and the test data in the same third shape data set learning examples are different from the test boards corresponding to the board deformation graph; and debugging the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model based on the plurality of second morphology data set learning examples and the plurality of third morphology data set learning examples.

7. The method of claim 6, wherein the debugging the first machine learning model, the second machine learning model, the first feature integration model, and the second feature integration model based on the plurality of second morphology data set learning examples and the plurality of third morphology data set learning examples comprises: performing characterization vector mining on the test data in the loaded learning example through the first machine learning model to obtain a test data characterization vector of the loaded learning example; the loaded learning example is a second morphological data set learning example or a third morphological data set learning example; performing characterization vector mining on the plate deformation graph in the loaded learning example through the second machine learning model to obtain a deformation graph characterization vector of the loaded learning example; integrating the test data characterization vector of the loaded learning example and the deformation graph characterization vector of the loaded learning example through the first feature integration model to obtain a first integrated characterization vector of the loaded learning example; integrating the test data characterization vector of the loaded learning example and the deformation graph characterization vector of the loaded learning example through the second feature integration model to obtain a second integrated characterization vector of the loaded learning example; predicting a first prediction test board corresponding to the loaded learning example based on the first integrated characterization vector of the loaded learning example; predicting a second prediction test board corresponding to the loaded learning example based on the second integrated characterization vector of the loaded learning example; if the loaded learning example is a second state data set learning example, and meanwhile, a first prediction test board corresponding to the second state data set learning example is different from the corresponding second prediction test board, optimizing model learnable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model, and enabling the first prediction test board redetermined for the second state data set learning example to be identical to the redetermined second prediction test board after optimizing the learnable variables; if the loaded learning example is a third morphology data set learning example, and meanwhile, the first prediction test board corresponding to the third morphology data set learning example is the same as the second prediction test board corresponding to the third morphology data set learning example, optimizing model variable variables of the first machine learning model, the second machine learning model, the first feature integration model and the second feature integration model, and enabling the first prediction test board redetermined for the third morphology data set learning example to be different from the second prediction test board redetermined after optimizing the variable variables.

8. The method according to claim 2, wherein the method further comprises: obtaining a third debugging sample, wherein the third debugging sample comprises a plurality of sample morphological data sets and associated coefficient annotation information of the sample morphological data sets, the sample morphological data sets comprise a fourth morphological data set learning example and a sample data extraction mark, and the associated coefficient annotation information is used for indicating a marked associated coefficient between the fourth morphological data set learning example and a sample target data extraction dimension indicated by the sample data extraction mark; performing characterization vector mining on the test data in the sample form data set through the first machine learning model to obtain a test data characterization vector of the sample form data set; performing characterization vector mining on the plate deformation graph in the sample form data set through the second machine learning model to obtain a deformation graph characterization vector of the sample form data set; integrating the test data characterization vector of the sample form data set and the deformation map characterization vector of the sample form data set through the first feature integration model to obtain a first integrated characterization vector of the sample form data set; integrating the test data characterization vector of the sample form data set and the deformation map characterization vector of the sample form data set through the second feature integration model to obtain a second integrated characterization vector of the sample form data set; combining the first integrated characterization vector of the sample morphology data set with the second integrated characterization vector of the sample morphology data set to obtain an integrated characterization vector of the sample morphology data set; outputting, by the third machine learning model, a predictive correlation coefficient between the fourth morphology data set learning example and the sample target data extraction dimension based on the integrated characterization vector of the sample morphology data set; determining a second cost based on the predicted association coefficient and the labeling association coefficient; and optimizing model-learnable variables of the first machine learning model, the second machine learning model, the first feature integration model, the second feature integration model and the third machine learning model based on the second cost until a second debugging stop requirement is met.

9. The utility model provides an artificial board bending creep test data analysis device which characterized in that includes: the board data acquisition module is used for acquiring a data extraction mark and a morphological data set of the target test board, wherein the data extraction mark is used for indicating a target data extraction dimension; the form data set comprises bending creep resistance test data and a plate deformation graph; the test data mining module is used for carrying out characterization vector mining on the bending creep resistance test data and the data extraction marks through a first machine learning model to obtain test data characterization vectors corresponding to each form data set; the image information mining module is used for carrying out characterization vector mining on the plate deformation graph through a second machine learning model to obtain deformation graph characterization vectors corresponding to each form data set; the characterization vector integration module is used for integrating the test data characterization vector and the deformation map characterization vector through a feature integration model to obtain an integrated characterization vector; the association coefficient determining module is used for outputting association coefficients between each morphological data set and the target data extraction dimension based on the integrated characterization vector through a third machine learning model; the test result determining module is used for determining board test results included in each form data set based on the integrated characterization vector through a fourth machine learning model; and the target result determining module is used for determining a board test result of the target test board in the target data extraction dimension from board test results included in the morphological data set based on the association coefficient.

10. A computer device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that the processor implements the steps of the method of any of claims 1 to 8 when the program is executed.