Twin correction method and system for aging performance of adhesive

Abstract

The invention discloses a twinning correction method and system for the aging performance of an adhesive, and relates to the technical field related to adhesives, and the method comprises the following steps: determining a plurality of actual aging performance data by introducing a plurality of aging conditions of the adhesive; and synchronizing the plurality of aging conditions to a simulation model for performance simulation, and generating a performance simulation data set. A twin network is constructed through deep learning, and a plurality of performance error coefficients are obtained. And dynamically correcting the simulation model according to the plurality of performance error coefficients to obtain a simulation correction model. And performing data correction on the performance simulation data set through the simulation correction model, optimizing the twin network according to the performance simulation correction data set, and generating a twin optimization network. The technical problems that in the prior art, the obtaining efficiency of the aging performance of the adhesive is low, and it is difficult to accurately predict the performance of the adhesive under different aging conditions are solved. The method has the technical effects that the precision of predicting the performance of the adhesive under different aging conditions is improved, and the prediction efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of adhesive technology, specifically to a twin-based method and system for correcting the aging performance of adhesives. Background Technology

[0002] In modern industrial manufacturing, adhesives are affected by various environmental factors in different application scenarios, including temperature, humidity, ultraviolet radiation intensity, and oxygen concentration in the air. These factors can induce aging of adhesives, thereby affecting their physical properties, such as bond strength, tensile strength, and fracture toughness. Therefore, aging performance testing of adhesives has become a crucial step in assessing their service life and reliability. However, most existing adhesive aging tests rely on accelerated aging experiments conducted in laboratories. While these methods can obtain actual aging performance data, they are often time-consuming and costly.

[0003] Therefore, the existing technology has technical problems such as low acquisition efficiency and difficulty in accurately predicting the performance of adhesives under different aging conditions when obtaining the aging properties of adhesives. Summary of the Invention

[0004] This application provides a twin correction method and system for adhesive aging performance, solving the technical problems of low acquisition efficiency and difficulty in accurately predicting adhesive performance under different aging conditions in the prior art. It significantly improves the accuracy and efficiency of predicting adhesive performance under different aging conditions.

[0005] This application provides a twin-based correction method for the aging performance of adhesives. The method includes: introducing multiple aging conditions for the adhesive; performing data sensing on the adhesive based on the multiple aging conditions to determine multiple actual aging performance data; synchronizing the multiple aging conditions to a simulation model for performance simulation to generate a performance simulation dataset; constructing a twin network using deep learning; transmitting the multiple actual aging performance data and the performance simulation dataset to the twin network for processing to obtain multiple performance error coefficients; dynamically correcting the simulation model according to the multiple performance error coefficients to obtain a corrected simulation model; correcting the performance simulation dataset using the corrected simulation model; and optimizing the twin network based on the corrected performance simulation dataset to generate an optimized twin network.

[0006] In this implementation, the multiple actual aging performance data and the performance simulation dataset are transmitted to the twin network for processing to obtain multiple performance error coefficients. The method includes: the twin network comprising a data simulation network and a data monitoring network; simulation output based on the data simulation network and the performance simulation dataset to generate a simulation performance prediction result; monitoring output based on the data monitoring network and the multiple actual aging performance data to generate an actual aging performance parameter set; and comparison of the simulation performance prediction result with the actual aging performance parameter set according to a self-checking cycle to generate multiple performance error coefficients.

[0007] In the implementation, the simulated performance prediction results are compared with the actual aging performance parameter set according to a self-check cycle to generate multiple performance error coefficients. The method includes: defining multiple application scenario parameters for the adhesive based on the multiple aging conditions; setting a self-check cycle according to the multiple application scenario parameters to iterate through the simulated performance prediction results and the actual aging performance parameter set for performance matching, determining multiple sets of predicted-actual performance data pairs; defining a loss function; calculating the loss function for each set of predicted-actual performance data pairs according to the self-check cycle to generate a performance error report; performing error analysis based on the performance error report to construct an error trend graph; and identifying the error coefficients of each set of predicted-actual performance data pairs according to the error trend graph to obtain the multiple performance error coefficients.

[0008] In the implementation method, the simulation model is dynamically corrected according to the multiple performance error coefficients to obtain a corrected simulation model. The method includes: continuously analyzing the error trend chart in conjunction with the multiple performance error coefficients to determine multiple continuously changing deviation points; classifying the sources of error based on these multiple continuously changing deviation points to obtain multiple error sources; traversing the simulation model according to the multiple error sources, formulating a correction strategy, and executing the correction strategy to dynamically correct the simulation model, thereby generating the corrected simulation model.

[0009] In the implementation, the simulation model is traversed based on the multiple error sources, a correction strategy is formulated, and the correction strategy is executed to dynamically correct the simulation model, generating the corrected simulation model. The method includes: traversing the simulation model based on the multiple error sources and determining whether there are nodes in the simulation model greater than or equal to a desired simulation threshold. If there are nodes in the simulation model greater than or equal to the desired simulation threshold, multiple nodes to be corrected are extracted. Dynamic correction instructions are generated based on the multiple error sources and the multiple performance error coefficients. The dynamic correction instructions are added to the correction strategy to dynamically mark the multiple nodes to be corrected, generating error distribution data. A performance parameter space is constructed, and the performance parameter space is dynamically searched according to the error distribution data. The simulation model is updated and corrected based on the search results to obtain the corrected simulation model.

[0010] In the implementation method, the performance parameter space is constructed by: analyzing the multiple actual aging performance data to determine the spatial dimensions, discretizing the data according to the spatial dimensions to generate multiple discrete sample values; combining the discrete sample values ​​of each spatial dimension to construct a multi-dimensional parameter space; and mapping the multiple actual aging performance data to the multi-dimensional parameter space based on the multiple application scenario parameters to construct the performance parameter space.

[0011] In the implementation, the performance simulation dataset is corrected using the simulation correction model, and the Siamese network is optimized based on the corrected performance simulation dataset to generate an optimized Siamese network. The method includes: dividing the performance simulation correction dataset into an optimized training dataset, an optimized validation dataset, and an optimized test dataset; performing backpropagation on the Siamese network using the optimized training dataset; updating the weights of the data simulation network and the data monitoring network based on the backpropagation results to generate multiple weight coefficients; evaluating the data simulation network and the data monitoring network using the optimized validation dataset in conjunction with the multiple weight coefficients; dynamically adjusting the Siamese network based on the evaluation results to generate an adjusted Siamese network result; and conducting adaptive testing on the adjusted Siamese network using the optimized test dataset in conjunction with a learning rate strategy. When the test passes, the optimized Siamese network is generated.

[0012] This application also provides a twin correction system for the aging performance of adhesives, including: The actual performance data acquisition module is used to introduce multiple aging conditions of the adhesive, perform data sensing on the adhesive based on the multiple aging conditions, and determine multiple actual aging performance data.

[0013] The simulation data acquisition module is used to synchronize the multiple aging conditions to the simulation model for performance simulation and generate a performance simulation dataset.

[0014] The error coefficient acquisition module is used to construct a Siamese network through deep learning, and transmit the multiple actual aging performance data and the performance simulation dataset to the Siamese network for processing to obtain multiple performance error coefficients.

[0015] The simulation correction model acquisition module is used to dynamically correct the simulation model according to the multiple performance error coefficients to obtain the simulation correction model.

[0016] The twin optimization network acquisition module is used to correct the performance simulation dataset using the simulation correction model, optimize the twin network based on the performance simulation correction dataset, and generate the twin optimization network.

[0017] This application proposes a twin-based correction method and system for adhesive aging performance. By introducing multiple aging conditions for the adhesive, multiple actual aging performance data are determined. These multiple aging conditions are synchronized to a simulation model for performance simulation, generating a performance simulation dataset. A twin network is constructed using deep learning to obtain multiple performance error coefficients. The simulation model is dynamically corrected according to these performance error coefficients to obtain a corrected simulation model. The performance simulation dataset is then corrected using the corrected simulation model, and the twin network is optimized based on the corrected performance simulation dataset to generate an optimized twin network. This solves the technical problems of low acquisition efficiency and difficulty in accurately predicting adhesive performance under different aging conditions in existing technologies. It improves the accuracy and efficiency of predicting adhesive performance under different aging conditions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0019] Figure 1 A schematic flowchart of a twin-based method for modifying the aging properties of adhesives provided in this application embodiment; Figure 2 This is a schematic diagram of a twin-based correction system for the aging performance of adhesives provided in an embodiment of this application.

[0020] Figure labeling: Actual performance data acquisition module 11, simulation data acquisition module 12, error coefficient acquisition module 13, simulation correction model acquisition module 14, twin optimization network acquisition module 15. Detailed Implementation

[0021] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, 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 embodiments of this application only.

[0024] This application provides a twin correction method and system for the aging performance of adhesives, such as... Figure 1 As shown, the method includes: Multiple aging conditions are introduced for the adhesive. Based on these conditions, data sensing is performed on the adhesive to determine multiple actual aging performance data. These aging conditions are then synchronized to a simulation model for performance simulation, generating a performance simulation dataset. A Siamese network is constructed using deep learning. The multiple actual aging performance data and the performance simulation dataset are transmitted to the Siamese network for processing to obtain multiple performance error coefficients.

[0025] To assess the aging performance of adhesives, multiple aging conditions need to be introduced, including temperature, humidity, UV radiation intensity, oxygen concentration in the air, and corresponding time. These aging conditions correspond to the application conditions of the adhesive in actual use. Actual aging performance data of the adhesive under these conditions are tested, including bond strength, tensile strength, and fracture toughness. Furthermore, these aging conditions are synchronized to a simulation model for performance simulation, generating a performance simulation dataset. The simulation model is a virtual model built using computer software based on finite element analysis tools, used to simulate the physical processes of the adhesive under multiple aging conditions and predict the performance data of the adhesive under different aging conditions, thereby generating the performance simulation dataset.

[0026] Furthermore, a Siamese network is constructed based on deep learning technology. The Siamese network consists of two sub-networks: a data simulation network and a data monitoring network, which respectively process the performance simulation dataset and multiple actual aging performance data. The Siamese network is constructed based on actual aging performance data over a full time period under different aging conditions in historical records, as well as the performance simulation dataset over a full time period. The actual aging performance data over a full time period includes bond strength, tensile strength, and fracture toughness data recorded at each time period node. The performance simulation dataset over a full time period includes bond strength, tensile strength, and fracture toughness data output from performance simulations performed by the simulation model at each time period node. Subsequently, the actual aging performance data under different aging conditions and corresponding full time periods are input into the deep neural network model for training. The model learns the time dependence and physical characteristics in the simulation data to predict the future aging performance of the adhesive until the model training is complete, thus obtaining the data monitoring network. Using the same construction method, the performance simulation dataset under different aging conditions and corresponding full time periods is used as training data to construct the data simulation network. Furthermore, the multiple actual aging performance data and the performance simulation dataset are transmitted to the Siamese network for processing to obtain multiple performance error coefficients.

[0027] The method provided in this application embodiment further includes: the twin network comprising a data simulation network and a data monitoring network. Simulation output is generated based on the data simulation network and the performance simulation dataset to produce simulation performance prediction results. Monitoring output is generated based on the data monitoring network and the multiple actual aging performance data to produce an actual aging performance parameter set. The simulation performance prediction results are compared with the actual aging performance parameter set according to a self-check cycle to generate multiple performance error coefficients.

[0028] The twin network comprises a data simulation network and a data monitoring network. Based on the data simulation network, aging conditions are simulated using the performance simulation dataset to predict the aging process of the adhesive, outputting simulation performance prediction results. These results include predicted parameters for bond strength, tensile strength, and fracture toughness at multiple future time points. Based on the data monitoring network, aging conditions are monitored using the multiple actual aging performance data to generate an actual aging performance parameter set. This set includes parameters for bond strength, tensile strength, and fracture toughness monitored at multiple future time points. The simulation performance prediction results are compared with the actual aging performance parameter set according to a self-checking cycle to generate multiple performance error coefficients.

[0029] The method provided in this application embodiment further includes: defining multiple application scenario parameters for the adhesive based on the multiple aging conditions, and setting the self-test cycle according to the multiple application scenario parameters. Iterating through the simulation performance prediction results and the actual aging performance parameter set to perform performance matching, and determining multiple sets of prediction-actual performance data pairs. Defining a loss function, and calculating for each set of prediction-actual performance data pairs according to the self-test cycle and the loss function, generating a performance error report. Performing error analysis based on the performance error report, and constructing an error trend graph. Identifying the error coefficients of each set of prediction-actual performance data pairs according to the error trend graph, and obtaining the multiple performance error coefficients.

[0030] Multiple application scenario parameters for the adhesive are defined based on the aforementioned aging conditions. A self-inspection cycle is set according to these parameters, which represent various environmental or operational conditions the adhesive may encounter during actual use, such as aging conditions like temperature, humidity, and ultraviolet radiation. The self-inspection cycle is set by professional technicians based on the actual application scenario parameters and can be every 12 hours, 24 hours, or weekly. A higher frequency of the cycle ensures more timely correction of the simulation model. Subsequently, the simulation performance prediction results are iterated and matched with the actual aging performance parameter set to determine multiple sets of prediction-actual performance data pairs. These pairs are the pairings between the adhesive performance data predicted by the simulation model and the actual aging performance data at corresponding time points, used to analyze the difference between prediction and reality. Each prediction-actual performance data pair consists of all data pairs for a single performance parameter. Furthermore, a loss function is defined as the ratio of the difference between the predicted and actual data in multiple sets of predicted-actual performance data pairs to the actual aging performance data. The loss function is applied to each set of predicted-actual performance data according to the self-inspection cycle, generating a performance error report. This report includes the error ratio parameter for each set of predicted-actual performance data pairs within the self-inspection cycle. Error analysis is performed based on the performance error report, and an error trend graph is constructed, recording the trend of error parameters changing over time. Finally, based on the error trend graph, the error coefficient for each set of predicted-actual performance data pairs is identified. The error coefficient is the slope of the error trend graph in each self-inspection cycle.

[0031] The simulation model is dynamically corrected according to the multiple performance error coefficients to obtain a corrected simulation model. The performance simulation dataset is then corrected using this corrected model, and the Siamese network is optimized based on the corrected performance simulation dataset to generate an optimized Siamese network.

[0032] The simulation model is dynamically corrected according to the multiple performance error coefficients to obtain a corrected simulation model. Further, the performance simulation dataset is corrected using the corrected simulation model to obtain corrected data. A Siamese network is then trained, validated, and tested based on the corrected performance simulation dataset to generate an optimized Siamese network. Finally, the optimized Siamese network is used to accurately predict the adhesive performance under different aging conditions. This solves the technical problems of low acquisition efficiency and difficulty in accurately predicting adhesive performance under different aging conditions in existing technologies. It improves the accuracy and efficiency of predicting adhesive performance under different aging conditions.

[0033] The method provided in this application embodiment further includes: performing continuous analysis based on the error trend graph and the multiple performance error coefficients to determine multiple continuously changing deviation points; performing source tracing and classification based on the multiple continuously changing deviation points to obtain multiple error sources; traversing the simulation model according to the multiple error sources, formulating a correction strategy, executing the correction strategy to dynamically correct the simulation model, and generating the corrected simulation model.

[0034] The simulation model is dynamically corrected according to the multiple performance error coefficients to obtain a corrected simulation model. The method includes: continuously analyzing the error trend chart in conjunction with the multiple performance error coefficients to determine multiple continuously changing deviation points. These continuously changing deviation points are the time nodes corresponding to the self-inspection cycle in which the error coefficient increases compared to the previous self-inspection cycle. The existence of multiple continuously changing deviation points indicates that the simulation model has a systematic bias in its prediction of the corresponding parameters, requiring focused correction. Further, technical personnel identify the specific sources of error for each of the multiple continuously changing deviation points, such as environmental factors like temperature. Based on these multiple continuously changing deviation points, source tracing and classification are performed to obtain multiple error sources. These error sources include: material parameter errors: such as inaccurate settings in the simulation model regarding the physical properties of adhesive materials (e.g., elastic modulus, shear strength); environmental factor errors: such as the effects of environmental conditions like temperature and humidity not being fully reflected in the simulation model; and time-dependent errors: such as the time-dependent aging process of adhesives, where the simulation model cannot accurately capture the trend of performance changes over time. The simulation model is iterated through according to the multiple error sources, a correction strategy is formulated, the correction strategy is executed to dynamically correct the simulation model, and the corrected simulation model is generated.

[0035] The method provided in this application embodiment further includes: traversing the simulation model based on the multiple error sources, and determining whether there are nodes in the simulation model that are greater than or equal to the expected simulation threshold. If there are nodes in the simulation model that are greater than or equal to the expected simulation threshold, then multiple nodes to be corrected are extracted. Dynamic correction instructions are generated based on the multiple error sources and the multiple performance error coefficients. The dynamic correction instructions are added to the correction strategy to dynamically mark the multiple nodes to be corrected, generating error distribution data. A performance parameter space is constructed, and the performance parameter space is dynamically searched according to the error distribution data. The simulation model is updated and corrected according to the search results to obtain the corrected simulation model.

[0036] The simulation model is traversed based on the multiple error sources to formulate a correction strategy. This correction strategy is then executed to dynamically correct the simulation model, generating a corrected simulation model. The method includes: after generating multiple performance error coefficients, the system traverses the simulation model, checking each node in the model one by one, and determining whether there are any nodes in the simulation model that are greater than or equal to a desired simulation threshold. The desired simulation threshold is a pre-set error coefficient judgment standard, based on the performance error coefficients. If the error between the simulation result and the actual data is greater than this threshold, the node needs to be corrected. If there are nodes in the simulation model greater than or equal to the desired simulation threshold, multiple nodes to be corrected are extracted. These multiple nodes to be corrected are multiple data nodes within the continuously changing deviation point detection period. Further, dynamic correction instructions are generated based on the multiple error sources and the multiple performance error coefficients. These dynamic correction instructions are specific instructions used to correct the nodes to be corrected in the simulation model, including the error sources and performance error coefficients, with the aim of adjusting the node parameters to make the simulation results closer to the actual data. The dynamic correction instructions are added to the correction strategy to dynamically mark the multiple nodes to be corrected, marking the correction strategies corresponding to the multiple nodes to be corrected, and generating error distribution data. The error distribution data contains the nodes to be corrected and their corresponding correction strategies. A performance parameter space is constructed, and the performance parameter space is dynamically searched according to the error distribution data to obtain the actual aging performance data corresponding to the error distribution data in the performance parameter space. Based on the search results, the simulation model is updated and corrected according to the actual aging performance data to obtain the simulation correction model.

[0037] The method provided in this application embodiment further includes: analyzing the multiple actual aging performance data to determine the spatial dimension, discretizing the data according to the spatial dimension to generate multiple discrete sample values; combining the discrete sample values ​​of each spatial dimension to construct a multi-dimensional parameter space; and mapping the multiple actual aging performance data to the multi-dimensional parameter space based on the multiple application scenario parameters to construct the performance parameter space.

[0038] The method for constructing the performance parameter space includes: analyzing multiple actual aging performance data to determine spatial dimensions. These spatial dimensions represent key parameters affecting adhesive aging, including temperature, humidity, UV intensity, stress, and time, as well as performance parameters such as bond strength, tensile strength, and fracture toughness. Subsequently, each spatial dimension is discretized, dividing a continuous parameter range into multiple fixed values ​​as discrete sample values, generating multiple discrete sample values. Each combination of sample values ​​represents a specific aging parameter and forms a point in the parameter space. Based on the multiple application scenario parameters, the multiple actual aging performance data are mapped to the multidimensional parameter space to obtain corresponding parameter space points. The performance parameter space containing all parameter space points is then constructed based on these parameter space points.

[0039] The method provided in this application embodiment further includes: dividing the performance simulation correction dataset into an optimized training dataset, an optimized validation dataset, and an optimized test dataset. Backpropagation is performed on the Siamese network using the optimized training dataset. Based on the backpropagation results, the weights of the data simulation network and the data monitoring network are updated to generate multiple weight coefficients. The data simulation network and the data monitoring network are evaluated using the optimized validation dataset in conjunction with the multiple weight coefficients. Based on the evaluation results, the Siamese network is dynamically adjusted to generate a Siamese network adjustment result. Based on the optimized test dataset, the Siamese network adjustment result is adaptively tested in conjunction with a learning rate strategy. When the test passes, the optimized Siamese network is generated.

[0040] The performance simulation dataset is corrected using the simulation correction model. The Siamese network is then optimized based on the corrected performance simulation dataset to generate an optimized Siamese network. The method includes: partitioning the performance simulation correction dataset to determine an optimized training dataset, an optimized validation dataset, and an optimized test dataset. These datasets are used for training, validating, and testing the Siamese network, respectively, to ensure the model's accuracy and generalization ability. After the dataset partitioning, the Siamese network is trained using the optimized training dataset. During training, the backpropagation algorithm is used to update the weights of the data simulation network and the data monitoring network based on the model's output error. Finally, multiple new weight coefficients are generated to improve the Siamese network's performance. Backpropagation is an optimization algorithm that calculates the network's output error and propagates it back to each layer of the network, thereby adjusting the network's weights and gradually reducing the prediction error. The Siamese network is evaluated using the optimized validation dataset to test its performance on data not used in training. The evaluation aims to determine whether the network is overfitting or underfitting, ensuring that the model's performance on new data is consistent with its performance on training data. If network performance deteriorates during the evaluation process, the Siamese network needs to be dynamically adjusted. This involves adjusting the model's training parameters in real time based on the evaluation results of the validation set to improve its generalization ability. After dynamic adjustment of the Siamese network, adaptive testing is performed using an optimized test dataset. Adaptive testing, combined with a learning rate strategy, dynamically adjusts the model's learning rate based on its performance on the test set, ensuring the model's learning efficiency at different stages. After adaptive testing on the optimized test dataset confirms stable network performance, the final optimized Siamese network is generated. This optimized network can more accurately predict the performance of adhesives under different aging conditions and has good generalization ability, making it suitable for various application scenarios.

[0041] In the above text, refer to Figure 1 A twinning correction method for the aging performance of adhesives according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A twin-based correction system for the aging properties of adhesives according to an embodiment of the present invention is described.

[0042] A twin-based correction system for adhesive aging performance according to an embodiment of the present invention solves the technical problem in the prior art of low acquisition efficiency and difficulty in accurately predicting adhesive performance under different aging conditions when acquiring adhesive aging performance data. It improves the accuracy and efficiency of predicting adhesive performance under different aging conditions. The twin-based correction system for adhesive aging performance includes: an actual performance data acquisition module 11, a simulation data acquisition module 12, an error coefficient acquisition module 13, a simulation correction model acquisition module 14, and a twin optimization network acquisition module 15.

[0043] The actual performance data acquisition module 11 is used to introduce multiple aging conditions of the adhesive, perform data sensing on the adhesive based on the multiple aging conditions, and determine multiple actual aging performance data.

[0044] The simulation data acquisition module 12 is used to synchronize the multiple aging conditions to the simulation model for performance simulation and generate a performance simulation dataset.

[0045] Error coefficient acquisition module 13 is used to construct a twin network through deep learning, transmit the multiple actual aging performance data and the performance simulation dataset to the twin network for processing, and obtain multiple performance error coefficients.

[0046] The simulation correction model acquisition module 14 is used to dynamically correct the simulation model according to the multiple performance error coefficients to obtain the simulation correction model.

[0047] The twin optimization network acquisition module 15 is used to correct the performance simulation dataset using the simulation correction model, optimize the twin network based on the performance simulation correction dataset, and generate the twin optimization network.

[0048] The specific configuration of the error coefficient acquisition module 13 will be described in detail below. The error coefficient acquisition module 13 may further include: transmitting the multiple actual aging performance data and the performance simulation dataset to the twin network for processing to obtain multiple performance error coefficients. The method includes: the twin network comprising a data simulation network and a data monitoring network; performing simulation output based on the data simulation network and the performance simulation dataset to generate a simulation performance prediction result; performing monitoring output based on the data monitoring network and the multiple actual aging performance data to generate an actual aging performance parameter set; and comparing the simulation performance prediction result with the actual aging performance parameter set according to a self-check cycle to generate multiple performance error coefficients.

[0049] The specific configuration of the error coefficient acquisition module 13 will be described in detail below. The error coefficient acquisition module 13 further includes: comparing the simulated performance prediction results with the actual aging performance parameter set according to a self-test cycle to generate multiple performance error coefficients. The method includes: defining multiple application scenario parameters for the adhesive based on the multiple aging conditions; setting the self-test cycle according to the multiple application scenario parameters to iterate through the simulated performance prediction results and the actual aging performance parameter set for performance matching, and determining multiple sets of predicted-actual performance data pairs; defining a loss function; calculating the loss function for each set of predicted-actual performance data pairs according to the self-test cycle to generate a performance error report; performing error analysis based on the performance error report to construct an error trend graph; and identifying the error coefficients of each set of predicted-actual performance data pairs according to the error trend graph to obtain the multiple performance error coefficients.

[0050] The specific configuration of the simulation correction model acquisition module 14 will be described in detail below. The simulation correction model acquisition module 14 may further include: dynamically correcting the simulation model according to the plurality of performance error coefficients to obtain a simulation correction model. The method includes: continuously analyzing the error trend graph in conjunction with the plurality of performance error coefficients to determine multiple continuously changing deviation points; tracing and classifying the sources of error based on the multiple continuously changing deviation points to obtain multiple error sources; traversing the simulation model according to the multiple error sources, formulating a correction strategy, and executing the correction strategy to dynamically correct the simulation model to generate the simulation correction model.

[0051] The specific configuration of the simulation correction model acquisition module 14 will be described in detail below. The simulation correction model acquisition module 14 further includes: traversing the simulation model based on the multiple error sources, formulating a correction strategy, executing the correction strategy to dynamically correct the simulation model, and generating the simulation correction model. The method includes: traversing the simulation model based on the multiple error sources and determining whether there are nodes in the simulation model greater than or equal to a desired simulation threshold. If there are nodes in the simulation model greater than or equal to the desired simulation threshold, multiple nodes to be corrected are extracted. Dynamic correction instructions are generated based on the multiple error sources and the multiple performance error coefficients. The dynamic correction instructions are added to the correction strategy to dynamically mark the multiple nodes to be corrected, generating error distribution data. A performance parameter space is constructed, and the performance parameter space is dynamically searched according to the error distribution data. The simulation model is updated and corrected based on the search results to obtain the simulation correction model.

[0052] The specific configuration of the simulation correction model acquisition module 14 will be described in detail below. The simulation correction model acquisition module 14 further includes: constructing the performance parameter space, the method of which includes: analyzing the multiple actual aging performance data to determine the spatial dimension, discretizing according to the spatial dimension to generate multiple discrete sample values; combining the discrete sample values ​​of each spatial dimension to construct a multi-dimensional parameter space; and mapping the multiple actual aging performance data to the multi-dimensional parameter space based on the multiple application scenario parameters to construct the performance parameter space.

[0053] The specific configuration of the Siamese optimization network acquisition module 15 will be described in detail below. The Siamese optimization network acquisition module 15 further includes: correcting the performance simulation dataset using the simulation correction model, optimizing the Siamese network based on the performance simulation correction dataset, and generating a Siamese optimization network. The method includes: dividing the performance simulation correction dataset into an optimized training dataset, an optimized validation dataset, and an optimized test dataset; performing backpropagation on the Siamese network using the optimized training dataset, updating the weights of the data simulation network and the data monitoring network based on the backpropagation results, and generating multiple weight coefficients; evaluating the data simulation network and the data monitoring network in conjunction with the multiple weight coefficients based on the optimized validation dataset, dynamically adjusting the Siamese network based on the evaluation results, and generating a Siamese network adjustment result; and conducting adaptive testing on the Siamese network adjustment result in conjunction with a learning rate strategy based on the optimized test dataset. When the test passes, the Siamese optimization network is generated.

[0054] The adhesive aging performance twin correction system provided in this embodiment of the invention can execute the adhesive aging performance twin correction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0055] While this application makes various references to certain modules in the system according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved. In addition, the specific names of each functional unit are only for easy distinction and are not intended to limit the scope of protection of this invention.

[0056] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for twinning correction of the aging properties of adhesives, characterized in that, The method includes: Multiple aging conditions for the adhesive are introduced, and data sensing is performed on the adhesive based on the multiple aging conditions to determine multiple actual aging performance data. The multiple aging conditions are synchronized to the simulation model for performance simulation, generating a performance simulation dataset. A twin network is constructed using deep learning, and the multiple actual aging performance data and the performance simulation dataset are transmitted to the twin network for processing to obtain multiple performance error coefficients. The simulation model is dynamically corrected according to the multiple performance error coefficients to obtain a corrected simulation model. The performance simulation dataset is corrected using the simulation correction model, and the Siamese network is optimized based on the performance simulation correction dataset to generate an optimized Siamese network.

2. The method for twinning correction of adhesive aging performance as described in claim 1, characterized in that, The method involves transmitting the multiple actual aging performance data and the performance simulation dataset to the twin network for processing to obtain multiple performance error coefficients. The twin network includes a data simulation network and a data monitoring network; Based on the data simulation network and the performance simulation dataset, simulation output is generated to produce simulation performance prediction results. Based on the data monitoring network, the monitoring output is combined with the multiple actual aging performance data to generate an actual aging performance parameter set; According to the self-inspection cycle, the simulation performance prediction results are compared with the actual aging performance parameter set to generate multiple performance error coefficients.

3. The method for twinning correction of adhesive aging performance as described in claim 2, characterized in that, According to the self-inspection cycle, the simulation performance prediction results are compared with the actual aging performance parameter set to generate multiple performance error coefficients. The method includes: Based on the aforementioned multiple aging conditions, multiple application scenario parameters for the adhesive are defined, and the self-inspection cycle is set according to the aforementioned multiple application scenario parameters; The simulation performance prediction results are traversed and matched with the actual aging performance parameter set to determine multiple sets of prediction-actual performance data pairs. Define a loss function, and calculate the loss function for each pair of predicted and actual performance data according to the self-testing period to generate a performance error report; Based on the performance error report, perform error analysis and construct an error trend graph; The error coefficients of each predicted-actual performance data pair are identified according to the error trend graph to obtain the plurality of performance error coefficients.

4. The method for twinning correction of adhesive aging performance as described in claim 3, characterized in that, The simulation model is dynamically corrected according to the multiple performance error coefficients to obtain a corrected simulation model. The method includes: By continuously analyzing the error trend chart and the multiple performance error coefficients, multiple continuously changing deviation points are identified. Based on the multiple continuously changing deviation points, source tracing and classification are performed to obtain multiple sources of error; The simulation model is iterated through according to the multiple error sources, a correction strategy is formulated, the correction strategy is executed to dynamically correct the simulation model, and the corrected simulation model is generated.

5. The method for twinning correction of adhesive aging performance as described in claim 4, characterized in that, The simulation model is iterated through based on the multiple error sources, a correction strategy is formulated, and the correction strategy is executed to dynamically correct the simulation model, generating the corrected simulation model. The method includes: Based on the multiple error sources, the simulation model is traversed, and it is determined whether there are nodes in the simulation model that are greater than or equal to the expected simulation threshold. If there are nodes in the simulation model that are greater than or equal to the expected simulation threshold, then multiple nodes to be corrected are extracted. Dynamic correction instructions are generated based on the multiple error sources and the multiple performance error coefficients. The dynamic correction instructions are added to the correction strategy to perform dynamic marking on the multiple nodes to be corrected, generating error distribution data; A performance parameter space is constructed, and the performance parameter space is dynamically retrieved according to the error distribution data. The simulation model is updated and corrected based on the retrieval results to obtain the corrected simulation model.

6. The method for twinning correction of adhesive aging performance as described in claim 5, characterized in that, The method for constructing the performance parameter space includes: Based on the analysis of the multiple actual aging performance data, the spatial dimension is determined, and the data is discretized according to the spatial dimension to generate multiple discrete sample values. By combining the discrete sampled values ​​of each spatial dimension, a multidimensional parameter space is constructed. Based on the multiple application scenario parameters, the multiple actual aging performance data are mapped to the multidimensional parameter space to construct the performance parameter space.

7. The method for twinning correction of adhesive aging properties as described in claim 2, characterized in that, The performance simulation dataset is corrected using the simulation correction model, and the Siamese network is optimized based on the corrected performance simulation dataset to generate an optimized Siamese network. The method includes: The performance simulation correction dataset is divided into optimized training dataset, optimized validation dataset, and optimized test dataset. The Siamese network is backpropagated using the optimized training dataset. Based on the backpropagation results, the weights of the data simulation network and the data monitoring network are updated to generate multiple weight coefficients. The data simulation network and the data monitoring network are evaluated based on the optimized verification dataset, and the multiple weight coefficients are combined. The Siamese network is dynamically adjusted based on the evaluation results to generate the Siamese network adjustment results. Based on the optimized test dataset, the Siamese network adjustment results are combined with a learning rate strategy for adaptive testing. When the test passes, the optimized Siamese network is generated.

8. A twin-based correction system for the aging properties of adhesives, characterized in that, The system includes: The actual performance data acquisition module is used to introduce multiple aging conditions of the adhesive, perform data sensing on the adhesive based on the multiple aging conditions, and determine multiple actual aging performance data. The simulation data acquisition module is used to synchronize the multiple aging conditions to the simulation model for performance simulation and generate a performance simulation dataset. The error coefficient acquisition module is used to construct a Siamese network through deep learning, and transmit the multiple actual aging performance data and the performance simulation dataset to the Siamese network for processing to obtain multiple performance error coefficients. The simulation correction model acquisition module is used to dynamically correct the simulation model according to the multiple performance error coefficients to obtain the simulation correction model. The twin optimization network acquisition module is used to correct the performance simulation dataset using the simulation correction model, optimize the twin network based on the performance simulation correction dataset, and generate the twin optimization network.

Images