Efficient migration method and device for electromagnetic structure performance prediction, equipment and medium

Abstract

The invention discloses an electromagnetic structure performance prediction-oriented efficient migration method, device and equipment and a medium, and the method comprises the steps: collecting a sample in a selected source scene, and obtaining an electromagnetic structure sample; simulating the electromagnetic structure sample by using industrial simulation software to obtain annotation information; training a proxy model by using a sample with annotation information for performance prediction of the electromagnetic structure; collecting a sample on the target scene; and migrating the proxy model trained on the source scene to the target scene by using the sample collected on the target scene. The electromagnetic field simulation information is injected into the agent model, label information expansion is achieved, the sample utilization efficiency is improved, and the precision and efficiency of the electromagnetic performance prediction agent model are remarkably improved. Besides, when the electromagnetic scene is changed, the agent model in the source scene can be migrated to the target scene without simulating the sample of the target scene again, so that the model migration efficiency in the complex scene is improved.

Description

Technical Field

[0001] The present invention relates to the field of performance prediction of electromagnetic structures, and particularly to an efficient migration method, device, equipment and medium for performance prediction of electromagnetic structures. Background Art

[0002] In the field of performance prediction of electromagnetic structures, traditional model training methods usually rely on a large amount of performance label data obtained through electromagnetic simulation calculations. Due to the complexity of electromagnetic structures and the high resource consumption of simulation calculations, this method incurs high data costs. Especially in the application of transfer learning, in order to ensure the performance prediction accuracy between the source scenario and the target scenario, a large number of labeled samples are required, which further increases the data cost. Therefore, improving the utilization efficiency of sampling samples has become an urgent problem to be solved.

[0003] In addition, there are many electromagnetic structure scenarios, and variable performance indicators may be faced under different application scenarios. The traditional "one-to-one" optimization method optimizes the changing scenarios from scratch and is difficult to cope with frequent scenario changes. If it can be ensured that the model can maintain performance and generalization ability in the new scenario, the number of simulations can be significantly reduced and the optimization efficiency can be improved. Therefore, how to quickly complete the migration of the performance predictor in complex and variable scenarios is also an unsolved problem.

[0004] The above two problems reveal the key limitations of the current migration technology in the field of electromagnetic structure performance prediction. Therefore, there is an urgent need for a new method that can improve the sample utilization efficiency and the efficiency of model migration in complex scenarios without increasing the data cost. Summary of the Invention

[0005] To at least partly solve one of the technical problems existing in the prior art, an object of the present invention is to provide an efficient migration method, device, equipment and medium for performance prediction of electromagnetic structures.

[0006] The first technical solution adopted by the present invention is as follows:

[0007] An efficient migration method for performance prediction of electromagnetic structures, comprising the following steps:

[0008] Collect samples in a selected source scenario to obtain electromagnetic structure samples;

[0009] Use industrial simulation software to simulate the electromagnetic structure samples to obtain annotation information;

[0010] Efficiently train a surrogate model using the samples with annotation information for performance prediction of electromagnetic structures;

[0011] Collect samples in the target scenario;

[0012] Using the samples collected in the target scenario, the surrogate model trained in the source scenario is migrated to the target scenario.

[0013] Further, collecting samples in the selected source scenario includes:

[0014] Using a preset scenario as the source scenario; the preset scenario includes a chip radiation suppression structure scenario or a miniaturized DGS common-mode filtering structure scenario;

[0015] In the source scenario, a random sampling method is used to obtain samples, and electromagnetic structure samples are obtained for subsequent training and construction of the surrogate model.

[0016] Further, simulating the electromagnetic structure samples using industrial simulation software to obtain annotation information includes:

[0017] Using a VBS script to batch simulate the collected electromagnetic structure samples;

[0018] After the simulation is completed, the VBS script continues to automatically export the performance curves and electromagnetic field information of each electromagnetic structure sample as annotation information.

[0019] Further, training the surrogate model using the samples with annotation information includes:

[0020] The electromagnetic structure samples collected in the source scenario are respectively input into two paths for performance curve prediction;

[0021] In the first path, the electromagnetic structure sample passes through an encoder and a decoder to predict the electromagnetic field information, and then the predicted electromagnetic field information passes through the decoder to predict the performance curve;

[0022] In the second path, the electromagnetic structure sample directly predicts the performance curve after passing through the encoder and the decoder;

[0023] Weights are assigned to the performance curves predicted by the two paths, and they are weighted and integrated to obtain the final predicted performance curve;

[0024] Using the sample data with annotation information to train the surrogate model, thereby adjusting the parameters of the surrogate model and the weights of the two paths to obtain the surrogate model trained and completed in the source scenario.

[0025] Further, collecting samples in the target scenario includes:

[0026] Using a preset scenario as the target scenario for migration; the preset scenario includes a chip radiation suppression structure scenario or a miniaturized DGS common-mode filtering structure scenario;

[0027] In the target scenario, a random sampling method is used to obtain samples.

[0028] Further, the method of migrating the surrogate model trained on the source scenario to the target scenario by using the samples collected on the target scenario includes:

[0029] Inputting the sample data without labeled information collected under the target scenario into the surrogate model trained on the source scenario to obtain the predicted performance curves under two paths;

[0030] Calculating the KL divergence loss of the predicted performance curves under two paths for predicting the result consistency perception;

[0031] Screening unreliable samples based on the prediction consistency and applying the entropy maximization loss to reduce the prediction confidence of the model;

[0032] Applying the entropy minimization loss to reliable samples to improve the prediction confidence of the model, thereby adjusting the surrogate model parameters and migrating them to the target scenario.

[0033] The second technical solution adopted by the present invention is:

[0034] An efficient migration device for electromagnetic structure performance prediction, including:

[0035] A first acquisition module, configured to acquire samples under a selected source scenario to obtain electromagnetic structure samples;

[0036] A simulation annotation module, configured to simulate the electromagnetic structure samples by using industrial simulation software to obtain annotation information;

[0037] A model training module, configured to train a surrogate model by using the samples with annotation information for electromagnetic structure performance prediction;

[0038] A second acquisition module, configured to acquire samples on the target scenario;

[0039] A target migration module, configured to migrate the surrogate model trained on the source scenario to the target scenario by using the samples collected on the target scenario.

[0040] The third technical solution adopted by the present invention is:

[0041] An electronic device, the electronic device includes a processor and a memory, and at least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement an efficient migration method for electromagnetic structure performance prediction as described above.

[0042] The fourth technical solution adopted by the present invention is:

[0043] A computer-readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement an efficient migration method for electromagnetic structure performance prediction as described above.

[0044] The fifth technical solution adopted by the present invention is:

[0045] A computer program product or a computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the above-mentioned efficient migration method for electromagnetic structure performance prediction.

[0046] The beneficial effects of the present invention are as follows: The present invention injects electromagnetic field simulation information into the surrogate model, realizes the expansion of label information, improves the sample utilization efficiency, and significantly improves the accuracy and efficiency of the electromagnetic performance prediction surrogate model. In addition, when the electromagnetic scenario changes, the present invention can migrate the surrogate model in the source scenario to the target scenario without re-simulating the samples in the target scenario, and improves the efficiency of model migration in complex scenarios without increasing the data cost. Description of the Drawings

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following introduces the drawings of the related technical solutions in the embodiments of the present invention or the prior art. It should be understood that the drawings introduced below are only for conveniently and clearly expressing some embodiments of the technical solutions in the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative efforts.

[0048] Figure 1 It is a flowchart of the steps of an efficient migration method for electromagnetic structure performance prediction in an embodiment of the present invention;

[0049] Figure 2 It is a schematic diagram of random sampling in an embodiment of the present invention;

[0050] Figure 3 It is a schematic diagram of obtaining sample annotation information in an embodiment of the present invention;

[0051] Figure 4 It is a schematic diagram of an electromagnetic structure performance prediction surrogate model in an embodiment of the present invention;

[0052] Figure 5 It is a schematic diagram of migrating the surrogate model to the target scenario in an embodiment of the present invention. Detailed implementation manners

[0053] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary only for explaining the present invention and should not be construed as limiting the present invention. For the step numbers in the following embodiments, they are only set for the convenience of description and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0054] In the description of the present invention, it should be understood that for the orientation description, such as up, down, front, back, left, right, etc., the orientation or positional relationship indicated is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.

[0055] In the description of the present invention, the meaning of "a number of" is one or more, the meaning of "a plurality of" is two or more, and understandings such as "greater than", "less than", "exceeding", etc. do not include the present number, and understandings such as "above", "below", "within", etc. include the present number. If there is a description of "first" and "second", it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features.

[0056] In the description of the present invention, unless otherwise clearly defined, words such as "set", "installed", "connected", etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0057] Term explanation:

[0058] VBS script: A VBS script is a script file written based on the Visual Basic Script language and is used for automated task processing. In the present invention, the VBS script is used for batch simulation operations. By calling the industrial simulation software interface, loading electromagnetic structure samples, performing simulations, exporting data, and recording task logs, the efficiency and consistency are improved.

[0059] Embodiment 1

[0060] As Figure 1 shown, this embodiment provides an efficient migration method for predicting the performance of electromagnetic structures, which can improve the sample utilization efficiency and the efficiency of model migration in complex scenarios without increasing the data cost. The method specifically includes the following steps:

[0061] S1. Collect samples under the selected source scenarios to obtain electromagnetic structure samples.

[0062] In some embodiments, step S1 specifically includes the following steps:

[0063] S1-1: First, select scenarios such as the chip radiation suppression structure scenario and the miniaturized DGS common-mode filtering structure scenario as the source scenarios for migration. The user should ensure the availability of the scenarios, and all samples collected on the source scenarios will be used for the construction of the subsequent surrogate model.

[0064] S1-2: Use random sampling to obtain samples in the source scenarios for the subsequent training and construction of the surrogate model. As Figure 2 shown, Figure 2 is a schematic diagram of random sampling.

[0065] S2. Use industrial simulation software to simulate the electromagnetic structure samples to obtain annotation information.

[0066] The embodiments of the present invention relate to using industrial simulation software to simulate the collected samples to obtain the necessary sample annotation information. Specifically, the method of this embodiment is mainly applied to specific scenarios such as chip radiation suppression structures and miniaturized DGS common-mode filtering structures. In this process, the VBS script technology is particularly introduced to batch process simulation operations and data export, thereby improving the efficiency and automation of the simulation process.

[0067] As an alternative implementation, step S2 specifically includes the following steps:

[0068] S2-1: Use the VBS script to batch simulate the selected electromagnetic structure samples, which cover various scenarios such as chip radiation suppression structures and miniaturized DGS common-mode filtering structures. During the simulation process, the VBS script automatically calls the computing resources of the simulation software (such as HFSS, CST) and executes the simulation tasks according to parameters such as the preset frequency range, grid density, and boundary conditions. The operation of the VBS script ensures the automation and batch execution ability of the tasks.

[0069] After the simulation is completed, the VBS script continues to automatically export the simulation results, including the following annotation information:

[0070] Performance curves: Reflect key performance indicators such as scattering parameters (such as S11, S21), insertion loss, and bandwidth of the samples at different frequencies.

[0071] Electromagnetic field information: Includes electromagnetic field intensity distribution, electric field vector direction, magnetic field distribution, and power density distribution. These data are saved in graphical and numerical forms to provide high-quality annotated samples for subsequent model training.

[0072] This information is used as labeled data for subsequent model training and validation, ensuring the accuracy and consistency of the data, as Figure 3 shown.

[0073] S3. Efficiently train a surrogate model using samples with labeled information for performance prediction of electromagnetic structures.

[0074] Exemplarily, after the samples collected in the source scenario are input into the surrogate model, the performance curve will be predicted through two paths, and the final prediction result is obtained by weighted integration of the prediction results of the two paths. See Figure 4 . In Path 1, it is necessary to first predict the electromagnetic field information after passing the electromagnetic structure through the encoder and decoder, and then pass the predicted electromagnetic field information through the decoder to predict the performance curve. In Path 2, the performance curve is directly predicted after the electromagnetic structure passes through the encoder and decoder. Assign corresponding weights to the performance curves predicted by the two paths, and the final predicted performance curve is obtained by weighted integration.

[0075] As an alternative implementation, in Path 1, the feature information of the electromagnetic structure sample is extracted after passing through the encoder, and then the electromagnetic field information, such as electric field distribution, magnetic field intensity, etc., is predicted through the decoder. Then, these electromagnetic field information are further used to calculate the performance curve through the decoder.

[0076] In Path 2, the electromagnetic structure sample directly passes through the joint processing of the encoder and decoder, skipping the intermediate step of electromagnetic field information, and directly predicting the performance curve.

[0077] The main differences between these two paths are as follows:

[0078] 1) Processing flow: Path 1 needs to generate intermediate electromagnetic field information, while Path 2 directly skips this step.

[0079] 2) Data dependence: Path 1 is more dependent on the accuracy of the intermediate field information, so it is suitable for scenarios with rich labeled information; Path 2 is more simplified and suitable for scenarios with limited data resources but high computational efficiency requirements.

[0080] 3) Curve difference: Path 1 can usually capture more detailed performance change details, while Path 2 may have a slight loss in accuracy but has an advantage in computational speed.

[0081] The final prediction result is weighted and integrated in the following way: weights w 1 and w 2 are respectively assigned to the performance curves generated by Path 1 and Path 2, and the weight assignment is determined according to the validation error during model training. For example, if the error of Path 1 on the training data is less than that of Path 2, a higher weight is assigned. The final performance curve is obtained through the following formula:

[0082] P final = w 1 ·P 1 + w 2 ·P 2

[0083] where P 1 and P 2 are the prediction curves of Path 1 and Path 2 respectively.

[0084] Using the sample data with annotation information collected in the above steps, train the surrogate model, thereby adjusting the parameters of the surrogate model and the weights of the two paths, and obtaining the surrogate model trained on the source scenario.

[0085] S4. Collect samples in the target scenario.

[0086] As an optional implementation manner, step S4 specifically includes the following steps:

[0087] S4-1: Select scenarios such as the chip radiation suppression structure scenario and the miniaturized DGS common-mode filtering structure scenario as the target scenarios for migration. The user should ensure the availability of the scenarios, and all samples collected in the target scenarios will be used for the subsequent migration of the surrogate model.

[0088] S4-2: Use random sampling in the target scenario to obtain samples for the subsequent migration of the surrogate model.

[0089] S5. Use the samples collected in the target scenario to migrate the surrogate model trained on the source scenario to the target scenario.

[0090] See Figure 5 As an optional implementation manner, step S5 specifically includes the following steps:

[0091] S5-1: Input the sample data without annotation information in the target scenario collected in step S4 into the surrogate model trained on the source scenario, and obtain the predicted performance curves under the two paths.

[0092] S5-2: Calculate the KL divergence loss of the predicted performance curves under the two paths for predicting the result consistency perception.

[0093] S5-3: Screen unreliable samples based on the prediction consistency, apply the entropy maximization loss to reduce the prediction confidence of the model. On the contrary, apply the entropy minimization loss to reliable samples to improve the prediction confidence of the model, thereby adjusting the parameters of the surrogate model and migrating it to the target scenario.

[0094] Specifically, the Kullback-Leibler divergence (KL divergence) is used to measure the distribution difference of the performance curve prediction results under two paths, and its calculation formula is as follows:

[0095]

[0096] Where: P 1(i) and P 2(i) are the predicted probability distributions of Path 1 and Path 2 at the i-th frequency point, respectively; The smaller the value, the closer the prediction results of the two paths are. During the model training process, the KL divergence is used for consistency perception to guide the model optimization objective.

[0097] As an alternative implementation, the process of screening samples based on consistency is as follows:

[0098] Calculate the consistency index: According to the KL divergence of the predicted performance curves of the two paths, set a threshold ∈. When holds, the sample is considered reliable; otherwise, it is an unreliable sample.

[0099] Apply the entropy maximization loss: For unreliable samples, increase the entropy of the prediction distribution and reduce the confidence of the model in them. The formula for entropy is as follows:

[0100]

[0101] Maximizing the entropy helps reduce the model's dependence on unreliable samples. Apply the entropy minimization loss: For reliable samples, minimize the entropy of their prediction distributions to improve the confidence of the model.

[0102] Generally speaking, in the field of performance prediction of electromagnetic structures, traditional model training methods usually rely on a large amount of performance label data obtained through electromagnetic simulation calculations. Due to the complexity of electromagnetic structures and the high resource consumption of simulation calculations, this method incurs high data costs. The present invention innovatively proposes an efficient transfer method for electromagnetic structure performance prediction, injecting electromagnetic field simulation information into the surrogate model, realizing the expansion of label information, improving the sample utilization efficiency, and significantly improving the accuracy and efficiency of the electromagnetic performance prediction surrogate model. In addition, there are many electromagnetic structure scenarios, and the performance indicators may vary under different application scenarios. The traditional "one-to-one" optimization method optimizes the changing scenarios from scratch and is difficult to cope with frequent scenario changes. At the same time, when the electromagnetic scenario changes in the present invention, it is not necessary to re-simulate the samples of the target scenario, and the surrogate model in the source scenario can be transferred to the target scenario. In summary, the present invention can improve the sample utilization efficiency and the efficiency of model transfer in complex scenarios without increasing the data cost.

[0103] Embodiment 2

[0104] This embodiment provides an efficient migration device for electromagnetic structure performance prediction, including:

[0105] A first acquisition module, configured to acquire samples in a selected source scenario to obtain electromagnetic structure samples;

[0106] A simulation annotation module, configured to use industrial simulation software to simulate the electromagnetic structure samples to obtain annotation information;

[0107] A model training module, configured to use the samples with annotation information to train a surrogate model for electromagnetic structure performance prediction;

[0108] A second acquisition module, configured to acquire samples in a target scenario;

[0109] A target migration module, configured to use the samples acquired in the target scenario to migrate the surrogate model trained in the source scenario to the target scenario.

[0110] Since this device is an efficient migration device for electromagnetic structure performance prediction in an embodiment of the present invention, and the principle of solving problems by this device is similar to that of the method, the implementation of this device can refer to the implementation process of the above method embodiment, and the repeated parts will not be described again.

[0111] Embodiment 3

[0112] An embodiment of the present invention further provides an electronic device, where the electronic device includes a processor and a memory, and at least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement Figure 1 An efficient migration method for electromagnetic structure performance prediction as shown.

[0113] It can be understood that the memory may include a random access memory (RAM), and may also include a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the above method embodiments, etc.; the data storage area may store data created according to the use of the server, etc.

[0114] The processor may include one or more processing cores. The processor uses various interfaces and circuits to connect various parts within the entire server. By running or executing instructions, programs, code sets, or instruction sets stored in the memory, and by invoking data stored in the memory, it performs various functions of the server and processes data. Optionally, the processor may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor may integrate a combination of one or several of a central processing unit (CPU) and a modem, etc. Among them, the CPU mainly processes the operating system, application programs, etc.; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor and may be implemented separately by a single chip.

[0115] Since this electronic device is the electronic device corresponding to an efficient migration method for electromagnetic structure performance prediction in an embodiment of the present invention, and the principle by which this electronic device solves problems is similar to that of this method, the implementation of this electronic device can refer to the implementation process of the above method embodiment, and repeated parts will not be elaborated.

[0116] Embodiment 4

[0117] An embodiment of the present invention further provides a computer-readable storage medium, in which at least one instruction, at least one segment of program, code set, or instruction set is stored, and the at least one instruction, the at least one segment of program, the code set, or the instruction set is loaded and executed by a processor to implement Figure 1 an efficient migration method for electromagnetic structure performance prediction as shown.

[0118] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. This program can be stored in a computer-readable storage medium, which includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM), or other optical disc memories, magnetic disk memories, tape memories, or any other computer-readable medium capable of carrying or storing data.

[0119] Since this storage medium is the storage medium corresponding to an efficient migration method for electromagnetic structure performance prediction in the embodiments of the present invention, and the principle of solving problems by this storage medium is similar to that of this method, the implementation of this storage medium can refer to the implementation process of the above method embodiments, and the repeated parts will not be described again.

[0120] Embodiment 5

[0121] In some possible implementation manners, each aspect of the method in the embodiments of the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a computer device, the program code is used to cause the computer device to execute the steps of an efficient migration method for electromagnetic structure performance prediction according to various exemplary implementation manners described above in this specification. Among them, the executable computer program code or "code" for executing each embodiment can be written in high-level programming languages such as C, C++, C#, Smalltalk, Java, JavaScript, Visual Basic, structured query language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0122] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0123] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0124] The above embodiments are only for illustrating the technical concept and features of the present invention, and the purpose is to enable those of ordinary skill in the art to understand the content of the present invention and implement it accordingly, and it cannot be used to limit the protection scope of the present invention. All equivalent changes or modifications made according to the essence of the content of the present invention should be covered within the protection scope of the present invention.

Claims

1. An efficient migration method for electromagnetic structure performance prediction, characterized in that: The following steps are involved: Collect samples in the selected source scene to obtain electromagnetic structure samples; Use industrial simulation software to simulate electromagnetic structure samples and obtain annotation information; Using samples with labeled information to train proxy models for performance prediction of electromagnetic structures; Collect samples on the target scene; The proxy model trained on the source scene is transferred to the target scene using samples collected on the target scene.

2. The efficient migration method for electromagnetic structure performance prediction according to claim 1, characterized in that: The collecting of samples in the selected source scene includes: A preset scene is used as a source scene; the preset scene includes a chip radiation suppression structure scene or a miniaturized DGS common mode filtering structure scene; Random sampling is used to obtain samples in the source scene to obtain electromagnetic structure samples for subsequent training and construction of the proxy model.

3. The efficient migration method for electromagnetic structure performance prediction according to claim 1, characterized in that: The use of industrial simulation software to simulate the electromagnetic structure sample to obtain annotation information includes: Use VBS scripts to perform batch simulation on the collected electromagnetic structure samples; After completing the simulation, the VBS script continues to automatically export the performance curves and electromagnetic field information of each electromagnetic structure sample as annotation information.

4. The efficient migration method for electromagnetic structure performance prediction according to claim 1, characterized in that: The method of training the proxy model using samples with labeled information includes: The electromagnetic structure samples collected on the source scene are input into two paths for performance curve prediction; In the first path, the electromagnetic structure sample is passed through an encoder and a decoder to predict the electromagnetic field information, and then the predicted electromagnetic field information is passed through a decoder to predict the performance curve; In the second path, the electromagnetic structure sample is passed through the encoder and decoder to directly predict the performance curve; The performance curves predicted by the two paths are assigned corresponding weights, and the final predicted performance curve is obtained by weighted integration. The proxy model is trained using sample data with labeled information, so as to adjust the parameters of the proxy model and the weights of the two paths to obtain a proxy model trained on the source scene.

5. The efficient migration method for electromagnetic structure performance prediction according to claim 1, characterized in that: The collecting of samples on the target scene includes: A preset scenario is used as a target scenario for migration; the preset scenario includes a chip radiation suppression structure scenario or a miniaturized DGS common mode filtering structure scenario; Use random sampling to obtain samples in the target scene.

6. The efficient migration method for electromagnetic structure performance prediction according to claim 1, characterized in that: The method of migrating the proxy model trained on the source scene to the target scene by using samples collected on the target scene includes: The sample data without labeled information collected in the target scene is input into the proxy model trained in the source scene to obtain the performance curves predicted under the two paths; Calculate the KL divergence loss of the performance curves predicted under the two paths for prediction result consistency perception; Filter unreliable samples based on prediction consistency, apply entropy maximization loss, and reduce the prediction confidence of the model; For reliable samples, entropy minimization loss is applied to improve the prediction confidence of the model, thereby adjusting the proxy model parameters and migrating to the target scene.

7. An efficient migration device for electromagnetic structure performance prediction, characterized in that: include: A first acquisition module is used to collect samples in a selected source scene to obtain electromagnetic structure samples; A simulation annotation module is used to simulate electromagnetic structure samples using industrial simulation software to obtain annotation information; A model training module is used to train a proxy model using samples with labeled information for performance prediction of electromagnetic structures; A second acquisition module is used to collect samples on the target scene; The target migration module is used to migrate the proxy model trained on the source scene to the target scene using samples collected on the target scene.

8. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

Images