An automatic prediction method for laser parameter combination for laser-induced graphene

Through machine learning, the series network is constructed, and laser parameter combination is predicted to achieve laser-induced graphene, which solves the problem of low efficiency in laser parameter combination discovery in the prior art and improves the accuracy and efficiency of electromagnetic shielding performance.

CN118520356BActive Publication Date: 2024-11-22HANGZHOU HENGYING TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410979419.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-22
Estimated Expiration
2044-07-22

AI Technical Summary

Technical Problem

The combination of laser parameters with high electromagnetic shielding performance in the prior art mainly depends on accidental discovery or artificial selection, with low generation efficiency and it is difficult to quickly find laser parameters that meet the target electromagnetic shielding performance.

Method used

Machine learning technology is used to build a series network, predict laser parameter combinations through preset target electromagnetic shielding performance data, and laser induced graphene is achieved by predicting laser parameter combinations, improving the discovery efficiency and accuracy of high-value laser parameter combinations.

Benefits of technology

The rapid search for laser parameter combinations based on the set electromagnetic shielding effect is realized, which improves the discovery efficiency and accuracy of the combination of high-value parameter combinations in laser-induced graphene, and reduces design costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118520356B_ABST
    Figure CN118520356B_ABST
Patent Text Reader

Abstract

The present invention discloses an automatic prediction method for laser parameter combinations for laser-induced graphene, and relates to the field of electromagnetic shielding materials. By obtaining preset target electromagnetic shielding effectiveness data; obtaining a pre-trained series network, the series network is used to determine the predicted laser parameter combination according to the electromagnetic shielding effectiveness data; outputting the predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination; inputting the target electromagnetic shielding effectiveness data into the series network, obtaining the predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realizing laser-induced graphene through the predicted laser parameter combination. The present invention adopts machine learning technology to construct a series network, and the series network can predict the corresponding laser parameter combination based on the set electromagnetic shielding effectiveness, thereby improving the efficiency of discovering high-value laser parameter combinations; it can also predict the corresponding electromagnetic shielding effectiveness based on the predicted laser parameter combination, and verify the accuracy and reliability of the predicted laser parameter combination.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of electromagnetic shielding materials, and in particular to an automatic prediction method for laser parameter combination used for laser-induced graphene. Background Art

[0002] Laser induced graphene (LIG) has become one of the most active topics in graphene research in recent years. In LIG manufacturing, the mechanism of graphene formation depends largely on the characteristics of the laser. Ultrashort pulse lasers can control the photon energy to be highly concentrated in the local area of ​​polyimide (PI), thereby producing a rapid heating process and precisely adjusting the generated temperature, breaking chemical bonds and rearranging carbon atoms into graphene. At the same time, the instantaneous generation of gas products will produce a high-pressure environment, which helps to form LIG with a porous structure. Laser induced graphene is controlled by the synergistic effect of multiple key laser parameters. At present, the combination of laser parameters with high electromagnetic shielding performance mainly relies on accidental discovery or artificial selection, and the generation efficiency is low.

[0003] Therefore, the existing technology still needs to be improved and developed. Summary of the invention

[0004] The technical problem to be solved by the present invention is that, in view of the above-mentioned defects of the prior art, a method for automatically predicting laser parameter combinations for laser-induced graphene is provided, aiming to solve the problem that the combination of laser parameters with high electromagnetic shielding performance in the prior art mainly relies on accidental discovery or human selection, and has low generation efficiency.

[0005] The technical solution adopted by the present invention to solve the problem is as follows:

[0006] In a first aspect, an embodiment of the present invention provides a method for automatically predicting a combination of laser parameters for laser-induced graphene, wherein the method comprises:

[0007] Obtaining preset target electromagnetic shielding effectiveness data;

[0008] Acquire a pre-trained series network, wherein the series network is used to: determine a predicted laser parameter combination according to the electromagnetic shielding effectiveness data; and output corresponding predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination;

[0009] The target electromagnetic shielding effectiveness data is input into the series network to obtain a predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and laser induced graphene is realized through the predicted laser parameter combination.

[0010] In one embodiment, the series network comprises:

[0011] An inverse design network is used to determine the predicted laser parameter combination based on the electromagnetic shielding effectiveness data;

[0012] The forward prediction network is used to determine the predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination.

[0013] In one implementation, the forward prediction network is a pre-trained network, and the inverse design network is a network obtained by reverse deduction based on the pre-trained forward prediction network.

[0014] In one embodiment, the forward prediction network includes an input layer, a hidden layer, and an output layer; the forward prediction network uses a rectified linear unit as an activation function and uses an adaptive learning rate optimization algorithm as an optimizer.

[0015] In one embodiment, the method further comprises:

[0016] Acquire a preset database, wherein the database includes a plurality of laser parameter combinations;

[0017] Each of the laser parameter combinations is divided into a training data set and a verification data set corresponding to the series network according to a preset ratio.

[0018] In one embodiment, the method further comprises:

[0019] Obtaining a data set outside of the database;

[0020] The prediction accuracy of the trained forward prediction network is determined according to the data set.

[0021] In one embodiment, the method further comprises:

[0022] Experimentally verifying the predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data;

[0023] The database is updated according to the experimental verification results.

[0024] In a second aspect, an embodiment of the present invention further provides a laser parameter combination automatic prediction device for laser induced graphene, wherein the device comprises:

[0025] A data acquisition module, used to acquire preset target electromagnetic shielding effectiveness data;

[0026] A network acquisition module, used to acquire a pre-trained series network, wherein the series network is used to: determine a predicted laser parameter combination according to electromagnetic shielding effectiveness data; and output corresponding predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination;

[0027] The network prediction module is used to input the target electromagnetic shielding effectiveness data into the series network to obtain a predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realize laser induced graphene through the predicted laser parameter combination.

[0028] In a third aspect, an embodiment of the present invention further provides a terminal, wherein the terminal includes a memory and one or more processors; the memory stores one or more programs; the program includes instructions for executing any of the above-described methods for automatically predicting laser parameter combinations for laser-induced graphene; and the processor is used to execute the program.

[0029] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a plurality of instructions stored thereon, wherein the instructions are suitable for being loaded and executed by a processor to implement the steps of any of the above-described methods for automatically predicting laser parameter combinations for laser-induced graphene.

[0030] Beneficial effects of the present invention: The embodiment of the present invention obtains preset target electromagnetic shielding effectiveness data; obtains a pre-trained series network, the series network is used to determine the predicted laser parameter combination according to the electromagnetic shielding effectiveness data; outputs the predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination; inputs the target electromagnetic shielding effectiveness data into the series network, obtains the predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realizes laser induced graphene through the predicted laser parameter combination. The present invention uses machine learning technology to construct a series network, and the series network can predict the corresponding laser parameter combination based on the set electromagnetic shielding effectiveness, thereby improving the efficiency of discovering high-value laser parameter combinations; it can also predict the corresponding electromagnetic shielding effectiveness based on the predicted laser parameter combination, and verify the accuracy and reliability of the predicted laser parameter combination. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 It is a basic flow chart of a method for automatically predicting a combination of laser parameters for laser-induced graphene provided in an embodiment of the present invention.

[0033] Figure 2 3 is a SEM image (scanning electron microscope image) of LIG under different laser parameters provided by an embodiment of the present invention.

[0034] Figure 3 It is a detailed flowchart of a method for automatically predicting a combination of laser parameters for laser-induced graphene provided in an embodiment of the present invention.

[0035] Figure 4 It is a schematic diagram of the forward prediction results of the forward prediction network provided by an embodiment of the present invention.

[0036] Figure 5 It is a schematic diagram comparing the input SET (total electromagnetic shielding effectiveness) of LIG-GLPD1 (Generative laser parameter design) provided by an embodiment of the present invention and the predicted SET of the series network.

[0037] Figure 6 It is a schematic diagram comparing the input SET of LIG-GLPD2 provided by an embodiment of the present invention and the predicted SET of the series network.

[0038] Figure 7 It is a schematic diagram of the electromagnetic shielding effectiveness of LIG-GLPD1 and LIG-GLPD2 prepared experimentally according to an embodiment of the present invention.

[0039] Figure 8 It is a module schematic diagram of a device for automatically predicting laser parameter combinations for laser-induced graphene provided by an embodiment of the present invention.

[0040] Fig. 9 It is a principle block diagram of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The present invention discloses a method for automatically predicting a combination of laser parameters for laser-induced graphene. To make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0042] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0043] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0044] In view of the above-mentioned defects of the prior art, the present invention provides a method for automatically predicting laser parameter combinations for laser-induced graphene, such as Figure 1 As shown, the method includes:

[0045] Step S100, obtaining preset target electromagnetic shielding effectiveness data;

[0046] Step S200, obtaining a pre-trained series network, wherein the series network is used to: determine a predicted laser parameter combination according to electromagnetic shielding effectiveness data; and output corresponding predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination;

[0047] Step S300: input the target electromagnetic shielding effectiveness data into the series network to obtain a predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realize laser induced graphene through the predicted laser parameter combination.

[0048] In short, the goal of this embodiment is to quickly find a laser parameter combination that can achieve the set electromagnetic shielding effectiveness based on the set electromagnetic shielding effectiveness, and then apply the predicted laser parameter combination to laser-induced graphene, so as to obtain a graphene material that meets the needs of users. Specifically, in order to improve the efficiency of discovering high-value laser parameter combinations in laser-induced graphene, this embodiment uses machine learning technology to pre-build and train a series network. The input data and output data of the series network are both electromagnetic shielding effectiveness data. In actual application scenarios, users can set an electromagnetic shielding effectiveness according to their own needs, that is, to obtain target electromagnetic shielding effectiveness data. The target electromagnetic shielding effectiveness data is input into the trained series network, and the series network first predicts the corresponding laser parameter combination based on the input electromagnetic shielding effectiveness data, that is, to obtain the predicted laser parameter combination. Then, based on the predicted laser parameter combination, the corresponding electromagnetic shielding effectiveness is evaluated, that is, the predicted electromagnetic shielding effectiveness data is obtained. Finally, the predicted electromagnetic shielding effectiveness data is output. The series network in this embodiment can not only predict the laser parameter combination that meets the electromagnetic shielding effectiveness set by the user, so as to improve the efficiency of discovering high-value laser parameter combinations in laser-induced graphene; it can also predict the corresponding electromagnetic shielding effectiveness based on the predicted laser parameter combination, so as to further verify the accuracy and reliability of the predicted laser parameter combination.

[0049] In one implementation, the predicted laser parameter combination includes several femtosecond laser parameters, for example, the femtosecond laser parameters may be laser wavelength, pulse duration, repetition frequency, laser average power, scanning speed, laser pulse, and energy density.

[0050] Laser wavelength: The laser wavelength can be near infrared 1030nm, visible light 515nm, or ultraviolet 343nm. Depending on the wavelength, the light absorption efficiency of PI is also different. When the laser beam irradiates the target material, a small part of the laser energy is absorbed and converted into heat energy to heat the target material, while the remaining energy is reflected or scattered. Based on the light absorption of PI at 1030nm, 515nm, and 343nm, it is judged that the formation of LIG is attributed to multiphoton absorption. 2 Compared with continuous laser wavelengths, shorter light sources can achieve higher resolution.

[0051] Pulse duration: Pulse duration refers to the pulse width, which can be divided into continuous wave, long pulse and ultrashort pulse. The pulse duration can be ultrashort pulse, adjustable from 290fs to 10ps; pulse lasers can emit photons in a short time to produce higher peak power and quickly reach the required carbonization and graphitization temperature; the size of the heat-affected zone is proportional to the square root of the pulse duration. When the pulse duration is less than the time required to distribute the pulse energy in the form of heat on the substrate, the thermal effect can be ignored.

[0052] Repetition rate: Repetition rate refers to the number of pulses emitted per second, and is one of the key parameters for controlling the heat accumulation process during the formation of laser-induced graphene. The repetition rate can be adjusted from 0Hz to 1111kHz; a high repetition rate can reduce thermal damage and burning to the substrate while converting PI to LIG without damaging the substrate; a low repetition rate causes the photothermal energy deposited by each pulse to disappear before heat accumulation, thus failing to produce LIG.

[0053] Laser average power: The choice of laser power depends on the carbon precursor, the desired LIG properties and other control parameters; the laser power must exceed a certain threshold to produce a photothermal effect, and appropriately increasing the laser average power can produce higher quality LIG;

[0054] However, when the laser power is too large, it will lead to excessive carbonization, increased resistance, reduced reflection loss, and also destroy the porous structure of LIG, reducing absorption loss.

[0055] Scanning speed: When the scanning speed is increased to a certain level based on the determined laser power, rapid preparation can be achieved; when the scanning speed is too fast, the contact time will be too short, resulting in insufficient base temperature for carbonization and failure to form LIG.

[0056] Laser pulse (burst), energy density (density): When working in pulse mode, because the pulse profile is not continuous, the energy density parameter is introduced to control the SCA (laser pulse device or system control application that facilitates the control of the laser processing process) to determine the carbonization unit, and the laser pulse parameter is introduced to control how many laser pulses the SCA outputs at a time; follow speed×density×burst<laser operating frequency to determine the size of the spot and the size of the heat-affected zone.

[0057] like Figure 2 As shown in the figure, the influence of these laser parameters on the thermal effect of the material can be more intuitively observed through the scanning electron microscope image.

[0058] In one implementation, the series network includes:

[0059] An inverse design network is used to determine the predicted laser parameter combination based on the electromagnetic shielding effectiveness data;

[0060] The forward prediction network is used to determine the predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination.

[0061] In order to realize the two prediction functions of laser parameter combination and electromagnetic shielding effectiveness, this embodiment uses two networks with different functions to construct a series network, namely, an inverse design network and a forward prediction network. Specifically, the input data of the inverse design network is the target electromagnetic shielding effectiveness data set by the user, and the output data is the predicted laser parameter combination that meets the target electromagnetic shielding effectiveness data. The inverse design network and the forward prediction network are in a cascade relationship, that is, the output data of the inverse design network is the input data of the forward prediction network, and the output data of the forward prediction network is the predicted electromagnetic shielding effectiveness data obtained based on the evaluation of the predicted laser parameter combination. After being trained with a large amount of training data, the inverse design network and the forward prediction network in this embodiment have learned the complex mapping relationship between input and output. Therefore, when both networks can accurately predict, it is possible to automatically and efficiently predict high-value laser parameter combinations in laser-induced graphene, and further evaluate and test the electromagnetic shielding effectiveness of the predicted laser parameter combination, thereby improving the accuracy and reliability of the prediction results. In addition, the series network obtained by integrating the inverse design network and the forward prediction network can also solve the non-uniqueness problem that an electromagnetic shielding effectiveness SET (dB) may have multiple corresponding laser parameters, thereby improving the convergence of the neural network.

[0062] For example, as shown in Table 1 below:

[0063] Table 1. Examples of predicted laser parameter combinations

[0064]

[0065] Combining Table 1 and Figure 5 By selecting the total electromagnetic shielding efficiency (SET) as high as 30.15 dB as the input data of the series network, the SET of the predicted laser parameter combination LIG-GLPD1 reaches 30.17 dB; combined with Table 1 and Figure 6 By selecting SET up to 31.29dB as the input data of the series network, the predicted laser parameter combination LIG-GLPD2 has a SET of 31.31dB. This proves that the method provided by the present application can automatically and quickly predict the laser parameter combination that meets the set electromagnetic shielding effectiveness.

[0066] In one implementation, the forward prediction network is a pre-trained network, and the inverse design network is a network obtained by reverse deduction based on the pre-trained forward prediction network.

[0067] Specifically, this embodiment first establishes a forward prediction network (FPN), which can accurately predict the corresponding electromagnetic shielding performance according to the input laser parameter combination. The forward prediction network is a pre-trained network trained based on a private data set. Reverse design / inverse design is performed through the forward prediction network to construct an inverse design network with the opposite input-output relationship of the forward prediction network. Finally, the inverse design network and the forward prediction network are integrated to obtain a series network.

[0068] In one implementation, the inverse design network includes a plurality of hidden layers, each of which includes a plurality of neurons. For example, the inverse design network may be composed of three hidden layers, with 1324, 811, and 528 neurons, respectively. The predicted laser parameter combination may be obtained from an interface layer, which is located between the output of the hidden layer of the inverse design network and the input of the forward prediction network.

[0069] In one implementation, the forward prediction network includes an input layer, a hidden layer, and an output layer; the forward prediction network uses a rectified linear unit as an activation function and uses an optimization algorithm of an adaptive learning rate as an optimizer.

[0070] Specifically, in order to realize the function of the forward prediction network, the present embodiment uses an input layer, a hidden layer and an output layer to construct a forward prediction network. For example, the forward prediction network can be composed of 1 input layer, 3 hidden layers and 1 output layer, with 7, 528, 811, 1324 and 334 neurons respectively. In addition, in the present embodiment, ReLU (Rectified Linear Unit) is used as an activation function in the forward prediction network, and an adaptive learning rate optimization algorithm (Adam) is used as an optimizer. The forward prediction network is iteratively trained using the optimizer until a preset number of iterations, such as 50 times, is reached. During the iterative training process, the continuous decrease in the mean square error (MSE) of the forward prediction network indicates that the performance of the forward prediction network (FPN) is gradually being optimized.

[0071] In one implementation, the method further includes:

[0072] Acquire a preset database, wherein the database includes a plurality of laser parameter combinations;

[0073] Each of the laser parameter combinations is divided into a training data set and a verification data set corresponding to the series network according to a preset ratio.

[0074] In order to obtain an accurate neural network model, this embodiment pre-establishes a database with rich and high-quality data sets. Specifically, Python (a computer program programming language) first randomly generates a total of 10,000 different laser parameter combinations, and a database is constructed based on these laser parameter combinations. Then the entire data set in the database is randomly divided into a training data set and a validation data set, for example, the split can be made in a ratio of 4:1. The forward prediction network is trained and verified using the training data set and the validation data set, and the MSE of the forward prediction network in the training data set and the validation data set is 4.9×10 -5 and 5.4×10 -5 .

[0075] In one implementation, the method further includes:

[0076] Obtaining a data set outside of the database;

[0077] The prediction accuracy of the trained forward prediction network is determined according to the data set.

[0078] Specifically, in order to visualize the prediction accuracy, this embodiment also selects a new data set that has never been seen by the forward prediction network from outside the data set in the database to test the prediction accuracy of the forward prediction network. Figure 4 As shown, the electromagnetic shielding effectiveness curve of the forward prediction network obtained using the new data set can reflect that the prediction result is close to the expected value, indicating that the prediction accuracy of the forward prediction network is high. Compared with the traditional experimental preparation + parameter measurement method, this embodiment can quickly obtain the spectrum, saving a lot of time and economic costs.

[0079] In one implementation, the method further includes:

[0080] Experimentally verifying the predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data;

[0081] The database is updated according to the experimental verification results.

[0082] Specifically, Figure 3 As shown, in order to verify and optimize the predicted effect of the series network, this embodiment also conducts experimental verification on the target electromagnetic shielding effectiveness data and its corresponding predicted laser parameter combination, and feeds back the experimental verification results to the database to update the database.

[0083] For example, two sets of candidate parameters of LIG-GLPD1 and LIG-GLPD2 were verified, laser-induced graphene was prepared, and the scattering parameters (S11 and S21) of LIG-GLPD1 and LIG-GLPD2 in the frequency range of 8.2-12.4 GHz (x-band) were measured at room temperature using a vector network analyzer. Before testing S11 and S21, the equipment was calibrated using the TRL calibration method (Thru-through calibration, Reflect-reflection calibration, Line-delay line calibration). The electromagnetic interference shielding effectiveness (EMI SE) of the sample was calculated by the following formula:

[0084] R=(S 11 ) 2 (1);

[0085] T=(S 21 ) 2 (2);

[0086] A+R+T=1(3);

[0087] SER = -10log(1-R) ​​(4);

[0088] SEA= =-10log(T / (1-R))(5);

[0089] SET = -10logT(6);

[0090] Among them, R, T, A are reflection coefficient, transmission coefficient and absorption coefficient respectively. SER, SEA and SET are reflection, absorption and total shielding effectiveness respectively. The sample size used for testing is 22.9×10.2mm 2 .

[0091] like Figure 7 As shown in the figure, the total shielding effectiveness SET of LIG-GLPD1 can reach 30.17dB, and the total shielding effectiveness SET of LIG-GLPD2 can reach 31.31dB. The results show that in the 8.2-12.4GHz band, the SET of the experimentally prepared laser parameters is in good agreement with the predicted SET. Both sets of laser parameters trained based on machine learning can generate laser-induced graphene, which can be fed back to the database later.

[0092] In one implementation, the training method of the series network includes:

[0093] Acquiring training data, wherein the training data includes an actual laser parameter combination and actual electromagnetic shielding effectiveness data corresponding to the actual laser parameter combination;

[0094] Inputting the actual electromagnetic shielding effectiveness data into the series network to obtain the predicted laser parameter combination and predicted electromagnetic shielding effectiveness data corresponding to the training data;

[0095] Determining a parameter prediction update value according to a difference between the actual laser parameter combination and the predicted laser parameter combination corresponding to the training data;

[0096] Determining an effectiveness prediction update value according to a difference between the actual electromagnetic shielding effectiveness data and the predicted electromagnetic shielding effectiveness data corresponding to the training data;

[0097] The network parameters of the series network are updated according to the parameter prediction update value and the performance prediction update value.

[0098] In simple terms, the training goal of the neural network is to converge the gap between the predicted data and the actual data, so the gap between the predicted data and the actual data can be used to guide the update of network parameters. Since the series network has two prediction functions, one is the function of predicting the laser parameter combination, and the other is the function of predicting the electromagnetic shielding effectiveness. Therefore, in order to effectively improve the overall prediction accuracy of the series network, this embodiment will calculate an update value based on the two prediction functions respectively, and update the network parameters of the series network through the two update values. Specifically, for each training data, this embodiment will calculate the parameter prediction update value according to the difference between the corresponding actual laser parameter combination and the predicted laser parameter combination, and the parameter prediction update value can reflect the quality of the series network's parameter prediction on the training data; and this embodiment will also calculate the performance prediction update value according to the difference between the corresponding actual electromagnetic shielding effectiveness data and the predicted electromagnetic shielding effectiveness data, and the performance prediction update value can reflect the quality of the series network's performance prediction on the training data. The parameter prediction update value and the performance prediction update value can be mutually corrected, tested, and referenced. For example, when the parameter prediction update value is large, it means that the series network has a large error in parameter prediction. Since the parameter prediction step is before the performance prediction step, the accuracy of the parameter prediction of the series network will affect its accuracy in performance prediction to a certain extent, and the data change trend of the performance prediction update value should be consistent with the parameter prediction update value. If the data change trend of the parameter prediction update value is different from that of the performance prediction update value, it means that one of the two may be inaccurate, thereby achieving mutual verification between the data. Finally, the parameter prediction update value and the performance prediction update value are used together to guide the network parameter update, which can improve the prediction accuracy of the series network faster and more accurately.

[0099] In one implementation, updating the network parameters of the series network according to the parameter prediction update value and the performance prediction update value includes:

[0100] When the actual electromagnetic shielding effectiveness data is greater than the predicted electromagnetic shielding effectiveness data corresponding to the training data, the parameter prediction update value is a parameter prediction loss value, and the effectiveness prediction update value is an effectiveness prediction loss value;

[0101] Correcting the performance prediction loss value according to the parameter prediction loss value to obtain a corrected performance prediction loss value;

[0102] The network parameters of the series network are updated according to the parameter prediction loss value and the modified performance prediction loss value.

[0103] Specifically, when the actual electromagnetic shielding effectiveness data is greater than the predicted electromagnetic shielding effectiveness data, it means that the currently predicted laser parameter combination does not meet the preset electromagnetic shielding effectiveness requirements. At this time, the parameter prediction update value reflects the error of the series network in parameter prediction, so it is defined as the parameter prediction loss value. In the case of errors in parameter prediction, it is difficult for the series network to predict the effectiveness that is consistent with the actual electromagnetic shielding effectiveness data. At this time, the effectiveness prediction update value reflects the error of the series network in effectiveness prediction, so it is defined as the effectiveness prediction loss value. The subsequent training goal of the series network is to converge the parameter prediction loss value and the effectiveness prediction loss value. In addition, since the parameter prediction loss value is too large, the effectiveness prediction loss value will be falsely high, that is, the height of the effectiveness prediction loss value does not entirely depend on the accuracy of the series network in effectiveness prediction, and it will also be affected by the parameter prediction loss value. Therefore, in order to truly reflect the accuracy of the series network in effectiveness prediction, this embodiment will correct the effectiveness prediction loss value based on the parameter prediction loss value. Finally, the network parameters of the series network are updated based on the parameter prediction loss value and the corrected performance prediction loss value, which can more accurately improve the prediction accuracy of the series network.

[0104] An example of a method for correcting the performance prediction loss value according to the parameter prediction loss value to obtain a corrected performance prediction loss value is as follows:

[0105] Obtain the pre-built mapping relationship between parameter prediction loss value and interference value;

[0106] Determining a target interference value according to the current parameter prediction loss value and the mapping relationship;

[0107] The target interference value is subtracted from the performance prediction loss value to obtain the modified performance prediction loss value.

[0108] In one implementation, updating the network parameters of the series network according to the parameter prediction update value and the performance prediction update value includes:

[0109] When the actual electromagnetic shielding effectiveness data is less than or equal to the predicted electromagnetic shielding effectiveness data corresponding to the training data, the effectiveness prediction update value is an effectiveness prediction reward value;

[0110] The network parameters of the series network are updated according to a weighted sum of the parameter prediction update value and the performance prediction reward value, wherein the weight of the performance prediction reward value is greater than the weight of the parameter prediction update value.

[0111] Specifically, when the actual electromagnetic shielding effectiveness data is less than or equal to its corresponding predicted electromagnetic shielding effectiveness data, it indicates that the currently predicted laser parameter combination has met or even exceeded the preset electromagnetic shielding effectiveness requirements. In order to make the series network iterate in the direction of predicting a laser parameter combination with better performance, this embodiment defines the performance prediction update value at this time as the performance prediction reward value, and assigns a higher weight value to the performance prediction reward value when calculating the network parameters used to update the series network. The subsequent training goal of the series network is to improve the performance prediction reward value.

[0112] The advantages of the present invention are that the interaction between femtosecond laser and material is nonlinear, and valuable laser parameter combinations in laser-induced graphene often rely on accidental discoveries. There are great limitations in artificially discovering high-value laser parameter combinations, and there is also a great selection bias in artificial selection, which increases the difficulty of discovering high-value laser parameter combinations. The present application can automatically, efficiently and quickly predict LIG with corresponding laser parameter combinations based on set electromagnetic shielding performance through machine learning, thereby reducing the cost of LIG design. Experimental verification has proved that the mean square error (MSE) of the forward prediction network established in the present invention on the validation set is as small as 5.4×10-5. The series network can successfully prepare LIG with a SET greater than 30dB in the range of 8.2~12.4GHz.

[0113] Based on the above embodiments, the present invention also provides a laser parameter combination automatic prediction device for laser induced graphene, such as Figure 8 As shown, the device comprises:

[0114] Data acquisition module 01, used to acquire preset target electromagnetic shielding effectiveness data;

[0115] The network acquisition module 02 is used to acquire a pre-trained series network, wherein the series network is used to: determine a predicted laser parameter combination according to the electromagnetic shielding effectiveness data; and output the corresponding predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination;

[0116] The network prediction module 03 is used to input the target electromagnetic shielding effectiveness data into the series network to obtain a predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realize laser induced graphene through the predicted laser parameter combination.

[0117] Based on the above embodiment, the present invention further provides a terminal, whose principle block diagram can be shown as follows: Fig. 9 As shown. The terminal includes a processor, a memory, a network interface, and a display screen connected through a system bus. Among them, the processor of the terminal is used to provide computing and control capabilities. The memory of the terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the terminal is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an automatic prediction method for laser parameter combinations for laser induced graphene is implemented. The display screen of the terminal can be a liquid crystal display screen or an electronic ink display screen.

[0118] Those skilled in the art will understand that Fig. 9 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the scheme of the present invention, and does not constitute a limitation on the terminal to which the scheme of the present invention is applied. The specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0119] In one implementation, the memory of the terminal stores one or more programs, and is configured to be executed by one or more processors. The one or more programs include instructions for performing a method for automatically predicting laser parameter combinations for laser-induced graphene.

[0120] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0121] In summary, the present invention discloses a method for automatically predicting laser parameter combinations for laser-induced graphene, the method obtaining preset target electromagnetic shielding effectiveness data; obtaining a pre-trained series network, wherein the series network is used to: determine the predicted laser parameter combination according to the target electromagnetic shielding effectiveness data; output the predicted electromagnetic shielding effectiveness data corresponding to the predicted laser parameter combination according to the predicted laser parameter combination; input the target electromagnetic shielding effectiveness data into the series network to obtain the predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data. The present invention uses machine learning technology to construct a series network, which can not only predict the corresponding laser parameter combination based on the set electromagnetic shielding effectiveness, thereby improving the efficiency of discovering high-value laser parameter combinations; it can also predict the corresponding electromagnetic shielding effectiveness based on the predicted laser parameter combination, and further verify the accuracy and reliability of the predicted laser parameter combination.

[0122] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for automatically predicting laser parameter combinations for laser-induced graphene, characterized in that: The method comprises: Obtaining preset target electromagnetic shielding effectiveness data; Acquire a pre-trained series network, wherein the series network is used to: determine a predicted laser parameter combination according to the electromagnetic shielding effectiveness data; and output corresponding predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination; Inputting the target electromagnetic shielding effectiveness data into the series network to obtain a predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realizing laser induced graphene through the predicted laser parameter combination; The series network comprises: An inverse design network is used to determine the predicted laser parameter combination based on the electromagnetic shielding effectiveness data; A forward prediction network, used to determine predicted electromagnetic shielding effectiveness data based on predicted laser parameter combinations; The forward prediction network is a pre-trained network, and the inverse design network is a network obtained by reverse deduction based on the pre-trained forward prediction network; The inverse design network includes a plurality of hidden layers, each of which includes a plurality of neurons; the predicted laser parameter combination is obtained from an interface layer, and the interface layer is located between the output of the hidden layer of the inverse design network and the input of the forward prediction network; The training method of the series network includes: Acquiring training data, wherein the training data includes an actual laser parameter combination and actual electromagnetic shielding effectiveness data corresponding to the actual laser parameter combination; Inputting the actual electromagnetic shielding effectiveness data into the series network to obtain the predicted laser parameter combination and predicted electromagnetic shielding effectiveness data corresponding to the training data; Determining a parameter prediction update value according to a difference between the actual laser parameter combination and the predicted laser parameter combination corresponding to the training data; Determining an effectiveness prediction update value according to a difference between the actual electromagnetic shielding effectiveness data and the predicted electromagnetic shielding effectiveness data corresponding to the training data; When the actual electromagnetic shielding effectiveness data is greater than the predicted electromagnetic shielding effectiveness data corresponding to the training data, the parameter prediction update value is a parameter prediction loss value, and the effectiveness prediction update value is an effectiveness prediction loss value; Obtain the pre-built mapping relationship between parameter prediction loss value and interference value; Determine a target interference value according to the current parameter prediction loss value and the mapping relationship; Subtracting the target interference value from the performance prediction loss value to obtain a modified performance prediction loss value; updating the network parameters of the series network according to the parameter prediction loss value and the modified performance prediction loss value; When the actual electromagnetic shielding effectiveness data is less than or equal to the predicted electromagnetic shielding effectiveness data corresponding to the training data, the effectiveness prediction update value is an effectiveness prediction reward value; The network parameters of the series network are updated according to a weighted sum of the parameter prediction update value and the performance prediction reward value, wherein the weight of the performance prediction reward value is greater than the weight of the parameter prediction update value.

2. The method for automatically predicting laser parameter combinations for laser-induced graphene according to claim 1, characterized in that: The forward prediction network includes an input layer, a hidden layer and an output layer; the forward prediction network adopts a rectified linear unit as an activation function and adopts an optimization algorithm of an adaptive learning rate as an optimizer.

3. The method for automatically predicting laser parameter combinations for laser-induced graphene according to claim 2, characterized in that: The method further comprises: Acquire a preset database, wherein the database includes a plurality of laser parameter combinations; Each of the laser parameter combinations is divided into a training data set and a verification data set corresponding to the series network according to a preset ratio.

4. The method for automatically predicting laser parameter combinations for laser-induced graphene according to claim 3, characterized in that: The method further comprises: Obtaining a data set outside of the database; The prediction accuracy of the trained forward prediction network is determined according to the data set.

5. The method for automatically predicting laser parameter combinations for laser-induced graphene according to claim 3, characterized in that: The method further comprises: Experimentally verifying the predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data; The database is updated according to the experimental verification results.

6. A laser parameter combination automatic prediction device for laser induced graphene, characterized in that: The device comprises: A data acquisition module, used to acquire preset target electromagnetic shielding effectiveness data; A network acquisition module, used to acquire a pre-trained series network, wherein the series network is used to: determine a predicted laser parameter combination according to electromagnetic shielding effectiveness data; and output corresponding predicted electromagnetic shielding effectiveness data according to the predicted laser parameter combination; A network prediction module, used for inputting the target electromagnetic shielding effectiveness data into the series network, obtaining a predicted laser parameter combination corresponding to the target electromagnetic shielding effectiveness data, and realizing laser induced graphene through the predicted laser parameter combination; The series network comprises: An inverse design network is used to determine the predicted laser parameter combination based on the electromagnetic shielding effectiveness data; A forward prediction network, used to determine predicted electromagnetic shielding effectiveness data based on predicted laser parameter combinations; The forward prediction network is a pre-trained network, and the inverse design network is a network obtained by reverse deduction based on the pre-trained forward prediction network; The inverse design network includes a plurality of hidden layers, each of which includes a plurality of neurons; the predicted laser parameter combination is obtained from an interface layer, and the interface layer is located between the output of the hidden layer of the inverse design network and the input of the forward prediction network; The training method of the series network includes: Acquiring training data, wherein the training data includes an actual laser parameter combination and actual electromagnetic shielding effectiveness data corresponding to the actual laser parameter combination; Inputting the actual electromagnetic shielding effectiveness data into the series network to obtain the predicted laser parameter combination and predicted electromagnetic shielding effectiveness data corresponding to the training data; Determining a parameter prediction update value according to a difference between the actual laser parameter combination and the predicted laser parameter combination corresponding to the training data; Determining an effectiveness prediction update value according to a difference between the actual electromagnetic shielding effectiveness data and the predicted electromagnetic shielding effectiveness data corresponding to the training data; When the actual electromagnetic shielding effectiveness data is greater than the predicted electromagnetic shielding effectiveness data corresponding to the training data, the parameter prediction update value is a parameter prediction loss value, and the effectiveness prediction update value is an effectiveness prediction loss value; Obtain the pre-built mapping relationship between parameter prediction loss value and interference value; Determine a target interference value according to the current parameter prediction loss value and the mapping relationship; Subtracting the target interference value from the performance prediction loss value to obtain a modified performance prediction loss value; updating the network parameters of the series network according to the parameter prediction loss value and the modified performance prediction loss value; When the actual electromagnetic shielding effectiveness data is less than or equal to the predicted electromagnetic shielding effectiveness data corresponding to the training data, the effectiveness prediction update value is an effectiveness prediction reward value; The network parameters of the series network are updated according to a weighted sum of the parameter prediction update value and the performance prediction reward value, wherein the weight of the performance prediction reward value is greater than the weight of the parameter prediction update value.

7. A terminal, characterized in that: The terminal includes a memory and one or more processors; the memory stores one or more programs; the program contains instructions for executing the method for automatically predicting laser parameter combinations for laser-induced graphene as described in any one of claims 1-5; and the processor is used to execute the program.

8. A computer-readable storage medium having a plurality of instructions stored thereon, characterized in that: The instructions are suitable for being loaded and executed by a processor to implement the steps of the method for automatically predicting laser parameter combinations for laser induced graphene as described in any one of claims 1 to 5 above.