A construction method and application of a design model for a metamaterial sensing chip based on deep learning

Abstract

The present solution provides a method for constructing a design model of a metamaterial sensing chip based on deep learning and its application. The set of transmission curve acquisition points is input into the input layer of the design model for normalization processing to obtain a one-dimensional sequence. The one-dimensional sequence is respectively input into multiple convolutional branches of a parallel convolutional layer for convolutional processing to obtain convolutional features. The multiple convolutional features are input into a linear layer for splicing to obtain a spliced sequence. The spliced sequence is input into the output layer for processing and outputting a structural parameter group, which can be used for the design of a metamaterial sensing chip.

Description

Technical Field

[0001] The present invention relates to the field of sensing chip design, and particularly to a method for constructing and applying a design model of a metamaterial sensing chip based on deep learning. Background Art

[0002] With the rapid development and breakthrough of deep learning and artificial neural networks, the fields of their application have been continuously expanding. In addition to the traditional field of computer science, they are also applied to physics, chemistry, materials science, etc.

[0003] The design of metamaterial structures often requires a rather complex design process and design cycle. The traditional calculation method starts from certain initial conditions and boundary conditions, uses computational electromagnetic simulation to solve the discrete Maxwell's equations in space and time, and by setting sufficient grids and iterative steps, the optical properties of a given structure can be accurately calculated. Often, it is necessary to fine-tune the geometric shape and iteratively perform simulations to gradually approach the target response. This process largely depends on the past experience of the design template, and due to the limitations of simulation capabilities and time, only a limited number of design parameters are adjusted when searching for the optimal structure.

[0004] Therefore, the implicit relationship between the physical structure parameters of the metamaterial and its electromagnetic response can be discovered through deep learning methods, thereby helping designers to quickly design metamaterial sensors on demand. However, currently, the speed of using deep learning models to design electromagnetic-induced transparency metamaterials is still not fast enough, and the accuracy is still not high enough. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for constructing and applying a design model of a metamaterial sensing chip based on deep learning, and construct a design model of a metamaterial sensing chip that can be reversely designed and can simultaneously generate three transparent window resonance peaks, so as to improve the design efficiency of metamaterial sensors.

[0006] To achieve the purpose, the present technical solution provides a method for constructing a design model of a metamaterial sensing chip based on deep learning, including the following steps:

[0007] Obtain a set of transmission curve acquisition points corresponding to a metamaterial sensing chip with different structural parameter groups as training data, where the upper layer structure of the metamaterial sensing chip includes π-shaped resonant units and rectangular resonant units arranged at intervals, and the π-shaped opening of the π-shaped resonant unit is arranged facing the rectangular resonant unit, and the structural parameter group includes the horizontal side length, vertical bar length, width of the π-shaped opening, length of the long side of the rectangular resonant unit, and length of the short side of the rectangular resonant unit;

[0008] The training data is input into the design framework for training until the training conditions are met. The design framework includes an input layer, a parallel convolutional layer, a linear layer, and an output layer connected in sequence. The parallel convolutional layer includes multiple convolutional branches, each convolutional branch includes convolutional layers with different convolutional dimensions, and each convolutional layer is followed by a ReLU activation function and a Pool pooling operation;

[0009] The set of transmission curve acquisition points is input into the input layer for normalization processing to obtain a one-dimensional sequence. The one-dimensional sequence is respectively input into multiple convolutional branches of the parallel convolutional layer for convolutional processing to obtain convolutional features. The multiple convolutional features are input into the linear layer for splicing to obtain a spliced sequence. The spliced sequence is input into the output layer for processing and the structural parameter group is output.

[0010] In a second aspect, the present solution provides a design method for a metamaterial sensing chip. The acquisition points of the transmission curve containing the resonance peaks of three transparent windows are input into the design model constructed by the method for constructing a design model of a metamaterial sensing chip based on deep learning, and the structural parameter group corresponding to the metamaterial sensing chip is output.

[0011] Compared with the prior art, the present technical solution has the following characteristics and beneficial effects:

[0012] The present technical solution provides a method for constructing and applying a design model of a metamaterial sensing chip based on deep learning. This design model can adapt to the complexity of metamaterial device design and has good generalization ability. The design of metamaterial sensors involves the interactive influence of many physical parameters and electromagnetic characteristics, which is a highly complex task. The parallel convolutional layer with a 1D architecture adopted in this design model can accurately simulate the complex input-output relationship, that is, the mapping from the input spectral sequence to the output structural parameter sequence, through parallel learning of different data features. In addition, the non-linear mapping ability of this design model stems from the activation function after each convolutional layer, and thus can fit complex non-linear function relationships. In the field of metamaterial sensor design, many physical phenomena are non-linear. This characteristic enables the design model to accurately capture hidden physical laws. Thanks to the diverse feature extraction and fusion of parallel convolutional branches, the design model can thus handle various changing situations in the design of sensing chips, such as different materials, structures, etc., making it possible to well use this design model for the design of metamaterial sensing chips. Description of the Drawings

[0013] Figure 1 It is a schematic diagram of a metamaterial sensing chip that can generate resonance peaks of three transparent windows.

[0014] Figure 2 It is a schematic diagram of a sensing unit of a metamaterial sensing chip that can generate resonance peaks of three transparent windows.

[0015] Figure 3 It is a transmission curve diagram of various parts of the metamaterial sensor chip that can produce three transparent window resonance peaks.

[0016] Figure 4 This is the electric field distribution diagram of the metamaterial sensor chip that can generate three transparent window resonance peaks.

[0017] Figure 5 It is a schematic diagram of the network framework of the design model.

[0018] Figure 6 is the loss of the test set and training set during the construction process of the designed model.

[0019] Figure 7 It is the transmission curve of the metamaterial sensor chip designed by the design model and the target transmission curve.

[0020] Figure 8 This is the result graph of the metamaterial sensor chip's response to the test objects with different refractive indices.

[0021] Figure 9 The fitting result graph is obtained by linearly fitting the refractive index change of each transparent window relative to the surrounding test object.

[0022] In the figure: 10-π-type resonant unit, 11-horizontal bar, 12-first vertical bar, 13-second vertical bar, 100-π-type opening, 20-rectangular resonant closed loop. DETAILED DESCRIPTION

[0023] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0024] Those skilled in the art should understand that, in the disclosure of the present invention, the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the above terms should not be understood as limiting the present invention.

[0025] Embodiment 1

[0026] This scheme provides a method for constructing a design model of a metamaterial sensor chip based on deep learning. The design model can be used to design the structural parameters of the metamaterial sensor chip, and the metamaterial sensor chip can generate three transparent window resonance peaks to achieve simultaneous detection of multiple antibiotics.

[0027] Specifically, the method for constructing a design model of a metamaterial sensor chip based on deep learning provided in this solution includes the following steps:

[0028] A set of transmission curve collection points corresponding to the metamaterial sensor chip with different structural parameter groups is obtained as training data, wherein the upper structure of the metamaterial sensor chip includes a π-type resonance unit and a rectangular resonance unit arranged at intervals, and the π-type opening of the π-type resonance unit is arranged opposite to the rectangular resonance unit, wherein the structural parameter group includes the horizontal side length, the vertical bar length, the width of the π-type opening, the long side length and the short side length of the rectangular resonance unit;

[0029] Input the training data into the design framework for training until the training conditions are met, wherein the design framework includes an input layer, a parallel convolution layer, a linear layer, and an output layer connected in sequence, and the parallel convolution layer includes multiple convolution branches, each convolution branch includes convolution layers of different convolution dimensions, and each convolution layer includes a ReLU activation function and a Pool operation;

[0030] The transmission curve collection point set is input into the input layer for normalization to obtain a one-dimensional sequence. The one-dimensional sequence is input into multiple convolution branches of the parallel convolution layer for convolution to obtain convolution features. Multiple convolution features are input into the linear layer for splicing to obtain a spliced ​​sequence. The spliced ​​sequence is input into the output layer for processing and outputs a structural parameter group.

[0031] The design model of the metamaterial sensor chip based on deep learning provided by this solution selects 1D architecture as the basic architecture, which is particularly suitable for processing one-dimensional sequence data, so as to capture local key features through convolution operations and mine nonlinear mapping relationships in one-dimensional sequence data. At the same time, this solution introduces parallel convolution layers based on the 1D architecture so that the design model can focus on multi-size features, thereby fully covering all kinds of key information in the design, making the design model have a more thorough understanding of complex mapping relationships.

[0032] Figure 5 This is the overall architecture diagram of the design framework provided by this solution, such as Figure 5As shown, the acquisition points of the transmission curve are multiple spectral data points collected from the transmission curve. The acquisition points of the transmission curve corresponding to each structural parameter group of the metamaterial sensing chip are input into the input layer for normalization processing to obtain a one-dimensional sequence. In an embodiment of this solution, the metamaterial sensing chip corresponding to each structural parameter group collects 201 acquisition points of the transmission curve, and the input layer normalizes the acquisition points of the transmission curve to obtain a one-dimensional sequence.

[0033] In some embodiments, the parallel convolutional layer includes 3 convolutional branches with a 1D architecture. Each convolutional branch includes 3 convolutional layers with different convolutional dimensions, and each layer is followed by a ReLU activation function and a Pool pooling operation.

[0034] Specifically, the convolutional dimensions of the three convolutional layers in each convolutional branch are 256, 128, and 32 respectively.

[0035] Specifically, the parallel convolutional layer includes a first convolutional branch, a second convolutional branch, and a third convolutional branch in parallel. The convolutional kernels of the three convolutional layers in the first convolutional branch are 7, 5, and 3 respectively. The convolutional kernels of the three convolutional layers in the second convolutional branch are 5, 5, and 5 respectively. The convolutional kernels of the three convolutional layers in the third convolutional branch are 3, 5, and 7 respectively.

[0036] It should be noted that the design model provided by this solution has efficient feature extraction capabilities. For a one-dimensional sequence, the convolutional kernel slides along the sequence dimension, and can quickly focus on key local patterns. In the relevant spectral data of the metamaterial sensing chip design, these local features may correspond to specific combinations of structural parameters, phased changes in electromagnetic responses, etc. The introduction of the parallel convolutional layer in the design model further strengthens this ability. The convolutional branches of different branches are configured with different sizes of convolutional kernels, each capturing local features of different scales, covering fine changes in a small range of input data to trend features with a larger span, making the feature information extracted by the network more comprehensive and rich.

[0037] At the same time, the network architecture of this design model can reduce the number of parameters and computational costs. The convolutional branches under the parallel convolutional layer perform relatively independent operations, and then the convolutional results are integrated through a linear layer, avoiding redundant calculations. Different convolutional branches focus on processing different types of features, and the computational resources are more reasonably allocated, improving the operation efficiency without sacrificing performance. Moreover, the convolutional branches with a 1D architecture greatly reduce the number of parameters to be trained through the receptive field and weight sharing.

[0038] In some embodiments, the linear layer includes a splicing linear layer and a fully connected linear layer. The convolutional features output by the three convolutional branches enter the splicing linear layer together for splicing to obtain a spliced feature, and the spliced feature is then mapped through fully connected linear layers with different dimensions for dimensionality reduction to obtain a dimensionality-reduced spliced sequence.

[0039] In a specific embodiment, the concatenated linear layer concatenates the convolution features of three convolutional branches to obtain a concatenated feature with a length of 96. The concatenated feature is then subjected to dimensionality reduction mapping through a fully connected linear layer with dimensions of 64 and 16 to obtain a concatenated sequence. The concatenated sequence is input into the output layer to obtain a set of structural parameters consisting of the horizontal side length of the π-shaped resonant unit, the vertical bar length, the width of the π-shaped opening, the length of the long side of the rectangular resonant unit, and the length of the short side.

[0040] Regarding the acquisition of training data:

[0041] This solution uses CST Studio Suite 2020 electromagnetic simulation software to model and simulate the metamaterial sensing chip, so as to obtain the data set required for the experiment. In the process of obtaining the data set, the frequency domain solver in the software is used. Based on the proposed metamaterial sensing chip, the five structural parameters of the horizontal side length of the π-shaped resonant unit, the vertical bar length, the width of the π-shaped opening, the length of the long side of the rectangular resonant unit, and the length of the short side are used as variables for simulation to complete the acquisition of the transmission curve acquisition points. In the process of data acquisition, for the transmission curve obtained by device simulation, 201 points are uniformly selected as the transmission curve acquisition points to represent the transmission curve.

[0042] In some specific embodiments, the total amount of collected data is 3125. 80% is used as the training set and 20% is used as the test set. A set of transmission curve acquisition points in the data set includes 201 sampling points of the transmission curve and the five structural parameters of the horizontal side length of the π-shaped resonant unit, the vertical bar length, the width of the π-shaped opening, the length of the long side of the rectangular resonant unit, and the length of the short side.

[0043] Regarding the training process:

[0044] In this solution, when the training set is input into the model, it is trained in batches, with 16 pieces of data in each batch. The training and validation losses of this designed model are as Figure 6 shown, Figure 6 where (a) corresponds to the training set loss. It can be seen that the training set loss converges to 0.018 after 400 epochs. Figure 6 where (b) corresponds to the test set loss. The test set loss converges to 0.025 after 250 epochs. The model reaches a stable and low loss at 200 epochs during training. While having a relatively fast training speed, its MSE of the test set is as low as 0.029, indicating that this designed model has a relatively excellent design ability for the metamaterial sensing chip.

[0045] Regarding the verification of the designed model:

[0046] After the MSE of the training set converges to 0.018, 4 transmission curves are randomly selected from the test set and input into the improved design model, and 4 corresponding sets of structural parameter groups [the horizontal side length of the π-shaped resonant unit, the vertical bar length, the width of the π-shaped opening, the length of the long side of the rectangular resonant unit, and the length of the short side] can be designed immediately. The 4 sets of structural parameters designed are: [279.165, 77.872, 131.098, 147.113, 84.065], [273.851, 86.069, 125.673, 157.687, 92.147], [279.423, 81.895, 120.568, 152.472, 87.451], [269.421, 85.732, 131.734, 146.673, 88.356]. To more intuitively compare their design effects, they are organized into Table 1 as follows:

[0047] Table 1 Designed structural parameter groups

[0048]

[0049] Among them, the target parameter (a) / target parameter (b) / target parameter (c) and target parameter (d) in Table 1 respectively correspond to the structural parameters of the metamaterial sensing chip corresponding to the 4 transmission curves selected from the test set, and the design parameters are the design values output by the design model. In this scheme, the 4 sets of structural parameters designed are simulated, and the obtained transmission curves are plotted together with the target curves input into the design model in Figure 7 It can be seen that the simulation results of the designed structural parameters are in good agreement with the target curves, and the design task of the target curves can be completed well.

[0050] Regarding the specific structure of the metamaterial sensing chip of this scheme, as shown in Figure 1 and Figure 2 This metamaterial sensing chip includes: at least one sensing unit, and each sensing unit includes a substrate layer and a π-shaped resonant unit 10 and a rectangular resonant unit 20 arranged at intervals on the substrate layer. The π-shaped resonant unit 10 includes a horizontally arranged horizontal bar 11 and two vertical bars arranged perpendicular to the horizontal bar 11. The vertical bars include a first vertical bar 12 and a second vertical bar 13, and a π-shaped opening 100 is formed between the first vertical bar 12 and the second vertical bar 13. The rectangular resonant unit 20 is a rectangular closed-loop structure, and the opening direction of the π-shaped opening 100 is set facing the rectangular resonant unit 20; multimode coupling occurs between the π-shaped resonant unit 10 and the rectangular resonant unit 20, and interference cancellation generates three transparent window resonance peaks at three different frequencies. In this scheme, the lengths of the first vertical bar 12 and the second vertical bar 12 are the same, and both are defined as the length of the vertical bar.

[0051] In some embodiments, the materials of the π-shaped resonant unit 10 and the rectangular resonant unit 20 are copper, and the substrate layer is quartz. Specifically, the conductivity of copper is 5.96×10 7 S / m, and the relative permittivity of quartz is 1.9. The reason for choosing a material with a low dielectric constant as the substrate layer material in this solution is to enhance the sensing performance of the triple-band electromagnetic induced transparency metamaterial sensing chip.

[0052] In some embodiments, the rectangular resonant unit 20 includes two long sides arranged in parallel and two short sides arranged in parallel, and the long sides and the short sides are perpendicular to each other, and the length of the long side is greater than the length of the short side. Specifically, the long side of the rectangular resonant unit 20 is parallel to the horizontal bar 11 of the π-shaped resonant unit 10, and the opening of the π-shaped opening 100 is facing the long side of the rectangular resonant unit 20.

[0053] As Figure 2 shown, define the length of the horizontal side of the π-shaped resonant unit of this metamaterial sensing chip as c, the length of the vertical bar as d, the width of the π-shaped opening as g, the length of the long side of the rectangular resonant unit as l, and the length of the short side as w. It is these five structural parameters that form a structural parameter group that affects the position of the transmission window resonance peak generated by the metamaterial sensing chip.

[0054] This solution uses CST Studio Suite electromagnetic simulation software to complete the simulation of this triple-band electromagnetic induced transparency metamaterial sensing chip. During the simulation, the boundary conditions of the triple-band electromagnetic induced transparency metamaterial sensing chip unit in the x and y directions are set as unit cell boundaries, and the z direction is an open (add space) boundary to simulate an infinite metamaterial periodic array. Select a THz plane wave with an incident direction of z as the excitation source, and the simulation frequency band is set to 0.3~1 THz.

[0055] This solution simulates the π-shaped resonant unit and the rectangular resonant unit separately. The simulation transmission curve diagram is as Figure 3 shown to study their contribution degrees to the three transparent windows. In Figure 3 , the transmission curve of the π-shaped resonant unit when simulated alone is represented by curve R1, the transmission curve of the rectangular resonant unit when simulated alone is represented by curve R2, and the transmission curve of the overall sensing chip simulation is represented by R3. It can be seen that the three transparent windows are formed by the coupling of multiple modes.

[0056] Next, analyze its formation mode in combination with the electric field distribution diagram. The electric field distribution diagrams at seven key points with increasing frequency are as Figure 4 shown. Figure 4 In (a) of Figure 4 represents the first wave valley, and in (b) ofFigure 4 In (c) represents the second trough, Figure 4 In (d) represents the second transparent window, Figure 4 In (e) represents the third trough, Figure 4 In (f) represents the third transparent window, Figure 4 In (g) represents the fourth trough. The electric field distribution is divided into three parts for analysis, namely the horizontal bar P1 of the π-shaped resonant unit, the first vertical bar and the second vertical bar P2 of the π-shaped resonant unit, and the overall rectangular resonant unit P3. It can be seen that P2 and P3 are excited at the first trough, only P1 is excited at the second trough, and all three are in the bright state before and after the first transparent window. Then the first transparent window is a bright-bright-bright coupling mode, and the electric field energy accumulates between the horizontal bar, the first vertical bar and the second vertical bar of the π-shaped resonant unit and the rectangular resonant unit. Only P2 is strongly excited at the third trough, and P3 is not excited before and after the second transparent window. Then the second transmission window is a bright-dark-bright coupling mode, and the electric field energy only accumulates between the horizontal bar and the two vertical bars of the π-shaped resonant unit. P1 is strongly excited at the fourth trough, and P3 is not excited before and after the third transparent window. Then the third transmission window is a bright-dark-bright coupling mode, and the electric field energy only accumulates at the upper side length and the two arms of the π-shaped resonator. In summary, the three transparent windows are bright-bright-bright coupling, bright-dark-bright coupling and bright-dark-bright coupling modes respectively.

[0057] Embodiment 2

[0058] This solution provides an application method of a design model of a metamaterial sensing chip based on deep learning constructed according to Embodiment 1, that is, Embodiment 2 provides a design method of a metamaterial sensing chip:

[0059] Input the acquisition points of the transmission curve containing the resonant peaks of the three transparent windows into the design model, and output the structural parameter group of the corresponding metamaterial sensing chip.

[0060] The content same as that in Embodiment 1 will not be elaborated here.

[0061] It should be noted that in a specific embodiment, this solution designs a metamaterial sensing chip that can be used for antibiotic detection by using the design method of this metamaterial sensing chip.

[0062] Specifically, input the acquisition points of the transmission curve containing the resonant peaks of the three transmission windows of 0.76 THz, 0.79 THz and 0.85 THz into the design model, and output the structural parameter group of the corresponding metamaterial sensing chip. This metamaterial sensing chip can be used to detect chlortetracycline, tetracycline and sodium penicillin, where 0.76 THz, 0.79 THz and 0.85 THz respectively correspond to the fingerprint spectra of chlortetracycline, tetracycline and sodium penicillin.

[0063] Specifically, the structural parameter set of the metamaterial sensing chip obtained by this solution is as follows: the length of the crossbar 11 of the π-shaped resonant unit 10 is 272 µm to 276 µm, the lengths of the first vertical bar 12 and the second vertical bar 13 are 80 µm to 84 µm, the width of the π-shaped opening 100 is 124 µm to 128 µm, the length of the long side of the rectangular resonant unit 20 is 150 µm to 154 µm, and the length of the short side is 86 µm to 90 µm.

[0064] Preferably, the length of the crossbar 11 of the π-shaped resonant unit 10 c is 274 µm, the lengths of the first vertical bar 12 and the second vertical bar 13 d are 82 µm, and the width of the π-shaped opening 100 g is 126 µm. The length of the long side of the rectangular resonant unit 20 l is 152 µm, and the length of the short side w is 88 µm. The side length of the substrate layer Px and Py is 180 µm.

[0065] In addition, the application team conducted performance tests on the preferred metamaterial sensing chip to study the response of the metamaterial sensing chip to test objects with different refractive indices, and the results are as Figure 8 shown. It can be seen from Figure 8 that when the refractive index of the test object increases from 1.0 to 1.3 in steps of 0.1, the resonant peaks of the three transmission windows all undergo frequency shifts to varying degrees. The resonant frequency of the first transmission window resonant peak moves from 0.755 THz to 0.706 THz, and for each test object with a corresponding refractive index, the frequency shift of the first transparent window resonant peak is 164, 163, 163 THz, resulting in a total redshift of 490 GHz; the resonant frequency of the second transmission window resonant peak moves from 0.798 THz to 0.770 THz, and for each test object with a corresponding refractive index, the frequency shift of the second transparent window resonant peak is 95, 93, 92 THz, resulting in a total redshift of 280 GHz; the resonant frequency of the third transmission window resonant peak moves from 0.848 THz to 0.802 THz, and the frequency shift of the third transparent window resonant peak is 154, 154, 152 THz, resulting in a total redshift of 460 GHz. By comparing with each other, it can be seen that the sensitivity of the resonant peaks of the three transmission windows to the change in the refractive index of the test object from low to high is the first transmission window resonant peak, the third transmission window resonant peak, and the second transmission window resonant peak in turn. The difference in sensitivity is caused by the different electromagnetic responses of the coupling fields of each transparent window to the test object.

[0066] Performing linear fitting on the refractive index changes of each transparent window relative to the surrounding test objects, the fitting results are as Figure 9As shown, S1, S2, and S3 respectively represent the first, second, and third transparent windows. The slope of the fitted line corresponds to the sensitivity of the transparent window. The slopes of the three fitted lines are represented by k1, k2, and k3 respectively, where k1 = -0.163, k2 = -0.093, k3 = -0.155, and their R 2 are 0.99977, 0.99931, and 0.9938 respectively. Then the sensitivities of the three transparent windows are S1 = 163 GHz / RIU, S2 = 93 GHz / RIU, and S3 = 155 GHz / RIU respectively. The central frequencies of the three transparent windows are 0.755 THz, 0.795 THz, and 0.848 THz respectively, and the full-width at half-maximum are 0.051 THz, 0.043 THz, and 0.032 THz respectively. Then the Q values are 14.8, 18.5, and 26.5 respectively, and the FOMs are 3.2, 2.2, and 4.8 respectively. The sensing performances of the three transparent windows are somewhat different, and specific sensing detections of chlortetracycline, tetracycline, and sodium penicillin can be realized at the corresponding frequencies respectively.

[0067] Those skilled in the art should understand that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0068] The above embodiments only represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for constructing a design model of a metamaterial sensing chip based on deep learning, characterized in that The method includes the following steps: Obtain a set of transmission curve acquisition points corresponding to metamaterial sensing chips with different structural parameter groups as training data. The upper layer structure of the metamaterial sensing chip includes π-shaped resonant units and rectangular resonant units arranged at intervals. The π-shaped opening of the π-shaped resonant unit is arranged facing the rectangular resonant unit. The π-shaped resonant unit includes a horizontally arranged horizontal bar and two vertically arranged vertical bars relative to the horizontal bar. The vertical bars include a first vertical bar and a second vertical bar arranged at intervals, and a π-shaped opening is formed between the first vertical bar and the second vertical bar. The rectangular resonant unit is a rectangular closed-loop structure. The structural parameter group includes the horizontal side length of the π-shaped resonant unit, the vertical bar length, the width of the π-shaped opening, the length of the long side of the rectangular resonant unit, and the length of the short side. Multimode coupling occurs between the π-shaped resonant unit and the rectangular resonant unit, and interference cancellation generates three transparent window resonance peaks at three different frequencies; Input the training data into the design framework for training until the training conditions are met. The design framework includes an input layer, a parallel convolutional layer, a linear layer, and an output layer connected in sequence. The parallel convolutional layer includes multiple convolutional branches. Each convolutional branch includes convolutional layers with different convolutional dimensions. Each convolutional layer is followed by a ReLU activation function and a Pool pooling operation. The parallel convolutional layer includes 3 convolutional branches with a 1D architecture. Each convolutional branch includes 3 convolutional layers with different convolutional dimensions, and each layer is followed by a ReLU activation function and a Pool pooling operation; The set of transmission curve acquisition points is input into the input layer for normalization processing to obtain a one-dimensional sequence. The one-dimensional sequence is respectively input into multiple convolutional branches of the parallel convolutional layer for convolutional processing to obtain convolutional features. The multiple convolutional features are input into the linear layer for splicing to obtain a spliced sequence. The spliced sequence is input into the output layer for processing and the structural parameter group is output.

2. The method for constructing a design model of a metamaterial sensing chip based on deep learning according to claim 1, wherein The convolutional dimensions of the three convolutional layers in each convolutional branch are 256, 128, and 32 respectively.

3. The method for constructing a design model of a metamaterial sensing chip based on deep learning according to claim 2, characterized in that The parallel convolutional layer includes a first convolutional branch, a second convolutional branch, and a third convolutional branch connected in parallel. The convolutional kernels of the three convolutional layers in the first convolutional branch are 7, 5, and 3 respectively. The convolutional kernels of the three convolutional layers in the second convolutional branch are 5, 5, and 5 respectively. The convolutional kernels of the three convolutional layers in the third convolutional branch are 3, 5, and 7 respectively.

4. The method for constructing a design model of a metamaterial sensing chip based on deep learning according to claim 1, characterized in that, The linear layer includes a splicing linear layer and a fully connected linear layer. The convolutional features output by the three convolutional branches enter the splicing linear layer together for splicing to obtain a spliced feature. The spliced feature is then subjected to dimensionality reduction mapping through fully connected linear layers with different dimensions to obtain a dimensionality-reduced spliced sequence.

5. The method for constructing a design model of a metamaterial sensing chip based on deep learning according to claim 4, wherein The splicing linear layer splices the convolutional features of the three convolutional branches to obtain a spliced feature with a length of 96. The spliced feature is then subjected to dimensionality reduction mapping through fully connected linear layers with dimensions of 64 and 16 to obtain a spliced sequence.

6. The construction method of the design model of the metamaterial sensing chip based on deep learning according to claim 1, characterized in that, The rectangular resonant unit includes two parallel long sides and two parallel short sides, and the long side and the short side are vertically arranged, and the length of the long side is greater than the length of the short side.

7. The method for constructing a design model of a metamaterial sensing chip based on deep learning according to claim 6, characterized in that, The long side of the rectangular resonant unit is arranged parallel to the horizontal bar of the π-shaped resonant unit, and the opening of the π-shaped opening faces the long side of the rectangular resonant unit.

8. The method for constructing a design model of a metamaterial sensing chip based on deep learning according to claim 1, wherein, The materials of the π-shaped resonant unit and the rectangular resonant unit are copper, and the substrate layer is quartz.

9. A design method of a metamaterial sensing chip, characterized in that, Input the acquisition points of the transmission curve with three transparent window resonance peaks into the design model constructed by the construction method of the design model of the metamaterial sensing chip based on deep learning according to any one of claims 1 to 8, and output the structural parameter group of the corresponding metamaterial sensing chip.

Images