Open resonator based on a hollow-core anti-resonant optical fiber and method for designing the same

Abstract

The application provides a kind of hollow-core anti-resonant optical fiber reverse design method and system based on series neural network, belongs to hollow-core anti-resonant optical fiber design technical field, obtains the expected limit loss and known structure parameter of the hollow-core anti-resonant optical fiber to be designed;Using the prediction model trained in advance, the expected limit loss and known structure parameter obtained are processed to obtain the reverse design structure parameter of the hollow-core anti-resonant optical fiber to be designed;Wherein, the prediction model includes reverse structure design network and forward performance prediction network;Reverse structure design network is used to process based on expected limit loss and known structure parameter to obtain reverse design structure parameter.The application shortens the optimization design time of hollow-core anti-resonant optical fiber, and can quickly realize the rapid design of the specified limit loss of hollow-core anti-resonant optical fiber structure.

Description

Technical Field

[0001] This invention relates to the field of hollow-core anti-resonant fiber design technology, specifically to a reverse design method and system for hollow-core anti-resonant fibers based on a series neural network. Background Technology

[0002] Hollow-core antiresonant fiber, due to its unique antiresonant light guiding mechanism that confines light within its hollow core, offers a wider transmission band, greater design freedom, and a relatively simpler cladding structure compared to traditional hollow-core photonic bandgap fiber. It is currently used in fiber optic communication, fiber optic sensors, and fiber lasers. With further development in these fields, the demand and requirements for hollow-core antiresonant fiber are increasing. Currently, the main numerical simulation methods used for performance prediction and structural design of specific hollow-core antiresonant fibers are still the finite element method (FEM) and multi-parameter multiple-scan optimization. Calculating the optical properties of hollow-core antiresonant fiber using the FEM and optimizing the design of hollow-core antiresonant fiber with specified optical properties using multi-parameter scanning methods are time-consuming and computationally resource-intensive, and the calculations of multiple parameter scans are not fully utilized. Summary of the Invention

[0003] The purpose of this invention is to provide a reverse design method and system for hollow anti-resonant optical fibers based on a series neural network, so as to solve at least one of the technical problems existing in the background art.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] On one hand, the present invention provides a reverse design method for hollow-core anti-resonant optical fiber, comprising:

[0006] Obtain the expected confinement loss and known structural parameters of the hollow antiresonant fiber to be designed;

[0007] Using a pre-trained prediction model, the obtained expected constraint loss and known structural parameters are processed to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a complete structural parameter vector for processing to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters.

[0008] Preferably, the known structural parameters include: the style of the first cladding tube, the style of the second cladding tube, the number of cladding tubes in each layer, the core diameter, the wall thickness of the first cladding tube, the wall thickness of the nested tubes in the first cladding tube, the wall thickness of the second cladding tube, and the wall thickness of the nested tubes in the second cladding tube.

[0009] Preferably, the reverse design structural parameters include: the radial axis length of the first cladding tube, the tangential axis length of the first cladding tube, the radial axis length of the second cladding tube, the tangential axis length of the second cladding tube, the radial axis length of the nested tube in the first cladding tube, the tangential axis length of the nested tube in the first cladding tube, the radial axis length of the nested tube in the second cladding tube, and the tangential axis length of the nested tube in the second cladding tube.

[0010] Preferably, the number of hidden layers and the number of nodes in each hidden layer of the positive performance prediction network are optimized at a learning rate of 0.0002. The number of hidden layers increases from 1 to 10 layers in increments of 1, and the number of nodes increases from 50 to 800 in increments of 50. The ReLU activation function and Adam optimizer are selected during training, and the number of epochs is 5000.

[0011] Preferably, an optimization loss function Loss = α × loss is constructed for optimizing the reverse design structure network. str +β×loss cl , where loss str and loss cl Let α and β represent the mean square error loss (MSE) of the reverse design structure and the corresponding constraint loss, respectively, with α and β being the corresponding coefficients, each with a value of 1.

[0012] Preferably, for the optimization of the reverse structure design network, the learning rate is increased from 0.0001 to 0.0020 in intervals of 0.0001, the number of hidden layers is increased from 1 to 10 in intervals of 1, the number of nodes is increased from 50 to 800 in intervals of 50, the activation function used during training is ReLU, the optimizer is Adam, and the training epochs are 5000.

[0013] Secondly, the present invention provides a reverse design system for hollow-core anti-resonant optical fiber, comprising:

[0014] The acquisition module is used to acquire the expected confinement loss and known structural parameters of the hollow anti-resonant fiber to be designed.

[0015] The prediction module is used to process the acquired expected constraint loss and known structural parameters using a pre-trained prediction model to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a complete structural parameter vector and process it to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters are.

[0016] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the reverse design method for hollow anti-resonant optical fiber based on a serial neural network as described above.

[0017] Fourthly, the present invention provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the reverse design method for hollow anti-resonant optical fiber based on a serial neural network as described above.

[0018] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the reverse design method of hollow-core anti-resonant optical fiber based on a serial neural network as described above.

[0019] The beneficial effects of this invention are: it shortens the optimization design time of hollow anti-resonant optical fibers and enables rapid reverse design of hollow anti-resonant optical fibers with specified characteristics.

[0020] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the hollow anti-resonant optical fiber structure when the number of cladding tubes in each cladding tube is 6, as described in an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the network structure of the positive performance prediction model described in an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of the training results of the positive performance prediction model described in an embodiment of the present invention.

[0025] Figure 4 This is a schematic diagram of the cascaded neural network structure described in an embodiment of the present invention.

[0026] Figure 5 This is a schematic diagram comparing the design results of the cascaded neural network described in an embodiment of the present invention. Detailed Implementation

[0027] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0028] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0030] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated 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, and / or groups thereof.

[0031] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0032] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0033] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0034] Example 1

[0035] In this embodiment 1, a reverse design system for hollow-core anti-resonant optical fiber is first provided, including:

[0036] The acquisition module is used to acquire the expected confinement loss and known structural parameters of the hollow anti-resonant fiber to be designed.

[0037] The prediction module is used to process the acquired expected constraint loss and known structural parameters using a pre-trained prediction model to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a structural parameter vector and process it to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters are.

[0038] In this embodiment 1, the above-described system is used to implement a reverse design method for hollow-core anti-resonant optical fibers, including:

[0039] The expected confinement loss and known structural parameters of the hollow anti-resonant fiber to be designed are obtained using the acquisition module.

[0040] Using a prediction module, based on a pre-trained prediction model, the obtained expected constraint loss and known structural parameters are processed to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a structural parameter vector and process it to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters.

[0041] The known structural parameters include: the style of the first cladding tube, the style of the second cladding tube, the number of cladding tubes in each layer, the core diameter, the wall thickness of the first cladding tube, the wall thickness of the nested tubes in the first cladding tube, the wall thickness of the second cladding tube, and the wall thickness of the nested tubes in the second cladding tube. The reverse-engineered structural parameters include: the radial axis length of the first cladding tube, the tangential axis length of the first cladding tube, the radial axis length of the second cladding tube, the tangential axis length of the nested tubes in the first cladding tube, the radial axis length of the nested tubes in the second cladding tube, and the tangential axis length of the nested tubes in the second cladding tube. The positive performance prediction network was trained using dataset 1. The number of hidden layers and the number of nodes per hidden layer were optimized with a learning rate of 0.0002. The number of hidden layers increased from 1 to 10 in increments of 1, and the number of nodes increased from 50 to 800 in increments of 50. The ReLU activation function and Adam optimizer were selected during training, with 5000 epochs. An optimization loss function, Loss = α × loss, was constructed to optimize the reverse design structure network. str +β×loss cl , where loss str and loss cl Let represent the mean squared error (MSE) of the reverse design structure and the corresponding constraint loss, respectively. α and β are the corresponding coefficients, with a value of 1. Based on datasets 1-1 to 3-3, the reverse structure design network was optimized. The learning rate increased from 0.0001 to 0.0020 in intervals of 0.0001, the number of hidden layers increased from 1 to 10 in intervals of 1, and the number of nodes increased from 50 to 800 in intervals of 50. The activation function used during training was ReLU, the optimizer was Adam, and the training epochs were 5000.

[0042] Example 2

[0043] In this embodiment 2, a method for designing a reverse structure of a hollow-core anti-resonant optical fiber is provided, targeting a hollow-core anti-resonant optical fiber structure such as... Figure 1As shown, the black area represents quartz, whose refractive index is determined by the Sellmeier equation, and the white area represents air, with a refractive index of 1. The structure of the hollow-core antiresonant fiber can be determined by the parameter vector [s1,s2,N,D]. core ,Ra 1c ,Ta 1c ,Ra 2c ,Ta 2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1n ,t 1c ,t 1n ,t 2c ,t 2n [Identify] where s1 and s2 represent the patterns of the first and second cladding tubes, N is the number of individual cladding tubes, and Ra 1c Ta 1c Ra represents the radial and tangential lengths of the first elliptical cladding tube, respectively. 2c and Ta 2c Ra represents the radial and tangential lengths of the second elliptical cladding tube, respectively. 1n Ta 1n Ra represents the lengths of the radial and tangential axes of the elliptical nested tubes within the first elliptical cladding tube, respectively. 2n and Ta 2n t1c and t1n represent the lengths of the radial and tangential axes of the elliptical nested tubes in the second elliptical cladding tube, respectively; t1c and t1n represent the wall thicknesses of the first cladding tube and the nested tubes therein, respectively; and t2c and t2n represent the wall thicknesses of the first cladding tube and the nested tubes therein, respectively.

[0044] Figure 1In the diagram, (a) to (l) are schematic diagrams of the hollow-core anti-resonant fiber structures corresponding to datasets 1-0, 1-1, 1-2, 1-3, 2-0, 2-1, 2-2, 2-3, 3-0, 3-1, 3-2, and 3-3, respectively. In this embodiment, the particle swarm optimization algorithm combined with the finite element method is used to calculate the confinement loss of the hollow-core anti-resonant fiber (Fanchao Meng, Xiaoting Zhao, Jinmin Ding, Yingli Niu, Xinghua Zhang, Lvyun Yang, Xin Wang, Shuqin Lou, Xinzhi Sheng, Guangming Tao, and Sheng Liang, "Discovering extremely low confinement-loss anti-resonant fibers via swarm."). intelligence," Opt. Express 29, 35544-35555 (2021)), where the inertia weight of the particle swarm optimization algorithm is 0.8, and the learning factor for both individuals and the swarm is 1.0. Considering the rationality of fiber fabrication, the parameter ranges for each structure are determined as follows: 5≤N≤10, D core =30μm, 20μm≤Ra 1c Ta 1c ≤40μm, 10μm≤Ra 2c Ta 2c ≤20μm, 0.3μm≤t 1c t 1n t 2c t 2n ≤0.7μm, define parameter η1=Ra 2n / Ra 2c ,η2=Ta 2n / Ra 2n η3=Ra 1n / Ra 1c ,η4=Ta 1n / Ra 1n The ratios between the radial and tangential axes of the nested tube and the cladding tube are described, with values ​​ranging from 0.3 to 0.8. Data with a limiting loss of less than 0.01 dB / m are selected to construct dataset 1. Based on dataset 1, datasets 1-0 to 3-3 are determined according to different hollow-core anti-resonant fiber cladding structure patterns.

[0045] like Figure 2 As shown in Example 2, the training of the positive performance prediction model is as follows:

[0046] Figure 2A schematic diagram of the network structure of the forward performance prediction model is given. The overall network structure consists of an input layer, a hidden layer, and an output layer. The input is a normalized parameter vector [s1,s2,N,D] describing the hollow anti-resonant fiber structure. core ,Ra 1c ,Ta 1c ,Ra 2c ,Ta 2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1n ,t 1c ,t 1n ,t 2c ,t 2n The output is the limited loss of a hollow antiresonant fiber.

[0047] The number of hidden layers and the number of nodes per hidden layer in the positive performance prediction model were optimized with a learning rate of 0.0002. The number of hidden layers increased from 1 to 10 layers in increments of 1, and the number of nodes increased from 50 to 800 in increments of 50. The ReLU activation function and Adam optimizer were selected during training, with 5000 epochs. The coefficient of determination R0 was chosen. 2 The mean squared error (MSE) is used to evaluate model performance. Figure 3 (a) and Figure 3 (b) MSE and R are given for different numbers of hidden layers and nodes. 2 The position indicated by the arrow represents the optimal result, where there are 6 hidden layers, 800 nodes, and a corresponding MSE of 0.0007 and R0.0007. 2 It is 0.9154.

[0048] In this second embodiment, the training of the cascaded neural network is as follows: Figure 4 A schematic diagram of the cascaded neural network is given. The cascaded neural network consists of an inverse structural design model and a pre-trained forward performance prediction model. The input to the inverse structural design model is the expected confinement loss and the known structural parameters [s1, s2, N, D] of the hollow-core antiresonant fiber. core ,t 1c ,t 1n ,t 2c ,t 2n The output is the reverse engineering structural parameters [Ra]. 1c ,Ta 1c ,Ra 2c ,Ta 2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1nThe known structural parameters of the hollow anti-resonant fiber and the structural parameters output by the inverse structural design model form a complete structural parameter vector of the hollow anti-resonant fiber. Then, the confined loss of the hollow anti-resonant fiber is obtained through a pre-trained forward performance prediction model.

[0049] Training a cascaded neural network is essentially an optimization of the inverse structure design model. The optimization parameters are the learning rate (increasing from 0.0001 to 0.0020 in 0.0001 increments), the number of hidden layers (increasing from 1 to 10 in 1-in increments), and the number of nodes (increasing from 50 to 800 in 50-in increments). The activation function and optimizer used during training are ReLU and Adam, respectively. The training epochs are 5000, and the expression for the loss function is as follows:

[0050] Loss = α × loss str +β×loss cl

[0051] The loss function is used to optimize the weights and biases of the inverse structure design model. str and loss cl Let MSE represent the limiting losses for the reverse structural design model and the forward performance prediction model, respectively, with α and β being the corresponding coefficients, each with a value of 1. Table 1 presents the optimal parameters and optimization results for the reverse design model under each dataset.

[0052] Table 1

[0053] Dataset 1-0 1-1 1-2 1-3 2-0 2-1 2-2 2-3 3-0 3-1 3-2 3-3 Hidden layer 6 10 10 10 8 10 10 10 10 10 7 10 Number of nodes 400 700 650 500 750 800 700 700 700 750 450 700 Learning rate 0.0010 0.0003 0.0009 0.0004 0.0009 0.0007 0.0006 0.0005 0.0005 0.0003 0.0010 0.0004 R2CL 0.9881 0.7112 0.9500 0.8346 0.8216 0.7621 0.9078 0.7301 0.9706 0.9426 0.9069 0.8343 <![CDATA[R 2 (Sun) 1c )]]> 0.8234 0.0406 0.7858 0.4277 0.5510 0.2555 0.9661 0.4761 0.6722 0.6817 0.8223 0.2492 <![CDATA[R 2 (Facing 1c )]]> 0.8316 0.2991 0.9391 0.4269 0.9445 0.5392 0.9722 0.6543 0.8922 0.9025 0.9937 0.6885 <![CDATA[R 2 (Sun) 2c )]]> - 0.2653 0.8394 0.3413 - 0.6321 0.9665 0.4051 - 0.7362 0.8953 0.4623 <![CDATA[R 2 (Facing 2c )]]> - 0.4203 0.7848 0.5629 - 0.4355 0.9680 0.6378 - 0.8133 0.9530 0.6361 <![CDATA[R 2 (Day 2n )]]> - - 0.8479 0.4052 - - 0.9685 0.3769 - - 0.7810 0.4473 <![CDATA[R 2 (Facing 2n )]]> - - 0.8527 0.5887 - - 0.9760 0.5101 - - 0.7886 0.5813 <![CDATA[R 2 (Sun) 1n )]]> - - - - 0.6528 0.4846 0.9959 0.3264 0.7507 0.7143 0.7812 0.3743 <![CDATA[R 2 (Facing 1n )]]> - - - - 0.6065 0.6204 0.9972 0.3354 0.8254 0.8178 0.8025 0.2561 <![CDATA[R 2 ]]> 0.8275 0.2563 0.8416 0.4587 0.6887 0.4945 0.9763 0.4652 0.7851 0.7776 0.8522 0.4618

[0054] Result verification:

[0055] like Figure 5 As shown, in order to further verify the effectiveness of the cascaded neural network in the reverse structure design of hollow anti-resonant optical fiber, the output characteristics of the cascaded neural network and the finite element calculation results of the predicted structure output by the reverse structure design network were compared under various datasets after a given design target. It was found that the three were basically consistent and the error was within an acceptable range, which proved the effectiveness of the cascaded neural network.

[0056] In summary, in this embodiment 2, a structure vector is used to describe the hollow-core anti-resonant fiber structure, specifically described as: [s1,s2,N,D core ,Ra 1c ,Ta 1c ,Ra 2c ,Ta 2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1n ,t 1c ,t1n ,t 2c ,t 2n ], where s1 and s2 represent the patterns of the first and second cladding tubes, N is the number of individual cladding tubes, and Ra 1c Ta 1c Ra 2c and Ta 2c Ra represents the radial and tangential lengths of the first and second elliptical cladding tubes, respectively. 1n Ta 1n Ra 2n and Ta 2n t represents the lengths of the radial and tangential axes of the elliptical nested tubes within the corresponding cladding tube, respectively. 1c t 1n t 2c and t 2n These represent the wall thicknesses of the corresponding cladding tube and nested tube, respectively.

[0057] In this embodiment, the model uses a loss limit of 0.01dB / m as the standard for dataset construction. Only data with a loss limit of less than 0.01dB / m are selected to form the dataset. At the same time, datasets 1-0, 1-1, 1-2, 1-3, 2-0, 2-1, 2-2, 2-3, 3-0, 3-1, 3-2 and 3-3 are constructed according to the structural patterns of the first cladding tube and the second cladding tube.

[0058] A fully connected neural network is constructed to predict the confinement loss of hollow antiresonant optical fiber, using the coefficient of determination R. 2 The model was evaluated using mean squared error loss (MSE). With a learning rate of 0.0002, ReLU activation function, Adam optimizer, and 5000 training epochs, the number of hidden layers and nodes per layer were optimized to improve model performance. The optimal result was 6 hidden layers and 800 nodes. A tandem neural network was constructed for the reverse design of hollow-core antiresonant fiber. The tandem neural network includes a reverse structure design model and a forward performance prediction model. The inputs are the expected limit loss of the hollow-core antiresonant fiber and the structural parameters [s1, s2, N, D]. core ,t 1c ,t 1n ,t 2c ,t 2n [Ra], which refers to the style of the first cladding tube, the style of the second cladding tube, the number of cladding tubes, the core diameter, the wall thickness of the first cladding tube and its nested tubes, and the wall thickness of the second cladding tube and its nested tubes. The reverse structural design model output is [Ra]. 1c ,Ta 1c ,Ra 2c ,Ta2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1n [s1, s2, N, D] refers to the radial and tangential axis lengths of the first cladding tube, the nested tubes within the first cladding tube, the second cladding tube, and the nested tubes within the second cladding tube. The input structural parameters [s1, s2, N, D] of the reverse engineering design model are... core ,t 1c ,t 1n ,t 2c ,t 2n ] and predicted output structure parameters [Ra 1c ,Ta 1c ,Ra 2c ,Ta 2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1n The complete parameter vector is input into the forward performance prediction model, and the output is the predicted constraint loss. For cascaded neural network optimization, it is actually optimizing the inverse structure design model. The optimization loss for constructing the inverse design structure model is the function Loss = α × loss. str +β×loss cl , where loss str and loss clLet MSE represent the reverse design structure and the corresponding constraint loss, respectively. α and β are the corresponding coefficients, each with a value of 1. Optimization was performed on different datasets in terms of learning rate, number of nodes, and number of hidden layers. The final network structure parameters of the reverse design model are as follows: Dataset 1-0, 6 hidden layers, 400 nodes, learning rate 0.001; Dataset 1-1, 10 hidden layers, 700 nodes, learning rate 0.0003; Dataset 1-2, 10 hidden layers, 650 nodes, learning rate 0.0009; Dataset 1-3, 10 hidden layers, 500 nodes, learning rate 0.0009; Dataset 2-0, 8 hidden layers, 750 nodes, learning rate 0.0009; Dataset 2- Dataset 1: 10 hidden layers, 800 nodes, learning rate 0.0007; Dataset 2-2: 10 hidden layers, 700 nodes, learning rate 0.0006; Dataset 2-3: 10 hidden layers, 700 nodes, learning rate 0.0005; Dataset 3-0: 10 hidden layers, 700 nodes, learning rate 0.0005; Dataset 3-1: 10 hidden layers, 750 nodes, learning rate 0.0003; Dataset 3-2: 7 hidden layers, 450 nodes, learning rate 0.001; Dataset 3-3: 10 hidden layers, 700 nodes, learning rate 0.0004. The activation function and optimizer are ReLU and Adam, respectively, and the training epochs are 5000.

[0059] Example 3

[0060] This embodiment 3 provides a non-transitory computer-readable storage medium for storing computer instructions. When these computer instructions are executed by a processor, they implement the reverse design method for hollow-core anti-resonant optical fiber based on a serial neural network as described above. The method includes:

[0061] Obtain the expected confinement loss and known structural parameters of the hollow antiresonant fiber to be designed;

[0062] Using a pre-trained prediction model, the obtained expected constraint loss and known structural parameters are processed to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a complete structural parameter vector for processing to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters.

[0063] Example 4

[0064] This embodiment 4 provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the reverse design method for hollow anti-resonant optical fibers based on a cascaded neural network as described above. The method includes:

[0065] Obtain the expected confinement loss and known structural parameters of the hollow antiresonant fiber to be designed;

[0066] Using a pre-trained prediction model, the obtained expected constraint loss and known structural parameters are processed to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a complete structural parameter vector for processing to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters.

[0067] Example 5

[0068] This embodiment 5 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the reverse design method of hollow-core anti-resonant optical fiber based on a series neural network as described above. The method includes:

[0069] Obtain the expected confinement loss and known structural parameters of the hollow antiresonant fiber to be designed;

[0070] Using a pre-trained prediction model, the obtained expected constraint loss and known structural parameters are processed to obtain the reverse design structural parameters of the hollow-core anti-resonant fiber to be designed. The prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected constraint loss and known structural parameters to obtain the reverse design structural parameters. The forward performance prediction network is used to combine the known structural parameters and the reverse design structural parameters into a complete structural parameter vector for processing to obtain the actual predicted constraint loss. The closer the actual predicted constraint loss is to the expected constraint loss, the better the obtained reverse design structural parameters.

[0071] In summary, the reverse design method and system for hollow-core antiresonant optical fibers based on a cascaded neural network described in this invention can significantly shorten the optimization design time for hollow-core antiresonant optical fibers. Furthermore, the forward performance prediction model in the cascaded neural network can quickly predict the performance of hollow-core antiresonant optical fibers compared to traditional numerical simulation methods such as the finite element method. Reverse design may yield optical fiber structures that differ significantly from manually designed structures, realizing artificial intelligence-assisted invention of optical fibers. Based on the analysis of the assisted invention structure, new optical fiber knowledge can be summarized. Existing reverse design methods for optical fibers mainly focus on photonic crystal fibers and few-mode fibers, with no known reverse design method for hollow-core antiresonant optical fibers. Therefore, this invention focuses on the implementation of a reverse design method for hollow-core antiresonant optical fibers. Moreover, compared to the previous method of simply cascading the reverse design structure model and the forward performance prediction model, modifications have been made to the cascaded neural network used for hollow-core antiresonant optical fibers, introducing constraints on some structural parameters.

[0072] In this embodiment of the invention, a parameter vector is used to describe the structural parameters of the hollow-core anti-resonant optical fiber. The structural parameter vector is [s1, s2, N, D]. core ,Ra 1c ,Ta 1c ,Ra 2c ,Ta 2c ,Ra 2n ,Ta 2n ,Ra 1n ,Ta 1n ,t 1c ,t 1n ,t 2c ,t 2nThose skilled in the art can arbitrarily increase or decrease the parameters based on the structural characteristics of hollow anti-resonant optical fibers and the specific optimization parameter range, and all such increases or decreases should be considered within the scope of protection of this invention. When constructing the dataset, a particle swarm optimization algorithm was used, and the dataset was filtered using 0.01 dB / m as a standard. The dataset was divided according to the structural styles of the first and second cladding tubes of different hollow anti-resonant optical fibers. Those skilled in the art can also divide the dataset based on other optical fiber structural characteristics, and all such divisions should be considered within the scope of protection of this invention. The neural network in this embodiment is a cascaded neural network built based on a fully connected neural network. Those skilled in the art can also use other neural networks to complete the construction of the cascaded neural network, such as convolutional neural networks, recurrent neural networks, reinforcement learning neural networks, etc., and all such divisions should be considered within the scope of protection of this invention. The activation function used in this embodiment is the ReLU activation function. Those skilled in the art can also use Sigmoid or other activation functions, and all such activations should be considered within the scope of protection of this invention. The Adam optimizer used in this embodiment can be replaced with other optimizers such as stochastic gradient descent (SGD), and all such optimizations should be considered within the scope of protection of this invention. The input to the cascaded neural network is the expected characteristics and known optical fiber structural parameters. Alternatively, the optical fiber structural parameters can be omitted, or the known structural parameters and the expected design structural parameters can be adjusted.

[0073] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A reverse design method for hollow-core anti-resonant optical fiber, characterized in that, include: Obtain the expected confinement loss and known structural parameters of the hollow antiresonant fiber to be designed; Using a pre-trained prediction model, the obtained expected limiting loss and known structural parameters are processed to obtain the reverse design structural parameters of the hollow anti-resonant fiber to be designed; wherein, the prediction model includes a reverse structural design network and a forward performance prediction network; The reverse structural design network is used to process the expected limitation loss and known structural parameters to obtain the reverse design structural parameters; the forward performance prediction network is used to process the known structural parameters and the reverse design structural parameters into a complete structural parameter vector to obtain the actual predicted limitation loss. The closer the actual predicted limitation loss is to the expected limitation loss, the better the obtained reverse design structural parameters are.

2. The reverse design method for hollow-core anti-resonant optical fiber according to claim 1, characterized in that, The known structural parameters include: the style of the first cladding tube, the style of the second cladding tube, the number of cladding tubes in each layer, the core diameter, the wall thickness of the first cladding tube, the wall thickness of the nested tubes in the first cladding tube, the wall thickness of the second cladding tube, and the wall thickness of the nested tubes in the second cladding tube.

3. The reverse design method for hollow-core anti-resonant optical fiber according to claim 1, characterized in that, The reverse engineering structural parameters include: the radial axis length of the first cladding tube, the tangential axis length of the first cladding tube, the radial axis length of the second cladding tube, the tangential axis length of the second cladding tube, the radial axis length of the nested tube in the first cladding tube, the tangential axis length of the nested tube in the first cladding tube, the radial axis length of the nested tube in the second cladding tube, and the tangential axis length of the nested tube in the second cladding tube.

4. The reverse design method for hollow-core anti-resonant optical fiber according to claim 1, characterized in that, The number of hidden layers and the number of nodes in each hidden layer of the positive performance prediction network were optimized with a learning rate of 0.0002. The number of hidden layers increased from 1 to 10 layers in increments of 1, and the number of nodes increased from 50 to 800 in increments of 50. The ReLU activation function and Adam optimizer were selected during training, and the number of epochs was 5000.

5. The reverse design method for hollow-core anti-resonant optical fiber according to claim 1, characterized in that, Construct an optimization loss function Loss = α × loss for optimizing the reverse design structure network. str +β×loss cl , where loss str and loss cl Let α and β represent the mean square error loss (MSE) of the reverse design structure and the corresponding constraint loss, respectively, with α and β being the corresponding coefficients, each with a value of 1.

6. The reverse design method for hollow-core anti-resonant optical fiber according to claim 5, characterized in that, For the optimization of the reverse structure design network, the learning rate was increased from 0.0001 to 0.0020 in intervals of 0.0001, the number of hidden layers was increased from 1 to 10 in intervals of 1, and the number of nodes was increased from 50 to 800 in intervals of 50. The activation function used during training was ReLU, the optimizer was Adam, and the training epochs were 5000.

7. A reverse design system for hollow-core anti-resonant optical fiber, characterized in that, include: The acquisition module is used to acquire the expected confinement loss and known structural parameters of the hollow anti-resonant fiber to be designed. The prediction module is used to process the obtained expected limiting loss and known structural parameters using a pre-trained prediction model to obtain the reverse design structural parameters of the hollow anti-resonant fiber to be designed; wherein, the prediction model includes a reverse structural design network and a forward performance prediction network. The reverse structural design network is used to process the expected limitation loss and known structural parameters to obtain the reverse design structural parameters; the forward performance prediction network is used to process the known structural parameters and the reverse design structural parameters into a complete structural parameter vector to obtain the actual predicted limitation loss. The closer the actual predicted limitation loss is to the expected limitation loss, the better the obtained reverse design structural parameters are.

8. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the reverse design method for hollow anti-resonant optical fiber as described in any one of claims 1-6.

9. A computer program product, characterized in that, Includes a computer program, which, when run on one or more processors, is used to implement the reverse design method for hollow anti-resonant optical fiber as described in any one of claims 1-6.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions to implement the reverse design method for hollow anti-resonant optical fiber as described in any one of claims 1-6.

Images