Reverse design method and system of one-dimensional photonic crystal

Through the combination of deep neural network and conditional variational automatic encoder, the problem of prestress not being considered in the reverse design of traditional one-dimensional phonon crystals is solved, and the topological distribution and prestress are achieved simultaneous design, which improves the diversity of design solutions. and computational efficiency.

CN119069045AInactive Publication Date: 2024-12-03HAINAN UNIV
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411105539.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The traditional one-dimensional phonon crystal reverse design method fails to effectively consider prestress, resulting in fewer design solutions and high computational complexity, making it difficult to solve large engineering problems.

Method used

Deep neural network is used to establish a forward model and a conditional variational automatic encoder to establish an inverse model. The topological distribution and prestress size of the phonon crystal are input and output through the target band gap input and output, and the difference between the predicted band gap and the target band gap is calculated. Select the appropriate design plan.

Benefits of technology

The topological distribution of prestressed and phononic crystals is realized as part of the design scheme, which increases the diversity of the design scheme, simplifies the calculation process, and is suitable for on-demand designs for large projects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119069045A_ABST
    Figure CN119069045A_ABST
Patent Text Reader

Abstract

The invention relates to a reverse design method and system for a one-dimensional photonic crystal. The method comprises the following steps: establishing a forward model through a deep neural network, and establishing a reverse model through a conditional variation automatic encoder; inputting the target band gap into the reverse model to output topological distribution and prestress of the photonic crystal; inputting the topological distribution and the size of the pre-stress into a forward model, and outputting a predicted band gap; and calculating a difference value between the predicted band gap and the target band gap, and screening corresponding topological distribution and pre-stress as a design scheme for storage. A deep learning model is selected as a basis for reversely designing the photonic crystal, and finally a reverse design method and system which take the target band gap as input and can take the pre-stress and topological distribution of the photonic crystal as output at the same time are established. The topological structure of the phononic crystal and the prestress borne by the phononic crystal are used as design schemes for reverse design at the same time, the diversity of the design schemes is greatly improved, and on-demand design is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of acoustic functional material structure design, and in particular to a reverse design method and system for a one-dimensional phononic crystal. Background Art

[0002] Phononic crystals are a new type of functional material formed by periodic arrangement of elastic solids in another solid or fluid medium. The design methods of one-dimensional phononic crystals mainly include reverse design methods and adjusting the band gap characteristics by changing the cross-sectional radius. At present, most of the reverse design methods for phononic crystals are topological optimization methods. When the number of iterations of traditional topological optimization methods increases, the memory and time required for computer calculations will increase exponentially, making it difficult for topological optimization methods to solve large engineering problems. In addition, the current reverse design only reverses the topological structure of the phononic crystal, without taking prestress as part of the design solution.

[0003] Therefore, the traditional one-dimensional phononic crystal reverse design does not take the influence of prestress into consideration, resulting in few design solutions that meet the requirements. Summary of the invention

[0004] Based on this, in order to solve the above technical problems, a reverse design method and system for one-dimensional phononic crystals are provided, which can simultaneously output the prestress and the topological distribution of the phononic crystal to achieve on-demand design.

[0005] A reverse design method for a one-dimensional phononic crystal, the method comprising:

[0006] The forward model is built through a deep neural network, and the reverse model is built through a conditional variational autoencoder;

[0007] Determine a target band gap, and input the target band gap into the inverse model, and output the topological distribution and prestress magnitude of the phononic crystal through the inverse model;

[0008] Inputting the topological distribution and the prestress magnitude into the forward model, and outputting a predicted band gap through the forward model;

[0009] The difference value between the predicted band gap and the target band gap is calculated, and if the difference value is within a threshold range, the topological distribution and prestress corresponding to the predicted band gap are saved as a design solution.

[0010] In one embodiment, the forward model includes a fully connected layer and an activation function layer.

[0011] In one embodiment, the fully connected layer and the activation function layer each have five layers; and the number of neurons in each layer of the fully connected layer is 196, 196, 196, 196, and 160 respectively.

[0012] In one embodiment, the topological distribution and prestress magnitude input into the forward model are 98-dimensional vectors;

[0013] The predicted band gap output from the forward model includes a first order band gap and a second order band gap;

[0014] The first-order band gap and the second-order band gap form a four-dimensional vector output;

[0015] The forward model is trained deterministically using a supervised training method.

[0016] In one embodiment, the four-dimensional vector includes a first-order band gap lower limit, a first-order band gap upper limit, a second-order band gap lower limit, and a second-order band gap upper limit.

[0017] In one of the embodiments, the conditional variational autoencoder includes an encoder and a decoder.

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

[0019] Determining design variables, wherein the design variables are used to represent the magnitude of prestress and the distribution of phononic crystal materials;

[0020] The step of inputting the target band gap into the inverse model and outputting the topological distribution and prestress magnitude of the phononic crystal through the inverse model includes:

[0021] Inputting the design variables and the target band gap into the encoder;

[0022] Mapping the design variables and the target band gap into latent variables through the encoder;

[0023] The hidden variables and the target band gap are input into the decoder to generate the topological distribution and prestress size of the phononic crystal.

[0024] In one embodiment, the method further comprises:

[0025] The inverse model is obtained by optimizing by minimizing reconstruction error and KL divergence calculation method.

[0026] A reverse design system for a one-dimensional phononic crystal, the system comprising:

[0027] Model building module, which is used to build forward models through deep neural networks and reverse models through conditional variational autoencoders;

[0028] An inverse model operation module, used for determining a target band gap, inputting the target band gap into the inverse model, and outputting the topological distribution and prestress magnitude of the phononic crystal through the inverse model;

[0029] A forward model operation module, used for inputting the topological distribution and the prestress magnitude into the forward model, and outputting a predicted band gap through the forward model;

[0030] The design scheme screening module is used to calculate the difference value between the predicted band gap and the target band gap. If the difference value is within a threshold range, the topological distribution and prestress corresponding to the predicted band gap are saved as a design scheme.

[0031] The above-mentioned one-dimensional phononic crystal reverse design method and system establishes a forward model through a deep neural network, establishes a reverse model through a conditional variational autoencoder, outputs a predicted band gap through the established model, and then calculates the difference between the predicted band gap and the target band gap, and selects the topological distribution and prestress as the design scheme to save; that is, the deep learning model is selected as the basis for reverse design of phononic crystals, and finally a reverse design method and system is established with the target band gap as input, which can simultaneously output the prestress and the topological distribution of the phononic crystal, and the topological structure of the phononic crystal and the prestress of the phononic crystal are simultaneously used as design schemes for reverse design, which greatly increases the diversity of design schemes and realizes on-demand design. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A diagram showing an application environment of a reverse design method for a one-dimensional phononic crystal in one embodiment;

[0033] Figure 2 A schematic diagram of a process flow of a reverse design method of a one-dimensional phononic crystal in one embodiment;

[0034] Figure 3 A schematic diagram of the structure of a reverse model in one embodiment;

[0035] Figure 4 A schematic diagram of the structure of a forward model in one embodiment;

[0036] Figure 5 is a structural block diagram of a reverse design system for a one-dimensional phononic crystal in one embodiment;

[0037] Figure 6 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0039] The reverse design method of the one-dimensional phononic crystal provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Figure 1 As shown, the application environment includes a computer device 110. The computer device 110 can establish a forward model through a deep neural network and establish an inverse model through a conditional variational autoencoder; the computer device 110 can determine the target band gap, and input the target band gap into the inverse model, and output the topological distribution and prestress of the phononic crystal through the inverse model; the computer device 110 can input the topological distribution and prestress into the forward model, and output the predicted band gap through the forward model; the computer device 110 can calculate the difference between the predicted band gap and the target band gap, and if the difference is within a threshold range, the topological distribution and prestress corresponding to the predicted band gap are saved as a design solution. Among them, the computer device 110 can be, but is not limited to, various personal computers, laptops, smart phones, robots, tablet computers and other devices.

[0040] In one embodiment, Figure 2 As shown, a reverse design method of a one-dimensional phononic crystal is provided, comprising the following steps:

[0041] Step 202: Establish a forward model through a deep neural network, and establish a reverse model through a conditional variational autoencoder.

[0042] Among them, the role of the forward model is to input the topological distribution of the sub-crystal and the size of the prestress it is subjected to into the forward model, and the model outputs the band gap; the role of the reverse model is to input the band gap into the reverse model, and the model outputs the topological distribution of the phononic crystal and the size of the prestress it should be subjected to.

[0043] Step 204, determine the target band gap, and input the target band gap into the inverse model, and output the topological distribution and prestress magnitude of the phononic crystal through the inverse model.

[0044] In one embodiment, Figure 3 As shown, the conditional variational autoencoder CVAE includes an encoder and a decoder.

[0045] In one embodiment, a reverse design method for a one-dimensional phononic crystal is provided, which may also include an output process of a reverse model, the specific process including: determining design variables, the design variables are used to represent the prestress and the distribution of the phononic crystal material; inputting the design variables and the target band gap into an encoder; mapping the design variables and the target band gap into hidden variables through the encoder; inputting the hidden variables and the target band gap into a decoder to generate the topological distribution and prestress of the phononic crystal. In one embodiment, the reverse model is optimized by minimizing the reconstruction error and the KL divergence calculation method.

[0046] Specifically, in CVAE, the design variables (i.e., the 97-dimensional vector representing the prestress and the distribution of the phononic crystal material) and the target band gap are first input into the encoder, and then the design variables and the target band gap are finally mapped to the latent variable z; the latent variable z follows a Gaussian distribution with an average value of α and a standard deviation of γ; then the target band gap and the latent variable z are input into the decoder to finally generate a design solution. In the training process, the trained forward model is used to assist in training the inverse model, so that the loss function L CVAE Reach the minimum. Its mathematical expression is: in:

[0047]

[0048]

[0049] Among them, m represents the number of training sample sets, α k and γ k They represent the mean and standard deviation of the kth element in the latent variable z, n represents the dimension of the latent variable z and is also the dimension of α and γ (i.e. 98), and N represents the dimension of the vector of the phononic crystal material distribution (i.e. 97).

[0050] In this embodiment, for the inverse model, the issues that need to be considered are more complex than those for the forward model. For the same bandgap structure, there are likely to be multiple design schemes that meet the requirements of the bandgap at the same time, so the inverse model needs to generate multiple design schemes for a given input. In order to make the inverse model have this effect, the conditional variational autoencoder (CVAE) is used as the basis for establishing the inverse model. As a neural network based on probabilistic supervised learning, CVAE optimizes the model by minimizing the reconstruction error and KL divergence; wherein the reconstruction error measures the difference between the reconstructed sample and the original sample, and the KL divergence is used to measure the difference between the distribution of the latent variable and the standard normal distribution. Through the optimization of these error terms, CVAE can better learn the distribution characteristics of the input samples, and can produce multiple outputs for a given input by running multiple times.

[0051] Step 206, input the topological distribution and the prestress magnitude into the forward model, and output the predicted band gap through the forward model.

[0052] In one embodiment, Figure 4 As shown in Figure 1, the forward model includes a fully connected layer and an activation function layer. There are five layers in each of the fully connected layers and the activation function layer, and the number of neurons in each layer of the fully connected layer is 196, 196, 196, 196, and 160 respectively.

[0053] In one embodiment, the topological distribution and prestress magnitude input into the forward model are 98-dimensional vectors; the predicted band gap output from the forward model includes the first-order band gap and the second-order band gap; the first-order band gap and the second-order band gap form a four-dimensional vector output; the forward model is deterministically trained using supervised training. The four-dimensional vector includes the first-order band gap lower limit, the first-order band gap upper limit, the second-order band gap lower limit, and the second-order band gap upper limit.

[0054] Specifically, the input of the forward model is the prestress σ0 plus the topological structure a of the phononic crystal. k (k = 1-97), is a 98-dimensional vector; the output is the predicted first-order bandgap lower limit Ω1 ’ l , the upper limit of the first-order band gap Ω1 ’ u , the lower limit of the second-order band gap Ω2 ’ l and the upper limit of the second-order band gap Ω2 ’ u During training, supervised training is used to perform deterministic training on the model to minimize the loss function L DNN The purpose of this is expressed mathematically as follows:

[0055]

[0056] In this embodiment, for the forward model, a unique output needs to be generated for a given input, because for the reverse design method of the one-dimensional phononic crystal in this embodiment, each design scheme must correspond to only one certain band gap structure. Therefore, for this problem, a deep neural network (DNN) is selected as the basis for establishing a forward model, because DNN, as a model based on deterministic supervised learning, will always produce the same output for a given input, which can well meet the requirements of the forward model.

[0057] Step 208, calculating the difference between the predicted band gap and the target band gap, and if the difference is within a threshold range, saving the topological distribution and prestress corresponding to the predicted band gap as a design solution.

[0058] In this embodiment, both the forward model and the reverse model are trained and the results are good. After the target band gap is determined, the target band gap frequency is input into the reverse model. After the reverse model outputs the design scheme, in order to make the output design scheme feasible, the design scheme needs to be screened. Specifically, the design scheme is input into the forward model as an input value, and the predicted band gap is finally output. The predicted band gap is compared with the target band gap. It is allowed that the predicted band gap is slightly different from the target band gap. As long as MAPE≤1% is satisfied, the design scheme corresponding to the band gap is retained.

[0059] It should be understood that, although the various steps in the above-mentioned flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above-mentioned flow chart may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0060] In one embodiment, Figure 5 As shown, a reverse design system for a one-dimensional phononic crystal is provided, comprising: a model building module 510, a reverse model operation module 520, a forward model operation module 530 and a design solution screening module 540, wherein:

[0061] A model building module 510, for building a forward model through a deep neural network and building a reverse model through a conditional variational autoencoder;

[0062] The inverse model operation module 520 is used to determine the target band gap, input the target band gap into the inverse model, and output the topological distribution and prestress magnitude of the phononic crystal through the inverse model;

[0063] A forward model operation module 530 is used to input the topological distribution and the prestress magnitude into the forward model, and output the predicted band gap through the forward model;

[0064] The design scheme screening module 540 is used to calculate the difference value between the predicted band gap and the target band gap. If the difference value is within a threshold range, the topological distribution and prestress corresponding to the predicted band gap are saved as a design scheme.

[0065] In one embodiment, the forward model includes a fully connected layer and an activation function layer.

[0066] In one embodiment, there are five fully connected layers and five activation function layers; and the number of neurons in each of the fully connected layers is 196, 196, 196, 196, and 160 respectively.

[0067] In one embodiment, the topological distribution and prestress magnitude input into the forward model are 98-dimensional vectors; the predicted band gap output from the forward model includes a first-order band gap and a second-order band gap; the first-order band gap and the second-order band gap form a four-dimensional vector output; the forward model is deterministically trained using supervised training.

[0068] In one embodiment, the four-dimensional vector includes a first-order band gap lower limit, a first-order band gap upper limit, a second-order band gap lower limit, and a second-order band gap upper limit.

[0069] In one embodiment, the conditional variational autoencoder includes an encoder and a decoder.

[0070] In one embodiment, the inverse model operation module 520 is also used to determine the design variables, which are used to represent the prestress size and the distribution of the phononic crystal material; the design variables and the target band gap are input into the encoder; the design variables and the target band gap are mapped to the hidden variables through the encoder; the hidden variables and the target band gap are input into the decoder to generate the topological distribution and prestress size of the phononic crystal.

[0071] In one embodiment, the inverse model is obtained by optimizing by minimizing the reconstruction error and calculating the KL divergence.

[0072] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device 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 computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a reverse design method of a one-dimensional phononic crystal is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0073] Those skilled in the art will understand that Figure 6The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0074] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps of a reverse design method of a one-dimensional phononic crystal when executing the computer program.

[0075] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of a reverse design method of a one-dimensional phononic crystal are implemented.

[0076] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods 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 in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can 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).

[0077] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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, they should be considered to be within the scope of this specification.

[0078] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A reverse design method for a one-dimensional phononic crystal, characterized in that: The method comprises: The forward model is built through a deep neural network, and the reverse model is built through a conditional variational autoencoder; Determine a target band gap, and input the target band gap into the inverse model, and output the topological distribution and prestress magnitude of the phononic crystal through the inverse model; Inputting the topological distribution and the prestress magnitude into the forward model, and outputting a predicted band gap through the forward model; The difference value between the predicted band gap and the target band gap is calculated, and if the difference value is within a threshold range, the topological distribution and prestress corresponding to the predicted band gap are saved as a design solution.

2. The reverse design method of one-dimensional phononic crystal according to claim 1, characterized in that: The forward model includes a fully connected layer and an activation function layer.

3. The reverse design method of one-dimensional phononic crystal according to claim 2, characterized in that: The fully connected layer and the activation function layer both have five layers; and the number of neurons in each layer of the fully connected layer is 196, 196, 196, 196, and 160 respectively.

4. The reverse design method of one-dimensional phononic crystal according to claim 2, characterized in that: The topological distribution and prestress magnitude input into the forward model are 98-dimensional vectors; The predicted band gap output from the forward model includes a first order band gap and a second order band gap; The first-order band gap and the second-order band gap form a four-dimensional vector output; The forward model is trained deterministically using a supervised training method.

5. The reverse design method of one-dimensional phononic crystal according to claim 4, characterized in that: The four-dimensional vector includes a first-order band gap lower limit, a first-order band gap upper limit, a second-order band gap lower limit, and a second-order band gap upper limit.

6. The reverse design method of one-dimensional phononic crystal according to claim 1, characterized in that: The conditional variational autoencoder includes an encoder and a decoder.

7. The reverse design method of one-dimensional phononic crystal according to claim 6, characterized in that: The method further comprises: Determining design variables, wherein the design variables are used to represent the magnitude of prestress and the distribution of phononic crystal materials; The step of inputting the target band gap into the inverse model and outputting the topological distribution and prestress magnitude of the phononic crystal through the inverse model includes: Inputting the design variables and the target band gap into the encoder; Mapping the design variables and the target band gap into latent variables through the encoder; The hidden variables and the target band gap are input into the decoder to generate the topological distribution and prestress size of the phononic crystal.

8. The reverse design method of one-dimensional phononic crystal according to claim 7, characterized in that: The method further comprises: The inverse model is obtained by optimizing by minimizing reconstruction error and KL divergence calculation method.

9. A one-dimensional phononic crystal reverse design system, characterized in that: The system comprises: Model building module, which is used to build forward models through deep neural networks and reverse models through conditional variational autoencoders; An inverse model operation module, used for determining a target band gap, inputting the target band gap into the inverse model, and outputting the topological distribution and prestress magnitude of the phononic crystal through the inverse model; A forward model operation module, used for inputting the topological distribution and the prestress magnitude into the forward model, and outputting a predicted band gap through the forward model; The design scheme screening module is used to calculate the difference value between the predicted band gap and the target band gap. If the difference value is within a threshold range, the topological distribution and prestress corresponding to the predicted band gap are saved as a design scheme.

Citation Information

Patent Citations

  • High-toughness high-temperature-resistant RAFM steel and design method thereof based on machine learning

    CN113073267A

  • Periodic super-barrier reverse design method based on deep learning

    CN114943162A

  • Phononic crystal with adjustable band gap and band gap adjusting method

    CN115273781A

  • Marine observation control method and device for adaptive sampling of multiple underwater robots

    CN116048106A

  • Phononic crystal band gap design method and design system based on intelligent coding

    CN116776752A