CN116776752A - Phonon crystal band gap design method and system based on intelligent coding - Google Patents
Phonon crystal band gap design method and system based on intelligent coding Download PDFInfo
- Publication number
- CN116776752A CN116776752A CN202311069161.1A CN202311069161A CN116776752A CN 116776752 A CN116776752 A CN 116776752A CN 202311069161 A CN202311069161 A CN 202311069161A CN 116776752 A CN116776752 A CN 116776752A
- Authority
- CN
- China
- Prior art keywords
- band gap
- data
- precision
- photonic crystal
- optimized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013461 design Methods 0.000 title claims abstract description 68
- 238000000034 method Methods 0.000 title claims abstract description 60
- 239000013078 crystal Substances 0.000 title claims description 50
- 239000013598 vector Substances 0.000 claims abstract description 92
- 239000004038 photonic crystal Substances 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 42
- 238000004364 calculation method Methods 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 14
- 238000003062 neural network model Methods 0.000 claims abstract description 14
- 238000011156 evaluation Methods 0.000 claims abstract description 10
- 238000004088 simulation Methods 0.000 claims abstract description 9
- 230000006870 function Effects 0.000 claims description 45
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 230000035772 mutation Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 239000006185 dispersion Substances 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000012512 characterization method Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 238000004806 packaging method and process Methods 0.000 claims description 3
- 230000007547 defect Effects 0.000 abstract description 4
- 238000005457 optimization Methods 0.000 description 26
- 210000000349 chromosome Anatomy 0.000 description 19
- 239000000463 material Substances 0.000 description 11
- 210000002569 neuron Anatomy 0.000 description 10
- 239000011159 matrix material Substances 0.000 description 8
- 230000004913 activation Effects 0.000 description 7
- 230000002068 genetic effect Effects 0.000 description 6
- 210000004027 cell Anatomy 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 108090000623 proteins and genes Proteins 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/26—Composites
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/10—Noise analysis or noise optimisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Computer Hardware Design (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Mathematical Physics (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Databases & Information Systems (AREA)
- Complex Calculations (AREA)
Abstract
The invention relates to the technical field of noise and vibration control, and discloses a photonic crystal band gap design method and a photonic crystal band gap design system based on intelligent coding, wherein the photonic crystal band gap design method comprises the following steps: training a convolutional neural network model and a deep neural network model based on a database to obtain a topological structure evaluation environment; inputting the target band gap distribution data and the target maximum band gap order into an iterative model for iteration to obtain an optimized feature vector; performing precision calculation on the band gap corresponding to the optimized feature vector to determine optimized image data; generating corresponding vector data based on the optimized image data, performing finite element simulation on the vector data to obtain optimized band gap distribution data, and performing precision calculation on the optimized band gap distribution data and band gaps required by users to determine the topological structure of the photonic crystal band gaps; the invention solves the problem that the existing photonic crystal band gap design method based on intelligent coding can not be realized to consider the user demand and avoid the non-uniqueness defect.
Description
Technical Field
The invention relates to the technical field of noise and vibration control, in particular to a photonic crystal band gap design method and system based on intelligent coding.
Background
The phonon crystal is a composite material with two or more materials which are periodically arranged according to a certain matching structure in space and has elastic wave band gap, and the phonon crystal has the capability of breaking through the upper limit of the vibration reduction performance of the traditional material due to the artificial design factors. However, the conventional photonic crystal design method uses numerical tools or theoretical derivation in an iterative trial-and-error manner to achieve excellent band gap characteristics. This dynamic design approach can be computationally expensive. Meanwhile, the band gap characteristic of the phonon crystal has strong nonlinear characteristics, and batch trial and error based on parameter gradiometry can miss the optimal parameter values among partial gradients.
In recent years, artificial intelligence and optimization algorithms are developed, and are applied to the field of photonic crystal structure design, and a disposable design concept that a user needs to input and a photonic crystal unit structure outputs is created. The on-demand design method for directly obtaining the photonic crystal topological structure by using the target band gap distribution as a guide through a modularized algorithm can greatly reduce the design cost.
The current phonon crystal on-demand design mainly involves the following two ideas: (1) The method belongs to unidirectional design of phonon crystals, only takes a certain characteristic of a band gap as a fixed and single optimization direction, and cannot perform personalized design on band gap distribution as required by users. At the same time, the way to obtain the characteristic frequency by theoretical calculation is complex, even impossible, and the simplification depends in part on subjective understanding of the performer. If numerical simulation is used as a tool for calculating the band gap, the increase of the calculation cost can make a large-scale optimization program difficult to perform; (2) Directly establishing nonlinear association of a photonic crystal topological structure and band gap distribution characteristics, directly taking band gap distribution requirements as input characteristics, obtaining a single-value solution of the topological structure, converting reverse design of the photonic crystal into a problem of classifying multi-component materials of a unit cell topological structure, and directly establishing a mapping relation between a band gap distribution one-dimensional matrix and material distribution based on an artificial neural network algorithm. The design mode belongs to single-value design of phononic crystals, has higher data requirements, and the model after training is completely limited to training sample data when engineering application is finished, so that expansibility is lacking. Meanwhile, the modeling mode of causal substitution cannot avoid a non-unique trap of the backscattering problem, and falls into a one-to-one easily-biased state completely judged by a model, so that a topological design result which is completely contrary to reality is easy to generate. Both the two ways can not realize the on-demand design which considers the demands of users and avoids non-unique defects, and can not provide substantial theoretical guidance for the engineering application design of phonon crystals in a real sense.
Disclosure of Invention
The invention provides a photonic crystal band gap design method and a photonic crystal band gap design system based on intelligent coding, which are used for solving the problem that the existing photonic crystal band gap design method based on intelligent coding cannot be used for considering the requirements of users and avoiding non-unique defects.
In order to achieve the above object, the present invention is realized by the following technical scheme:
in a first aspect, the present invention provides a photonic crystal bandgap design method based on intelligent coding, including:
acquiring basic configuration data of phononic crystals based on user requirements, constructing a database based on the basic configuration data, and constructing and training a convolutional neural network model by utilizing data in the database;
converting the image data in the basic configuration data through a trained convolutional neural network model to obtain a feature vector, and combining the feature vector with corresponding feature frequency data in a database to obtain training data;
constructing and training a deep neural network model by utilizing training data, storing and packaging the deep neural network model to form a topological structure evaluation environment, and obtaining band gap distribution data and a maximum band gap order based on the feature vector and the trained deep neural network model;
Constructing an iterative model based on a meta heuristic algorithm, constructing a defined fitness function based on a topological structure evaluation environment, setting iteration basic parameters, and inputting a characteristic vector of a target band gap and a maximum band gap order into the iterative model to obtain an optimized characteristic vector;
performing precision calculation on the optimized characteristic vector corresponding to the band gap and the characteristic vector of the band gap required by the user to obtain first precision, and determining optimized image data based on the first precision;
generating corresponding vector data based on the optimized image data, performing finite element simulation on the vector data to obtain optimized band gap distribution data, performing precision calculation on the optimized band gap distribution data and a band gap required by a user to obtain second precision, and taking the band gap distribution data as a topological structure of a phonon crystal band gap based on the second precision.
Optionally, the constructing a database based on the basic configuration data includes:
and constructing a basic configuration curve coordinate function based on the basic configuration data, generating image data and characteristic frequency data of the phonon crystal according to the basic configuration curve coordinate function, and constructing a database through the image data and the characteristic frequency data.
Optionally, the defining process of the fitness function includes:
the fitness function is defined in three parts:
the first part, the part of the user requirement band gap represented by the gamma interval is compared with the individual band gap of the stop band, each group of the pass bands can be correctly corresponding to each other, and 5 minutes are added;
setting an ideal upper limit, wherein the upper limit score is L/100 min, subtracting the difference between the upper limit score and the lower limit score, and taking the number of gamma intervals as a calculation unit;
and the third part, the upper limit is set as Q-2, and the score of the third part is obtained by subtracting the separated order difference value from the upper limit score.
Optionally, the iteration basic parameters include:
inheritance rate, crossover rate, mutation rate, population size and evolution number.
Optionally, the determining the optimized image data based on the first precision includes:
and comparing the first precision with the preset precision, inputting the optimized feature vector into the trained neural network again for a new iteration when the first precision is smaller than the preset precision, and inputting the optimized feature vector into the trained convolutional neural network model for inverse decoding when the first precision is larger than or equal to the preset precision to obtain optimized image data.
Optionally, the topology structure using the optimized band gap distribution data as the phonon crystal band gap based on the second precision includes:
and comparing the second precision with the preset precision, when the second precision is smaller than the preset precision, acquiring basic configuration data of the photonic crystal again based on the user requirement, and when the second precision is larger than or equal to the preset precision, taking the optimized band gap distribution data as the topological structure of the band gap of the photonic crystal.
Optionally, the convolutional neural network model includes: an encoder and a decoder;
the encoder converts the image data into feature vectors by extracting information in the image data, so that the structural information of phonon crystals can be expressed in a vector form and participate in band gap distribution prediction;
the decoder decodes the feature vector into image data, and outputs the image data as a topology of phonon crystals.
Optionally, the deep neural network model includes: a band gap distribution prediction model and a band gap main order prediction model;
the input characteristic of the band gap distribution prediction model is the characteristic vector of the phonon crystal structure, the output characteristic is the band gap distribution, and the band gap distribution is characterized in a mode of numerical coding of a passband and a stopband;
The input characteristic of the band gap main order prediction model is the characteristic vector of the phonon crystal structure, and the output characteristic is the order of the maximum band gap.
Optionally, the feature vector of the user-required band gap includes:
and carrying out band gap characterization on the user demand band gap to obtain a characteristic vector of the user demand band gap, wherein the gamma interval is a range interval obtained by setting the upper limit of a band gap range and setting the accuracy.
In a second aspect, an embodiment of the present application provides a photonic crystal bandgap design system based on intelligent encoding, including a processor and a memory;
a memory for storing a computer program;
a processor for implementing the method steps described in the first aspect when executing a program stored on a memory.
The beneficial effects are that:
according to the photonic crystal band gap design method based on intelligent coding, the feature vector is globally optimized based on the genetic algorithm, the excellent photonic crystal structure with good band gap distribution can be obtained without repeated iterative trial and error, operation resources are saved, in actual application, the optimal photonic crystal structure can be directly obtained by inputting a user basic configuration and a band gap required by a user into the design method according to requirements, the modularization degree is high, under the framework, other constraint conditions and optimization targets are added, and the photonic crystal band gap design method based on intelligent coding can be applied to different scenes and has high engineering value;
According to the photonic crystal band gap design method based on intelligent coding, provided by the invention, a nonlinear fitting model is built on the basis of an accurate forward prediction model instead of simply exchanging a characteristic value and a target value, so that the non-uniqueness defect of a backscattering problem is completely avoided.
The reverse search optimization process is different from the fitting calculation of a one-to-one model, and the easy deviation state depending on a single-value model is stopped. Particularly, when searching the singular band gap characteristics under the basic configuration required by the user, the upper limit of the fitness under the configuration can be searched in the optimization process, and the capability limit of the basic configuration of the user can be obtained by adjusting the key attention factors of the fitness.
In the genetic algorithm optimization process, the proposed fitness slipping strategy can gradually slip the band gap required by the user even if band gap distribution of all individuals is not intersected with the band gap required by the user in the initial iteration of the population, so that the optimization algorithm can smoothly transition to the middle and late stages. The provided coding expansion strategy can expand the range of the chromosome of the individual, expand the basic search space of the optimization algorithm, and is not limited by the original training sample data simply, so that the optimization algorithm has obvious expansibility.
The proposed on-demand design method has the capability of processing images and generating images for design objects, and can grasp key information of phonon crystal structures as feature vectors to participate in prediction and optimization. For design targets, band gap distribution and band gap orders can be comprehensively considered, the weight of the fitness function can be adjusted according to different user demands, so that personalized structural designs with different emphasis are realized, the flexibility is high, and meanwhile, manufacturability factors and engineering characteristic factors can be added into the fitness function design in engineering application, so that the adaptability is high.
Drawings
FIG. 1 is one of the flow charts of a photonic crystal bandgap design method based on smart coding in accordance with a preferred embodiment of the present invention;
FIG. 2 is a schematic diagram of database construction according to a preferred embodiment of the present invention;
FIG. 3 is a schematic diagram of a self-encoding module, a forward prediction module, and a backward prediction module in an on-demand design method according to a preferred embodiment of the present invention;
FIG. 4 is a schematic diagram of a band gap distribution prediction model according to a preferred embodiment of the present invention;
FIG. 5 is a second flowchart of a photonic crystal bandgap design method based on smart coding according to a preferred embodiment of the present invention.
Detailed Description
The following description of the present invention will be made clearly and fully, and it is apparent that the embodiments described are only some, but not all, of the embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Unless defined otherwise, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this invention belongs. The terms "first," "second," and the like, as used herein, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. Likewise, the terms "a" or "an" and the like do not denote a limitation of quantity, but rather denote the presence of at least one. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "upper", "lower", "left", "right", etc. are used merely to indicate a relative positional relationship, which changes accordingly when the absolute position of the object to be described changes.
Referring to fig. 1-5, an embodiment of the present application provides a photonic crystal bandgap design method based on intelligent coding, including:
acquiring basic configuration data of phononic crystals based on user requirements, constructing a database based on the basic configuration data, and constructing and training a convolutional neural network model by utilizing data in the database;
converting the image data in the basic configuration data through the trained convolutional neural network model to obtain a feature vector, and combining the feature vector with corresponding feature frequency data in a database to obtain training data;
constructing and training a deep neural network model by utilizing training data, storing and packaging the deep neural network model to form a topological structure evaluation environment, and obtaining band gap distribution data and a maximum band gap order based on feature vectors and the trained deep neural network model;
constructing an iterative model based on a meta heuristic algorithm, constructing a defined fitness function based on a topological structure evaluation environment, setting iteration basic parameters, and inputting a characteristic vector of a target band gap and a maximum band gap order into the iterative model to obtain an optimized characteristic vector;
performing precision calculation on the optimized characteristic vector corresponding to the band gap and the characteristic vector of the band gap required by the user to obtain first precision, and determining optimized image data based on the first precision;
Generating corresponding vector data based on the optimized image data, performing finite element simulation on the vector data to obtain optimized band gap distribution data, performing precision calculation on the optimized band gap distribution data and a band gap required by a user to obtain second precision, and taking the band gap distribution data as a topological structure of a phonon crystal band gap based on the second precision.
In the embodiment, a database is constructed according to the user requirements, and a topological structure evaluation environment is obtained by training a convolutional neural network model and a deep neural network model based on the database; inputting the target band gap distribution data and the target maximum band gap order into an iterative model for iteration to obtain an optimized feature vector; performing precision calculation on the feature vector of the band gap corresponding to the optimized feature vector and the feature vector of the band gap required by the user to determine optimized image data; generating corresponding vector data based on the optimized image data, performing finite element simulation on the vector data to obtain optimized band gap distribution data, and performing precision calculation on the optimized band gap distribution data and band gaps required by users to determine the topological structure of the photonic crystal band gaps.
The feature vector is globally optimized based on a genetic algorithm, so that an excellent phonon crystal structure with good band gap distribution can be obtained without repeated iterative trial and error, and operation resources are saved. In practical application, the optimal phonon crystal structure can be directly obtained by inputting the user basic configuration and the user requirement band gap into a design method according to the requirement, the modularization degree is high, and under the framework, other constraint conditions and optimization targets are added, so that the method can be applied to different scenes and has high engineering value.
The workflow is as shown in fig. 5:
step 1, firstly, setting an accuracy threshold P designed according to the requirement. The dimensions and basic configuration of the phononic crystal are determined according to the user's needs. And dividing equidistant variable parameter combinations according to the variable parameters acquired by the basic configuration, and acquiring training data in batches to form a database.
And 2, training the self-coding module by utilizing database data to obtain an encoder and a decoder. PNG data in a database are converted into feature vectors of data samples through an encoder, and are combined with CSV data representing band gap distribution in the database to obtain training data, and a band gap distribution prediction model and a band gap main order prediction model in a forward prediction module are trained to serve as evolutionary environments of genetic algorithms.
And 3, directly carrying out band gap characterization of a gamma interval on a band gap required by a user, and directly transmitting the band gap into an adaptability function in a reverse optimization module. Basic parameters of the genetic algorithm are set, including inheritance rate, crossover rate, mutation rate, population scale and evolution times. The band gap corresponding to the optimized individual is compared with the band gap required by the user, and the calculation accuracy is improvedIf the value is less than P, repeating the step 3, and carrying out the optimization of the reverse optimization module again until +. >Greater than P.
Step 4, inputting the optimal individual meeting the precision requirement into a decoder, restoring the optimal individual into PNG format data, converting the PNG format data into DXF format, guiding the PNG format data into Comsol for finite element simulation or performing entity experiment, comparing the obtained band gap distribution with the band gap required by the user, and calculating the precisionIf the data is smaller than P, repeating the step 1, and acquiring the data, training the model and optimizing the structure again until the +.>And (3) storing the topological structure of the phonon crystal to finish the design, wherein the topological structure is larger than P.
Optionally, constructing the database based on the basic configuration data includes:
and constructing a basic configuration curve coordinate function based on the basic configuration data, generating image data and characteristic frequency data of the phonon crystal according to the basic configuration curve coordinate function, and constructing a database through the image data and the characteristic frequency data.
Optionally, the defining process of the fitness function includes:
the fitness function is defined in three parts:
the first part, the part of the user requirement band gap represented by the gamma interval is compared with the individual band gap of the stop band, each group of the pass bands can be correctly corresponding to each other, and 5 minutes are added;
setting an ideal upper limit, wherein the upper limit score is L/100 min, subtracting the difference between the upper limit score and the lower limit score, and taking the number of gamma intervals as a calculation unit;
And the third part, the upper limit is set as Q-2, and the score of the third part is obtained by subtracting the separated order difference value from the upper limit score.
In the above embodiment, the fitness function is divided into three parts, the first part is a main body part, and is the corresponding degree of band gap distribution, that is, the part of the user requirement band gap represented by the gamma interval is a stop band, which is compared with the individual band gaps, and each group of the pass bands can correctly correspond to each other, and 5 minutes are added. However, in order to prevent the initial population from being 0 points, the population cannot normally evolve, and thus the fitness of the second part is set. The second part is the bandwidth of the main bandgap and the closeness of the lower bandgap boundary. The ideal situation is taken as an upper limit, the upper limit score is L/100 score and L/100 score respectively, the difference between the two scores is subtracted, and the number of gamma intervals is taken as a calculation unit. The second part can be used for initializing non-ideal population, and can also slowly slide in the designed band gap direction until the first part contains overlapping fraction, and then the evolution is continued based on the fitness of the first part. The third part is the degree of discretization with the target master. The upper limit of the part is set as Q-2, and the difference value of the separated orders subtracted from the upper limit fraction is the score of the third part. And finally, weighting the three scores according to engineering application requirements to obtain a final fitness function. The specific process is as follows:
；
Wherein the method comprises the steps of，/>，/>，/>The fitness score, the first part score, the second part score and the third part score are respectively; />，/>The user demand band gap of the ith gamma interval and the current individual band gap distribution situation are respectively 0 or 1; />The number of the gamma intervals; />，/>Respectively calculating positions of a target main band gap and a current individual main band gap according to a gamma interval sequence number; />，/>The widths of the target main band gap and the current main band gap are respectively in gamma regionCalculating the number of the intervals; />Is the maximum order of the dispersion curve; />，/>The target master orders and the master orders of the current individuals are respectively; />，/>，/>The three fractional weight coefficients are respectively.
Optionally, the iteration basic parameters include:
inheritance rate, crossover rate, mutation rate, population size and evolution number.
Optionally, determining the optimized image data based on the first precision includes:
and comparing the first precision with the preset precision, inputting the optimized feature vector into the trained neural network again for a new iteration when the first precision is smaller than the preset precision, and inputting the optimized feature vector into the trained convolutional neural network model for inverse decoding when the first precision is larger than or equal to the preset precision to obtain optimized image data.
Optionally, regarding the optimized band gap distribution data as a topology of phonon crystal band gaps based on the second precision includes:
and comparing the second precision with the preset precision, when the second precision is smaller than the preset precision, acquiring basic configuration data of the photonic crystal again based on the user requirement, and when the second precision is larger than or equal to the preset precision, taking the optimized band gap distribution data as the topological structure of the band gap of the photonic crystal.
Optionally, the convolutional neural network model includes: an encoder and a decoder;
the encoder converts the image data into feature vectors by extracting information in the image data, so that the structural information of phonon crystals can be expressed in a vector form and participate in band gap distribution prediction;
the decoder decodes the feature vector into image data, and outputs the image data as a topology of phonon crystals.
Optionally, the deep neural network model includes: a band gap distribution prediction model and a band gap main order prediction model;
the input characteristic of the band gap distribution prediction model is the characteristic vector of the phonon crystal structure, the output characteristic is the band gap distribution, and the band gap distribution is characterized in a mode of numerical coding of a passband and a stopband;
The input characteristic of the band gap main order prediction model is the characteristic vector of the phonon crystal structure, and the output characteristic is the order of the maximum band gap.
Optionally, the feature vector of the user-required band gap includes:
and carrying out band gap characterization on the user demand band gap to obtain a characteristic vector of the user demand band gap, wherein the gamma interval is a range interval obtained by setting the upper limit of a band gap range and setting the accuracy.
Mainly comprises the following steps: 1) Establishing a database; 2) And (5) establishing an on-demand design method. The database is used for creating a trainable data sample according to a basic configuration required by a user, and providing data support for an on-demand design method; firstly, a self-coding module, a forward prediction module and a reverse design module are established according to the design method, database data samples are respectively imported into model frames of the modules for training, and a comprehensive design-on-demand model which can be directly applied is formed; and finally, directly inputting the band gap required by the user into the comprehensive model designed according to the requirement, obtaining the topological structure with the highest degree of agreement with the band gap required by the user, and storing the topological structure.
1. And (3) establishing a database.
As shown in fig. 2, the database contains three format files: 1) DXF format file (called DXF hereinafter) is an open vector data format, and can be directly imported into numerical simulation software to calculate characteristic frequency; 2) PNG format file (PNG) is used for lossless compression of image file format, can characterize phonon crystal topology structure, and can be imported into a training model of a self-encoder; 3) The CSV format file (hereinafter referred to as CSV) is a comma separated value format file for storing data, and stores characteristic frequency data obtained after numerical simulation. The database data sample is generated according to the basic configuration required by the user, and can be directly put into the framework of the on-demand design method for training the entity model.
Firstly, according to a basic configuration required by a user, batch acquisition of DXF data is performed. The boundary type two-dimensional solid state phonon crystal basic configuration is selected as an example, but the protection scope of the claims of the application is not limited. The geometric numerical parameter in the basic configuration is called a variable parameter, the phonon crystal unit cell is set to be a square with the length L and composed of two materials, the two materials are mutually divided by a curve function, and the curve coordinate function of the basic configuration is defined as follows:
；
wherein x and y are the abscissa and ordinate of the boundary curve respectively;is an independent variable of angle +.>As a function of the curve of the variable parameter.
Constraint conditions of the curve coordinate function are as follows:
；
wherein the method comprises the steps ofIs the maximum area constraint value; />，/>For the maximum boundary constraint value, the cell side length L is taken in this embodiment. />For the minimum point distance constraint value, this embodiment takes one thousandth of L. />For the distance threshold value of non-adjacent points, the present embodiment takes +.>. The first area constraint is to ensure that the proportion of two materials is kept within a certain range, the second side length constraint and the third side length constraint ensure that the boundary cannot be exceeded when the boundary is drawn, and the fourth side length constraint is to ensure that sharp shapes cannot be caused by too close points which are not adjacent.
And generating DXF data in batches according to the curve coordinate function of the basic configuration by using Python language. For a pair ofEquidistant programming is carried out on the parameter values of (2), namely +.>All parameters in the range are equidistantly taken out of N parameter values in the respective range, and each group of parameter combinations is traversed. Under the combination of the parameters of each group, < > and->Taking 0 to 2->1000 equidistant intervals between the two points are used for continuously acquiring coordinate values and connecting the coordinate values in sequence, and converting a continuous function curve into a discrete point broken line. And (3) introducing an Ezdxf module into the Python environment, drawing a cell boundary and a material boundary curve, obtaining DXF data, and storing the DXF data in a local place by using a serial number as a file name.
PNG data is then acquired in bulk. And drawing a material curve boundary by using a path.add_polyline_path () method in a Python language environment, importing MatplotlibBackend to directly fill an area of an acquired curve, setting unified pixels of a PNG picture as D x D, and storing acquired PNG data in a local place with the file name of the same serial number.
And finally, acquiring the characteristic frequencies of different PNG data in batches, and storing the characteristic frequencies as CSV data. And using the function logic of the Matlab bottom layer to control the modeling operation of the Comsol integration. The template model is fixed firstly, and comprises a phonon crystal structure, external input and grid division which are selected randomly, and the template model is calculated and stored as a.M file and a.MPH file. Matlab interfaces with Comsol through ports and then opens the M file. And modifying the model and performing function operation by utilizing the variable socket MPH file and utilizing the Matlab command. And (3) circularly introducing DXF data into the Comsol one by one, re-dividing grids, operating a model, and setting the maximum order of the obtained dispersion curve as the Q order. And storing the characteristic frequency of the calculation result in a CSV form locally. Wherein the material property definition part does not need to be re-reconstructed by Matlab language, since the topology is always two closed areas.
2. And (5) establishing an on-demand design method.
The on-demand design method is to use a convolution self-encoder as an image data preprocessing tool, and a band gap distribution prediction model of the phonon crystal as a forward environment to design the topological structure of the phonon crystal on demand. As shown in fig. 3, the on-demand design method includes a self-encoding module, a forward prediction module, and a reverse optimization module. The self-coding module can realize information extraction of a topological structure and decoding of the feature vector, and provides continuous numerical features for the forward prediction module. The forward prediction module receives the feature vector as input, carries out regression prediction on the band gap distribution characteristic of the feature vector, and provides a fitness evaluation criterion for the reverse optimization module. The reverse optimization module performs search optimization on the feature vector according to the band gap required by the user, decodes and restores the topological structure of the phonon crystal by the self-encoding module, and the whole process is a closed loop of data flow. The reference numbers in the figures refer to the interrelationship between the various components.
First, a self-coding module is established, corresponding to the numbers a, b, c and d in fig. 3. The self-encoding module consists of a convolutional self-encoder algorithm, which interconnects the encoder and decoder, i.e. the output of the encoder is directly converted into the input of the decoder. During model training, the two are cooperatively trained through the flow of training data. The function of the encoder is to extract information from PNG data and convert it into feature vectors so that structural information of phonon crystals can be expressed in vector form and participate in the prediction of band gap distribution. The decoder has the function of decoding the feature vector into the topological structure of the phonon crystal, and ensures that the final output of the on-demand design method is the optimal structure. Therefore, the input and output of the encoder are PNG data, and no training label exists.
The self-encoder model structure includes an input layer in the encoder section that accepts PNG data in the shape of dxdx 1. The middle hidden layers are alternately stacked with the convolution layers and the pooling layers, the number of layers is preferably 8-12, and the filter shape is set to be 2 multiplied by 2. Converting the two-dimensional PNG data into a three-dimensional matrix, and acquiring a characteristic vector in the last layer of the encoder, wherein the shape is MxMxN. The model structure of the decoder part still adds an input layer in the first layer to accept the feature vector, the rest of the intermediate layers are mirror images of the encoder intermediate layers, and the feature vector is gradually restored into PNG data by using a convolution layer and a pooling layer. The encoder and decoder are connected to form a self-encoder, the implicit layer activation functions of which are all set to Relu. Since PNG pictures contain two materials belonging to a binary pixel matrix, the activation function of the output layer of the self-encoder is set to Sigmoid, the final matrix is converted into a binary pixel matrix, and the binary cross entropy is set as the loss function of the self-encoder accordingly.
The PNG data needs to be subjected to two-dimensional and binarization processing when training the self-encoder. Because the generated picture data is a three-channel pixel matrix, the first channel is directly selected as the matrix representative of each picture. And dividing the data in the two-dimensional matrix into 0 and 1 by using a median value to finish binarization operation. PNG data were processed according to 8:2 is divided into training set data and test set data, which are used as input data to be imported into a self-encoder together, parameters are set, including iteration times, batch data sample numbers and random factors, the model is trained, and the training result of the model is saved. After the self-encoder training is completed, the network structure and network parameters of the encoder and the decoder are respectively stored in a local mode in the format of an H5 file.
Next, a forward prediction module is established, corresponding to sequence numbers e, f and g in FIG. 3. The forward prediction module is used for fitting the association between the topological structure of the phonon crystal and the band gap distribution by using a data driving algorithm, the data is derived from a real sample of finite element numerical simulation, and the self-encoder module is subjected to characteristic transformation. The model result is directly used as the fitness computing environment of the reverse optimization module, and selection conditions are provided for population evolution of the reverse optimization module. The module is used as forward distribution of band gap distribution, predicts band gap distribution of phonon crystal under various topological configurations, accords with normal design logic, and lays a foundation for obtaining band gap performance upper limit of fixed structure mode.
The forward prediction module comprises two band gap prediction models, namely a band gap distribution prediction model and a band gap main order prediction model. The structure of the band gap distribution prediction model is shown in fig. 4, in this embodiment, a Deep Neural Network (DNN) is selected as a basic framework for classification prediction, and the input features are the output of the encoder, i.e. the feature vector of the photonic crystal structure, and the type is floating point type. The output characteristics are band gap distribution, which is characterized by the numerical coding mode of pass band and stop band, and the specific expression criteria are as follows: the upper limit of the band gap range of the study is set, and 60kHz is taken in the embodiment. The 0Hz to 60kHz is divided into L sections (set according to the study accuracy, this embodiment is set to 1000), called γ sections, so the value of one γ section represents the 60Hz band gap case. If the gamma interval is swept by the dispersion curve, namely, no band gap exists, the numerical value is set to be 0; if the gamma interval is not swept by the dispersion curve, i.e. is in the bandgap range, the value is set to 1. The model structure includes an input layer, an hidden layer, and an output layer. The input layer receives the feature vector and transmits the information to the hidden layer, relu is selected as an activation function of the input layer, the number of neurons is set to be MxMxN, and the numerical value type is a floating point type containing zero value. The hidden layer is set to 4-8, the number of the hidden layer neurons is expanded from the number of the input layer neurons and then contracted to the number of the output layer neurons, and Relu is selected as an activation function of the input layer. The number of the neurons of the output layer is L, and the activation function is set to Sigmoid corresponding to the band gap distribution condition of the gamma interval. The loss function of the model is set as binary cross entropy to achieve the purpose of two classifications.
CSV data in a database are converted according to a gamma interval band gap representation criterion, the CSV data are changed into two-dimensional vectors representing band gap distribution, longitudinal labels are topological structure samples, and transverse labels are band gap numerical representations of gamma intervals, and the two-dimensional vectors are called gamma band gaps. After the PNG data is converted into a feature vector by utilizing an encoder H5 file, the feature vector is directly integrated with a gamma band gap by taking a topological configuration as a longitudinal label to form a training table, and the training table is respectively set as input and output of a band gap distribution prediction model. Setting training parameters including iteration times, batch data sample numbers and random factors, and training the band gap distribution prediction model. After training, the network structure and network parameters of the model are respectively stored in a local mode in the format of an H5 file.
The band gap principal order prediction model selects DNN as a framework of regression prediction, and the input characteristic of the model is the output of an encoder, namely the characteristic vector of a phonon crystal structure, and the type is a floating point type. The output characteristic is the order of the maximum band gap, and the value method of the order k is as follows: the band gap with the widest band gap range in the range of 0Hz to 60kHz is located between the kth dispersion curve and the k+1 dispersion curve. The number of neurons of the output layer of the model is set to be MxMxN, the hidden layer is set to be 2-4, the number of neurons of the hidden layer is directly contracted to the number of neurons of the output layer from the number of neurons of the input layer, and Relu is selected as an activation function of the input layer and the hidden layer. The number of the neurons of the output layer is 1, and the activation function is set as linear corresponding to the main step of the band gap. The loss function of the model is set to Mean Absolute Error (MAE) and the optimizer is set to Adam. CSV data in a database is converted into band gap main order data, PNG data is converted into feature vectors by utilizing an encoder H5 file, and then is directly integrated with the main order data into a training table by taking a topological configuration as a longitudinal label, and the training table is respectively set as input and output of a band gap main order prediction model. Setting training parameters including iteration times, batch data sample numbers and random factors, and training the band gap distribution prediction model. After training, the network structure and network parameters of the model are respectively stored in a local mode in the format of an H5 file.
Finally, the reverse optimization module corresponds to the sequence numbers h, i and j in fig. 3. The reverse optimization module is based on the forward environment of the forward prediction module, takes the band gap of the user as a guide, proposes a fitness slipping strategy to search and optimize the topological structure of the photonic crystal, and the optimization result of the reverse optimization module is a characteristic vector of the photonic crystal, and the characteristic vector is imported into a decoder in the self-encoder module to restore the optimal topological structure so as to achieve the aim of designing the photonic crystal according to the requirement.
The reverse optimization module selects a meta heuristic algorithm to optimize the topological structure of the phonon crystal. In this embodiment, genetic algorithm is selected as the optimization framework. The main procedure is set as follows:
population chromosome coding: the feature vector is taken as a chromosome, and thus the chromosome length is m×m×n bits. And taking the obtained forward training data as a basis, wherein the chromosome range is the range corresponding to each column of feature vectors, and expanding the scaling factor up and down to form the chromosome coding range. The partial feature vector is in an all-zero column, and the special chromosome position coding range is set to be in an all-zero list. The coding mode adopts random floating point number coding. The coding expansion strategy is shown in the following formula:
；
Wherein the method comprises the steps ofCoding region after expansion for chromosome i, < > for chromosome i>For the expansion factor, this example takes 5%; />，The original coding ranges of the ith chromosome are respectively; />Initializing a coding value for the ith chromosome; />Is a random valued function.
Fitness function definition: the fitness function is divided into three parts, wherein the first part is a main body part and is the corresponding degree of band gap distribution, namely, the part of the user demand band gap represented by the gamma interval is a stop band, which is compared with the individual band gaps, and each group of the pass bands can be correctly corresponding to the pass band, and 5 points are added. However, in order to prevent the initial population from being 0 points, the population cannot normally evolve, and thus the fitness of the second part is set. The second part is the bandwidth of the main bandgap and the closeness of the lower bandgap boundary. The ideal situation is taken as an upper limit, the upper limit score is L/100 score and L/100 score respectively, the difference between the two scores is subtracted, and the number of gamma intervals is taken as a calculation unit. The second part can be used for initializing non-ideal population, and can also slowly slide in the designed band gap direction until the first part contains overlapping fraction, and then the evolution is continued based on the fitness of the first part. The third part is the degree of discretization with the target master. The upper limit of the part is set as Q-2, and the difference value of the separated orders subtracted from the upper limit fraction is the score of the third part. And finally, weighting the three scores according to engineering application requirements to obtain a final fitness function. The specific process is as follows:
；/>
Wherein the method comprises the steps of，/>，/>，/>Fitness scores respectivelyA first fractional portion, a second fractional portion, and a third fractional portion; />，/>The user demand band gap of the ith gamma interval and the current individual band gap distribution situation are respectively 0 or 1; />The number of the gamma intervals; />，/>Respectively calculating positions of a target main band gap and a current individual main band gap according to a gamma interval sequence number; />，/>The widths of the target main band gap and the current main band gap are calculated according to the number of gamma intervals; />Is the maximum order of the dispersion curve; />，/>The target master orders and the master orders of the current individuals are respectively; />，/>，/>Respectively three-part fractional weighting coefficients。
Individual fitness calculation: and integrating the individuals into an array, loading the H5 file of the forward prediction module, directly transmitting the H5 file into the forward prediction module to calculate the band gap distribution and the primary order, and transmitting the band gap distribution and the primary order to the upper fitness function for evaluating the individual fitness.
Inheritance, crossover and mutation operations: the inheritance operation directly sorts the individuals of the previous generation population according to the fitness, and reserves excellent individuals according to the inheritance rate of the population, so as to reserve the excellent individuals in the next generation population; the crossing operation adopts two-point crossing, namely, randomly acquiring chromosome positions with the number of chromosomes as an upper limit, and directly exchanging chromosome fragments between two points of a male parent and a female parent; variation was by triple randomization. The first weight is that each individual generates a random number between 0 and 1, and if the random number is lower than the mutation rate, the mutation operation is carried out. The second factor is to randomly select mutation positions, and randomly select chromosome positions for mutation operation. The third is a value that randomly changes to be different from the original chromosome in the coding range corresponding to the position. In the third random, the chromosome gene after mutation is different from the original gene by circulation, and if the chromosome coding range selected in the step is a fixed zero value, the chromosome coding range falls into an infinite circulation trap. Thus, in the first re-operation, the selected position is incremented by one if the selected chromosome-encoding range is zero.
The embodiment of the application also provides a photonic crystal band gap design system based on intelligent coding, which comprises a processor and a memory;
a memory for storing a computer program;
and the processor is used for realizing the steps in the photonic crystal band gap design method based on intelligent coding when executing the program stored in the memory.
The photonic crystal band gap design system based on intelligent coding can realize all the embodiments of the photonic crystal band gap design method based on intelligent coding and can achieve the same beneficial effects, and details are not repeated here.
The foregoing describes in detail preferred embodiments of the present application. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the application by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.
Claims (10)
1. A photonic crystal band gap design method based on intelligent coding is characterized by comprising the following steps:
acquiring basic configuration data of phononic crystals based on user requirements, constructing a database based on the basic configuration data, and constructing and training a convolutional neural network model by utilizing data in the database;
Converting the image data in the basic configuration data through a trained convolutional neural network model to obtain a feature vector, and combining the feature vector with corresponding feature frequency data in a database to obtain training data;
constructing and training a deep neural network model by utilizing training data, storing and packaging the deep neural network model to form a topological structure evaluation environment, and obtaining band gap distribution data and a maximum band gap order based on the feature vector and the trained deep neural network model;
constructing an iteration model based on a meta heuristic algorithm, defining a fitness function based on the topological structure evaluation environment, setting iteration basic parameters, and inputting target band gap distribution data and a target maximum band gap order into the iteration model for iteration to obtain an optimized feature vector;
inputting the optimized feature vector into a topological structure evaluation environment to obtain a corresponding band gap, performing precision calculation on the band gap feature vector and the feature vector of the band gap required by a user to obtain first precision, and determining optimized image data based on the first precision;
generating corresponding vector data based on the optimized image data, performing finite element simulation on the vector data to obtain optimized band gap distribution data, performing precision calculation on the optimized band gap distribution data and a band gap required by a user to obtain second precision, and taking the band gap distribution data as a topological structure of a phonon crystal band gap based on the second precision.
2. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the constructing a database based on the basic configuration data comprises:
and constructing a basic configuration curve coordinate function based on the basic configuration data, generating image data and characteristic frequency data of the phonon crystal according to the basic configuration curve coordinate function, and constructing a database through the image data and the characteristic frequency data.
3. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the fitness function defining process comprises:
the fitness function is defined in three parts:
the first part, the part of the user requirement band gap represented by the gamma interval is compared with the individual band gap of the stop band, each group of the pass bands can be correctly corresponding to each other, and 5 minutes are added;
setting an ideal upper limit, wherein the upper limit score is L/100 min, subtracting the difference between the two, and taking the number of gamma intervals as a calculation unit, wherein L is the number of the gamma intervals;
a third section, setting the upper limit as Q-2 score, subtracting the separated order difference from the upper limit score to obtain the score of the third section, wherein, Is the maximum order of the dispersion curve.
4. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the iteration basic parameters include:
inheritance rate, crossover rate, mutation rate, population size and evolution number.
5. The smart code based photonic crystal bandgap design method of claim 1, wherein said determining optimized image data based on said first accuracy comprises:
and comparing the first precision with the preset precision, inputting the optimized feature vector into the trained neural network again for a new iteration when the first precision is smaller than the preset precision, and inputting the optimized feature vector into the trained convolutional neural network model for inverse decoding when the first precision is larger than or equal to the preset precision to obtain optimized image data.
6. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the topology structure using the optimized band gap distribution data as the photonic crystal band gap based on the second precision comprises:
and comparing the second precision with the preset precision, when the second precision is smaller than the preset precision, acquiring basic configuration data of the photonic crystal again based on the user requirement, and when the second precision is larger than or equal to the preset precision, taking the optimized band gap distribution data as the topological structure of the band gap of the photonic crystal.
7. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the convolutional neural network model comprises: an encoder and a decoder;
the encoder converts the image data into feature vectors by extracting information in the image data, so that the structural information of phonon crystals can be expressed in a vector form and participate in band gap distribution prediction;
the decoder decodes the feature vector into image data, and outputs the image data as a topology of phonon crystals.
8. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the deep neural network model comprises: a band gap distribution prediction model and a band gap main order prediction model;
the input characteristic of the band gap distribution prediction model is the characteristic vector of the phonon crystal structure, the output characteristic is the band gap distribution, and the band gap distribution is characterized in a mode of numerical coding of a passband and a stopband;
the input characteristic of the band gap main order prediction model is the characteristic vector of the phonon crystal structure, and the output characteristic is the order of the maximum band gap.
9. The photonic crystal band gap design method based on intelligent coding according to claim 1, wherein the feature vector of the user-required band gap comprises:
And carrying out band gap characterization on the user demand band gap to obtain a characteristic vector of the user demand band gap, wherein the gamma interval is a range interval obtained by setting the upper limit of a band gap range and setting the accuracy.
10. The photonic crystal band gap design system based on intelligent coding is characterized by comprising a processor and a memory;
a memory for storing a computer program;
a processor for implementing the method steps of any one of claims 1-9 when executing a program stored on a memory.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311069161.1A CN116776752B (en) | 2023-08-24 | 2023-08-24 | Phonon crystal band gap design method and system based on intelligent coding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311069161.1A CN116776752B (en) | 2023-08-24 | 2023-08-24 | Phonon crystal band gap design method and system based on intelligent coding |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116776752A true CN116776752A (en) | 2023-09-19 |
| CN116776752B CN116776752B (en) | 2023-11-24 |
Family
ID=87986369
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311069161.1A Active CN116776752B (en) | 2023-08-24 | 2023-08-24 | Phonon crystal band gap design method and system based on intelligent coding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116776752B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120330632A1 (en) * | 2011-06-27 | 2012-12-27 | The Research Foundation Of State University Of New York | Method for predicting optimized crystal structures |
| WO2013055503A1 (en) * | 2011-10-14 | 2013-04-18 | The Trustees Of Princeton University | Narrow-band frequency filters and splitters, photonic sensors, and cavities having pre-selected cavity modes |
| CN103246767A (en) * | 2013-04-26 | 2013-08-14 | 北京工业大学 | Liquid-liquid photonic crystal topological optimization method based on genetic algorithm and finite element method |
| US20160211828A1 (en) * | 2015-01-15 | 2016-07-21 | National Institute Of Standards And Technology | Reticulated resonator, process for making and use of same |
| CN115035965A (en) * | 2022-06-24 | 2022-09-09 | 湘潭大学 | Single-phase phonon crystal plate band gap optimization method based on geometric shape optimization |
| CN115602256A (en) * | 2022-10-14 | 2023-01-13 | 西安近代化学研究所（Cn） | Design method of layered furazan energetic material crystal |
| CN116244997A (en) * | 2023-03-13 | 2023-06-09 | 北京工业大学 | Optimization design method of light energy-absorbing vibration-damping multifunctional lattice structure with phonon band gap |
-
2023
- 2023-08-24 CN CN202311069161.1A patent/CN116776752B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120330632A1 (en) * | 2011-06-27 | 2012-12-27 | The Research Foundation Of State University Of New York | Method for predicting optimized crystal structures |
| WO2013055503A1 (en) * | 2011-10-14 | 2013-04-18 | The Trustees Of Princeton University | Narrow-band frequency filters and splitters, photonic sensors, and cavities having pre-selected cavity modes |
| CN103246767A (en) * | 2013-04-26 | 2013-08-14 | 北京工业大学 | Liquid-liquid photonic crystal topological optimization method based on genetic algorithm and finite element method |
| US20160211828A1 (en) * | 2015-01-15 | 2016-07-21 | National Institute Of Standards And Technology | Reticulated resonator, process for making and use of same |
| CN115035965A (en) * | 2022-06-24 | 2022-09-09 | 湘潭大学 | Single-phase phonon crystal plate band gap optimization method based on geometric shape optimization |
| CN115602256A (en) * | 2022-10-14 | 2023-01-13 | 西安近代化学研究所（Cn） | Design method of layered furazan energetic material crystal |
| CN116244997A (en) * | 2023-03-13 | 2023-06-09 | 北京工业大学 | Optimization design method of light energy-absorbing vibration-damping multifunctional lattice structure with phonon band gap |
Non-Patent Citations (3)
| Title |
|---|
| GENGWANG YAN ET AL: "Multi-objective optimization of elastic metaplates for lightweight and ultrawide bandgaps", ELSEVIER:INTERNATIONAL JOURNAL OF MECHANICAL SCIENCES, pages 1 - 22 * |
| XIANG LI ET AL: "Designing phononic crystal with anticipated band gap through a deep learning based data-driven method", ELSEVIER:SCIENCE DIRECT, pages 1 - 20 * |
| 刘么和 等: "基于智能算法的声子晶体带隙优化设计", 振动与冲击, vol. 27, no. 09, pages 94 - 95 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116776752B (en) | 2023-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cheng et al. | Rethinking Bayesian learning for data analysis: The art of prior and inference in sparsity-aware modeling | |
| CN110390638B (en) | High-resolution three-dimensional voxel model reconstruction method | |
| Ryberg et al. | Metamodel-based multidisciplinary design optimization for automotive applications | |
| Zhang et al. | Enhanced ELITE-load: A novel CMPSOATT methodology constructing short-term load forecasting model for industrial applications | |
| Chen et al. | Mo-padgan: Reparameterizing engineering designs for augmented multi-objective optimization | |
| CN114329232A (en) | User portrait construction method and system based on scientific research network | |
| CN113343427B (en) | Structural topology configuration prediction method based on convolutional neural network | |
| CN113112607A (en) | Method and device for generating three-dimensional grid model sequence with any frame rate | |
| Babuška et al. | An expert-system-like feedback approach in the hp-version of the finite element method | |
| Yin et al. | Batude: Budget-aware neural network compression based on tucker decomposition | |
| JP3881510B2 (en) | Method and program for identifying curved surface model | |
| Heaney et al. | Applying convolutional neural networks to data on unstructured meshes with space-filling curves | |
| CN114386582A (en) | Human body action prediction method based on confrontation training attention mechanism | |
| CN114329900A (en) | Device parameter acquisition method and system based on neural network and related components | |
| CN116776752B (en) | Phonon crystal band gap design method and system based on intelligent coding | |
| Balazs et al. | Hierarchical-interpolative fuzzy system construction by genetic and bacterial memetic programming approaches | |
| Yang et al. | Selection of optimal material and operating conditions in composite manufacturing. Part I: computational tool | |
| CN116542080A (en) | Condition generation countermeasure network topology optimization method and system based on contrast learning | |
| CN112837420B (en) | Shape complement method and system for terracotta soldiers and horses point cloud based on multi-scale and folding structure | |
| CN115908697A (en) | Generation model based on point cloud probability distribution learning and method thereof | |
| CN115546050A (en) | Intelligent restoration network and restoration method for ceramic cultural relics based on point cloud completion | |
| CN112488117B (en) | Point cloud analysis method based on direction-induced convolution | |
| CN115239967A (en) | Image generation method and device for generating countermeasure network based on Trans-CSN | |
| Son et al. | A framework for multivariate data based floor plan retrieval and generation | |
| Ahn et al. | Multi-branch neural architecture search for lightweight image super-resolution |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |