CN111680453B - Grating incidence parameter inversion model structure and establishing method - Google Patents
Grating incidence parameter inversion model structure and establishing method Download PDFInfo
- Publication number
- CN111680453B CN111680453B CN202010486478.5A CN202010486478A CN111680453B CN 111680453 B CN111680453 B CN 111680453B CN 202010486478 A CN202010486478 A CN 202010486478A CN 111680453 B CN111680453 B CN 111680453B
- Authority
- CN
- China
- Prior art keywords
- grating
- incidence
- parameter inversion
- coupling efficiency
- output
- 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.)
- Active
Links
Images
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
- G02—OPTICS
- G02B—OPTICAL ELEMENTS, SYSTEMS OR APPARATUS
- G02B27/00—Optical systems or apparatus not provided for by any of the groups G02B1/00 - G02B26/00, G02B30/00
- G02B27/0012—Optical design, e.g. procedures, algorithms, optimisation routines
-
- 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
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Optics & Photonics (AREA)
- Optical Communication System (AREA)
- Optical Integrated Circuits (AREA)
Abstract
The invention relates to a grating incidence parameter inversion model structure, which comprises the following steps: the input layer, the hidden layer and the output layer are sequentially connected, wherein the input parameter of the grating incidence parameter inversion model structure is a grating coupler parameter, and the output parameter is grating coupling efficiency; the input parameters of the input layer comprise structural parameters of the grating coupler, an incidence angle, an incidence wavelength and an incidence polarization state of an optical signal of the grating coupler; the output parameters of the output layer comprise Z forward grating coupling efficiency, Z backward grating coupling efficiency and grating total coupling efficiency. The grating incidence parameter inversion model structure can obtain a more accurate quantitative relation between grating incidence parameters and optical coupling efficiency, and can provide more accurate incidence angle information for alignment of optical signals and optical antenna receivers.
Description
Technical Field
The invention belongs to the technical field of grating couplers, and particularly relates to a grating incidence parameter inversion model structure and an establishment method.
Background
Along with the continuous progress of modern information technology, modern communication equipment not only improves communication quality and efficiency, and towards miniaturized, miniaturized target development, in military equipment such as satellite communication, naval vessel communication and radar stealth, put forward higher requirement to antenna receiving end's size. The antenna plays a vital role in the communication quality as a front-end component in the communication device. Currently, free space laser communication antennas for small satellites and small lasers typically employ telescopes to collect the optical wave energy.
With the development of technology, a new configuration, namely a chip type optical antenna, is proposed, a silicon-based grating coupler is used as a signal receiver of the optical antenna, the appearance of the silicon-based grating coupler accords with the specification of an integrated circuit chip, and the silicon-based grating coupler and a telescope are interchangeable in terms of functions. The chip antenna replaces a telescope, which means that an optical device and an electronic device are integrated at the chip level, and the biggest benefit is that the defect of discrete components is eliminated, the volume of the optical antenna is greatly reduced, and the performance of the whole machine is improved.
The grating coupler is very sensitive to the change of the incident angle of the optical signal, and the coupling efficiency is drastically reduced due to the small change near the optimal incident angle, so that if the grating coupler is used as a receiving end for an optical antenna, the alignment of the communication optical signal and the receiving end coupler (i.e. the coupling efficiency reaches a peak value after the communication light is incident) is particularly important. Then, providing a reference rotation angle for the alignment of the optical signal transmitter and the receiving end coupler, and establishing a grating incidence parameter inversion model for researching the grating coupling efficiency and the incidence parameter is particularly important.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a grating incidence parameter inversion model structure and an establishment method. The technical problems to be solved by the invention are realized by the following technical scheme:
the invention provides a grating incidence parameter inversion model structure, which comprises the following steps: the input layer, the hidden layer and the output layer are sequentially connected, wherein the input parameter of the grating incidence parameter inversion model structure is a grating coupler parameter, and the output parameter is grating coupling efficiency;
the input parameters of the input layer comprise structural parameters of the grating coupler, an incidence angle, an incidence wavelength and an incidence polarization state of an optical signal of the grating coupler; the output parameters of the output layer comprise Z forward grating coupling efficiency, Z backward grating coupling efficiency and grating total coupling efficiency.
In one embodiment of the present invention, the grating coupler parameters include a structural parameter of the grating coupler, an incident angle of the optical signal, an incident wavelength, and an incident polarization state, and the grating coupling efficiency includes a Z-forward grating coupling efficiency, a Z-reverse grating coupling efficiency, and a grating total coupling efficiency.
In one embodiment of the invention, the input layer comprises n neurons, the hidden layer comprises p neurons, each neuron being provided with an activation function f, the output layer comprises q neurons, wherein,
the activation function f is a function of the activation,
/>
the number of neurons of the hidden layer is variable.
In one embodiment of the invention, the input layer comprises 4 neurons, the hidden layer comprises 8 neurons, and the output layer comprises 3 neurons.
The invention also provides a method for establishing the grating incidence parameter inversion model structure, which is used for establishing any grating incidence parameter inversion model structure in the embodiment, and comprises the following steps:
s1: fixing grating structure parameters, setting a plurality of groups of optical signals with different polarization states, incident wavelengths and incident angles, and calculating to obtain a plurality of groups of corresponding grating coupling efficiency data by a finite time domain difference method to serve as training samples;
s2: establishing a grating incidence parameter inversion neural network model;
s3: initializing the weight of the grating incidence parameter inversion neural network model;
s4: setting the learning times M and error precision epsilon of the grating incidence parameter inversion neural network model;
s5: and training the grating incidence parameter inversion neural network model according to the training sample to obtain a grating incidence parameter inversion model.
In one embodiment of the present invention, in step S1, the structural parameters of the grating are: grating period t=650 nm, etch depth h=130 nm, duty cycle f=0.5, width d=15 μm; the polarization states of the optical signals are set to TE and TM, and the incident wavelengths are set to 1550nm and 633nm; the incident angle is set to 8-18 degrees.
In one embodiment of the present invention, the S5 includes:
s501: inputting a group of training samples, and calculating the input and output of each layer of the grating incidence parameter inversion neural network model;
s502: calculating the training error of the grating incidence parameter inversion neural network model;
s503: according to the training error, calculating an adjustment value of the weight and updating the weight;
s504: repeating the steps S501-S503 until the training sample is used up to finish learning once, and executing the step S505;
s505: calculating to obtain a global error E, and executing steps S501-S504 when the global error E is larger than the error precision epsilon or the learning times are smaller than the learning times M; and when the global error E is smaller than the error precision epsilon or the learning times are larger than or equal to the learning times M, obtaining the grating incidence parameter inversion model.
In one embodiment of the present invention, in S505, a global error E is calculated according to the following formula,
wherein m represents the number of sample groups in the training sample, d o (k) Representing the expected output, yo, corresponding to the kth training sample o (k) And (3) inverting the output of the neural network model by using the grating incidence parameters of the k-th training sample, wherein q represents the number of neurons of the output layer.
Compared with the prior art, the invention has the beneficial effects that:
1. the grating incidence parameter inversion model structure can obtain a more accurate quantitative relation between grating incidence parameters and optical coupling efficiency, and can provide more accurate incidence angle information for alignment of optical signals and optical antenna receivers.
2. The grating incidence parameter inversion model structure is simple in structure and has good learning capacity and prediction capacity.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention, as well as the preferred embodiments thereof, together with the following detailed description of the invention, given by way of illustration only, together with the accompanying drawings.
Drawings
FIG. 1 is a schematic diagram of a grating incidence parameter inversion model structure according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of coupling efficiency of a grating along a Z direction according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of correlation of a predicted result of a grating incidence parameter inversion model according to an embodiment of the present invention;
fig. 4 is a schematic diagram of correlation of a grating incidence parameter inversion model to a predicted result of a physical grating coupler according to an embodiment of the present invention.
Detailed Description
In order to further explain the technical means and effects adopted by the invention to achieve the preset aim, the invention provides a grating incidence parameter inversion model structure and a construction method according to the invention, which are described in detail below with reference to the accompanying drawings and the detailed description.
The foregoing and other features, aspects, and advantages of the present invention will become more apparent from the following detailed description of the preferred embodiments when taken in conjunction with the accompanying drawings. The technical means and effects adopted by the present invention to achieve the intended purpose can be more deeply and specifically understood through the description of the specific embodiments, however, the attached drawings are provided for reference and description only, and are not intended to limit the technical scheme of the present invention.
Example 1
Referring to fig. 1, fig. 1 is a schematic diagram of a grating incident parameter inversion model structure according to an embodiment of the present invention, as shown in the drawing, the grating incident parameter inversion model structure of the present embodiment includes: the input layer, the hidden layer and the output layer are sequentially connected, wherein the input parameter of the grating incidence parameter inversion model structure is a grating coupler parameter, and the output parameter is grating coupling efficiency; the input parameters of the input layer comprise structural parameters of the grating coupler, an incidence angle, an incidence wavelength and an incidence polarization state of an optical signal of the grating coupler; the output parameters of the output layer comprise Z forward grating coupling efficiency, Z backward grating coupling efficiency and grating total coupling efficiency.
Specifically, the grating coupler parameters include structural parameters of the grating coupler, an incident angle, an incident wavelength and an incident polarization state of an optical signal, and the grating coupling efficiency includes a Z forward grating coupling efficiency, a Z reverse grating coupling efficiency and a grating total coupling efficiency.
Further, the input layer comprises n neurons, the hidden layer comprises p neurons, each neuron is provided with an activation function f, the output layer comprises q neurons, wherein the activation function f is,
the number of neurons of the hidden layer is variable.
Specifically, in the present embodiment, the input layer includes 4 neurons, the hidden layer includes 8 neurons, and the output layer includes 3 neurons.
Further, the network parameter determination of the grating incidence parameter inversion model structure of the embodiment is specifically described.
In the structure of the BP neural network, network parameter determination mainly comprises determining the number of hidden layers, the number of neurons of each layer and an activation function of each layer. From the practical point of view of the angle sensitivity problem of the grating coupler, since derivative is required in the transformation process of the hidden layer, the activation function of the hidden layer is selected from an S-type activation function, that is,
the transformation of the output layer mainly involves the inverse feedback of the target error, so the activation function of the output layer only needs to be selected as a linear function g (u) =u.
In the structure of the BP neural network, according to the problem to be solved and with the practical situation, the number of neurons of each layer tends to be refined and simplified, and factors with larger main relations are covered. Therefore, interference of secondary factors can be eliminated, the number of times of calculation is reduced, and the rate of constructing a model is improved. The problem to be solved by the grating incidence parameter inversion model structure of the embodiment is to obtain the corresponding grating incidence parameters by using the model according to the grating coupling efficiency under different incidence conditions. The training times of the whole model are increased by an order of magnitude by one neuron, so that the number of input quantity and output quantity is not required to be subdivided too much for faster and more effective model establishment, and is not required to be more than 3 according to experience. Therefore, the training times can be greatly reduced, and the training accuracy is improved.
In case the grating coupler is fabricated, the structural parameters of the grating coupler are certain, and then, in this embodiment, the incident parameters include the incident angle, the incident wavelength and the incident polarization state of the optical signal. In this embodiment, the grating coupling efficiency is simplified into a diffraction model of xoz plane, please refer to fig. 2, fig. 2 is a schematic diagram of the grating coupling efficiency along the Z direction according to an embodiment of the present invention, and as shown in the figure, the output coupling efficiency includes the Z forward grating coupling efficiency, the Z backward grating coupling efficiency and the total grating coupling efficiency.
Further, the fitting ability of the entire neural network is mainly affected by the number of hidden layer neurons. If the number of neurons in the hidden layer is too small, the established inversion model cannot reach the expected fitting precision and cannot accurately reflect the mapping relation between the input quantity and the output quantity; if the number of neurons in the hidden layer is too many, although the established inversion model has better fitting precision, the calculation time is greatly prolonged, and the universality is not high after the training sample is jumped out, so that the method is not applicable to other situations.
In this embodiment, the number of neurons in the hidden layer is determined by a trial and error method, and the minimum number of neurons is selected within the accuracy range through multiple trial and error fitting.
Firstly, determining the possible number of neurons of an hidden layer to be 4-13 by using an empirical formula, and then selecting the minimum number of neurons by using a trial-and-error method, wherein the empirical formula is as follows:
wherein p represents the number of neurons in the hidden layer; n represents the number of neurons of the input layer; q represents the number of neurons in the output layer; alpha is a constant between 1 and 10.
In the BP neural network architecture, training errors and inspection errors are two important evaluation parameters. The training error of the model reflects the approximation capability of the model to the training sample, and represents the strength of the fitting capability of the model to the training sample, and the learning capability of the whole model; the test error of the model is the approximation capability of the test sample, and represents the fitting capability of the model to the test sample, and the applicability of the whole model is high. From the above understanding, if the training error of the model is small and the inspection error is large, the model has strong learning ability and small applicability; if the training error of the model is large, the learning ability of the model is poor, and the learning ability of the model needs to be improved by increasing the number of hidden layers or the number of neurons of each hidden layer.
By trial and error calculation, when the number of neurons of the hidden layer is 8, the order of magnitude of the checking error is reduced to 10 -5 A relatively stable state is achieved and the convergence rate of the model training error is the fastest, so the number of hidden layer neurons is selected to be 8.
The grating incidence parameter inversion model structure of the embodiment can obtain a more accurate quantitative relation between grating incidence parameters and optical coupling efficiency, and can provide more accurate incidence angle information for alignment of optical signals and optical antenna receivers. And the structure is simple, and the learning ability and the prediction ability are good.
Example two
The embodiment provides a method for establishing a grating incidence parameter inversion model structure, which is used for establishing the grating incidence parameter inversion model structure in the first embodiment. The method for establishing the grating incidence parameter inversion model structure of the embodiment comprises the following steps:
s1: fixing grating structure parameters, setting a plurality of groups of optical signals with different polarization states, incident wavelengths and incident angles, and calculating to obtain a plurality of groups of corresponding grating coupling efficiency data by a finite time domain difference method to serve as training samples;
in this embodiment, the structural parameters of the grating are set as follows: grating period t=650 nm, etch depth h=130 nm, duty cycle f=0.5, width d=15 μm. The polarization states of the optical signals are set to TE and TM, the incident wavelengths are set to 1550nm and 633nm, and the incident angles are set to 8 DEG-18 deg. According to the design research of the grating coupler, the optimal incidence angle of the grating coupler is 12 degrees, the incidence angle coupling window of the grating coupler is 5-20 degrees, if the incidence angle is out of the range, the forward coupling efficiency along Z is too low, the observation cannot be carried out in actual measurement, and the incidence angle of an optical signal is selected to be 8-18 degrees in order to simplify a model. Referring to table 1, table 1 is a training sample of the grating-incidence parametric inversion model of the present embodiment.
TABLE 1 training samples of grating incidence parameter inversion model
S2: establishing a grating incidence parameter inversion neural network model;
according to the grating incidence parameter inversion model structure described in the first embodiment, a grating incidence parameter inversion neural network model is built, that is, 4 neurons are set in an input layer, 8 neurons are set in an hidden layer, 3 neurons are set in an output layer, and an S-type activation function is selected as an activation function of each neuron in the hidden layer
The activation function of the output layer is a linear function g (u) =u.
S3: initializing the weight of the grating incidence parameter inversion neural network model;
in this embodiment, the weights of the grating incidence parameter inversion neural network model include the connection weights w of the input layer and the hidden layer ih And connection weight w of hidden layer and output layer ho . Specifically, the connection weight w of the input layer and the hidden layer in the grating incidence parameter inversion neural network model ih And connection weight w of hidden layer and output layer ho The random values in (-1, 1) are given to the respective values, and the initialization is completed.
S4: setting the learning times M and error precision epsilon of the grating incidence parameter inversion neural network model;
in this embodiment, the learning frequency M of the grating incidence parameter inversion neural network model is 10000 times, and the error precision epsilon is 10 -5 。
S5: and training the grating incidence parameter inversion neural network model according to the training sample to obtain a grating incidence parameter inversion model.
Specifically, the method comprises the following steps:
s501: inputting a group of training samples, and calculating the input and output of each layer of the grating incidence parameter inversion neural network model;
in this embodiment, the kth group of samples in the training samples is taken as the input x of the input layer i (k) The inputs transmitted to the hidden layer through the input layer are:
wherein x is i (k) Input k=1, 2, …, n representing the number of neurons of the input layer, b h Representing the threshold value of each neuron of the hidden layer;
the output of the hidden layer after passing through the hidden layer is as follows:
ho h (k)＝f(hi h (k))h＝1,2,… (4)，
wherein, the liquid crystal display device comprises a liquid crystal display device,
an activation function representing an hidden layer; />
The inputs transmitted to the output layer are:
wherein p represents the number of neurons in the hidden layer, b o A threshold value representing each neuron of the output layer;
the output of the output layer after passing through the output layer is as follows:
yo o (k)＝g(yi o (k))o＝1,2,… (6)，
where g (u) =u, represents the activation function of the output layer.
S502: calculating the training error of the grating incidence parameter inversion neural network model;
the training error includes an output layer training error delta o (k) Implicit layer training error delta h (k) Specifically, it comprises:
an error e of the actual output from the desired output is calculated,
wherein d o (k) Representing expected outputs corresponding to the kth group of samples, q representing the number of neurons of the output layer;
calculating an output layer training error delta according to the error e of the actual output and the expected output o (k) Implicit layer training error delta h (k)，
S503: according to the training error, calculating an adjustment value of the weight and updating the weight;
specifically, according to the output layer training error delta o (k) Implicit layer training error delta h (k) Calculating to obtain an adjustment value Deltaw of the connection weight of the hidden layer and the output layer ho (k) And an adjustment value Deltaw of the connection weight of the input layer and the hidden layer ih (k)，
Where μ represents the learning rate of the grating incidence parameter inversion neural network model.
Obtaining the adjustment value Deltaw of the connection weight of the hidden layer and the output layer according to the calculation ho (k) And an adjustment value Deltaw of the connection weight of the input layer and the hidden layer ih (k) Updating the weight of the grating incidence parameter inversion neural network model.
S504: repeating the steps S501-S503 until the training sample is used up to finish learning once, and executing the step S505;
s505: calculating to obtain a global error E, and executing steps S501-S504 when the global error E is larger than the error precision epsilon or the learning times are smaller than the learning times M; and when the global error E is smaller than the error precision epsilon or the learning times are larger than or equal to the learning times M, obtaining the grating incidence parameter inversion model.
In this embodiment, the global error E is calculated according to the following formula,
wherein m represents the number of sample groups in the training sample, d o (k) Represents the kth groupExpected output, yo, corresponding to training samples o (k) And (3) inverting the output of the neural network model by using the grating incidence parameters of the k-th training sample, wherein q represents the number of neurons of the output layer.
Example III
The performance of the grating incidence parameter inversion model in the first embodiment is checked. The learning ability of the BP neural network is verified by training samples, the output of which is the expected amount of the model in the most ideal case. The most ideal case is not possible nor necessary in reality, because in this case, the fitting degree is highest, the generalization capability is not available, the practical application value is not available, and the mapping of the non-training samples is inaccurate or even wrong. Therefore, the learning capacity and the generalization capacity of the BP neural network are relative, and the embodiment establishes a grating incidence parameter inversion model, so that the model has practical generalization capacity while ensuring higher learning capacity, and therefore, the generalization capacity of the model needs to be evaluated by using a test sample.
In this embodiment, 12 sets of test samples, which are obtained by calculation in a finite-field differential method, and the structural parameters of the grating coupler of the test samples are consistent with those of the training samples, are brought into the model to verify the generalization adaptability of the grating incident parameter inversion model according to the first embodiment. Referring to table 2, table 2 is a test sample of the grating-incidence parametric inversion model of the present embodiment.
TABLE 2 inspection sample of grating incidence parameter inversion model
In this embodiment, a linear regression method is used to evaluate the predictive power (generalization power) of the grating incidence parametric inversion model. And (3) taking the test sample into a model to obtain a corresponding output predicted value, and comparing and normalizing the actual value and the predicted value to obtain a linear regression fit line of the grating incidence parameter inversion model:
y＝rx+b (13)，
where y represents a predicted value, x represents an actual value, r represents a correlation coefficient, and b represents a constant.
The correlation coefficient r reflects the correlation between the actual value and the predicted value, namely the mapping degree of the grating incidence parameter inversion model to the test sample, the value range of r is [0,1], the value of r intuitively indicates the correlation degree of the actual value and the predicted value, and the larger the r is, the higher the correlation degree of the two values is.
When evaluating the generalization capability (prediction capability) of the grating incidence parameter inversion model by using a linear regression method, if the correlation coefficient r is greater than 0.9, the correlation between the predicted value and the actual value of the grating incidence parameter inversion model is good, and the prediction capability of the model is good; if the correlation coefficient r is smaller than 0.9, the prediction capability of the grating incidence parameter inversion model is poor, the model structure, the hidden layer number and the neuron number of each layer are required to be adjusted, meanwhile, the accuracy of a training sample is required to be considered, and the mapping relation of input and output can be accurately reflected.
Referring to fig. 3, fig. 3 is a schematic diagram of correlation of a predicted result of a grating incident parameter inversion model provided by the embodiment of the present invention, as shown in the drawing, it can be seen from the figure that a linear regression fit line of correlation of test samples is y=0.972187x+0.0157, and a correlation coefficient r= 0.972187 indicates that the correlation between predicted values and actual output values of 12 groups of test samples is very high, and the established grating incident parameter inversion model has good prediction capability.
Further, the prediction capability of the grating incidence parameter inversion model is verified in experiments by using a grating coupler on a physical sheet. Considering the inversion range of the incidence angle of the grating incidence parameter inversion model and practical experimental conditions, selecting incident light with TM polarization 1550nm, respectively coupling the incident light into a grating coupler at incidence angles with actual values of 10 DEG, 15 DEG and 18 DEG, and inputting the measured grating coupling efficiency into the grating incidence parameter inversion model established in the first embodiment to obtain the predicted value of the model to the incidence angle. The correlation of the actual value and the predicted value is analyzed by a linear regression method.
Referring to fig. 4, fig. 4 is a schematic diagram showing the correlation of the grating incident parameter inversion model to the predicted result of the physical grating coupler according to the embodiment of the present invention, and as shown in the figure, the linear regression fit line of the correlation of the predicted result of the physical grating coupler is y= 0.909581x-0.0159, and the correlation coefficient 0.972187 > r= 0.909581 > 0.9 indicates that the correlation of the actual value of the incident angle of the physical grating coupler and the predicted value of the grating incident parameter inversion model is higher, but the prediction capability of the grating incident parameter inversion model to the incident angle of the physical grating coupler is weaker than that to the test sample. The obtained predicted value of the incidence angle is 8.89% different from the actual value in the range of the effective incidence angle of the grating coupler, and the error is in the specified range. Therefore, the grating incidence parameter inversion model established in the first embodiment also has good prediction capability on the coupling efficiency of the physical grating coupler.
The foregoing is a further detailed description of the invention in connection with the preferred embodiments, and it is not intended that the invention be limited to the specific embodiments described. It will be apparent to those skilled in the art that several simple deductions or substitutions may be made without departing from the spirit of the invention, and these should be considered to be within the scope of the invention.
Claims (6)
1. The method for establishing the grating incidence parameter inversion model structure is characterized by being used for constructing the grating incidence parameter inversion model structure, and the grating incidence parameter inversion model structure comprises the following steps: the input layer, the hidden layer and the output layer are sequentially connected, wherein the input parameter of the grating incidence parameter inversion model structure is a grating coupler parameter, and the output parameter is grating coupling efficiency; the input parameters of the input layer comprise structural parameters of the grating coupler, an incidence angle, an incidence wavelength and an incidence polarization state of an optical signal of the grating coupler; the output parameters of the output layer comprise Z forward grating coupling efficiency, Z backward grating coupling efficiency and grating total coupling efficiency;
the method for establishing the grating incidence parameter inversion model structure comprises the following steps:
s1: fixed grating structure parameterSetting a plurality of groups of optical signals with different polarization states, incident wavelengths and incident angles, and calculating to obtain a plurality of groups of corresponding grating coupling efficiency data by a finite time domain difference method to serve as training samples; wherein, the structural parameters of the grating are as follows: grating period t=650 nm, etch depth h=130 nm, duty cycle f=0.5, width d =15 μm; the polarization states of the optical signals are set to TE and TM, and the incident wavelengths are set to 1550nm and 633nm; the incident angle is set to 8-18 degrees;
s2: establishing a grating incidence parameter inversion neural network model;
s3: initializing the weight of the grating incidence parameter inversion neural network model;
s4: setting the learning times M and error precision epsilon of the grating incidence parameter inversion neural network model;
s5: and training the grating incidence parameter inversion neural network model according to the training sample to obtain a grating incidence parameter inversion model.
2. The method for building a grating-incidence-parameter inversion model structure according to claim 1, wherein the grating-coupler parameters include structural parameters of a grating coupler, an incidence angle of an optical signal, an incidence wavelength, and an incidence polarization state, and the grating coupling efficiency includes a Z-forward grating coupling efficiency, a Z-backward grating coupling efficiency, and a grating total coupling efficiency.
3. The method of building a grating-incidence parametric inversion model structure according to claim 1, wherein the input layer comprises n neurons, the hidden layer comprises p neurons, each neuron is provided with an activation function f, the output layer comprises q neurons, wherein,
the activation function f is a function of the activation,
the number of neurons of the hidden layer is variable.
4. A method of building a grating-incident parametric inversion model structure according to claim 3, wherein the input layer comprises 4 neurons, the hidden layer comprises 8 neurons, and the output layer comprises 3 neurons.
5. The method for building a grating-incidence-parameter inversion model structure according to claim 1, wherein S5 comprises:
s501: inputting a group of training samples, and calculating the input and output of each layer of the grating incidence parameter inversion neural network model;
s502: calculating the training error of the grating incidence parameter inversion neural network model;
s503: according to the training error, calculating an adjustment value of the weight and updating the weight;
s504: repeating the steps S501-S503 until the training sample is used up to finish learning once, and executing the step S505;
s505: calculating to obtain a global error E, and executing steps S501-S504 when the global error E is larger than the error precision epsilon or the learning times are smaller than the learning times M; and when the global error E is smaller than the error precision epsilon or the learning times are larger than or equal to the learning times M, obtaining the grating incidence parameter inversion model.
6. The method for building a grating-incident parametric inversion model structure according to claim 5, wherein in S505, the global error E is calculated according to the following formula,
wherein m represents the number of sample groups in the training sample, d o (k) Representing the expected output, yo, corresponding to the kth training sample o (k) Grating incidence parameter inversion neural network model output representing the kth training sample, q representing the number of neurons of the output layer。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010486478.5A CN111680453B (en) | 2020-06-01 | 2020-06-01 | Grating incidence parameter inversion model structure and establishing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010486478.5A CN111680453B (en) | 2020-06-01 | 2020-06-01 | Grating incidence parameter inversion model structure and establishing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111680453A CN111680453A (en) | 2020-09-18 |
| CN111680453B true CN111680453B (en) | 2023-05-30 |
Family
ID=72434605
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010486478.5A Active CN111680453B (en) | 2020-06-01 | 2020-06-01 | Grating incidence parameter inversion model structure and establishing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111680453B (en) |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104008427A (en) * | 2014-05-16 | 2014-08-27 | 华南理工大学 | Central air conditioner cooling load prediction method based on BP neural network |
| CN104376361A (en) * | 2014-10-15 | 2015-02-25 | 南京航空航天大学 | Nuclear accident source item inversion method based on BP neural network algorithm |
| CN110703386B (en) * | 2019-09-30 | 2020-07-28 | 西安交通大学 | Bragg concave diffraction grating type polarization-wavelength hybrid multiplexer |
-
2020
- 2020-06-01 CN CN202010486478.5A patent/CN111680453B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111680453A (en) | 2020-09-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111695295A (en) | Method for constructing incident parameter inversion model of grating coupler | |
| CN111695294A (en) | Construction method of grating incidence parameter inversion model based on BP neural network | |
| CN107390199B (en) | A kind of radar maneuvering target tracking waveform design method | |
| CN107807366B (en) | Atmospheric visibility calculation method, device, radar and system | |
| CN111578966B (en) | Hemisphere harmonic oscillator characteristic parameter identification method based on LMS algorithm | |
| CN105259548A (en) | Dispersion mismatch correction method used in FMCW absolute distance measurement technology | |
| Arockia Bazil Raj et al. | A direct and neural controller performance study with beam wandering mitigation control in free space optical link | |
| CN110672932B (en) | Automatic calibration method for multi-antenna navigation darkroom test signal level | |
| CN108563895A (en) | A kind of interval model modification method considering correlation | |
| CN101285862A (en) | All-digital compensation microwave radiometer | |
| CN111680453B (en) | Grating incidence parameter inversion model structure and establishing method | |
| CN103257340B (en) | Method for calibrating amplitude consistency of a plurality of ground receivers with radar satellite | |
| CN114624742A (en) | Amplitude-phase error calibration positioning method and device for polarization sensitive array | |
| CN109341690B (en) | Robust and efficient combined navigation self-adaptive data fusion method | |
| Zhao et al. | L1-norm constraint kernel adaptive filtering framework for precise and robust indoor localization under the internet of things | |
| CN113630197B (en) | Antenna phase adjustment method and device, storage medium and electronic equipment | |
| CN105352531B (en) | The detection method of the performance parameter of laser range finder | |
| Zhao et al. | Fusing vehicle trajectories and gnss measurements to improve gnss positioning correction based on actor-critic learning | |
| CN107483131B (en) | Method for generating high-speed aircraft double-satellite combined channel Markov state sequence | |
| CN109506683A (en) | A kind of FBG Fibre Optical Sensor demodulating system towards marine environmental monitoring | |
| CN114205009B (en) | Automatic phase matching method based on FPGA digital phased array antenna receiving channel | |
| CN110850366B (en) | Positioning method based on received signal strength under mixed line-of-sight and non-line-of-sight environment | |
| Elmer et al. | Beamformer design for radio astronomical phased array feeds | |
| Wu et al. | Improved adaptive iterated extended Kalman filter for GNSS/INS/UWB-integrated fixed-point positioning | |
| EP3580607A1 (en) | Methods and apparatus for optical polarization control |
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 |