Design method and system of grating coupler, computer equipment and storage medium

Abstract

The invention relates to the technical field of grating coupler design, and discloses a grating coupler design method and system, computer equipment and a storage medium, and the method comprises the steps: constructing a topology prediction model based on a deep neural network, and training the topology prediction model through employing a preset method; inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain structural parameters of the functional area of the grating coupler; and determining a topological structure of the functional region of the grating coupler according to the structural parameters so as to realize the design of the grating coupler. Through approximate processing of the training data set, a simple network structure and a complex nonlinear algorithm are combined, the computational power demand of the model is reduced, the prediction efficiency and effect of the model are improved, meanwhile, through a forward and reverse combined training method, the prediction precision of the model is further improved, and the prediction efficiency of the model is improved. Therefore, the design of the grating coupler with high coupling efficiency at any angle is realized, and the device performance of the grating coupler is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of grating coupler design, and in particular to a design method, system, computer equipment and storage medium of a grating coupler. Background Art

[0002] With the continuous development of information technology, applications such as big data, the Internet of Things, and cloud computing have led modern civilization into the era of massive data. Faced with the explosive growth of data volume, traditional electrical signal processing methods are gradually becoming inapplicable, and photonic integrated chips with all-optical signal processing as a typical application have received a lot of attention.

[0003] For silicon-based photonic integrated chips, a significant limiting factor is that silicon, as an indirect bandgap material, is difficult to emit light by itself, so it is necessary to introduce an off-chip light source or an on-chip hybrid integrated gain material. As a passive device that can couple the optical chip with an external light source (optical fiber), the grating coupler is flexible and can be located anywhere in the chip, so it can be used for wafer-level online testing. In addition, it is known for its large coupling alignment tolerance and easy packaging. However, the low coupling efficiency and narrow bandwidth of the grating coupler have greatly hindered its application in practice.

[0004] In order to improve the performance indicators of the device, the structural parameters of the device can be optimized. Currently, the commonly used structural optimization algorithms include genetic algorithms, particle swarm algorithms, topology optimization algorithms, etc., but these algorithms have certain limitations. On the one hand, as a global optimization algorithm, genetic algorithms are usually suitable for optimization scenarios with fewer parameters, and their optimization results improve as the parameter space becomes larger. However, the cost is the rapid increase in the amount of calculation and computing resource requirements. Therefore, this algorithm is not suitable for grating couplers that require multi-parameter parallel optimization. On the other hand, although the particle swarm algorithm can be applied to the design of silicon optical devices, its convergence speed is too slow and is not suitable for complex grating coupler optimization scenarios. In addition, since the topology optimization algorithm requires a strict mathematical expression between the optimization target and the structural parameters, and combines the steepest descent algorithm to execute the optimization process, its process is too complicated and the iterative convergence speed is slow, so it is not suitable for complex grating coupler optimization scenarios. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a design method, system, computer device and storage medium for a grating coupler, so as to solve the problems of large computing resource requirements, long iteration cycle, slow convergence speed and so on of the existing design methods, so as to achieve the effect of quickly and efficiently realizing the design of the grating coupler.

[0006] In a first aspect, the present invention provides a design method for a grating coupler, the method comprising:

[0007] A topology prediction model is constructed based on a deep neural network, and the topology prediction model is trained using a preset method. The topology prediction model includes an input layer, an output layer, and a plurality of hidden layers. The activation function of the topology prediction model is a Swish function. The training data set is constructed using approximate processing to adapt to the convergence of the Swish function.

[0008] Inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler;

[0009] According to the structural parameters, the topological structure of the functional area of ​​the grating coupler is determined to realize the design of the grating coupler.

[0010] Furthermore, the step of training the topology prediction model using a preset method includes:

[0011] Generate training data through simulation method, and perform approximate processing on the training data to obtain a training data set;

[0012] The topology prediction model is trained using the training data set.

[0013] Furthermore, the step of generating training data by a simulation method and performing approximate processing on the training data to obtain a training data set includes:

[0014] Generate training data with SOI structural consistency through finite-difference time-domain simulation, wherein the training data includes structural training data and spectral training data;

[0015] Dividing the functional area of ​​the grating coupler corresponding to the structural training data into a plurality of pixel areas according to a preset size;

[0016] Performing data conversion on each pixel region according to a region state to obtain the processed structure training data, wherein the region state includes an etched state and a non-etched state;

[0017] A training data set is formed according to the processed structural training data and the spectral training data.

[0018] Furthermore, the step of using the training data set to train the topology prediction model includes:

[0019] Constructing a forward training model with the same structure as the topology prediction model, wherein the input data of the forward training model are the structural parameters of the functional area of ​​the grating coupler, and the output data of the forward training model are the response spectrum of the grating coupler;

[0020] Using the training data set to train the forward training model to obtain a first network parameter configuration of the forward training model;

[0021] Connecting the trained forward training model and the untrained topology prediction model in series to obtain a series network model, and training the series network model using the training data set;

[0022] The second network parameter configuration of the topology prediction model is determined from the trained series network model to obtain the trained topology prediction model.

[0023] In a second aspect, the present invention provides a design system for a grating coupler, the system comprising:

[0024] A model construction module, used to construct a topology prediction model based on a deep neural network, and train the topology prediction model using a preset method, wherein the topology prediction model includes an input layer, an output layer, and a plurality of hidden layers, and the activation function of the topology prediction model is a Swish function, and the training data set is constructed using approximate processing to adapt to the convergence of the Swish function;

[0025] A parameter prediction module, used for inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler;

[0026] The grating design module is used to determine the topological structure of the functional area of ​​the grating coupler according to the structural parameters to realize the design of the grating coupler.

[0027] Further, the model building module includes a model training module;

[0028] The model training module is used to generate training data through a simulation method and perform approximate processing on the training data to obtain a training data set;

[0029] The topology prediction model is trained using the training data set.

[0030] Furthermore, the model training module also includes a data set construction module;

[0031] The data set construction module is used to generate training data with SOI structural consistency through finite difference time domain simulation, wherein the training data includes structural parameters and spectral parameters;

[0032] Dividing the functional area of ​​the grating coupler corresponding to the structural parameters into a plurality of pixel areas according to preset sizes;

[0033] Performing data conversion on each pixel region according to a region state to obtain the processed structural parameters, wherein the region state includes an etched state and a non-etched state;

[0034] A training data set is formed according to the processed structural parameters and spectral parameters.

[0035] Furthermore, the model training module is also used to construct a forward training model with the same structure as the topology prediction model, the input data of the forward training model is the structural parameters of the functional area of ​​the grating coupler, and the output data of the forward training model is the response spectrum of the grating coupler;

[0036] Using the training data set to train the forward training model to obtain a first network parameter configuration of the forward training model;

[0037] Connecting the trained forward training model and the untrained topology prediction model in series to obtain a series network model, and training the series network model using the training data set;

[0038] The second network parameter configuration of the topology prediction model is determined from the trained series network model to obtain the trained topology prediction model.

[0039] In a third aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0040] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0041] The present invention provides a design method, system, computer device and storage medium for a grating coupler. The present invention combines a simple network structure with a complex nonlinear algorithm through approximate processing of a training data set, so that the Swish function can converge under a simple network architecture, reduce the computing power requirements of the model, and improve the prediction efficiency and effect of the model. At the same time, through a forward and reverse combined training method, the prediction accuracy of the model can be improved, thereby realizing the design of a grating coupler with high coupling efficiency at any angle and ensuring the device performance of the grating coupler. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a schematic flow chart of a design method of a grating coupler in an embodiment of the present invention;

[0043] Figure 2is a schematic structural diagram of a grating coupler in an embodiment of the present invention;

[0044] Figure 3 is a schematic diagram of the structure of a topology prediction model in an embodiment of the present invention;

[0045] Figure 4 is a schematic diagram of the structure of a forward training model in an embodiment of the present invention;

[0046] Figure 5 is a schematic diagram of the structure of a series network model in an embodiment of the present invention;

[0047] Figure 6 is a schematic structural diagram of a design system for a grating coupler in an embodiment of the present invention;

[0048] Figure 7 It is a diagram of the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] See also Figure 1 , a design method of a grating coupler proposed in the first embodiment of the present invention, wherein the method comprises steps S10 to S30:

[0051] Step S10, constructing a topology prediction model based on a deep neural network, and training the topology prediction model using a preset method, wherein the topology prediction model includes an input layer, an output layer, and a plurality of hidden layers, and the activation function of the topology prediction model is a Swish function, and the training data set is constructed using approximate processing to adapt to the convergence of the Swish function;

[0052] Step S20, inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler;

[0053] Step S30, determining the topological structure of the functional area of ​​the grating coupler according to the structural parameters to realize the design of the grating coupler.

[0054] In this embodiment, a topology prediction model based on a deep neural network is used to predict the structural parameters of the functional area of ​​the grating coupler. The structure of the grating coupler is as follows: Figure 2As shown, the structure of the high coupling efficiency non-periodic grating coupler includes a silicon substrate, a silicon dioxide buried oxide layer, a silicon structure layer and a silicon dioxide cladding layer, and the topological structure of the functional area responsible for coupling in the silicon structure layer is the part to be designed by the present invention. The topological structure is obtained by etching the silicon structure layer. Therefore, through the prediction model provided by the present invention, what is obtained is a binary code that characterizes whether each structural parameter (that is, the pixel area) in the topological structure is etched.

[0055] In this embodiment, the topology prediction model is as follows: Figure 3 As shown, it includes an input layer, an output layer and several hidden layers. In order to distinguish it from the architecture of the forward training model in the subsequent training process, it is represented as the first input layer and the first output layer here. In order to reduce the demand for computing resources, shorten the iteration cycle of the model, improve the convergence speed of the model, and ensure the reliability and stability of the calculation, this embodiment sets the hidden layer to at least three layers and less than ten layers. Preferably, the hidden layer is set to four layers. The input data of the model is the response spectrum expected by the device to be designed, and the output is the structural parameters of the functional area of ​​the device to be designed. The input layer is used to respond to the input light. The spectrum is discretized and passed to the hidden layer; the hidden layer is used to extract spectral features from the response spectrum, perform linear or nonlinear processing through the activation function, obtain the state probability value of the structure parameter, and pass it to the output layer, that is, the task of the hidden layer is to process the input data passed from the input layer, obtain the pixel point probability of the pixel topological pattern and pass it to the output layer of the back end; the output layer is used to judge the input structure parameters and output binary-coded structure parameters. For example, when the obtained pixel point probability value is greater than 0.5, the corresponding structure parameter (pixel area) unit is encoded as 1, otherwise it is encoded as 0.

[0056] The topology prediction model in this embodiment adopts a relatively small-scale deep learning neural network architecture. ι The input vector x (l) Perform a linear transformation:

[0057] z (l) =W (l) x (l) +b (l)

[0058] Among them, W (l) is the weight vector of the lth layer; b (l) is the bias vector of the lth layer; z (l) is the linear output of the input vector of the lth layer.

[0059] In order to introduce nonlinear relationships and enable the network to approximate complex functions, a nonlinear activation function is used in each layer. The output after activation is expressed as:

[0060] a (l) =F(b (l) )

[0061] Among them, F represents the activation function, a (l) is the output after the activation function.

[0062] From the input layer to the output layer, linear transformation and activation function are performed layer by layer to finally obtain the output of the network:

[0063]

[0064] The goal of deep learning is to minimize the loss function through training. The loss function used in this embodiment is:

[0065]

[0066] in, and i They represent the i-th predicted value and actual value respectively, and n represents the number of predicted values.

[0067] The back-propagation algorithm is used to calculate the gradient of the loss function with respect to the weights and biases through chain derivation.

[0068] Assuming the loss function is L, the gradient of the weights and biases of each layer is:

[0069]

[0070] Update weights and biases using gradient descent:

[0071]

[0072] Where η is the learning efficiency.

[0073] The loss is calculated through forward propagation, and the weights and biases are updated through back propagation, which is repeated many times (called epochs) until the loss function converges. A deep learning neural network can be represented as a combination of linear transformations and nonlinear activations layer by layer. The output is obtained through forward propagation, and the weights are continuously adjusted through back propagation and gradient descent, ultimately achieving the desired mapping relationship on a given data set.

[0074] In the above model construction process, the activation function is an important part. Since the design of the grating coupler belongs to a multivariate optimization problem, in order to obtain more accurate optimization results, it is necessary to adopt a more advanced activation function suitable for complex reasoning tasks. For this reason, in this embodiment, the activation function of the topology prediction model adopts the Swish function:

[0075]

[0076] Here, x is the input.

[0077] The Swish function is a smooth, non-monotonic activation function that is suitable for improving model performance and gradient flow in deep neural networks. One of the biggest features of this function is its non-monotonicity, that is, as x increases in certain intervals, the function value may decrease. This feature helps the neural network model learn complex features. The Swish function uses the self-gating mechanism of the input to make it outperform traditional activation functions in many tasks, especially in deep models. The following table 1 compares the Swish function with other common functions:

[0078] Table 1 Comparison of Swish function and other common functions

[0079]

[0080] Through the above comparison, it can be clearly seen that the Swish activation function has relatively excellent performance and is suitable for complex reasoning tasks, but its cost is high computational complexity and the need for large-scale deep neural networks. The following Table 2 shows the difference in computational complexity between the Swish function and the commonly used ReLU function:

[0081] Table 2 Computational complexity of Swish function and ReLU function

[0082] characteristic ReLU Swish Mathematical Expression max(0,x) x·σ(x) Computational complexity O(1) O(1) Actual operation steps 1 comparison 1 exponentiation, 1 division, 1 multiplication, 1 addition Actual time (per neuron) 1ns 18ns

[0083] From Table 2 above, we can intuitively see the huge computing power requirements of the Swish function. Therefore, the Swish function is usually used in network models with complex architectures. The number of model layers is required to be dozens or even hundreds of layers in order for the model to achieve a good convergence effect. For the topology prediction model with only a few layers in this embodiment, when the Swish function is used as the activation function, the model is actually unable to converge during the training process. In other words, the Swish function cannot be directly used as the activation function of the topology prediction model. In order to utilize the excellent reasoning performance of the Swish function, the present invention solves the mismatch between the simple model structure in the topology prediction model and the activation function with high computing power requirements by preprocessing the training data set and using the processed data set to train the model. The specific steps include:

[0084] Generate training data through simulation method, and perform approximate processing on the training data to obtain a training data set;

[0085] The topology prediction model is trained using the training data set.

[0086] In this embodiment, an approximate processing method is used for processing the data set, and a relatively coarse precision is used to reduce the amount of calculation required by the algorithm. The steps of the approximate processing include:

[0087] Generate training data with SOI structural consistency through finite-difference time-domain simulation, wherein the training data includes structural training data and spectral training data;

[0088] Dividing the functional area of ​​the grating coupler corresponding to the structural training data into a plurality of pixel areas according to a preset size;

[0089] Performing data conversion on each pixel region according to a region state to obtain the processed structure training data, wherein the region state includes an etched state and a non-etched state;

[0090] A training data set is formed according to the processed structural training data and the spectral training data.

[0091] In this embodiment, a training data set is first generated through time-domain finite difference simulation. At this time, it is necessary to pay attention to the consistency of the SOI structure used in the data set to ensure the simplicity of the subsequent algorithm. The data set is a structure-performance data set of photonic devices, that is, structural training data and corresponding spectral training data.

[0092] Then, the structural training data is approximated so that the neural network can converge faster and obtain more stable results. The structural training data is actually the structural parameters of the functional area of ​​the grating coupler. In this embodiment, the functional area of ​​the grating coupler is divided into several pixel areas according to the preset size. For each pixel area, there are only two states: etching / not etching, and it is converted into a digital 0 or 1 according to the state. For example, with 50nm as the preset size, the 5um grating coupler functional area is divided into 300 pixel areas, thereby obtaining a binary code of 0-1. It can be seen that the pixel area division size in this embodiment is relatively large, that is, a lower precision is used in the data processing process. Generally speaking, a lower data precision will affect the accuracy of model training and prediction. The reason why a lower precision is used to process the data in this embodiment is that although the photonic device is theoretically very sensitive to structural parameters such as period and duty cycle, the difference in such fine structure cannot be accurately reflected in the actual manufacturing process, and the calculation of such fine structure often consumes a lot of computing resources. For example, whether the period of a grating coupler is 500nm or 505nm, in theory, its performance parameters will be significantly different, but in actual manufacturing, due to the existence of manufacturing errors, the period of the prepared grating coupler may be a value between 495 and 505nm. Therefore, unlike the conventional neural network prediction model, in this embodiment, combined with the characteristics of the photonic device itself, a relatively coarse precision is used to reduce the amount of calculation required by the algorithm, making it possible to use a relatively simple neural network architecture to carry a complex algorithm, and this approximate processing does not reduce the accuracy of the final prediction result.

[0093] In this embodiment, by using the processed data set, under the premise that the thickness of the silicon / silicon dioxide layer in the specified SOI structure remains unchanged, by constructing a relatively simple network structure and running a complex nonlinear algorithm, the overall amount of computation is kept within an acceptable range. The reduction in the amount of computation means a reduction in the demand for computing resources / running time, thereby achieving rapid convergence of the deep learning network and reducing the time and labor costs required for device design. That is, this embodiment realizes the combination of a simple neural network architecture and a complex algorithm by approximating the constructed data set, thereby reducing the computing power requirements of the model and improving the prediction efficiency and effect of the model.

[0094] Based on the above training data set, the model is trained to obtain a trained topology prediction model. The input of the model is the expected response spectrum of the grating coupler to be designed, and the output is the structural parameters corresponding to the response spectrum. However, in fact, when a conventional model training method is used, the trained model obtained may have multiple corresponding output data for one input data, that is, one response spectrum corresponds to several structural parameters. This situation is not conducive to the refined design of the grating coupler. In order to achieve a one-to-one correspondence between input data and output data, in this embodiment, a training method based on an inverse network model is preferably used to make the output structural parameters of the model the optimal result of the expected response spectrum. The specific steps include:

[0095] Constructing a forward training model with the same structure as the topology prediction model, wherein the input data of the forward training model are the structural parameters of the functional area of ​​the grating coupler, and the output data of the forward training model are the response spectrum of the grating coupler;

[0096] Using the training data set to train the forward training model to obtain a first network parameter configuration of the forward training model;

[0097] Connecting the trained forward training model and the untrained topology prediction model in series to obtain a series network model, and training the series network model using the training data set;

[0098] The second network parameter configuration of the topology prediction model is determined from the trained series network model to obtain the trained topology prediction model.

[0099] In this embodiment, based on the architecture of the topology prediction model, a forward training model with the same architecture as the model is constructed. For the convenience of description, the forward training model is called the forward network, the topology prediction model is called the reverse network, and the input layer and output layer in the reverse network are represented by the first input layer and the first output layer, and the input layer and output layer in the forward network are represented by the second input layer and the second output layer. Among them, the architecture of the forward network is as follows Figure 4 As shown, the forward network includes a second input layer, multiple hidden layers and a second output layer. The difference from the reverse network is that the input of the second input layer is the structural parameters of the functional area of ​​the device; the hidden layer is used to assign different weights to the input features; the second output layer is used to output the spectral response of the corresponding device. The preferred number of hidden layers of the forward network is four, and each layer has a different number of nodes.

[0100] For the constructed forward network, the forward network is trained using the above training data set to obtain the trained forward network parameter configuration. Then the trained forward network is connected end to end with the untrained reverse network and the configuration parameters of the forward network are locked to obtain a series network model, such as Figure 5 As shown, it can be seen that in the series network model, the data input to the reverse network is actually the data output by the forward network, and the input data of the forward network and the output data of the reverse network are both structural parameters. Then, the series network model is trained using the above training data set, and the deployed series network parameter configuration will be obtained, which includes the parameter configuration of the forward network and the reverse network. Further, the reverse network, that is, the network parameter configuration of the topology prediction model can be obtained, thereby obtaining a trained topology prediction model. In this way, the input and output of the topology prediction model can be optimally corresponded, thereby obtaining the optimal structural parameters corresponding to the expected response spectrum.

[0101] Through the topology prediction model, 1*N pixel 0-1 codes are obtained to characterize the structural parameters of the functional area of ​​the device. N represents the number of pixel areas divided in the training data set. Taking the above training data set divided into 300 pixel areas as an example, N is 300. Then, according to the structural parameter (pixel) unit code obtained by the output layer, it is determined whether to etch the area; when the corresponding pixel point code is 0, the pixel area is etched, and when it is 1, the pixel area is not etched, so as to obtain the following: Figure 2 The structure of the grating coupler shown. Furthermore, for the obtained structure, a finite difference time domain simulation can be performed to verify the performance indicators of the designed device, including the diffraction length distribution and the response spectrum, so as to determine whether the performance of the device corresponding to the predicted structural parameters meets the requirements, and the structural parameters are fine-tuned based on the performance indicators to further improve the performance of the device. The specific simulation steps can refer to the conventional steps, and will not be repeated here.

[0102] The present embodiment provides a method for designing a grating coupler. The present invention combines a simple network structure with a complex nonlinear algorithm through approximate processing of a training data set, thereby reducing the computing power requirements of the model and improving the prediction efficiency and effect of the model. At the same time, through a combined forward and reverse training method, the prediction accuracy of the model is further improved, thereby realizing the design of a grating coupler with high coupling efficiency at any angle and ensuring the device performance of the grating coupler.

[0103] See also Figure 6 Based on the same inventive concept, a design system for a grating coupler proposed in a second embodiment of the present invention includes:

[0104] A model construction module 10 is used to construct a topology prediction model based on a deep neural network, and to train the topology prediction model using a preset method. The topology prediction model includes an input layer, an output layer, and a plurality of hidden layers. The activation function of the topology prediction model is a Swish function, and the training data set is constructed using approximate processing to adapt to the convergence of the Swish function.

[0105] A parameter prediction module 20, for inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler;

[0106] The grating design module 30 is used to determine the topological structure of the functional area of ​​the grating coupler according to the structural parameters to realize the design of the grating coupler.

[0107] Further, the model building module 10 includes a model training module;

[0108] The model training module is used to generate training data through a simulation method and perform approximate processing on the training data to obtain a training data set;

[0109] The topology prediction model is trained using the training data set.

[0110] Furthermore, the model training module also includes a data set construction module;

[0111] The data set construction module is used to generate training data with SOI structural consistency through finite difference time domain simulation, wherein the training data includes structural parameters and spectral parameters;

[0112] Dividing the functional area of ​​the grating coupler corresponding to the structural parameters into a plurality of pixel areas according to preset sizes;

[0113] Performing data conversion on each pixel region according to a region state to obtain the processed structural parameters, wherein the region state includes an etched state and a non-etched state;

[0114] A training data set is formed according to the processed structural parameters and spectral parameters.

[0115] Furthermore, the model training module is also used to construct a forward training model with the same structure as the topology prediction model, the input data of the forward training model is the structural parameters of the functional area of ​​the grating coupler, and the output data of the forward training model is the response spectrum of the grating coupler;

[0116] Using the training data set to train the forward training model to obtain a first network parameter configuration of the forward training model;

[0117] Connecting the trained forward training model and the untrained topology prediction model in series to obtain a series network model, and training the series network model using the training data set;

[0118] The second network parameter configuration of the topology prediction model is determined from the trained series network model to obtain the trained topology prediction model.

[0119] The technical features and technical effects of the grating coupler design system proposed in the embodiment of the present invention are the same as those of the method proposed in the embodiment of the present invention, and are not described in detail here. Each module in the above-mentioned grating coupler design system can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0120] In addition, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0121] See also Figure 7 , an internal structure diagram of a computer device in one embodiment, the computer device can specifically be a terminal or a server. The computer device includes a processor, a memory, a network interface, a display and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a design method for a grating coupler is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, a trackball or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0122] It can be understood by those skilled in the art that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have the same component arrangement.

[0123] In addition, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0124] In summary, the embodiments of the present invention propose a design method, system, computer device and storage medium for a grating coupler. The method constructs a topology prediction model based on a deep neural network, and trains the topology prediction model using a preset method. The topology prediction model includes an input layer, an output layer and several hidden layers, and the activation function of the topology prediction model is a Swish function. The response spectrum of the grating coupler is input into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler. According to the structural parameters, the topological structure of the functional area of ​​the grating coupler is determined to achieve the design of the grating coupler. The present invention combines a simple network structure with a complex nonlinear algorithm through approximate processing of the training data set, reduces the computing power requirement of the model, improves the prediction efficiency and effect of the model, and further improves the prediction accuracy of the model through a forward and reverse combined training method, thereby achieving the design of a grating coupler with high coupling efficiency at any angle and ensuring the device performance of the grating coupler.

[0125] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0126] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A design method for a grating coupler, characterized in that: include: A topology prediction model is constructed based on a deep neural network, and the topology prediction model is trained using a preset method. The topology prediction model includes an input layer, an output layer, and a plurality of hidden layers. The activation function of the topology prediction model is a Swish function. The training data set is constructed using approximate processing to adapt to the convergence of the Swish function. Inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler; According to the structural parameters, the topological structure of the functional area of ​​the grating coupler is determined to realize the design of the grating coupler.

2. The design method of the grating coupler according to claim 1, characterized in that: The step of training the topology prediction model using a preset method comprises: Generate training data through simulation method, and perform approximate processing on the training data to obtain a training data set; The topology prediction model is trained using the training data set.

3. The design method of the grating coupler according to claim 2, characterized in that: The step of generating training data by simulation method and performing approximate processing on the training data to obtain the training data set comprises: Generate training data with SOI structural consistency through finite-difference time-domain simulation, wherein the training data includes structural training data and spectral training data; Dividing the functional area of ​​the grating coupler corresponding to the structural training data into a plurality of pixel areas according to a preset size; Performing data conversion on each pixel region according to a region state to obtain the processed structure training data, wherein the region state includes an etched state and a non-etched state; A training data set is formed according to the processed structural training data and the spectral training data.

4. The design method of the grating coupler according to claim 2, characterized in that: The step of using the training data set to train the topology prediction model comprises: Constructing a forward training model with the same structure as the topology prediction model, wherein the input data of the forward training model are the structural parameters of the functional area of ​​the grating coupler, and the output data of the forward training model are the response spectrum of the grating coupler; Using the training data set to train the forward training model to obtain a first network parameter configuration of the forward training model; Connecting the trained forward training model and the untrained topology prediction model in series to obtain a series network model, and training the series network model using the training data set; The second network parameter configuration of the topology prediction model is determined from the trained series network model to obtain the trained topology prediction model.

5. A grating coupler design system, characterized in that: include: A model construction module, used to construct a topology prediction model based on a deep neural network, and train the topology prediction model using a preset method, wherein the topology prediction model includes an input layer, an output layer, and a plurality of hidden layers, and the activation function of the topology prediction model is a Swish function, and the training data set is constructed using approximate processing to adapt to the convergence of the Swish function; A parameter prediction module, used for inputting the response spectrum of the grating coupler into the trained topology prediction model to obtain the structural parameters of the functional area of ​​the grating coupler; The grating design module is used to determine the topological structure of the functional area of ​​the grating coupler according to the structural parameters to realize the design of the grating coupler.

6. The design system of the grating coupler according to claim 5, characterized in that: The model building module includes a model training module; The model training module is used to generate training data through a simulation method and perform approximate processing on the training data to obtain a training data set; The topology prediction model is trained using the training data set.

7. The design system of the grating coupler according to claim 6, characterized in that: The model training module also includes a data set construction module; The data set construction module is used to generate training data with SOI structural consistency through finite difference time domain simulation, wherein the training data includes structural parameters and spectral parameters; Dividing the functional area of ​​the grating coupler corresponding to the structural parameters into a plurality of pixel areas according to preset sizes; Performing data conversion on each pixel region according to a region state to obtain the processed structural parameters, wherein the region state includes an etched state and a non-etched state; A training data set is formed according to the processed structural parameters and spectral parameters.

8. The design system of the grating coupler according to claim 6, characterized in that: The model training module is also used to construct a forward training model with the same structure as the topology prediction model, the input data of the forward training model is the structural parameters of the functional area of ​​the grating coupler, and the output data of the forward training model is the response spectrum of the grating coupler; Using the training data set to train the forward training model to obtain a first network parameter configuration of the forward training model; Connecting the trained forward training model and the untrained topology prediction model in series to obtain a series network model, and training the series network model using the training data set; The second network parameter configuration of the topology prediction model is determined from the trained series network model to obtain the trained topology prediction model.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images