Intracavity laser modeling and solving method and system based on physical information neural network and electronic equipment

Abstract

The invention relates to an intracavity laser modeling and solving method and system based on a physical information neural network and electronic equipment. The method comprises the following steps: acquiring intracavity laser system parameters and experimental data; initializing a neural network model; introducing physical constraints in a partial differential equation form into the neural network model to obtain an intracavity laser field solving model based on a physical information neural network; and on the basis of the intracavity laser parameters and experimental data, network parameter updating is carried out on the intracavity laser field solving model by using an adaptive optimization algorithm, when a preset constraint condition is met, modeling results of field intensity distribution and nonlinear effect evolution in the laser cavity are obtained, and modeling and solving tasks of the intracavity laser propagation behavior are completed. Compared with the prior art, the method has the advantages that the solution precision and efficiency of intracavity laser light field distribution and energy output are effectively improved, and the like.

Description

Technical Field

[0001] The present invention relates to the intersection of laser technology and artificial intelligence, and in particular to a method, system and electronic equipment for intracavity laser modeling and solving based on physical information neural network. Background Art

[0002] With the rapid development of optical technology and computing technology, the application demand of intracavity lasers in optical communications, nonlinear optics, quantum information and other fields is increasing. However, accurately simulating and solving the complex behavior of intracavity lasers has always been a technical difficulty. Traditional numerical methods are inefficient when dealing with high-dimensional nonlinear dynamic systems and have huge demands on computing resources. The technology based on Physics-Informed Neural Network (PINN) has become a potential solution for modeling and solving intracavity laser behavior because of its ability to integrate physical laws and efficiently handle complex problems.

[0003] Recently, the application of physical information neural networks in nonlinear optics and light wave propagation has begun to attract attention. It can significantly improve the accuracy and efficiency of optical system modeling by directly embedding physical equations. However, although physical information neural networks have shown superior performance in dealing with physical problems such as optics, how to achieve high-precision prediction of intracavity laser behavior in high-dimensional laser cavity solutions and nonlinear effect modeling is still a problem that needs to be solved in this field. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method, system and electronic equipment for intracavity laser modeling and solution based on physical information neural network, which can effectively improve the solution accuracy and efficiency of intracavity laser light field distribution and energy output.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] According to a first aspect of the present invention, a method for modeling and solving intracavity laser based on a physical information neural network is provided, comprising the following steps: obtaining parameters and experimental data of an intracavity laser system; initializing a neural network model; introducing physical constraints in the form of partial differential equations into the neural network model to obtain an intracavity laser field solution model based on a physical information neural network; updating network parameters of the intracavity laser field solution model using an adaptive optimization algorithm based on the intracavity laser parameters and experimental data, and obtaining modeling results of field intensity distribution and nonlinear effect evolution in the laser cavity when preset constraints are met, thereby completing the modeling and solution tasks of the intracavity laser propagation behavior.

[0007] As a preferred technical solution, the physical constraints in the form of partial differential equations include optical wave equations and nonlinear optical effect equations.

[0008] As a preferred technical solution, the physical constraint in the form of partial differential equation is expressed as:

[0009] Nu(x,t)=f(x,t)

[0010] Bu(x,t)=g(x,t)

[0011] u(x,0)=h(x)

[0012] Where x and t are spatial and temporal data, respectively; u(x, t) is the equation to be solved; N and B are differential operators; f(x, t), g(x, t) and h(x) are the source terms of the original equation, boundary condition equation and initial condition equation, respectively.

[0013] As a preferred technical solution, the loss function used in the updating process of the intracavity laser field solution model based on physical information neural network is expressed as:

[0014] L′=λ data L data +λ PDE L PDE +λ BC L BC +λ IC L IC

[0015] in

[0016]

[0017] Where, L data L is the loss term to measure data deviation. PDE is the loss term of the original equation deviation, L BC is the loss term of boundary condition deviation, L IC is the loss term of the initial condition deviation, λ data , PDE , BC , IC are the weight coefficients corresponding to each loss term, N d 、N f 、N g and N h They represent the number of data points or matching points used to calculate the corresponding loss term, and the subscripts d, f, g, and h represent the parameters and L, respectively. data , L PDE , L BC , L IC correspond.

[0018] As a preferred technical solution, the neural network model is constructed based on a multi-layer perceptron, uses a periodic activation function to capture high-frequency laser oscillation characteristics, and uses residual connections to optimize the network structure.

[0019] As a preferred technical solution, the neural network model also introduces a multi-scale network structure, which includes a shallow network and a deep network. The shallow network is used to capture the low-frequency behavior of the intracavity laser, and the deep network is used to model the high-frequency behavior.

[0020] As a preferred technical solution, the neural network model also embeds a weight normalization layer and a gradient clipping mechanism.

[0021] As a preferred technical solution, the neural network model also uses a multi-resolution data sampling strategy to adjust the network parameters to an optimal range.

[0022] According to a second aspect of the present invention, there is provided an intracavity laser modeling and solution system based on a physical information neural network, the system being used to implement the method described, comprising a data input module, a neural network modeling module, a physical constraint module and a network parameter optimization module; the data input module being used to obtain laser system parameters and experimental data; the neural network modeling module being used to construct and initialize a neural network model; the physical constraint module being used to introduce physical constraints in the form of partial differential equations into the neural network model, so as to obtain an intracavity laser field solution model based on a physical information neural network; the network parameter optimization module being used to update network parameters of the intracavity laser field solution model using an adaptive optimization algorithm based on the intracavity laser parameters and experimental data, and when the preset constraint conditions are met, the modeling results of the field intensity distribution and the evolution of the nonlinear effect in the laser cavity are obtained, thereby completing the modeling and solution tasks of the intracavity laser propagation behavior.

[0023] According to a third aspect of the present invention, there is provided an electronic device, comprising a memory, a processor, and a program stored in the memory, wherein the processor implements the method when executing the program.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] 1. The present invention proposes a model for solving the intracavity laser field based on a physical information neural network, which integrates physical constraints with the neural network model and optimizes the efficient training framework. It can effectively simulate the complex light field distribution and nonlinear effects in the laser cavity and achieve high-precision and fast light field calculation.

[0026] 2. The present invention utilizes the physical information neural network method and introduces physical constraints in the form of partial differential equations to achieve cross-scale optical solution capabilities from microscopic (gain medium distribution, intracavity mode distribution) to macroscopic (system resonant cavity design, light field energy transmission);

[0027] 3. The present invention updates the network parameters of the intracavity laser field solution model through an adaptive optimization algorithm, improves the model convergence efficiency and computing performance, can automatically adapt to different intracavity optical device configurations and parameter changes, is more robust to non-ideal conditions (such as thermal effects, foreign body intrusion, etc.), and has broad engineering application potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic flow chart of the method provided by the present invention;

[0029] Figure 2 Provide a schematic diagram of the structure and flow of the system for the present invention;

[0030] Figure 3 It is a specific flow chart of the neural network modeling module and the physical constraint module in the embodiment of the present invention. DETAILED DESCRIPTION

[0031] Aiming at the problem of laser propagation and behavior prediction in complex nonlinear optical systems, the present invention provides a method for modeling and solving intracavity lasers based on physical information neural networks. The method embeds physical constraints in the form of partial differential equations into the neural network model to achieve high-precision modeling and solving of the complex dynamic behaviors of intracavity lasers, while improving computational efficiency and reducing model training costs. By using a multi-layer perceptron (MLP) as the core modeling structure of the physical information neural network, the present invention can effectively capture the nonlinear dynamic behaviors of intracavity lasers in high-dimensional parameter space, and achieve accurate simulation and solving of complex laser systems through a combination of physical constraints and data-driven methods.

[0032] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0033] Example

[0034] like Figure 1 As shown, the intracavity laser modeling and solution method based on physical information neural network provided in this embodiment includes the following steps:

[0035] Step S1, obtaining intracavity laser system parameters and experimental data;

[0036] Step S2, initializing the neural network model;

[0037] Step S3, introducing physical constraints in the form of partial differential equations into the neural network model to obtain a solution model for the intracavity laser field based on a physical information neural network;

[0038] Step S4, based on the intracavity laser parameters and experimental data, the network parameters of the intracavity laser field solution model are updated using an adaptive optimization algorithm. When the preset constraints are met, the modeling results of the field intensity distribution and nonlinear effect evolution in the laser cavity are obtained, completing the modeling and solution tasks of the intracavity laser propagation behavior.

[0039] The above method can be based on Figure 2 The intracavity laser modeling and solution system shown in the figure is implemented, and the system includes a data input module, a neural network modeling module, a physical constraint module and a network parameter optimization module, which are respectively used to implement steps S1, S2, S3 and S4 of the above method. The system embeds the intracavity optical nonlinear effect equation (such as rate equation, wave equation, etc.) through the physical constraint module to accurately simulate the intracavity laser propagation behavior. The specific process is as follows:

[0040] Data input module

[0041] This module is used to process the input data of the system, including laser system parameters (such as gain medium distribution, cavity length, reflectivity, etc.) and experimentally obtained data, and standardize them into the input format required for network training.

[0042] Neural Network Modeling Module

[0043] like Figure 3 As shown in the figure, this module uses a multi-layer perceptron as the core modeling structure and optimizes the solution performance through the following designs:

[0044] Activation function optimization: Select a periodic activation function (such as sine function, cosine function, etc.) to capture the high-frequency laser oscillation characteristics.

[0045] Network structure optimization: The training stability of the deep network is enhanced by introducing residual connections (ResNet); further, a multi-scale network structure is introduced, in which the shallow network captures the low-frequency behavior of the intracavity laser, and the deep network is responsible for modeling the high-frequency behavior. In addition, weight normalization layers and gradient clipping mechanisms can be embedded in the network to enhance the generalization ability of the model.

[0046] Multi-resolution data sampling strategy: The spatial resolution and time step in the cavity are adjusted to the optimal parameter range that adapts to the physical information neural network, thereby reducing the model complexity.

[0047] Physical Constraint Module

[0048] like Figure 3 As shown, this module is used to explicitly introduce physical constraints in the form of partial differential equations (groups) into the model, including the optical wave equation, nonlinear optical effect equation (such as rate equation), etc. By introducing constraints on time derivatives and space derivatives, it ensures that the solution results are basically consistent with physical laws.

[0049] Specifically, consider the following set of physical constraint equations:

[0050] Nu(x,t)=f(x,t)

[0051] Bu(x,t)=g(x,t)

[0052] u(x,0)=h(x)

[0053] Where x and t are spatial and temporal data, respectively; u(x, t) is the equation to be solved; N and B are differential operators; f(x, t), g(x, t) and h(x) are the source terms of the original equation, boundary condition equation and initial condition equation, respectively.

[0054] In addition, the physical constraint module also constrains the model through the loss function, assigns reasonable weights to different loss terms, and obtains the loss function used by the corresponding intracavity laser field solution model based on physical information neural network during the update process:

[0055] L=λ data L data +λ PDE L PDE +λ BC L BC +λ IC L IC

[0056] in:

[0057]

[0058] Where, L data L is the loss term to measure data deviation. PDE is the loss term of the original equation deviation, L BC is the loss term of boundary condition deviation, L IC is the loss term of the initial condition deviation, λ data , PDE , BC , IC are the weight coefficients corresponding to each loss term, N d 、N f 、N g and N h They represent the number of data points or matching points used to calculate the corresponding loss term, and the subscripts d, f, g, and h represent the parameters and L, respectively.data , L PDE , L BC , L IC correspond.

[0059] Reasonable allocation of the weight ratio between physical constraints and data-driven items can ensure that the model satisfies both physical constraints and data fitting accuracy during the optimization process, thereby further improving modeling accuracy.

[0060] By calculating the loss function and passing the result to the network parameter optimization module and then updating the model, the physical constraint module can effectively improve the accuracy of the model solution results while making the results follow the real physical laws.

[0061] Figure 3 In this example, m represents the maximum number of partial derivatives of x involved in the partial differential equation to be solved, and θ represents the parameters in the network (u θ represents a network with θ as a parameter), n represents the number of data points, and d represents the dimension of each data point (x is an n*d matrix, n rows represent n data points, and each data point is a d-dimensional vector). c Represents the number of matching points.

[0062] Network parameter optimization module

[0063] This module uses adaptive optimization algorithms (such as Adam, L-BFGS, etc.) to update network parameters to improve model convergence speed and computational efficiency.

[0064] According to the results of the physical constraint module and the network parameter optimization module, it is determined whether the physical loss and data loss of the model are within the preset threshold range. If not, the network parameters are optimized iteratively until the constraints are met. Finally, the modeling results of the field intensity distribution and nonlinear effect evolution in the laser cavity are obtained, thereby completing the task of accurately modeling and solving the laser propagation behavior in the cavity.

[0065] In a simulation, the neural network modeling module uses a multi-layer perceptron as the core modeling structure, and uses a sine function as a periodic activation function, introduces a multi-scale network design, a residual connection, and uses a multi-resolution data sampling strategy. At the same time, the intracavity optical nonlinear effect equation and time and space derivative constraints are introduced. The adaptive optimization algorithm selects Adam optimization for parameter update. Specifically, the number of layers of the multi-layer perceptron in the neural network structure is 4, with 128 neurons in each layer. Based on this neural network structure, the aforementioned method is implemented, and the intracavity laser steady-state distribution is solved with high precision to achieve high-precision prediction of the laser field distribution. The final simulation error is less than 5%, indicating that the method provided in this embodiment can effectively capture the nonlinear dynamic behavior and high-frequency characteristics of the laser system.

[0066] In summary, the method provided by the present invention introduces a periodic activation function and a physical information neural network structure that supports multi-scale data analysis when solving the high-frequency problem of intracavity laser propagation, which can significantly improve the model's ability to capture high-frequency light field features. At the same time, through the adaptive optimization algorithm and multi-resolution data sampling strategy, the model training time and computing resource requirements can be reduced.

[0067] In some other embodiments, the present invention can be used in the fields of high-power laser cavity design, optical communication system optimization, and optical nonlinear effect research. For example, in the field of optical communication, it can be used to simulate the intracavity propagation behavior of ultrashort pulse lasers; in high-power laser systems, it can be used to evaluate the stability of lasers and calculate energy conversion efficiency.

[0068] Further, the present embodiment also provides an electronic device, including a memory, a processor, and a program stored in the memory, and the processor implements one or more steps of the aforementioned method when executing the program. The electronic device processor includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or computer program instructions loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for the operation of the device can also be stored. The CPU, ROM, and RAM are connected to each other through a bus. The input / output (I / O) interface is also connected to the bus. Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks. The processing unit performs the various methods and processes described above, such as one or more steps of the aforementioned method. For example, in some embodiments, one or more steps of the foregoing method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of the foregoing method may be executed. Alternatively, in other embodiments, the CPU may be configured to execute one or more steps of the foregoing method in any other appropriate manner (e.g., by means of firmware). The functions described above may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0069] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for modeling and solving intracavity laser based on physical information neural network, characterized in that: The following steps are involved: Obtain intracavity laser system parameters and experimental data; Initialize the neural network model; Introducing physical constraints in the form of partial differential equations into the neural network model to obtain a solution model for the intracavity laser field based on a physical information neural network; Based on the intracavity laser parameters and experimental data, an adaptive optimization algorithm is used to update the network parameters of the intracavity laser field solution model. When the preset constraints are met, the modeling results of the field intensity distribution and nonlinear effect evolution in the laser cavity are obtained, completing the modeling and solution tasks of the intracavity laser propagation behavior.

2. The intracavity laser modeling and solution method based on physical information neural network according to claim 1 is characterized in that: The physical constraints in the form of partial differential equations include optical wave equations and nonlinear optical effect equations.

3. The intracavity laser modeling and solving method based on physical information neural network according to claim 1 is characterized in that: The physical constraints in the form of partial differential equations are expressed as: Nu(x,t)=f(x,t) Bu(x,t)=g(x,t) u(x,0)=h(x) Where x and t are spatial and temporal data, respectively; u(x, t) is the equation to be solved; N and B are differential operators; f(x, t), g(x, t) and h(x) are the source terms of the original equation, boundary condition equation and initial condition equation, respectively.

4. The intracavity laser modeling and solution method based on physical information neural network according to claim 3 is characterized in that: The loss function used in the updating process of the intracavity laser field solution model based on physical information neural network is expressed as: L′=λ data L data +λ PDE L PDE +λ BC L BC +λ IC L IC in: Where, L data L is the loss term to measure data deviation. PDE is the loss term of the original equation deviation, L BC is the loss term of boundary condition deviation, L IC is the loss term of the initial condition deviation, λ data , PDE , BC , IC are the weight coefficients corresponding to each loss term, N d 、N f 、N g and N h They represent the number of data points or matching points used to calculate the corresponding loss term, and the subscripts d, f, g, and h represent the parameters and L, respectively. data , L PDE , L BC , L IC correspond.

5. The intracavity laser modeling and solving method based on physical information neural network according to claim 1 is characterized in that: The neural network model is built based on a multi-layer perceptron, uses a periodic activation function to capture high-frequency laser oscillation characteristics, and uses residual connections to optimize the network structure.

6. The intracavity laser modeling and solving method based on physical information neural network according to claim 5 is characterized in that: The neural network model also introduces a multi-scale network structure, which includes a shallow network and a deep network. The shallow network is used to capture the low-frequency behavior of the intracavity laser, and the deep network is used to model the high-frequency behavior.

7. The intracavity laser modeling and solving method based on physical information neural network according to claim 6 is characterized in that: The neural network model also embeds a weight normalization layer and a gradient clipping mechanism.

8. The intracavity laser modeling and solving method based on physical information neural network according to claim 7 is characterized in that: The neural network model also uses a multi-resolution data sampling strategy to adjust the network parameters to an optimal range.

9. An intracavity laser modeling and solution system based on physical information neural network, characterized in that: The system is used to implement the method according to any one of claims 1 to 8, comprising a data input module, a neural network modeling module, a physical constraint module and a network parameter optimization module; The data input module is used to obtain laser system parameters and experimental data; The neural network modeling module is used to build and initialize the neural network model; The physical constraint module is used to introduce physical constraints in the form of partial differential equations into the neural network model to obtain a solution model for the intracavity laser field based on a physical information neural network; The network parameter optimization module is used to update the network parameters of the intracavity laser field solution model based on the intracavity laser parameters and experimental data using an adaptive optimization algorithm. When the preset constraints are met, the modeling results of the field intensity distribution and nonlinear effect evolution in the laser cavity are obtained, thereby completing the modeling and solution tasks of the intracavity laser propagation behavior.

10. An electronic device comprising a memory, a processor, and a program stored in the memory, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

Images