CN118410725A - Laser reverse design method and system based on neural network algorithm - Google Patents
Laser reverse design method and system based on neural network algorithm Download PDFInfo
- Publication number
- CN118410725A CN118410725A CN202410876871.3A CN202410876871A CN118410725A CN 118410725 A CN118410725 A CN 118410725A CN 202410876871 A CN202410876871 A CN 202410876871A CN 118410725 A CN118410725 A CN 118410725A
- Authority
- CN
- China
- Prior art keywords
- laser
- physical characteristic
- characteristic parameters
- characteristic parameter
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013461 design Methods 0.000 title claims abstract description 128
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 72
- 238000000034 method Methods 0.000 title claims abstract description 43
- 239000011159 matrix material Substances 0.000 claims description 23
- 238000012549 training Methods 0.000 claims description 22
- 238000013139 quantization Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 15
- 238000012407 engineering method Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 230000000704 physical effect Effects 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004088 simulation Methods 0.000 claims description 6
- 238000012790 confirmation Methods 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 4
- 239000000463 material Substances 0.000 description 14
- 238000004891 communication Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000012938 design process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000005086 pumping Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000017525 heat dissipation Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Semiconductor Lasers (AREA)
Abstract
The invention relates to the technical field of laser design, and provides a laser reverse design method and system based on a neural network algorithm, wherein the method comprises the following steps: acquiring output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network; inputting at least one group of physical characteristic parameters of a laser to the laser design model to obtain actual output characteristic parameters of the laser design model; acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result; and adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser. The invention solves the problems of high design cost and low efficiency of the existing laser.
Description
Technical Field
The invention relates to the technical field of laser design, in particular to a laser reverse design method and system based on a neural network algorithm.
Background
A laser is a device capable of emitting a high intensity, monochromatic, coherent beam of light. There are various types and uses such as semiconductor lasers, gas lasers, solid state lasers. Lasers can be used in communication, medical, industrial fields. Lasers are required to possess different properties, including different thresholds, efficiencies, linewidths, based on different application objectives. While the materials and structure of the laser are the primary factors affecting the performance of the laser, the goal of the laser design is to optimize the active region, buried layer, grating, electrode, end-facet reflectivity, resonant cavity length components to produce a laser output with the desired wavelength, mode and power.
Laser design is a complex process requiring expertise in multiple areas of optics, material science, and thermal management. The general laser design process first requires determining the wavelength, power, mode and beam quality requirements of the desired laser. And selecting a proper gain medium and an optimal pumping mode according to the required wavelength and power output. The cavity is then designed and the heat dissipation mechanism is designed according to the requirements for beam mode and quality. Finally, multiple tests, adjustments and optimizations are required to be performed on the laser design prototype to improve the performance of the laser. Multiple modeling and simulation tools are required in this process to predict the behavior of the laser and optimize its performance. In addition, careful consideration of the materials and manufacturing processes of the laser components (e.g., gain medium, pump source, and optical elements) is required.
The current method of forward designing a laser requires multiple prototypes and tests to achieve the desired performance, which is a lengthy iterative process. The forward laser design requires consideration of a number of parameters and factors, including wavelength, power, mode, beam quality, pumping scheme, and cavity design, which are quite complex. There is an interdependence and complex correlation between these parameters, which requires comprehensive consideration and optimization. The design process can be cumbersome and time consuming, requiring substantial expertise and empirical support. In addition, laser forward design trial and error is costly and it is often necessary to verify the feasibility and performance of the design through prototyping and testing after the basic simulation is completed. If the design is not ideal or problematic, multiple iterations and adjustments may be required, resulting in higher trial-and-error costs. Especially for high power lasers or complex systems, the resources and time investment required for prototyping and testing are large.
Disclosure of Invention
The invention provides a laser reverse design method and system based on a neural network algorithm, which are used for solving the problems of high design cost and low efficiency of the existing laser.
The invention provides a laser reverse design method based on a neural network algorithm, which comprises the following steps:
Acquiring output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network;
Inputting at least one group of physical characteristic parameters of a laser to the laser design model to obtain actual output characteristic parameters of the laser design model;
Acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result;
And adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser.
According to the laser reverse design method based on the neural network algorithm provided by the invention, the physical characteristic parameters of at least one group of lasers are input into the laser design model, and the method further comprises the following steps: inputting initial physical characteristic parameters of a laser into the laser design model;
the adjusting the physical characteristic parameters of the laser based on the deviation, and completing the reverse design of the target laser further comprises:
Under the condition that the deviation is larger than a preset threshold, modifying and iterating the input initial physical characteristic parameters repeatedly by adopting a back propagation algorithm based on the differentiable characteristic of the neural network model; until the deviation is not greater than the preset threshold;
And taking the physical characteristic parameter corresponding to the current laser design model as the design parameter of the target laser to be designed under the condition that the deviation is not larger than the preset threshold value.
According to the laser reverse design method based on the neural network algorithm,
The step of obtaining the output characteristic parameters of the target laser is further to obtain the output characteristic parameters of the laser obtained by the existing simulation method or directly test the required output characteristic parameters of the laser;
and said inputting at least one set of physical characteristic parameters of the laser into the laser design model comprises:
inputting a plurality of groups of first physical characteristic parameters and second physical characteristic parameters into the laser design model; wherein,
The first physical characteristic parameter is a physical characteristic parameter that can be characterized by a quantization parameter;
the second physical characteristic parameters are physical characteristic parameters represented by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
According to the laser reverse design method based on the neural network algorithm,
The laser design model is formed by training a plurality of groups of physical characteristic parameters of lasers and corresponding output characteristic parameters based on a preset neural network, and further comprises:
Obtaining a plurality of groups of laser samples, wherein the laser samples comprise physical characteristics and corresponding output characteristics of each laser;
quantifying a plurality of groups of laser samples to obtain physical characteristic parameter samples and output characteristic parameter samples of the lasers;
storing the physical characteristic parameter sample and the output characteristic parameter sample of the laser according to a preset data structure to obtain a training data set;
Based on the training data set, training the neural network according to the relation between the physical characteristic parameters of the plurality of groups of lasers and the output characteristic parameters, and obtaining the laser design model.
According to the laser reverse design method based on the neural network algorithm,
Further among the obtaining physical characteristic parameter samples and output characteristic parameter samples of the laser,
Obtaining a first physical characteristic parameter sample and a second physical characteristic parameter sample of the laser; wherein,
The first physical characteristic parameter sample is a physical characteristic parameter sample which can be characterized by one quantization parameter;
The second physical characteristic parameter sample is characterized by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
According to the laser reverse design method based on the neural network algorithm, the second physical characteristic parameters can be obtained through the following steps:
Performing grid division on the laser corresponding to the physical characteristic parameter sample of the laser in a coordinate system comprising X, Y and Z coordinate axes, and coordinating each grid point to obtain coordinates (x, y, Z) of each grid point;
Defining a function ，Representing the physical property values at coordinates (x, y, z);
Acquiring a function value corresponding to each grid based on grid coordinates according to the physical characteristic parameter sample;
；
then the first time period of the first time period, Matrix elements serving as physical characteristic matrixes of the designated areas;
The matrix of a certain appointed physical characteristic of the laser sample is ；
The matrix M is a second physical characteristic parameter.
The invention also provides a laser reverse design system based on the neural network algorithm, which comprises:
The data acquisition module is used for acquiring the output characteristic parameters of the target laser and taking the output characteristic parameters as the output characteristic parameters of the trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network;
The output confirmation module is used for inputting at least one group of physical characteristic parameters of the laser to the laser design model and obtaining actual output characteristic parameters of the laser design model;
The deviation acquisition module is used for acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser and generating a deviation result;
And the reverse design module is used for adjusting physical characteristic parameters of the laser based on the deviation result to finish reverse design of the target laser.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the laser reverse design method based on the neural network algorithm when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a neural network algorithm-based laser reverse engineering method as described in any one of the above.
The invention also provides a computer program product comprising a computer program which when executed by a processor implements a laser reverse engineering method based on a neural network algorithm as described in any one of the above.
The invention provides a reverse design method and a reverse design system for a laser based on a neural network algorithm, which are characterized in that relatively simple parameters of the laser are directly used as input, a three-dimensional network is constructed according to an xyz coordinate system, relatively complex characteristics such as materials, doping concentration and geometric structure related characteristics are parameterized as functions of positions, a plurality of functions of the same characteristics form a characteristic matrix, a plurality of characteristic matrices are also used as the input of the neural network, the output characteristic of the laser is used as the output of the neural network, the obtained neural network algorithm is fixed according to the physical characteristics and the output characteristic of the existing laser as a training data set of the neural network, the output characteristic is modified by using a reverse propagation algorithm, and the output is used as the training parameter. The physical characteristics of the existing laser are changed under the condition of adjusting the output characteristics, so that the novel laser meeting the requirements is obtained.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a laser reverse design method based on a neural network algorithm.
Fig. 2 is a schematic diagram of module connection of a laser reverse design system based on a neural network algorithm.
Fig. 3 is a schematic structural diagram of an electronic device provided by the present invention.
Reference numerals: 110: a data acquisition module; 120: an output confirmation module; 130: a deviation acquisition module; 140: reverse design module; 310: a processor; 320: a communication interface; 330: a memory; 340: a communication bus.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The following describes a laser reverse design method based on a neural network algorithm with reference to fig. 1, which comprises the following steps: step 100, obtaining output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network.
In the present invention, the design is optimized and improved by neural network algorithms starting from existing laser systems or devices by inputting the required laser characteristics (threshold, efficiency, linewidth, single mode stability, wavelength stability, high temperature operation characteristics, high frequency modulation characteristics, anti-reflection characteristics). The complex process of forward design is simplified, the efficiency can be improved, the cost can be reduced, the stability can be improved, the rapid iteration can be realized, and the research and development and push-out speeds of products can be increased.
Specifically, the output characteristics of the laser include threshold, power, efficiency, linewidth characteristics, single mode stability, wavelength stability, high temperature operation characteristics, high frequency modulation characteristics, antireflection light characteristics. As the output of the neural network, there areThen there is。
Step 200, inputting at least one group of physical characteristic parameters of the laser to the laser design model, and obtaining actual output characteristic parameters of the laser design model.
Specifically, inputting initial physical characteristic parameters of a laser into the laser design model;
the adjusting the physical characteristic parameters of the laser based on the deviation, and completing the reverse design of the target laser further comprises:
Under the condition that the deviation is larger than a preset threshold, modifying and iterating the input initial physical characteristic parameters repeatedly by adopting a back propagation algorithm based on the differentiable characteristic of the neural network model until the deviation is not larger than the preset threshold;
And taking the physical characteristic parameter corresponding to the current laser design model as the design parameter of the target laser to be designed under the condition that the deviation is not larger than the preset threshold value.
The method comprises the steps of obtaining output characteristic parameters of a target laser, wherein the output characteristic parameters of the target laser are further obtained by an existing simulation method, or the required output characteristic parameters of the laser are directly tested;
and said inputting at least one set of physical characteristic parameters of the laser into the laser design model comprises:
Inputting a plurality of groups of first physical characteristic parameters and second physical characteristic parameters into the laser design model; wherein the first physical characteristic parameter is a physical characteristic parameter that can be characterized by a quantization parameter;
the second physical characteristic parameters are physical characteristic parameters represented by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
The laser design model is based on a preset neural network and is trained by a plurality of groups of physical characteristic parameters of lasers and corresponding output characteristic parameters, and further comprises:
Obtaining a plurality of groups of laser samples, wherein the laser samples comprise physical characteristics and corresponding output characteristics of each laser;
quantifying a plurality of groups of laser samples to obtain physical characteristic parameter samples and output characteristic parameter samples of the lasers;
storing the physical characteristic parameter sample and the output characteristic parameter sample of the laser according to a preset data structure to obtain a training data set;
Based on the training data set, training the neural network according to the relation between the physical characteristic parameters of the plurality of groups of lasers and the output characteristic parameters, and obtaining the laser design model.
Further, a first physical characteristic parameter sample and a second physical characteristic parameter sample of the laser are obtained; wherein,
The first physical characteristic parameter sample is a physical characteristic parameter sample which can be characterized by one quantization parameter;
The second physical characteristic parameter sample is characterized by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
The second physical characteristic parameter may be obtained by the following steps.
And carrying out grid division on the laser corresponding to the physical characteristic parameter sample of the laser in a coordinate system comprising X, Y and Z coordinate axes, and carrying out the coordinated treatment on each grid point to obtain the coordinates (x, y, Z) of each grid point.
Defining a function，Representing the physical property values at coordinates (x, y, z).
And acquiring the function value corresponding to each grid based on the grid coordinates according to the physical characteristic parameter sample.
。
Then the first time period of the first time period,As matrix elements of the physical property matrix of the designated area.
The matrix of a certain appointed physical characteristic of the laser sample is。
The matrix M is a second physical characteristic parameter.
And 300, obtaining deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result.
And 400, adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser.
The first physical characteristic parameter sample specifically includes:
the structural features of the laser include: buried layer, active region, grating, electrode and resonant cavity length;
and carrying out grid division on the structural features, and carrying out coordinated treatment on each grid point to obtain coordinate information of each grid point.
Factors affecting laser characteristics can be largely divided into the following: buried layers (structure, material, shape, size, doping concentration), active regions (quantum well material composition, well thickness, number of layers, strain amount, barreer layer thickness), gratings (binding coefficient, displacement amount, distribution area), electrodes (shape, material), resonant cavity length. For simpler features, such as resonant cavity length, number of quantum well layers, can be directly input as a parameter. For relatively complex parameters, a reasonably designed parameterization process is needed, because the geometric structure of the laser is complex, the structures and performances represented by different positions are different, for convenience of parameterization, the laser is divided into grids in an X, Y and Z coordinate system, each grid point is coordinated, and the coordinates (X, Y and Z) of each grid point are obtained.
In a specific embodiment, taking the active region quantum well material composition as an example, after meshing, the active region is partitioned into multiple meshes, and the material of each mesh can be expressed as a function of (x, y, z), such as M (x, y, z). Multiple grid point materials can be written asEach grid point of the active region can be used as a matrix element of an active region material matrix, and then the matrix existsThe material composition of the active region is parameterized. Similarly, for complex features, similar parameterization methods can be used, and features of each point corresponding to the X, Y, and Z coordinate systems are used as matrix elements of the features to form a matrix describing the features. Thus, a plurality of parameterized expressions describing the characteristics of the complete laser are obtained, and the matrixes and simple characteristic parameters are used as inputs of neural network training, namely the parameters can be expressed as。
In the present invention, neural network training is performed by parameterizing the physical characteristics and output characteristics of existing lasers as the inputs and outputs of the neural network. And then, under the condition of fixing the weight of the trained neural network, continuously optimizing the input of any given output based on the neural network as a trainable parameter by using a back propagation algorithm, thereby obtaining the novel laser meeting the requirements.
The invention provides a reverse design method of a laser based on a neural network algorithm, which takes relatively simple parameters of the laser as input, constructs a three-dimensional network according to an xyz coordinate system, takes relatively complex characteristics such as material, doping concentration and geometric structure related characteristic parameters as functions of positions, forms a characteristic matrix by a plurality of functions of the same characteristic, takes a plurality of characteristic matrices as the input of the neural network, takes output characteristics of the laser as the output of the neural network, takes the physical characteristics and the output characteristics of the existing laser as training data sets of the neural network, fixes the obtained neural network algorithm, utilizes a reverse propagation algorithm to modify the output characteristics, and takes the output as training parameters. The physical characteristics of the existing laser are changed under the condition of adjusting the output characteristics, so that the novel laser meeting the requirements is obtained.
Referring to fig. 2, the invention also discloses a laser reverse design system based on the neural network algorithm, which comprises:
The data acquisition module 110 is configured to acquire an output characteristic parameter of a target laser, and take the output characteristic parameter as an output characteristic parameter of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network;
an output confirmation module 120, configured to input at least one set of physical characteristic parameters of the laser to the laser design model, and obtain actual output characteristic parameters of the laser design model;
The deviation obtaining module 130 is configured to obtain a deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generate a deviation result;
And the reverse design module 140 is used for adjusting physical characteristic parameters of the laser based on the deviation result to complete reverse design of the target laser.
Wherein inputting at least one set of physical characteristic parameters of the laser into the laser design model is further: inputting initial physical characteristic parameters of a laser into the laser design model;
the adjusting the physical characteristic parameters of the laser based on the deviation, and completing the reverse design of the target laser further comprises:
Under the condition that the deviation is larger than a preset threshold, modifying and iterating the input initial physical characteristic parameters repeatedly by adopting a back propagation algorithm based on the differentiable characteristic of the neural network model; until the deviation is not greater than the preset threshold;
And taking the physical characteristic parameter corresponding to the current laser design model as the design parameter of the target laser to be designed under the condition that the deviation is not larger than the preset threshold value.
Acquiring the output characteristic parameters of the target laser further comprises acquiring the output characteristic parameters of the laser obtained by the existing simulation method or directly testing the required output characteristic parameters of the laser;
and said inputting at least one set of physical characteristic parameters of the laser into the laser design model comprises:
Inputting a plurality of groups of first physical characteristic parameters and second physical characteristic parameters into the laser design model; wherein the method comprises the steps of
The first physical characteristic parameter is a physical characteristic parameter that can be characterized by a quantization parameter;
the second physical characteristic parameters are physical characteristic parameters represented by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
The laser design model is formed by training a plurality of groups of physical characteristic parameters of lasers and corresponding output characteristic parameters based on a preset neural network, and further comprises:
Obtaining a plurality of groups of laser samples, wherein the laser samples comprise physical characteristics and corresponding output characteristics of each laser;
quantifying a plurality of groups of laser samples to obtain physical characteristic parameter samples and output characteristic parameter samples of the lasers;
storing the physical characteristic parameter sample and the output characteristic parameter sample of the laser according to a preset data structure to obtain a training data set;
Based on the training data set, training the neural network according to the relation between the physical characteristic parameters of the plurality of groups of lasers and the output characteristic parameters, and obtaining the laser design model.
The obtaining of the physical characteristic parameter sample and the output characteristic parameter sample of the laser further comprises obtaining a first physical characteristic parameter sample and a second physical characteristic parameter sample of the laser;
Wherein the first physical characteristic parameter sample is a physical characteristic parameter sample that can be characterized by a quantization parameter;
The second physical characteristic parameter sample is characterized by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
The second physical characteristic parameter may be obtained by:
And carrying out grid division on the laser corresponding to the physical characteristic parameter sample of the laser in a coordinate system comprising X, Y and Z coordinate axes, and carrying out the coordinated treatment on each grid point to obtain the coordinates (x, y, Z) of each grid point.
Defining a function，Representing the physical property values at coordinates (x, y, z).
And acquiring the function value corresponding to each grid based on the grid coordinates according to the physical characteristic parameter sample.
。
Wherein the method comprises the steps ofAs matrix elements of the physical property matrix of the designated area.
The matrix of a certain appointed physical characteristic of the laser sample is; The matrix M is a second physical characteristic parameter.
The invention provides a laser reverse design system based on a neural network algorithm, which takes relatively simple parameters of a laser as input, constructs a three-dimensional network according to an xyz coordinate system, takes relatively complex characteristics such as material, doping concentration and geometric structure related characteristic parameters as functions of positions, forms a characteristic matrix by a plurality of functions of the same characteristic, takes a plurality of characteristic matrices as the input of the neural network, takes output characteristics of the laser as the output of the neural network, takes the output characteristics of the existing laser as a training data set of the neural network, fixes the obtained neural network algorithm according to the physical characteristics and the output characteristics of the existing laser, and utilizes a reverse propagation algorithm to modify the output characteristics and take the output as the training parameters. The physical characteristics of the existing laser are changed under the condition of adjusting the output characteristics, so that the novel laser meeting the requirements is obtained.
Fig. 3 illustrates a physical schematic diagram of an electronic device, as shown in fig. 3, where the electronic device may include: processor 310, communication interface (Communications Interface) 320, memory 330 and communication bus 340, wherein processor 310, communication interface 320 and memory 330 communicate with each other via communication bus 340. The processor 310 may invoke logic instructions in the memory 330 to perform a neural network algorithm based laser reverse design method comprising: acquiring output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network; inputting at least one group of physical characteristic parameters of a laser to the laser design model to obtain actual output characteristic parameters of the laser design model; acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result; and adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser.
Further, the logic instructions in the memory 330 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, where the computer program product includes a computer program, where the computer program can be stored on a non-transitory computer readable storage medium, where the computer program when executed by a processor can perform a method for designing a reverse direction of a laser based on a neural network algorithm provided by the above methods, where the method includes: acquiring output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network; inputting at least one group of physical characteristic parameters of a laser to the laser design model to obtain actual output characteristic parameters of the laser design model; acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result; and adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser.
In yet another aspect, the present invention provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the neural network algorithm-based laser reverse engineering method provided by the above methods, the method comprising: acquiring output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network; inputting at least one group of physical characteristic parameters of a laser to the laser design model to obtain actual output characteristic parameters of the laser design model; acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result; and adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The reverse laser design method based on the neural network algorithm is characterized by comprising the following steps of:
Acquiring output characteristic parameters of a target laser, and taking the output characteristic parameters as output characteristic parameters of a trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network;
Inputting at least one group of physical characteristic parameters of a laser to the laser design model to obtain actual output characteristic parameters of the laser design model;
Acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser, and generating a deviation result;
And adjusting physical characteristic parameters of the laser based on the deviation result to finish the reverse design of the target laser.
2. The neural network algorithm-based laser reverse engineering method of claim 1, wherein inputting the at least one set of physical characteristic parameters of the laser into the laser design model is further as follows: inputting initial physical characteristic parameters of a laser into the laser design model;
the adjusting the physical characteristic parameters of the laser based on the deviation, and completing the reverse design of the target laser further comprises:
Under the condition that the deviation is larger than a preset threshold, modifying and iterating the input initial physical characteristic parameters repeatedly by adopting a back propagation algorithm based on the differentiable characteristic of the neural network model; until the deviation is not greater than the preset threshold;
And taking the physical characteristic parameter corresponding to the current laser design model as the design parameter of the target laser to be designed under the condition that the deviation is not larger than the preset threshold value.
3. The neural network algorithm-based laser reverse design method according to claim 2, wherein the obtaining the output characteristic parameters of the target laser is further obtaining the output characteristic parameters of the laser or directly testing the required output characteristic parameters of the laser by the existing simulation method;
and said inputting at least one set of physical characteristic parameters of the laser into the laser design model comprises:
inputting a plurality of groups of first physical characteristic parameters and second physical characteristic parameters into the laser design model; wherein,
The first physical characteristic parameter is a physical characteristic parameter that can be characterized by a quantization parameter;
the second physical characteristic parameters are physical characteristic parameters represented by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
4. The neural network algorithm-based laser reverse design method according to any one of claims 1 to 3, wherein the laser design model is trained based on a preset neural network by physical characteristic parameters of a plurality of groups of lasers and corresponding output characteristic parameters, and further comprising:
Obtaining a plurality of groups of laser samples, wherein the laser samples comprise physical characteristics and corresponding output characteristics of each laser;
quantifying a plurality of groups of laser samples to obtain physical characteristic parameter samples and output characteristic parameter samples of the lasers;
storing the physical characteristic parameter sample and the output characteristic parameter sample of the laser according to a preset data structure to obtain a training data set;
Based on the training data set, training the neural network according to the relation between the physical characteristic parameters of the plurality of groups of lasers and the output characteristic parameters, and obtaining the laser design model.
5. The method for reverse engineering a laser based on a neural network algorithm according to claim 4, wherein the obtaining a physical characteristic parameter sample and an output characteristic parameter sample of the laser is further,
Obtaining a first physical characteristic parameter sample and a second physical characteristic parameter sample of the laser; wherein,
The first physical characteristic parameter sample is a physical characteristic parameter sample which can be characterized by one quantization parameter;
The second physical characteristic parameter sample is characterized by quantization parameters corresponding to the positions of all coordinate points after the given area of the laser forms coordinates based on grid division.
6. The neural network algorithm-based laser reverse engineering method of claim 5, wherein the second physical characteristic parameter is obtained by:
Performing grid division on the laser corresponding to the physical characteristic parameter sample of the laser in a coordinate system comprising X, Y and Z coordinate axes, and coordinating each grid point to obtain coordinates (x, y, Z) of each grid point;
Defining a function ，Representing the physical property values at coordinates (x, y, z);
Acquiring a function value corresponding to each grid based on grid coordinates according to the physical characteristic parameter sample;
；
then the first time period of the first time period, As matrix elements of the physical property matrix of the designated area,
The matrix of a certain appointed physical characteristic of the laser sample is；
The matrix M is a second physical characteristic parameter.
7. A neural network algorithm-based laser reverse engineering system, the system comprising:
The data acquisition module is used for acquiring the output characteristic parameters of the target laser and taking the output characteristic parameters as the output characteristic parameters of the trained laser design model; the laser design model is trained by a plurality of groups of physical characteristic parameter samples of lasers and corresponding output characteristic parameter samples based on a preset neural network;
The output confirmation module is used for inputting at least one group of physical characteristic parameters of the laser to the laser design model and obtaining actual output characteristic parameters of the laser design model;
The deviation acquisition module is used for acquiring the deviation between the actual output characteristic parameter and the output characteristic parameter of the target laser and generating a deviation result;
And the reverse design module is used for adjusting physical characteristic parameters of the laser based on the deviation result to finish reverse design of the target laser.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the neural network algorithm-based laser reverse engineering method of any one of claims 1 to 6 when the program is executed.
9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the neural network algorithm-based laser reverse engineering method of any one of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements a laser reverse engineering method based on a neural network algorithm according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410876871.3A CN118410725B (en) | 2024-07-02 | 2024-07-02 | Laser reverse design method and system based on neural network algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410876871.3A CN118410725B (en) | 2024-07-02 | 2024-07-02 | Laser reverse design method and system based on neural network algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118410725A true CN118410725A (en) | 2024-07-30 |
| CN118410725B CN118410725B (en) | 2024-09-17 |
Family
ID=92033009
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410876871.3A Active CN118410725B (en) | 2024-07-02 | 2024-07-02 | Laser reverse design method and system based on neural network algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118410725B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112989701A (en) * | 2021-03-15 | 2021-06-18 | 华中科技大学 | Wavelength tuning method of SGDBR tunable semiconductor laser |
| CN114322751A (en) * | 2020-09-30 | 2022-04-12 | 广东博智林机器人有限公司 | Target measuring method, target measuring device, computer equipment and storage medium |
| CN115719043A (en) * | 2022-12-02 | 2023-02-28 | 合肥工业大学 | Method and system for designing optical fiber pulse vortex laser based on neural network |
| US20230118061A1 (en) * | 2020-03-25 | 2023-04-20 | Bystronic Laser Ag | Quality control of a laser machining process using machine learning |
| CN118036370A (en) * | 2024-02-02 | 2024-05-14 | 香港中文大学（深圳） | Design method and device of photonic crystal surface emitting laser and intelligent terminal |
-
2024
- 2024-07-02 CN CN202410876871.3A patent/CN118410725B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20230118061A1 (en) * | 2020-03-25 | 2023-04-20 | Bystronic Laser Ag | Quality control of a laser machining process using machine learning |
| CN114322751A (en) * | 2020-09-30 | 2022-04-12 | 广东博智林机器人有限公司 | Target measuring method, target measuring device, computer equipment and storage medium |
| CN112989701A (en) * | 2021-03-15 | 2021-06-18 | 华中科技大学 | Wavelength tuning method of SGDBR tunable semiconductor laser |
| CN115719043A (en) * | 2022-12-02 | 2023-02-28 | 合肥工业大学 | Method and system for designing optical fiber pulse vortex laser based on neural network |
| CN118036370A (en) * | 2024-02-02 | 2024-05-14 | 香港中文大学（深圳） | Design method and device of photonic crystal surface emitting laser and intelligent terminal |
Non-Patent Citations (1)
| Title |
|---|
| 杨阳;程寅;徐明顺;王石语;过振;蔡德芳;李兵斌;: "神经网络在调Q激光器优化设计上的应用", 中国激光, no. 03, 10 March 2010 (2010-03-10) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118410725B (en) | 2024-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tovar et al. | Topology optimization using a hybrid cellular automaton method with local control rules | |
| CN113298257A (en) | Method and device for creating a model of a technical system from measurements | |
| CN111900716A (en) | Random power flow uncertainty quantification method based on sparse chaotic polynomial approximation | |
| EP3732593A1 (en) | Method for applying generative design to the configuration of mechanical assemblies | |
| CN114647081A (en) | Diffraction optical element based on neural network and design method thereof | |
| CN118410725B (en) | Laser reverse design method and system based on neural network algorithm | |
| Viegas et al. | Distributed controller design and performance optimization for discrete‐time linear systems | |
| CN112394640B (en) | Parameter setting method and device, storage medium and parameter setting unit | |
| Kianifar et al. | Application of permutation genetic algorithm for sequential model building–model validation design of experiments | |
| CN117349585B (en) | Operator performance optimization method based on accelerator constraint | |
| CN114065387A (en) | Frame optimization method and device, computer equipment and storage medium | |
| CN115310209B (en) | Pneumatic shape migration optimization method and related device based on VAE | |
| Zhao et al. | Automatic optimal block decomposition for structured mesh generation using genetic algorithm | |
| CN113673042B (en) | Ship noise reduction system design method, system, medium and terminal based on acoustic environment simulation | |
| CN116661135A (en) | Wavefront shaping method, apparatus, computer device and storage medium | |
| CN118657055A (en) | Laser design method and system based on neural network | |
| CN111062159B (en) | Virtual test method for unmanned aerial vehicle | |
| CN115587480A (en) | Digital simulation method and digital simulation device | |
| CN114567007A (en) | Opportunity constraint-based static voltage stability probability evaluation method for power system | |
| CN113033062A (en) | Wind generating set bearing block optimization design method, storage medium and bearing block | |
| Kamiya et al. | Theoretical proof of edge search strategy applied to power plant start-up scheduling | |
| de Carvalho et al. | Application of the Quasi-Newton in interior point methods for solving the predispatch problem | |
| Zhelyazov et al. | Algorithms for System Identification | |
| CN118797821B (en) | Vehicle energy absorption box preparation method and device, electronic equipment and storage medium | |
| CN118734458B (en) | Manifold nerve operator construction method based on regularized Laplacian basis |
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 |