Electron gun focus stabilizing and current stabilizing method, system and equipment based on self-adaptive control and medium

Abstract

The invention discloses an adaptive control-based electron gun focus stabilization and current stabilization method, system and device, and a medium. The method comprises the following steps: S1, constructing a deep learning model for simulating a relationship between control parameters of an electron gun and electron beam current; s2, inputting control parameters of the electron gun into the deep learning model, and outputting predicted electron beam current; s3, a back propagation algorithm and an optimizer are adopted to update parameters of the deep learning model, a loss function is minimized, the loss function is used for measuring an error between the predicted electron beam current and the target electron beam current, and the performance of the deep learning model is optimized to obtain an optimized model; s4, determining a target function according to the stability of the electron beam current and the stability of the focus, and finding an optimal control parameter by adopting an optimization algorithm; and S5, adjusting the control parameters of the electron gun according to the optimal control parameters. According to the method, the electronic gun can efficiently adjust the focus position and the current output in a variable environment, and the target of current stabilization and focus stabilization is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of electron guns, and in particular to a method, system, device and medium for stabilizing focus and current of an electron gun based on adaptive control. Background Art

[0002] A microfocus electron gun is a highly precise electron beam source widely used in scanning electron microscopes (SEMs), transmission electron microscopes (TEMs), focused ion beams (FIBs), and other equipment, fulfilling the core tasks of high-resolution imaging and precision machining. Its main advantage lies in its ability to focus the electron beam within a very small area, thereby achieving extremely high resolution and precise imaging and machining. However, the stability of the microfocus electron gun, particularly the stability of the focus and current, directly affects imaging quality and machining accuracy.

[0003] In practical applications, despite the sophisticated design of microfocus electron guns, their focus and current stability are still subject to temperature fluctuations, power supply fluctuations, mechanical vibrations, and electromagnetic interference. These factors often cause the focus and current of the electron gun to be unstable in actual use, resulting in blurred or distorted images and even affecting the long-term reliability and performance of the system.

[0004] Optimization methods for current and focus stabilization typically employ techniques such as feedback control systems, adaptive control, and real-time adjustment to monitor and adjust system parameters in real time, thereby maintaining stable focus position and current. Closed-loop feedback control algorithms monitor focus and current status in real time, feeding the results back to the control system to automatically adjust system operating parameters. Common control strategies include PID control (proportional-integral-derivative control). Adaptive control automatically adjusts control parameters based on changes in system status. It is suitable for systems with large uncertainties and offers improved immunity to complex disturbances such as mechanical vibration and power supply fluctuations. Digital signal processing technology is used to process signals such as focus position and current in real time to optimize focus and current stability. Filtering and noise reduction are often used to process sensor signals and reduce noise and interference. Closed-loop feedback control algorithms suffer from limited control accuracy, long response times, and poor adaptability to disturbances. Adaptive algorithms are complex, difficult to adjust parameters, and computationally expensive, increasing the burden on the system. Digital signal processing requires significant computing and processing power, adding to system complexity and cost. Furthermore, it relies heavily on hardware, increasing the physical cost and design complexity of the system. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, system, device and medium for stabilizing focus and current of an electron gun based on adaptive control, so that the electron gun can efficiently adjust the focus position and current output in a changing environment to achieve the goal of stabilizing current and focus.

[0006] The present invention is achieved through the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for stabilizing focus and current of an electron gun based on adaptive control, comprising the following steps:

[0008] S1: Build a deep learning model to simulate the relationship between the control parameters of the electron gun and the electron beam current;

[0009] S2: Input the control parameters of the electron gun into the deep learning model and output the predicted electron beam current;

[0010] S3: Using a backpropagation algorithm and an optimizer to update the parameters of the deep learning model, minimize a loss function, where the loss function is used to measure the error between the predicted electron beam current and the target electron beam current, and optimize the performance of the deep learning model to obtain an optimized model;

[0011] S4: Determine the objective function based on the stability of the electron beam current and the stability of the focus, and use the optimization algorithm to find the optimal control parameters;

[0012] S5: Adjust the control parameters of the electron gun according to the optimal control parameters.

[0013] Furthermore, the deep learning model adopts the LSTM model, and the control parameters of the electron gun include the filament current I f , gate voltage U g , cathode voltage U c and anode high voltage U a .

[0014] Furthermore, the input of the LSTM model is the control parameter sequence X(t) of the electron gun = [I f (T),U g ,U c ,U a ], output the predicted electron beam current I a , the hidden state update calculation formula of the LSTM model is:

[0015] h t =LSTM(X(t),h t-1 );

[0016] Among them, X(t) is the input sequence, h t-1 is the hidden state at the previous moment, h t is the hidden state at the current moment.

[0017] Furthermore, the LSTM model includes a forget gate, an input gate, a memory unit, and an output gate. The mathematical expression of the forget gate is:

[0018] f t =σ(W f ·[Xt ,h t-1 ]+b f );

[0019] Among them, W f is the weight matrix of the forget gate, b f is the bias term, σ is the sigmoid activation function;

[0020] The mathematical expression of the input gate is:

[0021] i t =σ(W i ·[X t ,h t-1 ]+b i );

[0022]

[0023] Among them, i t is the output of the input gate, is the candidate memory cell state, W i is the weight of the input gate, W C is the memory weight of the candidate memory unit, b i is the bias term of the input gate, b c is the bias term of the candidate memory unit;

[0024] The mathematical expression of the memory unit is:

[0025]

[0026] Among them, C t is the current state of the memory unit, C t-1 is the state of the memory unit at the previous moment;

[0027] The mathematical expression of the output gate is:

[0028] o t =σ(W o ·[X t ,h t-1 ]+b o );

[0029] h t =o t tanh(C t );

[0030] Among them, t is the output of the output gate, W o is the candidate weight of the output gate, b o is the bias term;

[0031] The output of the LSTM model is mapped to the electron beam current I through a fully connected layer a (t), the electron beam current I a The mathematical expression of (t) is:

[0032] I a (t)=W1·h t +b1;

[0033] Among them, W1 is the fully connected weight matrix and b1 is the bias term.

[0034] Furthermore, the loss function The mathematical expression is:

[0035]

[0036] in, is the target electron beam current;

[0037] The objective function is determined based on the stability of the electron beam current and the stability of the focus. The formula is:

[0038]

[0039] in, is the ratio of the gate voltage to the cathode voltage, is the target grid voltage to cathode voltage ratio, λ1 and λ2 are weighting coefficients.

[0040] Furthermore, the optimization algorithm adopts a Bayesian optimization algorithm, which specifically includes:

[0041] The Gaussian process is used to model the objective function, and the Gaussian process describes x and the output objective function Assuming that the function f(x) obeys a Gaussian process, the prior distribution of the Gaussian process is:

[0042]

[0043] Where μ(x) is the mean function, k(x,x′) is the kernel function, and the kernel function includes the RBF kernel function. The formula is:

[0044]

[0045] in, is the signal variance, l is the length dimension;

[0046] The acquisition function is used to select the next test point for evaluation. The expected improvement formula of the acquisition function is:

[0047]

[0048] Among them, f * (x) is the current optimal objective function value, f(x) is the Gaussian process prediction value for the current point, Indicates expectations for the expression.

[0049] Furthermore, the specific method of using the optimization algorithm to find the optimal control parameters includes:

[0050] S41: Select initial data from prior knowledge;

[0051] S42: Construct a Gaussian process model, and use the initial data to train the Gaussian process model to obtain a proxy model;

[0052] S43: Calculate the acquisition function value of each possible point according to the proxy model;

[0053] S44: Select the point with the maximum acquisition function value to conduct an experiment and obtain a new objective function value;

[0054] S45: updating the Gaussian process model using the new objective function value;

[0055] S46: Repeat steps S43-S45 until the stop condition is met.

[0056] In a second aspect, an embodiment of the present invention provides an electron gun focus and current stabilization system based on adaptive control, comprising: a model construction module, a prediction module, a model optimization module, a control parameter optimization module, and an adjustment module, wherein the model construction module is configured to construct a deep learning model that simulates the relationship between the control parameters of the electron gun and the electron beam current;

[0057] The prediction module is configured to input the control parameters of the electron gun into the deep learning model and output the predicted electron beam current;

[0058] The model optimization module is configured to use a back-propagation algorithm and an optimizer to update the parameters of the deep learning model, minimize a loss function, wherein the loss function is used to measure the error between the predicted electron beam current and the target electron beam current, and optimize the performance of the deep learning model to obtain an optimized model;

[0059] The control parameter optimization module is configured to determine an objective function according to the stability of the electron beam current and the stability of the focus, and to find the optimal control parameters using an optimization algorithm;

[0060] The adjustment module is configured to adjust the control parameters of the electron gun according to the optimal control parameters.

[0061] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor, an input device, an output device and a memory, wherein the processor is connected to the input device, the output device and the memory respectively, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method described in the above embodiment.

[0062] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the method described in the above embodiment.

[0063] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0064] The embodiments of the present invention provide a method, system, device and medium for stabilizing focus and current of an electron gun based on adaptive control. The method simulates and predicts the dynamic behavior of the electron gun through a deep learning model, and adopts an optimization algorithm to adjust the relevant control parameters of the electron gun in real time, so that the electron gun can efficiently adjust the focal position and current output in a changing environment, thereby achieving the goal of stabilizing focus and current. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the examples. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. In the drawings:

[0066] Figure 1 A flowchart of a method for stabilizing focus and current of an electron gun based on adaptive control provided by the first embodiment of the present invention;

[0067] Figure 2 A structural block diagram of an electron gun focus and current stabilization system based on adaptive control provided in a second embodiment of the present invention;

[0068] Figure 3 This is a structural block diagram of an electronic device provided by the third embodiment of the present invention. DETAILED DESCRIPTION

[0069] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0070] Example 1

[0071] like Figure 1 FIG. 1 is a flowchart of a method for stabilizing focus and current of an electron gun based on adaptive control provided by a first embodiment of the present invention. The method includes the following steps:

[0072] S1: Build a deep learning model to simulate the relationship between the control parameters of the electron gun and the electron beam current;

[0073] S2: Input the control parameters of the electron gun into the deep learning model and output the predicted electron beam current;

[0074] S3: Using a backpropagation algorithm and an optimizer to update the parameters of the deep learning model, minimize a loss function, where the loss function is used to measure the error between the predicted electron beam current and the target electron beam current, and optimize the performance of the deep learning model to obtain an optimized model;

[0075] S4: Determine the objective function based on the stability of the electron beam current and the stability of the focus, and use the optimization algorithm to find the optimal control parameters;

[0076] S5: Adjust the control parameters of the electron gun according to the optimal control parameters.

[0077] In this embodiment, in order to achieve the electron beam current I a To achieve stability, we first need to build an LSTM (long short-term memory network) model to simulate the control parameters of the electron gun and the electron beam current I a The relationship between the control parameters includes: filament current I f , gate voltage U g , cathode voltage U c and anode high voltage U a When processing time series data, LSTM can capture the dynamic impact of control parameter changes on the electron beam current.

[0078] The input of the LSTM model is the control parameter sequence of the electron gun X(t) = [I f (T),U g ,U c ,U a ], the output is the predicted electron beam current I a , the basic process is as follows:

[0079] 1. LSTM hidden state update:

[0080] For the input sequence X(t) and the hidden state h at the previous moment t-1 ,LSTM calculates the hidden state h at the current moment through the following formula t :

[0081] ht =LSTM(X(t),h t-1 );

[0082] Among them, X(t) is the input sequence, h t-1 is the hidden state at the previous moment, h t is the hidden state at the current moment.

[0083] 2. Gating mechanism

[0084] LSTM regulates the flow of information through three gating mechanisms: input gate, forget gate, and output gate. The following is the detailed structure of LSTM:

[0085] The forget gate determines which information needs to be discarded from the memory cell. It uses a sigmoid activation function to output a value between 0 and 1, indicating the degree of retention or discard. The mathematical expression of the forget gate is:

[0086] f t =σ(W f ·[X t ,h t-1 ]+b f );

[0087] Among them, W f is the weight matrix of the forget gate, b f is the bias term, and σ is the sigmoid activation function.

[0088] The input gate determines which new information needs to be written to the memory cell. It consists of two parts: a sigmoid layer that determines which information to update, and a tanh layer that generates candidate information, representing possible new information. The mathematical expression of the input gate is:

[0089] i t =σ(W i ·[X t ,h t-1 ]+b i ),

[0090]

[0091] Among them, i t is the output of the input gate, is the candidate memory cell state, W i is the weight of the input gate, W C is the memory weight of the candidate memory unit, b i is the bias term of the input gate, b c is the bias term of the candidate memory unit.

[0092] The state of the memory unit is updated through the forget gate and the input gate. The mathematical expression of the memory unit is:

[0093]

[0094] Among them, C t is the current state of the memory unit, C t-1 is the state of the memory unit at the previous moment.

[0095] The output gate determines the hidden state h of the current time step t It consists of two parts: a sigmoid layer that determines what information to output, and a tanh layer that scales the memory cell state. The mathematical expression of the output gate is:

[0096] o t =σ(W o ·[X t ,h t-1 ]+b o ),

[0097] h t =o t tanh(C t ),

[0098] Among them: t is the output of the output gate, W o is the candidate weight of the output gate, b o is the bias term.

[0099] 3. Electron beam current prediction

[0100] The final output of LSTM is mapped to the electron beam current I through a fully connected layer a (t) above:

[0101] I a (t)=W1·h t +b1,

[0102] Among them, W1 is the fully connected weight matrix and b1 is the bias term.

[0103] 4. Loss Function and Training

[0104] In order to optimize the LSTM model, we need to define a loss function to measure the gap between the model's prediction results and the target. The mean square error is usually used to measure the prediction error of the electron beam current:

[0105]

[0106] in: is the target electron beam current, the loss function It is a measure of the error between the predicted and target values ​​of the electron beam current. The training process uses a backpropagation algorithm and an optimizer to update the parameters of the LSTM model, minimize the loss function, and optimize the model performance.

[0107] The LSTM model can address the vanishing gradient problem: through its memory cells and gating mechanism, it can better capture long-term dependencies. It also offers high flexibility: LSTM can process variable-length sequence data and excels in tasks such as natural language processing and time series prediction. The LSTM model can also simulate and predict the dynamic behavior of electron guns.

[0108] After training the LSTM model, we used a Bayesian optimization algorithm to adjust the control parameters to stabilize the electron beam current. Bayesian optimization is a global optimization method that approximates the target using a Gaussian process model and selects the next optimal point for evaluation using an acquisition function.

[0109] During the optimization process, an objective function needs to be defined The goal is to optimize both the stability of the electron beam current and the stability of the focus. The electron beam current stability is achieved by minimizing the electron beam current I a (t) and target value The difference between the grid voltage and the cathode voltage ensures the stability of the electron beam current. Focus stability is achieved by ensuring that the ratio of the grid voltage to the cathode voltage is U g / U c remains constant to ensure focus stability.

[0110] In Bayesian optimization it is defined as:

[0111]

[0112] in, is the target electron beam current, is the ratio of the gate voltage to the cathode voltage, is the target grid voltage to cathode voltage ratio, λ1 and λ2 are weighting coefficients that control the relative importance of electron beam current and focus stability.

[0113] Bayesian optimization uses Gaussian processes to model the objective function Gaussian process is a non-parametric probability model used to describe the relationship between input point x and output The relationship between , the prior distribution of the Gaussian process is:

[0114]

[0115] Where: μ(x) is the mean function, which represents the expected value of the function f(x) and is usually set to zero. k(x, x′) is the covariance function (or kernel function), which represents the correlation between the function f(x) at two input points x and x′. Common kernel functions include the RBF kernel function (radial basis function), which is mathematically expressed as:

[0116]

[0117] in, is the signal variance, which controls the amplitude of the function. l is the length scale, which controls the smoothness of the function.

[0118] Bayesian optimization uses an acquisition function to select the next experimental point for evaluation. The acquisition function is used to select the next evaluation point in each iteration to balance exploration and exploitation. Exploration: Sampling is performed in areas with large uncertainty in the objective function to discover potential global optimal solutions. Exploitation: Sampling is performed in areas with high estimated mean values ​​of the objective function to approximate the currently known optimal solution. Common acquisition functions include expected improvement and confidence upper bound. In this embodiment, expected improvement is used, and its formula is:

[0119]

[0120] Among them, f * (x) is the current optimal objective function value, f(x) is the Gaussian process prediction value for the current point, Indicates expectations for the expression.

[0121] By selecting the point with the largest expected improvement, Bayesian optimization can effectively explore the unknown area and find the optimal control parameters. The specific method of using the optimization algorithm to find the optimal control parameters is as follows:

[0122] S41: Initialization: Select the initial experimental points based on prior knowledge.

[0123] S42: Constructing a Gaussian process model: Using the initial data to train the Gaussian process model, we can obtain a proxy model. A surrogate model is a mathematical model used to approximate the target function.

[0124] S43: Calculate acquisition function: Calculate the acquisition function value of each possible point based on the current proxy model.

[0125] S44: Select the next experimental point: select the point with the maximum acquisition function value to conduct the experiment and obtain a new objective function value.

[0126] S45: Update model: get new objective function value through experiment Update the Gaussian process model.

[0127] S46: Repeat S43-S45 until the stop condition is met.

[0128] Bayesian optimization establishes the relationship between control parameters and electron beam current through Gaussian process regression, and guides the optimization process through acquisition function, so that the system can achieve the goal of stable flow and stable focus.

[0129] An embodiment of the present invention provides an electron gun focus and current stabilization method based on adaptive control. The method simulates and predicts the dynamic behavior of the electron gun through a deep learning model, and uses an optimization algorithm to adjust the relevant control parameters of the electron gun in real time, so that the electron gun can efficiently adjust the focal position and current output in a changing environment, thereby achieving the goal of stabilizing the current and focusing.

[0130] Example 2

[0131] like Figure 2 As shown, another embodiment of the present invention provides an electron gun focus and current stabilization system based on adaptive control, which is used to implement the electron gun focus and current stabilization method based on adaptive control described in the first embodiment. The system includes: a model construction module, a prediction module, a model optimization module, a control parameter optimization module and an adjustment module. The model construction module is configured to construct a deep learning model that simulates the relationship between the control parameters of the electron gun and the electron beam current;

[0132] The prediction module is configured to input the control parameters of the electron gun into the deep learning model and output the predicted electron beam current;

[0133] The model optimization module is configured to use a back-propagation algorithm and an optimizer to update the parameters of the deep learning model, minimize a loss function, and optimize the performance of the deep learning model to obtain an optimized model.

[0134] The control parameter optimization module is configured to determine an objective function based on the stability of the electron beam current and the stability of the focus, and to find the optimal control parameters using an optimization algorithm;

[0135] The adjustment module is configured to adjust the control parameters of the electron gun according to the optimal control parameters.

[0136] An electron gun focus and current stabilization system based on adaptive control provided in an embodiment of the present invention is based on the same inventive concept and has the same beneficial effects as the electron gun focus and current stabilization method based on adaptive control described in the first embodiment, and will not be described in detail here.

[0137] Example 3

[0138] like Figure 3As shown, a structural block diagram of an electronic device according to another embodiment of the present invention is shown, the device includes a processor, an input device, an output device and a memory, the processor is connected to the input device, the output device and the memory respectively, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method described in the first embodiment above.

[0139] It should be understood that in the embodiments of the present invention, the processor referred to may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0140] Input devices may include a touchpad, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and output devices may include a display (LCD, etc.), a speaker, etc.

[0141] The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store information about the device type.

[0142] In specific implementations, the processor, input device, and output device described in the embodiments of the present invention can execute the implementation methods described in the method embodiments provided by the embodiments of the present invention, and can also execute the implementation methods of the system embodiments described in the embodiments of the present invention, which will not be repeated here.

[0143] Example 4

[0144] Another embodiment of the present invention further provides a computer-readable storage medium storing a computer program. The computer program includes program instructions. When executed by a processor, the program instructions enable the processor to execute the method described in the first embodiment.

[0145] The computer-readable storage medium may be the internal storage unit of the terminal described in the aforementioned embodiment, such as the hard disk or memory of the terminal. The computer-readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Furthermore, the computer-readable storage medium may also include both the internal storage unit of the terminal and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the terminal. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0146] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the composition and steps of each example according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0147] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the terminals and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0149] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for stabilizing focus and current of an electron gun based on adaptive control, characterized in that: The following steps are involved: S1: Build a deep learning model to simulate the relationship between the control parameters of the electron gun and the electron beam current; S2: Input the control parameters of the electron gun into the deep learning model and output the predicted electron beam current; S3: Using a backpropagation algorithm and an optimizer to update the parameters of the deep learning model, minimize a loss function, where the loss function is used to measure the error between the predicted electron beam current and the target electron beam current, and optimize the performance of the deep learning model to obtain an optimized model; S4: Determine the objective function based on the stability of the electron beam current and the stability of the focus, and use the optimization algorithm to find the optimal control parameters; S5: Adjust the control parameters of the electron gun according to the optimal control parameters.

2. The method for stabilizing focus and current of an electron gun based on adaptive control according to claim 1, characterized in that: The deep learning model adopts the LSTM model, and the control parameters of the electron gun include the filament current I f , gate voltage U g , cathode voltage U c and anode high voltage U a .

3. The method for stabilizing focus and current of an electron gun based on adaptive control according to claim 2, characterized in that: The input of the LSTM model is the control parameter sequence of the electron gun X(t) = [I f (T),U g ,U c ,U a ], output the predicted electron beam current I a , the hidden state update calculation formula of the LSTM model is: h t =LSTM(X(t),h t-1 ); Among them, X(t) is the input sequence, h t-1 is the hidden state of the previous moment, h t is the hidden state at the current moment.

4. The method for stabilizing focus and current of an electron gun based on adaptive control according to claim 3, characterized in that: The LSTM model includes a forget gate, an input gate, a memory unit and an output gate. The mathematical expression of the forget gate is: f t =σ(W f ·[X t ,h t-1 ]+b f ); Among them, W f is the weight matrix of the forget gate, b f is the bias term, σ is the sigmoid activation function; The mathematical expression of the input gate is: i t =σ(W i ·[X t ,h t-1 ]+b i ); Among them, i t is the output of the input gate, is the candidate memory cell state, W i is the weight of the input gate, W C is the memory weight of the candidate memory unit, b i is the bias term of the input gate, b c is the bias term of the candidate memory unit; The mathematical expression of the memory unit is: Among them, C t is the current state of the memory unit, C t-1 is the state of the memory unit at the previous moment; The mathematical expression of the output gate is: the t =σ(W o ·[X t ,h t-1 ]+b o ); h t =o t ·tanh(C t ); Among them, t is the output of the output gate, W o is the candidate weight of the output gate, b o is the bias term; The output of the LSTM model is mapped to the electron beam current I through a fully connected layer a (t), the electron beam current I a The mathematical expression of (t) is: I a (t)=W1·h t +b1; Among them, W1 is the fully connected weight matrix and b1 is the bias term.

5. The method for stabilizing focus and current of an electron gun based on adaptive control according to claim 4, characterized in that: The loss function The mathematical expression is: in, is the target electron beam current; The objective function is determined based on the stability of the electron beam current and the stability of the focus. The formula is: in, is the ratio of the gate voltage to the cathode voltage, is the target grid voltage to cathode voltage ratio, λ1 and λ2 are weighting coefficients.

6. The method for stabilizing focus and current of an electron gun based on adaptive control according to claim 5, characterized in that: The optimization algorithm adopts the Bayesian optimization algorithm, which specifically includes: The Gaussian process is used to model the objective function, and the Gaussian process describes x and the output objective function Assuming that the function f(x) obeys a Gaussian process, the prior distribution of the Gaussian process is: Among them, μ(x) is the mean function, which represents the expected value of the function f(x), and k(x, x′) is the kernel function, which represents the correlation between the function f(x) at two input points x and x′. The kernel function includes the RBF kernel function. The mathematical expression is: in, is the signal variance, l is the length dimension; The acquisition function is used to select the next test point for evaluation. The expected improvement formula of the acquisition function is: Among them, f * (x) is the current optimal objective function value, f(x) is the Gaussian process prediction value for the current point, Indicates expectations for the expression.

7. The method for stabilizing focus and current of an electron gun based on adaptive control according to claim 6, characterized in that: The specific method of using the optimization algorithm to find the optimal control parameters includes: S41: Select initial data from prior knowledge; S42: Construct a Gaussian process model, and use the initial data to train the Gaussian process model to obtain a proxy model; S43: Calculate the acquisition function value of each possible point according to the proxy model; S44: Select the point with the maximum acquisition function value to conduct an experiment and obtain a new objective function value; S45: updating the Gaussian process model using the new objective function value; S46: Repeat steps S43-S45 until the stop condition is met.

8. An electron gun focus and current stabilization system based on adaptive control, characterized in that: include: a model building module, a prediction module, a model optimization module, a control parameter optimization module, and an adjustment module, wherein the model building module is configured to build a deep learning model that simulates the relationship between the control parameters of the electron gun and the electron beam current; The prediction module is configured to input the control parameters of the electron gun into the deep learning model and output the predicted electron beam current; The model optimization module is configured to use a back-propagation algorithm and an optimizer to update the parameters of the deep learning model, minimize a loss function, wherein the loss function is used to measure the error between the predicted electron beam current and the target electron beam current, and optimize the performance of the deep learning model to obtain an optimized model; The control parameter optimization module is configured to determine an objective function according to the stability of the electron beam current and the stability of the focus, and to find the optimal control parameters using an optimization algorithm; The adjustment module is configured to adjust the control parameters of the electron gun according to the optimal control parameters.

9. An electronic device comprising a processor, an input device, an output device, and a memory, wherein the processor is connected to the input device, the output device, and the memory respectively, and the memory is used to store a computer program, wherein the computer program comprises program instructions, and wherein: The processor is configured to call the program instructions and execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions. When the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 7.

Images