Self-optimization control method and system for mode locking state of solid laser

Abstract

The invention relates to a self-optimization control method and system for a mode-locked state of a solid laser. Firstly, an initial data set containing at least one of pulse width, frequency, temperature and vibration data is obtained, then the initial data set is preprocessed, and a mode locking stability feature set is extracted from the initial data set. And then, inputting data in the mode locking stability feature set into a support vector machine model so as to obtain a mode locking strength predicted value. And then, calculating a deviation value between the mode locking strength predicted value and the real-time data, generating an optimization decision value by using Kalman filtering, and if the deviation between the optimization decision value and the real-time data exceeds a preset threshold value, adjusting operation parameters. And finally, inputting the adjusted operation parameters into a mapping relation model optimized by using a gradient descent algorithm, and further generating a dynamic power configuration scheme. According to the method, the stability and automatic regulation and control of the mode locking state are realized, the efficient utilization of resources is realized, and the performance and adaptability of the solid laser are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the field of laser technology, and particularly relates to a method and system for self-optimizing control of the mode-locking state of a solid-state laser. Background Art

[0002] With the development of laser technology, there has emerged a technology for self-optimizing control of the mode-locking state of solid-state lasers. As a key means of generating ultrashort laser pulses, the mode-locking technology is widely used in the field of solid-state lasers. However, in actual operation, solid-state lasers are susceptible to various factors such as temperature, vibration, and pump source stability, resulting in an unstable mode-locking state. An unstable mode-locking state not only reduces the quality of laser pulses, limits their applications in material processing, medical treatment, scientific research, etc., but also causes fluctuations in the output power of the laser, affecting the overall performance of the system. At the same time, traditional mode-locking control methods mostly rely on manual experience and preset parameters, and it is difficult to adapt to complex and changing working environments in real time and dynamically. Summary of the Invention

[0003] Based on this, in view of the above technical problems, it is necessary to provide a method and system for self-optimizing control of the mode-locking state of a solid-state laser that can achieve stable and automatic regulation of the mode-locking state.

[0004] In a first aspect, the present application provides a method for self-optimizing control of the mode-locking state of a solid-state laser, including:

[0005] Obtaining an initial data set; extracting a mode-locking stability feature set after preprocessing the initial data set; the initial data set includes at least one of pulse width, frequency, temperature, and vibration data.

[0006] Inputting the data in the mode-locking stability feature set into a support vector machine model to obtain a predicted value of the mode-locking intensity.

[0007] Calculating the deviation between the predicted value of the mode-locking intensity and the real-time data and generating an optimization decision value using a Kalman filter; if the deviation between the optimization decision value and the real-time data exceeds a preset threshold, adjust the operating parameters.

[0008] Inputting the adjusted operating parameters into a mapping relationship model optimized by a gradient descent algorithm to generate a dynamic power configuration scheme.

[0009] In one embodiment, inputting the data in the mode-locking stability feature set into a support vector machine model to obtain a predicted value of the mode-locking intensity includes:

[0010] Obtaining the time-frequency domain hybrid data in the mode-locking stability feature set; the time-frequency domain hybrid data includes the amplitude jitter of the pulse sequence and the spectral signal-to-noise ratio parameter.

[0011] Perform feature dimensionality reduction processing based on principal component analysis on the time-domain and frequency-domain mixed data to obtain a set of low-dimensional feature vectors.

[0012] Input the set of low-dimensional feature vectors into a pre-trained support vector machine model to obtain an error distribution matrix; the error distribution matrix includes the residual sequence between the preliminary prediction value and the measured value.

[0013] Use the sliding window mechanism to perform dynamic threshold segmentation on the residual sequence to obtain abnormal fluctuation interval marking data.

[0014] Align the abnormal fluctuation interval marking data with the time-domain and frequency-domain mixed data in terms of timestamps to generate a mode-locking stability sample set with abnormal annotations.

[0015] Update the training weight parameters of the support vector machine model according to the mode-locking stability sample set with abnormal annotations to obtain a trained support vector machine model.

[0016] Input the set of low-dimensional feature vectors into the trained support vector machine model to obtain the predicted value of the mode-locking intensity.

[0017] In one embodiment, calculating the deviation between the predicted value of the mode-locking intensity and the real-time data and generating an optimized decision value using Kalman filtering includes:

[0018] Obtain the time series difference between the predicted value of the mode-locking intensity and the real-time data.

[0019] Update to obtain a covariance matrix according to the time series difference and using the state estimation equation of the Kalman filter; the covariance matrix is used to calculate the confidence interval of the optimized decision value.

[0020] Judge the relative distance between the confidence interval and the mean value of the real-time data based on a preset threshold. If it exceeds the preset threshold, generate a parameter adjustment signal; the parameter adjustment signal includes a target parameter identifier and correction amount data.

[0021] Input the parameter adjustment signal into a dynamic threshold decision model to obtain an optimized decision value.

[0022] In one embodiment, updating to obtain a covariance matrix according to the time series difference and using the state estimation equation of the Kalman filter includes:

[0023] Obtain difference data based on the time series data and use the Kalman filter to filter the difference data to obtain an initial state estimation value.

[0024] Based on the initial state estimation value, combine with the time series to judge the change trend of the state estimation to obtain an estimation result.

[0025] Among them, if the change trend exceeds the preset threshold, the state estimation equation of the Kalman filter is used to iteratively adjust the estimation result to obtain an optimized state estimation value.

[0026] The stability of the verification result is analyzed based on the optimized state estimation value to obtain the final estimation result.

[0027] Key features are extracted from the final estimation result and a correction value of the covariance matrix is generated using matrix operations to obtain the adjusted covariance matrix data.

[0028] Based on the adjusted covariance matrix data and the difference data of the time series, the deviation degree of the data input is judged to obtain the deviation correction result.

[0029] The covariance matrix is extracted from the deviation correction result to obtain the updated covariance matrix.

[0030] In one of the embodiments, the adjusted operating parameters are input into the mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme, including:

[0031] The adjusted operating parameters are input into the dynamic mapping model optimized based on the gradient descent to obtain the initial power configuration matrix; the operating parameter set includes device state variables and environmental monitoring indicators.

[0032] Based on the initial power configuration matrix, the eigenvector related to the real-time load in the matrix is extracted to obtain the load eigenvector.

[0033] The load eigenvector is input into the preset constraint conditions and the dynamic balance equation to obtain the optimized power sequence.

[0034] According to the optimized power sequence, the activation layer parameters of the dynamic mapping model are updated to generate the updated power configuration matrix.

[0035] The power configuration matrix is feature-matched with the historical operation records to screen out the configuration schemes with a matching degree higher than the threshold to obtain the dynamic power configuration scheme; the dynamic power configuration scheme includes power values and execution time interval parameters.

[0036] In one of the embodiments, the constraint conditions include:

[0037]

[0038]

[0039] Among them, L represents the load eigenvector, n represents the vector dimension, α i represents the weight coefficient, V iLet \(\boldsymbol{\alpha}\) denote the feature components, \(\lambda\) denote the constraint parameter, \(C\) denote the constraint condition, \(Q\) denote the time-domain constraint function, \(T\) denote the time period, \(\gamma(t)\) denote the time-varying coefficient, \(R(t)\) denote the response function, \(\eta\) denote the regularization parameter, and \(W\) denote the weight matrix.

[0040] In a second aspect, the present application also provides a self-optimizing control system for the mode-locking state of a solid-state laser. The system includes:

[0041] A data processing module, which is used to obtain an initial data set; and is also used to extract a mode-locking stability feature set after preprocessing the initial data set; the initial data set includes at least one of pulse width, frequency, temperature, and vibration data.

[0042] A parameter optimization module, which is used to input the data in the mode-locking stability feature set into a support vector machine model to obtain a predicted value of the mode-locking intensity; and is also used to calculate the deviation between the predicted value of the mode-locking intensity and the real-time data and generate an optimization decision value by using Kalman filtering; if the deviation between the optimization decision value and the real-time data exceeds a preset threshold, the operating parameters are adjusted.

[0043] A dynamic configuration module, which is used to input the adjusted operating parameters into a mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme.

[0044] In a third aspect, the present application also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the foregoing method is implemented.

[0045] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the foregoing method is implemented.

[0046] For the foregoing self-optimizing control method and system for the mode-locking state of a solid-state laser, first, an initial data set including at least one of pulse width, frequency, temperature, and vibration data is obtained. Then, the initial data set is preprocessed and a mode-locking stability feature set is extracted therefrom. Subsequently, the data in the mode-locking stability feature set is input into a support vector machine model to obtain a predicted value of the mode-locking intensity. After that, the deviation between the predicted value of the mode-locking intensity and the real-time data is calculated, and an optimization decision value is generated by using Kalman filtering. If the deviation between this optimization decision value and the real-time data exceeds a preset threshold, the operating parameters are adjusted. Finally, the adjusted operating parameters are input into a mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme. The foregoing method and system achieve the stable and automatic regulation of the mode-locking state, realize the efficient utilization of resources, and greatly improve the performance and adaptability of the solid-state laser. Description of the Drawings

[0047] To more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the accompanying drawings required for the description of the embodiments or related technologies. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0048] Figure 1 It is a flowchart of the self-optimizing control method for the mode-locking state of a solid-state laser provided by an embodiment of the present invention.

[0049] Figure 2 It is a flowchart of inputting the data in the mode-locking stability feature set into a support vector machine model to obtain the predicted value of the mode-locking intensity provided by an embodiment of the present invention.

[0050] Figure 3 It is a flowchart of inputting the adjusted operating parameters into a mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme provided by an embodiment of the present invention.

[0051] Figure 4 It is a structural block diagram of the self-optimizing control system for the mode-locking state of a solid-state laser provided by an embodiment of the present invention. Specific Embodiments

[0052] In order to make the objectives, technical solutions, and advantages of the present application clearer, the following further details the present application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] In one of the embodiments, as Figure 1 shown, the present application provides a self-optimizing control method for the mode-locking state of a solid-state laser, which may include the following steps:

[0054] Step S101, obtain an initial data set; extract a mode-locking stability feature set after preprocessing the initial data set; the initial data set includes at least one of pulse width, frequency, temperature, and vibration data.

[0055] First, through professional sensing devices, collect data covering at least one type of data among pulse width, frequency, temperature, and vibration data to construct an initial data set. These data reflect the real-time operating state of the solid-state laser from multiple dimensions. For example, the pulse width and frequency are directly related to the characteristics of laser pulses, while the temperature and vibration affect the stability and optical performance of the laser cavity. Subsequently, perform preprocessing on the initial data set, including data cleaning to remove noise and outliers, data normalization to unify the dimension, etc., to improve the data quality. Based on the preprocessed data, use specific algorithms and technologies to extract a feature set that can characterize the mode-locking stability.

[0056] In step S102, the data in the mode-locking stability feature set is input into the support vector machine model to obtain the mode-locking intensity prediction value.

[0057] The data in the mode-locking stability feature set is passed as input to a pre-trained support vector machine model. During the training phase, this model learns the complex relationship between features and mode-locking intensity through a large amount of labeled historical data, thereby constructing an accurate mapping function. When new mode-locking stability feature data is received, the model predicts the mode-locking intensity based on the learned mapping relationship and outputs the mode-locking intensity prediction value. This prediction value can provide an important reference for subsequent judgments on whether the mode-locking state is stable and whether parameter adjustment is required, and helps to achieve self-optimizing control of the mode-locking state of the solid-state laser.

[0058] In step S103, calculate the deviation between the mode-locking intensity prediction value and the real-time data and generate an optimized judgment value using Kalman filtering; if the deviation between the optimized judgment value and the real-time data exceeds the preset threshold, adjust the operating parameters.

[0059] By comparing the mode-locking intensity prediction value with the real data obtained through real-time monitoring, calculate the deviation between the two. The acquisition of the deviation can intuitively reflect the degree of deviation between the model prediction value and the actual situation. To effectively suppress data noise and improve the accuracy and reliability of the judgment, the Kalman filtering algorithm is introduced to process the deviation data. Based on the Kalman filtering algorithm, the deviation data is operated on to generate an optimized judgment value. This optimized judgment value comprehensively reflects the dynamic changes of the system. Subsequently, compare the optimized judgment value with the real-time data. If the deviation between the two exceeds the preset threshold, it indicates that the mode-locking state of the solid-state laser has a large deviation and is in an unstable state. To ensure the continuous and stable operation of the laser, it is necessary to make targeted adjustments to its operating parameters, such as pump power, cavity mirror position, etc., so as to achieve self-optimizing control of the mode-locking state and ensure the stability and reliability of the laser output performance.

[0060] In step S104, input the adjusted operating parameters into the mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration plan.

[0061] Input the corrected operating parameters, such as key parameters reflecting the operating state of the laser, such as pump current, cavity mirror angle, etc., into the mapping relationship model optimized by the gradient descent algorithm. After being optimized by the gradient descent algorithm, this mapping relationship model can deeply explore the potential connection between the operating parameters and the required power configuration. Based on the input adjusted operating parameters, the model performs a series of complex calculations and analyses under the framework of preset constraint conditions and objective functions. Based on the operation results, a dynamic power configuration scheme that fits the real-time operating conditions of the laser is generated. This scheme is not statically set, but is adjusted in real time with the change of the laser operating parameters, which can achieve precise control of the power of the solid-state laser, while ensuring the stability of the mode-locking state, greatly improving the energy efficiency of the laser, reducing energy consumption, extending the service life of the equipment, and making the laser better adapt to the needs of different working scenarios.

[0062] For the above self-optimizing control method for the mode-locking state of the solid-state laser, first obtain an initial data set including at least one of pulse width, frequency, temperature, and vibration data, and then preprocess the initial data set and extract the mode-locking stability feature set from it. Subsequently, input the data in the mode-locking stability feature set into the support vector machine model to obtain the mode-locking intensity prediction value. Then, calculate the deviation between the mode-locking intensity prediction value and the real-time data, and use the Kalman filter to generate an optimization decision value. If the deviation between this optimization decision value and the real-time data exceeds the preset threshold, adjust the operating parameters. Finally, input the adjusted operating parameters into the mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme. The above method realizes the stability and automatic control of the mode-locking state, realizes the efficient utilization of resources, and greatly improves the performance and adaptability of the solid-state laser.

[0063] In one embodiment, as Figure 2 shown, inputting the data in the mode-locking stability feature set into the support vector machine model to obtain the mode-locking intensity prediction value may include the following steps:

[0064] Step S201, obtain the time-frequency domain hybrid data in the mode-locking stability feature set; the time-frequency domain hybrid data includes the amplitude jitter of the pulse sequence and the spectral signal-to-noise ratio parameter.

[0065] Step S202, perform feature dimensionality reduction processing on the time-frequency domain hybrid data based on principal component analysis to obtain a low-dimensional feature vector set.

[0066] Step S203, input the low-dimensional feature vector set into the pre-trained support vector machine model to obtain an error distribution matrix; the error distribution matrix includes the residual sequence of the preliminary prediction value and the measured value.

[0067] Step S204, use the sliding window mechanism to perform dynamic threshold segmentation on the residual sequence to obtain the abnormal fluctuation interval marking data.

[0068] Step S205: Align the time stamps of the abnormal fluctuation interval marking data and the time-frequency domain mixed data to generate a mode-locking stability sample set with abnormal annotations.

[0069] Step S206: Update the training weight parameters of the support vector machine model according to the mode-locking stability sample set with abnormal annotations to obtain a trained support vector machine model.

[0070] Step S207: Input the low-dimensional feature vector set into the trained support vector machine model to obtain the predicted value of the mode-locking intensity of the mode-locking intensity.

[0071] Specifically, first, extract the time-frequency domain mixed data containing the pulse sequence amplitude jitter and the spectral signal-to-noise ratio parameters from the mode-locking stability feature set. Then, to reduce the data dimension, reduce the computational complexity, and avoid overfitting, use the principal component analysis method to perform feature dimensionality reduction on the mixed data to obtain a low-dimensional feature vector set. Subsequently, input this low-dimensional feature vector set into the pre-trained support vector machine model to obtain an error distribution matrix containing the residual sequence of the preliminary predicted value and the measured value. After that, use the sliding window mechanism to perform dynamic threshold segmentation on the residual sequence to mark the abnormal fluctuation interval. Align the time stamps of the abnormal fluctuation interval marking data and the time-frequency domain mixed data to generate a mode-locking stability sample set with abnormal annotations. Based on this sample set, update the training weight parameters of the support vector machine model to obtain a trained support vector machine model. Finally, input the low-dimensional feature vector set into the trained model again to obtain the predicted value of the mode-locking intensity.

[0072] In this embodiment, feature dimensionality reduction is performed through principal component analysis, which can effectively reduce data redundancy, improve computational efficiency, retain key information at the same time, and enhance the generalization ability of the model. The sliding window mechanism is used to mark the abnormal fluctuation interval, and the support vector machine model is updated in combination with abnormal annotations, enabling the model to better adapt to abnormal situations in the data and improving the accuracy and robustness of the mode-locking intensity prediction. Through this refined processing and optimization, the mode-locking state of the solid-state laser can be more accurately grasped, providing a reliable basis for subsequent self-optimization control, and helping to improve the performance and stability of the solid-state laser.

[0073] In one of the embodiments, calculating the deviation between the predicted value of the mode-locking intensity and the real-time data and generating an optimized decision value using the Kalman filter may include the following steps:

[0074] Step S301: Obtain the time series difference between the predicted value of the mode-locking intensity and the real-time data.

[0075] Step S302: Update and obtain the covariance matrix according to the time series difference and using the state estimation equation of the Kalman filter; the covariance matrix is used to calculate the confidence interval of the optimized decision value.

[0076] Step S303, judging the relative distance between the confidence interval and the mean of the real-time data based on a preset threshold, and generating a parameter adjustment signal if it exceeds the preset threshold; the parameter adjustment signal includes a target parameter identifier and correction amount data.

[0077] Step S304: input the dynamic threshold determination model according to the parameter adjustment signal to obtain an optimized determination value.

[0078] First, by calculating the difference between the predicted value of the locking intensity and the real-time data, the time series difference between the two is obtained. This difference intuitively reflects the deviation between the predicted value and the actual operating data. Then, with the help of the state estimation equation of the Kalman filter, the covariance matrix is updated according to the time series difference. The covariance matrix can provide key data support for calculating the confidence interval of the optimized judgment value, thereby evaluating the reliability of the predicted data. Subsequently, the relative distance between the confidence interval and the mean of the real-time data and the preset threshold are compared. Once the relative distance exceeds the preset threshold, a parameter adjustment signal is generated, which contains the target parameter identifier and correction data to clarify the parameters that need to be adjusted and the adjustment range. Finally, the parameter adjustment signal is input into the dynamic threshold judgment model to obtain the optimized judgment value.

[0079] On the one hand, the use of Kalman filter effectively suppresses the noise interference in the data, accurately estimates the system state, makes the calculation of the covariance matrix more accurate, and greatly improves the reliability of subsequent judgments. On the other hand, through the dynamic threshold judgment model combined with the judgment mechanism of the preset threshold, it is possible to capture abnormal changes in the mode-locking state in a timely manner, generate parameter adjustment signals in a targeted manner, and realize real-time and precise control of the operating parameters of the solid-state laser. This not only ensures the stability of the mode-locking state of the solid-state laser and improves the quality of laser output, but also enhances the adaptability of the laser under complex working conditions and extends the service life of the equipment.

[0080] In one embodiment, the covariance matrix is updated based on the time series difference and using the state estimation equation of the Kalman filter, which may include the following steps:

[0081] Step S401, obtaining difference data based on time series data and filtering the difference data using a Kalman filter to obtain an initial state estimate.

[0082] Step S402, judging the change trend of the state estimation based on the initial state estimation value and the time series, and obtaining the estimation result.

[0083] If the change trend exceeds the preset threshold, the state estimation equation of the Kalman filter is used to iteratively adjust the estimation result to obtain the optimized state estimation value.

[0084] Step S403, analyze the stability of the verification result based on the optimized state estimation value to obtain the final estimation result.

[0085] Step S404, extract key features from the final estimation result and generate a correction value of the covariance matrix through matrix operations to obtain the adjusted covariance matrix data.

[0086] Step S405, based on the adjusted covariance matrix data and combined with the difference data of the time series, judge the deviation degree of the data input to obtain the deviation correction result.

[0087] Step S406, extract the covariance matrix from the deviation correction result to obtain the updated covariance matrix.

[0088] First, extract the difference data based on the time series data, and use the Kalman filter to perform filtering processing on it to obtain the initial state estimation value. This step effectively removes the noise in the data through the characteristics of the Kalman filter, laying a good foundation for subsequent state estimation. Then, combined with the time series, judge the change trend of the state estimation based on the initial state estimation value, and then obtain the estimation result. Once the change trend exceeds the preset threshold, use the state estimation equation of the Kalman filter to iteratively adjust the estimation result to obtain the optimized state estimation value. Subsequently, based on the optimized state estimation value, deeply analyze the stability of the verification result to determine the final estimation result. After that, extract the key features from the final estimation result, and generate a correction value of the covariance matrix through matrix operations to obtain the adjusted covariance matrix data. According to the adjusted covariance matrix data, combined with the difference data of the time series, judge the deviation degree of the data input to obtain the deviation correction result. Finally, extract the covariance matrix from the deviation correction result to obtain the updated covariance matrix.

[0089] On the one hand, the multiple applications of the Kalman filter can continuously and effectively suppress noise interference in complex time series data, improve the accuracy of state estimation, and ensure that the system's understanding of its own state is accurate and reliable. On the other hand, through multi-step analysis, judgment and adjustment, the dynamic optimization of the covariance matrix is realized. This not only enables the system to capture the deviation of data input more sensitively, correct it in a timely manner, enhances the stability and anti-interference ability of the system, but also provides more accurate data support for various decision-making and control operations based on the covariance matrix, significantly improving the operation performance and reliability of the entire system.

[0090] In one embodiment, as Figure 3 shown, input the adjusted operating parameters into the mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme, which may include the following steps:

[0091] Step S501: Input the adjusted operating parameters into the dynamic mapping model optimized by gradient descent to obtain the initial power configuration matrix. The set of operating parameters includes device state variables and environmental monitoring indicators.

[0092] Step S502: Extract the eigenvector related to the real-time load from the initial power configuration matrix to obtain the load eigenvector.

[0093] Step S503: Input the load eigenvector into the preset constraint conditions and dynamic balance equations to obtain the optimized power sequence.

[0094] Step S504: Update the activation layer parameters of the dynamic mapping model according to the optimized power sequence to generate the updated power configuration matrix.

[0095] Step S505: Perform feature matching between the power configuration matrix and the historical operation records to screen out the configuration schemes with a matching degree higher than the threshold to obtain the dynamic power configuration scheme. The dynamic power configuration scheme includes power values and execution time interval parameters.

[0096] First, input the adjusted operating parameters including device state variables and environmental monitoring indicators into the dynamic mapping model optimized by gradient descent to obtain the initial power configuration matrix. This step uses the gradient descent algorithm to optimize the model so that the model can better fit the relationship between the operating parameters and the power configuration. Then, extract the eigenvector related to the real-time load from the initial power configuration matrix to obtain the load eigenvector, which provides a key basis for accurately adjusting the power subsequently. Subsequently, substitute the load eigenvector into the preset constraint conditions and dynamic balance equations, and calculate to obtain the optimized power sequence to ensure that the power configuration meets the system operation requirements. After that, update the activation layer parameters of the dynamic mapping model according to the optimized power sequence to generate the updated power configuration matrix, enabling the model to adapt to real-time changes. Finally, perform feature matching between the power configuration matrix and the historical operation records, and screen out the configuration schemes with a matching degree higher than the threshold to form a dynamic power configuration scheme including power values and execution time interval parameters.

[0097] Through the dynamic mapping model optimized by gradient descent in this embodiment, the complex relationship between the operating parameters and the power configuration can be accurately captured, improving the accuracy and adaptability of the power configuration. Using the preset constraint conditions and dynamic balance equations to generate the optimized power sequence ensures the efficient use of resources while the power configuration meets the system operation requirements. The feature matching mechanism with the historical operation records enables the system to draw on past experience, further optimize the power configuration scheme, and enhance the stability and reliability of the system. The dynamic power configuration scheme can be adjusted in real time according to the device state and environmental changes, effectively reducing energy consumption, extending the service life of the device, and improving the operation efficiency and performance of the solid-state laser under different working conditions.

[0098] In one embodiment, the constraint conditions may include:

[0099]

[0100] where L represents the load feature vector, n represents the vector dimension, α i represents the weight coefficient, V i represents the feature component, λ represents the constraint parameter, C represents the constraint condition, Q represents the time-domain constraint function, T represents the time period, γ(t) represents the time-varying coefficient, R(t) represents the response function, η represents the regularization parameter, and W represents the weight matrix.

[0101] This embodiment improves the scientificity and reliability of the dynamic power configuration of the solid-state laser. Multiple elements such as the load feature vector and the vector dimension work together to provide comprehensive load information for the system. Through the integrated analysis of these elements, the power configuration can be accurately adjusted to avoid resource waste and abnormal equipment operation caused by over- or under-configuration. The time-varying coefficient and the time-domain constraint function enable the system to adapt to complex working conditions that change over time, ensuring the rationality of the power configuration at different times. The introduction of the response function and the regularization parameter not only ensures the response speed of the system but also prevents model overfitting, improving the generalization ability and stability of the model. The weight matrix flexibly adjusts the influence degree of each element, enhancing the adaptability of the constraint conditions. The power configuration scheme generated under these constraint conditions not only meets the real-time operation requirements of the solid-state laser but also improves the energy utilization efficiency, reduces the operation cost of the system, extends the overall service life of the equipment, and helps the solid-state laser to perform better in fields such as material processing and scientific experiments.

[0102] In one embodiment, as Figure 4 shown, the present application also provides a self-optimizing control system for the mode-locking state of a solid-state laser. The system may include:

[0103] A data processing module 601, configured to obtain an initial data set; and also configured to extract a mode-locking stability feature set after preprocessing the initial data set; the initial data set includes at least one of pulse width, frequency, temperature, and vibration data.

[0104] A parameter optimization module 602, configured to input the data in the mode-locking stability feature set into a support vector machine model to obtain a mode-locking intensity prediction value; and also configured to calculate the deviation between the mode-locking intensity prediction value and the real-time data and generate an optimization determination value using a Kalman filter; if the deviation between the optimization determination value and the real-time data exceeds a preset threshold, adjust the operating parameters.

[0105] A dynamic configuration module 603, configured to input the adjusted operating parameters into a mapping relationship model optimized using a gradient descent algorithm to generate a dynamic power configuration scheme.

[0106] The above self-optimizing control system for the mode-locking state of a solid-state laser includes a data processing module, a parameter optimization module, and a dynamic configuration module. The data processing module obtains an initial data set containing at least one of pulse width, frequency, temperature, and vibration data, preprocesses it, and extracts a mode-locking stability feature set; the parameter optimization module inputs the feature set data into a support vector machine model to obtain a predicted value of the mode-locking intensity, calculates the deviation between this predicted value and the real-time data, and uses Kalman filtering to generate an optimization decision value. If the deviation between this decision value and the real-time data exceeds a preset threshold, the operating parameters are adjusted; the dynamic configuration module inputs the adjusted operating parameters into a mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme, thereby realizing the self-optimizing control of the mode-locking state of the solid-state laser. The above system realizes the stable and automatic regulation of the mode-locking state, realizes the efficient utilization of resources, and greatly improves the performance and adaptability of the solid-state laser.

[0107] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown sequentially in the direction of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least some of the steps or stages in other steps or other steps.

[0108] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the self-optimizing control method and system for the mode-locking state of the solid-state laser as described above are implemented.

[0109] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0110] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the descriptions of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. A person of ordinary skill in the art can understand and implement it without creative efforts.

[0111] The above embodiments only represent several implementation manners of the embodiments of the present application. The descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A self-optimizing control method for the mode-locked state of a solid-state laser, characterized in that, The method includes: Obtain an initial data set; extract a mode-locking stability feature set after preprocessing the initial data set; the initial data set includes at least one of pulse width, frequency, temperature, and vibration data; Input the data in the mode-locking stability feature set into a support vector machine model to obtain a mode-locking intensity prediction value; Calculate the deviation between the mode-locking intensity prediction value and the real-time data and generate an optimized decision value using Kalman filtering; if the deviation between the optimized decision value and the real-time data exceeds a preset threshold, adjust the operating parameters; Input the adjusted operating parameters into a mapping relationship model optimized using the gradient descent algorithm to generate a dynamic power configuration scheme.

2. The method according to claim 1, characterized in that, The step of inputting the data in the mode-locking stability feature set into a support vector machine model to obtain a mode-locking intensity prediction value includes: Obtain the time-domain and frequency-domain hybrid data in the mode-locking stability feature set; the time-domain and frequency-domain hybrid data includes amplitude jitter of the pulse sequence and spectral signal-to-noise ratio parameters; Perform feature dimensionality reduction processing on the time-domain and frequency-domain hybrid data based on principal component analysis to obtain a low-dimensional feature vector set; Input the low-dimensional feature vector set into a pre-trained support vector machine model to obtain an error distribution matrix; the error distribution matrix includes a residual sequence of the preliminary prediction value and the measured value; Use a sliding window mechanism to perform dynamic threshold segmentation on the residual sequence to obtain abnormal fluctuation interval marking data; Align the time stamps of the abnormal fluctuation interval marking data and the time-domain and frequency-domain hybrid data to generate a mode-locking stability sample set with abnormal annotations; Update the training weight parameters of the support vector machine model according to the mode-locking stability sample set with abnormal annotations to obtain a trained support vector machine model; Input the low-dimensional feature vector set into the trained support vector machine model to obtain a mode-locking intensity prediction value of the mode-locking intensity.

3. The method according to claim 1, wherein The step of calculating the deviation between the mode-locking intensity prediction value and the real-time data and generating an optimized decision value using Kalman filtering includes: Obtain the time series difference between the mode-locking intensity prediction value and the real-time data; Update to obtain a covariance matrix according to the time series difference and using the state estimation equation of the Kalman filter; the covariance matrix is used to calculate the confidence interval of the optimized decision value; Based on a preset threshold, judge the relative distance between the confidence interval and the mean value of the real-time data. If it exceeds the preset threshold, generate a parameter adjustment signal; the parameter adjustment signal includes a target parameter identifier and correction amount data; Input the parameter adjustment signal into a dynamic threshold decision model to obtain an optimized decision value.

4. The method according to claim 3, characterized in that The step of updating to obtain a covariance matrix according to the time series difference and using the state estimation equation of the Kalman filter includes: Obtain difference data based on time series data and use the Kalman filter to filter the difference data to obtain an initial state estimation value; Based on the initial state estimation value, combine with the time series to judge the change trend of the state estimation to obtain an estimation result; Wherein, if the change trend exceeds a preset threshold, use the state estimation equation of the Kalman filter to iteratively adjust the estimation result to obtain an optimized state estimation value; Analyze the stability of the verification result based on the optimized state estimation value to obtain the final estimation result; Extract key features from the final estimation result and generate a corrected value of the covariance matrix using matrix operations to obtain the adjusted covariance matrix data; Based on the adjusted covariance matrix data and the difference data of the time series, judge the deviation degree of the data input to obtain the deviation correction result; Extract the covariance matrix from the deviation correction result to obtain the updated covariance matrix.

5. The method according to claim 1, wherein Inputting the adjusted operating parameters into the mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme, including: Input the adjusted operating parameters into the dynamic mapping model optimized by gradient descent to obtain the initial power configuration matrix; the set of operating parameters includes device state variables and environmental monitoring indicators; Extract the eigenvector related to the real-time load from the initial power configuration matrix to obtain the load eigenvector; Input the load eigenvector into the preset constraint conditions and the dynamic balance equation to obtain the optimized power sequence; Update the activation layer parameters of the dynamic mapping model according to the optimized power sequence to generate the updated power configuration matrix; Match the features of the power configuration matrix with the historical operation records to screen out the configuration schemes with a matching degree higher than the threshold to obtain the dynamic power configuration scheme; the dynamic power configuration scheme includes power values and execution time interval parameters.

6. The method according to claim 5, characterized in that, The constraint conditions include: Among them, L represents the load feature vector, n represents the vector dimension, α i represents the weight coefficient, V i represents the feature component, λ represents the constraint parameter, C represents the constraint condition, Q represents the time-domain constraint function, T represents the time period, γ(t) represents the time-varying coefficient, R(t) represents the response function, η represents the regularization parameter, and W represents the weight matrix.

7. Solid laser mode-locking state self-optimizing control system, characterized in that, The system includes: A data processing module, configured to obtain an initial data set; and also configured to extract a mode-locking stability feature set after preprocessing the initial data set; the initial data set includes at least one of pulse width, frequency, temperature, and vibration data; A parameter optimization module, configured to input the data in the mode-locking stability feature set into a support vector machine model to obtain a mode-locking intensity prediction value; and also configured to calculate the deviation amount between the mode-locking intensity prediction value and the real-time data and generate an optimized determination value using Kalman filtering; if the deviation between the optimized determination value and the real-time data exceeds a preset threshold, adjust the operating parameters; A dynamic configuration module, configured to input the adjusted operating parameters into the mapping relationship model optimized by the gradient descent algorithm to generate a dynamic power configuration scheme.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 6.

Images