An intelligent harmonic mode-locking method and system based on deep learning guided adaptive search

Abstract

The application discloses an intelligent harmonic mode-locked method and system based on deep learning guided adaptive search, and belongs to the technical field of ultrafast fiber lasers and automatic control. The method comprises the following steps: collecting spectrum signals, time domain signals and radio frequency signals output by a laser, identifying the current working state of the laser to distinguish a continuous light state, a Q-switched mode-locked state and a fundamental frequency mode-locked state; when the state is the continuous light state or the Q-switched mode-locked state, searching and adjusting a polarization control parameter and a pump power, and applying the adjusted parameters to the laser; when the state is the fundamental frequency mode-locked state, jointly judging based on the time domain signals and the radio frequency signals, if the target harmonic mode-locked state is not reached, searching and adjusting the polarization control parameter and the pump power, and if the target harmonic mode-locked state is reached, outputting a mode-locked result and corresponding parameters. The application realizes full automation control of the fiber laser, improves the generality of the algorithm, and has a simple structure and is easy to implement.

Description

Technical Field

[0001] This invention relates to the field of ultrafast fiber lasers and automatic control technology, and in particular to an intelligent harmonic mode-locking method and system based on deep learning-guided adaptive search. Background Technology

[0002] Fiber lasers, due to their small size, low cost, excellent heat dissipation, and high energy conversion efficiency, have become an important platform in the field of ultrafast photonics. They play a crucial role in numerous fields such as thermal imaging, spectral analysis, optical communication, precision sensing, biomedicine, materials processing, environmental monitoring, and military detection, and are particularly suitable for applications requiring extremely high temporal resolution and control precision. Through mode-locking technology, fiber lasers can output ultrashort pulses in the picosecond to femtosecond range, significantly expanding their functional boundaries in high-performance applications.

[0003] Mode-locking methods mainly include active mode-locking, passive mode-locking, and hybrid mode-locking. Among them, passive mode-locking technology based on nonlinear polarization rotation (NPR) has become a key research focus due to its advantages such as simple structure, narrow output pulse width, and wide and flat spectrum. However, such lasers are highly sensitive to intracavity polarization state and pump power, and are easily affected by environmental disturbances, device drift, and aging, resulting in significant uncertainty in their output state. Especially in high-harmonic mode-locking states, lasers have higher requirements for parameter matching conditions, often facing problems such as difficulty in automatic acquisition, short stability maintenance time, and difficulty in rapid recovery after mode lock loss. Traditional methods relying on manual polarizer adjustment are not only inefficient but also fail to meet the requirements of rapid mode-locking and long-term stability. While some existing automatic mode-locking schemes have improved in response speed, most schemes are still mainly aimed at the fundamental frequency mode-locking process, and dedicated technical solutions for the identification, determination, automatic search, and recovery from mode lock loss in high-harmonic modes are still relatively lacking. Therefore, there is an urgent need to develop a fully automated, highly responsive, and intelligent mode-locking control system capable of achieving the target harmonic mode-locking state, in order to further improve the practical performance and robustness of ultrafast fiber lasers. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent harmonic mode-locking method and system based on deep learning-guided adaptive search, so as to solve the problems in passive mode-locked fiber lasers where polarization control parameters and pump power adjustment rely on manual experience, the degree of automation is insufficient, and the high-order harmonic mode-locking state is difficult to establish and maintain quickly and stably.

[0005] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0006] In a first aspect, the present invention provides an intelligent harmonic mode-locking method based on deep learning-guided adaptive search, comprising:

[0007] Monitor and acquire spectral, time-domain, and radio frequency signals output by a passively mode-locked fiber laser based on nonlinear polarization rotation;

[0008] The current operating state of the passively mode-locked fiber laser is identified based on the spectral signal to distinguish between continuous light state, Q-switched mode-locked state, and fundamental frequency mode-locked state.

[0009] When the passively mode-locked fiber laser is currently operating in continuous light mode or Q-switched mode, the polarization control parameters and pump power are searched and adjusted, and the adjusted polarization control parameters and pump power are applied to the passively mode-locked fiber laser for re-monitoring and identification.

[0010] When the passively mode-locked fiber laser is currently in the fundamental frequency mode-locked state, the time domain signal and the radio frequency signal are jointly judged to determine whether the target harmonic mode-locked state has been reached.

[0011] If the target harmonic mode-locking state is not achieved, the polarization control parameters and pump power are searched and adjusted, and monitoring and judgment are performed again.

[0012] If the target harmonic mode-locking state is achieved, the mode-locking result, corresponding polarization control parameters, and pump power are output.

[0013] Preferably, the step of identifying the current operating state of the passively mode-locked fiber laser based on the spectral signal includes:

[0014] Acquire spectral signal data of a passively mode-locked fiber laser based on nonlinear polarization rotation under different polarization control parameters and different pump powers;

[0015] The acquired spectral signal data is preprocessed and divided into training, validation, and test sets.

[0016] The pre-built laser state recognition model is trained using the training set to obtain a trained laser state recognition model for recognizing continuous light state, Q-switched mode-locked state, and fundamental frequency mode-locked state.

[0017] The real-time acquired spectral signal data is input into the trained laser state recognition model, which outputs the corresponding laser operating state.

[0018] When the output result is in continuous light state or Q-switched mode-locked state, the output signal does not reach the base frequency mode-locked state.

[0019] When the output result is a baseband mode-locked state, a baseband mode-locked success signal is output.

[0020] Preferably, the laser state recognition model employs a residual neural network.

[0021] Preferably, the time-domain signal and the radio frequency signal are jointly judged to determine whether the target harmonic mode-locking state has been reached, including:

[0022] Calculate the repetition frequency in the fundamental frequency mode-locked state. ; By utilizing the time-domain and radio-frequency signals output from the laser, pulse peak detection is performed on the time-domain signal, and the interval between adjacent pulse peaks is extracted. And calculate the average pulse interval. : ,in, This represents the number of pulse peaks currently detected. Calculate the current repetition frequency based on the average pulse interval. : ;

[0023] Determine the number of pulse peaks in the time-domain signal:

[0024] When the number of currently detected pulse peaks Less than the preset threshold When the harmonic mode-locking criterion is not met, a signal indicating that the target harmonic mode-locking state has not been reached is output.

[0025] When the number of currently detected pulse peaks Greater than or equal to the preset threshold At the same time, the position of the main harmonic peak and the harmonic structure features are extracted from the radio frequency signal, and the time domain signal and the radio frequency signal are jointly determined, as follows:

[0026] Calculate the standard deviation of pulse peak interval And calculate the relative jitter. : ;

[0027] When the current repetition frequency Repetition frequency in mode-locked state with fundamental frequency Satisfying integer multiple relationships and relative jitter Less than the preset threshold And the number of consecutive detections exceeds a preset threshold. When the target harmonic mode-locking state is reached, a harmonic mode-locking success signal is output, along with the mode-locking result, corresponding polarization control parameters, and pump power.

[0028] When the current repetition frequency Repetition frequency in mode-locked state with fundamental frequency Not satisfying integer multiple relationship, or relative jitter amount Not less than the preset threshold Or the number of consecutive detections does not exceed a preset threshold. When the target harmonic mode-locking state is not reached, a signal indicating that the target harmonic mode-locking state has not been reached is output.

[0029] Preferably, the process of searching and adjusting the polarization control parameters and pump power includes:

[0030] Get the current parameter set The parameter set Including pump power and polarization control parameters , represented as ;

[0031] Based on the current parameter set Initialize the search space and initial search step size ;

[0032] pump power and polarization control parameters Loaded into the laser system;

[0033] Determine the current parameter set Whether the target harmonic mode-locked state is hit. If not, the parameter set is perturbed as follows:

[0034] When the target harmonic mode-locked state is hit, the hit result and the corresponding parameter set are output. And stop the current search process;

[0035] When the target harmonic mode-locked state is not hit, determine the current parameter set. Whether a candidate state is hit; the candidate state is an intermediate state that has not reached the target harmonic mode-locking state, but is closer to the target harmonic mode-locking state than the current working state;

[0036] When a candidate state is hit, the current parameter set is... Save as a candidate working point;

[0037] When no candidate state is found, the current parameter set is... Perform random perturbation to obtain the updated parameter set. : ,in, This represents the perturbation amount in the current iteration;

[0038] The similarity score is calculated based on the parameter sets before and after the random perturbation. : ,in, This is the state evaluation function;

[0039] when When the current parameter update direction is determined to be the improvement direction, the current parameter update direction is maintained; the parameter update direction is the parameter increment vector between the current parameter set and the updated parameter set.

[0040] when When the current parameter update direction is determined to be a non-improvement direction, the current parameter update direction is corrected.

[0041] The pump power and polarization control parameters are reset according to the updated parameter update direction, and the next round of search and adjustment is carried out until the preset termination condition is reached, at which point the search is terminated and the current search result is output.

[0042] Preferably, the state evaluation function is expressed as:

[0043] ,

[0044] in, For time-domain evaluation, Frequency domain evaluation metrics Spectral evaluation quantity, This indicates the first step in the parameter search and adjustment process. iteration , , Let be the weighting coefficient, and satisfy: .

[0045] Preferably, the time-domain evaluation quantity Represented as:

[0046] ,

[0047] in, For the first The number of pulse peaks detected in each iteration This serves as a normalized reference value for the number of pulse peaks. For the first The relative jitter in each iteration This serves as a reference value for jitter normalization. and For time-domain weighting coefficients.

[0048] Preferably, the frequency domain evaluation quantity Represented as:

[0049] ,

[0050] in, For the first The repetition frequency of each iteration For the target harmonic order, For the first The intensity difference between the main harmonic peak and the background in the next iteration This is a normalized reference value. and These are the frequency domain weighting coefficients.

[0051] Preferably, the spectral evaluation quantity Represented as:

[0052] ,

[0053] in, For the first The spectral bandwidth of the next iteration is 3 dB. For the first The spectral peak of the next iteration and These are the spectral weighting coefficients.

[0054] Secondly, the present invention provides an intelligent harmonic mode-locking system based on deep learning-guided adaptive search, used to implement the above-mentioned intelligent harmonic mode-locking method based on deep learning-guided adaptive search, the system comprising:

[0055] The data monitoring module is used to monitor and acquire the spectral signal, time domain signal, and radio frequency signal output by the passively mode-locked fiber laser based on nonlinear polarization rotation; and to send the spectral signal, time domain signal, and radio frequency signal to the fundamental frequency mode-locking module, the harmonic mode-locking module, and the parameter search and adjustment module.

[0056] The fundamental frequency mode-locking module is used to identify the current operating state of the passively mode-locked fiber laser based on the spectral signal, so as to distinguish between continuous light state, Q-switched mode-locking state and fundamental frequency mode-locking state.

[0057] The parameter search and adjustment module is used to search and adjust the polarization control parameters and pump power when the passive mode-locked fiber laser is currently in continuous light mode or Q-switched mode-locked mode, and then apply the adjusted polarization control parameters and pump power to the passive mode-locked fiber laser.

[0058] The harmonic mode-locking module is used to jointly determine whether the target harmonic mode-locking state has been reached when the passively mode-locked fiber laser is currently in the fundamental frequency mode-locking state. If the target harmonic mode-locking state has not been reached, the parameter search and adjustment module is invoked to search and adjust the polarization control parameters and pump power. If the target harmonic mode-locking state has been reached, the mode-locking result and the corresponding polarization control parameters and pump power are output.

[0059] The beneficial effects achieved by this invention are as follows:

[0060] This invention utilizes a deep learning model to automatically identify the laser's state and combines it with an adaptive optimization algorithm to intelligently search for optimal parameters, achieving fully automated and rapid control of an ultrafast fiber laser from fundamental frequency mode-locking to target high-repetition-frequency harmonic mode-locking. This invention effectively solves the problems of traditional methods, such as reliance on manual labor, low efficiency, and difficulty in rapidly establishing and maintaining stable harmonic mode-locking, significantly improving the automation, stability, and practicality of laser systems. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of the intelligent harmonic mode-locking method based on deep learning-guided adaptive search provided by the present invention;

[0062] Figure 2 This is a schematic diagram of the laser current working state identification process provided by the present invention;

[0063] Figure 3 This is a schematic diagram of the harmonic mode-locking process provided by the present invention;

[0064] Figure 4 This is a schematic diagram of the parameter search and adjustment process provided by the present invention;

[0065] Figure 5 This is a schematic diagram of the harmonic mode-locked spectrum results of a preferred embodiment of the present invention;

[0066] Figure 6 This is a schematic diagram of the harmonic mode-locking time domain results of a preferred embodiment of the present invention;

[0067] Figure 7 This is a schematic diagram of the harmonic mode-locked frequency domain results of a preferred embodiment of the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0069] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0070] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0071] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0072] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0073] It should be emphasized here that the step markers mentioned below are not a limitation on the order of the steps, but should be understood as meaning that the steps can be executed in the order mentioned in the embodiments, or in a different order than in the embodiments, or several steps can be executed simultaneously.

[0074] This invention provides an intelligent harmonic mode-locking method based on deep learning-guided adaptive search, achieving fully automated identification of the spectral state of fiber lasers and rapid high-repetition-rate (HRNR) harmonic mode-locking. This method accurately determines the current spectral state and generates corresponding control parameters using an optimization algorithm. These parameters then drive the electrically controlled polarization controller and pump system to perform real-time adjustments, completing HNR harmonic mode-locking and significantly improving the automation and response speed of harmonic mode-locked laser state control. For detailed implementation details, please refer to [link to implementation details]. Figure 1 ,include:

[0075] Monitor and acquire spectral, time-domain, and radio frequency signals output by a passively mode-locked fiber laser based on nonlinear polarization rotation;

[0076] The current operating state of the laser is identified based on the collected spectral signals to distinguish between continuous light state, Q-switched mode-locked state, and fundamental frequency mode-locked state.

[0077] When the laser is currently operating in continuous light mode or Q-switched mode-locked mode, the polarization control parameters and pump power are searched and adjusted, and the adjusted polarization control parameters and pump power are applied back to the passively mode-locked fiber laser. Then, the process is repeated for monitoring and identification.

[0078] When the laser is currently operating in the fundamental frequency mode-locked state, the time domain signal and radio frequency signal are jointly judged to determine whether the target harmonic mode-locked state has been reached.

[0079] If the target harmonic mode-locking state is not achieved, the polarization control parameters and pump power are searched and adjusted, and the monitoring and judgment are repeated.

[0080] If the target harmonic mode-locking state is achieved, output the mode-locking result and the corresponding polarization control parameters and pump power.

[0081] It should be noted that the establishment of harmonic mode-locking state is premised on fundamental frequency mode-locking state. Only after the laser has first achieved stable fundamental frequency mode-locking, and further adjustments to the pump power and polarization control parameters, can the system enter the target harmonic mode-locking state. Therefore, in this invention, the current operating state of the laser is first identified based on the acquired spectral signal, the specific implementation process of which is as follows: Figure 2 As shown, it includes:

[0082] S11. Obtain spectral signal data of a passively mode-locked fiber laser based on nonlinear polarization rotation under different polarization control parameters and different pump powers;

[0083] S12. Preprocess the acquired spectral signal data. The preprocessing includes at least one or more of the following: normalization, cropping, and filtering.

[0084] S13. Divide the preprocessed spectral signal data into training set, validation set and test set;

[0085] S14. The pre-built laser state recognition model is trained using the training set to obtain a trained laser state recognition model for recognizing continuous light state, Q-switched mode-locked state and fundamental frequency mode-locked state.

[0086] S15. Input the real-time acquired spectral signal data into the trained laser state recognition model and output the corresponding laser working state.

[0087] S16. Determine the current state of the laser based on the laser operating state category;

[0088] When the output result is in continuous light state or Q-switched mode-locked state, the output signal does not reach the base frequency mode-locked state.

[0089] When the output result is a baseband mode-locked state, a baseband mode-locked success signal is output.

[0090] Preferably, the laser state recognition model employs a residual neural network.

[0091] To address the sensitivity of the polarization state of harmonic mode-locked lasers in optical fibers to environmental influences, this invention, when the laser is currently operating in a fundamental frequency mode-locked state, jointly analyzes the time-domain and radio-frequency signals to determine whether the target harmonic mode-locked state has been achieved. For details, please refer to [link to relevant documentation]. Figure 3 ,include:

[0092] S21. Receive the baseband mode-locking success signal and calculate the repetition frequency in the baseband mode-locking state. ;

[0093] S22. Using the time-domain signal and radio-frequency signal output from the laser, perform pulse peak detection on the time-domain signal and extract the interval between adjacent pulse peaks. And calculate the average pulse interval. ,in, This represents the number of pulse peaks currently detected.

[0094] S23. Calculate the current repetition frequency based on the average pulse interval. Its expression is ;

[0095] S24. Determine the number of pulse peaks in the time-domain signal:

[0096] When the number of currently detected pulse peaks Less than the preset threshold When the harmonic mode-locking criterion is not met, a signal indicating that the target harmonic mode-locking state has not been reached is output.

[0097] When the number of currently detected pulse peaks Greater than or equal to the preset threshold Calculate the standard deviation of the pulse peak interval. And calculate the relative jitter. Its expression is ;

[0098] S25. Set the current repetition frequency. Repetition frequency in mode-locked state with fundamental frequency Compare them to determine whether the integer multiple relationship is satisfied, that is... ,in, The target harmonic order;

[0099] S26. Combine the radio frequency signal to extract the position of the main harmonic peak and the harmonic structure characteristics, and make a joint judgment on the time domain signal and the radio frequency signal.

[0100] When the current repetition frequency Repetition frequency in mode-locked state with fundamental frequency Satisfying integer multiple relationships and relative jitter Less than the preset threshold And the number of consecutive detections exceeds a preset threshold. When the target harmonic mode-locking state is reached, a harmonic mode-locking success signal is output, along with the mode-locking result, corresponding polarization control parameters, and pump power.

[0101] When the current repetition frequency Repetition frequency in mode-locked state with fundamental frequency Not satisfying integer multiple relationship, or relative jitter amount Not less than the preset threshold Or the number of consecutive detections does not exceed a preset threshold. When the target harmonic mode-locking state is not reached, a signal indicating that the target harmonic mode-locking state has not been reached is output.

[0102] For subsequent adjustments and recovery, in this invention, if the target harmonic mode-locking state is not achieved, the polarization control parameters and pump power are searched and adjusted, and monitoring and judgment are performed again. The specific implementation process is as follows: Figure 4 As shown, it includes:

[0103] S31, Receive the current parameter set Parameter set Including pump power and polarization control parameters ,Right now ;

[0104] S32. Based on the current parameter set Initialize the search space and initial search step size ;

[0105] S33, Pump power and polarization control parameters Loaded into the laser system;

[0106] S34. Determine the current parameter set Whether the target harmonic mode-locked state is hit. If not, the parameter set is perturbed as follows:

[0107] When the target harmonic mode-locked state is hit, the hit result and the corresponding parameter set are output. And stop the current search process;

[0108] If the target harmonic mode-locked state is not found, further judgment is made on the current parameter set. Whether the laser output hits the candidate state; where the candidate state is an intermediate state that is closer to the target harmonic mode-locking state than the current working state, although it has not reached the target harmonic mode-locking criterion.

[0109] When a candidate state is hit, the current parameter set is... Save as a candidate working point;

[0110] When no candidate state is found, the current parameter set is... Perform random perturbation to obtain the updated parameter set. : ,in, This represents the perturbation amount in the current iteration;

[0111] S35. Calculate the similarity score based on the parameter set before and after the random perturbation. Its expression is ,in, This is the state evaluation function;

[0112] when When the current parameter update direction is determined to be the direction of improvement, the current parameter update direction is maintained.

[0113] when When the current parameter update direction is determined to be a non-improvement direction, the current parameter update direction is corrected.

[0114] The parameter update direction is defined as the parameter increment vector between the current parameter set and the updated parameter set, i.e.:

[0115] ,

[0116] in, Indicates the first The parameter update direction corresponding to the next iteration;

[0117] S36. Based on the updated parameter update direction, reset the pump power and polarization control parameters, and perform the next round of search and adjustment;

[0118] S37. Determine whether the preset termination condition has been met;

[0119] When the preset termination condition is met, the search will terminate and the current search results will be output.

[0120] In this invention, the specific expression of the state evaluation function is as follows:

[0121] ,

[0122] in For time-domain evaluation, Frequency domain evaluation metrics Spectral evaluation quantity, This indicates the first step in the parameter search and adjustment process. iteration , , Let be the weighting coefficient, and satisfy:

[0123] ,

[0124] Time-domain evaluation quantity The expression is:

[0125] ,

[0126] in For the first The number of pulse peaks detected in each iteration This serves as a normalized reference value for the number of pulse peaks. For the first The relative jitter in each iteration This serves as a reference value for jitter normalization. and For time-domain weighting coefficients.

[0127] Frequency domain evaluation quantity The expression is:

[0128] ,

[0129] in For the first The repetition frequency of each iteration For the target harmonic order, For the first The intensity difference between the main harmonic peak and the background in the next iteration This is a normalized reference value. and These are the frequency domain weighting coefficients.

[0130] Spectral evaluation quantity The expression is:

[0131] ,

[0132] in For the first The spectral bandwidth of the next iteration is 3 dB. For the first The spectral peak of the next iteration and These are the spectral weighting coefficients.

[0133] Figure 5 The diagram shows the harmonic mode-locked spectrum results achieved by the method of this invention. It can be seen that the spectral center is located at approximately 1569.8 nm, the bandwidth is 2.8 nm, and obvious Kelly sidebands are visible on both sides of the spectrum, indicating that the laser is in a stable mode-locked state.

[0134] Figure 6 This is a schematic diagram of the time-domain results corresponding to the harmonic mode-locked pulses. The output pulses exhibit a stable periodic distribution on the time axis, with an interval of approximately 3.2 ns between adjacent pulses, indicating that the laser has formed a stable periodic pulse train.

[0135] Figure 7 This is a schematic diagram of the frequency domain results of the corresponding harmonic mode-locked pulse. The repetition frequency is as high as 309.96MHz, and the difference in intensity from the background noise is 33dB, indicating that the output pulse has good periodic stability and high mode-locking quality.

[0136] Based on the same inventive concept, this invention also provides an intelligent harmonic mode-locking system based on deep learning-guided adaptive search, used to implement the above-mentioned intelligent harmonic mode-locking method based on deep learning-guided adaptive search. The system includes:

[0137] The data monitoring module is used to monitor and acquire the spectral signal, time domain signal, and radio frequency signal output by the passively mode-locked fiber laser based on nonlinear polarization rotation; and to send the spectral signal, time domain signal, and radio frequency signal to the fundamental frequency mode-locking module, the harmonic mode-locking module, and the parameter search and adjustment module.

[0138] The fundamental frequency mode-locking module is used to identify the current operating state of the passively mode-locked fiber laser based on the spectral signal, so as to distinguish between continuous light state, Q-switched mode-locking state and fundamental frequency mode-locking state.

[0139] The parameter search and adjustment module is used to search and adjust the polarization control parameters and pump power when the passive mode-locked fiber laser is currently in continuous light mode or Q-switched mode-locked mode, and then apply the adjusted polarization control parameters and pump power to the passive mode-locked fiber laser.

[0140] The harmonic mode-locking module is used to jointly determine whether the target harmonic mode-locking state has been reached when the passively mode-locked fiber laser is currently in the fundamental frequency mode-locking state. If the target harmonic mode-locking state has not been reached, the parameter search and adjustment module is invoked to search and adjust the polarization control parameters and pump power. If the target harmonic mode-locking state has been reached, the mode-locking result and the corresponding polarization control parameters and pump power are output.

[0141] It is worth noting that the system embodiment corresponds to the above method embodiment. The implementation methods of the above method embodiments are all applicable to the system embodiment and can achieve the same or similar technical effects, so they will not be described in detail here.

[0142] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0143] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0144] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0145] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0147] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A smart harmonic mode-locking method based on deep learning-guided adaptive search, characterized in that, include: Monitor and acquire spectral, time-domain, and radio frequency signals output by a passively mode-locked fiber laser based on nonlinear polarization rotation; The current operating state of the passively mode-locked fiber laser is identified based on the spectral signal to distinguish between continuous light state, Q-switched mode-locked state, and fundamental frequency mode-locked state. When the passively mode-locked fiber laser is currently operating in continuous light mode or Q-switched mode-locked mode, the polarization control parameters and pump power are searched and adjusted, and the adjusted polarization control parameters and pump power are applied to the passively mode-locked fiber laser for re-monitoring and identification. When the passively mode-locked fiber laser is currently in the fundamental frequency mode-locked state, the time domain signal and the radio frequency signal are jointly judged to determine whether the target harmonic mode-locked state has been reached. If the target harmonic mode-locking state is not achieved, the polarization control parameters and pump power are searched and adjusted, and monitoring and judgment are performed again. If the target harmonic mode-locking state is achieved, the mode-locking result, corresponding polarization control parameters, and pump power are output.

2. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 1, characterized in that, The process of identifying the current operating state of the passively mode-locked fiber laser based on the spectral signal includes: Acquire spectral signal data of a passively mode-locked fiber laser based on nonlinear polarization rotation under different polarization control parameters and different pump powers; The acquired spectral signal data is preprocessed and divided into training, validation, and test sets. The pre-built laser state recognition model is trained using the training set to obtain a trained laser state recognition model for recognizing continuous light state, Q-switched mode-locked state, and fundamental frequency mode-locked state. The real-time acquired spectral signal data is input into the trained laser state recognition model, which outputs the corresponding laser operating state. When the output result is in continuous light state or Q-switched mode-locked state, the output signal does not reach the base frequency mode-locked state. When the output result is a baseband mode-locked state, a baseband mode-locked success signal is output.

3. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 2, characterized in that, The laser state recognition model employs a residual neural network.

4. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 1, characterized in that, The time-domain signal and the radio frequency signal are jointly evaluated to determine whether the target harmonic mode-locking state has been reached, including: Calculate the repetition frequency in the fundamental frequency mode-locked state ; By utilizing the time-domain and radio-frequency signals output from the laser, pulse peak detection is performed on the time-domain signal, and the interval between adjacent pulse peaks is extracted. And calculate the average pulse interval. : ,in, This represents the number of pulse peaks currently detected. Calculate the current repetition frequency based on the average pulse interval. : ; Determine the number of pulse peaks in the time-domain signal: When the number of currently detected pulse peaks Less than the preset threshold When the harmonic mode-locking criterion is not met, a signal indicating that the target harmonic mode-locking state has not been reached is output. When the number of currently detected pulse peaks Greater than or equal to the preset threshold At the same time, the position of the main harmonic peak and the harmonic structure features are extracted from the radio frequency signal, and the time domain signal and the radio frequency signal are jointly determined, as follows: Calculate the standard deviation of pulse peak interval And calculate the relative jitter. : ; When the current repetition frequency Repetition frequency in mode-locked state with fundamental frequency Satisfying integer multiple relationships and relative jitter Less than the preset threshold And the number of consecutive detections exceeds a preset threshold. When the target harmonic mode-locking state is reached, a harmonic mode-locking success signal is output, along with the mode-locking result, corresponding polarization control parameters, and pump power. When the current repetition frequency Repetition frequency in mode-locked state with fundamental frequency Not satisfying integer multiple relationship, or relative jitter amount Not less than the preset threshold Or the number of consecutive detections does not exceed a preset threshold. When the target harmonic mode-locking state is not reached, a signal indicating that the target harmonic mode-locking state has not been reached is output.

5. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 1, characterized in that, The process of searching and adjusting polarization control parameters and pump power includes: Get the current parameter set The parameter set Including pump power and polarization control parameters , represented as ; Based on the current parameter set Initialize the search space and initial search step size ; pump power and polarization control parameters Loaded into the laser system; Determine the current parameter set Whether the target harmonic mode-locked state is hit. If not, the parameter set is perturbed as follows: When the target harmonic mode-locked state is hit, the hit result and the corresponding parameter set are output. And stop the current search process; When the target harmonic mode-locked state is not hit, determine the current parameter set. Whether a candidate state is hit; the candidate state is an intermediate state that has not reached the target harmonic mode-locking state, but is closer to the target harmonic mode-locking state than the current working state; When a candidate state is hit, the current parameter set is... Save as a candidate working point; When no candidate state is found, the current parameter set is... Perform random perturbation to obtain the updated parameter set. : ,in, This represents the perturbation amount in the current iteration; The similarity score is calculated based on the parameter sets before and after the random perturbation. : ,in, This is the state evaluation function; when When the current parameter update direction is determined to be the improvement direction, the current parameter update direction is maintained; the parameter update direction is the parameter increment vector between the current parameter set and the updated parameter set. when When the current parameter update direction is determined to be a non-improvement direction, the current parameter update direction is corrected. The pump power and polarization control parameters are reset according to the updated parameter update direction, and the next round of search and adjustment is carried out until the preset termination condition is reached, at which point the search is terminated and the current search result is output.

6. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 5, characterized in that, The state evaluation function is expressed as follows: ， in, For time-domain evaluation, Frequency domain evaluation metrics Spectral evaluation quantity, This indicates the first step in the parameter search and adjustment process. iteration , , Let be the weighting coefficient, and satisfy: .

7. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 6, characterized in that, The time-domain evaluation quantity Represented as: ， in, For the first The number of pulse peaks detected in each iteration This serves as a normalized reference value for the number of pulse peaks. For the first The relative jitter in each iteration This serves as a reference value for jitter normalization. and For time-domain weighting coefficients.

8. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 6, characterized in that, The frequency domain evaluation quantity Represented as: ， in, For the first The repetition frequency of each iteration For the target harmonic order, For the first The intensity difference between the main harmonic peak and the background in the next iteration This is a normalized reference value. and These are the frequency domain weighting coefficients.

9. The intelligent harmonic mode-locking method based on deep learning-guided adaptive search according to claim 6, characterized in that, The spectral evaluation quantity Represented as: ， in, For the first The spectral bandwidth of the next iteration is 3 dB. For the first The spectral peak of the next iteration and These are the spectral weighting coefficients.

10. An intelligent harmonic mode-locking system based on deep learning-guided adaptive search, characterized in that, The system for implementing the intelligent harmonic mode-locking method based on deep learning-guided adaptive search as described in claim 1, the system comprising: The data monitoring module is used to monitor and acquire the spectral signal, time domain signal, and radio frequency signal output by the passively mode-locked fiber laser based on nonlinear polarization rotation; and to send the spectral signal, time domain signal, and radio frequency signal to the fundamental frequency mode-locking module, the harmonic mode-locking module, and the parameter search and adjustment module. The fundamental frequency mode-locking module is used to identify the current operating state of the passively mode-locked fiber laser based on the spectral signal, so as to distinguish between continuous light state, Q-switched mode-locking state and fundamental frequency mode-locking state. The parameter search and adjustment module is used to search and adjust the polarization control parameters and pump power when the passive mode-locked fiber laser is currently in continuous light mode or Q-switched mode-locked mode, and then apply the adjusted polarization control parameters and pump power to the passive mode-locked fiber laser. The harmonic mode-locking module is used to jointly determine whether the target harmonic mode-locking state has been reached when the passively mode-locked fiber laser is currently in the fundamental frequency mode-locking state. If the target harmonic mode-locking state has not been reached, the parameter search and adjustment module is invoked to search and adjust the polarization control parameters and pump power. If the target harmonic mode-locking state has been reached, the mode-locking result and the corresponding polarization control parameters and pump power are output.

Images