A femtosecond pulse energy self-stabilizing device and method based on Bessel beams and AI

Abstract

This invention discloses a femtosecond pulse energy self-stabilization device and method based on Bessel beams and AI, relating to the technical field of lasers. The device operates as follows: an initial femtosecond Gaussian pulse is output from a femtosecond seed source and converted into a non-diffraction Bessel beam by a Bessel beam generation module; pulse energy fluctuations are converted into spectral width variations through a nonlinear medium; energy fluctuations are predicted using AI, and compensation commands are generated using the ADRC algorithm to drive a frequency domain shutter module and a pump module for correction; a high-power stable femtosecond pulse is output through an amplification module, and feedback is fed back through a feedback detection module to form a closed-loop control. This invention, through predictive fluctuation compensation, controls pulse energy jitter to less than 0.1%, effectively solving the technical problems of lag and device damage under high power in traditional PID control.

Description

Technical Field

[0001] This invention relates to the technical field of lasers, and specifically to a femtosecond pulse energy self-stabilization device and method based on Bessel beams and AI. Background Technology

[0002] Femtosecond lasers possess irreplaceable advantages in precision machining, biomedicine, and quantum technology due to their ultrashort pulse widths and ultra-high peak power. As high-power ultrafast solid-state lasers develop towards kilowatt-level output and MHz-level high repetition rates, the stability of pulse energy has become a core bottleneck restricting their application accuracy.

[0003] Currently, femtosecond pulse energy stabilization mainly adopts the traditional PID feedback control method, which has the following prominent technical pain points: First, the feedback lag is significant, with compensation delay typically ranging from 10 to 100 μs, making it unable to cope with transient interferences such as thermal drift and mechanical vibration under high-power conditions, resulting in pulse energy jitter that is difficult to control below 1%; Second, it is prone to damaging devices under high power conditions, as existing systems often use Gaussian beams as the detection and transmission carrier, which are prone to self-focusing effects, leading to the burnout of the crystal medium; Third, it has poor adaptability, and for high-repetition-rate, kilowatt-level femtosecond lasers, traditional stabilization schemes cannot balance stability and power output. Summary of the Invention

[0004] The purpose of this invention is to solve the problems mentioned in the background art, such as control lag, easy damage to devices under high power, and excessive energy fluctuations in traditional control methods, and to propose a femtosecond pulse energy self-stabilization device and method based on Bessel beams and AI.

[0005] A first aspect of this invention provides a femtosecond pulse energy self-stabilizing device based on Bessel beams and AI, the device comprising: sequentially arranged along the optical path direction: Femtosecond seed source, used to output femtosecond Gaussian pulses; A Bessel beam generation module is used to convert the femtosecond Gaussian pulse into a diffraction-free Bessel beam; The nonlinear detection module is used to convert the pulse energy fluctuations of the Bessel beam into spectral width changes and to unfold the changed spectrum. The frequency domain shutter module is used to perform spectral trimming on the Bessel beam to achieve coarse correction of the pulse energy; The amplification module is used to amplify the power of the corrected Bessel beam and output it. The feedback detection module is used to acquire spectral width data and pulse energy data; The AI ​​prediction control module is connected to the feedback detection module, the frequency domain shutter module, and the femtosecond seed source, respectively. It is used to generate compensation instructions by running a preset AI algorithm based on the collected spectral width data and pulse energy data, and drive the frequency domain shutter module and the femtosecond seed source to perform energy compensation.

[0006] Optionally, the Bessel beam generation module includes a conical lens and a low-dispersion lens; wherein: The conical lens is used to shape a Gaussian beam into a Bessel beam. The low-dispersion lens is used to correct the dispersion introduced by the cone lens, ensuring the diffraction-free characteristics of the Bessel beam.

[0007] Optionally, the cone angle of the cone lens is 1-3°, the diameter of the central main lobe of the generated Bessel beam is 2-5mm, and the width of the annular side lobes is 0.5-1mm.

[0008] Optionally, the nonlinear detection module includes a nonlinear medium and a grating; wherein: The nonlinear medium is fused silica or sapphire with a thickness of 5-20 mm, used to convert the pulse energy fluctuations of the Bessel beam into spectral width changes through self-phase modulation effect; The grating is a blazed grating with a diffraction efficiency greater than 95%, used to expand the changed spectrum.

[0009] Optionally, the frequency domain shutter module includes a high-speed liquid crystal tunable filter; the response time of the high-speed liquid crystal tunable filter is less than... The passband adjustment range is 1-10nm.

[0010] Optionally, the feedback detection module includes a spectral detector and an energy detector; wherein: The spectral detector is used to collect the expanded spectral width data, and the sampling frequency is not less than 10MHz; The energy detector is used to collect the final pulse energy data with an accuracy of 0.01 mJ.

[0011] Optionally, the AI ​​prediction control module includes an industrial control computer integrating FPGA and GPU; wherein: The FPGA is used to acquire spectral data in real time, with a sampling frequency greater than 10MHz; The GPU is used to run a neural network prediction model and an active disturbance rejection control algorithm. The neural network prediction model receives spectral width data and outputs a predicted fluctuation in pulse energy. The active disturbance rejection control algorithm generates a final compensation instruction to drive the frequency domain shutter module and the femtosecond seed source based on the predicted fluctuation and the current real pulse energy measured by the feedback detection module.

[0012] Optionally, the amplification module is a Yb:YAG solid-state slab amplifier with a bidirectional pumping structure on the side and end face, and a pumping wavelength of 940nm.

[0013] A second aspect of this invention provides a femtosecond pulse energy self-stabilization method based on Bessel beams and AI, the method comprising: S1: Convert the femtosecond Gaussian pulse output from the femtosecond seed source into a diffraction-free Bessel beam through the Bessel beam generation module; S2: Pass the Bessel beam through the nonlinear medium, use the self-phase modulation effect to convert the pulse energy fluctuation into a change in spectral width, and unfold the changed spectrum through a grating; S3: Real-time acquisition of spectral width data after grating expansion using a spectral detector, and transmission of the data to the AI ​​prediction and control module; S4: The AI ​​predictive control module uses a neural network prediction model to predict the amount and trend of energy fluctuations, and generates compensation commands through an active disturbance rejection control algorithm to drive the frequency domain shutter module to perform spectral trimming on the Bessel beam and adjust the pump power of the femtosecond seed source to achieve energy correction. S5: The corrected Bessel beam is amplified by the amplification module and then output; S6: Collect pulse energy data using an energy detector and feed it back to the AI ​​predictive control module. When the pulse energy fluctuation exceeds a preset threshold, repeat step S4 until the stability requirement is met, forming a complete closed-loop control.

[0014] The beneficial effects of this invention are: 1. By utilizing the non-diffraction characteristics of Bessel beams to extend the nonlinear action distance, the self-phase modulation effect is fully accumulated, and even small energy fluctuations can trigger detectable changes in spectral width, thus achieving highly sensitive non-contact energy fluctuation monitoring.

[0015] 2. By using a neural network prediction model to predict energy fluctuation trends and compensate in advance, the problem of delayed compensation for transient interference under high-power conditions is solved.

[0016] 3. Bessel beams avoid the self-focusing effect of high-power Gaussian beams, protect nonlinear media and subsequent optical components from damage, and extend the system's lifespan.

[0017] 4. Through a dual-layer compensation mechanism of coarse correction by frequency domain shutter spectral clipping and fine correction by pump power adjustment, combined with closed-loop control formed by energy detector feedback, the output pulse energy jitter can be stably controlled below 0.1%, meeting the stringent requirements for femtosecond laser energy stability in fields such as precision machining and quantum technology. Attached Figure Description

[0018] Figure 1A schematic diagram of a femtosecond pulse energy self-stabilizing device based on Bessel beam and AI provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a femtosecond pulse energy self-stabilization method based on Bessel beams and AI, provided for an embodiment of the present invention.

[0019] Figure labeling: 1-Femtosecond seed source; 2-Bessel beam generation module; 21-Conical lens; 22-Low dispersion lens; 3-Nonlinear detection module; 4-Frequency domain shutter module; 5-AI predictive control module; 6-Amplification module; 7-Feedback detection module; 8-Femtosecond Gaussian pulse; 9-Bessel beam; 10-High-power stable femtosecond pulse output. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention is provided in conjunction with the accompanying drawings and preferred embodiments.

[0021] This invention provides a femtosecond pulse energy self-stabilizing device based on Bessel beams and AI. See also... Figure 1 , Figure 1 This is a schematic diagram of a femtosecond pulse energy self-stabilizing device based on a Bessel beam and AI, provided as an embodiment of the present invention. The device includes: Femtosecond seed source 1 is used to output femtosecond Gaussian pulse 8.

[0022] The Bessel beam generation module 2 is set in the output optical path of the femtosecond seed source 1 and is used to convert the femtosecond Gaussian pulse 8 into a non-diffraction Bessel beam 9.

[0023] The nonlinear detection module 3 is set in the output optical path of the Bessel beam generation module 2. It is used to convert the pulse energy fluctuation of the Bessel beam 9 into a change in spectral width and to expand the changed spectrum.

[0024] The frequency domain shutter module 4 is set in the output optical path of the nonlinear detection module 3 and is used to perform spectral trimming on the Bessel beam to achieve coarse correction of the pulse energy.

[0025] Amplification module 6 is set on the output optical path of frequency domain shutter module 4 and is used to amplify the power of the corrected Bessel beam and output it.

[0026] The feedback detection module 7 is connected to the output terminals of the nonlinear detection module 3 and the frequency domain shutter module 4, and is used to acquire spectral width data and pulse energy data.

[0027] The AI ​​prediction control module 5 is connected to the feedback detection module 7, the frequency domain shutter module 4, and the femtosecond seed source 1, respectively. It is used to generate compensation instructions by running a preset AI algorithm based on the collected spectral width data and pulse energy data, and drive the frequency domain shutter module 4 and the femtosecond seed source 1 to perform energy compensation.

[0028] This invention provides a femtosecond pulse energy self-stabilizing device based on a Bessel beam and artificial intelligence (AI). By leveraging the non-diffraction propagation characteristics of the Bessel beam, the effective distance for nonlinear detection is extended several times, ensuring high-sensitivity detection while avoiding damage to the nonlinear medium from high-power lasers. Simultaneously, AI is used to predict energy fluctuation trends, combined with active disturbance rejection control for correction, ensuring that pulse energy jitter can be stably controlled below 0.1%, meeting the stringent energy stability requirements of kilowatt-level, megahertz-frequency femtosecond lasers.

[0029] In one implementation, the Bessel beam generation module 2 includes a conical lens 21 and a low-dispersion lens 22. Wherein: Conical lens 21 is used to shape a Gaussian beam into a Bessel beam. Specifically, the cone angle of conical lens 21 is 1-3°. The specific cone angle parameter of 1-3° of the conical lens can generate a Bessel beam with a central main lobe diameter of 2-5 mm and annular sidelobe width of 0.5-1 mm. This parameter range can ensure that the beam has sufficient power density in the nonlinear medium to excite the self-phase modulation effect, while avoiding excessive energy concentration that could lead to medium damage.

[0030] The low-dispersion lens 22 is used to correct the dispersion introduced by the conical lens 21, ensuring the diffraction-free characteristics of the Bessel beam and maintaining a stable transverse light field distribution over a long propagation distance, thus providing uniform and controllable beam conditions for subsequent nonlinear detection.

[0031] In one implementation, the nonlinear detection module 3 includes a nonlinear medium and a grating. Wherein: Nonlinear media are used to convert the pulse energy fluctuations of the Bessel beam 9 into spectral width variations through a self-phase modulation effect. Specifically, the nonlinear media are fused silica or sapphire with a thickness of 5-20 mm.

[0032] The grating is used to expand the changed spectrum, facilitating subsequent data acquisition. Specifically, the grating is a blazed grating with a diffraction efficiency greater than 95%.

[0033] In one implementation, the frequency domain shutter module 4 includes a high-speed liquid crystal tunable filter (LCTF). Specifically, the response time of this high-speed liquid crystal tunable filter is less than... With a passband adjustment range of 1-10nm, it features fast response and fine adjustment capabilities. By adjusting the filter passband width in real time, precise spectral trimming of the Bessel beam can be performed in the frequency domain, directly translating changes in spectral width into energy compensation, thus achieving coarse correction of pulse energy.

[0034] In one implementation, the feedback detection module 7 includes a spectral detector and an energy detector. Wherein: The spectrometer is used to acquire the expanded spectral width data. Specifically, the spectrometer has a sampling frequency of not less than 10MHz, enabling real-time high-speed acquisition of the expanded spectral width data, ensuring complete capture of spectral change information and without losing transient characteristics of rapid fluctuations.

[0035] An energy detector is used to collect pulse energy data. Specifically, the energy detector has an accuracy of 0.01 mJ, enabling it to accurately measure the output pulse energy and provide a true energy feedback signal to the AI ​​predictive control module, ensuring the accuracy and reliability of closed-loop control.

[0036] In one implementation, the AI ​​predictive control module 5 includes an industrial control computer integrating an FPGA and a GPU. The FPGA is used to acquire spectral data in real time, with a sampling frequency of not less than 10MHz. The GPU is used to run the neural network predictive model and the active disturbance rejection control (ADRC) algorithm. Wherein: The neural network prediction model receives spectral width data and outputs a predicted value for pulse energy fluctuations. Specifically, a generative adversarial network (GAN) is used to simulate transient disturbances (mechanical vibration, thermal drift, etc.) under high-power conditions, generating spectral-energy correlation samples corresponding to the disturbances. The prediction model is trained using both real-collected samples and simulated samples to establish a mapping relationship between spectral width changes and pulse energy fluctuations.

[0037] The active disturbance rejection control algorithm generates the final compensation command to drive the frequency domain shutter module 4 and the femtosecond seed source 1 based on the fluctuation prediction and the current real pulse energy measured by the feedback detection module 7, thereby achieving precise correction of the pulse energy.

[0038] In one implementation, the amplification module 6 is a Yb:YAG solid-state slab amplifier with a bidirectional pumping structure on the side and end face. The pumping wavelength is 940nm, which is used to amplify the corrected Bessel beam pulse and output a high-power stable femtosecond pulse.

[0039] This invention provides a method for self-stabilizing the energy of femtosecond pulses based on Bessel beams and AI. See also... Figure 2 , Figure 2 A flowchart illustrating a femtosecond pulse energy self-stabilization method based on Bessel beams and AI, provided as an embodiment of the present invention. The method includes: S1: Bessel beam generation: Femtosecond seed source 1 outputs a femtosecond Gaussian pulse 8, which passes sequentially through the conical lens 21 and the low-dispersion lens 22 of the Bessel beam generation module 2. The conical lens 21 converts the femtosecond Gaussian pulse 8 into a Bessel beam 9, and the low-dispersion lens 22 corrects the dispersion introduced by the conical lens 21, ensuring the diffraction-free characteristics of the Bessel beam and keeping the beam spot size stable with very little change during propagation.

[0040] S2: Energy to Spectrum Conversion: The Bessel beam 9 passes through the nonlinear medium (fused silica or sapphire) of the nonlinear detection module 3. Utilizing the self-phase modulation (SPM) effect, the minute fluctuations in pulse energy are converted into changes in spectral width: as the energy increases, the spectral width broadens; as the energy decreases, the spectral width narrows, achieving precise energy-spectrum mapping. The changed spectrum is then unfolded through a grating.

[0041] S3: Spectral Acquisition and Signal Transmission: The spectral detector of the feedback detection module 7 acquires spectral width data in real time and transmits the data to the FPGA of the AI ​​prediction control module. The sampling frequency is not less than 10MHz to ensure the real-time performance of the signal.

[0042] S4: AI Prediction and Compensation: The system receives the spectral width signal transmitted by the spectral detector in real time, predicts the pulse energy fluctuation through a neural network prediction model, generates a compensation amount based on the predicted fluctuation, and then the Active Disturbance Rejection Control (ADRC) unit generates control commands based on the predicted compensation amount. On the one hand, it drives the high-speed liquid crystal tunable filter (LCTF) to adjust the passband: when the energy is too high, the spectral edge portion is cut off; when the energy is insufficient, the passband is widened, realizing spectral clipping and coarse energy correction. On the other hand, it fine-tunes the pump power of the femtosecond seed source 1 to achieve fine energy correction. Furthermore, the energy detector of the feedback detection module 7 collects the corrected pulse energy data and transmits it to the AI ​​prediction control module 5. If the energy jitter is greater than a preset threshold, such as 0.1%, step S4 is repeated until the stability requirement is met, forming a complete closed-loop control.

[0043] S5: Amplified Output: The corrected Bessel beam pulse enters the amplification module 6, and is amplified by bidirectional pumping of the Yb:YAG solid-state slab amplifier to output a kilowatt-level high-power femtosecond pulse.

[0044] This invention proposes a technical solution that deeply integrates the non-diffraction and long-range characteristics of Bessel beams with AI prediction and active disturbance rejection control (ADRC) techniques, forming a closed-loop stable system of "physical conversion - spectral detection - AI prediction - precise compensation". This invention utilizes a conical lens to convert Gaussian pulses into non-diffraction Bessel beams, extending the nonlinear operating distance many times over. This avoids crystal damage caused by high-power lasers and improves the mapping sensitivity between energy fluctuations and spectral changes. Through the self-phase modulation effect of the nonlinear medium, minute fluctuations in pulse energy are transformed into detectable changes in spectral width, achieving non-contact, highly sensitive energy fluctuation monitoring. Combined with AI prediction and ADRC, it overcomes the lag bottleneck of traditional PID control, achieving pulse energy self-stabilization.

[0045] In one embodiment, to verify the effectiveness of the femtosecond pulse energy self-stabilization device and method based on Bessel beams and AI proposed in this invention, the following experiment was conducted.

[0046] This embodiment provides an AI predictive control femtosecond pulse energy self-stabilization device and method based on Bessel beams. The specific parameters and implementation process are as follows: (1) Device parameters: Femtosecond seed source 1: Outputs 8 femtosecond Gaussian pulses with parameters of 1030nm / 50fs / 1MHz and output power of 30mW.

[0047] Bessel beam generation module 2: The cone angle of the cone lens 21 is 2° and the focal length of the low dispersion lens 22 is 50mm, which converts the Gaussian pulse into a Bessel beam 9 with a central main lobe diameter of 3mm and annular side lobe width of 0.8mm.

[0048] Nonlinear detection module 3: The nonlinear medium is sapphire with a thickness of 10mm; the grating is a blazed grating with a diffraction efficiency of 96%.

[0049] Frequency domain shutter module 4: LCTF response time 500ns, passband adjustment range 1-10nm.

[0050] AI Predictive Control Module 5: Industrial computer integrating FPGA (sampling frequency 10MHz) and GPU.

[0051] Amplification Module 6: Yb:YAG solid-state slab amplifier, side + end face bidirectional pumping, pump wavelength 940nm, output power 100W.

[0052] Feedback detection module 7: Spectrometer sampling frequency 10MHz, energy detector accuracy 0.01mJ.

[0053] (2) Implementation process: Step 1: The femtosecond seed source 1 outputs a femtosecond Gaussian pulse 8 of 1030nm / 50fs / 1MHz, which enters the Bessel beam generation module 2 and is converted into a diffraction-free Bessel beam 9 by the conical lens 21 and the low-dispersion lens 22. Step 2: When the Bessel beam 9 passes through a 10mm thick sapphire crystal, the spectral width expands from 10nm to 10.2nm due to the self-phase modulation effect when the pulse energy fluctuates from 30mW to 30.03mW (fluctuation of 0.1%); when the energy fluctuates to 29.97mW (fluctuation of -0.1%), the spectral width narrows from 10nm to 9.8nm. Step 3: The grating unfolds the changed spectrum, the spectral detector collects the spectral width data, and transmits it to the FPGA of the AI ​​prediction control module 5; Step 4: The prediction model receives the spectral width signal, predicts the energy fluctuation, and the ADRC unit generates compensation instructions: when the energy is too high, the LCTF is driven to narrow the passband from 10nm to 9.8nm, cutting off the spectral edge portion; when the energy is insufficient, the passband is widened to 10.2nm, and the pump power of the femtosecond seed source 1 is finely adjusted to correct the energy fluctuation. The energy detector collects energy data and feeds it back to the AI ​​prediction control module 5 to form a closed-loop control, ensuring that the energy fluctuation is stable within 0.09%.

[0054] Step 5: The calibrated Bessel beam 9 enters the Yb:YAG solid-state slab amplifier and is amplified to 100W output.

[0055] (3) Implementation effect In this embodiment, under 100W and 1MHz operating conditions, the pulse energy jitter is 0.09%, the pulse width drift is 0.8fs after 1200 hours of continuous operation, and the beam quality M... 2 =1.25, which can meet the requirements of ultra-precision etching of SiC / GaN chips (line width ≤100nm), while the sapphire medium does not burn out, and the lifespan of the laser is increased by more than 30%.

[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention are within the scope of the claims of the present invention.

Claims

1. A femtosecond pulse energy self-stabilizing device based on Bessel beams and AI, characterized in that, The device comprises the following components arranged sequentially along the optical path: Femtosecond seed source (1) is used to output femtosecond Gaussian pulses (8); A Bessel beam generation module (2) is used to convert the femtosecond Gaussian pulse (8) into a diffraction-free Bessel beam (9). The nonlinear detection module (3) is used to convert the pulse energy fluctuation of the Bessel beam (9) into a change in spectral width and to unfold the changed spectrum. The frequency domain shutter module (4) is used to perform spectral trimming on the Bessel beam to achieve coarse correction of the pulse energy; Amplification module (6) is used to amplify the power of the corrected Bessel beam and output it; The feedback detection module (7) is used to acquire spectral width data and pulse energy data; The AI ​​prediction control module (5) is connected to the feedback detection module (7), the frequency domain shutter module (4) and the femtosecond seed source (1) respectively. It is used to run a preset AI algorithm to generate compensation instructions based on the collected spectral width data and pulse energy data, and drive the frequency domain shutter module (4) and the femtosecond seed source (1) to perform energy compensation.

2. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 1, characterized in that, The Bessel beam generation module (2) includes a conical lens (21) and a low-dispersion lens (22); wherein: The conical lens (21) is used to shape the Gaussian beam into a Bessel beam; The low-dispersion lens (22) is used to correct the dispersion introduced by the cone lens (21) and ensure the diffraction-free characteristics of the Bessel beam.

3. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 2, characterized in that, The cone angle of the cone lens (21) is 1-3°, the diameter of the central main lobe of the generated Bessel beam is 2-5mm, and the width of the annular side lobe is 0.5-1mm.

4. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 1, characterized in that, The nonlinear detection module (3) comprises a nonlinear medium and a grating; wherein: The nonlinear medium is fused silica or sapphire with a thickness of 5-20 mm, used to convert the pulse energy fluctuation of the Bessel beam (9) into a spectral width change through a self-phase modulation effect; The grating is a blazed grating with a diffraction efficiency greater than 95%, used to expand the changed spectrum.

5. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 1, characterized in that, The frequency domain shutter module (4) includes a high-speed liquid crystal tunable filter; the response time of the high-speed liquid crystal tunable filter is less than... The passband adjustment range is 1-10nm.

6. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 1, characterized in that, The feedback detection module (7) includes a spectral detector and an energy detector; wherein: The spectral detector is used to collect the expanded spectral width data, and the sampling frequency is not less than 10MHz; The energy detector is used to collect the final pulse energy data with an accuracy of 0.01 mJ.

7. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 1, characterized in that, The AI ​​prediction control module (5) includes an industrial control computer integrating FPGA and GPU; wherein: The FPGA is used to acquire spectral data in real time, with a sampling frequency greater than 10MHz; The GPU is used to run a neural network prediction model and an active disturbance rejection control algorithm. The neural network prediction model receives spectral width data and outputs a predicted fluctuation of pulse energy. The active disturbance rejection control algorithm generates a final compensation instruction to drive the frequency domain shutter module (4) and the femtosecond seed source (1) based on the predicted fluctuation and the current real pulse energy measured by the feedback detection module (7).

8. The femtosecond pulse energy self-stabilizing device based on Bessel beam and AI according to claim 1, characterized in that, The amplification module (6) is a Yb:YAG solid-state slab amplifier with a bidirectional pumping structure on the side and end face, and a pumping wavelength of 940nm.

9. A method for self-stabilizing the energy of femtosecond pulses based on Bessel beams and AI, characterized in that, The method includes: Step 1: Convert the femtosecond Gaussian pulse (8) output by the femtosecond seed source (1) into a non-diffraction Bessel beam (9) through the Bessel beam generation module (2). Step 2: Pass the Bessel beam (9) through the nonlinear medium, use the self-phase modulation effect to convert the pulse energy fluctuation into a change in spectral width, and unfold the changed spectrum through a grating; Step 3: Use a spectral detector to collect the spectral width data after grating expansion in real time, and transmit the data to the AI ​​prediction control module (5). Step 4: The AI ​​prediction control module (5) uses a neural network prediction model to predict the amount of energy fluctuation and generates compensation instructions through an active disturbance rejection control algorithm to drive the frequency domain shutter module (4) to perform spectral trimming on the Bessel beam and adjust the pump power of the femtosecond seed source (1) to achieve energy correction. Step 5: The corrected Bessel beam is amplified by the amplification module (6) and then output; Step 6: Use an energy detector to collect pulse energy data and feed it back to the AI ​​predictive control module (5) to form a closed-loop control.

10. The femtosecond pulse energy self-stabilization method based on Bessel beams and AI according to claim 9, characterized in that, The closed-loop control includes: When the pulse energy fluctuation exceeds the preset threshold, repeat step four until the stability requirement is met, thus forming a complete closed-loop control.

Images