Optical system full-automatic collimation method, system, medium and equipment

Abstract

The invention relates to the technical field of optical systems, and provides a full-automatic collimation method and system for an optical system, a medium and equipment, and the method comprises the steps: extracting pulse features of a pulse sequence, calculating pulse interval jitter, an energy fluctuation rate and a deviation between a pulse repetition rate and a theoretical value, and judging a mode locking state; for the output laser, reconstructing a spectral curve of the output laser, and extracting spectral features; obtaining a mode locking stability score through a deep learning model; the basic parameters of the laser cavity are optimized by maximizing the power and the collimation degree of the output laser; basic parameters of the laser cavity are used as priori, and parameters of all optical elements of the resonant cavity are adjusted by maximizing a mode locking stability score and a target pulse width score; and on the premise of the adjusted parameters of the optical elements of the resonant cavity, the pose of the lens is adjusted by maximizing the power and the collimation of the output laser. High-efficiency and progressive optimization of optical system parameters is realized, and the collimation degree is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical systems, and in particular relates to a fully automatic alignment method, system, medium and equipment for an optical system. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Currently, the construction and alignment of optical systems still rely primarily on manual operations by researchers, which is highly dependent on the technical skills of technicians and lacks unified operating methods and operating standards. This also consumes a lot of work. Therefore, there is a great demand for methods to automatically build and align optical systems.

[0004] Currently, there has been some research work on self-adjusting the alignment of optical devices. However, since many optical devices in use today are designed for manual manipulation by researchers, the R&D and production costs of redesigning fully automatic motorized optical devices are very high. Moreover, this solution cannot be carried out in collaboration with experienced researchers, making it difficult to implement.

[0005] More importantly, existing research on automatic mode-locking technology is mainly concentrated in the field of fiber lasers. Related technologies mostly use pulse sequence analysis combined with dispersion Fourier transform (DFT) for automatic mode-locking, which has a high degree of realization. However, for solid-state lasers, due to the significant differences in structure and working mechanism from fiber lasers, it is very difficult to directly transplant the above-mentioned automatic mode-locking technology. At present, the research on automatic mode-locking of solid-state lasers mainly focuses on using CCD to collect output spot images and distinguish the mode-locking state based on neural networks. However, this method is limited by image resolution, recognition accuracy and real-time performance, and has high requirements for the experimental environment. Therefore, the existing methods for automatic mode-locking of solid-state lasers have low efficiency in identifying and controlling the mode-locking state, and it is difficult to meet the requirements of high precision, high stability and high real-time performance. Summary of the Invention

[0006] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a fully automatic alignment method, system, medium and equipment for an optical system. After the locking state has been preliminarily classified, a quantitative evaluation is performed. On this basis, the entire optical device posture parameter space is divided into multiple optimization levels such as pump optical path, resonant cavity, output optical path, etc. according to the physical structure and function. The optimization relationship between the parameters and the target performance is established at each layer, thereby realizing efficient and progressive optimization of the optical system parameters and improving the collimation.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: A first aspect of the present invention provides a method for fully automatic alignment of an optical system, comprising: For the constructed optical system, obtain the pulse sequence of the output laser; Extract pulse characteristics from the pulse sequence, and based on the pulse characteristics, calculate the pulse interval jitter, energy fluctuation rate, and the deviation of the pulse repetition rate from the theoretical value to determine the mode-locking state; For the output laser, after time-domain stretching through the fiber dispersion unit, the fast Fourier transform combined with the dispersion mapping relationship is applied to reconstruct the spectral curve of the output laser and extract the spectral characteristics; Based on pulse characteristics, mode-locking status, spectral characteristics, pulse interval jitter, energy fluctuation rate, and the deviation of pulse repetition rate from theoretical value, a mode-locking stability score is obtained through a deep learning model; The basic parameters of the laser cavity are optimized by maximizing the power and collimation of the output laser. Taking the basic parameters of the laser cavity as a priori, the parameters of each optical component of the resonant cavity are adjusted by maximizing the locking mode stability score and the target pulse width score. Based on the adjusted parameters of each optical component of the resonant cavity, the lens position is adjusted by maximizing the power and collimation of the output laser.

[0008] Furthermore, the optical system is constructed using a collaborative robot.

[0009] Furthermore, the fiber dispersion unit is parameter optimized by a multi-objective genetic algorithm, and the objective function of the multi-objective genetic algorithm is to minimize the reconstruction error of the stretched time domain signal and maximize the spectral resolution.

[0010] Furthermore, the step of reconstructing the spectral information of the output laser includes: performing amplitude normalization and baseline correction on the time domain signal after time domain stretching, and then using a compensation algorithm to perform inverse time domain transformation to correct the signals of each frequency band based on the total dispersion parameters obtained by dispersive fiber calibration; applying fast Fourier transform to the corrected time domain signal to obtain the corresponding spectrum; and converting the spectrum into a spectral curve in actual physical units through a preset dispersion mapping relationship.

[0011] Furthermore, the mode-locking stability score is a weighted sum of a pulse interval jitter score, an energy fluctuation score, a spectral purity score, a pulse repetition frequency stability score, and a pulse shape consistency score.

[0012] Furthermore, the spectral characteristics include central wavelength, 3dB bandwidth, spectral profile and main peak energy ratio.

[0013] Furthermore, the pulse characteristics include pulse interval, inter-pulse jitter, intensity distribution, energy fluctuation rate, pulse envelope shape and consistency parameters, A second aspect of the present invention provides an optical system fully automatic collimation system, comprising: A data acquisition module is configured to: acquire a pulse sequence of an output laser from the constructed optical system; A mode-locking state judgment module is configured to: extract pulse characteristics from the pulse sequence, and based on the pulse characteristics, calculate the pulse interval jitter, energy fluctuation rate, and the deviation of the pulse repetition rate from the theoretical value to judge the mode-locking state; A spectral feature extraction module is configured to: for the output laser, after time domain stretching through the fiber dispersion unit, apply fast Fourier transform combined with dispersion mapping to reconstruct the spectral curve of the output laser and extract the spectral features; A mode-locking stability scoring module is configured to obtain a mode-locking stability score based on pulse characteristics, mode-locking status, spectral characteristics, pulse interval jitter, energy fluctuation rate, and deviation of the pulse repetition rate from the theoretical value through a deep learning model; The collimation module is configured to: optimize the basic parameters of the laser cavity by maximizing the power and collimation of the output laser; adjust the parameters of each optical component of the resonant cavity by maximizing the locking stability score and the target pulse width score based on the basic parameters of the laser cavity; and adjust the lens position by maximizing the power and collimation of the output laser based on the adjusted parameters of each optical component of the resonant cavity.

[0014] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for fully automatic alignment of an optical system.

[0015] The fourth aspect of the present invention provides a computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein when the processor executes the program, the steps in the method for fully automatic alignment of an optical system as described above are implemented.

[0016] Compared with the prior art, the present invention has the following beneficial effects: After the locking state has been preliminarily classified, the present invention conducts a quantitative evaluation. On this basis, the entire optical device posture parameter space is divided into multiple optimization levels such as pump optical path, resonant cavity, and output optical path according to physical structure and function. The optimization relationship between parameters and target performance is established at each level to achieve efficient and progressive optimization of optical system parameters and improve collimation.

[0017] The optical system of the present invention is constructed and preliminarily aligned using a collaborative robot, and can be directly adapted to optical devices currently commonly used in academia and industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0019] Figure 1 This is a flow chart of a fully automatic collimation method for an optical system according to a first embodiment of the present invention; Figure 2 It is a structural diagram of a computer device according to the fourth embodiment of the present invention. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0021] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0022] Example 1 This embodiment provides a fully automatic alignment method for an optical system.

[0023] This embodiment provides a fully automatic alignment method for an optical system, such as Figure 1 As shown, the following steps are included: Step 1: Optical system initialization and target setting.

[0024] Step 2: Collect engineer adjustment action data and train the deep learning model.

[0025] Step 3: After three-dimensional visual recognition of the optical components, the robot automatically builds the optical path system and performs preliminary alignment.

[0026] Step 301: Collect optical device information.

[0027] (1) Three-dimensional pose information acquisition: The three-dimensional vision module uses a high-precision depth camera or lidar to collect three-dimensional point cloud data of the optical device; the point cloud is processed by the PointNet++ deep learning network to extract the key pose information of the optical device, including: spatial coordinates: the center position of the optical device; rotation matrix: describes the angular information of the device; translation vector: describes the movement direction of the device.

[0028] Accuracy optimization: Use a denoising autoencoder based on self-supervised learning to optimize point cloud data, filter out noise points and improve pose recognition accuracy.

[0029] (2) Optical parameter detection: Detection range: The optical detection module covers spectral detection from ultraviolet to mid-infrared bands, and can monitor the spot area, light intensity distribution, spectral characteristics and pulse width of the output light path in real time; Feature extraction: For spectra, the Transformer-based Spectral Transformer network is used in combination with a multi-head attention mechanism to extract optical features, including the light intensity distribution map within the band range and the light spot energy distribution data.

[0030] (3) Information transmission and analysis: The collected three-dimensional pose and optical feature data are transmitted to the terminal processing control unit to form a multimodal input data stream, which serves as the basis for subsequent adjustment strategy calculations.

[0031] Step 302: Deep learning model calculation adjustment strategy.

[0032] (1) The deep learning model adopts a multi-task joint learning framework, which mainly consists of the following parts: Input layer: receives 3D pose information (point cloud feature vector) and optical features (spectrum and output parameters) as input; Feature extraction layer: Pose feature extraction: Use the PointNet++ model to process 3D point cloud data; Optical feature extraction: Use the Spectral Transformer model to extract spectral and pulse light features; Feature fusion layer: Uses a multi-head attention mechanism to fuse pose and optical features to generate a high-dimensional feature representation of the global state of the optical system.

[0033] Prediction layer: Outputs the adjustment strategy for each optical device, including rotation (Roll, Pitch, Yaw) and displacement (X, Y, Z).

[0034] Predict the output light parameters of the adjusted optical system for subsequent error evaluation.

[0035] (2) Dynamic optimization: Loss function design: Posture loss: calculates the difference between the target pose and the current pose of the device based on the Euclidean distance; Optical parameter loss: based on the mean square error (MSE) between the target spectrum, spot area, etc. and the actual value; Optimization method: Use the AdamW optimizer to update the parameters, and combine the learning rate scheduler to dynamically adjust the learning rate.

[0036] (3) Reinforcement learning-assisted optimization: During the adjustment process, deep reinforcement learning algorithms (such as Proximal Policy Optimization, PPO) are used to further optimize the adjustment path to ensure adjustment efficiency and accuracy.

[0037] Step 303: Adjust the position of the optical device.

[0038] Based on the adjustment parameters output by the deep learning model, the terminal processing control unit generates motion instructions for the flexible robotic arm. The instructions include position adjustment (X, Y, Z directions) and rotation adjustment (Roll, Pitch, Yaw), completing the construction of the optical path system and preliminary alignment.

[0039] Step 2: Automatic mode locking of solid-state laser and alignment of optical system based on pulse sequence analysis.

[0040] In step 2, an oscilloscope is used to obtain the spectral information of the output light of the constructed optical system in real time and at high speed. Compared with the traditional spectrometer detecting spectral information (more than ten seconds each time), the speed is much faster and can achieve real-time output.

[0041] In step 2, the solid-state laser is automatically mode-locked and aligned with the optical system by detecting the pulse sequence with an oscilloscope and combining it with the dispersion Fourier transform (DFT). The specific implementation steps are as follows: Step 201: Pulse sequence acquisition and processing.

[0042] High-speed oscilloscope acquisition system: A high-speed oscilloscope with a bandwidth of ≥20 GHz is used to acquire the pulse sequence output by the laser of the constructed optical system. The sampling rate reaches above 50 GSa / s, ensuring that ultrashort pulses in the range of picoseconds to femtoseconds can be accurately captured.

[0043] Pulse extraction algorithm: A multi-resolution analysis method based on wavelet transform is used to accurately extract the pulse characteristics of the pulse sequence from the noise background, including pulse interval, pulse energy, pulse repetition rate, intensity distribution and pulse envelope shape. The extracted pulse characteristics serve as the core parameters for subsequent mode-locking stability evaluation and deep learning model input.

[0044] Pulse stability assessment: Based on pulse characteristics such as pulse interval and energy (intensity) distribution, the timing jitter and energy fluctuation of N consecutive pulses (usually N ≥ 1000) are calculated, and the pulse repetition rate is compared with the theoretical repetition rate to achieve real-time assessment of mode-locking stability and accurate identification of mode-locking status.

[0045] Specifically, pulse stability assessment is performed using quantitative indicators. Multi-dimensional measurements include pulse interval jitter, energy fluctuation, and the deviation of the pulse repetition rate from the theoretical repetition rate. The specific implementation method is as follows: (1) Pulse interval jitter ( ).

[0046] Assume that the arrival time of N pulses collected continuously is , the measured pulse interval is , the theoretical cavity length round trip time is , then the pulse interval jitter is defined as follows: .

[0047] (2) Energy fluctuation rate ( ).

[0048] Assume that the energy of each pulse is , the average energy is , the energy fluctuation rate is defined as: .

[0049] (3) Deviation between pulse repetition rate and theoretical value ( ).

[0050] Measured main repetition rate , theoretical repetition rate , where c is the speed of light and L is the cavity length; the deviation is given by: .

[0051] In summary, the locking state can be determined based on the above indicators: (1) No locking state at all: and ,and .in, The pulse judgment threshold is set to 0.1 (i.e. 10%), which can be adjusted according to the specific laser cavity parameters. The energy judgment threshold is set to 0.1 (i.e. 10%), which can be adjusted according to the specific laser cavity parameters. is the repetition frequency deviation threshold, for example 0.10 (10%).

[0052] (2) Q-switched pulse state: no pulse output for a long time, occasional high energy and large interval ( ) pulse, which appears as a partial period Maximum (i.e., there is greater than the set value), the pulse energy is much higher than the average value (that is, the difference between the pulse energy and the average value is greater than the threshold).

[0053] (3) Multi-pulse (harmonic mode locking) state: the pulse interval is an integer component relationship, and the main repetition rate is an integer multiple of the cavity length round trip time. Lower than , but the spectrum features multiple peaks.

[0054] (4) Unstable mode locking state: and Between stable mode locking and no mode locking (typical value 0.02 < <0.1 , 0.05< <0.2), there is a main period drift phenomenon, and energy mutation phenomenon. Among the M (M≥1000) pulses measured, when ≥5% of the pulse intervals deviate from 0.02 < When the pulse energy deviates from the average value by 3% or more, it is judged as the main period drift. More than 0.02 When , it is determined to be an energy mutation.

[0055] (5) Stable locking state: satisfying , , , the main period shift and the energy mutation pulse ratio are both less than 1%.

[0056] Step 202: Reconstruct spectrum information based on real-time spectrum analysis using Dispersive Fourier Transform (DFT).

[0057] (1) Fiber dispersion unit: Using a specially designed dispersion structure (such as a chirped fiber Bragg grating), the output light of the optical system built in step 1 is coupled into a long-distance dispersion-compensating fiber (a dispersion unit with known dispersion parameters) to perform time-domain stretching on the ultrashort pulse and the output laser pulse, thereby realizing the dispersion Fourier transform (DFT) process. This process spreads the different spectral components of the laser pulse on the time axis, so that the spectral information forms a one-to-one correspondence with the time domain waveform. The dispersion parameters of the dispersion unit are precisely calibrated to ensure that the mapping from spectrum to time domain is reversible and highly accurate, providing a basis for subsequent spectral reconstruction.

[0058] As an embodiment, the stretching ratio may be preset to a fixed value; As another embodiment, the stretching ratio can automatically set the stretching ratio of the dispersion unit according to the spectral characteristics and resolution requirements to be detected, so as to achieve sub-nanometer spectral resolution. Specifically: according to the central wavelength, pulse width and output spectral width of different laser systems, dynamic switching and optimization are achieved through programmable or automatically adjustable multi-segment dispersion units to ensure that time domain stretching matches the signal acquisition bandwidth and resolution requirements, and is fully adaptable to different types of ultrafast laser systems.

[0059] The dispersion unit parameter optimization uses a multi-objective optimization algorithm to minimize the spectrum reconstruction error and ensure the signal integrity of the acquisition system. The specific process is as follows: Input: laser center wavelength, expected pulse width, output spectrum width, sampling bandwidth and other parameters; Objective function: Minimize the reconstruction error of the stretched time-domain signal, maximize the spectral resolution, and ensure that the pulse is not distorted. That is, ensure that the correlation coefficient between the pulse shape and the theoretical shape is greater than 0.95 and the pulse width variation rate is less than 5%.

[0060] Optimization method: Adopting an adaptive multi-objective genetic algorithm, the cumulative dispersion of each dispersion unit is dynamically adjusted to ensure that the final stretching ratio meets the performance requirements of different laser systems. Output: Optimal dispersion unit parameter configuration, used to automatically set the dispersion fiber segment and achieve adaptive matching of the system.

[0061] (2) Real-time spectral reconstruction: A high-speed oscilloscope is used to collect the time domain signal after being stretched by the dispersion unit. The collected signal is amplitude normalized and baseline corrected, and dispersion compensation is performed based on the calibration parameters of the dispersion unit. Subsequently, according to the dispersion mapping relationship, the corrected time domain signal is converted into a spectral curve to obtain the spectral characteristics such as the central wavelength, spectral bandwidth and spectral profile of the output laser. For scenarios requiring higher resolution, the fast Fourier transform (FFT) can be further applied to the corrected time domain signal to achieve high-precision reconstruction of the spectral information. The above process realizes high-speed and real-time acquisition of the output laser spectral information.

[0062] The specific steps of reconstructing the spectral information of the output light are as follows: For the time domain signal S(t) after dispersion fiber stretching, amplitude normalization and baseline correction are first performed; Based on the total dispersion parameter D obtained by dispersive fiber calibration, the compensation algorithm is used to correct the signal of each frequency band by inverse time domain transformation; Apply fast Fourier transform (FFT) to the corrected time domain signal to obtain the corresponding spectrum S(f); Through the preset mapping relationship, S(f) is converted into a spectral curve (central wavelength, bandwidth, etc.) under actual physical units to achieve real-time monitoring.

[0063] This embodiment achieves high-resolution analysis of the entire spectral range (eg, 750-850 nm) through single-shot pulse acquisition, without requiring a complex mechanical structure of a scanning spectrometer.

[0064] Step 203: Deep learning-assisted mode-locking state optimization.

[0065] In this embodiment, a deep learning model based on a hybrid architecture of convolutional neural network (CNN) and long short-term memory network (LSTM) is adopted to perform multi-dimensional comprehensive scoring on the mode-locking quality. This model takes the mode-locking state, time-domain pulse characteristics, spectral characteristics (spectral information), and the evaluation results of pulse stability as inputs to achieve real-time quantitative evaluation of the mode-locking state.

[0066] The input features include but are not limited to: Mode-locking state; Spectral characteristics (central wavelength, 3dB bandwidth, spectral profile, proportion of main peak energy, etc.) reconstructed by an oscilloscope and DFT algorithm; Time-domain characteristics (pulse interval, jitter between pulses , intensity distribution, energy volatility , pulse envelope shape, consistency parameter, etc.) analyzed by the "pulse extraction algorithm"; Evaluation results of pulse stability, including the deviation of pulse repetition rate from the theoretical cavity length and so on.

[0067] The comprehensive model normalizes the above features and then sends them into the neural network. After training, a set of scoring parameters is output, including: pulse interval jitter score , energy fluctuation score , spectral purity score , pulse repetition frequency stability score , pulse shape consistency score .

[0068] Each of the above scoring items is normalized to the range of 0 to 1. The larger the value, the better the mode-locking state.

[0069] The comprehensive score S can be obtained by weighted summation with preset weights: ; where ~ are the weights of each scoring component, which can be set according to system requirements or obtained adaptively through training.

[0070] The comprehensive score S is used to judge the mode-locking state in real time and is directly fed back to the hierarchical Bayesian optimization control module as the basis for adjustment decisions, improving the accuracy and efficiency of automatic mode-locking. The following grading criteria can be specifically set: for example, S>0.95 represents excellent mode-locking, 0.9<S≤0.95 represents generally acceptable mode-locking, and S≤0.9 represents mode-locking state to be optimized, etc. This scoring mechanism realizes the upgrade from single classification to multi-dimensional detailed quantitative evaluation of the mode-locking state, and can more accurately drive the subsequent automatic adjustment and fine-tuning optimization process.

[0071] Mode-locking stability scoring: The CNN- and LSTM-based mode-locking state classification network outputs a score based on the input multimodal pulse and spectral characteristics to comprehensively assess the current mode-locking quality. The neural network outputs scores for each of the aforementioned key dimensions, which serve as quantitative criteria to guide optical system adjustments in real time. This scoring provides real-time assessment of mode-locking quality. The purpose of this scoring is to provide a more detailed quantitative evaluation after the initial classification of the mode-locking state, providing optimization targets and criteria for multi-objective optimization and subsequent fine-tuning. This scoring complements the previously mentioned "pulse stability assessment": the former is used to initially determine whether mode-locking is established, while the latter (here) accurately quantifies the quality of the mode-locked pulse and guides robot adjustments.

[0072] Step 204: Multi-objective layered optical device adjustment strategy.

[0073] To improve overall parameter optimization efficiency and physical interpretability, this embodiment employs an adaptive hierarchical Bayesian optimization approach. This approach divides the entire optical device pose parameter space into multiple optimization levels, such as the pump optical path, resonant cavity, and output optical path, based on physical structure and function. At each level, a Bayesian optimization relationship is established between the parameters and target performance (e.g., mode-locked state, spectral width, output power, etc.). Historical data and real-time measurement information are incorporated as a priori information during the optimization process at each level, enabling efficient, progressive optimization.

[0074] The adaptive hierarchical Bayesian optimization process is as follows: (1) The optical system is divided into several optimization layers, including the pump optical path, resonant cavity, and output optical path, based on the physical and functional structure.

[0075] (2) The optimization process is carried out from the inside out and layer by layer: (a) First layer: Pump optical path optimization. This layer optimizes the basic laser cavity parameters (such as pump power and alignment angle) in the optical system to achieve continuous laser output. The objective function is to maximize the output power and collimation accuracy of the laser system and maximize the success probability of the subsequent mode locking process (i.e., maximize the comprehensive score S).

[0076] (b) Second layer: Optimization of the resonator optical elements (such as parallel mirrors, dispersion compensating mirrors, and output coupling mirrors), with mode-locking stability and pulse width as the primary evaluation targets. The mode-locking stability (comprehensive score) is calculated in steps 201-203. This layer uses the optimal parameters of the pump optical path as a priori input and the result as a fixed condition. Bayesian optimization is then used to further adjust the fine-tuning parameters of the resonator optical elements to achieve a global optimum for both the mode-locking stability score and the target pulse width score. Multiple rounds of sampling, evaluation, and updating are used to gradually converge to the optimal resonator configuration.

[0077] (c) The third layer: output optical path optimization, which primarily improves collimation quality and transmission efficiency. Output optical path optimization is premised on achieving the optimal solution for the mode-locked state in the previous layer. The objective function is to maximize the collimation of the output beam (e.g., minimizing the far-field divergence angle and circularity) and transmission efficiency (output power). Specifically, by collecting output beam quality indicators in real time and comparing them with the results of the optimal simulation model, the output lens position is continuously fine-tuned to achieve more precise performance output.

[0078] In each layer, the Bayesian optimization algorithm is used to find the optimal combination of parameters of this layer, and the optimization results of each layer are used as the prior for the optimization of the next layer, which are continuously updated to improve the global convergence speed and precision.

[0079] After all layers are optimized, the results are comprehensively evaluated, and global assessments and small-scale adjustments are made based on key performance indicators. If a local system disturbance is detected, the affected single-layer parameters can be immediately re-optimized to ensure long-term stability and environmental adaptability.

[0080] Step 205: Continuous optimization mechanism of closed-loop feedback.

[0081] (1) Real-time parameter monitoring: Continuously monitor the pulse sequence and reconstructed spectral information to provide real-time feedback. The optical detection module performs a comprehensive test on the adjusted optical path, including parameters such as spectral range, spot area, and pulse width. The test results are compared with the preset target values ​​to determine whether the optical path has achieved the optimization goal. If the output optical parameters do not reach the target value, the terminal processing control unit will feed back the error to the deep learning model, recalculate the adjustment strategy, and perform fine-tuning. This cycle continues until the optical path meets the preset requirements.

[0082] (2) Gradual fine-tuning: Specifically, starting with the optimal device parameters obtained through the hierarchical Bayesian optimization process, key performance indicators such as the laser output pulse sequence and spectral characteristics are monitored in real time. If environmental disturbances, random system fluctuations, or small parameter drifts are detected, submicron-level fine adjustments are immediately made to relevant key components (such as the resonator lens and output optical path lens) near the parameters obtained through the hierarchical optimization process.

[0083] Through this progressive fine-tuning based on the hierarchical optimization results, it is possible to dynamically respond to external and internal disturbances, maintaining and improving the overall performance and locking stability of the optical system.

[0084] Environmental Disturbance Compensation: By establishing a correlation model between environmental factors such as temperature, humidity, and vibration and system drift, this system actively compensates for the optical device position and laser output parameters executed by the flexible collaborative robot in real time. Specifically, this system monitors environmental parameters and dynamically corrects optical device adjustment instructions and optical system control strategies (such as automatically fine-tuning device position) to reduce error drift caused by environmental changes and improve system robustness and accuracy.

[0085] Self-learning capability: Record the effect of each adjustment, continuously update and improve the internal model, and improve the efficiency and accuracy of subsequent adjustments.

[0086] Compared to existing solutions that rely on expensive spot imaging systems or complex spectral analysis instruments, this invention utilizes high-speed oscilloscopes, DFT, and deep learning analysis to significantly reduce overall costs while enhancing the intelligence and automation of mode-locking operations. Data and decision-making within each algorithmic step form a continuous closed loop, ensuring real-time operation, robustness, and technical scalability.

[0087] This embodiment provides a fully automatic alignment method for an optical system, which has achieved remarkable results: Improved mold locking reliability: Compared with traditional manual adjustment methods, the automatic mold locking success rate has increased from approximately 70% to over 95%, significantly reducing mold locking failures.

[0088] Improved time efficiency: The adjustment time of the complete Ti:Sapphire laser system is shortened from 3-5 hours with traditional methods to 30-45 minutes, which is about 5-6 times more efficient.

[0089] Spectral accuracy assurance: Through DFT technology, a sub-nanometer 0.2nm spectral resolution is achieved, which can accurately monitor and control the laser output spectral characteristics.

[0090] Pulse width control accuracy: The pulse width can be controlled within ±5% of the target value. For a target pulse width of 100 fs, the actual control accuracy reaches ±5 fs.

[0091] Adaptability and Compatibility: The system is successfully compatible with a variety of commercial optical components and laser systems, including solid-state lasers with different gain media, such as Ti:Sapphire, Yb:YAG, and Er:fiber. Furthermore, due to its environmental perturbation compensation mechanism, the system can adapt to working environments with greater environmental variability, compared to some current solid-state laser automatic mode-locking methods, which can only operate in laboratory environments with constant temperature and humidity and isolation from environmental disturbances.

[0092] Example 2 This embodiment provides an optical system fully automatic collimation system, comprising: A data acquisition module is configured to: acquire a pulse sequence of an output laser from the constructed optical system; A mode-locking state judgment module is configured to: extract pulse characteristics from the pulse sequence, and based on the pulse characteristics, calculate the pulse interval jitter, energy fluctuation rate, and the deviation of the pulse repetition rate from the theoretical value to judge the mode-locking state; A spectral feature extraction module is configured to: for the output laser, after time domain stretching through the fiber dispersion unit, apply fast Fourier transform combined with dispersion mapping to reconstruct the spectral curve of the output laser and extract the spectral features; A mode-locking stability scoring module is configured to obtain a mode-locking stability score based on pulse characteristics, mode-locking status, spectral characteristics, pulse interval jitter, energy fluctuation rate, and deviation of the pulse repetition rate from the theoretical value through a deep learning model; The collimation module is configured to: optimize the basic parameters of the laser cavity by maximizing the power and collimation of the output laser; adjust the parameters of each optical component of the resonant cavity by maximizing the locking stability score and the target pulse width score based on the basic parameters of the laser cavity; and adjust the lens position by maximizing the power and collimation of the output laser based on the adjusted parameters of each optical component of the resonant cavity.

[0093] It should be noted here that the various modules in this embodiment correspond one-to-one to the various steps in Example 1, and the specific implementation processes are the same, which will not be repeated here.

[0094] Example 3 This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of the method for fully automatic alignment of an optical system as described in the first embodiment above are implemented.

[0095] Example 4 This embodiment provides a computer device, such as Figure 2 As shown, the present invention includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001. The processor 1001, the communication interface 1002, and the computer-readable storage medium 1003 may be connected via a bus or other means. The communication interface 1002 is configured to receive and transmit data, and when the processor 1001 executes the program, the steps of the method for fully automatic alignment of an optical system as described in the first embodiment are implemented.

[0096] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A fully automatic alignment method for an optical system, characterized in that: include: For the constructed optical system, obtain the pulse sequence of the output laser; Extract pulse characteristics from the pulse sequence, and based on the pulse characteristics, calculate the pulse interval jitter, energy fluctuation rate, and the deviation of the pulse repetition rate from the theoretical value to determine the mode-locking state; For the output laser, after time-domain stretching through the fiber dispersion unit, the fast Fourier transform combined with the dispersion mapping relationship is applied to reconstruct the spectral curve of the output laser and extract the spectral characteristics; Based on pulse characteristics, mode-locking status, spectral characteristics, pulse interval jitter, energy fluctuation rate, and the deviation of pulse repetition rate from theoretical value, a mode-locking stability score is obtained through a deep learning model; The basic parameters of the laser cavity are optimized by maximizing the power and collimation of the output laser. Taking the basic parameters of the laser cavity as a priori, the parameters of each optical component of the resonant cavity are adjusted by maximizing the locking mode stability score and the target pulse width score. Based on the adjusted parameters of each optical component of the resonant cavity, the lens position is adjusted by maximizing the power and collimation of the output laser.

2. The method for fully automatic alignment of an optical system according to claim 1, wherein: The optical system is constructed using a collaborative robot.

3. The method for fully automatic alignment of an optical system according to claim 1, wherein: The fiber dispersion unit is parameter optimized by a multi-objective genetic algorithm, wherein the objective function of the multi-objective genetic algorithm is to minimize the reconstruction error of the stretched time domain signal and to maximize the spectral resolution.

4. The method for fully automatic alignment of an optical system according to claim 1, wherein: The step of reconstructing the spectral information of the output laser includes: performing amplitude normalization and baseline correction on the time domain signal after time domain stretching, and then correcting the signals of each frequency band using a compensation algorithm and time domain inverse transformation based on the total dispersion parameter obtained by dispersive fiber calibration; applying a fast Fourier transform to the corrected time domain signal to obtain the corresponding spectrum; and converting the spectrum into a spectral curve in actual physical units through a preset dispersion mapping relationship.

5. The method for fully automatic alignment of an optical system according to claim 1, wherein: The mode-locking stability score is a weighted sum of a pulse interval jitter score, an energy fluctuation score, a spectral purity score, a pulse repetition frequency stability score, and a pulse shape consistency score.

6. The method for fully automatic alignment of an optical system according to claim 1, wherein: The spectral characteristics include central wavelength, 3dB bandwidth, spectral profile and main peak energy ratio.

7. The method for fully automatic alignment of an optical system according to claim 1, wherein: The pulse characteristics include pulse interval, inter-pulse jitter, intensity distribution, energy fluctuation rate, pulse envelope shape and coherence parameters.

8. An optical system fully automatic collimation system, characterized in that: include: A data acquisition module is configured to: acquire a pulse sequence of an output laser from the constructed optical system; A mode-locking state judgment module is configured to: extract pulse characteristics from the pulse sequence, and based on the pulse characteristics, calculate the pulse interval jitter, energy fluctuation rate, and the deviation of the pulse repetition rate from the theoretical value to judge the mode-locking state; A spectral feature extraction module is configured to: for the output laser, after time domain stretching through the fiber dispersion unit, apply fast Fourier transform combined with dispersion mapping to reconstruct the spectral curve of the output laser and extract the spectral features; A mode-locking stability scoring module is configured to obtain a mode-locking stability score based on pulse characteristics, mode-locking status, spectral characteristics, pulse interval jitter, energy fluctuation rate, and deviation of the pulse repetition rate from the theoretical value through a deep learning model; The collimation module is configured to: optimize the basic parameters of the laser cavity by maximizing the power and collimation of the output laser; adjust the parameters of each optical component of the resonant cavity by maximizing the locking stability score and the target pulse width score based on the basic parameters of the laser cavity; and adjust the lens position by maximizing the power and collimation of the output laser based on the adjusted parameters of each optical component of the resonant cavity.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the fully automatic alignment method of an optical system according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein: When the processor executes the program, the steps of the fully automatic alignment method for an optical system according to any one of claims 1 to 7 are implemented.

Images