A mid-infrared ultra-short pulse laser state automatic identification and switching device and method

Abstract

A kind of mid-infrared ultra-short pulse laser state automatic identification and switching device and method, device: the pump input arm of combiner is connected with pump source, one end of gain fiber is connected with the signal output arm of combiner, one end of single-mode fiber is connected with the other end of gain fiber, one input arm of output coupler is connected with the other end of single-mode fiber, its one output arm is connected with spectrometer, the input end of isolator is connected with the other output arm of output coupler, the input end of polarizer is connected with the output end of isolator, the input end of electric polarization controller is connected with the output end of polarizer, its output end is connected with the signal input arm of combiner.Method: spectrometer receives the output signal of mode-locked laser, and sends to control terminal, control terminal is optimized according to the matching degree of output laser pulse state and set target pulse state Control parameter.This device and method can realize the automatic adjustment and switching of noise-like mode-locked pulse and soliton mode-locked pulse.

Description

Technical Field

[0001] This invention relates to an automatic identification and switching device for mid-infrared ultrashort pulse laser status, belonging to the field of mode-locked laser and automatic control technology. Background Technology

[0002] Ultrafast fiber lasers have important applications in communication, sensing, and precision machining. How to improve the power and energy of laser pulses, understand and utilize the complex nonlinearities of ultrashort pulse fiber lasers, and conduct key technology research on high-energy, high-peak-power all-fiber ultrashort pulse laser applications has become an important scientific problem in fiber pulse lasers.

[0003] Mode-locking is one of the important means to realize ultrafast lasers. Nonlinear polarization-rotation mode-locked lasers are a type of passively mode-locked fiber laser, attracting much attention due to their simple structure. Compared with other mode-locked lasers, they firstly possess an all-fiber structure, eliminating the need for external electrical signals for modulation. This structure not only enables all-optical transmission but also allows them to operate without modulation bandwidth limitations, achieving mode-locked pulse output over a wide wavelength range. Furthermore, their response speed is determined by the polarization origin of the fiber's nonlinear effects, with output pulse widths reaching the femtosecond level. These superior properties exhibited by mode-locked lasers have led to extensive research on them.

[0004] Noise-like mode-locking (NMR) is a special type of pulse generated by mode-locked lasers under certain conditions. It features high pulse envelope energy, high average power, low coherence, and good environmental stability. NMR pulses not only have applications in fundamental physics research but also hold significant value in practical production due to their unique properties. Low coherence is an inherent characteristic of NMR pulses, which is particularly advantageous for typical low-coherence spectral interferometry techniques such as fiber grating demodulation and fiber optic information storage and reproduction. Utilizing dispersion management techniques, Raman effects, birefringence, and soliton self-frequency shifting, NMR pulses can also achieve ultrawide flat spectra. In recent years, with the development of fiber amplification technology and the continuous increase in pump pulse power, the experimentally achievable NMR pulse energy has been steadily rising, with single-pulse energies of several hundred nanojoules now achievable in fiber lasers. High-energy NMR pulses can be effectively applied not only to machining (micromachining) operations but also as pump sources to generate supercontinuum spectra.

[0005] Soliton mode-locking is a special type of pulse generated by a mode-locked laser under certain conditions. The formation of optical solitons in optical fibers is based on the balance between the anomalous group velocity dispersion (GVD) and self-phase modulation (SPM) in the fiber's nonlinear effects. Furthermore, optical soliton pulses can be transmitted over long distances in optical fibers without distortion. Due to these characteristics, optical soliton communication systems have significant advantages in ultra-long-distance, ultra-high-capacity optical communication. Both theoretical analysis and experimental research on optical solitons are still ongoing. Various types of optical solitons and their applications are being explored and studied in fields such as optical communication and signal processing.

[0006] In a soliton mode-locked laser, the soliton pulse circulates within the laser resonant cavity. If the effects of dispersion and nonlinearity in the cavity are weak, the pulse will experience some weak periodic perturbations due to the discrete nature of dispersion and nonlinearity. Some additional interference may also be due to periodic losses or amplification within the resonant cavity. This periodic interference causes the soliton to propagate along with the dispersive wave. This often does not have a significant impact because only the soliton experiences nonlinearity, causing the relative phase between the soliton and the dispersive wave to change continuously. However, in fiber lasers, resonant coupling (or quasi-phase matching) sometimes occurs, causing the phase between the dispersive wave and the optical soliton to change by integer multiples of 2π in each cycle at certain frequencies. This manifests as a series of paired narrow peaks in the soliton spectrum, known as Kelly bands. In the Fourier transform-limited soliton state, the distance between the Kelly bands reflects the magnitude of dispersion within the cavity.

[0007] However, polarization control of mode-locked fiber lasers based on the nonlinear polarization rotation effect has always been problematic. Current control methods are mostly manual, and achieving mode-locked laser output depends heavily on the user's experience. Furthermore, mode-locked fiber lasers based on the nonlinear polarization rotation effect are highly sensitive to temperature and noise, and control parameters under different conditions cannot guarantee consistent output. Even a stable mode-locked fiber laser can lose its mode-locked state when the environment changes significantly, such as with an increase in temperature. Once lost, it requires professional maintenance personnel to restore its state, greatly limiting the use of mode-locked fiber lasers in both military and civilian fields.

[0008] To address the aforementioned issues, a few experiments utilizing electrically controlled polarization for automatic mode-locking have been reported in recent years. U. Andral et al. from the University of Burgundy and RI Woodward et al. from Imperial College London have successively achieved automatic mode-locking using genetic algorithms combined with electrically controlled polarization technology. However, the former's experimental structure is cumbersome, requiring two electrically controlled polarization controllers and six voltage controls; the latter's mode-locking recognition process is extremely complex, requiring the simultaneous use of time-domain, frequency-domain, and spectral information for comprehensive recognition. Furthermore, the aforementioned methods all target the control and recognition of noise-like mode-locking in a single mode-locking state. Chinese invention patent CN108539571A, published on September 14, 2018, proposes a fast automatic mode-locking method encompassing multi-state pulse recognition. It uses a high-speed oscilloscope as the detection device and adjusts the polarization state according to an optimized algorithm, solving the polarization control problem in passively mode-locked lasers based on nonlinear polarization evolution. The signal detected by the high-speed oscilloscope remains a time-domain signal; both noise-like mode-locking and soliton mode-locking are in the mode-locked state, exhibiting highly consistent time-domain characteristics. However, noise-like pulses and soliton mode-locking share the same pulse envelope. Limited by the low bandwidth of photodetectors (especially in the 2-3 micrometer mid-infrared band, where the bandwidth of time-domain detectors is strictly limited; the maximum bandwidth of a 2-micrometer commercially available detector is 22 GHz, and the maximum bandwidth of a 3-micrometer detector is limited to 1 GHz), it is impossible to directly distinguish between these two states using existing photodetector systems. Chinese invention patent CN106329303A, published on January 11, 2017, proposes an automatic mode-locking control method for an automatic mode-locked fiber laser. This method suggests that the current operating state of the laser can be determined by counting and level detection of the signal from the photodetector, but it still cannot effectively identify and control noise-like mode-locking and soliton mode-locking.

[0009] Therefore, there is an urgent need to develop an intelligent identification method that can achieve precise polarization control and effectively identify noise-like pulses and soliton pulses. This method would solve the problem of automatic identification and switching between noise-like mode-locked states and soliton mode-locked states, enabling different ultrafast laser outputs and thus expanding the application scenarios of mode-locked lasers. Summary of the Invention

[0010] To address the problems existing in the prior art, this invention provides an automatic identification and switching device and method for mid-infrared ultrashort pulse laser states. The device has a compact structure and can output high-power, high-beam-quality, high-efficiency, and high-stability ultrashort pulse lasers. Simultaneously, it facilitates the automatic adjustment and switching of noise-like mode-locked pulses and soliton mode-locked pulses. The method can automatically and accurately identify the states of noise-like pulses and soliton pulses, solving the problem of state misjudgment in mid-infrared ultrafast fiber laser pulse applications, and enabling automatic adjustment and switching between the two pulse states.

[0011] To achieve the above objectives, the present invention provides an automatic identification and switching device for mid-infrared ultrashort pulse laser status, comprising a pump source, a mode-locked laser, a spectrometer, and a control terminal;

[0012] The mode-locked laser consists of a beam combiner, a gain fiber, a single-mode fiber, an output coupler, an isolator, a polarizer, and an electrically powered polarization controller.

[0013] The pump input arm of the combiner is connected to the output end of the pump source via a passive optical fiber. One end of the gain fiber is connected to the signal output arm of the combiner. One end of the single-mode fiber is connected to the other end of the gain fiber. One input arm of the output coupler is connected to the other end of the single-mode fiber, and one output arm is connected to the input end of the spectrometer. The input end of the isolator is connected to the other output arm of the output coupler. The input end of the polarizer is connected to the output end of the isolator. The input end of the electric polarization controller is connected to the output end of the polarizer, and its output end is connected to the signal input arm of the combiner.

[0014] The control terminal is connected to the pump source, the electric polarization controller, and the spectrometer, respectively.

[0015] As a preferred embodiment, the control terminal is a computer.

[0016] As a preferred embodiment, the spectrometer has a measurable range of 1200nm to 2400nm, can measure optical power of -70dBm, can achieve dynamic measurement of 55dB at a resolution of 0.05nm, has a maximum scan speed of 0.5s over a 100nm span, and uses a logarithmic coordinate mode. The spectrometer is equipped with GPIB, RS232 and Ethernet interfaces, supporting remote control and data acquisition by an external PC.

[0017] As a preferred embodiment, the electric polarization controller is a paddle-based polarization controller with an outer diameter of 18 mm for the winding disk, each paddle can rotate 170°, and the minimum step size is 0.12°, which can cover any polarization state on the Bonga sphere.

[0018] As a preferred embodiment, the pump source supports external level signal control of pump power output, with an adjustable power range of 0–30W.

[0019] As a preferred embodiment, the gain fiber 3 is a thulium-doped gain fiber.

[0020] In this invention, the pump light and signal light can be coupled into the same optical fiber by using a beam combiner; the gain fiber can be used to excite micrometer-wavelength laser light after receiving light from the pump source; the output coupler can split the light in the fiber ring cavity into two beams, one beam for easy spectral data acquisition by the spectrometer, and the other beam continues to oscillate and provide feedback within the cavity; the isolator and polarizer enable unidirectional transmission of the pump light and signal light within the fiber ring cavity. The electric polarization controller alters the saturable absorption effect caused by nonlinear polarization rotation within the fiber ring cavity, thereby achieving stable mode-locking. Connecting the spectrometer to one output arm of the output coupler facilitates sampling of the output signal to obtain spectral data; simultaneously, connecting the spectrometer to a control terminal allows the spectrometer to send the obtained spectral data to the control terminal for data processing. The control terminal is also connected to both the pump source and the electric polarization controller. This not only facilitates the control terminal in sending control parameters to the pump source and the electric polarization controller for execution, but also allows for easy switching between noise-like mode-locked pulses and soliton mode-locked pulses by controlling the electric polarization controller. Furthermore, this connection method forms a closed-loop feedback control structure, which facilitates the adjustment of the laser's output signal state, promoting stable laser operation at the target state. This compact device can achieve the output of high-power, high-beam-quality, high-efficiency, and high-stability ultrashort pulse lasers, while also facilitating the automatic adjustment and switching of noise-like mode-locked pulses and soliton mode-locked pulses.

[0021] This invention also provides a method for automatic identification and switching of mid-infrared ultrashort pulse laser status, comprising the following steps:

[0022] Step 1: Select the target pulse through the control terminal and control the mode-locked laser to start working;

[0023] Step 2: Use a spectrometer to receive the output signal of the mode-locked laser and send the sampled spectral data to the control terminal;

[0024] Step 3: The control terminal uses a trained neural network model to identify the mode-locked state of the spectral data and calculates the matching degree between the measured pulse spectrum and the target spectrum. If the matching degree between the output laser pulse state and the set target pulse state is greater than 80%, it is considered to be in agreement, and Step 5 is executed directly; if the matching degree between the output laser pulse state and the set target pulse state is less than or equal to 80%, it is considered to be in agreement, and Step 4 is executed directly.

[0025] Step 4: The control terminal generates corresponding control parameters based on the genetic optimization algorithm and sends the control parameters to the pump source and the electric polarization controller via the serial communication protocol. The control parameters include the pump power of the pump source and the blade angle of the electric polarization controller.

[0026] The specific method of the genetic optimization algorithm is as follows:

[0027] S1: The control terminal randomly generates 20 sets of initial control parameters and sends them to the pump source and the electric polarization controller respectively, so that the pump source and the electric polarization controller execute the received control parameters.

[0028] S2: Calculation, using the control terminal to calculate the matching degree between the spectral data corresponding to the 20 sets of initial control parameters in S1 and the target pulse;

[0029] S3: Sorting, using the control terminal to sort the 20 sets of initial parameters according to the degree of matching;

[0030] S4: Select the top 6 sets of control parameters with the highest matching degree using the control terminal;

[0031] S5: Genetics, using the control terminal to randomly cross-combine 6 sets of parameters to generate 10 new sets of control parameters;

[0032] S6: Mutation. In the newly generated control parameters in S5, the control terminal randomly selects 4 sets of control parameters and randomly modifies the values ​​of the control parameters to generate 20 new sets of control parameters.

[0033] S7: Send the newly generated 20 sets of control parameters to the pump source and electric polarization controller via the control terminal;

[0034] Step 5: If the output laser pulse state matches the target pulse, then pulse state monitoring is performed. Pulse state monitoring refers to comparing the current pulse state with the set target pulse state at regular intervals and determining whether the current output matches the set value. If the output laser state changes abruptly and does not match the target pulse, then jump to step 3 and start searching for the target state again.

[0035] Step Six: If the control terminal switches the target pulse, proceed to Step Two;

[0036] Furthermore, in order to effectively improve the recognition speed and accuracy, in step three, the neural network model is trained with a total of 1000 samples, all of which are spectral data. Among them, 600 are simulation data and 400 are experimental data. Of these, 300 are noise-like mode-locked pulse data, 300 are soliton mode-locked pulse data, and the rest are mode-free state data.

[0037] As a preferred embodiment, in step one, the target pulse is a selected noise mode-locked pulse or a soliton mode-locked pulse.

[0038] This invention utilizes an electric polarization controller as the polarization control device in a mode-locked laser, and a spectrometer as the measuring device to receive the output signal of the mode-locked laser. Simultaneously, a neural network model on the control terminal identifies the mode-locking state of the sampled spectral data. The matching degree between the output laser pulse state and the set target pulse state determines whether optimization processing is required. The optimization process employs a genetic algorithm, which effectively ensures the laser output signal remains stable in the target state through a closed-loop feedback structure and effective state identification. Furthermore, the electric polarization controller can control the switching between noise-like mode-locked pulses and soliton mode-locked pulses. Since noise-like mode-locked pulses and soliton mode-locked pulses have highly consistent time-domain characteristics, they are difficult to distinguish effectively using only an oscilloscope or photodetector. This invention, through enhanced training of the spectral state of ultrafast laser pulses using a convolutional neural network, can automatically achieve accurate and efficient identification of noise-like pulse and soliton pulse states, solving the problem of state misjudgment in mid-infrared ultrafast fiber laser pulse applications and realizing automatic adjustment and switching between the two pulse states. The neural network training algorithm employed in this invention offers advantages such as fast recognition speed, high recognition accuracy, and intelligence, significantly improving the reliability of mid-infrared ultrafast fiber lasers in industrial applications. The genetic algorithm used for intelligent control of laser pulse states enables rapid and precise switching between various laser states, including complex noise-like pulse states and soliton states. Compared to existing manual control methods, the resulting laser states are more efficient, significantly reducing the cost of manual control. Compared to existing mid-infrared ultrafast laser systems, this invention uses ordinary commercial single-mode fiber instead of the expensive polarization-maintaining fiber used in mainstream methods, greatly reducing costs. The intelligent recognition and control method employed in this invention also eliminates the need for complex temperature control, mechanical vibration control, and environmental disturbance prevention control systems, allowing for rapid adjustment according to environmental changes and greatly enhancing environmental adaptability. The fully automatic intelligent control system can achieve various pulse outputs, including multiple types of noise, solitons, and harmonic mode-locked solitons, and can also generate various customized laser pulse outputs according to customer needs. This greatly expands the application range of the laser system, and the rapid switching between different pulse states will solve the problem of high-precision laser processing of complex structural materials. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the structure of the present invention;

[0040] Figure 2 This is a schematic diagram of setting noise-like mode-locking as the target pulse on the LabVIEW platform.

[0041] Figure 3 This is the noise-like mode-locked pulse spectral measurement signal in this invention;

[0042] Figure 4This is the noise-like mode-locked pulse time-domain measurement signal in this invention;

[0043] Figure 5 A schematic diagram showing how to set soliton mode-locking as the target pulse on the LabVIEW platform.

[0044] Figure 6 This is the soliton mode-locked pulse spectral measurement signal in this invention;

[0045] Figure 7 It is the soliton mode-locked pulse time-domain measurement signal in this invention.

[0046] In the diagram: 1. Pump source, 2. Bundle combiner, 3. Gain fiber, 4. Single-mode fiber, 5. Output coupler, 6. Isolator, 7. Polarizer, 8. Motorized polarization controller, 9. Spectrometer, 10. Control terminal. Detailed Implementation

[0047] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0048] Ultrafast fiber lasers can generate ultrashort pulses with high peak power and wide spectrum, playing an indispensable role in both basic scientific research and industrial applications. For example, they are used as a light source for nonlinear optical imaging systems in the biomedical field; and high-energy short-pulse fiber lasers can be used for laser-based fine microstructure processing.

[0049] Mode locking is one method for generating ultrashort pulses in lasers. Similar to Q-switching, mode locking also modulates the laser cavity, splitting the originally continuous wave (CW) to generate pulses. However, mode locking and Q-switching are only similar in physical phenomena; their underlying principles are completely different. Mode locking techniques are generally divided into active mode locking and passive mode locking. Active mode locking uses a method of periodically modulating the resonant cavity parameters. This is based on using a modulator within the resonant cavity, controlled by an external signal, to periodically change the cavity loss or optical path (amplitude modulation and phase modulation) at a certain modulation frequency. When the selected modulation frequency is equal to the interval between longitudinal modes, the modulation of each mode will generate sidebands with frequencies consistent with the frequencies of two adjacent longitudinal modes. Due to the interaction between the longitudinal modes, all modes achieve synchronization under sufficiently strong modulation, and the modes will coherently superimpose to form a mode-locked pulse sequence. Another effective method for generating ultrashort pulses is passive mode locking. This method involves placing a saturable absorber within the laser resonant cavity. A saturable absorber is a nonlinear absorbing medium that exhibits absorption transitions at laser frequencies and has a large absorption cross-section. When a laser pulse is incident on this absorber, the absorber molecules absorb the laser radiation. As the laser intensity increases, the number of particles in its upper energy level also increases. When the laser intensity exceeds the absorber's saturation intensity, the absorber becomes saturated, allowing the most intense laser pulse to pass through it freely with minimal loss, thus obtaining a very strong mode-locked pulse. It is similar to a passive Q-switch, but with some differences. Passive mode-locking requires the saturable absorber to have an extremely short upper energy level lifetime and must be placed close to a total reflection mirror within the cavity.

[0050] Nonlinear polarization rotation mode-locking is a type of passive mode-locking. Nonlinear polarization rotation (NPR), also known as NPR, uses a typical unidirectional annular cavity. The mode-locking element is based on the nonlinear polarization rotation effect of light in weakly birefringent optical fibers and two polarization controllers or waveplate groups (polarizer and analyzer) with mutually perpendicular polarization states. By adjusting the waveplates or polarization controllers, the polarization state of the light can be changed from elliptical to linear and then back to elliptical. During pulse transmission, the phase shift induced by self-phase modulation and cross-phase modulation effects acts on the orthogonal polarization components, causing their polarization vectors to rotate continuously along the fiber. Due to the intensity dependence of the nonlinear polarization rotation angle, when the pulse intensity is low, the polarization vector hardly rotates as the nonlinear effect is small, and the transmission loss is high because the polarization states of the two polarization controllers are perpendicular. However, when the central portion of the light pulse with higher intensity passes through the fiber, the nonlinear effect is strong, the polarization vector rotates, and thus passes through the polarizer with less intensity. The result is high loss at the edges of the low-intensity pulse and low loss at the center of the high-intensity pulse, repeating this cycle and narrowing the pulse. Although femtosecond-level all-fiber mode-locked laser pulses have been achieved using nonlinear polarization rotation technology, the polarization-dependent nature of pulse formation and output, coupled with the extreme sensitivity of polarization states to the environment, causes instability in the nonlinear polarization rotation environment, significantly limiting its commercial production. A major constraint is that the optimal polarization setting changes with temperature, requiring readjustment, making it difficult to quantify and achieve specific modulation depths and saturation powers experimentally.

[0051] Noise-like pulses (NLPs) are special pulses generated by mode-locked lasers under certain conditions, characterized by high energy, wide pulse width, and low coherence. NLPs not only have applications in fundamental physics research but also possess significant practical value in manufacturing due to their unique properties. Low coherence is an inherent characteristic of NLPs, which is particularly advantageous for typical low-coherence spectral interferometry techniques such as fiber grating demodulation and fiber optic information storage and reproduction. Utilizing dispersion management techniques, Raman effects, birefringence, and soliton self-frequency shifting, NLPs can also achieve ultrawide flat spectra. In recent years, with the development of fiber amplification technology and the continuous increase in pump pulse power, the experimentally achievable energy of NLPs has been steadily rising; single-pulse energies of several hundred nanojoules can now be achieved in fiber lasers. High-energy NLPs can be effectively applied not only to machining (micromachining) but also as pump sources to generate supercontinuum spectra.

[0052] In most cases, noise-like pulses cannot be considered ideal high-energy pulses due to their wide substrate and low signal-to-noise ratio. However, mode-locking using nonlinear fiber ring mirrors in the anomalous dispersion region can produce noise-like pulses with a high signal-to-noise ratio. These pulses have enormous application potential in high-energy femtosecond pulses, with applications extending to lidar, electric field-molecular interactions, and other fields. Theoretically, dispersion management techniques can also be used to stabilize and control the dynamic evolution of nonlinear optical systems.

[0053] A soliton, also known as a solitary wave, is a special form of ultrashort pulse, or a pulsed traveling wave whose shape, amplitude, and velocity remain constant during propagation. An optical soliton is such a light pulse that can propagate in optical fibers while maintaining its shape, amplitude, and velocity for an extended period. The properties of optical solitons enable ultra-long-distance, ultra-high-capacity optical communication. In soliton mode-locked lasers, the soliton pulse circulates within the laser resonant cavity. If the effects of dispersion and nonlinearity within the resonant cavity are weak, the pulse will experience some weak periodic perturbations due to the discrete nature of dispersion and nonlinearity. Some additional interference may also be due to periodic losses or amplification within the resonant cavity. In fiber lasers, resonant coupling (or quasi-phase matching) sometimes occurs, causing the phase between the dispersive wave and the optical soliton to change by an integer multiple of 2π in each cycle at certain frequencies.

[0054] The characteristics of optical solitons determine their promising applications in the field of communications. The fundamental-order optical soliton is typically used for communication because it remains unchanged throughout propagation. Optical soliton communication has the following characteristics:

[0055] (1) Large capacity: The transmission rate can generally reach 20Gb / s, and can reach more than 100Gb / s;

[0056] (2) Low bit error rate and strong anti-interference ability: The fundamental optical soliton remains unchanged during transmission and the thermal insulation characteristics of the optical soliton determine that the bit error rate of optical soliton transmission is much lower than that of conventional optical fiber communication, and even error-free optical fiber communication with a bit error rate of less than 10-12 can be achieved.

[0057] (3) No repeater station is required: As long as the fiber loss is compensated for, the optical signal can be transmitted over a very long distance without distortion, thus eliminating the complex processes of photoelectric conversion, reshaping and amplification, error checking, electro-optical conversion, and retransmission.

[0058] Optical solitons can maintain their shape during propagation and remain stable even under slight disturbances, thus they have long been considered an ideal carrier of optical information. The self-stabilizing property of optical solitons stems from the balance between the dispersion effect and nonlinear effect of light, and is a concrete realization of the "attractor" concept in nonlinear science in optical systems.

[0059] A Convolutional Neural Network (CNN) is a type of feedforward neural network whose artificial neurons can respond to a portion of the surrounding units within their coverage area, making it excellent for large-scale image processing. A CNN consists of one or more convolutional layers and a fully connected layer at the top (corresponding to a classic neural network), as well as associated weights and pooling layers.

[0060] LabVIEW (Laboratory Virtual Instrument Engineering Workbench) is a graphical programming language development environment widely accepted by industry, academia, and research laboratories as a standard data acquisition and instrument control software. LabVIEW integrates all the functions required to communicate with hardware and data acquisition cards that meet GPIB, VXI, RS-232, and RS-485 protocols. It also includes built-in library functions for easy application of software standards such as TCP / IP and ActiveX. It is a powerful and flexible software. It allows for easy creation of your own virtual instruments, and its graphical interface makes programming and use both engaging and fun.

[0061] A virtual instrument is a computer-based instrument. The close integration of computers and instruments is a significant direction in current instrument development. Broadly speaking, this integration takes two forms: one is to embed the computer into the instrument, a typical example being so-called intelligent instruments. As computer functions become increasingly powerful and their size shrinks, the functions of these instruments are also becoming more powerful; instruments with embedded systems are now emerging. The other approach is to integrate the instrument into a computer, relying on general-purpose computer hardware and operating systems to realize various instrument functions.

[0062] like Figures 1 to 7 As shown, the present invention provides an automatic identification and switching device for mid-infrared ultrashort pulse laser status, including a pump source 1, a mode-locked laser, a spectrometer 9, and a control terminal 10;

[0063] The mode-locked laser consists of a beam combiner 2, a gain fiber 3, a single-mode fiber 4, an output coupler 5, an isolator 6, a polarizer 7, and an electric polarization controller 8.

[0064] The pump input arm of the combiner 2 is connected to the output end of the pump source 1 via a passive optical fiber. One end of the gain fiber 3 is connected to the signal output arm of the combiner 2. One end of the single-mode fiber 4 is connected to the other end of the gain fiber 3. One input arm of the output coupler 5 is connected to the other end of the single-mode fiber 4, and one output arm is connected to the input end of the spectrometer 9. The input end of the isolator 6 is connected to the other output arm of the output coupler 5. The input end of the polarizer 7 is connected to the output end of the isolator 6. The input end of the electric polarization controller 8 is connected to the output end of the polarizer 7, and its output end is connected to the signal input arm of the combiner 2.

[0065] The control terminal 10 is connected to the pump source 1, the electric polarization controller 8, and the spectrometer 9, respectively.

[0066] As a preferred embodiment, the beam combiner 2 is a (2+1)×1 high-power multimode pump + signal light beam combiner with an operating wavelength of 1960-2020nm;

[0067] As a preferred embodiment, the gain fiber 3 is a 2m thulium-doped gain fiber, specifically a thulium fiber with a 4% doping concentration.

[0068] As a preferred embodiment, the single-mode fiber 4 is an SMF28e fiber;

[0069] As a preferred embodiment, the output coupler 5 is a 2×2 broadband fiber coupler, 2000±200nm, 90:10 coupling ratio, SM2000 fiber;

[0070] As a preferred embodiment, the polarizer 7 is a coaxial fiber polarizer, 2000±50nm, SM / SM pigtail;

[0071] As a preferred embodiment, the isolator 6 is model IO-F-2000, 2000nm, single-mode, 10W;

[0072] Preferably, the control terminal 10 is a computer with an AMD Ryzen 7 4800H CPU, 16GB of RAM, and a LabVIEW 2018 platform. The LabVIEW platform provides communication interfaces with the electric polarization controller 8, the pump source 1 in the mode-locked laser, and the spectrometer 9. An optimization algorithm framework and program framework are built within the LabVIEW platform, including a target pulse setting window, a control parameter display window, and an output laser measurement window. The LabVIEW platform is connected to the electric polarization controller 8 via a USB Type-A to Micro USB Type-B cable, to the spectrometer 9 via a USB Type-A to GP-IB cable, and to the pump source 1 in the mode-locked laser via a USB Type-A to RS232 cable.

[0073] The spectrometer 9 has a measurable range of 1200nm to 2400nm, can measure optical power up to -70dBm, and can achieve dynamic measurement of 55dB at a resolution of 0.05nm. Its maximum scan speed is 0.5s over a 100nm span, and it uses a logarithmic coordinate mode. The spectrometer 9 is equipped with GPIB, RS232, and Ethernet interfaces, supporting remote control and data acquisition by an external PC. Preferably, the spectrometer 9 is an AQ6375, with a wavelength range of 1200nm to 2400nm, a range of +20dBm to -70dBm, high wavelength resolution (0.05nm), and a large dynamic range (55dB). It is equipped with GPIB, RS-232, and Ethernet (10 / 100Base-T) interfaces for connecting to an external PC for remote control and building an automated testing system.

[0074] As a preferred embodiment, the electric polarization controller is a paddle-based polarization controller, model MPC320, used for... The outer diameter of the sheathed optical fiber and its winding spool The device features three propellers, each capable of rotating 170° with a minimum step size of 0.12°, covering any polarization state on the Poincaré sphere. The motorized polarization controller is based on the principle of stress-induced birefringence, altering the polarization state of light passing through a single-mode fiber. By winding the fiber onto two or three independent winding disks, two or three independent waveplates (fiber delayers) are formed, with an effective operating wavelength from 300 nm to 2100 nm.

[0075] As a preferred embodiment, the pump source 1 is a 793nm semiconductor laser, and the output pigtail connected to it is 1.5m long. It supports external level signal control of pump power output, and its adjustable power range is 0 to 30W, with a maximum output power of 30W.

[0076] In this invention, the pump light and signal light can be coupled into the same optical fiber by using a beam combiner; the gain fiber can be used to excite micrometer-wavelength laser light after receiving light from the pump source; the output coupler can split the light in the fiber ring cavity into two beams, one beam for easy spectral data acquisition by the spectrometer, and the other beam continues to oscillate and provide feedback within the cavity; the isolator and polarizer enable unidirectional transmission of the pump light and signal light within the fiber ring cavity. The electric polarization controller alters the saturable absorption effect caused by nonlinear polarization rotation within the fiber ring cavity, thereby achieving stable mode-locking. Connecting the spectrometer to one output arm of the output coupler facilitates sampling of the output signal to obtain spectral data; simultaneously, connecting the spectrometer to a control terminal allows the spectrometer to send the obtained spectral data to the control terminal for data processing. The control terminal is also connected to both the pump source and the electric polarization controller. This not only facilitates the control terminal in sending control parameters to the pump source and the electric polarization controller for execution, but also allows for easy switching between noise-like mode-locked pulses and soliton mode-locked pulses by controlling the electric polarization controller. Furthermore, this connection method forms a closed-loop feedback control structure, which facilitates the adjustment of the laser's output signal state, promoting stable laser operation at the target state. This compact device can achieve the output of high-power, high-beam-quality, high-efficiency, and high-stability ultrashort pulse lasers, while also facilitating the automatic adjustment and switching of noise-like mode-locked pulses and soliton mode-locked pulses.

[0077] This invention also provides a method for automatic identification and switching of mid-infrared ultrashort pulse laser status, with specific embodiments as follows:

[0078] Example 1:

[0079] An automatic identification and switching device for mid-infrared ultrashort pulse laser states is used to identify and control noise-like pulses in a 2-micron band fiber laser. A thulium-doped fiber laser based on nonlinear polarization rotation is used as the controlled object. The specific steps include:

[0080] Step 1: Select the target pulse on the LabVIEW platform in the control terminal 10 and control the mode-locked laser to start working;

[0081] The LabVIEW platform block diagram is attached. Figure 2As shown, either a noise-based mode-locked pulse or a soliton mode-locked pulse can be selected. Specifically, the target pulse type, the target pulse center wavelength range, and the target pulse spectral full width at half maximum (FWHM) range can be set. If a noise-based mode-locked pulse is selected, the 3dB bandwidth of the target pulse spectrum in logarithmic coordinates can be selected from 10-30 nm. If a soliton mode-locked pulse is selected, the 3dB bandwidth of the target pulse spectrum in logarithmic coordinates can be selected from 4-10 nm. A reference spectral signal is generated based on the set parameters, as shown in the attached figure. Figure 2 As shown, in this embodiment, the target pulse type is set to noise-like mode-locked, the desired center wavelength is 1972nm, the 3dB bandwidth of the spectrum in logarithmic coordinates is 20nm, the spectral resolution is set to 0.05nm, and the single scan range is 100nm.

[0082] Step 2: Use spectrometer 9 to receive the output signal of the mode-locked laser and send the sampled spectral data to the LabVIEW platform in control terminal 10;

[0083] The output signal of the mode-locked laser has a wavelength range of 1800-2940nm, a pulse repetition rate of 10kHz-1GHz, and a pulse width range of 100fs-50ps.

[0084] Step 3: The control terminal 10 uses a trained neural network model on the LabVIEW platform to identify the mode-locked state of the spectral data and calculate the matching degree between the measured pulse spectrum and the target spectrum. If the matching degree between the output laser pulse state and the set target pulse state is greater than 80%, it is considered to be in agreement, and step 5 is executed directly; if the matching degree between the output laser pulse state and the set target pulse state is less than or equal to 80%, it is considered to be in agreement, and step 4 is executed directly.

[0085] The neural network model was trained with a total of 1000 samples, all of which were spectral data, including 600 simulation data and 400 experimental data. Of these, 300 were noise-like mode-locked pulse data and 300 were soliton mode-locked pulse data, with the remainder being mode-free state data. The neural network model was trained using a convolutional neural network, with 900 data points used as the training set and 100 data points used as the test set, achieving accuracies of 99.8% and 97%, respectively.

[0086] Step 4: The control terminal 10 generates corresponding control parameters according to the genetic optimization algorithm, and sends the control parameters to the pump source 1 and the electric polarization controller 8 through the serial communication protocol. The control parameters include the pump power of the pump source 1 and the blade angle of the electric polarization controller 8.

[0087] The specific method of the genetic optimization algorithm is as follows:

[0088] S1: The control terminal 10 randomly generates 20 sets of initial control parameters and sends them to the pump source 1 and the electric polarization controller 8 respectively, so that the pump source 1 and the electric polarization controller 8 execute the received control parameters.

[0089] S2: Calculation, using control terminal 10 to calculate the matching degree between the spectral data corresponding to the 20 sets of initial control parameters in S1 and the target pulse;

[0090] S3: Sorting, using control terminal 10 to sort the 20 sets of initial parameters according to the degree of matching;

[0091] S4: Select the top 6 sets of control parameters with the highest matching degree using the control terminal 10;

[0092] S5: Genetics, using control terminal 10 to randomly cross-combine 6 sets of parameters to generate 10 new sets of control parameters;

[0093] S6: Mutation. In the newly generated control parameters in S5, control terminal 10 randomly selects 4 sets of control parameters and randomly modifies the control parameter values ​​to generate 20 new sets of control parameters.

[0094] S7: Send the newly generated 20 sets of control parameters to the pump source 1 and the electric polarization controller 8 through the control terminal 10;

[0095] Step 5: If the output laser pulse state matches the target pulse, then pulse state monitoring is performed. Pulse state monitoring refers to comparing the current pulse state with the set target pulse state at regular intervals and determining whether the current output matches the set value. If the output laser state changes abruptly and does not match the target pulse, then jump to step 3 and start searching for the target state again.

[0096] Step 6: If the control terminal 10 switches the target pulse, proceed to step 2;

[0097] In step three, the neural network model is trained with a total of 1,000 samples, all of which are spectral data. Among them, 600 are simulation data and 400 are experimental data. Of these, 300 are noise-like mode-locked pulse data, 300 are soliton mode-locked pulse data, and the rest are mode-free state data.

[0098] The final output pulse spectrum measurement signal is shown in the attached figure. Figure 4 As shown, the corresponding control parameters are a pump power of 9.5W, and the three blade angles of the electric polarization controller are 7.9 degrees, 16.0 degrees, and 6.0 degrees, respectively. The matching degree calculated by the neural network model is 95.6%, as shown in the attached figure. Figure 2 As shown;

[0099] Use an oscilloscope to verify the time-domain state of the output pulse. The output laser is input to a high-speed photoelectric sensor, which transmits the signal to the oscilloscope, as shown in the attached diagram. Figure 5 As shown.

[0100] Example 2:

[0101] An automatic identification and switching device for mid-infrared ultrashort pulse laser states is used to identify and control soliton-like pulses in a 2-micron band fiber laser. A thulium-doped fiber laser based on nonlinear polarization rotation is used as the controlled object. The specific steps include:

[0102] Step 1: Select the target pulse on the LabVIEW platform in the control terminal 10 and control the mode-locked laser to start working;

[0103] The LabVIEW platform block diagram is attached. Figure 2 As shown, either a noise-based mode-locked pulse or a soliton mode-locked pulse can be selected. Specifically, the target pulse type, the target pulse center wavelength range, and the target pulse spectral full width at half maximum (FWHM) range can be set. If a noise-based mode-locked pulse is selected, the 3dB bandwidth of the target pulse spectrum in logarithmic coordinates can be selected from 10-30 nm. If a soliton mode-locked pulse is selected, the 3dB bandwidth of the target pulse spectrum in logarithmic coordinates can be selected from 4-10 nm. A reference spectral signal is generated based on the set parameters, as shown in the attached figure. Figure 5 As shown, in this embodiment, the target pulse type is set to soliton mode-locked, the desired center wavelength is 1960nm, the 3dB bandwidth of the spectrum in logarithmic coordinates is 6nm, the spectral resolution is set to 0.05nm, and the single scan range is 100nm.

[0104] Step 2: Use spectrometer 9 to receive the output signal of the mode-locked laser and send the sampled spectral data to the LabVIEW platform in control terminal 10;

[0105] The output signal of the mode-locked laser has a wavelength range of 1800-2940nm, a pulse repetition rate of 10kHz-1GHz, and a pulse width range of 100fs-50ps.

[0106] Step 3: The control terminal 10 uses a trained neural network model on the LabVIEW platform to identify the mode-locked state of the spectral data and calculate the matching degree between the measured pulse spectrum and the target spectrum. If the matching degree between the output laser pulse state and the set target pulse state is greater than 80%, it is considered to be in agreement, and step 5 is executed directly; if the matching degree between the output laser pulse state and the set target pulse state is less than or equal to 80%, it is considered to be in agreement, and step 4 is executed directly.

[0107] The neural network model was trained with a total of 1000 samples, all of which were spectral data, including 600 simulation data and 400 experimental data. Of these, 300 were noise-like mode-locked pulse data and 300 were soliton mode-locked pulse data, with the remainder being mode-free state data. The neural network model was trained using a convolutional neural network, with 900 data points used as the training set and 100 data points used as the test set, achieving accuracies of 99.8% and 97%, respectively.

[0108] Step 4: The control terminal 10 generates corresponding control parameters according to the genetic optimization algorithm, and sends the control parameters to the pump source 1 and the electric polarization controller 8 through the serial communication protocol. The control parameters include the pump power of the pump source 1 and the blade angle of the electric polarization controller 8.

[0109] The specific method of the genetic optimization algorithm is as follows:

[0110] S1: The control terminal 10 randomly generates 20 sets of initial control parameters and sends them to the pump source 1 and the electric polarization controller 8 respectively, so that the pump source 1 and the electric polarization controller 8 execute the received control parameters.

[0111] S2: Calculation, using control terminal 10 to calculate the matching degree between the spectral data corresponding to the 20 sets of initial control parameters in S1 and the target pulse;

[0112] S3: Sorting, using control terminal 10 to sort the 20 sets of initial parameters according to the degree of matching;

[0113] S4: Select the top 6 sets of control parameters with the highest matching degree using the control terminal 10;

[0114] S5: Genetics, using control terminal 10 to randomly cross-combine 6 sets of parameters to generate 10 new sets of control parameters;

[0115] S6: Mutation. In the newly generated control parameters in S5, control terminal 10 randomly selects 4 sets of control parameters and randomly modifies the control parameter values ​​to generate 20 new sets of control parameters.

[0116] S7: Send the newly generated 20 sets of control parameters to the pump source 1 and the electric polarization controller 8 through the control terminal 10;

[0117] Step 5: If the output laser pulse state matches the target pulse, then pulse state monitoring is performed. Pulse state monitoring refers to comparing the current pulse state with the set target pulse state at regular intervals and determining whether the current output matches the set value. If the output laser state changes abruptly and does not match the target pulse, then jump to step 3 and start searching for the target state again.

[0118] Step 6: If the control terminal 10 switches the target pulse, proceed to step 2;

[0119] In step three, the neural network model is trained with a total of 1,000 samples, all of which are spectral data. Among them, 600 are simulation data and 400 are experimental data. Of these, 300 are noise-like mode-locked pulse data, 300 are soliton mode-locked pulse data, and the rest are mode-free state data.

[0120] The final output pulse spectrum measurement signal is shown in the attached figure. Figure 6 As shown, the corresponding control parameters are a pump power of 5W, and the three blade angles of the electric polarization controller are 86.0 degrees, 65.8 degrees, and 71.5 degrees respectively. The matching degree calculated by the neural network model is 86.5%, as shown in the attached figure. Figure 5 As shown;

[0121] Use an oscilloscope to verify the time-domain state of the output pulse. The output laser is input to a high-speed photoelectric sensor, which transmits the signal to the oscilloscope, as shown in the attached diagram. Figure 7 As shown.

[0122] This invention utilizes an electric polarization controller as the polarization control device in a mode-locked laser, and a spectrometer as the measuring device to receive the output signal of the mode-locked laser. Simultaneously, a neural network model on the control terminal identifies the mode-locking state of the sampled spectral data. The matching degree between the output laser pulse state and the set target pulse state determines whether optimization processing is required. The optimization process employs a genetic algorithm, which effectively ensures the laser output signal remains stable in the target state through a closed-loop feedback structure and effective state identification. Furthermore, the electric polarization controller can control the switching between noise-like mode-locked pulses and soliton mode-locked pulses. Since noise-like mode-locked pulses and soliton mode-locked pulses have highly consistent time-domain characteristics, they are difficult to distinguish effectively using only an oscilloscope or photodetector. This invention, through enhanced training of the spectral state of ultrafast laser pulses using a convolutional neural network, can automatically achieve accurate and efficient identification of noise-like pulse and soliton pulse states, solving the problem of state misjudgment in mid-infrared ultrafast fiber laser pulse applications and realizing automatic adjustment and switching between the two pulse states. The neural network training algorithm employed in this invention offers advantages such as fast recognition speed, high recognition accuracy, and intelligence, significantly improving the reliability of mid-infrared ultrafast fiber lasers in industrial applications. The genetic algorithm used for intelligent control of laser pulse states enables rapid and precise switching between various laser states, including complex noise-like pulse states and soliton states. Compared to existing manual control methods, the resulting laser states are more efficient, significantly reducing the cost of manual control. Compared to existing mid-infrared ultrafast laser systems, this invention uses ordinary commercial single-mode fiber instead of the expensive polarization-maintaining fiber used in mainstream methods, greatly reducing costs. The intelligent recognition and control method employed in this invention also eliminates the need for complex temperature control, mechanical vibration control, and environmental disturbance prevention control systems, allowing for rapid adjustment according to environmental changes and greatly enhancing environmental adaptability. The fully automatic intelligent control system can achieve various pulse outputs, including multiple types of noise, solitons, and harmonic mode-locked solitons, and can also generate various customized laser pulse outputs according to customer needs. This greatly expands the application range of the laser system, and the rapid switching between different pulse states will solve the problem of high-precision laser processing of complex structural materials.

Claims

1. A method for automatic identification and switching of mid-infrared ultrashort pulse laser status, characterized in that, This method employs an automatic identification and switching device for mid-infrared ultrashort pulse laser states. The device includes a pump source (1), a mode-locked laser, a spectrometer (9), and a control terminal (10). The mode-locked laser consists of a beam combiner (2), a gain fiber (3), a single-mode fiber (4), an output coupler (5), an isolator (6), a polarizer (7), and an electric polarization controller (8). The pump input arm of the beam combiner (2) is connected to the output end of the pump source (1) via a passive optical fiber. One end of the gain fiber (3) is connected to the signal output arm of the beam combiner (2). The single-mode fiber (4) is connected to the output end of the pump source (1) via a passive optical fiber. One end of the output coupler (5) is connected to the other end of the gain fiber (3), one input arm of the output coupler (5) is connected to the other end of the single-mode fiber (4), and one output arm of the output coupler (5) is connected to the input end of the spectrometer (9). The input end of the isolator (6) is connected to the other output arm of the output coupler (5), the input end of the polarizer (7) is connected to the output end of the isolator (6), the input end of the electric polarization controller (8) is connected to the output end of the polarizer (7), and its output end is connected to the signal input arm of the combiner (2). The control terminal (10) is connected to the pump source (1), the electric polarization controller (8), and the spectrometer (9) respectively. The method includes the following steps: Step 1: Select the target pulse through the control terminal (10) and control the mode-locked laser to start working; Step 2: Use a spectrometer (9) to receive the output signal of the mode-locked laser and send the sampled spectral data to the control terminal (10). Step 3: The control terminal (10) uses a trained neural network model to identify the mode-locked state of the spectral data and calculates the matching degree between the measured pulse spectrum and the target spectrum. If the matching degree between the output laser pulse state and the set target pulse state is greater than 80%, it is considered to be consistent and Step 5 is executed directly; if the matching degree between the output laser pulse state and the set target pulse state is less than or equal to 80%, it is considered to be inconsistent and Step 4 is executed directly. Step 4: The control terminal (10) generates the corresponding control parameters according to the genetic optimization algorithm and sends the control parameters to the pump source (1) and the electric polarization controller (8) through the serial communication protocol. The control parameters include the pump power of the pump source (1) and the blade angle of the electric polarization controller (8). The specific method of the genetic optimization algorithm is as follows: S1: The control terminal (10) randomly generates 20 sets of initial control parameters and sends them to the pump source (1) and the electric polarization controller (8) respectively, so that the pump source (1) and the electric polarization controller (8) execute the received control parameters. S2: Calculate the degree of matching between the spectral data corresponding to the 20 initial control parameters in S1 and the target pulse using the control terminal (10); S3: Sorting, using the control terminal (10) to sort the 20 sets of initial parameters according to the degree of matching; S4: Select the top 6 control parameters with the highest matching degree using the control terminal (10); S5: Genetics, using the control terminal (10) to randomly cross-combine 6 sets of parameters to generate 10 new sets of control parameters; S6: Mutation, in the newly generated control parameters in S5, use the control terminal (10) to randomly select 4 sets of control parameters and randomly modify the control parameter values ​​to generate 20 new sets of control parameters; S7: Send the newly generated 20 sets of control parameters to the pump source (1) and the electric polarization controller (8) through the control terminal (10); Step 5: If the output laser pulse state matches the target pulse, then pulse state monitoring is performed. Pulse state monitoring refers to comparing the current pulse state with the set target pulse state at regular intervals and determining whether the current output matches the set value. If the output laser state changes abruptly and does not match the target pulse, then jump to step 3 and start searching for the target state again. Step 6: If the control terminal (10) switches the target pulse, jump to step 2.

2. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, In step three, the neural network model is trained with a total of 1,000 samples, all of which are spectral data. Among them, 600 are simulation data and 400 are experimental data. Of these, 300 are noise-like mode-locked pulse data, 300 are soliton mode-locked pulse data, and the rest are mode-free state data.

3. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, In step one, the target pulse is a selected noise mode-locked pulse or a soliton mode-locked pulse.

4. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, The control terminal (10) is a computer.

5. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, The spectrometer (9) has a measurable range of 1200nm~2400nm, can measure optical power of -70dBm, can achieve dynamic measurement of 55dB at a resolution of 0.05nm, and has a maximum scanning speed of 0.5s over a span of 100nm. It uses a logarithmic coordinate mode. The spectrometer (9) is equipped with GPIB, RS232 and Ethernet interfaces, and supports remote control and data acquisition by an external PC.

6. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, The electric polarization controller is a paddle-based polarization controller with an outer diameter of 18 mm for the winding disk. Each paddle can rotate 170°, with a minimum step size of 0.12°, and can cover any polarization state on the Bonga sphere.

7. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, The pump source (1) supports external level signal control of pump power output, and the adjustable power range is 0 to 30W.

8. The method for automatic identification and switching of mid-infrared ultrashort pulse laser state according to claim 1, characterized in that, The gain fiber (3) is a thulium-doped gain fiber.

Images