Laser with microfluidic heat dissipation cooperative mode-locking mechanism

Abstract

The invention discloses a laser with a microfluidic heat dissipation cooperative mode locking mechanism, and relates to the technical field of laser. The laser comprises a mode-locked laser resonant cavity, an optical fiber beam splitter connected with the mode-locked laser resonant cavity, an optical fiber Bragg grating sensor, an online spectrum analyzer, a cooperative control processor, a precise microfluid pump and an intracavity electro-optical modulator, the mode-locked laser resonant cavity comprises a gain optical fiber, a semiconductor saturable absorber and an intracavity electro-optical modulator, the gain optical fiber is thermally coupled with the micro-fluidic heat dissipation channel, and the micro-fluidic heat dissipation channel is communicated with a fluid loop of the precise micro-fluid pump. According to the method, the cooperative control model is constructed, the heat dissipation state and the mode locking quality are correlated, and the thermal state prediction model is utilized to carry out feedforward type active adjustment, so that the inherent limitation of traditional separated hysteresis control is overcome, and precise closed-loop control of the mode locking state of the laser is realized; and the stability, the reliability and the environmental adaptability of the laser under complex working conditions are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of laser technology, and more specifically to a laser with a microfluidic heat dissipation and mode-locking mechanism. Background Technology

[0002] A laser is an optical device that generates a highly coherent beam based on the principle of stimulated emission. Its core structure includes a gain medium, a pump source, and an optical resonant cavity. The gain medium achieves population inversion under the excitation of the pump source, converting energy into photons through stimulated emission and amplifying the output. The optical resonant cavity consists of at least two mirrors. Through a feedback mechanism, photons of specific wavelengths and propagation directions are selected, causing the light to oscillate within the optical resonant cavity and form a stable standing wave field. Finally, a laser beam with high directionality, high monochromaticity, and high energy density is output through a partially transmitted mirror. Lasers have been widely used in materials processing, medical diagnosis, optical communication, and precision measurement. However, existing technologies still have optimization requirements in terms of energy conversion efficiency, beam quality stability, and environmental adaptability.

[0003] To address the issue of unstable mode-locking states caused by internal heat fluctuations in mode-locked lasers, existing technologies separate the microfluidic cooling system from the intracavity control system for mode-locking, allowing them to operate independently. However, this approach suffers from several drawbacks. The microfluidic cooling system can only passively maintain the temperature, while the intracavity control system only provides hysteresis compensation after spectral characteristics deteriorate. This prevents the two systems from achieving coordinated control and feedforward adjustment, resulting in insufficient suppression of thermal disturbances and hindering proactive, predictive, and precise control of the mode-locking state. Consequently, it limits the stability and reliability of output performance. To resolve these issues, a laser with a microfluidic cooling-coordinated mode-locking mechanism is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a laser with a microfluidic heat dissipation and mode-locking mechanism to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a laser with a microfluidic heat dissipation and collaborative mode-locking mechanism, the laser including a mode-locked laser resonant cavity, an optical fiber beam splitter connected to the mode-locked laser resonant cavity, an optical fiber Bragg grating sensor, an online spectrometer, a collaborative control processor, a precision microfluidic pump, and an intracavity electro-optic modulator; The mode-locked laser resonant cavity includes a gain fiber, a semiconductor saturable absorber, and an in-cavity electro-optic modulator. The gain fiber is thermally coupled to a microfluidic heat dissipation channel, which is connected to the fluid circuit of a precision microfluidic pump. Fiber Bragg grating sensors are integrated into sections of the gain fiber to collect temperature data from the gain fiber. The online spectrometer is connected to the output of the mode-locked laser resonator via an optical fiber beam splitter to acquire the spectral data of the laser. The collaborative control processor connects to the fiber Bragg grating sensor, online spectrometer, precision microfluidic pump, and intracavity electro-optic modulator via a data and control bus. The collaborative control processor is used to receive and preprocess temperature data and spectral data, extract the average temperature and temperature drift rate from the preprocessed temperature data, and extract the spectral width and center wavelength from the preprocessed spectral data. The collaborative control processor is further used to run a thermal state prediction model based on average temperature and temperature drift rate to generate a predicted temperature, and to run a mode-lock quality evaluation model based on spectral width and center wavelength to generate a mode-lock quality score. The collaborative control processor is further used to train the collaborative control model based on the preset optimization target, and input the predicted temperature and mode-locking quality score into the collaborative control model to generate flow rate control commands for the precision microfluidic pump and voltage control commands for the intracavity electro-optic modulator. The collaborative control processor sets up the decision logic internally and transmits the control commands that meet the decision logic to the precision microfluidic pump and the intracavity electro-optic modulator for execution. The collaborative control processor is further used to provide early warnings of deviations from the safe operating range in predicted temperature and mold clamping quality scores, and to display status data through a graphical user interface.

[0006] A further improvement to the technical solution of this invention lies in that: the fiber Bragg grating sensor is integrated into the gain fiber section for acquiring temperature data of the gain fiber, and the online spectrometer is connected to the output end of the mode-locked laser resonator via a fiber beam splitter. The process for acquiring spectral data of the laser includes: Within a predetermined section of the fiber core of the gain fiber, a grating region is formed by ultraviolet laser exposure. To query the reflection spectrum of the grating region, a broadband light source is input at the first port of the fiber circulator, the second port is connected to the gain fiber containing the grating region, and the third port is connected to the spectral detection unit. The spectral detection unit is signal-connected to the co-control processor to form a temperature data acquisition link. The optical signal emitted by the broadband light source propagates in the gain fiber. When it passes through the grating region, the optical signal that satisfies the Bragg wavelength condition is reflected. The magnitude of the Bragg wavelength depends on the effective refractive index and the grating period of the grating. When the temperature of the gain fiber changes, the effective refractive index and grating period of the fiber core are altered through thermo-optical effects and thermal expansion, resulting in a shift in the reflected Bragg wavelength. The spectral detection unit receives the optical signal reflected back from the third port of the fiber optic circulator and measures the drifted Bragg wavelength. The collaborative control processor calculates the Bragg wavelength drift based on the drifted Bragg wavelength and, based on the linear relationship between the Bragg wavelength drift and the temperature change, obtains the temperature data of the gain fiber. The fiber optic beam splitter separates a small portion of the optical signal and imports it into the input port of the online spectrometer. A small portion of the optical signal entering the online spectrometer is collimated into a parallel beam by the collimation system inside the online spectrometer, and the parallel beam is incident on the diffraction grating. A diffraction grating uses the dispersion effect to spread parallel beams of different wavelengths in space, so that the parallel beams exit at different diffraction angles, and are projected and imaged onto the photosensitive surface of a photodetector array through a focusing system. The photosensitive surface of the photodetector array is divided into pixel units. The online spectral analyzer collects the electrical signals output by the pixel units, generates spectral data, and transmits the spectral data to the collaborative control processor.

[0007] A further improvement to the technical solution of this invention lies in the following: the collaborative control processor receives and preprocesses temperature data and spectral data, extracts the average temperature and temperature drift rate from the preprocessed temperature data, and extracts the spectral width and center wavelength from the preprocessed spectral data, the process of which includes: The collaborative control processor performs noise filtering and timestamp alignment on the temperature and spectral data to obtain preprocessed temperature and spectral data. Within a preset time window, the arithmetic mean of continuous temperature data points of the preprocessed temperature data is calculated to obtain the average temperature within the time window. Within the time window, linear regression analysis is performed on continuous temperature data points of the preprocessed temperature data, and the slope of the fitted line of the linear regression analysis is taken as the temperature drift rate. The center wavelength of the spectrum was calculated using the power-weighted average method after preprocessing the spectral data. The calculation process is as follows: ; in, Indicating the first spectral data Optical power at each spectral data point Indicates the relationship with the first The wavelength corresponding to each spectral data point This represents the total number of spectral data points in the spectral data; For the preprocessed spectral data, the full width at half maximum (FWHM) method is used to calculate the spectral width. The FWHM method involves taking the half-peak power of the preprocessed spectral data, finding two wavelengths on both sides of the spectral profile that correspond to the half-peak power, and taking the absolute value of the difference between the two wavelengths as the spectral width.

[0008] A further improvement to the technical solution of this invention lies in the following: the process of generating the predicted temperature by running a thermal state prediction model based on the average temperature and temperature drift rate includes: At each discrete time step, the cooperative control processor combines the average temperature and temperature drift rate into an input vector, and inputs the continuous input vectors into a long short-term memory network model with a long short-term memory network architecture in time order, wherein the long short-term memory network architecture is composed of memory units connected in sequence. At each time step, the memory unit updates its internal long-term memory state based on the current input vector and the memory unit output of the previous time step through the forget gate, input gate and output gate inside the memory unit. The update process includes forgetting the historical information of the long-term memory state, storing the new information of the current input vector, and selecting information from the updated long-term memory state as the output result of the current time step. After the input vector passes through the memory unit, the last layer of the Long Short-Term Memory network architecture outputs a feature vector containing historical information. This feature vector is then input into a fully connected layer, which performs a linear transformation on the feature vector to generate the predicted temperature.

[0009] A further improvement to the technical solution of this invention lies in the process of generating a mode-locking quality score based on a mode-locking quality evaluation model that operates on spectral width and center wavelength, including: For each sample consisting of spectral width and center wavelength, the corresponding mode-locking state category label is manually assigned to form a labeled training set. The mode-locking state category label includes stable mode-locking and unstable mode-locking. A support vector machine (SVM) classification algorithm is used to train the labeled dataset. In a two-dimensional feature space consisting of spectral width and center wavelength, the optimal decision boundary is found that maximizes the separation of sample points corresponding to different mode-locking state category labels. After training, the trained mode-locking quality evaluation model and its parameters are obtained, and the model parameters, including support vectors, are stored in the collaborative control processor. and its coefficients ; During the real-time operation of the laser, the cooperative control processor combines the currently calculated spectral width and center wavelength into a real-time feature vector. ; The mode-locking quality assessment model receives real-time feature vectors. And through a preset kernel function Calculate real-time feature vectors With stored support vectors The relationship between them; The mold clamping quality evaluation model is based on relationships and pre-stored coefficients. Calculate the decision function value The calculation process is as follows: ; in, Indicates the first The label of the mode-locked state category to which each support vector belongs. Indicates the bias term. This represents the total number of support vectors; decision function value The model-locking quality score is generated by normalizing the model using a preset mapping function.

[0010] A further improvement to the technical solution of this invention lies in the following: the process of training the cooperative control model according to a preset optimization objective includes: Defined in the collaborative control model to maximize the mode-locking quality score. And will predict temperature The reward function aims to maintain the temperature within a preset optimal range, and calculates the instantaneous reward value. The calculation process is as follows: ; in, and The preset weighting coefficients, Let be the penalty function, when When within the optimal temperature range, ,when When outside the optimal temperature range, ; Using a cooperative control model as the agent, the agent learns by interacting with the simulation environment of the laser. In each interaction step, the agent selects an action to output to the simulation environment based on the current state and obtains the reward value and the next state from the simulation environment. By using a reinforcement learning algorithm, the agent updates the parameters of its internal policy network until it learns the optimal policy that maximizes the cumulative expected reward. The state consists of the predicted temperature and the mode-locking quality score, and the action consists of the flow rate control command and the voltage control command.

[0011] A further improvement to the technical solution of this invention lies in the following: the process of inputting the predicted temperature and mode-locking quality score into the collaborative control model to generate flow rate control commands for the precision microfluidic pump and voltage control commands for the intracavity electro-optic modulator includes: The predicted temperature generated by the thermal state prediction model and the mode-locking quality score generated by the mode-locking quality evaluation model are combined into a state vector and input into the cooperative control model that has learned the optimal strategy. The cooperative control model that learns the optimal strategy performs forward calculation on the input state vector and outputs an action vector, which includes flow rate control commands for the precision microfluidic pump and voltage control commands for the intracavity electro-optic modulator.

[0012] A further improvement to the technical solution of this invention lies in the following: the process of setting up decision logic within the collaborative control processor and transmitting control commands that satisfy the decision logic to the precision microfluidic pump and the intracavity electro-optic modulator for execution includes: Control commands include flow rate control commands and voltage control commands; The system reads the currently executing control instruction from the registers of the cooperative control processor control interface, calculates the difference between the newly generated control instruction and the currently executing control instruction, and compares the difference with a preset execution threshold. If the difference is greater than the execution threshold, the newly generated control instruction is considered a valid update instruction, and the valid update instruction is transmitted to the precision microfluidic pump and the intracavity electro-optic modulator for update execution. If the difference does not exceed the execution threshold, the currently executing control instruction will remain unchanged, and no update operation will be performed.

[0013] A further improvement to the technical solution of this invention lies in the following: the process of providing early warning of deviations from the safe operating range in the predicted temperature and mold clamping quality score includes: Pre-set the limits for the safe working range, including the upper limit of temperature, the lower limit of temperature, and the lower limit of mold clamping quality score. Compare the predicted temperature and mold clamping quality score with the limits. When the predicted temperature exceeds the upper limit of the temperature range, the lower limit of the temperature range, and the mold-locking quality score is lower than the lower limit of the mold-locking quality score, and the duration of the deviation exceeds the preset warning trigger duration, a corresponding warning signal is generated.

[0014] A further improvement to the technical solution of this invention lies in that: the process of displaying status data through a graphical user interface includes: The status display data includes spectral data, predicted temperature, mold-locking quality score, flow rate control commands, voltage control commands, and warning signals; Based on the spectral data, it is rendered as a two-dimensional spectral curve with wavelength as the horizontal axis and optical power as the vertical axis; The predicted temperature and mold-locking quality score, which change over time, are rendered as a real-time updated time series curve. Display the current values ​​of the flow rate control command and voltage control command as digital readings; When a warning signal is received, the graphical user interface state changes. The two-dimensional spectral curve, digital readings, and changes in the graphical user interface are combined into a graphical user interface image, which is then transmitted to the laser's display device, where the graphical user interface is presented to the user.

[0015] Due to the adoption of the above technical solution, the technical progress achieved by this invention compared to the prior art is as follows: 1. This invention provides a laser with a microfluidic heat dissipation collaborative mode-locking mechanism. By constructing a collaborative control model, the microfluidic heat dissipation state is correlated with the laser mode-locking quality, realizing the collaborative action of heat dissipation regulation and mode-locking state regulation. This overcomes the limitation of the two being separated and unable to be optimized in traditional technology, and improves the coordination and efficiency of the overall control of the laser.

[0016] 2. This invention provides a laser with a microfluidic heat dissipation and collaborative mode-locking mechanism. By using a thermal state prediction model and a mode-locking quality evaluation model, the laser predicts the trend of mode-locking state changes caused by thermal effects and performs feedforward active adjustment. This effectively overcomes the response delay problem caused by hysteresis compensation in the prior art and significantly enhances the laser's ability to suppress thermal disturbances and its output stability.

[0017] 3. This invention provides a laser with a microfluidic heat dissipation collaborative mode-locking mechanism. It makes intelligent decisions through a data-driven collaborative control model and automatically generates flow rate control commands and voltage control commands based on preset optimization targets. This achieves precise closed-loop control of the mode-locking state, improves the automation level and control accuracy of the laser, and enhances the robustness and adaptability of the laser in different working environments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a schematic diagram of a laser with a microfluidic heat dissipation and collaborative mode-locking mechanism provided by the present invention.

[0020] In the figure, 1-mode-locked laser resonator, 2-graphical user interface, 3-fiber Bragg grating sensor, 4-online spectrometer, 5-cooperative control processor, 6-precision microfluidic pump, 7-intracavity electro-optic modulator, 8-gain fiber, 9-saturable absorber, 10-microfluidic heat dissipation channel, 11-temperature data, 12-spectral data, 13-flow rate control command, 14-voltage control command, 15-status display data, 16-data and control bus. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1, as Figure 1 As shown, the present invention provides a laser with a microfluidic heat dissipation and collaborative mode-locking mechanism. The laser includes a mode-locked laser resonant cavity 1, an optical fiber beam splitter connected to the mode-locked laser resonant cavity 1, an optical fiber Bragg grating sensor 3, an online spectrometer 4, a collaborative control processor 5, a precision microfluidic pump 6, and an intracavity electro-optic modulator 7.

[0023] The mode-locked laser resonant cavity 1 includes a gain fiber 8, a semiconductor saturable absorber 9, and an intracavity electro-optic modulator 7. The gain fiber 8 is thermally coupled to the microfluidic heat dissipation channel 10, and the microfluidic heat dissipation channel 10 is connected to the fluid circuit of the precision microfluidic pump 6.

[0024] The fiber Bragg grating sensor 3 is integrated into a section of the gain fiber 8 and is used to collect temperature data 11 from the gain fiber 8.

[0025] In some embodiments, a grating region is formed in a predetermined section of the fiber core of the gain fiber 8 by ultraviolet laser exposure. To query the reflection spectrum of the grating region, a broadband light source is input at the first port of the fiber optic circulator, the second port is connected to the gain fiber 8 containing the grating region, and the third port is connected to the spectral detection unit. The spectral detection unit is signal-connected to the co-control processor 5 to form the acquisition link of temperature data 11.

[0026] In some embodiments, the optical signal emitted by the broadband light source propagates in the gain fiber 8. When it passes through the grating region, the optical signal that satisfies the Bragg wavelength condition is reflected, wherein the magnitude of the Bragg wavelength depends on the effective refractive index and the grating period of the grating.

[0027] In some embodiments, when the temperature of the gain fiber 8 changes, the effective refractive index and grating period of the fiber core are altered by thermo-optical effects and thermal expansion effects, resulting in a drift of the reflected Bragg wavelength.

[0028] In some embodiments, the spectral detection unit receives the optical signal reflected back from the third port of the fiber optic circulator and measures the drifted Bragg wavelength.

[0029] In some embodiments, the cooperative control processor 5 calculates the Bragg wavelength drift based on the drifted Bragg wavelength and obtains the temperature data 11 of the gain fiber 8 based on the linear relationship between the Bragg wavelength drift and the temperature change.

[0030] The online spectrometer 4 is connected to the output end of the mode-locked laser resonator 1 via an optical fiber beam splitter, and is used to collect the spectral data 12 of the laser.

[0031] In some embodiments, the fiber optic beam splitter separates a small portion of the optical signal and imports this small portion of the optical signal into the input port of the online spectrometer 4.

[0032] In some embodiments, a small portion of the optical signal entering the online spectrometer 4 is collimated by the collimation system inside the online spectrometer 4 to form a parallel beam, and the parallel beam is incident on the diffraction grating.

[0033] In some embodiments, the diffraction grating utilizes the dispersion effect to spread parallel beams of different wavelengths in space, causing the parallel beams to exit at different diffraction angles, and through a focusing system, project and image onto the photosensitive surface of the photodetector array.

[0034] In some embodiments, the photosensitive surface of the photodetector array is divided into pixel units, the online spectrometer 4 collects the electrical signals output by the pixel units, generates spectral data 12, and transmits the spectral data 12 to the collaborative control processor 5.

[0035] The collaborative control processor 5 is connected to the fiber Bragg grating sensor 3, the online spectrometer 4, the precision microfluidic pump 6, and the intracavity electro-optic modulator 7 via the data and control bus 16.

[0036] The collaborative control processor 5 is used to receive and preprocess temperature data 11 and spectral data 12, extract the average temperature and temperature drift rate from the preprocessed temperature data 11, and extract the spectral width and center wavelength from the preprocessed spectral data 12.

[0037] In some embodiments, the collaborative control processor 5 performs noise filtering and timestamp alignment on the temperature data 11 and the spectral data 12 to obtain preprocessed temperature data 11 and spectral data 12.

[0038] In some embodiments, within a preset time window, the arithmetic mean of the continuous temperature data 11 points of the preprocessed temperature data 11 is calculated to obtain the average temperature within the time window.

[0039] In some embodiments, within a time window, linear regression analysis is performed on the continuous temperature data 11 points of the preprocessed temperature data 11, and the slope of the fitted line of the linear regression analysis is taken as the temperature drift rate.

[0040] In some embodiments, the center wavelength of the spectrum is calculated using the power-weighted average method on the preprocessed spectral data 12. The specific calculation formula is as follows:

[0041] in, This indicates the 12th spectral data. Optical power at 12 points of spectral data. Indicates the relationship with the first The wavelengths corresponding to 12 points of spectral data. This represents the total number of spectral data points 12 in spectral data 12.

[0042] In some embodiments, the spectral width is calculated using the full width at half maximum (FWHM) method for the preprocessed spectral data 12. The FWHM method includes taking the half-peak power of the preprocessed spectral data 12, finding two wavelengths corresponding to the half-peak power on both sides of the spectral profile, and taking the absolute value of the difference between the two wavelengths as the spectral width.

[0043] The collaborative control processor 5 is further used to run a thermal state prediction model based on average temperature and temperature drift rate to generate a predicted temperature, and to run a mode-locking quality evaluation model based on spectral width and center wavelength to generate a mode-locking quality score.

[0044] In some embodiments, at each discrete time step, the cooperative control processor 5 combines the average temperature and the temperature drift rate into an input vector, and inputs the continuous input vectors in time sequence into a long short-term memory network model employing a long short-term memory network architecture, wherein the long short-term memory network architecture is composed of memory units connected in sequence.

[0045] In some embodiments, at each time step, the memory unit updates its internal long-term memory state based on the current input vector and the memory unit output of the previous time step through the forget gate, input gate and output gate inside the memory unit. The update process includes forgetting the historical information of the long-term memory state, storing the new information of the current input vector, and selecting information from the updated long-term memory state as the output result of the current time step.

[0046] In some embodiments, after the input vector passes through the memory unit, the last layer of the Long Short-Term Memory network architecture outputs a feature vector containing historical information. The feature vector is then input into a fully connected layer, which performs a linear transformation on the feature vector to generate a predicted temperature.

[0047] In some embodiments, for each sample consisting of spectral width and center wavelength, the mode-locking state category label corresponding to the sample is manually labeled to form a labeled training set. The mode-locking state category label includes stable mode-locking and unstable mode-locking.

[0048] In some embodiments, a support vector machine (SVM) classification algorithm is used to train the labeled dataset. In a two-dimensional feature space composed of spectral width and center wavelength, the optimal decision boundary is found that separates sample points corresponding to different mode-locking state category labels with the maximum interval. After training, the trained mode-locking quality evaluation model and its parameters are obtained, and the model parameters are then stored in the collaborative control processor 5. The model parameters include support vectors. and its coefficients .

[0049] In some embodiments, during real-time operation of the laser, the cooperative control processor 5 combines the currently calculated spectral width and center wavelength into a real-time feature vector. .

[0050] In some embodiments, the mode-locking quality assessment model receives real-time feature vectors. And through a preset kernel function Calculate real-time feature vectors With stored support vectors The relationship between them.

[0051] In some embodiments, the mold clamping quality evaluation model is based on relationships and pre-stored coefficients. Calculate the decision function value The specific calculation formula is as follows:

[0052] in, Indicates the first The label of the mode-locked state category to which each support vector belongs. Indicates the bias term. This represents the total number of support vectors.

[0053] In some embodiments, the decision function value The model-locking quality score is generated by normalizing the model using a preset mapping function.

[0054] The collaborative control processor 5 is further used to train the collaborative control model according to the preset optimization target, and input the predicted temperature and mode-locking quality score into the collaborative control model to generate the flow rate control command 13 for the precision microfluidic pump 6 and the voltage control command 14 for the intracavity electro-optic modulator 7.

[0055] In some embodiments, a method is defined in the collaborative control model to maximize the mode-locking quality score. And will predict temperature The reward function aims to maintain the temperature within a preset optimal range, and calculates the instantaneous reward value. The specific calculation formula is as follows:

[0056] in, and The preset weighting coefficients, Let be the penalty function, when When within the optimal temperature range, ,when When outside the optimal temperature range, .

[0057] In some embodiments, a cooperative control model is used as an agent to learn by interacting with the simulation environment of the laser. In each interaction step, the agent selects an action to output to the simulation environment based on the current state and obtains a reward value and the next state from the simulation environment. By employing a reinforcement learning algorithm, the agent updates the internal policy network parameters until it learns the optimal policy that maximizes the cumulative expected reward. The state consists of the predicted temperature and the mode-locking quality score, and the action consists of the flow rate control command 13 and the voltage control command 14.

[0058] In some embodiments, the predicted temperature generated by the thermal state prediction model and the mode-locking quality score generated by the mode-locking quality evaluation model are combined into a state vector and input into the cooperative control model that has learned the optimal strategy.

[0059] In some embodiments, the cooperative control model that has learned the optimal strategy performs forward calculation on the input state vector and outputs an action vector, which includes a flow rate control command 13 for the precision microfluidic pump 6 and a voltage control command 14 for the intracavity electro-optic modulator 7.

[0060] The collaborative control processor 5 internally sets the judgment logic and transmits the control instructions that meet the judgment logic to the precision microfluidic pump 6 and the intracavity electro-optic modulator 7 for execution.

[0061] In some embodiments, the control commands include flow rate control command 13 and voltage control command 14.

[0062] In some embodiments, the currently executing control instruction is read from the register of the control interface of the cooperative control processor 5, the difference between the newly generated control instruction and the currently executing control instruction is calculated, and the difference is compared with a preset execution threshold.

[0063] In some embodiments, if the difference is greater than the execution threshold, the newly generated control instruction is considered a valid update instruction, and the valid update instruction is transmitted to the precision microfluidic pump 6 and the intracavity electro-optic modulator 7 for update execution.

[0064] In some embodiments, if the difference does not exceed the execution threshold, the currently executing control instruction remains unchanged, and no update operation is performed.

[0065] The collaborative control processor 5 is further used to provide early warnings of deviations from the safe operating range for predicted temperature and mold clamping quality scores, and to display status display data 15 through the graphical user interface 2.

[0066] In some embodiments, a safe operating range limit is preset, which includes an upper temperature limit, a lower temperature limit, and a lower limit for mold clamping quality score, and the predicted temperature and mold clamping quality score are compared with the limit.

[0067] In some embodiments, when the predicted temperature exceeds the upper limit of the temperature range, the lower limit of the temperature range, and the mold-locking quality score is lower than the lower limit of the mold-locking quality score, and the duration of the deviation exceeds a preset warning triggering time, a corresponding warning signal is generated.

[0068] In some embodiments, the status display data 15 includes spectral data 12, predicted temperature, mold-locking quality score, flow rate control command 13, voltage control command 14, and warning signal.

[0069] In some embodiments, based on the spectral data 12, it is rendered as a two-dimensional spectral curve with wavelength as the abscissa and optical power as the ordinate.

[0070] In some embodiments, the predicted temperature and mold-locking quality score that change over time are rendered as a time-series curve that is updated in real time.

[0071] In some embodiments, the current values ​​of the flow rate control command 13 and the voltage control command 14 are displayed as digital readings.

[0072] In some embodiments, when a warning signal is received, the graphical user interface 2 is triggered to change its state.

[0073] In some embodiments, the two-dimensional spectral curve, digital readings, and the changing state of the graphical user interface 2 are combined into a graphical user interface 2 image, and the graphical user interface 2 image is transmitted to the display device of the laser, so that the display device presents the graphical user interface 2 to the user.

[0074] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism, characterized in that, The laser includes a mode-locked laser resonant cavity, an optical fiber beam splitter connected to the mode-locked laser resonant cavity, an optical fiber Bragg grating sensor, an online spectrometer, a collaborative control processor, a precision microfluidic pump, and an intracavity electro-optic modulator. The mode-locked laser resonant cavity includes a gain fiber, a semiconductor saturable absorber, and an in-cavity electro-optic modulator. The gain fiber is thermally coupled to a microfluidic heat dissipation channel, which is connected to the fluid circuit of the precision microfluidic pump. The fiber Bragg grating sensor is integrated into a section of the gain fiber and is used to collect temperature data from the gain fiber. The online spectrometer is connected to the output end of the mode-locked laser resonator via the fiber optic beam splitter, and is used to collect the spectral data of the laser. The collaborative control processor is connected to the fiber Bragg grating sensor, the online spectrometer, the precision microfluidic pump, and the intracavity electro-optic modulator via a data and control bus. The collaborative control processor is used to receive and preprocess the temperature data and the spectral data, extract the average temperature and temperature drift rate from the preprocessed temperature data, and extract the spectral width and center wavelength from the preprocessed spectral data. The collaborative control processor is further used to run a thermal state prediction model based on the average temperature and the temperature drift rate to generate a predicted temperature, and to run a mode-locking quality evaluation model based on the spectral width and the center wavelength to generate a mode-locking quality score. The collaborative control processor is further used to train a collaborative control model based on a preset optimization target, and input the predicted temperature and the mode-locking quality score into the collaborative control model to generate the flow rate control command for the precision microfluidic pump and the voltage control command for the intracavity electro-optic modulator. The collaborative control processor internally sets up a decision logic and transmits the control command that satisfies the decision logic to the precision microfluidic pump and the intracavity electro-optic modulator for execution; The collaborative control processor is further used to issue early warnings for deviations from the safe working range of the predicted temperature and the mold-locking quality score, and to display status data through a graphical user interface.

2. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The fiber Bragg grating sensor is integrated into the gain fiber section and is used to collect temperature data of the gain fiber. The online spectrometer is connected to the output end of the mode-locked laser resonator via the fiber beam splitter. The process of collecting spectral data of the laser includes: Within a predetermined section of the fiber core of the gain fiber, a grating region is formed by ultraviolet laser exposure. To query the reflection spectrum of the grating region, a broadband light source is input at the first port of the fiber circulator, the second port is connected to the gain fiber containing the grating region, and the third port is connected to the spectral detection unit. The spectral detection unit is signal-connected to the cooperative control processor, forming the acquisition link for the temperature data. The optical signal emitted by the broadband light source propagates in the gain fiber, and when it passes through the grating region, the optical signal that satisfies the Bragg wavelength condition is reflected. When the temperature of the gain fiber changes, the effective refractive index and grating period of the fiber core are altered by the thermo-optic effect and thermal expansion effect, resulting in a drift of the reflected Bragg wavelength. The collaborative control processor calculates the Bragg wavelength drift based on the drifted Bragg wavelength, and obtains the temperature data of the gain fiber based on the linear relationship between the Bragg wavelength drift and the temperature change. The fiber optic beam splitter separates a small portion of the optical signal and imports the small portion of the optical signal into the input port of the online spectrometer. The small portion of the optical signal entering the online spectrometer is collimated by the collimation system inside the online spectrometer to form a parallel beam, and the parallel beam is incident on the diffraction grating. The diffraction grating utilizes the dispersion effect to expand the parallel beams of different wavelengths in space, causing the parallel beams to exit at different diffraction angles. Through the focusing system, the beams are projected and imaged onto the photosensitive surface of the photodetector array. The photosensitive surface of the photodetector array is divided into pixel units. The online spectral analyzer collects the electrical signals output by the pixel units, generates the spectral data, and transmits the spectral data to the collaborative control processor.

3. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The collaborative control processor is used to receive and preprocess the temperature data and the spectral data. The process of extracting the average temperature and temperature drift rate from the preprocessed temperature data and extracting the spectral width and center wavelength from the preprocessed spectral data includes: The collaborative control processor performs noise filtering and timestamp alignment on the temperature data and the spectral data to obtain preprocessed temperature data and spectral data. Within a preset time window, the arithmetic mean of continuous temperature data points of the preprocessed temperature data is calculated to obtain the average temperature within the time window. Within the time window, linear regression analysis is performed on continuous temperature data points of the preprocessed temperature data, and the slope of the fitted line of the linear regression analysis is taken as the temperature drift rate. The center wavelength of the spectrum is calculated using the power-weighted average method for the preprocessed spectral data. The spectral width is calculated using the full width at half maximum (FWHM) method for the preprocessed spectral data. The FWHM method includes taking the half-peak power of the preprocessed spectral data, finding two wavelengths corresponding to the half-peak power on both sides of the spectral profile, and taking the absolute value of the difference between the two wavelengths as the spectral width.

4. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The process of generating a predicted temperature by running a thermal state prediction model based on the average temperature and the temperature drift rate includes: At each discrete time step, the cooperative control processor combines the average temperature and the temperature drift rate into an input vector, and inputs the consecutive input vectors into the long short-term memory network model with a long short-term memory network architecture in chronological order, wherein the long short-term memory network architecture is composed of sequentially connected memory units; At each time step, the memory unit updates its internal long-term memory state based on the current input vector and the memory unit output of the previous time step through the forget gate, input gate and output gate inside the memory unit. The update process includes forgetting the historical information of the long-term memory state, storing the new information of the current input vector, and selecting information from the updated long-term memory state as the output result of the current time step. After the input vector passes through the memory unit, the last layer of the long short-term memory network architecture outputs a feature vector containing the historical information. The feature vector is then input into a fully connected layer, which performs a linear transformation on the feature vector to generate the predicted temperature.

5. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The process of generating a mode-lock quality score by running a mode-lock quality evaluation model based on the spectral width and the center wavelength includes: For each sample consisting of the spectral width and the center wavelength, the mode-locking state category label corresponding to the sample is manually labeled to form a labeled training set. The mode-locking state category label includes stable mode-locking and unstable mode-locking. The labeled dataset is trained using a support vector machine (SVM) classification algorithm. In the two-dimensional feature space defined by the spectral width and the center wavelength, the optimal decision boundary is found that separates sample points corresponding to different mode-locking state category labels with the maximum interval. After training, the trained mode-locking quality evaluation model and its parameters are obtained, and the model parameters, including support vectors and their coefficients, are stored in the collaborative control processor. During real-time operation of the laser, the cooperative control processor combines the currently calculated spectral width with the center wavelength into a real-time feature vector; The model lock quality evaluation model receives the real-time feature vector and calculates the relationship between the real-time feature vector and the stored support vector through a preset kernel function. The mold-locking quality evaluation model calculates the decision function value based on the relationship and the pre-stored coefficients; The decision function value is normalized using a preset mapping function to generate the mode-locking quality score.

6. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The process of training the collaborative control model based on the preset optimization objective includes: In the collaborative control model, a reward function is defined to maximize the mode-locking quality score and maintain the predicted temperature within a preset optimal temperature range as the optimization objective, and an instantaneous reward value is calculated. Using the cooperative control model as an agent, the agent learns by interacting with the laser's simulation environment. In each interaction step, the agent selects an action to output to the simulation environment based on the current state, and obtains the reward value and the next state from the simulation environment. By employing a reinforcement learning algorithm, the agent updates its internal policy network parameters until it learns the optimal policy that maximizes the cumulative expected reward. The state is composed of the predicted temperature and the mode-locking quality score, and the action is composed of the flow rate control command and the voltage control command.

7. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 6, characterized in that: The process of inputting the predicted temperature and the mode-locking quality score into the collaborative control model to generate the flow rate control command for the precision microfluidic pump and the voltage control command for the intracavity electro-optic modulator includes: The predicted temperature generated by the thermal state prediction model and the mode-locking quality score generated by the mode-locking quality evaluation model are combined into a state vector and input into the cooperative control model that has learned the optimal strategy. The cooperative control model that has learned the optimal strategy performs forward calculation on the input state vector and outputs an action vector, which includes the flow rate control command for the precision microfluidic pump and the voltage control command for the intracavity electro-optic modulator.

8. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The process by which the collaborative control processor internally sets up decision logic and transmits the control command that satisfies the decision logic to the precision microfluidic pump and the intracavity electro-optic modulator for execution includes: The control commands include the flow rate control command and the voltage control command; The control instruction currently being executed is read from the register of the control interface of the cooperative control processor; the difference between the newly generated control instruction and the currently being executed control instruction is calculated; and the difference is compared with a preset execution threshold. If the difference is greater than the execution threshold, the newly generated control instruction is considered a valid update instruction, and the valid update instruction is transmitted to the precision microfluidic pump and the intracavity electro-optic modulator for update execution. If the difference does not exceed the execution threshold, the currently executing control instruction remains unchanged, and no update operation is performed.

9. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The process of issuing a warning for deviations from the safe operating range in the predicted temperature and the mold-locking quality score includes: Pre-set limits for a safe working range, including an upper temperature limit, a lower temperature limit, and a lower limit for mold clamping quality score; compare the predicted temperature, the mold clamping quality score, and the limits. When the predicted temperature exceeds the upper limit of the temperature range, the lower limit of the temperature range, and the mold-locking quality score is lower than the lower limit of the mold-locking quality score, and the duration of the deviation exceeds the preset warning trigger duration, a corresponding warning signal is generated.

10. A laser with a microfluidic heat dissipation and collaborative mode-locking mechanism according to claim 1, characterized in that: The process of displaying status data through a graphical user interface includes: The status display data includes the spectral data, the predicted temperature, the mode-locking quality score, the flow rate control command, the voltage control command, and the warning signal; Based on the spectral data, it is rendered as a two-dimensional spectral curve with wavelength as the horizontal axis and optical power as the vertical axis; The predicted temperature and the mold-locking quality score, which change over time, are rendered as a real-time updated time series curve. The current values ​​of the flow rate control command and the voltage control command are displayed as digital readings; When the warning signal is received, the graphical user interface state is changed. The two-dimensional spectral curve, the digital reading, and the changing state of the graphical user interface are combined into a graphical user interface image, and the graphical user interface image is transmitted to the display device of the laser, whereby the display device presents the graphical user interface to the user.

Images