Ultra-short pulse laser cold machining self-adaptive regulation and control system and method

Abstract

The invention discloses an ultra-short pulse laser cold machining self-adaptive regulation and control system and method. The self-adaptive regulation and control system comprises a signal processing and feedback control module, an ultra-short pulse laser, a real-time phase change monitoring module and a motion control platform, wherein the ultra-short pulse laser, the real-time phase change monitoring module and the motion control platform are electrically connected with the signal processing and feedback control module. A spectrograph, a high-speed infrared camera, an interferometer and a laser processing device are integrally designed, a high-precision phase change monitoring system is integrated, and synchronous monitoring of three different physical quantities of spectrum, temperature and morphology of a processed area of a workpiece of an ultra-short pulse laser is achieved; through real-time signal processing and feedback, real-time sensing and feedback control of the phase change behavior in the material machining process are achieved, so that it is ensured that the machining process is always in a cold machining mode, the machining precision, the surface quality and the efficiency are improved, and the method is particularly suitable for precision machining of heat-sensitive materials or high-reflection materials or composite materials.

Description

Technical Field

[0001] This invention relates to the field of laser processing technology, specifically to an adaptive control system and method for ultrashort pulse laser "cold processing". Background Technology

[0002] Ultrashort pulse laser (picosecond, femtosecond laser) processing technology is hailed as a paradigm of "cold processing." Due to its extremely short pulse width, energy is injected within a timeframe far shorter than the material's thermal diffusion time. Theoretically, this allows for material removal through direct sublimation or plasmaification, thus minimizing thermal damage such as melting, microcracks, and recast layers. This characteristic makes it a promising candidate for cutting-edge manufacturing fields with strict limitations on thermal effects, such as precision medical devices, aerospace components, and semiconductor devices. Ideal "cold processing" requires the entire process to be strictly controlled within a region of the material that is neither thermally nor thermally sensitive.

[0003] However, in practical industrial applications, the "cold processing" effect of ultrashort pulse lasers is highly dependent on precise process parameter windows. When laser parameters deviate slightly due to factors such as equipment fluctuations, uneven material microstructure, or overlapping scanning paths, unexpected heat accumulation can easily occur in localized areas, causing the material to revert from the ideal "sublimation" removal mechanism to "melting" removal, forming a microscale heat-affected zone. This unreliability of the process and uncertainty of the results severely restrict the large-scale, industrial application of ultrashort pulse lasers in high-end manufacturing. Therefore, developing an intelligent control system capable of sensing the processing status in real time and dynamically adjusting parameters is key to overcoming the bottleneck in the application of ultrashort pulse laser "cold processing" technology.

[0004] Existing technical solutions mainly focus on offline optimization of process parameters and offline detection of processing results. For example, a database of process parameters is established through extensive experiments, or the processed surface is observed after processing using equipment such as scanning electron microscopes and white light interferometers to infer the quality of the processing. This approach is inefficient, costly, and cannot intervene in instantaneous anomalies that occur during processing. Some advanced solutions attempt to introduce process monitoring, such as the invention patent application CN202510042306.1, "A Femtosecond Laser Processing System, Method, and Equipment with Temperature Monitoring Function." This device monitors the temperature distribution of the processing area in real time using an infrared thermal imager, which can effectively control the heat accumulation effect during processing, avoid the formation of a heat-affected zone, achieve "cold processing" in the general sense, and achieve high processing accuracy and excellent surface quality.

[0005] However, existing technologies like CN202510042306.1 can only monitor a single parameter, temperature, and cannot comprehensively reflect the complex phase transition process of materials under the action of ultrashort pulse lasers. They also lack the comprehensive monitoring capability for key phase transition indicators such as plasma formation and spectral characteristics. In addition, the current lack of intelligent decision-making algorithms based on multi-parameter fusion makes it impossible to achieve adaptive optimization and control of process parameters, and manual intervention is still required to adjust parameters, which limits the improvement of processing efficiency and automation level. Summary of the Invention

[0006] The purpose of this invention is to construct an intelligent processing closed-loop system with the capabilities of "real-time perception, accurate judgment, and rapid execution" by integrating highly sensitive real-time phase transition monitoring technology with intelligent adaptive control algorithms. This fundamentally solves the above-mentioned problems and ensures high consistency, reliability, and extreme processing quality in the ultrashort pulse laser "cold processing" process.

[0007] To achieve the above objectives, this invention designs an adaptive control system for "cold processing" of ultrashort pulse lasers. The adaptive control system includes a signal processing and feedback control module, an ultrashort pulse laser, a real-time phase transition monitoring module, and a motion control platform, all electrically connected to the signal processing and feedback control module. The motion control platform includes a worktable and a controller, and the observation probes of the ultrashort pulse laser, the beam transmission and focusing module, and the real-time phase transition monitoring module are physically fixed together. A beam transmission and focusing module is provided between the laser output of the ultrashort pulse laser and the workpiece being processed on the worktable. The beam transmission and focusing module includes at least a focusing lens assembly (the lens assembly is existing optical technology), and the monitoring point is located near the focal plane of the lens assembly or within 2 times the focal length. The real-time phase change monitoring module includes a spectrometer, an infrared thermal imager, and an interferometer, which synchronously receive spectral, temperature, and morphological signals of the processed area of ​​the workpiece.

[0008] Furthermore, the interferometer reference beam source includes an independent path reference beam split from a laser source by a beam splitter, or one of the beams used as a reference source employing wavelength division multiplexing technology.

[0009] Furthermore, the spectrometer, infrared thermal imager, and interferometer are symmetrically mounted around the focusing lens assembly, with the observation direction all focused on the area of ​​the workpiece being processed.

[0010] Furthermore, the beam transmission and focusing module includes at least a beam splitting assembly, which includes at least a first semi-transparent mirror disposed in the optical path between the laser output and the focusing lens assembly.

[0011] Furthermore, the beam splitting assembly also includes a second semi-transparent and a third ...

[0012] Furthermore, the beam splitting assembly also includes a fourth semi-transparent mirror disposed in the optical path between the first semi-transparent mirror and the focusing lens assembly, and a fifth semi-transparent mirror disposed in the reflected optical path of the fourth semi-transparent mirror. The reflected light from the target area enters the interferometer sequentially through the fourth, first, second, third, and fifth semi-transparent mirrors, and the reference light enters the interferometer sequentially through the first, fourth, and fifth semi-transparent mirrors. The reference beam originates directly from the laser source. The reflected light from the target area enters the infrared thermal imager sequentially through the fourth, first, second, and third semi-transparent mirrors, and then sequentially through the fourth, first, and second semi-transparent mirrors and the first reflector to enter the spectrometer. The positions of the spectrometer and the infrared thermal imager can be interchanged.

[0013] Furthermore, a third and fourth reflecting mirror, as well as a sixth semi-transparent and semi-reflective mirror, are designed on the original laser reflection path of the first semi-transparent and semi-reflective mirror.

[0014] On the other hand, an adaptive control method for ultrashort pulse laser "cold processing" is provided. This adaptive control method relies on joint monitoring data from a spectrometer, infrared thermal imager, and interferometer to obtain more accurate and comprehensive information on the laser cold processing state of the processed area, including plasma spectra, temperature fields, and morphological changes. It also captures the phase transition state of the material under laser irradiation in real time. The method includes the following steps: S1. System initialization: Set basic laser parameters and phase transition monitoring thresholds according to the workpiece material; if the initialization data is adjusted based on feedback data received in step S4 during processing, proceed directly to step S3 after the adjustment is completed. S2. Calibration and Positioning: Start the motion control platform, position the laser focus to the workpiece to be processed area, and calibrate the sensor alignment of the monitoring module; S3. Real-time processing and monitoring: The ultrashort pulse laser is turned on for processing, while the phase transition monitoring module continuously collects joint monitoring data from the spectrometer, infrared thermal imager and interferometer of the processing area, and transmits it to the signal processing unit in the signal processing and feedback control module. S4. Signal Processing and Feedback: The signal processing unit analyzes the data in real time and identifies the phase transition state by combining plasma spectrum, temperature field, and morphological changes. If an anomaly is detected, the feedback control loop is triggered to adjust the laser parameters and scanning path. If the initialization parameters need to be adjusted, the process returns to step S1; otherwise, it proceeds to step S5. S5. Adaptive control cycle: Adjust the relative processing position of the worktable and / or laser based on the feedback signal, i.e., adjust the processing path, return to step S3, and continue to execute the monitoring-analysis-control cycle until the processing task is completed, ensuring that the whole process is in "cold processing" mode.

[0015] S6. Post-processing and data recording: After processing is completed, the system automatically saves the process data for quality traceability and process optimization.

[0016] Furthermore, the signal processing employs machine learning algorithms for feature classification and phase transition state prediction, outputting the phase transition level, including the following steps: S41. Input of real-time phase change monitoring data; S42. Feature extraction: Machine learning algorithms are used to extract spectral feature vectors, temperature feature vectors, and morphological feature vectors respectively. S43. Phase transition state identification: Based on the interrelationship between the spectral characteristics, temperature characteristics, and morphological characteristics of the material during laser cold processing, establish a relationship model between each feature vector and the phase transition state. The relationship model includes spectral feature functions, temperature feature functions, and morphological feature functions. The relationship model includes a theoretical relationship model or an empirical relationship model. S44. Parameter comparison generates control instructions, the control instructions including laser parameter adjustment instructions and / or motion path update instructions.

[0017] Furthermore, the signal processing method for jointly identifying phase transition states through plasma spectroscopy, temperature field, and morphological changes includes: a smart phase transition state identification method based on multi-dimensional feature fusion; The intelligent phase transition state identification method based on multi-dimensional feature fusion differs from traditional methods that rely solely on a single temperature or spectral intensity threshold for simple judgment. This invention employs a weighted fusion algorithm to comprehensively evaluate three types of feature parameters, establishing a phase transition state evaluation function (also known as an empirical relational model evaluation function). S=α×f(X_spectrum)+β×g(X_thermal)+γ×h(X_morphology); Among them, the weighting coefficients α, β, and γ are adaptively adjusted according to the material type and processing conditions; f(X_spectrum), g(X_thermal), and h(X_morphology) are the spectral feature function, temperature feature function, and morphological feature function, respectively, and X_spectrum, X_thermal, and X_morphology are the spectral feature vector, temperature feature vector, and morphological feature vector, respectively.

[0018] Furthermore, the spectral feature vector is defined as X_spectrum=[Te,ne,R_AI,W_line], where Te is the plasma electron temperature calculated by fitting the atomic spectral line intensity ratio through the Boltzmann distribution, ne is the electron density determined based on the Stark broadening effect, R_AI is the ratio of atomic spectral line to ion spectral line intensity (reflecting whether the material removal mechanism is dominated by sublimation or ionization), and W_line is the Lorentz broadening of the feature spectral line (characterizing the stability of the plasma state). The temperature feature vector is defined as X_thermal=[T_max,∇T,dT / dt,A_HAZ], which includes the peak temperature T_max of the processing area measured by the infrared thermal imager, the radial temperature gradient ∇T, the temperature rise rate dT / dt, and the heat-affected zone expansion area A_HAZ. The morphology feature vector is defined as X_morphology=[ΔZ,C_edge,η_removal], which includes the surface height change ΔZ measured by the confocal interferometer, the processing edge sharpness coefficient C_edge, and the material removal efficiency η_removal.

[0019] Furthermore, the corresponding function expression adopts a normalized weighted form based on an empirical relational model. The spectral characteristic function: f(X_spectrum)=w1×sigmoid(Te / Te_critical)+w2×log(ne / ne_threshold)+w3×(1-R_AI / R_ideal)+w4×(W_line / W_reference), Where Te_critical is the critical electron temperature for material sublimation, and ne_threshold is the density threshold for stable plasma formation. The temperature characteristic function: g(X_thermal)=w5×tanh(T_max / T_melting)+w6×(∇T / ∇T_ideal)^0.5+w7×|dT / dt| / |(dT / dt)_cold|+w8×(A_HAZ / A_laser_spot); Ensure that the temperature is always below the material's melting point T_melting and that the heat-affected zone is strictly limited; The morphological feature function: h(X_morphology)=w9×(ΔZ / Z_target)+w10×C_edge+w11×η_removal; Among them, the weight coefficients w1-w11 are obtained by machine learning training on experimental data of "cold working" of different materials. When the S value is maintained within the preset target range of "cold working" [S_min, S_max], it indicates that the material is removed by direct sublimation or plasma treatment, avoiding the melting and resolidification process.

[0020] The advantages and beneficial effects of this invention are as follows:

[0021] The purpose of this invention is to provide an adaptive control system and method for ultrashort pulse laser "cold processing" based on real-time phase transition monitoring. Through the integrated design of a spectrometer, high-speed infrared camera, interferometer, and laser processing device, a high-precision phase transition monitoring system is integrated to achieve simultaneous monitoring of three different physical quantities—spectral data, temperature, and morphology—of the workpiece processing area using an ultrashort pulse laser. This enables the backend system to analyze and judge the actual state of laser cold processing from the perspective of different physical phenomena, providing a fundamental guarantee for the accuracy and timeliness of on-site state judgment, and facilitating subsequent improvements in the efficiency and quality of laser cold processing. Through real-time signal processing and feedback, real-time perception and feedback control of phase transition behavior during material processing are achieved. This invention utilizes the ultrashort pulse characteristics of femtosecond or picosecond lasers, combined with plasma spectroscopy, interferometry, or thermal radiation monitoring technology, to capture the phase transition state of materials under laser irradiation in real time. Based on the monitoring data, parameters such as laser power, pulse frequency, and scanning speed are dynamically adjusted to ensure that the processing process is always in "cold processing" mode, thereby improving processing accuracy, surface quality, and efficiency. It is particularly suitable for the precision processing of heat-sensitive materials, highly reflective materials, or composite materials.

[0022] Compared with the prior art, the present invention also has the following advantages: (1) Through real-time phase change monitoring by multi-sensor fusion, the instantaneous phase change state of the material under laser action can be captured in situ and accurately, realizing direct diagnosis and early warning of the "cold processing" process, fundamentally avoiding the lag and non-specificity of traditional temperature monitoring, and significantly improving the observability and controllability of the processing process.

[0023] (2) Based on intelligent algorithm adaptive feedback control, the laser power, repetition frequency and other key parameters can be dynamically adjusted according to the real-time phase change state to form a fast and accurate process closed loop, effectively locking the processing process within the ideal "cold processing" window, ensuring extremely high consistency and repeatability of processing quality, and completely eliminating the generation of thermal damage.

[0024] (3) The system has the ability to learn and optimize process parameters, adapt to dynamic working conditions such as material differences and environmental fluctuations, automatically broaden the process window for stable processing, reduce the reliance on manual experience and tedious trial and error, and significantly improve the intelligence level, process robustness and overall production efficiency of ultra-short pulse laser "cold processing".

[0025] The problems actually solved by this invention also include: First, there is the unpredictability and lag in processing conditions. In current ultrashort pulse laser processing, process parameters are usually preset based on experience and offline testing. However, during actual processing, many dynamic factors, such as the inhomogeneity of the material's microstructure, minute changes in surface condition, the thermal accumulation effect in the overlapping area of ​​the scanning path, and fluctuations in the laser's output power, can cause instantaneous energy density deviations from the ideal "cold processing" window in local areas, leading to unexpected melting phenomena. Due to the lack of real-time, in-situ monitoring methods for the phase transformation state of the material in the processing area, operators cannot immediately perceive these deviations from the ideal processing state and can only discover defects through offline inspection after processing. By this time, the workpiece is already scrapped, resulting in a huge waste of resources and time.

[0026] Secondly, there is the extreme narrowness of the ideal "cold processing" window and the blind nature of parameter optimization. For ultrashort pulse lasers to achieve true "cold processing," the process parameter window is extremely narrow and exceptionally sensitive to parameter fluctuations. Current technologies heavily rely on operator experience and tedious, time-consuming trial-and-error methods for offline parameter optimization. This method is inefficient, costly, and prone to failure if processing conditions change even slightly, leading to unstable processing quality. Current technologies lack an intelligent method that can automatically identify, accurately lock onto, and dynamically track this narrow optimal process window during processing.

[0027] Finally, there is the passive and open-loop nature of the control strategy. Even if some systems integrate certain monitoring functions, their response modes are mostly limited to simple "threshold alarms" or "emergency shutdowns," failing to achieve proactive, adaptive closed-loop control. When an abnormal signal is detected, the system cannot accurately determine the root cause of the abnormality and therefore cannot execute precise and timely parameter correction commands, nor can it automatically and quickly pull the processing from the deviation state back to the ideal "cold processing" track. Attached Figure Description

[0028] Figure 1 This is a block diagram illustrating the principle of an adaptive control system for ultrashort pulse laser "cold processing" according to the present invention. Figure 2 This is a block diagram of another adaptive control system for "cold processing" using ultrashort pulse lasers. Figure 3 This is a block diagram of another adaptive control system for "cold processing" using ultrashort pulse lasers. Figure 4 This is a block diagram of another type of adaptive control system for ultrashort pulse laser "cold processing"; Figure 5This is a flowchart of an adaptive control method for "cold processing" using ultrashort pulse lasers; Figure 6 This is the flowchart for the signal processing and feedback module.

[0029] Marked in the image: 1. Focusing lens assembly; 2. Spectrometer assembly; 4. Spectrometer; 5. Infrared thermal imager; 6. Interferometer; 21. The first half-mirror; 22. The second half-mirror; 23. The third half-mirror; 24. The fourth half-mirror; 25. The fifth half-mirror; 26. The sixth half-mirror; 31. First reflecting mirror; 32. Second reflecting mirror; 33. Third reflecting mirror; 34. Fourth reflecting mirror. Detailed Implementation

[0030] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and examples. The following examples are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0031] First, some terms will be explained.

[0032] Phase transition monitoring: a key technology in laser material processing, refers to the use of sensors to capture in real time the characteristic physical signals generated when materials undergo phase transitions under laser irradiation. By analyzing changes in the thermal radiation spectrum of the molten pool, plasma plume morphology, or acoustic emission signals, it is possible to accurately determine whether the material has reached the critical point of phase transition such as melting or vaporization. This technology provides a core basis for achieving adaptive adjustment of process parameters and ensuring consistent processing quality.

[0033] Ultrashort pulse lasers refer to lasers with pulse durations ranging from picoseconds to femtoseconds. Their core characteristics are extremely high peak power and extremely short energy-time interaction. Energy is injected instantaneously within a timeframe far shorter than the material's thermal diffusion time, allowing the material to be removed through direct sublimation or plasmaification, with almost no heat transfer to the surrounding environment, thus achieving so-called "cold processing." This technology effectively avoids damage such as thermal melting and microcracks, and has key applications in precision machining, medical, and scientific research fields.

[0034] Adaptive control: The laser processing system monitors the processing status in real time (such as molten pool morphology and plasma signal), compares the collected data with a preset model, and then dynamically adjusts process parameters such as laser power and scanning speed, forming a closed-loop control of "perception-decision-execution". This technology can automatically compensate for interference factors in processing and is the core of achieving high consistency and intelligent manufacturing.

[0035] In the field of laser cold processing technology, existing real-time monitoring and control technologies, such as the "A Femtosecond Laser Processing System, Method, and Equipment with Temperature Monitoring Function" (application number CN202510042306.1), although introducing the concept of process monitoring and attempting to improve processing quality by monitoring temperature, have the following key drawbacks in achieving truly reliable ultrashort pulse laser "cold processing." First, the temperature monitoring signal relied upon by this scheme is a severely lagging and indirect physical quantity for determining the critical state of "cold processing." Second, the monitoring target of this scheme has a low correlation with the core physical process of cold processing—the instantaneous selection of the material phase transition path—and lacks specific identification capabilities. Third, the preset static threshold control logic of this scheme is difficult to adapt to the strong nonlinearity and dynamic time-varying nature of the ultrashort pulse laser processing process.

[0036] To address the systemic shortcomings of existing technologies, this invention aims to construct a precision "cold processing" control system capable of achieving full-chain intelligence from "sensing" to "decision-making" to "execution." By abandoning traditional, lagging temperature monitoring, it adopts a real-time monitoring strategy that fuses multimodal information highly sensitive to phase transition processes (such as plasma plume characteristic spectra and transient changes in material surface reflectivity). This enables in-situ, precise diagnosis of the instantaneous phase transition state and path of the material, providing proactive and direct criteria for control. Based on this, the system establishes a rapid adaptive closed loop based on accurate real-time phase transition state discrimination. It can proactively and dynamically adjust laser process parameters (such as pulse energy and repetition frequency) at the initial stage when minute signs of deviation from the ideal "cold processing" trend appear in the processing state, rapidly correcting and stabilizing the processing within the ideal parameter space, thereby eliminating thermal damage at its source. Ultimately, by introducing an intelligent control algorithm with self-learning capabilities, the system can learn the complex nonlinear relationship between process parameters and phase transition states online, autonomously calculate the optimal parameter adjustment strategy based on real-time monitoring information, significantly enhance robustness to dynamic working conditions, effectively broaden the stable "cold processing" process window, and ensure extremely high consistency, repeatability, and reliability throughout the entire process.

[0037] Example 1: This invention is an adaptive control system for ultrashort pulse laser "cold processing," such as... Figure 1As shown, the adaptive control system includes a signal processing and feedback control module, and an ultrashort pulse laser, a real-time phase transition monitoring module, and a motion control platform, all electrically connected to the signal processing and feedback control module. The motion control platform includes a worktable and a controller. The observation probes of the ultrashort pulse laser, beam transmission and focusing module, and real-time phase transition monitoring module are physically fixed and move as a whole relative to the workpiece to ensure the stability of the optical path. The monitoring and control are connected to the signal processing and feedback control module via electrical signals. The workpiece is located on the worktable. The emphasis here on the integrated fixed connection of the observation probe with the ultrashort pulse laser and beam transmission and focusing module is mainly based on the requirements of minimum optical path design and optical path stability. Back-end processing components that affect the overall size of the optical path device can be set up separately. In fact, this is not the only option. Existing miniaturization technologies for cameras, spectrometers, etc., are already quite good, and they can be directly fixed and installed on the periphery of existing optical path components. If a beam splitting optical path is used, only the fixed connection characteristics need to be ensured, and there are no special requirements for the size of the sensor. A beam transmission and focusing module is provided between the laser output of the ultrashort pulse laser and the workpiece being processed on the worktable. The beam transmission and focusing module includes at least a focusing lens assembly (1 (the lens assembly is existing optical technology), and the monitored point is located near the focal plane of the lens assembly or within 2 times the focal length). The real-time phase transition monitoring module includes a spectrometer 4, an infrared thermal imager 5, and an interferometer 6. These instruments synchronously receive spectral, temperature, and morphological signals from the processed area of ​​the workpiece. In this embodiment, the spectrometer (spectral sensor) collects the plasma emission spectrum of the processed area and identifies the type and degree of material phase transition through characteristic spectral lines (such as atomic and ionic lines). The infrared thermal imager is a high-speed infrared imager, primarily monitoring the temperature distribution at the workpiece processing point to prevent the expansion of the heat-affected zone due to heat accumulation. The interferometer is a confocal interferometer that measures the surface morphological changes of the workpiece in real time, identifying melting, vaporization, or solidification states. The monitoring data is transmitted via optical fiber to the signal processing and feedback control module, where machine learning algorithms are used for feature classification and phase transition state prediction, outputting the phase transition level.

[0038] The system works as follows: an ultrashort pulse laser generates femtosecond or picosecond laser pulses, which are then applied to the workpiece surface after passing through a beam transmission and focusing module; a real-time phase change monitoring module collects signals such as plasma spectrum, temperature field, and morphological changes in the processing area through multi-sensor fusion; a signal processing and feedback control module extracts features and identifies the phase change state of the collected signals, compares them with preset thresholds, and generates control commands; the motion control platform adjusts the laser parameters and processing path according to the control commands to achieve adaptive control.

[0039] After receiving the monitoring data, the signal processing and feedback control module first performs signal filtering and normalization to extract feature parameters. Then, based on a preset phase transition model and real-time data, it determines whether the current processing state deviates from the "cold processing" window. If it does, an adjustment command is generated using a PID controller or fuzzy logic algorithm to dynamically adjust parameters such as the laser's pulse energy, repetition frequency, and scanning speed. Simultaneously, the motion control platform adjusts the focusing position and processing trajectory to ensure processing consistency and accuracy.

[0040] Preferably, the interferometer reference beam source includes an independent path reference beam split from a laser source by a beam splitter, or one of the beams used as a reference source employing wavelength division multiplexing (WDM) technology; the other path split from the laser source by the beam splitter is the measurement beam, and the laser beam splitting technology can employ a classic Michelson beam splitting structure. The reference beam enters the interferometer through a folded-back optical path. Figure 1 (details omitted). In this embodiment, wavelength division multiplexing technology is used to introduce the reference beam into the interferometer through an optical fiber channel, avoiding mutual interference with the main optical path during long-distance transmission and when there is significant vibration interference.

[0041] Preferred, such as Figure 1 As shown in the figure, in this embodiment, the spectrometer 4, infrared thermal imager 5, and interferometer 6 are symmetrically mounted around the focusing lens assembly 1 (the spectrometer, infrared thermal imager, and interferometer are at an angle of 120 degrees to each other; due to the inconvenience of using text to express this in the figure, ...). Figure 1 The figures are indicated by labels, while the other figures are indicated by modular text. The observation direction is focused on the area of ​​the workpiece being processed.

[0042] This embodiment features a compact design for the ultrashort pulse laser, beam transmission and focusing module, and real-time phase transition monitoring module. The optical path and system hardware structure are relatively simple. The reference beam source for the interferometer can be directly introduced from the laser source. The disadvantage is that the direction of the reflected beam is not normal to the stage, resulting in oblique diffuse reflection and poor energy focusing. However, since the tilt angle is not large, the impact is minimal and there are no problems in engineering applications. In particular, the impact is even smaller for spectrometers and infrared thermal imagers. For the interferometer, it may be necessary to adjust the energy of the reference beam to obtain better interferometric measurement results.

[0043] This solution is, in fact, a convenient and easy-to-implement one. By integrating a spectrometer, a high-speed infrared camera, an interferometer, and a laser processing device, it enables simultaneous monitoring of three different physical quantities—spectral density, temperature, and morphology—of the workpiece's processed area. This allows the backend system to analyze and judge the true state of laser cold processing from the perspective of different physical phenomena, providing a fundamental guarantee for the accuracy and timeliness of on-site condition assessment, and facilitating subsequent improvements in the efficiency and quality of laser cold processing.

[0044] Example 2: The difference from Example 1 is that, as Figure 2 , 3 As shown in Figure 4, the beam transmission and focusing module of this embodiment includes at least a beam splitting assembly 2, and the beam splitting assembly includes at least a first semi-transparent mirror 21 disposed in the optical path between the laser output and the focusing lens assembly 1.

[0045] It should be further explained that: the term "semi-transparent and semi-reflective mirror" is a simplified name, and in some places it is also described as a dichroic mirror. For example, the first semi-transparent and semi-reflective mirror 21 can be a dichroic mirror, mainly to achieve the separation of the processing laser and the returned signal light. It mainly transmits the specific wavelength of the ultrashort pulse processing laser; in addition, it reflects the monitoring signal light band, mainly from the mid-far infrared band to the visible light band. Therefore, the dichroic mirrors used in the second semi-transparent and semi-reflective mirror 22 and the third semi-transparent and semi-reflective mirror 23 mainly transmit the mid-far infrared band to the high-speed infrared thermal camera and reflect the visible light band to the spectrometer. Specifically, different lenses should be selected according to the type of transmitted and reflected light waves. The design principle is the same for other embodiments.

[0046] Preferred, such as Figure 2 As shown, the beam-splitting assembly described in this embodiment further includes a second semi-transparent semi-reflective mirror 22 and a third semi-transparent semi-reflective mirror 23 sequentially arranged in the reflection optical path of the first semi-transparent semi-reflective mirror 21. A reflector is disposed in at least one reflection optical path of the second semi-transparent semi-reflective mirror 22 and the third semi-transparent semi-reflective mirror 23. Figure 2 As shown, a first reflector 31 is set in the reflected light path of the second semi-transparent and semi-reflective mirror 22, and a second reflector 32 is set in the reflected light path of the third semi-transparent and semi-reflective mirror 23. The reflected light from the first semi-transparent and semi-reflective mirror passes through the second and third semi-transparent and semi-reflective mirrors in sequence and enters a sensor such as a high-speed infrared camera in the real-time phase transition monitoring module. The reflected light from the first and second reflectors enters two sensors such as a spectrometer and an interferometer (the positions can be interchanged) in the real-time phase transition monitoring module, respectively. The reference beam source of the interferometer can be directly introduced from the laser source.

[0047] Compared to Embodiment 1, the real-time phase change monitoring module in this embodiment belongs to the reflective surface normal monitoring design mode. Because of the beam splitting assembly, the beam transmission and focusing module and the real-time phase change monitoring module in this embodiment are slightly less compact.

[0048] Example 3: The difference from Example 2 is that, as Figure 3As shown, in order to achieve another type of normal interferometry measurement, the beam splitting assembly further includes a fourth semi-transparent mirror 24 disposed in the optical path between the first semi-transparent mirror 21 and the focusing lens assembly 1, and a fifth semi-transparent mirror 25 disposed in the reflected optical path of the fourth semi-transparent mirror 24. The reflected light from the target area enters the interferometer sequentially through the fourth, first, second, third, and fifth semi-transparent mirrors, and the reference light enters the interferometer sequentially through the first, fourth, and fifth semi-transparent mirrors. The reference beam is directly introduced from the laser source. The reflected light from the target area enters the infrared thermal imager sequentially through the fourth, first, second, and third semi-transparent mirrors, and then sequentially through the fourth, first, and second semi-transparent mirrors and the first reflecting mirror 31 to enter the spectrometer. The positions of the spectrometer and the infrared thermal imager can be interchanged.

[0049] Note that, in addition to the differences mentioned above, the installation directions of the second and third semi-transparent and semi-reflective mirrors, as well as the installation position of the first reflector, are different in this embodiment compared to Embodiment 2. In fact, this embodiment suffers from significant energy loss, especially for interferometric reflected light, making it one of the designs with the greatest energy loss. However, for the laser pulse in this case, due to its high optical power, the impact on monitoring is not significant, and the project remains practical and feasible. In fact, in Embodiment 3, the installation direction of the second semi-transparent and semi-reflective mirror could be rotated by 90 degrees. In this way, the fifth semi-transparent and semi-reflective mirror would directly reflect the target light reflected by the second semi-transparent and semi-reflective mirror, reducing the energy loss of the first semi-transparent and semi-reflective mirror in the interferometer's optical path. Of course, the direction of the third semi-transparent and semi-reflective mirror and the position of the first reflector should also be adjusted appropriately, further reducing the energy of the reflected light from the target area received by the spectrometer and infrared thermal imager.

[0050] Example 4: like Figure 4 As shown, the difference between this embodiment and Embodiment 2 is that a third reflecting mirror 33, a fourth reflecting mirror 34, and a sixth semi-transparent reflecting mirror 26 are designed in the original laser reflection path of the first semi-transparent reflecting mirror 21. After the original laser beam is split by the first semi-transparent reflecting mirror 21, the reference light enters the interferometer through the third reflecting mirror 33, the fourth reflecting mirror 34, and the sixth semi-transparent reflecting mirror 26. Energy attenuation or phase adjustment elements can also be set in the optical path to match the energy and phase of the reference light with the reflected light from the target area. The reflected light from the target area of ​​the workpiece enters the interferometer through the first semi-transparent reflecting mirror 21 and the second semi-transparent reflecting mirror 22, and then enters the interferometer through the sixth semi-transparent reflecting mirror 26. The reflected light from the target area of ​​the workpiece enters the infrared thermal imager through the third semi-transparent reflecting mirror 23, and then enters the spectrometer through the first reflecting mirror 31. The positions of the spectrometer and the infrared thermal imager can be interchanged. In addition to the optical path designs of Embodiments 1 to 4, other optical path designs are also possible, which will not be listed here.

[0051] Example 5: An adaptive control method for ultrashort pulse laser "cold processing", such as Figure 5 As shown, the adaptive control method is characterized by using joint monitoring data from a spectrometer, an infrared thermal imager, and an interferometer to obtain more accurate and comprehensive information on the laser cold processing state, such as the plasma spectrum, temperature field, and morphological changes of the processed area, and to capture the phase transition state of the material under laser irradiation in real time. The method includes the following steps: S1. System initialization: Set basic laser parameters and phase transition monitoring thresholds according to the workpiece material; if the initialization data is adjusted based on feedback data received in step S4 during processing, proceed directly to step S3 after the adjustment is completed. The basic laser parameters include laser power, pulse frequency, scanning speed, and the angle of the lenses inside the laser. For specific definitions of phase transition levels, please refer to: Harilal, SS "Spectroscopic characterization of laser-induced plasmas for industrial applications" Applied Spectroscopy Reviews, Vol. 53, 2018, pp. 1-47, specifically pages 15-18: Spectral criteria for phase transition states, pages 22-25: Plasma classification standards, and pages 28-30: Threshold settings for industrial applications (the phase transition level threshold can be set with reference to this).

[0052] S2. Calibration and Positioning: Start the motion control platform, position the laser focus to the workpiece to be processed area, and calibrate the sensor alignment of the monitoring module; In this embodiment, a standard metal mesh sample is used. The characteristics of the mesh intersections are observed simultaneously by each sensor, the spatial offset matrix is ​​calculated and the coordinate transformation relationship is established to achieve precise alignment of the monitoring area; S3. Real-time processing and monitoring: The ultrashort pulse laser is turned on for processing, while the phase transition monitoring module continuously collects joint monitoring data from the spectrometer, infrared thermal imager and interferometer of the processing area, and transmits it to the signal processing unit in the signal processing and feedback control module. S4. Signal Processing and Feedback: The signal processing unit analyzes the data in real time and identifies the phase transition state by combining plasma spectrum, temperature field, and morphological changes. If an anomaly is detected, the feedback control loop is triggered to adjust the laser parameters and scanning path. If the initialization parameters need to be adjusted, the process returns to step S1; otherwise, it proceeds to step S5. After receiving the monitoring data, the signal processing and feedback module first performs signal filtering and normalization to extract feature parameters. Then, based on a preset phase transition model and real-time data, it determines whether the current processing state deviates from the "cold processing" window. If it does, an adjustment command is generated through a PID controller or fuzzy logic algorithm to dynamically adjust parameters such as the laser's pulse energy, repetition frequency, and scanning speed. Simultaneously, the motion control platform adjusts the focus position and processing trajectory to ensure processing consistency and accuracy. This includes dynamic focus position adjustment, adaptive processing trajectory adjustment, and synchronous adjustment by the motion control platform, all automatically implemented through the program. The focus position adjustment is crucial because the femtosecond laser processing output optical path contains many lenses and reflectors; the deflection angle of the laser needs to be changed by controlling the angle of the reflectors.

[0053] S5. Adaptive control cycle: Adjust the relative processing position of the worktable and / or laser based on the feedback signal, i.e., adjust the processing path, return to step S3, and continue to execute the monitoring-analysis-control cycle until the processing task is completed, ensuring that the whole process is in "cold processing" mode.

[0054] S6. Post-processing and data recording: After processing is completed, the system automatically saves the process data for quality traceability and process optimization.

[0055] In this embodiment, signal processing and feedback control are primarily based on machine learning methods. The training of the machine learning model mainly employs supervised learning. First, through numerous experiments involving ultrashort pulse laser processing of different materials, multi-dimensional feature data such as spectra, temperature, and morphology are systematically collected within a preset process parameter range. Simultaneously, offline detection methods such as scanning electron microscopy and surface profilometry are used to evaluate the processing results, establishing a training dataset containing feature vectors and processing quality labels. Subsequently, gradient boosting decision tree (GBDT) or deep neural network algorithms are used to train the dataset. The weight coefficients w1-w11 and fusion coefficients α, β, and γ are optimized using backpropagation to minimize the mean square error between the output value of the phase transition state evaluation function S and the actual "cold processing" quality evaluation result. Cross-validation is used during training to prevent overfitting, and an incremental learning mechanism enables the system to continuously update model parameters based on new processing data. Ultimately, an intelligent decision-making model is formed that can accurately predict the phase transition state under different working conditions and guide adaptive parameter control, realizing the transformation from offline experience-based dependence to online intelligent control.

[0056] This embodiment utilizes a spectrometer, an infrared thermal imager, and an interferometer to jointly monitor the laser "cold processing" process. It collects real-time and synchronous information on the plasma spectrum, temperature field, and morphological changes of the workpiece's processed area. This allows for more accurate real-time capture and judgment of the material's phase transition state under laser action. Compared to post-processing analysis without sensors and the uncertainty and incompleteness of information from a single sensor, the phase transition state judgment of this invention is more timely, accurate, and precise, providing a more basis for control. Therefore, the overall processing efficiency is higher and the quality is better.

[0057] Preferred, such as Figure 6 As shown, the feature is that the signal processing employs machine learning algorithms for feature classification and phase transition state prediction, outputting the phase transition level, including the following steps: S41. Input of real-time phase change monitoring data; S42. Feature extraction: Machine learning algorithms are used to extract spectral feature vectors, temperature feature vectors, and morphological feature vectors respectively. S43. Phase transition state identification: Based on the interrelationship between the spectral characteristics, temperature characteristics, and morphological characteristics of the material during laser cold processing, establish a relationship model between each feature vector and the phase transition state. The relationship model includes spectral feature functions, temperature feature functions, and morphological feature functions. The relationship model includes a theoretical relationship model or an empirical relationship model.

[0058] Based on the theoretical relationship model, this embodiment uses a 3-judgment-3 mode (three monitoring results represent the same state) to determine the phase transition state. Of course, a 3-judgment-2 mode can also be used (at least two of the three monitoring results represent the same state). That is, the phase transition state is determined independently based on three sets of feature vector values, and then the determination is made jointly. S44. Parameter comparison generates control instructions, the control instructions including laser parameter adjustment instructions and / or motion path update instructions.

[0059] The machine learning algorithm described in this embodiment uses a neural network prediction model based on a dual-temperature model for feature extraction. The model type is: using the Anisimov dual-temperature model combined with a deep neural network to establish a phase transition prediction model that couples electronic temperature and lattice temperature (for the dual-temperature model, see Anisimov et al. Soviet Physics JETP 1974, pp. 375-377). Input parameters: The 12-dimensional feature vector includes laser pulse energy (mJ), repetition frequency (kHz), scanning speed (mm / s), defocusing amount (μm), lattice temperature Tl (K), electron-phonon coupling coefficient, material thermal diffusivity, electron temperature Te (K), spectral intensity ratio R_ion_atom, plasma electron density, surface roughness Ra, ablation depth, etc. Connections: The 12 neurons in the input layer receive feature parameters, which are then transformed non-linearly through 3 hidden layers (64-32-16 neurons), with each layer using the ReLU activation function; Output results: The intermediate layer outputs the spectral feature vector [Te, ne, R_AI, W_line]; the temperature feature vector [T_max, ∇T, dT / dt, A_HAZ]; and the eigenvalues ​​of the morphological feature vector [ΔZ, C_edge, η_removal]. Finally, the probability distribution of the four phase transition levels is obtained through the softmax output layer.

[0060] The theoretical relationship model can be referenced as follows: Phase transition threshold model: Energy density criterion model Bulgakova et al. J Laser Micro / Nanoeng 2007, pp. 78-80; Plasma formation model: Keldysh ionization model, Gamaly Prog Quantum Electron 2013, pp. 280-295; Spectral prediction model: Plasma spectral emission model Griem Spectral Line Broadening by Plasmas 1974, pp. 89-156; Surface morphology evolution model: ablation depth prediction model Nolte et al. J Opt Soc Am B 1997, pp. 2716-2722; Machine learning augmentation model: Neural network phase transition prediction Ramprasad et al. npj Comput Mater 2017, Article 54; Online learning model: Shalev-Shwartz Found Trends ML 2012, pp. 125-180.

[0061] etc.

[0062] Example 6: The difference from Example 5 is that the phase transition state identification in this example is based on an empirical relationship model. The method of identifying the phase transition state by combining plasma spectrum, temperature field, and morphological changes includes: a smart phase transition state identification method based on multi-dimensional feature fusion. The intelligent phase transition state identification method based on multi-dimensional feature fusion differs from traditional methods that rely solely on a single temperature or spectral intensity threshold for simple judgment. This invention employs a weighted fusion algorithm to comprehensively evaluate three types of feature parameters, establishing a phase transition state evaluation function (also known as an empirical relational model evaluation function). S=α×f(X_spectrum)+β×g(X_thermal)+γ×h(X_morphology); Among them, the weighting coefficients α, β, and γ are adaptively adjusted according to the material type and processing conditions; f(X_spectrum), g(X_thermal), and h(X_morphology) are the spectral feature function, temperature feature function, and morphological feature function, respectively, and X_spectrum, X_thermal, and X_morphology are the spectral feature vector, temperature feature vector, and morphological feature vector, respectively.

[0063] Preferably, the spectral feature vector is defined as X_spectrum=[Te,ne,R_AI,W_line], where Te is the plasma electron temperature calculated by fitting the atomic spectral line intensity ratio through the Boltzmann distribution, ne is the electron density determined based on the Stark broadening effect, R_AI is the ratio of atomic spectral line to ion spectral line intensity (reflecting whether the material removal mechanism is dominated by sublimation or ionization), and W_line is the Lorentz broadening of the feature spectral line (characterizing the stability of the plasma state). The temperature feature vector is defined as X_thermal=[T_max,∇T,dT / dt,A_HAZ], which includes the peak temperature T_max of the processing area measured by the infrared thermal imager, the radial temperature gradient ∇T, the temperature rise rate dT / dt, and the heat-affected zone expansion area A_HAZ. The morphology feature vector is defined as X_morphology=[ΔZ,C_edge,η_removal], which includes the surface height change ΔZ measured by the confocal interferometer, the processing edge sharpness coefficient C_edge, and the material removal efficiency η_removal.

[0064] Preferably, the corresponding function expression adopts the normalized weighted form of the empirical relation model. The spectral characteristic function: f(X_spectrum)=w1×sigmoid(Te / Te_critical)+w2×log(ne / ne_threshold)+w3×(1-R_AI / R_ideal)+w4×(W_line / W_reference), Where Te_critical is the critical electron temperature for material sublimation, and ne_threshold is the density threshold for stable plasma formation. The temperature characteristic function: g(X_thermal)=w5×tanh(T_max / T_melting)+w6×(∇T / ∇T_ideal)^0.5+w7×|dT / dt| / |(dT / dt)_cold|+w8×(A_HAZ / A_laser_spot); Ensure that the temperature is always below the material's melting point T_melting and that the heat-affected zone is strictly limited; The morphological feature function: h(X_morphology)=w9×(ΔZ / Z_target)+w10×C_edge+w11×η_removal; Among them, the weight coefficients w1-w11 are obtained by machine learning training on experimental data of "cold working" of different materials. When the S value is maintained within the preset target range of "cold working" [S_min, S_max], it indicates that the material is removed by direct sublimation or plasma treatment, avoiding the melting and resolidification process.

[0065] The intelligent phase transition state identification method based on multi-dimensional feature fusion described in this embodiment overcomes the limitations of traditional single-sensor simple threshold judgment and establishes a precise phase transition state identification system based on the synergy of multi-physics field information. Specifically, the system first acquires plasma emission spectra in real time through a spectral sensor, extracting key spectral feature parameters such as the intensity ratio of atomic spectral lines to ion spectral lines, electron temperature Te, and electron density ne; simultaneously, it uses a high-speed infrared thermal imager to monitor the temperature distribution gradient of the processing area, the expansion rate of the heat-affected zone boundary, and the spatial uniformity index of the temperature field; and it uses a confocal interferometer to detect morphological features such as surface roughness changes, melt pool boundary clarity, and material removal efficiency; finally, it uses feature functions and the self-function of the machine learning model to weightedly calculate the phase transition state value and the corresponding confidence level.

[0066] Phase transition state values ​​can also be defined with specific grade values. Generally, phase transition grades 0-3 represent four states: solid-state processing, critical melting, vaporization-dominant, and plasmaization, respectively. The model also provides predicted values ​​for electron temperature and lattice temperature. When the Te / Tl ratio exceeds 3, it is determined to be a non-equilibrium "cold-processed" state.

[0067] Additional explanation: The Te / Tl ratio reflects the degree of thermodynamic non-equilibrium between free electrons in the plasma and the material lattice. This ratio is directly related to the electron temperature Te in the input spectral feature vector and is compared with the lattice temperature Tl in the temperature feature vector. When the Te / Tl ratio exceeds 3, it indicates that the energy of the electronic system is much higher than that of the lattice system. At this time, the material mainly transforms directly from solid to gas through electron excitation and plasmaification, skipping the liquid melting stage, which is determined to be a non-equilibrium "cold working" state, which is the ideal state for ultrashort pulse laser processing. Under extremely high power density conditions, the Te / Tl ratio may indeed far exceed 3, indicating stronger non-equilibrium characteristics, but the ratio will not be less than 1 (because the electron temperature cannot be lower than the lattice temperature). When the ratio is close to 1, it indicates that the system is approaching thermal equilibrium. At this time, traditional melting processing is likely to occur, deviating from the technical goal of "cold working". Therefore, the system provides accurate feedback signals for the dynamic control of laser parameters by real-time monitoring of the Te / Tl ratio change trend and combining the phase transition level determination results.

[0068] This method not only retains the rapid judgment mechanism based on dynamic thresholds, but more importantly, it introduces a forward-looking prediction algorithm based on multi-parameter correlation analysis. By monitoring the changing trends and interrelationships of various physical quantities, it can identify deviations in the processing state before the phase transition critical point is reached, thus achieving proactive control. Compared with traditional single-sensor methods, which are susceptible to noise interference, have judgment lag, and high false positive rates, the multi-dimensional fusion identification method of this invention significantly improves the accuracy, real-time performance, and robustness of phase transition state identification through information redundancy and cross-validation, providing a reliable technical foundation for achieving precise "cold processing" process control.

[0069] The relevant technical references are as follows: Spectral feature extraction algorithm: Pedarnig, JD et al. "Characterization of laser-induced plasma by optical emission spectroscopy" Applied Surface Science, Vol. 197-198, 2002, pp. 273-278. Page 275: Detailed description of the calculation method for the intensity ratio of Fe I at 538.3 nm / 526.9 nm; Page 276: Plasma temperature extraction algorithm; Image feature extraction: Haralick, RM "Textural features for image classification", IEEE Transactions on Systems, Man, and Cybernetics, Vol. 3, 1973, pp. 610-621. Pages 614-616: Detailed description of the Gray-Level Co-occurrence Matrix (GLCM) algorithm; Page 617: Definition of texture feature parameters; Classification Algorithm: Support Vector Machine (SVM) Classification, Liu, P. et al. "Deep learning for laser-induced breakdown spectroscopy: Review and prospects" TrAC Trends in Analytical Chemistry, Vol. 143, 2021, 116385, page 4: Application of SVM in LIBS Spectral Classification; Random Forest Algorithm: Wang, C. et al. "Machine learning for laser processing parameter optimization" Journal of Manufacturing Processes, Vol. 75, 2022, pp.1203-1215. Page 1208: Application of Random Forest in Laser Parameter Prediction; Page 1210: Feature Importance Ranking Results; Deep learning solution: Convolutional Neural Network (CNN). Zhang, T. et al. "Real-time defect detection in laser welding with deep neural networks" Journal of Manufacturing Systems, Vol. 62, 2022, pp. 800-814, pp. 804-806: CNN network architecture design; pp. 808: training strategy and hyperparameter settings; Long Short-Term Memory Network (LSTM): Chen, L. et al. "Deep learning approaches for laser welding quality prediction" Welding Journal, Vol. 100, 2021, pp. 285s-295s, pp. 288s-290s: LSTM temporal modeling method, pp. 292s: phase transition state temporal prediction algorithm; Phase transition level definition: Harilal, SS "Spectroscopic characterization of laser-induced plasmas for industrial applications" Applied Spectroscopy Reviews, Vol. 53, 2018, pp. 1-47, pp. 15-18: Spectroscopic criteria for phase transition states, pp. 22-25: Plasma classification standards, pp. 28-30: Threshold setting for industrial applications; Real-time prediction systems: Shalev-Shwartz, S. "Online Learning and Online ConvexOptimization" Foundations and Trends in Machine Learning, Vol. 4, 2012, pp. 107-194, pp. 125-140: Online learning algorithms. pp. 165-180: Concept drift detection.

[0070] The basic principle of this invention is:

[0071] This invention uses multi-sensor fusion to collect plasma spectrum, thermal radiation and surface morphology signals of the processing area in real time, jointly identify the phase transition state of the material, and dynamically adjust the laser parameters based on intelligent algorithms, so that the processing process is always maintained in a non-thermal-dominated "cold processing" mode.

[0072] The system includes an ultrashort pulse laser, a beam transmission and focusing module, a real-time phase change monitoring module, a signal processing and feedback control module, and a motion control platform. (1) The real-time phase change monitoring module integrates a spectral sensor, a high-speed infrared thermal imager, and a confocal interferometer to realize the synchronous acquisition and fusion analysis of multi-physical field signals in the processing area; (2) The signal processing and feedback control module has a built-in machine learning algorithm that can identify the phase change state based on real-time monitoring data and compare it with the preset "cold processing" threshold to generate laser parameter adjustment instructions; (3) The system forms a "sensing-analysis-control" closed loop and can adaptively adjust parameters such as laser pulse energy, repetition frequency, and scanning speed in real time according to the phase change state; (4) The system has the ability to learn process parameters and can optimize the control strategy based on historical processing data to improve the adaptability and robustness to dynamic working conditions.

[0073] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principle of the present invention, such as the design of the beam splitting optical path, other integrated installation methods of the spectrometer, infrared thermal imager, interferometer and lens assembly, different signal fusion processing methods, etc. These improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An adaptive control system for ultrashort pulse laser "cold processing", characterized in that, The adaptive control system includes a signal processing and feedback control module, and an ultrashort pulse laser, a real-time phase transition monitoring module, and a motion control platform, which are electrically connected to the signal processing and feedback control module respectively. The motion control platform includes a worktable and a controller, and the observation probes of the ultrashort pulse laser, the beam transmission and focusing module, and the real-time phase transition monitoring module are physically fixed. A beam transmission and focusing module is provided between the laser output of the ultrashort pulse laser and the workpiece being processed on the worktable. The beam transmission and focusing module includes at least a focusing lens assembly (1). The real-time phase change monitoring module includes a spectrometer (4), an infrared thermal imager (5), and an interferometer (6). The spectrometer, infrared thermal imager, and interferometer synchronously receive the spectral, temperature, and morphological signals of the processed area of ​​the workpiece.

2. The adaptive control system for ultrashort pulse laser "cold processing" according to claim 1, characterized in that, The interferometer reference beam source includes an independent path reference beam split from a laser source by a beam splitter, or one of the beams used as a reference source by wavelength division multiplexing (WDM).

3. The adaptive control system for ultrashort pulse laser "cold processing" according to claim 1, characterized in that, The spectrometer (4), infrared thermal imager (5), and interferometer (6) are symmetrically mounted around the focusing lens assembly (1).

4. The adaptive control system for ultrashort pulse laser "cold processing" according to claim 1, characterized in that, The beam transmission and focusing module includes at least a beam splitting assembly (2), and the beam splitting assembly includes at least a first semi-transparent mirror (21) disposed in the optical path between the laser output and the focusing lens assembly (1).

5. The adaptive control system for ultrashort pulse laser "cold processing" according to claim 4, characterized in that, The beam splitter assembly further includes a second semi-transparent semi-reflective mirror (22) and a third semi-transparent semi-reflective mirror (23) sequentially arranged in the reflection optical path of the first semi-transparent semi-reflective mirror (21), wherein at least one reflector is arranged in the reflection optical path of the second semi-transparent semi-reflective mirror (22) and the third semi-transparent semi-reflective mirror (23).

6. The adaptive control system for ultrashort pulse laser "cold processing" according to claim 4, characterized in that, The beam splitting assembly also includes a fourth semi-transparent mirror (24) disposed in the optical path between the first semi-transparent mirror (21) and the focusing lens assembly (1), and a fifth semi-transparent mirror (25) disposed in the reflected optical path of the fourth semi-transparent mirror (24).

7. The adaptive control system for ultrashort pulse laser "cold processing" according to claim 4, characterized in that, A third mirror (33) and a fourth mirror (34) and a sixth mirror (26) are designed on the original laser reflection path of the first semi-transparent and semi-reflective mirror (21).

8. An adaptive control method for "cold processing" with ultrashort pulse lasers, characterized in that, The adaptive control method, based on joint monitoring data from a spectrometer, an infrared thermal imager, and an interferometer, obtains information on the plasma spectrum, temperature field, and morphological changes of the processed area during laser cold processing, and captures the phase transition state of the material under laser irradiation in real time. The method includes the following steps: S1. System initialization: Set basic laser parameters and phase transition monitoring thresholds according to the workpiece material; if the initialization data is adjusted based on feedback data received in step S4 during processing, proceed directly to step S3 after the adjustment is completed. S2. Calibration and Positioning: Start the motion control platform, position the laser focus to the workpiece to be processed area, and calibrate the sensor alignment of the monitoring module; S3. Real-time processing and monitoring: The ultrashort pulse laser is turned on for processing, while the phase transition monitoring module continuously collects joint monitoring data from the spectrometer, infrared thermal imager and interferometer of the processing area and transmits it to the signal processing unit. S4. Signal Processing and Feedback: The signal processing unit analyzes the data in real time and identifies the phase transition state by combining plasma spectrum, temperature field, and morphological changes. If an anomaly is detected, the feedback control loop is triggered to adjust the laser parameters and scanning path. If the initialization parameters need to be adjusted, the process returns to step S1; otherwise, it proceeds to step S5. S5. Adaptive control cycle: Adjust the relative processing position of the worktable and / or laser based on the feedback signal, i.e., adjust the processing path, return to step S3, and continue to execute the monitoring-analysis-control cycle until the processing task is completed.

9. The adaptive control method for ultrashort pulse laser "cold processing" according to claim 8, characterized in that, The signal processing employs machine learning algorithms for feature classification and phase transition state prediction, outputting the phase transition level, including the following steps: S41. Input of real-time phase change monitoring data; S42. Feature extraction: Machine learning algorithms are used to extract spectral feature vectors, temperature feature vectors, and morphological feature vectors respectively. S43. Phase transition state identification: Based on the interrelationship between the spectral characteristics, temperature characteristics, and morphological characteristics of the material during laser cold processing, establish a relationship model between each feature vector and the phase transition state. The relationship model includes spectral feature function, temperature feature function, and morphological feature function. S44. Parameter comparison generates control instructions, the control instructions including laser parameter adjustment instructions and / or motion path update instructions.

10. The adaptive control method for ultrashort pulse laser "cold processing" according to claim 9, characterized in that, The signal processing method for jointly identifying phase transition states through plasma spectroscopy, temperature field, and morphological changes includes: a smart phase transition state identification method that integrates multi-dimensional features; The intelligent phase transition state recognition method based on multi-dimensional feature fusion establishes a phase transition state evaluation function: S=α×f(X_spectrum)+β×g(X_thermal)+γ×h(X_morphology); Among them, the weighting coefficients α, β, and γ are adaptively adjusted according to the material type and processing conditions; f(X_spectrum), g(X_thermal), and h(X_morphology) are the spectral feature function, temperature feature function, and morphological feature function, respectively, and X_spectrum, X_thermal, and X_morphology are the spectral feature vector, temperature feature vector, and morphological feature vector, respectively.

11. The adaptive control method for "cold processing" of ultrashort pulse lasers according to claim 9, characterized in that, The spectral feature vector is defined as X_spectrum=[Te,ne,R_AI,W_line], where Te is the plasma electron temperature calculated by fitting the atomic spectral line intensity ratio through the Boltzmann distribution, ne is the electron density determined based on the Stark broadening effect, R_AI is the ratio of atomic spectral line intensity to ion spectral line intensity, and W_line is the Lorentz broadening of the feature spectral line. The temperature feature vector is defined as X_thermal=[T_max,∇T,dT / dt,A_HAZ], which includes the peak temperature T_max of the processing area measured by the infrared thermal imager, the radial temperature gradient ∇T, the temperature rise rate dT / dt, and the heat-affected zone expansion area A_HAZ. The morphology feature vector is defined as X_morphology=[ΔZ,C_edge,η_removal], which includes the surface height change ΔZ measured by the confocal interferometer, the processing edge sharpness coefficient C_edge, and the material removal efficiency η_removal.

12. The adaptive control method for "cold processing" of ultrashort pulse lasers according to claim 9, characterized in that, The spectral characteristic function: f(X_spectrum)=w1×sigmoid(Te / Te_critical)+w2×log(ne / ne_threshold)+w3×(1-R_AI / R_ideal)+w4×(W_line / W_reference); Where Te_critical is the critical electron temperature for material sublimation, and ne_threshold is the density threshold for stable plasma formation. The temperature characteristic function: g(X_thermal)=w5×tanh(T_max / T_melting)+w6×(∇T / ∇T_ideal)^0.5+w7×|dT / dt| / |(dT / dt)_cold|+w8×(A_HAZ / A_laser_spot); The morphological feature function: h(X_morphology)=w9×(ΔZ / Z_target)+w10×C_edge+w11×η_removal; The weighting coefficients w1-w11 were obtained through machine learning training on experimental data of "cold working" of different materials.

Images