Method and device for adaptive adjustment of parameters of picosecond infrared laser glass filamentation cutting

Abstract

The application discloses a picosecond infrared laser glass filamentation cutting parameter adaptive adjustment method and device, comprising: obtaining plasma radiation spectrum signal and acoustic emission time-frequency signal and carrying out baseline correction and adaptive noise reduction; based on the corrected signal data, analyzing the nonlinear absorption coefficient distribution and reconstructing the thermal influence zone temperature field evolution data; inputting the corrected signal data into a physical information neural agent model to obtain multi-physical field coupling prediction parameters; based on the multi-physical field coupling prediction parameters, the nonlinear absorption coefficient distribution and the thermal influence zone temperature field evolution data, obtaining the filamentation uniformity index and the crack deflection amount; modulating the laser focal spot energy distribution with different preset cone angles to obtain the focal spot energy topology parameters; and executing model predictive control optimization according to the filamentation uniformity index, the crack deflection amount and the focal spot energy topology parameters to generate laser cutting parameter adjustment instructions. The application effectively improves the precision and reliability of the picosecond infrared laser glass filamentation cutting parameter adaptive adjustment.

Description

Technical Field

[0001] This application relates to the field of laser precision machining technology, and in particular to a method and apparatus for adaptive adjustment of parameters for picosecond infrared laser glass filament cutting. Background Technology

[0002] Picosecond infrared laser glass filament cutting technology utilizes ultrashort pulse lasers to induce nonlinear absorption within transparent materials, forming a through-type modified filament. High-precision separation is then achieved through mechanical cleaving, offering significant advantages such as no taper, no edge chipping, and high surface quality. It has become the mainstream technology for processing glass components in industries such as consumer electronics, photovoltaics, and display panels. However, due to the dynamic influence of factors such as the heterogeneity of glass materials, thickness fluctuations, coating differences, and thermal accumulation effects, maintaining consistent filament morphology with fixed cutting parameters is difficult. This can easily lead to process defects such as discontinuous filament formation, crack deflection, and excessively large heat-affected zones, resulting in decreased cutting yield and material scrap. Existing parameter adjustment methods largely rely on operator experience or offline process testing, lacking real-time perception and closed-loop control capabilities for the transient physical processes of laser-material interaction. This makes it unsuitable for the flexible manufacturing needs of multi-variety, small-batch production, hindering the large-scale application of picosecond laser glass filament cutting technology in high-end manufacturing. Summary of the Invention

[0003] The main objective of this application is to provide a method and apparatus for adaptive adjustment of parameters for picosecond infrared laser glass filament cutting, which aims to solve the technical problem that the cutting parameters cannot be adaptively adjusted and the filament quality is unstable in the existing picosecond laser glass filament cutting process due to the lack of real-time perception and predictive control of the transient process coupled with multiple physical fields.

[0004] To achieve the above objectives, this application proposes an adaptive adjustment method for picosecond infrared laser glass filament cutting parameters. The adaptive adjustment method for picosecond infrared laser glass filament cutting parameters includes: The plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser interaction region are acquired and baseline correction and adaptive noise reduction are performed to obtain the corrected signal data; Based on the corrected signal data, the nonlinear absorption coefficient distribution is analyzed using a filamentation energy deposition model, and the temperature field evolution data of the heat-affected zone is reconstructed. The corrected signal data is input into a pre-trained physical information neural agent model to obtain multi-physics field coupling prediction parameters, which include the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate. Based on the multiphysics field coupling prediction parameters, the nonlinear absorption coefficient distribution, and the temperature field evolution data of the heat-affected zone, the filament morphology is predicted and crack propagation is determined, resulting in the filament uniformity index and crack deflection amount. The laser focal spot energy topology parameters are obtained by modulating the laser focal spot energy distribution with different preset cone angles using a Bessel beam shaping unit. Based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters, a model prediction control optimization is performed to generate laser cutting parameter adjustment instructions.

[0005] Optionally, the step of acquiring the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser interaction region and performing baseline correction and adaptive noise reduction to obtain corrected signal data includes: The plasma radiation spectrum signal of the picosecond infrared laser-actuated region is acquired by the plasma spectrum acquisition unit, and the intensity value of the plasma characteristic spectral line is output after dark current subtraction. Acoustic emission time-frequency signals in the picosecond infrared laser action area are acquired by an acoustic emission sensor array, and the acoustic emission time-frequency signals are divided into multiple time windows along the cutting direction. The root mean square amplitude and spectral centroid of each time window are determined, and an acoustic emission feature vector is generated. Based on the plasma characteristic spectral line intensity values ​​and the acoustic emission characteristic vector, an adaptive gain scheduling algorithm is used to generate spectral integration time parameters, acoustic emission sampling rate parameters, and signal amplification parameters. Based on the spectral integration time parameter, the acoustic emission sampling rate parameter, and the signal amplification factor parameter, baseline correction is performed on the plasma radiation spectral signal and the acoustic emission time-frequency signal to obtain baseline-de-baseline signal data. The baseline-degraded signal data is sequentially processed by empirical mode decomposition, variational mode decomposition, and wavelet packet thresholding to obtain the corrected signal data.

[0006] Optionally, the step of analyzing the nonlinear absorption coefficient distribution and reconstructing the temperature field evolution data of the heat-affected zone based on the corrected signal data using a filamentation energy deposition model includes: Obtain the pre-calibrated time-domain waveform parameters of the laser pulse, the nonlinear optical coefficient of the glass material, and the thermal property parameters; Based on the laser pulse time-domain waveform parameters and the nonlinear optical coefficients of the glass material, the inverse problem of the plasma radiation spectrum in the correction signal data is solved to obtain the multiphoton absorption coefficient and the avalanche ionization coefficient. Based on the multiphoton absorption coefficient and the avalanche ionization coefficient, the deposition distribution of laser energy along the propagation path is calculated using a filamentation energy deposition model to obtain the nonlinear absorption coefficient distribution. A transient heat conduction equation is established based on the aforementioned thermal property parameters and the aforementioned nonlinear absorption coefficient distribution. The spatiotemporal distribution of the temperature field in the heat-affected zone is solved using the finite difference method to obtain the temperature field evolution data of the heat-affected zone. The temperature gradient distribution and cooling rate distribution are calculated based on the temperature field evolution data of the heat-affected zone, and a thermodynamic characteristic map is generated.

[0007] Optionally, inputting the corrected signal data into a pre-trained physical information neural agent model to obtain multi-physics coupling prediction parameters includes: A proxy model based on physical information neural network is constructed. The physical information neural proxy model includes a spatiotemporal coordinate embedding layer, a multi-scale feature encoder, and a multi-task decoder for filamentation depth prediction, microcrack density prediction, and stress relaxation prediction, respectively. Plasma radiation spectrum signals and acoustic emission time-frequency signals were collected under different glass grades, thicknesses and cutting speeds. The filamentation depth was marked by microscopic profiles, the microcrack density was marked by scanning electron microscopy, and the residual stress was marked by Raman spectroscopy. A multiphysics dataset was constructed, and the multiphysics dataset was divided into a training set and a validation set according to a preset ratio. The physical information neural agent model is trained using the training set. The nonlinear Schrödinger equation and the heat conduction equation are embedded as physical constraints into the loss function to optimize the network parameters until the root mean square error of the prediction on the validation set reaches a preset accuracy threshold, thus obtaining the pre-trained physical information neural agent model. The plasma radiation spectrum feature vector and acoustic emission time-frequency feature vector in the corrected signal data are input into the pre-trained physical information neural agent model, and the multi-task decoder outputs the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate as multi-physics coupling prediction parameters.

[0008] Optionally, the step of predicting filament morphology and determining crack propagation based on the multiphysics coupling prediction parameters, the nonlinear absorption coefficient distribution, and the temperature field evolution data of the heat-affected zone, to obtain the filament uniformity index and crack deflection, includes: The melting front trajectory coordinates are extracted from the temperature field evolution data of the heat-affected zone, and the energy deposition centerline is determined based on the nonlinear absorption coefficient distribution. Based on the melting front trajectory coordinates and the energy deposition centerline, calculate the filament cross-sectional geometric parameters and generate a filament morphology feature vector; The filamentation morphology feature vector is compared with the preset filamentation morphology template feature vector to determine the morphology deviation. If the morphology deviation exceeds the preset deviation threshold in multiple consecutive pulse cycles, the morphology deviation is used as the filamentation uniformity index. The stress intensity factor at the crack tip is determined based on the microcrack density spectrum in the multiphysics coupling prediction parameters. The crack propagation path is simulated based on the stress intensity factor and the cooling rate distribution in the temperature field evolution data of the heat-affected zone to obtain the reference deflection. The reference deflection amount is corrected by depth weighting based on the filamentation depth probability distribution to obtain the crack deflection amount.

[0009] Optionally, the step of obtaining the focal spot energy topology parameters by modulating the laser focal spot energy distribution with different preset cone angles using a Bessel beam shaping unit includes: The relative displacement of the cone lens group in the Bessel beam shaping unit is adjusted by the first preset cone angle to generate axial depth of field parameters without diffraction focal spots. The rotation angle of the axis-cone mirror in the Bessel beam shaping unit is adjusted by the second preset cone angle to generate the radial energy ring distribution parameters of the focal spot. The axial depth of field parameter is compared with a preset depth of field threshold. When the axial depth of field parameter meets the preset depth of field threshold, the depth of field qualification mark is set to a valid state. The annular energy uniformity of the radial energy distribution parameters of the focal spot is detected. When the annular energy non-uniformity is lower than the preset non-uniformity threshold, the annular uniformity indicator is set to an effective state. A logical AND operation is performed on the depth-of-field qualification mark and the ring uniformity mark. When the operation result is true, the focal spot energy topology parameter is calculated based on the axial depth-of-field parameter and the focal spot radial energy ring distribution parameter.

[0010] Optionally, the step of performing model prediction control optimization based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters to generate laser cutting parameter adjustment instructions includes: Process capability mapping is performed on the filament uniformity index to generate filament quality grade parameters; The crack deflection amount is mapped to a deflection threshold to generate crack risk level parameters; The annular energy concentration parameter and axial energy gradient parameter in the focal spot energy topology parameters are jointly discriminated to generate energy distribution level parameters; The filament quality grade parameter, the crack risk grade parameter, and the energy distribution grade parameter are weighted and summed according to preset weighting coefficients to obtain a comprehensive process score. Based on the comprehensive process score, a model predictive control optimization problem is constructed. Laser power, pulse energy, repetition frequency, scanning speed and auxiliary gas pressure are used as control variables, and filament uniformity index and crack deflection are used as controlled outputs. The optimal control sequence is solved to generate laser cutting parameter adjustment instructions.

[0011] Furthermore, to achieve the above objectives, this application also proposes a picosecond infrared laser glass filament cutting parameter adaptive adjustment device, which includes: The acquisition module is used to acquire the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser action area and perform baseline correction and adaptive noise reduction to obtain the corrected signal data; The analysis module is used to analyze the nonlinear absorption coefficient distribution based on the correction signal data using a filamentation energy deposition model, and to reconstruct the temperature field evolution data of the heat-affected zone. The prediction module is used to input the correction signal data into a pre-trained physical information neural agent model to obtain multi-physics field coupling prediction parameters, which include the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate. The determination module is used to predict the filament morphology and determine the crack propagation based on the multi-physics field coupling prediction parameters, the nonlinear absorption coefficient distribution and the temperature field evolution data of the heat-affected zone, and to obtain the filament uniformity index and crack deflection amount. The modulation module is used to modulate the laser focal spot energy distribution with different preset cone angles through the Bessel beam shaping unit to obtain the focal spot energy topology parameters; The optimization module is used to perform model prediction control optimization based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters, and generate laser cutting parameter adjustment instructions.

[0012] In addition, to achieve the above objectives, this application also proposes a picosecond infrared laser glass filament cutting parameter adaptive adjustment device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the picosecond infrared laser glass filament cutting parameter adaptive adjustment method as described above.

[0013] In addition, to achieve the above objectives, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the adaptive adjustment method for picosecond infrared laser glass filament cutting parameters as described above.

[0014] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the picosecond infrared laser glass filament cutting parameter adaptive adjustment method described above.

[0015] One or more technical solutions proposed in this application, by acquiring the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser interaction region and performing baseline correction and adaptive noise reduction, can effectively suppress the interference of plasma radiation fluctuations and acoustic emission environmental noise on signal quality. By analyzing the nonlinear absorption coefficient distribution and reconstructing the temperature field evolution data of the heat-affected zone through a filamentation energy deposition model, quantitative characterization of the transient physical process of laser-glass interaction is achieved. Combined with the multi-physics coupling prediction parameters output by the physical information neural agent model, simultaneous prediction of filamentation depth, microcrack density, and stress relaxation rate is realized, based on melt... By integrating the prediction of filamentation morphology based on the leading edge trajectory and energy deposition centerline, as well as the simulation of crack propagation path based on stress intensity factor, the precise determination of filamentation uniformity and crack deflection was achieved. Dynamic control of focal spot energy topology was realized through the cone angle modulation of the Bessel beam shaping unit. Finally, the filamentation uniformity index, crack deflection, and focal spot energy topology parameters were incorporated into the model prediction control optimization framework to generate laser cutting parameter adjustment instructions. This enabled multi-physics field coupled sensing, prediction, and closed-loop optimization of the picosecond infrared laser glass filamentation cutting process, effectively improving the accuracy and reliability of adaptive adjustment of glass filamentation cutting parameters under complex working conditions. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an embodiment of the method for adaptive adjustment of parameters for picosecond infrared laser glass filament cutting in this application. Figure 2 This is a schematic diagram of the module structure of the picosecond infrared laser glass filament cutting parameter adaptive adjustment device according to an embodiment of this application; Figure 3 This is a schematic diagram of the hardware operating environment of the picosecond infrared laser glass filament cutting parameter adaptive adjustment device in the embodiments of this application.

[0019] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0021] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0022] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a picosecond infrared laser glass filament cutting parameter adaptive adjustment device. The following description uses a picosecond infrared laser glass filament cutting parameter adaptive adjustment device as an example to illustrate this embodiment and the subsequent embodiments.

[0023] Based on this, embodiments of this application provide an adaptive adjustment method for picosecond infrared laser glass filament cutting parameters, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the picosecond infrared laser glass filament cutting parameter adaptive adjustment method of this application.

[0024] In this embodiment, the method for adaptive adjustment of picosecond infrared laser glass filament cutting parameters includes steps S10~S60: Step S10: Acquire the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser action area, and perform baseline correction and adaptive noise reduction to obtain the corrected signal data.

[0025] It should be noted that the plasma radiation spectral signal refers to the digital sequence obtained by analog-to-digital conversion of the analog electrical signal characterizing the intensity distribution of plasma radiation induced during the interaction between picosecond infrared laser and glass, acquired by a plasma spectral acquisition unit deployed beside the laser processing head. The plasma spectral acquisition unit is a photoelectric conversion device based on a fiber optic spectrometer and a photomultiplier tube array, with a spectral response range covering the ultraviolet to near-infrared band and a time resolution of less than 1 ns, capable of capturing the transient spectral characteristics of plasma radiation in real time during laser pulse action. The acoustic emission time-frequency signal refers to the digital sequence obtained by analog-to-digital conversion of the elastic wave signal characterizing material fracture and stress release during laser filamentation, acquired by an acoustic emission sensor array attached to the edge of the glass workpiece. The acoustic emission sensor array is an acoustic sensor device based on the piezoelectric ceramic resonance principle, with a frequency response range covering 20 kHz to 2 MHz, capable of capturing the acoustic characteristics of microcrack initiation and propagation in real time during laser scanning. Baseline correction is a preprocessing operation to eliminate signal baseline drift and dark current noise. Adaptive noise reduction is a signal optimization method that dynamically adjusts filtering parameters according to the signal-to-noise ratio. The corrected signal data is a plasma spectral feature vector and acoustic emission time-frequency feature vector with a stable baseline and high signal-to-noise ratio, output after baseline correction and adaptive noise reduction.

[0026] In actual execution, a plasma spectral acquisition module deployed beside the laser processing head synchronously acquires the plasma radiation spectral signal of the picosecond infrared laser action area with nanosecond-level time resolution. This module uses a 2048-pixel linear CCD spectral sensor to capture the transient spectral distribution of plasma radiation. Simultaneously, an acoustic emission sensor array synchronously acquires the acoustic emission time-frequency signal of the glass workpiece edge at a sampling rate of 10MHz. The acoustic emission time-frequency signal is divided into 128 sampling point time windows along the cutting direction, and the root mean square amplitude and spectral centroid of each time window are calculated to generate an acoustic emission feature vector. Based on the intensity deviation between the plasma characteristic spectral line intensity value and the preset intensity threshold, and the feature deviation between the acoustic emission feature vector and the target feature vector, an adaptive gain scheduling algorithm generates spectral integration time parameters, acoustic emission sampling rate parameters, and signal amplification factor parameters. The spectral integration time is dynamically adjusted within the range of 1ms to 100ms, the acoustic emission sampling rate is dynamically adjusted within the range of 1MHz to 10MHz, and the signal amplification factor is dynamically adjusted within the range of 20dB to 60dB. Based on the above parameters, baseline correction is performed on the plasma radiation spectrum signal and the acoustic emission time-frequency signal to obtain baseline-reduced signal data. Then, empirical mode decomposition is performed on the baseline-reduced signal data in sequence to separate the intrinsic mode functions of the signal, variational mode decomposition is performed to extract the modal components of specific frequency bands, and wavelet packet thresholding is performed to suppress high-frequency random noise, finally outputting the corrected signal data.

[0027] In one feasible implementation, step S10 may include: acquiring the plasma radiation spectrum signal of the picosecond infrared laser-affected region through a plasma spectral acquisition unit, and outputting the plasma characteristic spectral line intensity value after dark current subtraction processing; acquiring the acoustic emission time-frequency signal of the picosecond infrared laser-affected region through an acoustic emission sensor array, dividing the acoustic emission time-frequency signal into multiple time windows along the cutting direction, determining the root mean square amplitude and spectral centroid of each time window, and generating an acoustic emission feature vector; generating spectral integration time parameters, acoustic emission sampling rate parameters, and signal amplification factor parameters through an adaptive gain scheduling algorithm based on the plasma characteristic spectral line intensity value and the acoustic emission feature vector; performing baseline correction on the plasma radiation spectrum signal and the acoustic emission time-frequency signal based on the spectral integration time parameters, the acoustic emission sampling rate parameters, and the signal amplification factor parameters to obtain baseline-degraded signal data; and sequentially performing empirical mode decomposition, variational mode decomposition, and wavelet packet threshold denoising processing on the baseline-degraded signal data to obtain corrected signal data.

[0028] It should be noted that dark current subtraction is a digital signal processing operation used to eliminate dark current noise generated by spectral sensors under no-light conditions. The intensity value of plasma characteristic spectral lines is a scalar parameter output after dark current subtraction, which can stably reflect the current plasma radiation intensity and serves as a feedforward input for signal gain scheduling.

[0029] Understandably, a time window refers to the set of signal segments selected along the cutting direction in the acoustic emission time-frequency signal for evaluating local acoustic characteristics. The root mean square (RMS) amplitude is the root mean square value of the acoustic emission signal amplitude within the time window, reflecting the energy level of the acoustic event. The spectral centroid is the frequency centroid of the acoustic emission signal power spectrum within the time window, reflecting the dominant frequency characteristic of the acoustic event. The acoustic emission eigenvector is a two-dimensional eigenvector composed of the RMS amplitude and the spectral centroid, serving as the feedback input for signal gain scheduling.

[0030] Understandably, the adaptive gain scheduling algorithm is an adaptive control method that calculates signal acquisition parameters based on the deviations of plasma characteristic spectral line intensity values ​​and acoustic emission characteristic vectors relative to their respective target values ​​using a gain scheduling control law. The spectral integration time parameter refers to the duration for which the spectral sensor collects the optical signal. The acoustic emission sampling rate parameter refers to the rate at which the acoustic emission signal is digitized. The signal amplification factor parameter refers to the factor by which the analog signal output by the sensor is amplified. Baseline correction refers to the process of configuring the registers of the plasma spectral acquisition module and the acoustic emission sensor array based on the generated spectral integration time parameter, acoustic emission sampling rate parameter, and signal amplification factor parameter, thereby changing the system's signal response characteristics. Baseline-decorrected signal data refers to the plasma spectral signal and acoustic emission time-frequency signal output by the multi-sensor acquisition unit after baseline correction, characterized by a stable baseline and signal intensity approaching the target range.

[0031] Understandably, Empirical Mode Decomposition (EMD) is a data-driven decomposition method that adaptively decomposes a non-stationary signal into the sum of a finite number of eigenmode functions. Variational Mode Decomposition (VMD) is a signal decomposition method that decomposes a signal into a predetermined number of narrowband modal components using a variational framework. Wavelet packet thresholding is a method that uses wavelet packet transform to perform multi-resolution decomposition of the signal and suppresses noise wavelet coefficients through thresholding. The corrected signal data is the signal feature data output after the above three-step concatenated processing, which combines trend separation, frequency band extraction, and noise suppression.

[0032] In the specific implementation, the plasma spectral acquisition unit is installed beside the laser processing head, with its fiber optic probe aligned with the laser action area. The sensor operates in trigger mode, acquiring raw plasma radiation spectrum data with nanosecond-level time resolution during the laser pulse emission period. Since the plasma radiation intensity varies drastically with laser pulse energy and glass material, and the sensor exhibits dark current noise, dark current subtraction processing is required on the raw data. After processing, the output plasma characteristic spectral line intensity values ​​effectively eliminate dark current drift while preserving the true intensity variation trend of plasma radiation. Acoustic emission time-frequency signals are acquired through an edge processing unit within the multi-sensor acquisition unit. The acoustic emission time-frequency signals are divided into M non-overlapping time windows along the cutting direction, each time window having a length of N sampling points, for example, 1024 sampling points. To prioritize the acoustic monitoring quality of the filamentation area, the coverage of the time windows is limited to the laser pulse action period and its aftereffect region. For each time window, the root mean square amplitude and spectral centroid of the acoustic emission signal are determined, and these two parameters are integrated into a two-dimensional acoustic emission feature vector, which is then sent to the adaptive gain scheduling module. The intensity deviation is obtained by subtracting the preset plasma characteristic spectral line intensity threshold from the currently acquired intensity value. Similarly, the characteristic deviation is obtained by subtracting the current acoustic emission characteristic vector from the preset target acoustic emission characteristic vector. These two deviations are then substituted into a pre-designed proportional-integral gain (PIG) ​​scheduling control law to calculate the spectral integration time parameter, acoustic emission sampling rate parameter, and signal amplification factor parameter that need adjustment. The configurations of the spectral acquisition unit and acoustic emission acquisition unit are updated according to the calculated parameters, and the signal is reacquired to complete baseline correction, resulting in baseline-stable, intensity-within-a-reasonable debasement signal data. Next, empirical mode decomposition (EMD) is performed on the debasement signal data to separate multiple intrinsic mode functions at different scales, removing mode components corresponding to low-frequency baseline drift. Variational mode decomposition (VMD) is then performed on the remaining components to extract narrowband mode components corresponding to plasma radiation and acoustic emission characteristics. Finally, wavelet packet transform is performed on the extracted mode components, and the wavelet packet coefficients are thresholded to remove coefficients corresponding to random noise. The reconstructed signal is the corrected signal data.

[0033] Step S20: Based on the corrected signal data, the nonlinear absorption coefficient distribution is analyzed using the filamentation energy deposition model, and the temperature field evolution data of the heat-affected zone is reconstructed.

[0034] It should be noted that the filamentary energy deposition model is a physical and mathematical model describing the deposition of laser energy into a transparent material through nonlinear optical effects such as multiphoton absorption and avalanche ionization as an ultrashort pulse laser propagates within the material. The nonlinear absorption coefficient distribution is a spatial distribution function characterizing the absorption of laser energy along the propagation path, encompassing both the multiphoton absorption coefficient and the avalanche ionization coefficient. The temperature field evolution data of the heat-affected zone is a four-dimensional data field characterizing the temporal and spatial variations in the material's internal temperature after laser energy deposition.

[0035] In practical implementation, laser pulse temporal waveform parameters, glass material nonlinear optical coefficients, and thermal property parameters, pre-calibrated offline using an ultrafast photodetector and thermal property analyzer, are acquired. The laser pulse temporal waveform parameters include pulse width, pulse shape, and peak power; the glass material nonlinear optical coefficients include multiphoton absorption cross-section, avalanche ionization collision time, and effective conduction band electron mass; and the thermal property parameters include thermal conductivity, specific heat capacity, and thermal diffusivity. Based on the laser pulse temporal waveform parameters and the glass material nonlinear optical coefficients, the inverse problem of the plasma radiation spectrum in the correction signal data is solved. Utilizing the proportional relationship between plasma radiation intensity and free electron density, the multiphoton absorption coefficient and avalanche ionization coefficient are inverted to obtain them. Based on the multiphoton absorption coefficient and avalanche ionization coefficient, the deposition distribution of laser energy along the propagation path is calculated using a filamentation energy deposition model, yielding the nonlinear absorption coefficient distribution. A transient heat conduction equation is established based on the thermal property parameters and the nonlinear absorption coefficient distribution. The spatiotemporal distribution of the temperature field in the heat-affected zone is solved using the finite difference method, obtaining the temperature field evolution data of the heat-affected zone. The temperature gradient distribution and cooling rate distribution are calculated based on the temperature field evolution data of the heat-affected zone, and a thermodynamic characteristic map is generated.

[0036] In one feasible implementation, step S20 may include: acquiring pre-calibrated laser pulse time-domain waveform parameters, glass material nonlinear optical coefficients, and thermophysical parameters; based on the laser pulse time-domain waveform parameters and glass material nonlinear optical coefficients, solving the inverse problem of the plasma radiation spectrum in the correction signal data to obtain the multiphoton absorption coefficient and avalanche ionization coefficient; based on the multiphoton absorption coefficient and avalanche ionization coefficient, determining the deposition distribution of laser energy along the propagation path using a filamentation energy deposition model to obtain the nonlinear absorption coefficient distribution; establishing a transient heat conduction equation based on the thermophysical parameters and the nonlinear absorption coefficient distribution, and using the finite difference method to solve the spatiotemporal distribution of the temperature field in the heat-affected zone to obtain the temperature field evolution data of the heat-affected zone; determining the temperature gradient distribution and cooling rate distribution based on the temperature field evolution data of the heat-affected zone, and generating a thermodynamic characteristic spectrum.

[0037] It should be noted that solving the inverse problem is a mathematical inversion method that infers physical model parameters from observational data. The multiphoton absorption coefficient is a nonlinear optical parameter characterizing the probability of a material simultaneously absorbing multiple photons and transitioning to the conduction band. The avalanche ionization coefficient is a parameter characterizing the multiplication effect of conduction band free electrons absorbing photons through inverse bremsstrahlung and gaining sufficient energy before colliding with and ionizing lattice atoms to produce secondary electrons. The nonlinear absorption coefficient distribution is a spatial distribution function characterizing the absorption of laser energy by the material, jointly determined by the multiphoton absorption coefficient and the avalanche ionization coefficient.

[0038] Understandably, the transient heat conduction equation is a partial differential equation describing the conduction of heat within a material under unsteady conditions. The finite difference method is a method that discretizes a continuous partial differential equation into a system of algebraic equations for numerical solution. The temperature field evolution data of the heat-affected zone is a four-dimensional data field characterizing the temperature variation with time and space, output after being solved using the finite difference method. The temperature gradient distribution is a vector field characterizing the drastic spatial variation of the temperature field. The cooling rate distribution is a scalar field characterizing the rate at which the temperature decreases over time. The thermodynamic characteristic map is a feature image composed of the temperature gradient distribution and the cooling rate distribution, used to characterize the thermodynamic state of the heat-affected zone.

[0039] In the specific implementation, the inverse problem of the plasma radiation spectrum in the correction signal data is solved. The plasma radiation intensity I(λ) and the free electron density n... e The relationship is:

[0040] Where g(λ,T) is the spectral shape function of plasma radiation, and T is the electron temperature.

[0041] By measuring the relative intensity distribution of the plasma radiation spectrum and combining the Saha equation and charge conservation conditions, the free electron density n can be obtained. e And electron temperature T.

[0042] Based on the evolution data of free electron density, a rate equation for nonlinear absorption is established: in, Let K be the absorption coefficient of the photon. Let η be the laser intensity, η be the avalanche ionization coefficient, and β be the electron recombination coefficient.

[0043] The multiphoton absorption coefficient is obtained by fitting the experimentally measured free electron density with the numerical solution of the rate equation. And the avalanche ionization coefficient η. Then, based on the multiphoton absorption coefficient and the avalanche ionization coefficient, the nonlinear absorption coefficient distribution α(r,z) is calculated:

[0044] Where r is the radial coordinate, z is the axial coordinate, I(r,z) is the laser intensity distribution, and n e (r,z) represents the free electron density distribution.

[0045] Based on the thermophysical parameters and the nonlinear absorption coefficient distribution, the transient heat conduction equation is established: Where ρ is the material density, For specific heat capacity, Let Q(r,z,t) be the thermal conductivity, and Q(r,z,t) be the heat source term. Let represent the instantaneous intensity distribution of the laser at time t.

[0046] The transient heat conduction equation is discretized by spatial grid and iterated by time step. The temperature field at each time step is calculated using an alternating direction implicit scheme to obtain the temperature values ​​of each spatial grid node at different times. The temperature data of all nodes can be integrated to generate the temperature field evolution data of the heat-affected zone. Further, spatial gradient calculation and time difference calculation are performed on the temperature field to obtain the temperature gradient distribution and cooling rate distribution. The two results are normalized into feature data under the same spatial grid to output the thermodynamic feature spectrum.

[0047] Step S30: Input the corrected signal data into the pre-trained physical information neural agent model to obtain multi-physics field coupling prediction parameters, which include the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate.

[0048] It should be noted that the physical information neural surrogate model is a surrogate model based on a physical information neural network architecture, trained and optimized using large-scale laser processing experimental data. It embeds the nonlinear Schrödinger equation and heat conduction equation as physical constraints into the network loss function, and includes a spatiotemporal coordinate embedding layer, a multi-scale feature encoder, and a multi-task decoder. The multiphysics coupling prediction parameters are a set of structured prediction information output by the model after forward inference of the correction signal. The filamentation depth probability distribution is a probability density function characterizing the depth of the laser-modified filament within the material. The microcrack density spectrum is a spectral function characterizing the distribution of the number of microcracks as a function of their size. The stress relaxation rate is a physical quantity characterizing the rate at which residual stress decays over time.

[0049] In practical implementation, a proxy model based on a physical information neural network is constructed. The spatiotemporal coordinate embedding layer uses Fourier feature mapping to map spatiotemporal coordinates to a high-dimensional space. The multi-scale feature encoder uses residual connections and attention mechanisms for multi-scale feature extraction. The multi-task decoder is configured with a filamentation depth prediction head, a microcrack density prediction head, and a stress relaxation prediction head. Plasma radiation spectral signals and acoustic emission time-frequency signals are collected under different glass grades, thicknesses, and cutting speeds. Microscopic profile annotations are used to annotate the filamentation depth, scanning electron microscopy statistical annotations are used to annotate the microcrack density, and Raman spectral annotations are used to annotate the residual stress, constructing a multiphysics dataset. This dataset is then divided into a training set and a validation set at a predetermined ratio of 8:2. The physical information neural proxy model is trained using the training set. The nonlinear Schrödinger equation and the heat conduction equation are embedded as physical constraint terms in the loss function to optimize the network parameters until the root mean square error of the predictions on the validation set reaches a predetermined accuracy threshold of 5%, resulting in a pre-trained physical information neural proxy model. The plasma radiation spectrum feature vector and acoustic emission time-frequency feature vector in the corrected signal data are input into the pre-trained physical information neural agent model. The multi-task decoder outputs the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate as multi-physics coupling prediction parameters.

[0050] Step S40: Based on the multi-physics field coupling prediction parameters, the nonlinear absorption coefficient distribution, and the temperature field evolution data of the heat-affected zone, perform filament morphology prediction and crack propagation determination to obtain the filament uniformity index and crack deflection amount.

[0051] It should be noted that filament morphology prediction is achieved by analyzing the melt front trajectory and energy deposition distribution to predict the three-dimensional geometry of the laser-modified filament. Crack propagation determination is achieved by analyzing the stress intensity factor and the cooling rate of the heat-affected zone to determine the propagation path and degree of deflection of microcracks. The filament uniformity index is a quantitative indicator characterizing the degree to which the filament morphology deviates from the ideal state. Crack deflection is a physical quantity characterizing the degree to which the crack propagation path deviates from the cutting direction.

[0052] In actual implementation, the melting front trajectory coordinates are extracted from the temperature field evolution data of the heat-affected zone, and the energy deposition centerline is determined based on the nonlinear absorption coefficient distribution. Geometric parameters of the filamentation cross-section, including filament diameter, filament roundness, and filament taper, are calculated based on the melting front trajectory coordinates and the energy deposition centerline, generating a filamentation morphology feature vector. This feature vector is compared with a preset filamentation morphology template feature vector to determine the morphology deviation. If the morphology deviation exceeds a preset deviation threshold of 0.1 μm for 100 consecutive pulse cycles, the morphology deviation is output as the filamentation uniformity index, indicating a filamentation anomaly. The crack tip stress intensity factor is determined based on the microcrack density spectrum in the multiphysics coupling prediction parameters. Crack propagation path simulation was performed based on the cooling rate distribution in the stress intensity factor and temperature field evolution data of the heat-affected zone. The extended finite element method was used to calculate the crack propagation path under the coupled action of thermal and mechanical stresses, and a reference deflection was obtained. The reference deflection was then corrected by depth weighting according to the filamentation depth probability distribution, resulting in the corrected crack deflection.

[0053] In one feasible implementation, step S40 may include: extracting the melting front trajectory coordinates from the temperature field evolution data of the heat-affected zone; determining the energy deposition centerline based on the nonlinear absorption coefficient distribution; determining the filamentation cross-sectional geometric parameters based on the melting front trajectory coordinates and the energy deposition centerline; generating a filamentation morphology feature vector; comparing the filamentation morphology feature vector with a preset filamentation morphology template feature vector to determine the morphology deviation; if the morphology deviation exceeds a preset deviation threshold in multiple consecutive pulse cycles, using the morphology deviation as the filamentation uniformity index; determining the crack tip stress intensity factor based on the microcrack density spectrum in the multiphysics coupling prediction parameters; performing crack propagation path simulation based on the stress intensity factor and the cooling rate distribution in the temperature field evolution data of the heat-affected zone to obtain a reference deflection; and performing depth weight correction on the reference deflection based on the filamentation depth probability distribution to obtain the crack deflection.

[0054] It should be noted that the melting front trajectory coordinates refer to the coordinates of the intersection line between the isothermal surface where the material's melting point is reached and the cut section in the temperature field evolution data of the heat-affected zone. The energy deposition centerline refers to the line connecting the maximum absorption coefficients in the nonlinear absorption coefficient distribution. The filamentation cross-sectional geometric parameters are scalar indicators that quantitatively describe the morphology of the filament cross-section, including filament diameter, filament roundness, and filament taper. The filamentation morphology feature vector is a multidimensional feature vector formed by combining the above geometric parameters in a specific order, used to characterize the overall morphological state of the filament.

[0055] The preset filamentation morphology template feature vector is a reference feature vector characterizing a specific standard filamentation morphology benchmark of the equipment, calculated by collecting multiple sets of data under standard cutting process conditions during the initial installation and operation or periodic calibration phases of the equipment. Morphology deviation is a quantitative indicator measuring the degree of difference between the current filamentation morphology feature vector and the preset filamentation morphology template feature vector, calculated using weighted Euclidean distance. Multiple consecutive pulse cycles refer to a fixed-length laser pulse sequence window with a causal relationship in the time dimension, such as 100 pulse cycles. The preset deviation threshold is a boundary value determined based on the statistical characteristics of normal filamentation fluctuations, used to distinguish between random fluctuations and true filamentation anomalies. The filamentation uniformity index is a quantitative output characterizing the severity of filamentation anomalies and triggering parameter adjustments when the morphology deviation exceeds the threshold for multiple consecutive pulse cycles.

[0056] The microcrack density spectrum is a spectral function output by the physical information neural agent model, characterizing the distribution of the number of microcracks as a function of size. The crack tip stress intensity factor is a fracture mechanics parameter characterizing the singularity of the stress field at the crack tip. Crack propagation path simulation is a numerical simulation method based on fracture mechanics and the finite element method to calculate the trajectory of crack propagation in a stress field. The reference deflection is a physical quantity characterizing the degree to which the crack propagation path deviates from the ideal cutting direction. Depth weighting correction is a process of weighting and adjusting the reference deflection based on the influence of crack deflection on cutting quality at different filamentation depths.

[0057] In the specific implementation, the coordinates of the melting front trajectory are extracted from the temperature field evolution data T(r,z,t) of the heat-affected zone. The material melting point Tm is set, and at each time step, an isothermal surface with a temperature equal to Tm is searched. The coordinates of the intersection of the isothermal surface and the cutting section are extracted to obtain the sequence of melting front trajectory coordinates. Based on the nonlinear absorption coefficient distribution α(r,z), the coordinates of the maximum absorption coefficient are searched, and the energy deposition centerline is fitted to obtain the result.

[0058] The geometric parameters of the filamentation section are calculated based on the coordinates of the melt front trajectory and the energy deposition centerline. The filamentation diameter D is the maximum radial span of the melt front trajectory; the filamentation roundness C is the ratio of the radius of the minimum circumcircle to the maximum incircle of the melt front trajectory; the filamentation taper T... aper This is the ratio of the difference in diameter between the two ends of the energy deposition centerline to its length. The above geometric parameters are combined into a filamentation morphology feature vector f. shape =[D,C,T aper ].

[0059] The process of generating the preset filament morphology template feature vector is as follows: During the first run of the equipment, N samples are continuously collected under standard cutting process conditions. template Sets of data, such as N template =500, determine the filament morphology feature vector f for each set of data. shape (k), k=1,2,…,N template Thus, the template mean vector μ is determined. shape The template mean vector is saved as a preset filament morphology template feature vector to the device's local storage. The weighted Euclidean distance between the current filament morphology feature vector and the template feature vector is used as the morphology deviation. Considering the dimensional differences and anomaly sensitivity of different geometric parameters, different weights are assigned to the diameter, roundness, and taper parameters. A judgment window of 100 consecutive pulse cycles is set. When the morphology deviation of all groups within the window exceeds the preset deviation threshold of 0.1 μm, the morphology deviation of the current group is output as the filament uniformity index, triggering parameter adjustment.

[0060] Based on the microcrack density spectrum ρ(a), where a is the crack size, calculate the stress intensity factor at the crack tip. : Where Y is the geometric correction factor and σ is the far-field stress. This represents the critical crack size.

[0061] Based on the cooling rate distribution in the stress intensity factor and temperature field evolution data of the heat-affected zone, the extended finite element method (EPM) is used to simulate the crack propagation path. The EPM achieves numerical simulation of crack propagation without re-meshing by introducing discontinuous enrichment functions into the conventional finite element shape functions. The simulation yields the crack propagation path under the coupled action of thermal and mechanical stresses, and the angle between the propagation path and the ideal cutting direction is calculated as a reference deflection.

[0062] The reference deflection amount is corrected by depth weighting based on the probability distribution P(d) of the filamentation depth. A depth weighting function W(d) is defined; the greater the filamentation depth, the greater the hazard of crack deflection. For example, W(d) = 1 + 0.1·d, where d is the filamentation depth. The crack deflection amount is δ. corrected =W(d)·δ, where δ is the reference deflection.

[0063] Step S50: Modulate the laser focal spot energy distribution with different preset cone angles using a Bessel beam shaping unit to obtain the focal spot energy topology parameters.

[0064] It should be noted that a Bessel beam shaping unit is an optical module that converts a Gaussian beam into a Bessel beam and achieves dynamic modulation of the focal spot energy distribution through a combination of a conical lens group and an axial conical lens. The focal spot energy topology parameters are a set of parameters characterizing the three-dimensional energy distribution of the Bessel beam focal spot, including axial depth of field and radial energy zonation.

[0065] In actual execution, the relative displacement of the cone lens group in the Bessel beam shaping unit is adjusted by a first preset cone angle to generate axial depth-of-field parameters without diffraction focal spots. The rotation angle of the axial cone mirror in the Bessel beam shaping unit is adjusted by a second preset cone angle to generate radial energy zonation parameters of the focal spot. The axial depth-of-field parameters are compared with a preset depth-of-field threshold. When the axial depth-of-field parameters meet the preset depth-of-field threshold, the depth-of-field qualification indicator is set to a valid state. The zonation energy uniformity of the radial energy zonation parameters of the focal spot is detected. When the zonation energy non-uniformity is lower than a preset non-uniformity threshold, the zonation uniformity indicator is set to a valid state. A logical AND operation is performed on the depth-of-field qualification indicator and the zonation uniformity indicator. When the operation result is true, the focal spot energy topology parameters are calculated based on the axial depth-of-field parameters and the radial energy zonation parameters of the focal spot.

[0066] In one feasible implementation, step S50 may include: adjusting the relative displacement of the cone lens group in the Bessel beam shaping unit by a first preset cone angle to generate an axial depth-of-field parameter without diffraction focal spot; adjusting the rotation angle of the axial cone mirror in the Bessel beam shaping unit by a second preset cone angle to generate a focal spot radial energy ring distribution parameter; comparing the axial depth-of-field parameter with a preset depth-of-field threshold, and setting the depth-of-field qualification mark to a valid state when the axial depth-of-field parameter meets the preset depth-of-field threshold; performing a ring energy uniformity detection on the focal spot radial energy ring distribution parameter, and setting the ring energy uniformity mark to a valid state when the ring energy non-uniformity is lower than a preset non-uniformity threshold; performing a logical AND operation on the depth-of-field qualification mark and the ring energy uniformity mark, and determining the focal spot energy topology parameter based on the axial depth-of-field parameter and the focal spot radial energy ring distribution parameter when the operation result is true.

[0067] It should be noted that the first preset cone angle refers to the optical cone angle equivalent to the relative displacement of the cone lens group, used to control the axial depth of field of the Bessel beam. The axial depth of field parameter is a physical quantity characterizing the diffraction-free propagation distance of the Bessel beam. The second preset cone angle refers to the optical cone angle equivalent to the rotation angle of the axial cone lens, used to control the distribution pattern of the radial energy rings of the focal spot. The radial energy ring distribution parameter of the focal spot is a function characterizing the energy distribution with radius within the cross-section of the focal spot.

[0068] The preset depth-of-field threshold is the minimum depth-of-field value determined based on the glass thickness to ensure that the filament penetrates the entire thickness of the glass. The annular energy non-uniformity is an index characterizing the uniformity of energy distribution across the radial annular zones of the focal spot, calculated as the ratio of the standard deviation to the mean of the energy in each annular zone. The preset non-uniformity threshold is the maximum non-uniformity value determined based on filament quality requirements to ensure uniform energy distribution. The focal spot energy topology parameters are a structured set of parameters integrating axial depth of field and radial energy distribution, used to characterize the three-dimensional energy distribution features of the Bessel beam focal spot.

[0069] In its implementation, the Bessel beam shaping unit consists of two conical lenses and one axial conical lens. The relative displacement Δx of the conical lens group is related to the first preset cone angle. The relationship is:

[0070] Where n is the refractive index of the lens, α is the cone angle of the conical lens, and f is the focal length of the lens.

[0071] Axial depth of field Z max With the first preset cone angle The relationship is: in, Let be the radius of the waist spot of the incident beam.

[0072] By adjusting the relative displacement Δx of the conical lens group, the first preset cone angle is changed. Thus controlling the axial depth of field Z max .

[0073] The relationship between the rotation angle φ of the axial cone mirror and the second preset cone angle θ2 is as follows: θ2=φ·(n-1)·β Where β is the bottom angle of the axial cone mirror.

[0074] The relationship between the radial energy ring distribution of the focal spot and the second preset cone angle θ2 is determined by the zeroth-order Bessel function J0: in, Let be the light intensity at radial position r. Let J0 be the total intensity of the incident light, J0 be the zeroth-order Bessel function of the first kind, k be the wave number, and r be the radial coordinate.

[0075] By adjusting the rotation angle φ of the axial cone mirror to change the second preset cone angle θ2, different morphologies of radial energy ring distribution of the focal spot can be obtained.

[0076] The axial depth of field parameter Z max Compared with the preset depth of field threshold Z th Compare the results when Z max ≥Z th At that time, the depth-of-field qualification mark is set to the valid state. The radial energy annular distribution parameters of the focal spot are tested for annular energy uniformity, and the annular energy non-uniformity U = σI / μI is calculated, where σI is the standard deviation of energy in each annular zone, and μI is the mean energy in each annular zone. When U th When the uniform ring marking is active, the uniform ring marking is set to active status. A logical AND operation is performed on the depth-of-field qualified marking and the uniform ring marking. When both markings are active simultaneously, the focal spot energy topology parameter E is calculated based on the axial depth-of-field parameter and the focal spot radial energy ring distribution parameter. topo =[Z max ,U].

[0077] Step S60: Execute model prediction control optimization based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters to generate laser cutting parameter adjustment instructions.

[0078] It should be noted that model predictive control optimization is an advanced control method that uses a dynamic model of the controlled object to solve for the optimal control sequence in a finite time domain. The laser cutting parameter adjustment command is the final control output that integrates three types of information: filament uniformity, crack deflection, and focal spot energy topology. It includes adjustment values ​​for laser power, pulse energy, repetition frequency, scanning speed, and auxiliary gas pressure.

[0079] ​In actual implementation, the filament uniformity index is mapped to process capability, and filament quality grade parameters are generated based on morphological deviation, divided into four levels: Level 1 Excellent, Level 2 Good, Level 3 Qualified, and Level 4 Abnormal. The degree of deflection is classified into four levels: No Deflection, Slight Deflection, Moderate Deflection, and Severe Deflection, based on the numerical range of crack deflection. The filament quality grade, degree of deflection grade, and focal spot energy topology parameters are input into the trained predictive control model. Within the constraints, rolling optimization is performed to obtain the optimal adjustment amount for each laser cutting parameter. The adjustment amounts of each parameter are integrated to generate laser cutting parameter adjustment commands, which are then output to the motion control unit to complete the adaptive adjustment of the cutting parameters.

[0080] In one feasible implementation, step S60 may include: mapping the filament uniformity index to process capability to generate filament quality grade parameters; mapping the crack deflection amount to a deflection threshold to generate crack risk grade parameters; jointly discriminating the annular energy concentration parameter and the axial energy gradient parameter in the focal spot energy topology parameters to generate energy distribution grade parameters; performing weighted summation on the filament quality grade parameters, the crack risk grade parameters, and the energy distribution grade parameters according to preset weight coefficients to obtain a comprehensive process score; constructing a model predictive control optimization problem based on the comprehensive process score, using laser power, pulse energy, repetition frequency, scanning speed, and auxiliary gas pressure as control variables, and filament uniformity index and crack deflection amount as controlled outputs, solving for the optimal control sequence, and generating laser cutting parameter adjustment instructions.

[0081] It should be noted that the process capability mapping is a mapping rule that maps continuous morphological deviation values ​​to discrete quality levels, facilitating subsequent multi-parameter weighted scoring. The deflection threshold mapping is a mapping rule that maps continuous crack deflection amounts to discrete risk levels, used to quantify the potential harm of crack propagation deflection to cutting quality. The energy distribution level parameter is a discrete level parameter that integrates depth-of-field compliance and energy uniformity, used to characterize the degree to which the current focal spot energy modulation result meets process requirements.

[0082] The weighting coefficients are pre-calibrated through process experiments. The weight for filament quality level is typically 0.4, for crack risk level 0.35, and for energy distribution level 0.25. These weights can be adjusted based on the actual thickness and material type of the glass being cut. The state-space model is a linear state equation identified from historical process data, describing the dynamic impact of changes in control variables on the filament uniformity index and crack deflection. The objective function of the quadratic programming optimization problem is to maximize the comprehensive process score within the prediction time domain, with constraints set within the allowable adjustment range of each laser cutting parameter.

[0083] In the specific implementation, a filament quality grade mapping table is set, where the deviation range corresponds to the grade as follows: deviation less than 0.02μm corresponds to Grade 1 Excellent, 0.02μm to 0.05μm corresponds to Grade 2 Good, 0.05μm to 0.1μm corresponds to Grade 3 Acceptable, and greater than 0.1μm corresponds to Grade 4 Abnormal. Similarly, a crack deflection grade mapping table is set, where deflection less than 0.5° corresponds to no deflection, 0.5° to 2° corresponds to slight deflection, 2° to 5° corresponds to moderate deflection, and greater than 5° corresponds to severe deflection. An energy distribution grade mapping table is set, where depth of field is acceptable and zonal non-uniformity is less than 0.05 corresponds to Grade 1 Excellent, depth of field is acceptable but non-uniformity is 0.05 to 0.1 corresponds to Grade 2 Good, depth of field is unacceptable but non-uniformity is less than 0.1 corresponds to Grade 3 Acceptable, and depth of field is unacceptable and non-uniformity is greater than 0.1 corresponds to Grade 4 Abnormal. The discrete levels are converted into corresponding scores: Level 1 Excellent (100 points), Level 2 Good (80 points), Level 3 Pass (60 points), and Level 4 Abnormal (40 points). Substituting these scores into the weighted summation formula yields the comprehensive process score. Then, the parameter adjustment amount is calculated with maximizing the comprehensive process score as the optimization objective.

[0084] A predictive control optimization problem is constructed based on a comprehensive process score. A state-space model of the controlled object is established:

[0085] Wherein, the state vector =[η(k),δ(k)] T η(k) is the filament uniformity index in the k-th control cycle, δ(k) is the crack deflection in the k-th control cycle, and the control vector is... =[P(k),Ep(k),f(k),v(k),p_g(k)]T, where P(k) is the laser power, Ep(k) is the single pulse energy, f(k) is the pulse repetition frequency, v(k) is the laser scanning speed, and p_g(k) is the auxiliary gas pressure. v(k) is the process noise vector, v(k) is the observation noise vector, and matrices A, B, and C are parameter matrices identified through historical process data.

[0086] In satisfying i=0,1,…,N c-1 Under the constraints, construct the optimization objective function: in, For prediction in the time domain, N c To control the time domain, S(i) is the comprehensive process score for the i-th prediction step. For terminal weight, Step size weight, A comprehensive process score for predicting the time-domain endpoint.

[0087] The optimal control sequence Δu is obtained by solving a quadratic programming problem. The first control increment Δu(0) is taken as the parameter adjustment amount for the current cycle. The laser cutting parameter adjustment instructions ΔP, ΔEp, Δf, Δv, and Δp_g are integrated and sent to the motion control module to complete the parameter update.

[0088] This embodiment provides an adaptive adjustment method for picosecond infrared laser glass filament cutting parameters. By acquiring the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser interaction region and performing baseline correction and adaptive noise reduction, the interference of plasma radiation fluctuations and acoustic emission environmental noise on signal quality can be effectively suppressed. The nonlinear absorption coefficient distribution is analyzed using a filamentation energy deposition model, and the temperature field evolution data of the heat-affected zone is reconstructed, enabling a quantitative characterization of the transient physical process of laser-glass interaction. Combined with the multi-physics coupling prediction parameters output by the physical information neural surrogate model, the filamentation depth, microcrack density, and stress relaxation rate are adjusted. Simultaneous prediction, based on the melting front trajectory and energy deposition centerline filament morphology prediction, and based on the stress intensity factor crack propagation path simulation, accurately determined the filament uniformity and crack deflection. Dynamic control of the focal spot energy topology was achieved through the cone angle modulation of the Bessel beam shaping unit. Finally, the filament uniformity index, crack deflection, and focal spot energy topology parameters were incorporated into the model prediction control optimization framework to generate laser cutting parameter adjustment instructions. This enabled multi-physics field coupled sensing, prediction, and closed-loop optimization of the picosecond infrared laser glass filament cutting process, effectively improving the accuracy and reliability of adaptive adjustment of glass filament cutting parameters under complex working conditions.

[0089] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Based on this, step S20 may include steps S301 to S304:

[0090] Step S301: Construct a proxy model based on a physical information neural network. The physical information neural proxy model includes a spatiotemporal coordinate embedding layer, a multi-scale feature encoder, and a multi-task decoder for filamentation depth prediction, microcrack density prediction, and stress relaxation prediction, respectively.

[0091] It should be noted that the Physical Information Neural Network (PIN) is a machine learning architecture that embeds physical laws as constraints into the neural network loss function, enabling it to satisfy physical conservation laws while being data-driven. The spatiotemporal coordinate embedding layer is a network component that uses Fourier feature mapping to map spatiotemporal coordinates to a high-dimensional frequency space. The multi-scale feature encoder is a network component that uses residual connections and attention mechanisms for multi-scale feature extraction. The multi-task decoder consists of three parallel functional branches that output the filamentation depth probability distribution, microcrack density spectrum, and stress relaxation rate, respectively.

[0092] In actual implementation, the spatiotemporal coordinate embedding layer employs a stochastic Fourier feature map: γ(x)=[sin(2πB·x),cos(2πB·x)], where B is a random frequency matrix sampled from a normal distribution, mapping the input spatiotemporal coordinate x to a high-dimensional space. The multi-scale feature encoder uses a stacked structure of 5 residual blocks, each containing two fully connected layers and skip connections, outputting multi-scale feature vectors. The multi-task decoder contains three parallel fully connected branches: the filamentation depth prediction head outputs the probability distribution parameters of the filamentation depth, including the mean and variance; the microcrack density prediction head outputs the log-normal distribution parameters of the microcrack density; and the stress relaxation prediction head outputs the exponentially decaying parameters of the stress relaxation rate.

[0093] Step S302: Collect plasma radiation spectrum signals and acoustic emission time-frequency signals for different glass grades, thicknesses, and cutting speeds; perform microscopic profile annotation on the filamentation depth; perform scanning electron microscopy statistical annotation on the microcrack density; perform Raman spectroscopy annotation on the residual stress; construct a multiphysics dataset; and divide the multiphysics dataset into a training set and a validation set according to a preset ratio.

[0094] It should be noted that "different glass grades" refers to a collection of glass types from different manufacturers and with different chemical compositions, including Corning Gorilla Glass, Schott Borosilicate Glass, and Asahi Glass. "Different thickness specifications" refers to a range of glass thicknesses from 0.1 mm to 10 mm. "Different cutting speeds" refers to a range of laser scanning speeds from 10 mm / s to 1000 mm / s. Microscopic profile annotation is a method of marking by measuring the filamentation depth under a microscope after preparing cross-sectional samples through cutting, grinding, and polishing. Scanning electron microscopy statistical annotation is a method of marking by observing the morphology of microcracks and statistically analyzing the number and size of cracks using a scanning electron microscope. Raman spectroscopy annotation is a method of marking by measuring the spectral shift caused by residual stress within the glass using a Raman spectrometer and converting it into stress values.

[0095] In the specific implementation, picosecond infrared laser filament cutting experiments were conducted under various combinations of glass grades, thicknesses, and cutting speeds, collecting approximately 2000 sets of experimental data. Microscopic profiles were prepared for each set of experimental samples, and the filament depth was measured under an optical microscope and recorded as the true value of the filament depth. Scanning electron microscopy was performed on each set of experimental samples, capturing high-magnification images of the cracked areas. Image processing methods were used to statistically analyze the number and size distribution of microcracks, recording the true value of the microcrack density spectrum. Raman spectroscopy measurements were performed on each set of experimental samples, and the residual stress value was calculated based on the calibration relationship between Raman spectral line shift and stress, recording the true value of the stress relaxation rate. The labeled multiphysics dataset was divided into training and validation sets at a predetermined ratio of 8:2, using stratified random sampling to ensure consistent proportional distribution across categories and process conditions.

[0096] Step S303: Train the physical information neural agent model using the training set, embed the nonlinear Schrödinger equation and heat conduction equation as physical constraints into the loss function, optimize the network parameters until the root mean square error of the prediction on the validation set reaches the preset accuracy threshold, and obtain the pre-trained physical information neural agent model.

[0097] It should be noted that the nonlinear Schrödinger equation is a wave equation describing the propagation of ultrashort pulse lasers in transparent media, considering Kerr self-focusing and multiphoton absorption effects. The heat conduction equation is a diffusion equation describing the conduction of heat in a material. The physical constraint term is a component of the loss function, incorporating the residuals of the above partial differential equations as a penalty term.

[0098] In actual implementation, a training set is used to train the physical information neural agent model. The loss function consists of a data-driven term and a physical constraint term:

[0099] L=Ldata+λ·Lphysics Wherein, the data-driven term Ldata is the mean square error between the predicted value and the true value, the physical constraint term Lphysics is the sum of squared residuals of the nonlinear Schrödinger equation and the heat conduction equation, and λ is the physical constraint weight coefficient.

[0100] The Adam optimizer was used for training, with a learning rate of 0.001, a batch size of 64, and 5000 training epochs. During training, physical constraints ensured that the network output not only fit the experimental data but also satisfied the physical laws of laser propagation and heat conduction. Training was stopped when the root mean square error of predictions on the validation set reached a preset accuracy threshold of 5%, resulting in a pre-trained physical information neural agent model.

[0101] Step S304: Input the plasma radiation spectrum feature vector and acoustic emission time-frequency feature vector in the corrected signal data into the pre-trained physical information neural agent model, and output the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate through the multi-task decoder as multi-physics coupling prediction parameters.

[0102] It should be noted that the plasma radiation spectral feature vector is a low-dimensional feature vector extracted from the calibration signal data, characterizing the distribution of plasma radiation intensity. The acoustic emission time-frequency feature vector is a low-dimensional feature vector extracted from the calibration signal data, characterizing the time-frequency characteristics of the acoustic emission signal.

[0103] In actual implementation, the plasma radiation spectrum feature vector and the acoustic emission time-frequency feature vector are concatenated into a joint feature vector, which is then input into the pre-trained physical information neural agent model. The spatiotemporal coordinate embedding layer performs Fourier feature mapping on the input coordinates, the multi-scale feature encoder extracts multi-scale feature representations, and the multi-task decoder outputs the following respectively: the filamentation depth prediction head outputs the mean depth μd and the standard deviation depth σd, forming the filamentation depth probability distribution N(μd,σd²); the microcrack density prediction head outputs the logarithmic mean μρ and the logarithmic standard deviation σρ, forming the microcrack density log-normal distribution LN(μρ,σρ²); and the stress relaxation prediction head outputs the relaxation time constant τ and the initial stress σ0, forming the stress relaxation rate σ(t) = σ0·exp(-t / τ).

[0104] In this embodiment, by employing the physical constraint architecture of a physical information neural network, the nonlinear Schrödinger equation and the heat conduction equation are embedded into the loss function, so that the model output simultaneously satisfies data fitting and physical conservation, thereby improving the model's generalization ability and extrapolation performance under conditions of few samples. The multi-task decoder realizes the simultaneous prediction of filamentation depth, microcrack density and stress relaxation, thereby improving prediction efficiency and consistency.

[0105] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the adaptive adjustment method of picosecond infrared laser glass filament cutting parameters of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0106] This application also provides an adaptive adjustment device for picosecond infrared laser glass filament cutting parameters. Please refer to [reference needed]. Figure 2 The picosecond infrared laser glass filament cutting parameter adaptive adjustment device includes: The acquisition module 10 is used to acquire the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser action area and perform baseline correction and adaptive noise reduction to obtain the corrected signal data.

[0107] The analysis module 20 is used to analyze the nonlinear absorption coefficient distribution based on the correction signal data through the filamentation energy deposition model, and reconstruct the temperature field evolution data of the heat-affected zone.

[0108] The prediction module 30 is used to input the correction signal data into a pre-trained physical information neural agent model to obtain multi-physics field coupling prediction parameters, which include the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate.

[0109] The determination module 40 is used to predict the filament morphology and determine the crack propagation based on the multi-physics field coupling prediction parameters, the nonlinear absorption coefficient distribution and the temperature field evolution data of the heat-affected zone, and to obtain the filament uniformity index and crack deflection amount.

[0110] The modulation module 50 is used to modulate the laser focal spot energy distribution with different preset cone angles through the Bessel beam shaping unit to obtain the focal spot energy topology parameters.

[0111] The optimization module 60 is used to perform model prediction control optimization based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters to generate laser cutting parameter adjustment instructions.

[0112] The picosecond infrared laser glass filament cutting parameter adaptive adjustment device provided in this application adopts the picosecond infrared laser glass filament cutting parameter adaptive adjustment method in the above embodiments. It can solve the technical problem that existing picosecond laser glass filament cutting processes lack real-time sensing and predictive control of the transient process coupled with multi-physics fields, resulting in the inability to adaptively adjust cutting parameters and unstable filament quality. Compared with the prior art, the beneficial effects of the picosecond infrared laser glass filament cutting parameter adaptive adjustment device provided in this application are the same as those of the picosecond infrared laser glass filament cutting parameter adaptive adjustment method provided in the above embodiments. Furthermore, other technical features in the picosecond infrared laser glass filament cutting parameter adaptive adjustment device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0113] This application provides a picosecond infrared laser glass filament cutting parameter adaptive adjustment device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the picosecond infrared laser glass filament cutting parameter adaptive adjustment method in the above embodiment 1.

[0114] The following is for reference. Figure 3The diagram illustrates a structural schematic suitable for implementing the picosecond infrared laser glass filament cutting parameter adaptive adjustment device in the embodiments of this application. The picosecond infrared laser glass filament cutting parameter adaptive adjustment device in the embodiments of this application can include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 3 The illustrated picosecond infrared laser glass filament cutting parameter adaptive adjustment device is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0115] like Figure 3 As shown, the picosecond infrared laser glass filament cutting parameter adaptive adjustment device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in ROM (Read Only Memory) 1002 or a program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the picosecond infrared laser glass filament cutting parameter adaptive adjustment device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the picosecond infrared laser glass filament cutting parameter adaptive adjustment device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a picosecond infrared laser glass filament cutting parameter adaptive adjustment device with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0116] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0117] 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.

[0118] This application provides a non-transitory computer-readable storage medium storing a computer program thereon, characterized in that, when the computer program is executed by a processor, it implements the adaptive adjustment method for picosecond infrared laser glass filament cutting parameters as described above.

[0119] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for adaptive adjustment of parameters for picosecond infrared laser glass filament cutting, characterized in that, The method includes: The plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser interaction region are acquired and baseline correction and adaptive noise reduction are performed to obtain the corrected signal data; Based on the corrected signal data, the nonlinear absorption coefficient distribution is analyzed using a filamentation energy deposition model, and the temperature field evolution data of the heat-affected zone is reconstructed. The corrected signal data is input into a pre-trained physical information neural agent model to obtain multi-physics field coupling prediction parameters, which include the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate. Based on the multiphysics field coupling prediction parameters, the nonlinear absorption coefficient distribution, and the temperature field evolution data of the heat-affected zone, the filament morphology is predicted and crack propagation is determined, resulting in the filament uniformity index and crack deflection amount. The laser focal spot energy topology parameters are obtained by modulating the laser focal spot energy distribution with different preset cone angles using a Bessel beam shaping unit. Based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters, a model prediction control optimization is performed to generate laser cutting parameter adjustment instructions.

2. The method as described in claim 1, characterized in that, The process involves acquiring the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser's active region, performing baseline correction and adaptive noise reduction, to obtain corrected signal data, including: The plasma radiation spectrum signal of the picosecond infrared laser-actuated region is acquired by the plasma spectrum acquisition unit, and the intensity value of the plasma characteristic spectral line is output after dark current subtraction. Acoustic emission time-frequency signals in the picosecond infrared laser action area are acquired by an acoustic emission sensor array, and the acoustic emission time-frequency signals are divided into multiple time windows along the cutting direction. The root mean square amplitude and spectral centroid of each time window are determined, and an acoustic emission feature vector is generated. Based on the plasma characteristic spectral line intensity values ​​and the acoustic emission characteristic vector, an adaptive gain scheduling algorithm is used to generate spectral integration time parameters, acoustic emission sampling rate parameters, and signal amplification parameters. Based on the spectral integration time parameter, the acoustic emission sampling rate parameter, and the signal amplification factor parameter, baseline correction is performed on the plasma radiation spectral signal and the acoustic emission time-frequency signal to obtain baseline-de-baseline signal data. The baseline-degraded signal data is sequentially processed by empirical mode decomposition, variational mode decomposition, and wavelet packet thresholding to obtain the corrected signal data.

3. The method as described in claim 1, characterized in that, The process of analyzing the nonlinear absorption coefficient distribution using a filamentation energy deposition model based on the corrected signal data and reconstructing the temperature field evolution data of the heat-affected zone includes: Obtain the pre-calibrated time-domain waveform parameters of the laser pulse, the nonlinear optical coefficient of the glass material, and the thermal property parameters; Based on the laser pulse time-domain waveform parameters and the nonlinear optical coefficients of the glass material, the inverse problem of the plasma radiation spectrum in the correction signal data is solved to obtain the multiphoton absorption coefficient and the avalanche ionization coefficient. Based on the multiphoton absorption coefficient and the avalanche ionization coefficient, the deposition distribution of laser energy along the propagation path is determined by the filamentation energy deposition model, and the nonlinear absorption coefficient distribution is obtained. A transient heat conduction equation is established based on the aforementioned thermal property parameters and the aforementioned nonlinear absorption coefficient distribution. The spatiotemporal distribution of the temperature field in the heat-affected zone is solved using the finite difference method to obtain the temperature field evolution data of the heat-affected zone. Based on the temperature field evolution data of the heat-affected zone, the temperature gradient distribution and cooling rate distribution are determined, and a thermodynamic characteristic spectrum is generated.

4. The method as described in claim 1, characterized in that, The step of inputting the corrected signal data into a pre-trained physical information neural agent model to obtain multi-physics coupling prediction parameters includes: A proxy model based on physical information neural network is constructed. The physical information neural proxy model includes a spatiotemporal coordinate embedding layer, a multi-scale feature encoder, and a multi-task decoder for filamentation depth prediction, microcrack density prediction, and stress relaxation prediction, respectively. Plasma radiation spectrum signals and acoustic emission time-frequency signals were collected under different glass grades, thicknesses and cutting speeds. The filamentation depth was marked by microscopic profiles, the microcrack density was marked by scanning electron microscopy, and the residual stress was marked by Raman spectroscopy. A multiphysics dataset was constructed, and the multiphysics dataset was divided into a training set and a validation set according to a preset ratio. The physical information neural agent model is trained using the training set. The nonlinear Schrödinger equation and the heat conduction equation are embedded as physical constraints into the loss function to optimize the network parameters until the root mean square error of the prediction on the validation set reaches a preset accuracy threshold, thus obtaining the pre-trained physical information neural agent model. The plasma radiation spectrum feature vector and acoustic emission time-frequency feature vector in the corrected signal data are input into the pre-trained physical information neural agent model, and the multi-task decoder outputs the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate as multi-physics coupling prediction parameters.

5. The method as described in claim 1, characterized in that, The process of predicting filament morphology and determining crack propagation based on the multiphysics coupling prediction parameters, the nonlinear absorption coefficient distribution, and the temperature field evolution data of the heat-affected zone, to obtain the filament uniformity index and crack deflection, includes: The melting front trajectory coordinates are extracted from the temperature field evolution data of the heat-affected zone, and the energy deposition centerline is determined based on the nonlinear absorption coefficient distribution. Based on the melting front trajectory coordinates and the energy deposition centerline, the geometric parameters of the filamentation section are determined, and the filamentation morphology feature vector is generated. The filamentation morphology feature vector is compared with the preset filamentation morphology template feature vector to determine the morphology deviation. If the morphology deviation exceeds the preset deviation threshold in multiple consecutive pulse cycles, the morphology deviation is used as the filamentation uniformity index. The stress intensity factor at the crack tip is determined based on the microcrack density spectrum in the multiphysics coupling prediction parameters. The crack propagation path is simulated based on the stress intensity factor and the cooling rate distribution in the temperature field evolution data of the heat-affected zone to obtain the reference deflection. The reference deflection amount is corrected by depth weighting based on the filamentation depth probability distribution to obtain the crack deflection amount.

6. The method as described in claim 1, characterized in that, The process of modulating the laser focal spot energy distribution with different preset cone angles using a Bessel beam shaping unit to obtain focal spot energy topology parameters includes: The relative displacement of the cone lens group in the Bessel beam shaping unit is adjusted by the first preset cone angle to generate axial depth of field parameters without diffraction focal spots. The rotation angle of the axis-cone mirror in the Bessel beam shaping unit is adjusted by the second preset cone angle to generate the radial energy ring distribution parameters of the focal spot. The axial depth of field parameter is compared with a preset depth of field threshold. When the axial depth of field parameter meets the preset depth of field threshold, the depth of field qualification mark is set to a valid state. The annular energy uniformity of the radial energy distribution parameters of the focal spot is detected. When the annular energy non-uniformity is lower than the preset non-uniformity threshold, the annular uniformity indicator is set to an effective state. A logical AND operation is performed on the depth-of-field qualification mark and the ring uniformity mark. When the operation result is true, the focal spot energy topology parameter is determined based on the axial depth-of-field parameter and the focal spot radial energy ring distribution parameter.

7. The method as described in claim 1, characterized in that, The step of performing model prediction control optimization based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters to generate laser cutting parameter adjustment instructions includes: Process capability mapping is performed on the filament uniformity index to generate filament quality grade parameters; The crack deflection amount is mapped to a deflection threshold to generate crack risk level parameters; The annular energy concentration parameter and axial energy gradient parameter in the focal spot energy topology parameters are jointly discriminated to generate energy distribution level parameters; The filament quality grade parameter, the crack risk grade parameter, and the energy distribution grade parameter are weighted and summed according to preset weighting coefficients to obtain a comprehensive process score. Based on the comprehensive process score, a model predictive control optimization problem is constructed. Laser power, pulse energy, repetition frequency, scanning speed and auxiliary gas pressure are used as control variables, and filament uniformity index and crack deflection are used as controlled outputs. The optimal control sequence is solved to generate laser cutting parameter adjustment instructions.

8. A device for adaptive adjustment of parameters for picosecond infrared laser glass filament cutting, characterized in that, The device includes: The acquisition module is used to acquire the plasma radiation spectrum signal and acoustic emission time-frequency signal of the picosecond infrared laser action area and perform baseline correction and adaptive noise reduction to obtain the corrected signal data; The analysis module is used to analyze the nonlinear absorption coefficient distribution based on the correction signal data using a filamentation energy deposition model, and to reconstruct the temperature field evolution data of the heat-affected zone. The prediction module is used to input the correction signal data into a pre-trained physical information neural agent model to obtain multi-physics field coupling prediction parameters, which include the filamentation depth probability distribution, microcrack density spectrum and stress relaxation rate. The determination module is used to predict the filament morphology and determine the crack propagation based on the multi-physics field coupling prediction parameters, the nonlinear absorption coefficient distribution and the temperature field evolution data of the heat-affected zone, and to obtain the filament uniformity index and crack deflection amount. The modulation module is used to modulate the laser focal spot energy distribution with different preset cone angles through the Bessel beam shaping unit to obtain the focal spot energy topology parameters; The optimization module is used to perform model prediction control optimization based on the filament uniformity index, the crack deflection amount, and the focal spot energy topology parameters, and generate laser cutting parameter adjustment instructions.

9. A picosecond infrared laser glass filament cutting parameter adaptive adjustment device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method for adaptive adjustment of picosecond infrared laser glass filament cutting parameters as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the adaptive adjustment method for picosecond infrared laser glass filament cutting parameters as described in any one of claims 1 to 7.

Images