A laser cleaning closed-loop automatic control method and system

Abstract

The application relates to the technical field of laser processing, and discloses a laser cleaning closed-loop automatic control method and system, the method comprising the following steps: obtaining ultrasonic echo data of a surface of a metal substrate to be cleaned and corresponding collection space coordinates to obtain an initial thickness distribution matrix; obtaining a partition energy requirement range according to the initial thickness distribution matrix; performing evaluation and sorting according to the partition energy requirement range to obtain a high-risk area adjustment sequence; performing scheduling according to the high-risk area adjustment sequence to obtain a basic global cleaning plan containing initial laser driving parameters; quantifying surface infrared thermal image signals and spectral reflection signals to obtain real-time surface state characteristics; comparing the real-time surface state characteristics to obtain a local over-temperature early warning instruction, and then performing energy reduction calibration to obtain a safe laser output instruction and an updated physical execution path. The method realizes high-precision dynamic distribution of cleaning energy and extreme closed-loop blocking of thermal overload damage.

Description

Technical Field

[0001] This invention relates to the field of laser processing technology, and in particular to a closed-loop automated control method and system for laser cleaning. Background Technology

[0002] Currently, in the fields of precision manufacturing and industrial maintenance, the quality of material surface cleaning directly affects the subsequent processing performance and service life of components. Especially in the removal of oxide layers, rust layers, or coatings on metal surfaces, laser cleaning, with its non-contact nature and small heat-affected zone, has become a key link in promoting the full-process automation and green transformation of intelligent heat treatment production lines.

[0003] In a current technology, a laser cleaning scheme based on preset static process parameters is typically employed. This involves setting a fixed laser output power, pulse frequency, and scanning speed through manual experience or preliminary sampling, and driving the actuator to complete the surface treatment along a predetermined trajectory. However, this existing technology exhibits significant limitations when facing complex and variable working conditions. Because the distribution of contaminants on metal surfaces often exhibits randomness and thickness heterogeneity, a preset constant energy output is difficult to adapt to the constantly fluctuating cleaning requirements. If the system lacks real-time sensing capabilities of the instantaneous physical state of the cleaning interface, it cannot dynamically adjust energy distribution according to the reduction of contaminants, resulting in poor consistency in cleaning quality. Furthermore, due to the lack of closed-loop monitoring and early warning feedback logic for the cleaning status, the system struggles to effectively avoid the risk of thermal damage to the substrate caused by instantaneous energy overload while ensuring cleaning efficiency.

[0004] Existing technologies suffer from technical problems such as inaccurate dynamic energy distribution and mismatch in cleaning status feedback during the laser cleaning process. Summary of the Invention

[0005] This invention provides a closed-loop automated control method and system for laser cleaning to solve the technical problems of inaccurate dynamic energy distribution and mismatch in cleaning status feedback in the existing laser cleaning process.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a closed-loop automated control method for laser cleaning, comprising: Acquire ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned, and perform spatial interpolation on the ultrasonic echo data based on the acquisition spatial coordinates to obtain an initial thickness distribution matrix; The energy demand range of each zone is obtained by dividing the region according to the initial thickness distribution matrix and the preset thickness energy correlation model; The regional energy fluctuation parameter is obtained by statistically evaluating the amplitude of the energy demand range of the partition. The regional energy fluctuation parameter is compared with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates. The high-risk physical areas and boundary coordinates are weighted and sorted to obtain the high-risk area adjustment sequence. Based on the high-risk zone adjustment sequence and the preset total power constraint of the equipment hardware, the cleaning resources are coordinated and scheduled to obtain a basic global cleaning plan including the initial laser driving parameters. Based on the basic global cleaning plan, surface infrared thermal imaging signals and spectral reflection signals are obtained during the physical cleaning process. The surface infrared thermal imaging signals and spectral reflection signals are then quantized using physical features to obtain real-time surface state features. A local over-temperature warning command is obtained by comparing the real-time surface state characteristics with a preset substrate damage temperature threshold. Based on the local over-temperature warning command, the initial laser driving parameters are adaptively calibrated to reduce energy and obtain a safe laser output command and an updated physical execution path.

[0007] Secondly, the present invention provides a closed-loop automated control system for laser cleaning, comprising: The initial distribution interpolation module is used to acquire ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned, and to perform spatial interpolation on the ultrasonic echo data based on the acquisition spatial coordinates to obtain an initial thickness distribution matrix. The regional energy segmentation module is used to segment regions based on the initial thickness distribution matrix and a preset thickness energy correlation model to obtain the energy demand range of each region. The risk area screening module is used to perform amplitude statistical evaluation based on the energy demand range of the partition to obtain the regional energy fluctuation parameter, compare the regional energy fluctuation parameter with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates, and perform weighted sorting based on the high-risk physical areas and boundary coordinates to obtain the high-risk area adjustment sequence. The resource scheduling and planning module is used to coordinate and schedule cleaning resources according to the high-risk area adjustment sequence and the preset total power constraints of equipment hardware, so as to obtain a basic global cleaning plan including initial laser driving parameters. The real-time state quantization module is used to acquire surface infrared thermal image signals and spectral reflection signals during the physical cleaning process according to the basic global cleaning plan, and to perform physical feature quantization on the surface infrared thermal image signals and the spectral reflection signals to obtain real-time surface state features. The energy reduction calibration update module is used to compare the real-time surface state characteristics with the preset substrate damage temperature threshold to obtain a local over-temperature warning command, and to perform adaptive energy reduction calibration on the initial laser driving parameters based on the local over-temperature warning command to obtain a safe laser output command and an updated physical execution path.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains an initial thickness distribution matrix by spatially interpolating ultrasonic echo data from the surface of the metal substrate to be cleaned. It then uses a thickness-energy correlation model and amplitude statistical evaluation to screen out high-risk physical regions and boundary coordinates. Finally, based on the total power constraints of the equipment hardware, it coordinates and schedules cleaning resources to obtain a basic global cleaning plan. This approach overcomes the shortcomings of blindly or uniformly distributing energy in traditional cleaning schemes. It can proactively detect and lock high-risk areas prone to cleaning loss of control under complex conditions where the thickness distribution of contaminants on the surface is heterogeneous. This allows for feedforward planning of initial laser drive parameters from a global perspective. Ultimately, it achieves precise dynamic pre-allocation of laser cleaning energy under complex conditions, improving the overall resource utilization efficiency of the equipment and the spatial consistency of cleaning quality.

[0009] (2) This invention acquires surface infrared thermal imaging signals and spectral reflectance signals during the physical cleaning process based on a basic global cleaning plan, and quantifies the physical characteristics of these signals to obtain real-time surface state features. This multimodal real-time physical sensing mechanism breaks through the black-box dilemma of missing state feedback in traditional cleaning processes. Infrared thermal imaging signals can accurately reflect the thermodynamic accumulation on the macroscopic interface, while spectral reflectance signals can effectively characterize the physical stripping progress of contaminants at the microscopic level. The combination of the two transforms the abstract cleaning process into a concrete and controllable quantitative indicator. Ultimately, it achieves comprehensive and high-precision dynamic capture of the instantaneous physical state of laser cleaning, providing reliable underlying mapping data support for subsequent closed-loop parameter adjustment.

[0010] (3) This invention obtains a local overheat warning command by comparing real-time surface state characteristics with a preset substrate damage temperature threshold, and then adaptively reduces the energy of the initial laser drive parameters accordingly to obtain a safe laser output command and an updated physical execution path. This underlying adaptive calibration mechanism constructs a strict physical closed-loop control logic. When a local overheating risk is detected, the system does not require manual intervention and directly performs millisecond-level energy reduction blocking on the laser output at the drive layer, while simultaneously compensating for the speed of the physical execution path of the device, thereby effectively cutting off the conduction of heat overload to the deep layers of the metal substrate. Ultimately, it solves the hidden danger of substrate thermal damage caused by instantaneous energy overload during the cleaning process, maximizing the cleaning and stripping efficiency while achieving safe protection of the metal substrate. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the closed-loop automated control method for laser cleaning provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the closed-loop automated control system for laser cleaning provided in the second embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides a closed-loop automated control method for laser cleaning, comprising the following steps: S1, acquire ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned, and perform spatial interpolation on the ultrasonic echo data based on the acquisition spatial coordinates to obtain an initial thickness distribution matrix. S2, based on the initial thickness distribution matrix and the preset thickness energy correlation model, the region is segmented to obtain the energy demand range of the partition; S3, perform amplitude statistical evaluation based on the energy demand range of the partition to obtain regional energy fluctuation parameters, compare the regional energy fluctuation parameters with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates, and perform weighted sorting based on the high-risk physical areas and boundary coordinates to obtain a high-risk area adjustment sequence; S4. Based on the high-risk area adjustment sequence and the preset total power constraint of the equipment hardware, the cleaning resources are coordinated and scheduled to obtain a basic global cleaning plan including the initial laser driving parameters. S5. Based on the basic global cleaning plan, obtain the surface infrared thermal image signal and spectral reflection signal during the physical cleaning process, and perform physical feature quantization on the surface infrared thermal image signal and the spectral reflection signal to obtain real-time surface state features. S6. Based on the real-time surface state characteristics, a local over-temperature warning command is obtained by comparing them with the preset substrate damage temperature threshold. Based on the local over-temperature warning command, the initial laser driving parameters are adaptively calibrated to reduce energy and obtain a safe laser output command and an updated physical execution path.

[0014] In step S1, ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned are obtained, and an initial thickness distribution matrix is ​​obtained by spatial interpolation of the ultrasonic echo data based on the acquisition spatial coordinates.

[0015] The initial thickness distribution matrix is ​​obtained by spatial interpolation of the ultrasonic echo data based on the acquired spatial coordinates, including: The ultrasonic echo data and the corresponding acquisition spatial coordinates are smoothed and filled to obtain a continuous two-dimensional thickness scalar field. The initial thickness distribution matrix is ​​obtained by discretizing and sampling the two-dimensional thickness scalar field.

[0016] In one implementation, ultrasonic echo data and corresponding spatial coordinates of the surface of the metal substrate to be cleaned are acquired. In this embodiment, an ultrasonic sensor array positioned at the end of a mobile robotic arm performs a gridded scan of the metal substrate surface, recording the transmission time of the detection pulses and the echo reception time. The motion control module of the mobile robotic arm is deployed in an industrial control microprocessor. By reading the encoder values ​​of the servo motor at the base of the mobile robotic arm and combining them with a forward kinematic model constructed based on DH parameters, the joint angles of each servo motor are converted into spatial coordinates in a global Cartesian coordinate system. The lateral and longitudinal coordinates of each thickness detection point are calculated to form the acquisition spatial coordinates. The system performs a subtraction operation between the echo reception time and the transmission time to obtain a time difference value. This time difference value is then multiplied by a preset sound velocity parameter of the attachment to obtain the thickness value as the ultrasonic echo data.

[0017] It should be noted that the method for determining the preset sound velocity parameter of the adhering material is as follows: a standard calibration block of the same material as the contaminant to be cleaned is obtained, and the actual round-trip time of the sound wave in the standard calibration block is measured by ultrasonic pulse at a constant reference room temperature. The known absolute thickness of the standard calibration block is divided by 1 / 2 of the actual round-trip time, and the quotient obtained is calibrated as the preset sound velocity parameter of the adhering material.

[0018] It is worth noting that in this embodiment, the ultrasonic echo data and the corresponding acquisition spatial coordinates are smoothly filled to obtain a continuous two-dimensional thickness scalar field. All the ultrasonic echo data and the acquisition spatial coordinates are extracted, the spatial hysteresis distance between any two acquisition spatial coordinates is calculated, and the semivariogram values ​​of the corresponding ultrasonic echo data are statistically analyzed to construct an empirical semivariogram function. Based on the least squares method, the empirical semivariogram function is fitted using a spherical model.

[0019] During the fitting process, the optimization objective is to minimize the sum of squared residuals between the empirical half-variable and the model prediction, iteratively determining the sill value, nugget constant, and range parameters. Within the physical boundary of the substrate defined by the global rectangular coordinate system, a spatially continuous two-dimensional function expression is constructed. For any unsampled spatial point within the physical boundary of the substrate, based on the spherical model and the distance between the unsampled spatial point and the surrounding known acquisition spatial coordinates, the ordinary Kriging interpolation algorithm is used to establish the spatial covariance matrix and the Lagrange multiplier equations, and the unbiased optimal interpolation weights are calculated. The ultrasonic echo data corresponding to the surrounding known acquisition spatial coordinates are weighted and summed with the unbiased optimal interpolation weights to generate a continuous thickness numerical mapping relationship covering the entire substrate surface. This continuous thickness numerical mapping relationship is defined as the continuous two-dimensional thickness scalar field.

[0020] In one implementation, this embodiment discretizes and samples the two-dimensional thickness scalar field to obtain an initial thickness distribution matrix. A preset resampling resolution is obtained; the physical boundary of the continuous two-dimensional thickness scalar field is divided into multiple two-dimensional rectangular grid cells of the same size according to the preset resampling resolution; the geometric center coordinates of each two-dimensional rectangular grid cell are extracted, and these coordinates are substituted into the two-dimensional function expression of the continuous two-dimensional thickness scalar field to calculate the corresponding discrete thickness representative value; according to the increasing row and column numbers of the two-dimensional rectangular grid cells in the physical boundary, all the discrete thickness representative values ​​are concatenated and combined to generate the initial thickness distribution matrix.

[0021] It should be noted that the preset resampling resolution is determined by extracting the minimum physical spatial distance between adjacent detection points in the acquisition spatial coordinates, multiplying the minimum physical spatial distance by 0.5, and using the product as the preset resampling resolution. The multiplication by 0.5 is based on the Nyquist spatial sampling theorem, which states that to avoid aliasing distortion of high-frequency spatial features and retain the true pollutant thickness topological details during discretization resampling, the grid resolution size must be less than or equal to half the original sampling distance. All the above calculations and interpolation logic are performed by a parallel computing matrix implemented using a hardware description language by a field-programmable gate array (FPGA) to ensure the real-time performance of large-scale scalar field processing.

[0022] For example, this embodiment uses an ultrasonic sensor array containing 100 units to scan a substrate with an area of ​​1 square meter, acquiring ultrasonic echo data with a thickness ranging from 0.1 mm to 5.0 mm. The minimum physical spatial distance between adjacent detection points in the array is 1.0 cm. This embodiment constructs an empirical semivariogram and fits it to a spherical model with a nugget constant of 0.05 and a sill value of 1.2. Interpolation weights are calculated to smoothly fill the discrete detection points into a spatially continuous thickness surface, i.e., a continuous two-dimensional thickness scalar field. Subsequently, this embodiment calculates a preset resampling resolution of 0.5 cm based on the minimum physical spatial distance. The continuous two-dimensional thickness scalar field is then divided into two-dimensional grids and discretized according to the resampling resolution of 0.5 cm to generate an initial thickness distribution matrix with 200 rows and 200 columns.

[0023] In step S2, the energy demand range of each zone is obtained by dividing the region according to the initial thickness distribution matrix and the preset thickness energy correlation model, including: Extract the thickness values ​​from the initial thickness distribution matrix; The thickness value is compared with a preset thickness range threshold to obtain low, medium, and high thickness regions; The energy demand range of each zone is obtained by mapping the preset energy density gradients corresponding to the low, medium, and high thickness regions.

[0024] In one implementation, the thickness values ​​are extracted from the initial thickness distribution matrix. Specifically, each discrete grid cell in the initial thickness distribution matrix is ​​traversed, and floating-point data stored at the corresponding row and column index addresses is read to obtain the pollutant thickness parameters corresponding to each spatial sampling point as the thickness value. Both the matrix traversal and data extraction operations are performed by the memory management unit of the industrial control microprocessor via direct memory access to reduce data throughput overhead.

[0025] It should be noted that the preset thickness range threshold includes a first thickness boundary value and a second thickness boundary value. In this embodiment, the extracted thickness value is compared logically with both the first and second thickness boundary values. If the thickness value is less than or equal to the first thickness boundary value, the corresponding mesh cell is marked as a low thickness category; if the thickness value is greater than the first thickness boundary value and less than or equal to the second thickness boundary value, it is marked as a medium thickness category; if the thickness value is greater than the second thickness boundary value, it is marked as a high thickness category. By performing a connected component extraction operation on spatially adjacent meshes with the same thickness category, the distribution range of the low, medium, and high thickness regions on the physical substrate surface is determined.

[0026] It is worth noting that the method for determining the preset thickness range threshold is as follows: Obtain the photothermal ablation characteristic curve of the contaminant material to be cleaned at a preset laser frequency; extract the derivative change characteristics of the ablation depth and energy conversion efficiency from the photothermal ablation characteristic curve; select two thickness characteristic values ​​where the efficiency decay exceeds a preset proportion, and label them as the first thickness boundary value and the second thickness boundary value, respectively. The preset proportion is determined through offline cleaning yield testing experiments, taking the efficiency decay inflection point that causes the substrate surface roughness to increase beyond a certain value. For example, when the substrate surface roughness is greater than... When the laser energy has been applied excessively to the substrate, the corresponding photothermal ablation efficiency attenuation is usually in the range of 15% to 25%. In this embodiment, the average value of this range is taken as the preset ratio.

[0027] It is worth noting that the extraction of the two thickness feature values ​​is achieved by calculating the first derivative of the photothermal ablation characteristic curve, taking the thickness corresponding to the point where the absolute value of the derivative is the largest as the first thickness boundary value, which represents the starting point of the fastest decline in energy conversion efficiency; then, starting from the first thickness boundary value, along the direction of thickness increase, the thickness corresponding to the first time when the efficiency decay reaches a preset proportion is found as the second thickness boundary value, for example, 20 percent; if the efficiency decay does not reach the preset proportion, the thickness at the end of the curve is taken as the second thickness boundary value.

[0028] In one implementation, the energy requirement range for each zone is obtained by mapping based on the preset energy density gradients corresponding to the low, medium, and high thickness regions. Specifically, the divided low, medium, and high thickness regions are input as independent variables into a preset thickness-energy correlation model. This preset thickness-energy correlation model stores linear regression mapping relationships between different thickness levels and laser energy output indicators. The construction process of the linear regression mapping relationship involves collecting 500 sets of offline cleaning experimental data for a specific material to be cleaned to construct a training set, using contaminant thickness as the independent variable and the optimal energy output required to achieve the preset cleanliness level as the dependent variable, and then using the least squares method to fit and obtain the linear regression equation:

[0029] in, For energy density, For thickness, To characterize the slope of energy conversion rate, The system compensates for the intercept of the system environment and extracts the slope and intercept of the linear regression equation. Based on the linear regression mapping relationship, the system assigns corresponding energy output intervals as the preset energy density gradients for the low, medium, and high thickness regions, and summarizes them to generate the partitioned energy demand range that includes the spatial location mapping relationship.

[0030] It should be noted that the method for determining the preset energy density gradient is as follows: The melting point temperature of the metal substrate and the latent heat of vaporization of surface contaminants are obtained; a heat conduction model is constructed based on the one-dimensional transient Fourier partial differential equation of heat conduction; the minimum laser radiation energy required to completely peel off the target thickness of contaminants per unit area is calculated; simultaneously, the limiting energy value that causes irreversible thermal damage to the substrate surface is extracted as a safety upper limit. Within the numerical range formed by the minimum laser radiation energy and the safety upper limit, the range is discretized by combining the rated power output level of the laser to obtain the preset energy density gradient.

[0031] For example, this embodiment extracts thickness values ​​ranging from 0.1 mm to 5.0 mm from the initial thickness distribution matrix. Based on calibrated thickness interval thresholds, the 0.1 to 1.5 mm interval is defined as a low-thickness region, the 1.5 to 3.0 mm interval as a medium-thickness region, and the 3.0 to 5.0 mm interval as a high-thickness region. Mapping is performed using a thickness-energy correlation model to determine the partitioned energy demand range for the low-thickness region as 10 to 20 joules / cm², for the medium-thickness region as 20 to 35 joules / cm², and for the high-thickness region as 35 to 50 joules / cm². The above energy range data is correlated with the region segmentation coordinates to generate a digital mask matrix for precise matching of subsequent cleaning power. All the above region segmentation and mapping logic is deployed in a field-programmable gate array (FPGA), and the mask matrix is ​​updated in real time through hardware logic circuits.

[0032] In step S3, the regional energy fluctuation parameter is obtained by performing amplitude statistical evaluation based on the energy demand range of the partition. The regional energy fluctuation parameter is compared with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates. The high-risk physical areas and boundary coordinates are then weighted and sorted to obtain the high-risk area adjustment sequence.

[0033] Among them, the regional energy fluctuation parameters are obtained by performing amplitude statistical evaluation based on the energy demand range of the partition, including: The standard deviation and coefficient of variation of energy demand for each region are obtained by statistical deviation calculation based on the energy demand range of the partition. By performing weighted statistics based on the standard deviation of energy demand and the coefficient of variation, a regional energy fluctuation parameter reflecting the degree of discrete distribution of energy demand in the corresponding region is obtained.

[0034] In one implementation, the standard deviation and coefficient of variation of energy demand for each region are obtained by statistical deviation calculation based on the energy demand range of the partition. This embodiment extracts the energy demand values ​​corresponding to all discrete grid cells within the low, medium, and high thickness regions; for each thickness region, the arithmetic mean of all energy demand values ​​within that region is calculated; the square of the difference between each energy demand value within that region and the arithmetic mean is extracted, and the arithmetic mean of all squared values ​​is calculated and then the square root is taken to obtain the standard deviation of energy demand; the standard deviation of energy demand is divided by the arithmetic mean, and the quotient is determined as the coefficient of variation. The above statistical deviation calculation process is executed in the floating-point unit of the industrial control microprocessor, completing the physical calculation of the matrix data distribution index.

[0035] It is worth noting that this embodiment uses weighted statistics based on the standard deviation of energy demand and the coefficient of variation to obtain a regional energy fluctuation parameter reflecting the degree of dispersion of energy demand in the corresponding region. Specifically, a preset standard deviation transformation coefficient and a preset variation weight coefficient are extracted; the standard deviation of energy demand is multiplied by the standard deviation transformation coefficient, and then the coefficient of variation is multiplied by the variation weight coefficient; the accumulated result is determined as the regional energy fluctuation parameter. The preset standard deviation transformation coefficient and the preset variation weight coefficient are determined by obtaining an offline cleaning process dataset of the substrate and extracting the variance of the surface residue thickness after cleaning as the dependent variable. The corresponding historical energy demand standard deviation was extracted as the first independent variable. Historical coefficient of variation was extracted as the second independent variable. Construct a multiple linear regression equation:

[0036] in, The intercept is... and These are partial regression coefficients. The random error term is used; the multiple linear regression equation is solved using the least squares method, and the two partial regression coefficients obtained are... and Perform proportional summation and normalization processing, that is, divide the absolute value of each partial regression coefficient by the sum of the absolute values ​​of the two partial regression coefficients, and label the two dimensionless values ​​obtained as the preset standard deviation transformation coefficient and the preset variation weight coefficient, respectively.

[0037] It should be noted that in this embodiment, the regional energy fluctuation parameter is compared with a preset fluctuation safety threshold to filter out high-risk physical regions and boundary coordinates. It is determined whether the regional energy fluctuation parameter is greater than the preset fluctuation safety threshold; if it is less than or equal to the preset fluctuation safety threshold, the corresponding thickness region is marked as a normal safe region, and the initial resource scheduling priority of that region is maintained; if it is greater, the thickness region corresponding to the regional energy fluctuation parameter is marked as a high-risk physical region; the set of geometric center coordinates of the outermost grid cells of the high-risk physical region is extracted, and the set of geometric center coordinates is determined as the boundary coordinates.

[0038] The method for determining the preset fluctuation safety threshold is as follows: a normally distributed fluctuation laser energy pulse matrix is ​​continuously applied to a test sample of the same material as the metal substrate to be cleaned; the critical fluctuation parameter value when the microscopic thermal ablation pool first appears on the surface of the test sample is recorded; the critical fluctuation parameter value is multiplied by a preset safety margin coefficient, and the product is calibrated as the preset fluctuation safety threshold. The preset safety margin coefficient is determined based on the following: the output of the underlying hardware laser generator has a background energy fluctuation error of 5% to 10%, and there is physical hysteresis in the heat conduction process; the thermal accumulation buffer space tolerance is calibrated based on thermodynamic finite element simulation analysis, and the value of the safety margin coefficient is determined to be 0.8.

[0039] The normally distributed wave laser energy pulse matrix is ​​generated by setting the nominal energy density of the laser output and randomly generating the energy density of each pulse according to a normal distribution with a mean of 10% of the nominal value and a standard deviation of 10% of the nominal value, within a range of 70% to 130% of the nominal value. At the same time, the pulses are arranged in a two-dimensional grid with no less than 20 rows and columns, and each grid point corresponds to a random energy pulse. The pulse matrix is ​​applied to the entire surface of the test sample, and the surface thermal effect is monitored in real time using an infrared thermal imager. The standard deviation of the pulse energy density corresponding to the first observation of the microscopic thermal ablation molten pool is the critical wave parameter.

[0040] The high-risk area adjustment sequence is obtained by weighting and sorting the high-risk physical areas and their boundary coordinates, including: The first risk score is obtained by multiplying the regional energy fluctuation parameters according to the preset fluctuation amplitude weighting coefficient. The second risk score is calculated by multiplying the coverage area of ​​the high-risk physical area according to the preset regional proportion weighting coefficient. The high-risk area adjustment sequence is obtained by summing and sorting the first risk score and the second risk score.

[0041] In one implementation, a first risk score is obtained by multiplying the regional energy fluctuation parameter according to a preset fluctuation amplitude weighting coefficient, and a second risk score is obtained by multiplying the coverage area of ​​the high-risk physical region according to a preset regional proportion weighting coefficient. Based on the difference in physical dimensions between the energy fluctuation parameter and the coverage area, this embodiment first uses a range standardization method to map the regional energy fluctuation parameter and the coverage area of ​​all regions to a dimensionless interval of 0 to 1, obtaining normalized fluctuation parameters and normalized coverage areas. Subsequently, the preset fluctuation amplitude weighting coefficient is multiplied by the normalized fluctuation parameter, and the resulting scalar value is determined as the first risk score; the preset regional proportion weighting coefficient is multiplied by the normalized coverage area, and the resulting scalar value is determined as the second risk score.

[0042] It should be noted that the calculation of the coverage area of ​​the high-risk physical area is performed by counting the total number of grid cells contained in the high-risk physical area and multiplying the total number of grid cells by the physical area constant of a single grid cell.

[0043] It should be noted that the method for determining the preset fluctuation amplitude weight coefficient and the preset area proportion weight coefficient is as follows: A historical cleaning defect judgment database is acquired. The construction method of the historical cleaning defect judgment database involves collecting at least 1000 sets of historical post-cleaning surface images of substrates of the same material, performing adaptive threshold image segmentation on the historical post-cleaning surface images using the Otsu method (Otsu's maximum inter-class variance method), and combining this with a connected component area statistics algorithm to extract the surface residue or thermal damage area as the actual cleaning defect area for labeling and storing it in a relational database; extracting historical fluctuation parameters and historical area proportion data corresponding to the unqualified cleaning areas from the historical cleaning defect judgment database; calculating the correlation coefficient between the historical fluctuation parameters and the actual cleaning defect area, and the correlation coefficient between the historical area proportion data and the actual cleaning defect area, respectively, using the Pearson correlation coefficient method; dividing the two correlation coefficients by their arithmetic sum, and assigning the two quotients as the preset fluctuation amplitude weight coefficient and the preset area proportion weight coefficient, respectively.

[0044] It is worth noting that the high-risk area adjustment sequence is obtained by accumulating and sorting the first risk score and the second risk score. In this embodiment, the first risk score (after eliminating the dimension difference) is added to the second risk score to calculate a dimensionless comprehensive risk score. In the case of multiple high-risk physical areas, the comprehensive risk scores corresponding to all high-risk physical areas are compared. The comprehensive risk scores are arranged in descending order to generate a sequence containing high-risk physical area identifiers, which is then determined as the high-risk area adjustment sequence. The above sorting results and mask marking instructions are written into the shared memory controlled by the field-programmable gate array (FPGA) as input for the subsequent cleaning resource coordination and scheduling module.

[0045] For example, this embodiment calculates that the energy demand fluctuation amplitude in the high-thickness region is relatively large, and after statistical evaluation, its regional energy fluctuation parameter is 5.2 joules / square centimeter. The critical fluctuation parameter value recorded in advance through destructive testing samples is 3.75 joules / square centimeter. Combined with a safety margin coefficient of 0.8 based on the hardware background fluctuation, the preset fluctuation safety threshold is calibrated to be 3.0 joules / square centimeter. After comparison, 5.2 is greater than 3.0, and the high-thickness region is marked as a high-risk physical region, and the corresponding two-dimensional boundary coordinates are extracted. After mapping the fluctuation parameter and coverage area of ​​each high-risk physical region to the dimensionless interval of 0 to 1 using the range standardization method, the normalized fluctuation parameter of the high-thickness region is 0.85, and the normalized coverage area is 0.60. The system pre-calculates the calibrated fluctuation amplitude weight coefficient to 0.6 and the regional proportion weight coefficient to 0.4 using the Pearson correlation coefficient method. The high-risk physical area has a first risk score of 0.51 (0.85 multiplied by 0.6) and a second risk score of 0.24 (0.60 multiplied by 0.4). Adding 0.24 to 0.51 yields a comprehensive risk score of 0.75. The system uses the same algorithm to simultaneously calculate the comprehensive risk score for medium-thickness areas (0.45) and low-thickness areas (0.21). These scores are then arranged in descending order of value, generating a high-risk area adjustment sequence with the high-thickness area at the top.

[0046] In step S4, cleaning resources are comprehensively scheduled based on the high-risk area adjustment sequence and the preset total power constraint of the equipment hardware to obtain a basic global cleaning plan including initial laser drive parameters, including: The target energy density parameter is obtained by mapping and matching the high-risk area adjustment sequence with the partition energy demand range. The energy matching coefficient is obtained by solving the constraint solution based on the target energy density parameter and the preset total power constraint of the device hardware. The initial laser driving parameters are obtained by allocating energy according to the energy matching coefficient. Based on the initial laser driving parameters, a node search planning is performed to obtain a basic global cleaning plan that includes the initial physical execution path and the initial laser driving parameters.

[0047] In one implementation, the target energy density parameter is obtained by mapping and matching the high-risk area adjustment sequence with the partitioned energy demand range. In this embodiment, physical region identifiers arranged in the high-risk area adjustment sequence are extracted sequentially. These physical region identifiers are used as query indexes to perform position comparison retrieval in the mask matrix containing the partitioned energy demand range. The maximum value within the boundary of the energy demand range corresponding to the retrieved physical region is extracted, and this maximum value is determined as the target energy density parameter corresponding to that physical region. The above scheduling and mapping logic is deployed in a field-programmable gate array (FPGA) and interacts with the laser generator driver board via an industrial bus.

[0048] It should be noted that the method for determining the preset total power constraint of the equipment hardware is as follows: read the hardware nameplate data parameters of the laser generator in the cleaning equipment, extract the maximum rated output power value that allows continuous safe operation, and directly set the maximum rated output power value as the preset total power constraint of the equipment hardware.

[0049] It is worth noting that in this embodiment, the energy matching coefficient is obtained by constraining the target energy density parameter and the preset total power constraint of the equipment hardware. Specifically, the target energy density parameter of each physical region is multiplied by its corresponding coverage area, and an accumulation operation is performed to obtain the global theoretical total energy demand. The difference between the global theoretical total energy demand and the preset total power constraint of the equipment hardware is compared. If the global theoretical total energy demand is less than or equal to the preset total power constraint of the equipment hardware, the energy matching coefficient of all physical regions is assigned a constant of 1. If the global theoretical total energy demand is greater than the preset total power constraint of the equipment hardware, a linear programming equation is constructed with maximizing the energy satisfaction rate of high-priority regions as the objective function. When constructing the linear programming equation, a bottom-level physical baseline constraint is introduced, that is, the actual allocated energy density of each region must be greater than or equal to the minimum laser radiation energy required to completely remove contaminants of the corresponding thickness. The system uses the simplex method to solve the constraints and outputs the proportional coefficients between 0 and 1 for each physical region, which are then determined as the energy matching coefficients. If the linear programming equation becomes unsolvable globally due to extremely stringent total power constraints, the system automatically generates a global physical deceleration command to proportionally relax the power constraint limits.

[0050] In one implementation, initial laser driving parameters are obtained by allocating energy according to the energy matching coefficient. In this embodiment, the target energy density parameter of each physical region is multiplied by its corresponding energy matching coefficient to obtain the actual allocated energy density. Using a pre-built energy driving mapping table, the underlying hardware control instructions are retrieved and matched according to the actual allocated energy density. The corresponding pulse width modulation duty cycle and pulse repetition frequency values ​​are extracted, and the pulse width modulation duty cycle and pulse repetition frequency values ​​are combined to generate the initial laser driving parameters.

[0051] It should be noted that the method for constructing the pre-built energy-driven mapping table is as follows: In offline testing, the laser generator is driven to traverse different combinations of pulse width modulation duty cycle and pulse repetition frequency at a preset step size. A laser power meter is used to continuously measure and record the actual radiation energy density value reaching the metal substrate surface for each output combination at a fixed working focal length. The actual radiation energy density value is used as the key, and the corresponding combination of pulse width modulation duty cycle and pulse repetition frequency value is used as the value to construct a one-to-one data retrieval dictionary, which is then stored in memory to generate the pre-built energy-driven mapping table.

[0052] It is worth noting that this embodiment uses the initial laser driving parameters to perform node search planning to obtain a basic global cleaning plan that includes the initial physical execution path and the initial laser driving parameters. Specifically, the A-heuristic search algorithm is used to perform connection planning on the spatial nodes to be traversed extracted from the discrete grid cells. To eliminate dimensional differences, this embodiment first uses the extreme value method to normalize the actual Euclidean distance between two points to a dimensionless interval of 0 to 1 to obtain the spatial distance cost; simultaneously, the normalized weighted sum of the absolute values ​​of the difference in pulse width modulation duty cycle and the absolute values ​​of the difference in pulse repetition frequency between adjacent nodes is calculated to obtain the initial laser driving parameter switching cost. The comprehensive cost function of the A-heuristic search algorithm is defined as the weighted sum of the spatial distance cost and the initial laser driving parameter switching cost. The ratio of the weighting coefficient of the spatial distance cost to the weighting coefficient of the initial laser drive parameter switching cost is set to 1:20. The physical basis for this is that the electrical response and stabilization time of the laser generator's underlying hardware to duty cycle or frequency jumps is far greater than the mechanical time required for the actuator to perform millimeter-level spatial displacement. Giving the switching cost a higher penalty weight effectively avoids hardware overload and cleaning artifacts caused by high-frequency jumps. In this embodiment, the adjacent node with the lowest overall cost is selected successively as the next row's coordinate until all mesh cells are traversed, generating the spatially continuous initial physical execution path.

[0053] For example, this embodiment solves the linear programming equation using the simplex method. While ensuring that the energy in all regions meets the minimum stripping threshold, it prioritizes high-risk areas, resulting in an energy matching coefficient of 1 for the high-risk areas. Subsequently, The heuristic search algorithm starts planning from the high-risk zone. By setting a weight ratio of 1:20, the algorithm is extremely averse to frequent changes in laser parameters, thus actively avoiding the backtracking path with frequent changes in driving parameters, and generating an initial physical execution path with a total length of 1.2 meters, an estimated execution time of 15 seconds, and smooth parameter transitions.

[0054] In step S5, the surface infrared thermal image signal and spectral reflection signal during the physical cleaning process are obtained according to the basic global cleaning plan, and the surface infrared thermal image signal and the spectral reflection signal are quantized by physical features to obtain real-time surface state features.

[0055] Specifically, the real-time surface state characteristics are obtained by physical feature quantization of the surface infrared thermal image signal and the spectral reflectance signal, including: The surface temperature uniformity index is obtained by performing spatial distribution statistics on the infrared thermal image signal of the surface. The real-time surface reflectance is obtained by fitting the temporal characteristics of the spectral reflectance signal; The real-time surface state features are obtained by feature splicing of the surface temperature uniformity index and the real-time surface reflectivity.

[0056] In one implementation, surface infrared thermal imaging signals and spectral reflectance signals are acquired during the physical cleaning process according to the basic global cleaning plan. This embodiment analyzes the physical execution path coordinates and corresponding time axis sequences in the basic global cleaning plan. An industrial control microprocessor (MCU), via an industrial real-time Ethernet bus, such as EtherCAT, synchronously sends a hard trigger signal to a servo infrared thermal imager and a hyperspectral analyzer coaxially mounted with the laser head when the underlying actuator reaches the physical execution path coordinates. The high-speed acquisition interface of the field-programmable gate array (FPGA) reads the two-dimensional thermal radiation matrix sequence captured by the servo infrared thermal imager on the corresponding substrate surface as the surface infrared thermal imaging signal; simultaneously, it reads the echo intensity array acquired by the hyperspectral analyzer within a specific laser wavelength band as the spectral reflectance signal.

[0057] It is worth noting that in this embodiment, the surface temperature uniformity index is obtained by performing spatial distribution statistics on the surface infrared thermal image signal. The physical temperature values ​​corresponding to all pixels in a single frame of the surface infrared thermal image signal are extracted. In the floating-point arithmetic unit of the industrial control microprocessor, the arithmetic mean of all physical temperature values ​​in the single frame of the surface infrared thermal image signal is calculated. This arithmetic mean is synchronously represented as the real-time regional temperature and is encapsulated in subsequent features for over-temperature early warning determination. The standard deviation is also calculated. The standard deviation is divided by the arithmetic mean to obtain the coefficient of variation scalar, and the coefficient of variation scalar is determined as the surface temperature uniformity index.

[0058] In one implementation, the real-time surface reflectance is obtained by fitting the spectral reflectance signal with time-series features. This embodiment extracts the continuous observation amplitude sequence of the spectral reflectance signal within a preset time window to construct a one-dimensional time series. The Savitzky-Gore smoothing algorithm is used to perform local polynomial least squares fitting on the one-dimensional time series. Specifically, a local quadratic polynomial fitting equation is constructed:

[0059] in, This represents the fitted reflected light intensity value. This is the relative time index within the sliding window. The polynomial coefficients are to be determined. The optimal polynomial coefficients under the current sliding window are obtained by minimizing the sum of squared residuals between the actual continuous observed amplitudes and the output values ​​of the fitted equation using the least squares method. The current time point after fitting is then extracted. The constant term corresponding to the center point As a steady-state value of reflected light intensity, the steady-state value of reflected light intensity is mapped and determined as the real-time surface reflectivity.

[0060] It should be noted that the method for determining the preset time window is as follows: The duration of a single laser pulse from the bottom cleaning device is obtained; the duration of the single laser pulse is multiplied by a preset pulse period multiple, and the resulting product is defined as the preset time window. The preset pulse period multiple is determined based on the physical delay between the thermal response and photo-induced plasma flash caused by a single pulse. Through offline hyperspectral sampling experiments, it was determined that 10 consecutive pulse periods can precisely cover a complete local heat accumulation and dissipation cycle, thus accurately filtering out transient interference and capturing steady-state reflection characteristics. Therefore, the preset pulse period multiple is set to 10. The Savitzky-Gore smoothing algorithm extracts a preset odd number of consecutive sampling points as sliding window parameters; the selection of the preset odd number of consecutive sampling points is based on the product of the hyperspectral analyzer's sampling frequency and the preset time window being closest to an odd number.

[0061] It is worth noting that in this embodiment, the surface temperature uniformity index and the real-time surface reflectivity are concatenated to obtain the real-time surface state features. Specifically, the following steps are taken: the current absolute timestamp when the surface infrared thermal image signal and the spectral reflectance signal are captured are extracted, and the corresponding current spatial physical coordinates are extracted; in the shared memory of the FPGA, the current absolute timestamp, the current spatial physical coordinates, the surface temperature uniformity index, the real-time surface reflectivity, and the arithmetic mean calculated in the previous steps are concatenated according to a predefined data structure protocol to generate a one-dimensional feature vector containing spatiotemporal features and multimodal physical parameters, and the one-dimensional feature vector is determined as the real-time surface state feature.

[0062] For example, in this embodiment, during the physical cleaning process, the system analyzes the basic global cleaning plan and drives the sensors to collect data synchronously. A servo infrared thermal imager captures the two-dimensional thermal radiation matrix of the target area. The arithmetic mean temperature is calculated to be 30.2 degrees Celsius, and the temperature standard deviation is 1.11 degrees Celsius. Dividing 1.11 by 30.2 yields a surface temperature uniformity index of approximately 0.037. Simultaneously, a hyperspectral analyzer collects a sequence of echo light intensity within a preset time window. This preset time window is calculated to be 20 milliseconds based on the pulse period multiple of 10 (2 milliseconds for a single pulse duration). This window length precisely covers the complete local heat accumulation and dissipation cycle. After filtering out transient noise using the Savitzky-Gore smoothing algorithm, which includes quadratic polynomial equations, the coefficients of the constant term are solved. The system fits the steady-state light intensity value at the current moment and maps it to obtain a real-time surface reflectivity of 0.53. Subsequently, the system performs multi-dimensional feature concatenation on the current spatial coordinates, absolute timestamp, and scalar values ​​of 30.2 degrees Celsius, 0.037, and 0.53, and outputs a one-dimensional vector of real-time surface state features that includes comprehensive evaluation attributes.

[0063] In step S6, a local over-temperature warning command is obtained by comparing the real-time surface state characteristics with a preset substrate damage temperature threshold. Based on the local over-temperature warning command, the initial laser driving parameters are adaptively calibrated to reduce energy and obtain a safe laser output command and an updated physical execution path.

[0064] The process of adaptively reducing the energy of the initial laser drive parameters based on the local over-temperature warning command to obtain a safe laser output command and an updated physical execution path includes: Extract the actual temperature value corresponding to the local over-temperature warning command, and calculate the temperature difference by subtracting the actual temperature value from the preset substrate damage temperature threshold. Using a proportional-integral-derivative (PID) control algorithm, a safe laser output command is obtained by dynamically adjusting the laser duty cycle or pulse frequency in the initial laser drive parameters based on the temperature difference. The energy response time is obtained by performing time delay analysis based on the safe laser output command, and the updated physical execution path is obtained by performing scan speed compensation calculation on the initial physical execution path based on the energy response time.

[0065] In one implementation, this embodiment analyzes the real-time surface state features and extracts the real-time physical temperature value characterizing the thermodynamic state of the region. In the logic operation unit of the industrial control microprocessor (MCU), the real-time physical temperature value is compared with a preset substrate damage temperature threshold. If the real-time physical temperature value is less than or equal to the preset substrate damage temperature threshold, the current physical cleaning state is determined to be safe, and the cleaning task continues to be executed according to the initial physical execution path and the initial laser driving parameters. If the real-time physical temperature value is greater than the preset substrate damage temperature threshold, a local over-temperature warning command containing the coordinates exceeding the threshold and the real-time physical temperature value is generated. Further, the absolute temperature scalar in the local over-temperature warning command is extracted as the actual temperature value, and the positive difference obtained by subtracting the preset substrate damage temperature threshold from the actual temperature value is calibrated as the temperature difference value.

[0066] It should be noted that the method for determining the preset substrate damage temperature threshold is as follows: Gradient heating destructive testing is performed on standard metal substrates from the same batch. High-precision thermocouples are used to record the critical temperature values ​​at which microscopic lattice phase transitions or irreversible thermodynamic deformation occur on the substrate surface. This critical temperature value is multiplied by a redundancy factor of 0.9, and the resulting product is calibrated as the preset substrate damage temperature threshold. The basis for setting this 10% redundancy factor is that, based on the thermodynamic conduction delay characteristic in the thermal diffusion equation, heat has already accumulated inside the substrate when the temperature measurement module detects the surface temperature. This safety margin is specifically used to absorb the residual heat inertia during the period from triggering the warning to the energy reduction command taking effect, thereby achieving proactive intervention.

[0067] It is worth noting that this embodiment utilizes a proportional-integral-derivative (PID) control algorithm to dynamically adjust the laser duty cycle or pulse frequency in the initial laser drive parameters based on the temperature difference to obtain a safe laser output command. The temperature difference is used as the input deviation; preset proportional coefficients, preset integral coefficients, and preset derivative coefficients are extracted; and proportional, integral, and derivative operations are performed and summed in the closed-loop control module of the field-programmable gate array (FPGA). The algorithm output is defined as a normalized energy reduction ratio.

[0068] Subsequently, based on the energy reduction ratio coefficient and combined with the underlying hardware control mapping relationship, priority energy reduction mapping is performed. The energy reduction ratio coefficient is first mapped to the percentage reduction range of the current laser duty cycle to calculate the actual physical reduction amount. If the reduced laser duty cycle is greater than or equal to the preset lower limit of the laser hardware maintenance duty cycle, such as 10%, the physical reduction amount is directly used to reduce the duty cycle in the initial laser drive parameters individually. If the reduced laser duty cycle is lower than the preset lower limit of the laser hardware maintenance duty cycle, the lower limit of the duty cycle is locked, and the remaining energy reduction ratio coefficient is mapped to the Hertz reduction range of the pulse frequency to calculate the frequency reduction amount. The corresponding physical reduction amount is subtracted from the initial laser drive parameters to generate a safe laser output command.

[0069] It should be noted that the basis for setting the duty cycle lower limit to 10% is that when the pulse width modulation duty cycle is lower than this physical threshold, the pump source energy will not be able to maintain the population inversion and stable laser plasma excitation in the resonant cavity, which will lead to laser output interruption and nonlinear energy reduction. Therefore, it is necessary to lock the bottom line.

[0070] It should be noted that the preset proportional coefficient, preset integral coefficient, and preset differential coefficient are determined by the Ziegler-Nichols rule, based on the algebraic calculation and fine-tuning of the critical oscillation period and critical proportional gain measured during the offline initialization phase of the system, to ensure that the attenuation ratio is stable at 1 / 4.

[0071] In one implementation, the energy response time is obtained by performing delay analysis based on the safe laser output command. This embodiment reads the data packet bit length of the safe laser output command and the transmission baud rate of the underlying industrial real-time Ethernet bus, and calculates the command transmission delay by dividing the data packet bit length by the transmission baud rate; it then obtains the pre-calibrated capacitor discharge time constant of the laser generator driver motherboard; it adds the command transmission delay to the capacitor discharge time constant to obtain the energy response time reflecting the actual lag in the underlying hardware operation; the capacitor discharge time constant is obtained by connecting the output terminal of the laser generator driver motherboard to a standard resistive load, measuring the falling edge waveform of the drive signal using an oscilloscope probe, inputting a single-pulse width modulation falling edge trigger signal to the motherboard, and measuring the time difference during which the output voltage drops from 90% to 37% of its rated value. This time difference is the capacitor discharge time constant. This measurement is repeated ten times, and the arithmetic mean is taken as the final calibration value.

[0072] It is worth noting that in this embodiment, the updated physical execution path is obtained by performing scan speed compensation calculation on the initial physical execution path based on the energy response time. Specifically, the target radiation energy density after energy reduction is extracted according to the safe laser output command; the target radiation energy density is divided by the initial radiation energy density corresponding to the initial laser driving parameters to obtain the energy reduction ratio coefficient. In this embodiment, the initial scan speed configured in the initial physical execution path is multiplied by the energy reduction ratio coefficient to obtain the accelerated compensated scan speed; the initial intervention coordinates are delayed by a physical buffer distance along a predetermined trajectory, the physical buffer distance being equal to the initial scan speed multiplied by the energy response time; the accelerated compensated scan speed is used to overwrite the path speed attributes of all associated spatial nodes after the physical buffer distance delay, generating the updated physical execution path containing global new speed configuration information.

[0073] For example, in this embodiment, the actual temperature value detected is 39.2 degrees Celsius, and the system calculates the temperature difference to be 1.2 degrees Celsius. The industrial control microprocessor calls the control algorithm to calculate a normalized energy reduction ratio coefficient of 0.2. Based on this ratio coefficient, the system adjusts the initial laser drive parameters, dynamically reducing the laser energy density from 2.5 joules / cm² to 2.0 joules / cm². The system further analyzes the bus transmission and hardware discharge time, obtaining a true energy response time of 8 milliseconds. The system calculates the energy reduction ratio coefficient to be 0.8, which is 2.0 divided by 2.5. To quickly pass through the overheated zone and prevent heat accumulation caused by slow speed, the system multiplies the initial physical execution path scanning speed of 0.1 m / s by 0.8, calculating a compensated scanning speed of 0.08 m / s after speed-up. Combining an 8-millisecond energy response time and a 0.8-millimeter physical buffer distance caused by an initial velocity of 0.1 m / s, the system reconstructs and overwrites the velocity attributes of subsequent nodes starting 0.8 mm behind the current coordinates, and performs a lower-power cleaning operation at a reduced velocity of 0.08 m / s, thereby achieving perfect thermal damage prevention and physical-level hard-core closed-loop protection within milliseconds.

[0074] It is worth noting that the 38°C substrate damage temperature threshold used in the above example is merely an illustrative low value set for ease of understanding the technical principle and does not constitute an actual limitation on the present invention. For metal substrates in actual engineering applications, the substrate damage temperature threshold should be determined according to the thermodynamic properties of the specific material. For example, for ordinary carbon steel, the threshold can be set to 550°C to 650°C; for aluminum alloys, it can be set to 250°C to 350°C. Any damage temperature threshold corresponding to any actual material used based on the closed-loop energy reduction calibration principle of the present invention falls within the protection scope of the present invention.

[0075] In summary, this invention generates an initial thickness distribution matrix by interpolating ultrasonic echo data from the surface of the metal substrate to be cleaned. It then accurately screens high-risk physical regions using a thickness-energy correlation model and amplitude statistical evaluation mechanism. Under the constraint of total hardware power, it feeds forward to coordinate and schedule a basic global cleaning plan including initial laser drive parameters. Furthermore, during the physical cleaning execution phase, this invention integrates surface infrared thermography and spectral reflectance signals to quantify real-time surface state characteristics, constructing a low-level adaptive energy reduction calibration mechanism based on a proportional-integral-derivative control algorithm. Simultaneously, it compensates for the scanning speed of the physical execution path by incorporating hardware energy response time. This invention overcomes the limitations of the static energy allocation and the invisibility of process states in traditional laser cleaning, constructing a complete automated control chain from macroscopic feedforward resource planning to microscopic millisecond-level physical closed-loop intervention. Ultimately, under complex conditions of heterogeneous contaminant thickness distribution, it achieves high-precision dynamic allocation of laser cleaning energy and absolute physical blocking of instantaneous thermal overload risks, maximizing stripping efficiency while ensuring spatial consistency of cleaning quality and ultimate safety of the metal substrate.

[0076] Reference Figure 2 The second embodiment of the present invention provides a closed-loop automated control system for laser cleaning, comprising: The initial distribution interpolation module is used to acquire ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned, and to perform spatial interpolation on the ultrasonic echo data based on the acquisition spatial coordinates to obtain an initial thickness distribution matrix. The regional energy segmentation module is used to segment regions based on the initial thickness distribution matrix and a preset thickness energy correlation model to obtain the energy demand range of each region. The risk area screening module is used to perform amplitude statistical evaluation based on the energy demand range of the partition to obtain the regional energy fluctuation parameter, compare the regional energy fluctuation parameter with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates, and perform weighted sorting based on the high-risk physical areas and boundary coordinates to obtain the high-risk area adjustment sequence. The resource scheduling and planning module is used to coordinate and schedule cleaning resources according to the high-risk area adjustment sequence and the preset total power constraints of equipment hardware, so as to obtain a basic global cleaning plan including initial laser driving parameters. The real-time state quantization module is used to acquire surface infrared thermal image signals and spectral reflection signals during the physical cleaning process according to the basic global cleaning plan, and to perform physical feature quantization on the surface infrared thermal image signals and the spectral reflection signals to obtain real-time surface state features. The energy reduction calibration update module is used to compare the real-time surface state characteristics with the preset substrate damage temperature threshold to obtain a local over-temperature warning command, and to perform adaptive energy reduction calibration on the initial laser driving parameters based on the local over-temperature warning command to obtain a safe laser output command and an updated physical execution path.

[0077] It should be noted that the laser cleaning closed-loop automated control system provided in this embodiment of the invention is used to execute all process steps of the laser cleaning closed-loop automated control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0078] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0079] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A closed-loop automated control method for laser cleaning, characterized in that, include: Acquire ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned, and perform spatial interpolation on the ultrasonic echo data based on the acquisition spatial coordinates to obtain an initial thickness distribution matrix; The energy demand range of each zone is obtained by dividing the region according to the initial thickness distribution matrix and the preset thickness energy correlation model; The regional energy fluctuation parameter is obtained by statistically evaluating the amplitude of the energy demand range of the partition. The regional energy fluctuation parameter is compared with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates. The high-risk physical areas and boundary coordinates are weighted and sorted to obtain the high-risk area adjustment sequence. Based on the high-risk zone adjustment sequence and the preset total power constraint of the equipment hardware, the cleaning resources are coordinated and scheduled to obtain a basic global cleaning plan including the initial laser driving parameters. Based on the basic global cleaning plan, surface infrared thermal imaging signals and spectral reflection signals are obtained during the physical cleaning process. The surface infrared thermal imaging signals and spectral reflection signals are then quantized using physical features to obtain real-time surface state features. A local over-temperature warning command is obtained by comparing the real-time surface state characteristics with a preset substrate damage temperature threshold. Based on the local over-temperature warning command, the initial laser driving parameters are adaptively calibrated to reduce energy and obtain a safe laser output command and an updated physical execution path.

2. The laser cleaning closed-loop automated control method according to claim 1, characterized in that, The step of spatially interpolating the ultrasonic echo data based on the acquired spatial coordinates to obtain the initial thickness distribution matrix includes: The ultrasonic echo data and the corresponding acquisition spatial coordinates are smoothed and filled to obtain a continuous two-dimensional thickness scalar field. The initial thickness distribution matrix is ​​obtained by discretizing and sampling the two-dimensional thickness scalar field.

3. The laser cleaning closed-loop automated control method according to claim 1, characterized in that, The step of dividing the region into energy demand ranges based on the initial thickness distribution matrix and a preset thickness-energy correlation model includes: Extract the thickness values ​​from the initial thickness distribution matrix; The thickness value is compared with a preset thickness range threshold to obtain low, medium, and high thickness regions; The energy demand range of each zone is obtained by mapping the preset energy density gradients corresponding to the low, medium, and high thickness regions.

4. The laser cleaning closed-loop automated control method according to claim 1, characterized in that, The process of obtaining regional energy fluctuation parameters by performing amplitude statistical evaluation based on the energy demand range of the partition includes: The standard deviation and coefficient of variation of energy demand for each region are obtained by statistical deviation calculation based on the energy demand range of the partition. By performing weighted statistics based on the standard deviation of energy demand and the coefficient of variation, a regional energy fluctuation parameter reflecting the degree of discrete distribution of energy demand in the corresponding region is obtained.

5. The laser cleaning closed-loop automated control method according to claim 1, characterized in that, The step of obtaining the high-risk area adjustment sequence by weighted sorting based on the high-risk physical areas and boundary coordinates includes: The first risk score is obtained by multiplying the regional energy fluctuation parameters according to the preset fluctuation amplitude weighting coefficient. The second risk score is calculated by multiplying the coverage area of ​​the high-risk physical area according to the preset regional proportion weighting coefficient. The high-risk area adjustment sequence is obtained by summing and sorting the first risk score and the second risk score.

6. The laser cleaning closed-loop automated control method according to claim 1, characterized in that, The step of adjusting the sequence based on the high-risk area and the preset total power constraint of the equipment hardware to coordinate and schedule cleaning resources yields a basic global cleaning plan including initial laser drive parameters, including: The target energy density parameter is obtained by mapping and matching the high-risk area adjustment sequence with the partition energy demand range. The energy matching coefficient is obtained by solving the constraint solution based on the target energy density parameter and the preset total power constraint of the device hardware. The initial laser driving parameters are obtained by allocating energy according to the energy matching coefficient. Based on the initial laser driving parameters, a node search planning is performed to obtain a basic global cleaning plan that includes the initial physical execution path and the initial laser driving parameters.

7. The laser cleaning closed-loop automated control method according to claim 1, characterized in that, The step of quantizing the surface infrared thermal image signal and the spectral reflectance signal to obtain real-time surface state features includes: The surface temperature uniformity index is obtained by performing spatial distribution statistics on the infrared thermal image signal of the surface. The real-time surface reflectance is obtained by fitting the temporal features of the spectral reflectance signal; The real-time surface state features are obtained by feature stitching together the surface temperature uniformity index and the real-time surface reflectivity.

8. The laser cleaning closed-loop automated control method according to claim 6, characterized in that, The step of adaptively reducing the energy of the initial laser drive parameters to obtain a safe laser output command and an updated physical execution path based on the local over-temperature warning command includes: Extract the actual temperature value corresponding to the local over-temperature warning command, and calculate the temperature difference by subtracting the actual temperature value from the preset substrate damage temperature threshold. Using a proportional-integral-derivative control algorithm, a safe laser output command is obtained by dynamically adjusting the laser duty cycle or pulse frequency in the initial laser drive parameters based on the temperature difference. The energy response time is obtained by performing time delay analysis based on the safe laser output command, and the updated physical execution path is obtained by performing scan speed compensation calculation on the initial physical execution path based on the energy response time.

9. A closed-loop automated control system for laser cleaning, characterized in that, include: The initial distribution interpolation module is used to acquire ultrasonic echo data and corresponding acquisition spatial coordinates of the surface of the metal substrate to be cleaned, and to perform spatial interpolation on the ultrasonic echo data based on the acquisition spatial coordinates to obtain an initial thickness distribution matrix. The regional energy segmentation module is used to segment regions based on the initial thickness distribution matrix and a preset thickness energy correlation model to obtain the energy demand range of each region. The risk area screening module is used to perform amplitude statistical evaluation based on the energy demand range of the partition to obtain the regional energy fluctuation parameter, compare the regional energy fluctuation parameter with the preset fluctuation safety threshold to screen out high-risk physical areas and boundary coordinates, and perform weighted sorting based on the high-risk physical areas and boundary coordinates to obtain the high-risk area adjustment sequence. The resource scheduling and planning module is used to coordinate and schedule cleaning resources according to the high-risk area adjustment sequence and the preset total power constraints of equipment hardware, so as to obtain a basic global cleaning plan including initial laser driving parameters. The real-time state quantization module is used to acquire surface infrared thermal image signals and spectral reflection signals during the physical cleaning process according to the basic global cleaning plan, and to perform physical feature quantization on the surface infrared thermal image signals and the spectral reflection signals to obtain real-time surface state features. The energy reduction calibration update module is used to compare the real-time surface state characteristics with the preset substrate damage temperature threshold to obtain a local over-temperature warning command, and to perform adaptive energy reduction calibration on the initial laser driving parameters based on the local over-temperature warning command to obtain a safe laser output command and an updated physical execution path.

Images