A method and system for adaptive adjustment of laser cleaning process parameters

Abstract

This invention relates to the field of laser surface treatment technology, and discloses a method and system for adaptive adjustment of laser cleaning process parameters. The method includes acquiring image data and scanning path planning data of the laser-cleaned surface; performing semantic segmentation processing on the image data to obtain a contaminant distribution map; performing coordinated adjustment processing on the scanning path planning data, the contaminant distribution map, real-time coverage data, and substrate surface temperature data to obtain cleaning execution parameters including secondary adjustment power and final scanning speed; and performing combined issuance operations based on the cleaning execution parameters to obtain the adjustment result. This method can achieve adaptive coordinated adjustment of laser power, scanning speed, and cleaning progress, effectively eliminating local overheating and improving cleaning uniformity and efficiency.

Description

Technical Field

[0001] This invention relates to the field of laser surface treatment technology, and in particular to a method and system for adaptive adjustment of laser cleaning process parameters. Background Technology

[0002] Currently, with the deep integration of industrial manufacturing and high-end surface treatment technologies, laser cleaning technology has been widely used in various equipment maintenance and precision component refurbishment due to its advantages such as non-contact operation, no consumables, and high precision. Especially in upstream and downstream processes of complex industrial scenarios such as intelligent heat treatment production lines, the precise removal of oxide layers, rust, or contaminants from substrate surfaces has become a core element in ensuring the quality of subsequent processing and material performance.

[0003] In existing technologies, a continuous cleaning mode based on fixed process parameters or a step-by-step independent adjustment method relying solely on single-state feedback is typically employed. These systems often pre-set laser power and scanning speed, maintaining a constant output or abruptly switching based on threshold triggers throughout the entire operation cycle. However, in complex real-world conditions, the distribution of contaminants on the target surface often exhibits significant spatial non-uniformity. Due to a lack of precise perception of the microscopic residual distribution and real-time multi-dimensional quantification of cleaning progress, the pre-set rigid parameters are ill-suited to cope with dynamically changing process loads, easily leading to incomplete cleaning of heavily contaminated areas or excessive stripping of lightly contaminated areas. More critically, existing solutions fail to establish a fundamental dynamic coupling mechanism between laser energy input and the actuator's motion state. When a local area faces the risk of physical melting due to rapid heat accumulation, the system cannot perform real-time coordinated parameter intervention and adaptive reconstruction, easily causing thermal stress damage and irreversible quality degradation to the substrate surface.

[0004] Existing technologies have the problem of balancing cleaning uniformity with the thermal safety of the substrate. Summary of the Invention

[0005] This invention provides a method and system for adaptive adjustment of laser cleaning process parameters to solve the problem in the prior art that it is difficult to balance cleaning uniformity and substrate thermal safety.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for adaptive adjustment of laser cleaning process parameters, comprising: Image data of the laser-cleaned surface is acquired, and semantic segmentation processing is performed on the image data to obtain a contaminant distribution map; The initial adjustment power is obtained by performing a total amount assessment based on the preset density threshold and the mapping of pollutant distribution. The system acquires the preset scanning path planning data for the laser cleaning task, performs parameter estimation processing based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed, and performs speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command. Real-time coverage data of the cleaning area is acquired, and the real-time coverage data is processed to calculate the progress change rate. When the progress change rate is lower than the preset change rate threshold, compensation feedback processing is performed based on the pollutant distribution mapping and the progress change rate to obtain the power compensation factor. The initial adjusted power is modified according to the power compensation factor to obtain the secondary adjusted power; Acquire substrate surface temperature data, and perform deviation conversion processing based on the substrate surface temperature data and a preset protection threshold to obtain a temperature risk coefficient; Based on the temperature risk coefficient, the speed of the optimized speed command is adjusted for safety to obtain the final scanning speed. The secondary adjustment power and the final scanning speed are then combined and sent out to obtain the cleaning execution parameters.

[0007] Secondly, the present invention provides an adaptive adjustment system for laser cleaning process parameters, comprising: The image segmentation module is used to acquire image data of the laser-cleaned surface, perform semantic segmentation processing on the image data, and obtain a contaminant distribution map. The total quantity assessment module is used to perform total quantity assessment processing based on a preset density threshold and the mapping of pollutant distribution to obtain the initial adjustment power; The variable speed matching module is used to acquire the preset scanning path planning data of the laser cleaning task, perform parameter estimation processing based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed, and perform variable speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command. The compensation feedback module is used to acquire real-time coverage data of the cleaning area, perform progress calculation processing on the real-time coverage data to obtain the progress change rate, and when the progress change rate is lower than the preset change rate threshold, perform compensation feedback processing based on the pollutant distribution mapping and the progress change rate to obtain the power compensation factor. The parameter correction module is used to perform parameter correction processing on the initial adjusted power according to the power compensation factor to obtain the secondary adjusted power; The deviation conversion module is used to acquire substrate surface temperature data, and perform deviation conversion processing based on the substrate surface temperature data and a preset protection threshold to obtain a temperature risk coefficient. The combined delivery module is used to adjust the speed of the optimized speed command according to the temperature risk coefficient to obtain the final scanning speed, and to combine the secondary adjustment power and the final scanning speed to obtain the cleaning execution parameters.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains a contaminant distribution map by acquiring image data of the laser-cleaned surface and performing semantic segmentation processing. Based on this map, an initial adjustment power is obtained by performing total quantity assessment processing with a preset density threshold, thus achieving accurate quantitative perception of the microscopic contamination state of the target surface. This method breaks through the limitations of blindly outputting traditional static fixed parameters, ensuring dynamic matching between the initial laser energy feed and the actual contamination load, and avoiding surface damage to the substrate caused by excessive energy or residual plaque problems caused by insufficient energy from the source.

[0009] (2) This invention estimates the basic operating speed by introducing scanning path planning data and performs speed reduction matching in areas with high pollutant density. Simultaneously, it acquires real-time coverage data during the cleaning process to calculate the progress change rate, triggering compensation feedback processing for the initial power adjustment. This constructs a parameter dynamic adaptive correction mechanism driven by the actual processing progress. This mechanism effectively solves the problem of equipment operating parameters being disconnected from the actual physical load under complex dynamic working conditions, significantly improving the overall cleaning uniformity and dynamic operation efficiency.

[0010] (3) This invention obtains a temperature risk coefficient by acquiring the surface temperature data of the substrate and converting the deviation. When the system faces a potential overheating threat, it performs a safe speed-up operation on the optimized speed command based on this risk coefficient, thus constructing a tightly self-consistent thermal safety protection closed loop with rigorous underlying physical logic. This operation shortens the unit residence time of the high-energy laser in the local area by actively speeding up, effectively curbing the accumulation of heat distortion in the substrate. While ensuring the cleaning quality, it eliminates the hidden danger of irreversible thermal stress damage and physical melting of the substrate. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of a laser cleaning process parameter adaptive adjustment method provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of an adaptive adjustment system for laser cleaning process parameters 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 method for adaptive adjustment of laser cleaning process parameters, including the following steps: S1, acquire image data of the laser-cleaned surface, perform semantic segmentation processing on the image data, and obtain a contaminant distribution map; S2, perform total quantity assessment processing based on the preset density threshold and the pollutant distribution mapping to obtain the initial adjustment power; S3, acquire the preset scanning path planning data for the laser cleaning task, perform parameter estimation processing based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed, and perform speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command; S4, acquire real-time coverage data of the cleaning area, perform progress calculation processing on the real-time coverage data to obtain the progress change rate, and when the progress change rate is lower than the preset change rate threshold, perform compensation feedback processing based on the pollutant distribution mapping and the progress change rate to obtain the power compensation factor. S5, perform parameter correction processing on the initial adjustment power according to the power compensation factor to obtain the secondary adjustment power; S6, acquire substrate surface temperature data, and perform deviation conversion processing based on the substrate surface temperature data and the preset protection threshold to obtain the temperature risk coefficient; S7, adjust the speed of the optimized speed command according to the temperature risk coefficient to obtain the final scanning speed, and combine the secondary adjustment power and the final scanning speed to obtain the cleaning execution parameters.

[0014] In step S1, image data of the laser-cleaned surface is acquired, and semantic segmentation processing is performed on the image data to obtain a contaminant distribution map; The image data is semantically segmented to obtain a pollutant distribution map, including: The image data is subjected to image normalization processing to obtain a preprocessed image; The preprocessed image is input into a pre-trained U-Net semantic segmentation model for feature inference and extraction to obtain the pollutant probability distribution features; The pixel proportion of the pollutant probability distribution features is calculated to obtain the pollutant coverage rate, and the pollutant probability distribution features are mapped into a two-dimensional heat map in combination with a preset spatial coordinate system. Pollutant concentration areas are extracted from the two-dimensional heat map, and the K-means algorithm is used to perform cluster analysis on the pollutant concentration areas to obtain the coordinates of the cluster center points. The two-dimensional heat map and the coordinates of the cluster center points are then bound together to obtain the pollutant distribution mapping.

[0015] In one implementation, an industrial vision sensor deployed above the cleaning surface collects light reflection signals from the physical surface, generates a digital matrix with three color channels (red, green, and blue), and identifies this matrix as image data of the laser-cleaned surface. This embodiment performs channel dimensionality reduction on the image data, employing a fixed-ratio weighted average method. The values ​​of the red, green, and blue color channels are multiplied by preset red, green, and blue weights respectively, and then summed to convert them into single-channel grayscale values. Regarding the determination of the preset red, green, and blue weights, this embodiment extracts and normalizes the integral values ​​of the spectral sensitivity response function distribution of a standard optical sensor for different wavelengths of visible light. The calculated objective distribution ratios are strictly set as follows: red weight 0.299, green weight 0.587, and blue weight 0.114. Subsequently, this embodiment extracts the maximum and minimum grayscale values ​​in the current single-channel image, and uses a maximum-minimum normalization algorithm to linearly map the grayscale values ​​of all pixels to a floating-point range from zero to one, completing the image standardization process and outputting a single-channel preprocessed image.

[0016] It should be noted that, regarding the pre-trained U-Net semantic segmentation model, this embodiment employs a fully convolutional neural network based on an encoder-decoder architecture. Regarding the training process of this model, this embodiment extracts multi-source surface images from historical laser cleaning scenarios, and generates corresponding real contaminant distribution mask matrices from historical annotations. This embodiment performs data augmentation operations on the multi-source surface images with random angle geometric rotation and horizontal flipping to generate an augmented image set. The augmented image set is then combined with the corresponding real contaminant distribution mask matrix to construct a training sample set. This embodiment uses the above training sample set as input, employs a binary cross-entropy loss function to calculate the quantization error between the network output probability and the real annotations, and uses an adaptive moment estimation optimization algorithm to iteratively update the network weight parameters along the gradient descent direction. Regarding the determination of preset hyperparameters during training, this embodiment extracts the upper limit of the GPU memory capacity of the deployed computing nodes, rounds down the number of samples that cause the memory to be used to 90%, and objectively sets the batch size to a value of sixteen. Simultaneously, this embodiment uses a grid search method to perform multiple rounds of trial calculations on the independent test set to extract floating-point values ​​that result in the loss function having the largest negative gradient in the initial stage without diverging oscillations, and objectively sets the initial learning rate to 0.001. When the calculated validation set loss value stops decreasing for a preset number of rounds, this embodiment stops training, extracts the current network weight parameters, and outputs the pre-trained U-Net semantic segmentation model. The preset number of rounds is set based on the convergence variance fluctuation range of the validation set loss function during historical model training. This embodiment extracts the iteration span number covering three typical local oscillation cycles and objectively sets it to a value of ten to avoid premature stopping due to local oscillations and to prevent overfitting.

[0017] Specifically, 5000 surface images of laser cleaning scenarios were collected, covering three substrates: carbon steel, stainless steel, and aluminum alloy. Contaminant types included oxide scale, thin rust layers (<50μm thickness), and oil stains. Each image had a resolution of 512×512 pixels. Pixel-level semantic annotation was performed using the LabelMe annotation tool, labeling contaminated areas as foreground (1) and clean areas as background (0). The images were divided into a training set (4000 images), a validation set (500 images), and a test set (500 images) in an 8:1:1 ratio. The training set images were randomly rotated (-15° to +15°), horizontally flipped, and their brightness adjusted (0.8-1.2 times). Gaussian noise enhancement was applied, resulting in an equivalent training sample size of 24,000 images. The U-Net encoder consisted of 5 convolutional layers with 3×3 kernels and a stride of 1, followed by max pooling with a stride of 2 after each layer. The decoder consisted of 5 upsampling layers with skip connections. The output layer used the Sigmoid activation function. The batch size was 16, the initial learning rate was 0.001, and the Adam optimizer was used. The loss function was binary cross-entropy. The validation set loss was calculated after each iteration. If the validation set loss did not decrease for 10 consecutive epochs, training was stopped. In this embodiment, training was stopped at the 48th epoch, where the validation set loss stabilized at 0.023, and the average intersection-over-union ratio (IoU) of the test set was 0.87.

[0018] It is worth noting that in this embodiment, the preprocessed image generated in the preceding steps is input into the pre-trained U-Net semantic segmentation model. This model extracts features pixel-by-pixel through a non-linear activation function at the output layer, outputting a floating-point confidence probability of contaminants present on the physical surface for each pixel. This constitutes a probability matrix with a resolution completely consistent with the original image. In this embodiment, this probability matrix is ​​strictly defined as the contaminant probability distribution feature. This embodiment reads the confidence probabilities of all pixels in the contaminant probability distribution feature and performs an addition and summation operation. The summation result is divided by the total number of pixels in the image to calculate the contaminant coverage rate. Regarding the preset spatial coordinate system, this embodiment obtains the intrinsic and extrinsic parameter matrices of the industrial vision sensor using a checkerboard calibration method, constructs an affine transformation relationship between the pixel two-dimensional plane and the substrate physical plane, and sets this as the preset spatial coordinate system. This embodiment performs a geometric perspective transformation on the matrix coordinates of the contaminant probability distribution feature according to this preset spatial coordinate system, and linearly amplifies the confidence probabilities ranging from zero to one to a thermal scale of zero to one hundred using scalar multiplication, generating a two-dimensional heatmap characterizing the regional density distribution attribute.

[0019] It should be noted that the pollutant coverage rate is defined as the percentage of pixels with a pollutant probability greater than 0.5 out of the total number of pixels, used for macroscopic assessment of the proportion of polluted area; while the average residual density value in the subsequent steps refers to the arithmetic mean of the probability density values ​​of all pixels (range 0~1), and the relationship between the two depends on the probability distribution pattern; for example, when the probability of all polluted areas is 1, the coverage rate is equal to the average residual density; when the probability shows a gradual distribution, the average residual density value is usually less than the coverage rate; in this embodiment, both are used simultaneously to provide multidimensional information, but the initial adjustment power only depends on the average residual density value, and the coverage rate is only used as an auxiliary monitoring indicator.

[0020] In one implementation, a preset concentration threshold is set. Regarding the determination of this threshold, this embodiment extracts a two-dimensional heatmap distribution histogram of historical cleaned surface data, uses Otsu's method to calculate the adaptive segmentation value when the inter-class variance reaches its maximum value, and sets this value as the preset concentration threshold. This embodiment extracts the set of all physical grid coordinates in the two-dimensional heatmap where the heat value is greater than the preset concentration threshold, and identifies these as contaminant concentration areas. Subsequently, this embodiment uses the K-means algorithm to perform cluster analysis on the contaminant concentration areas. Regarding the method for determining the number of clusters K in the K-means algorithm, this embodiment uses the silhouette coefficient method, traversing candidate K values ​​from two to ten, calculating the global average silhouette coefficient for each classification case, and objectively setting the candidate value corresponding to the largest global average silhouette coefficient as the current number of clusters K. Based on this K value, this embodiment randomly initializes cluster centers within the contaminant concentration area and performs alternating distance iterative optimization until the Euclidean distance displacement of all cluster centers approaches zero. This embodiment extracts the final converged cluster center position values ​​and determines them as the cluster center point coordinates. In this embodiment, the numerical vector of the cluster center point coordinates is structurally spliced ​​and key-value pair combined with the two-dimensional grid matrix of the two-dimensional heat map to complete the data binding operation, generating a composite data structure containing the global thermal distribution field and local extreme value positions, which is strictly labeled as the pollutant distribution map.

[0021] It should be noted that the thermal value of the two-dimensional thermal map is a dimensionless relative pollution index, ranging from 0 to 100. In order to convert the thermal value into a pollutant density with actual physical meaning, in grams per square centimeter, this embodiment establishes a calibration curve through offline experiments, prepares pollution layers with different areal densities on the same substrate, measures their actual areal densities (weighing method) and simultaneously acquires images, and obtains the thermal value of the corresponding region through semantic segmentation; the conversion coefficient is obtained through linear fitting, and the actual density is equal to the thermal value multiplied by 0.003 grams per square centimeter; for example, a thermal value of 20 corresponds to 0.06 grams per square centimeter; in this embodiment, the real-time pollutant density used in step S4 is 0.3 grams per square centimeter, corresponding to a thermal value of 100, which belongs to a heavily polluted area.

[0022] In step S2, a total quantity assessment is performed based on a preset density threshold and the mapping between the pollutant distribution to obtain the initial adjustment power, including: Extract the density values ​​of all pixels in the two-dimensional heatmap corresponding to the pollutant distribution map; The density values ​​of all the pixels are added together to obtain the total density. The total density is then divided by the total number of pixels to obtain the average residual density value. When the average residual density value is greater than the preset density threshold, the difference between the average residual density value and the preset density threshold is calculated to obtain the excess density deviation. The excess density deviation, the preset base power, and the preset adjustment coefficient are multiplied to obtain the power increment. The base power and the power increment are added to obtain the initial adjustment power.

[0023] In one implementation, a digital scalar characterizing the degree of contamination of pixels is extracted from the contaminant distribution map generated in step S1. Specifically, in this embodiment, the value corresponding to each grid cell in the two-dimensional heatmap matrix is ​​read, and the value is determined as the density value of the corresponding pixel.

[0024] It should be noted that the preset density threshold is used to define the critical contamination level for laser equipment trigger power compensation. This embodiment uses the percentile method to objectively calibrate the preset density threshold. This embodiment extracts the density samples from the full thermal maps corresponding to historical qualified cleaning batches and calculates the cumulative distribution function of this sample set; the density value corresponding to the 95th percentile is objectively set as the preset density threshold to eliminate sporadic high-value interference caused by sensor noise; specifically, 200 thermal maps are randomly selected from historical qualified cleaning batches, all pixel density values ​​are extracted, and their cumulative distribution function is calculated; the density value corresponding to the 95th percentile is 0.05, and this embodiment uses 0.05 as the preset density threshold. If the distribution of the sampled data changes, it can be recalculated every quarter.

[0025] It is worth noting that in this embodiment, a traversal and accumulation operation is performed on all the extracted density values ​​to obtain a total density. Subsequently, this embodiment extracts the total number of grids in the two-dimensional heatmap as the total number of pixels, and divides the total density by the total number of pixels to calculate the average residual density value reflecting the global pollution level. When the average residual density value is greater than the preset density threshold, this embodiment performs a subtraction operation to obtain the excess density deviation.

[0026] In one implementation, the laser output intensity is dynamically excited based on the excess density deviation. Regarding the preset base power, this embodiment calibrates it based on the thermal damage threshold and initial physical state of the substrate material. Specifically, it obtains the safe operating upper limit per unit area by consulting a preset material energy efficiency database and sets it as the preset base power. The preset material energy efficiency database is a mapping table established by recording the maximum laser surface energy density without surface micro-melting and significant expansion of the heat-affected zone (HAZ) through gradient laser energy irradiation experiments on substrates of different standard grades. Regarding the preset adjustment coefficient, it quantifies the linear mapping relationship between density deviation and electrical energy conversion. This embodiment uses a system identification method, employing least squares regression analysis on the average residual density reduction rate after historical power adjustment as the cleaning effect to calculate the proportional slope that makes the cleaning efficiency curve tend to a steady state, and determines it as the preset adjustment coefficient. This embodiment multiplies the excess density deviation, the preset base power, and the preset adjustment coefficient to obtain the power increment, and then adds this power increment to the preset base power to output the final initial adjusted power.

[0027] Specifically, the construction of the material energy efficiency database involves selecting typical substrates, such as carbon steel Q235, stainless steel 304, and aluminum alloy 6061, and preparing five 50 mm x 50 mm samples for each material. A continuous laser is used to scan the samples at different powers, such as 50 watts to 150 watts, with a step size of 10 watts. After scanning, the surface is observed using a metallographic microscope. The highest power at which no micro-melting pits or significant expansion of the heat-affected zone appears is recorded as the safe upper limit for that material. The material grade and the safe upper limit power are stored in the database table. In this embodiment, the safe upper limit for 304 stainless steel is 100 watts, which is the preset base power.

[0028] In step S3, the preset scanning path planning data of the laser cleaning task is obtained, and parameter estimation processing is performed based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed. The speed matching processing is performed based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command.

[0029] The process of estimating parameters based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed includes: The difference between the initial adjustment power and the preset initial power is calculated to obtain the power deviation; The speed increment is obtained by multiplying the power deviation by a preset power adjustment factor. The preset base scanning speed is extracted from the scanning path planning data, and the base scanning speed is summed with the speed increment to obtain the base running speed.

[0030] In one implementation, a geometric trajectory description file for the current cleaning operation is retrieved from the host computer control system of the laser equipment and used as the preset scanning path planning data for the laser cleaning task. The scanning path planning data includes the boundary coordinates of the cleaning area, the total path length, and the suggested movement reference along that path.

[0031] It should be noted that the preset initial power is determined using a device calibration method in this embodiment. Specifically, this embodiment extracts the maximum output electrical energy of the laser device without thermal damage when performing dry scanning on a non-contaminated substrate under standard laboratory conditions, and objectively sets this as the preset initial power. Regarding the preset power adjustment factor, it quantifies the coupling ratio between energy increment and mechanical motion speed compensation. This embodiment uses a system identification method, performing linear regression fitting on the power change and the displacement increment required for cleaning per unit depth in historical operating data, calculating the slope of the fitted curve, and determining this as the preset power adjustment factor.

[0032] It is worth noting that in this embodiment, the power deviation is obtained by subtracting the preset initial power from the initial adjustment power using an algebraic difference. Then, this power deviation is multiplied by the preset power adjustment factor to calculate the speed increment reflecting the power compensation requirement. In this embodiment, a preset base scanning speed, which is tied to the current substrate material and process requirements, is read from the scanning path planning data. This base scanning speed is determined in advance by performing a single-pass cleaning of a standard contaminated sample at the base power using an orthogonal experimental method, taking the center speed range at which the cleaning residue rate reaches the process standard. Finally, the preset base scanning speed is summed with the speed increment to output a base operating speed that satisfies global energy density balance.

[0033] For example, in this embodiment, the initial adjustment power is 110 watts, and the preset initial power is 100 watts. This embodiment performs a subtraction operation to obtain a power deviation of 10 watts. The preset power adjustment factor is extracted as 2. Multiplying 10 by 2 yields a speed increment of 20 mm / s. The preset base scan speed corresponding to a path length of 5000 mm is extracted from the path planning file as 180 mm / s. Adding 180 and 20 yields a base running speed of 200 mm / s.

[0034] The optimized speed command is obtained by performing variable speed matching processing based on the pollutant distribution mapping and the basic operating speed, including: Extract the set of pixels with a density value greater than a preset high-density threshold from the pollutant distribution map, and calculate the ratio of the area of ​​the pixel set to the total cleaning area to obtain the proportion of high-density areas. The high-density region ratio is multiplied by a preset speed reduction coefficient to obtain the speed reduction ratio; Subtract the speed reduction ratio from the value to obtain the safety multiplier. Multiply the base operating speed by the safety multiplier to obtain the optimized speed value. The optimized speed value is converted into a hardware drive control signal to obtain the optimized speed command.

[0035] In one implementation, local feature extraction is performed on the pollutant distribution map output in step S1. It should be noted that this embodiment uses the percentile method for the preset high-density threshold. This embodiment extracts the probability density function of the density distribution of all pixels in the historical pollution image, calculates the value of the corresponding 90th percentile, and objectively sets it as the preset high-density threshold. This embodiment extracts the coordinates of all pixels with density values ​​greater than this threshold from the two-dimensional heatmap to form a pixel set, and counts the total number of pixels in this set; dividing this total number of pixels by the total number of pixels representing the total cleaning area yields the proportion of high-density areas.

[0036] It is worth noting that this embodiment uses this proportion to nonlinearly suppress the movement speed. Regarding the preset speed reduction coefficient, this embodiment employs a performance-driven method, using a grid search algorithm to iteratively optimize on a historical cleaning quality sample set, extracting the optimal proportion parameter that makes the residual rate in high-density areas less than 0.1%, and determining this as the preset speed reduction coefficient. This embodiment multiplies the proportion of high-density areas by the preset speed reduction coefficient to obtain the speed reduction ratio. Subsequently, the calculated value is subtracted from this ratio to obtain the safety multiplier.

[0037] It should be noted that in this embodiment, the aforementioned basic operating speed is multiplied by the safety multiplier to calculate the optimized speed value considering localized heavy pollution. Subsequently, this embodiment maps this value to the physical control domain of the laser galvanometer or translation stage. This embodiment extracts a preset signal conversion coefficient. Regarding the determination of this coefficient, this embodiment obtains the rated linear ratio between the input voltage range of the actuator driver and the corresponding output physical speed, calculates the required level deviation per unit speed, and sets it as the signal conversion coefficient. This embodiment multiplies the optimized speed value by this signal conversion coefficient to generate a corresponding pulse width modulation signal or analog voltage signal, which is then determined as the optimized speed command.

[0038] In step S4, real-time coverage data of the cleaning area is acquired, and progress calculation processing is performed on the real-time coverage data to obtain the progress change rate. When the progress change rate is lower than a preset change rate threshold, compensation feedback processing is performed based on the pollutant distribution mapping and the progress change rate to obtain a power compensation factor.

[0039] The process of calculating the progress of the real-time coverage data to obtain the progress change rate includes: Extract the current coverage area at the current moment from the real-time coverage data, and extract the historical coverage area at the previous moment; The difference between the current coverage area and the historical coverage area is calculated to obtain the area increment; The progress change rate is obtained by dividing the area increment by the time span parameter between the current time and the previous time.

[0040] In one implementation, the real-time deflection angle of the galvanometer and the displacement coordinates of the transmission mechanism are read by the underlying motion control card of the laser device. The coordinate data is then mapped to the three-dimensional physical model of the substrate and subjected to a geometric polygon Boolean union operation to calculate a scalar value representing the physical area of ​​the scanned area. This scalar value is then determined as the real-time coverage data of the cleaning area.

[0041] This embodiment uses the discrete grid method to approximate the covered area. The area to be cleaned is divided into square grids with a side length of one millimeter, and a two-dimensional Boolean array is established. Whenever the galvanometer scan passes through a grid cell, the cell is marked as one. The current covered area is equal to the number of grid cells marked as one multiplied by one square millimeter. The historical covered area is taken from the same Boolean array at the previous sampling time, and the sampling period is fixed at 0.1 seconds. The error of this method is less than one percent, which meets the engineering requirements.

[0042] It should be noted that this embodiment uses a physical boundary constraint reverse calculation method to determine the fixed sampling clock frequency of the equipment's main control system. Specifically, this embodiment extracts the maximum rated scanning line speed of the laser galvanometer and the minimum spot pitch resolution preset by the cleaning process; divides the maximum rated scanning line speed by the minimum spot pitch resolution to calculate the critical sampling frequency that meets the spatial continuity requirement; and performs an up-rounding mapping operation on the critical sampling frequency using the hardware timer frequency division rule of the main control chip to objectively set it as the fixed sampling clock frequency. Based on this fixed sampling clock frequency, this embodiment extracts the current coverage area corresponding to the current moment from the real-time coverage data, and extracts the historical coverage area of ​​the previous moment stored in the local cache sequence.

[0043] It should be noted that this embodiment calculates the algebraic difference between the current coverage area and the historical coverage area to obtain the area increment. Simultaneously, this embodiment extracts the system timestamp difference between the current moment and the previous moment and sets it as a time span parameter. This embodiment divides the area increment by the time span parameter to calculate the progress change rate, reflecting the instantaneous operational speed of the equipment.

[0044] It is worth noting that this embodiment uses statistical analysis to determine the preset rate of change threshold. This embodiment extracts the progress change rate data of all cleaning tasks in the steady-state execution phase from historical qualified production batches to form a sample set, and calculates the arithmetic mean and standard deviation of this sample set. The arithmetic mean is then subtracted by twice the standard deviation, and this subtracted value is objectively set as the preset rate of change threshold. This embodiment compares the currently calculated progress change rate with the preset rate of change threshold.

[0045] The power compensation factor is obtained by performing compensation feedback processing based on the pollutant distribution mapping and the progress change rate, including: Extract the real-time pollutant density of the current scanned area from the pollutant distribution map; The real-time pollutant density is multiplied by a preset area weight to obtain a compensation molecule variable; The power compensation factor is obtained by dividing the compensation numerator variable by the rate of change of progress.

[0046] In one implementation, when the rate of change of progress is lower than the preset rate of change threshold, this embodiment extracts the real-time spatial coordinates of the laser focus on the physical substrate at the current moment. Using these real-time spatial coordinates as the search key, this embodiment searches for the corresponding two-dimensional thermal map grid cell from the pollutant distribution map pre-generated in step S1, extracts the thermal scalar value recorded in the grid cell, and strictly determines it as the real-time pollutant density of the current scanning area; when the rate of change of progress is not lower than the preset rate of change threshold, the power compensation factor is assigned a value of zero, and no additional compensation is made for the initial adjustment power.

[0047] It should be noted that the preset area weight is objectively calibrated using a multivariate data fitting method in this embodiment. This embodiment extracts offline experimental data from historical laser surface treatments, sets different combinations of scanning advance speed and power gradient, and records the combination with no residue and no heat-affected zone observed under a microscope as the optimal laser energy intrusion amount, then performs regression fitting. A ternary nonlinear mapping dataset containing the relationship between contaminant density, scanning advance speed, and optimal laser energy intrusion amount is constructed; a multivariate linear regression algorithm is used to fit this dataset, calculating the dimensionless regression coefficient that balances the substrate heat input and contaminant stripping amount, and this regression coefficient is objectively set as the preset area weight.

[0048] Specifically, on 304 stainless steel samples, real-time contaminant densities were set to 0.1, 0.2, and 0.3 grams per square centimeter, with progress change rates of 80, 100, and 120 square millimeters per second, respectively. The optimal laser power increment was determined through orthogonal experiments to ensure no residue and no overheating after cleaning. Using the product of contaminant density and progress change rate as the independent variable and the optimal power increment as the dependent variable, a linear regression equation was fitted. The power increment was calculated as the ratio of contaminant density to progress change rate multiplied by 0.15. Therefore, the area weight was determined to be 0.15. In this example, 0.5 was used for simplification; in actual applications, this value should be adjusted according to the material.

[0049] It is worth noting that in this embodiment, the real-time pollutant density extracted in the preceding steps is multiplied by the preset area weight to calculate the compensation molecule variable. Subsequently, this embodiment divides the compensation molecule variable by the rate of change of progress to calculate a value reflecting the additional compensation of laser output energy required due to the slow progress, and outputs this value as the power compensation factor.

[0050] In step S5, the initial adjusted power is modified according to the power compensation factor to obtain the secondary adjusted power, including: The power compensation factor is converted into the corresponding adjustment suggestion difference through multiplication, and the initial adjustment power is added to the adjustment suggestion difference to obtain the secondary adjustment power.

[0051] In one implementation, the power compensation factor calculated and output in the preceding step (i.e., step S4) is extracted. This embodiment constructs the mapping calculation logic between this dimensionless factor and actual physical electrical energy, converting it into a value containing the dimension of absolute power.

[0052] It should be noted that, regarding the process of converting the power compensation factor into the corresponding adjustment suggestion difference through multiplication, this embodiment introduces a preset power conversion coefficient. The preset power conversion factor is determined using an offline engineering calibration method. Specifically, this embodiment extracts energy step response test data of historical laser cleaning equipment under different materials and contamination levels, calculates the additional physical electrical power input required to cause a change in unit peel depth under a unit change in scanning speed, and uses the least squares method to perform linear fitting on multiple sets of test data, extracting the slope proportionality constant of the fitted line, and objectively setting it as the preset power conversion coefficient.

[0053] Specifically, with a fixed pollutant density of 0.2 grams per square centimeter, the progress change rate was set to 60, 80, 100, and 120 square millimeters per second, and the laser power was gradually increased until the cleaning was complete. Calculations showed that the power compensation factor, which is the ratio of the theoretically required power increment to the initial power, has a linear relationship with the reciprocal of the progress change rate, and the absolute value of the fitting slope is 2000. Therefore, the adjustment suggestion is to multiply the power compensation factor by 2000; the unit of this coefficient is implicitly watts.

[0054] It is worth noting that in this embodiment, the power compensation factor and the preset power conversion coefficient are multiplied by a scalar to calculate an absolute physical quantity value with a specific power unit attribute, which is then strictly determined as the adjustment suggestion difference. Subsequently, this embodiment extracts the initial adjustment power calculated in step S2, performs an algebraic addition operation on the scalar value of the initial adjustment power and the adjustment suggestion difference, completes the addition compensation operation, calculates the latest control parameters after feedback correction for the current scanning progress, and outputs it as the secondary adjustment power.

[0055] In step S6, the surface temperature data of the substrate is obtained, and the deviation is calculated based on the surface temperature data of the substrate and the preset protection threshold to obtain the temperature risk coefficient.

[0056] Specifically, a temperature risk coefficient is obtained by performing deviation conversion processing based on the substrate surface temperature data and a preset protection threshold, including: Extract the highest temperature value from the substrate surface temperature data, and determine whether the highest temperature value is greater than the preset protection threshold. When the highest temperature value is greater than the preset protection threshold, the difference between the highest temperature value and the preset protection threshold is calculated to obtain the over-temperature deviation value. The over-temperature deviation value is divided by the preset protection threshold and converted into a percentage format to obtain the temperature risk coefficient. When the highest temperature value is not greater than the preset protection threshold, the temperature risk coefficient is set to zero.

[0057] In one implementation, an array of infrared thermal imaging sensors deployed around the laser cleaning processing area collects the thermal radiation infrared wavelength signal of the physical surface. This thermal radiation infrared wavelength signal is converted into a corresponding Celsius temperature matrix, which is then determined as the substrate surface temperature data. In this embodiment, all two-dimensional grid elements in the Celsius temperature matrix are traversed, and the element with the largest value is extracted and objectively determined as the highest point temperature value.

[0058] It should be noted that this embodiment uses an offline material thermophysical experimental calibration method to determine the preset protection threshold. Specifically, this embodiment extracts metallographic test samples of similar physical substrates under different temperature gradients, and tests their surface yield strength and microcrystalline lattice deformation; it extracts the critical physical temperature value at which the microcrystalline lattice deformation undergoes a nonlinear abrupt change or the surface yield strength decreases by more than 5%, and subtracts a preset safety redundancy tolerance from this critical physical temperature value, strictly setting it as the preset protection threshold. Regarding the determination of this preset safety redundancy tolerance, this embodiment extracts the maximum steady-state measurement error range of the aforementioned infrared thermal imaging sensor array under a standard blackbody radiation source, and sets its absolute value as the preset safety redundancy tolerance.

[0059] It is worth noting that this embodiment compares the extracted highest temperature value with the preset protection threshold. When the highest temperature value is greater than the preset protection threshold, this embodiment performs algebraic subtraction to subtract the preset protection threshold from the highest temperature value, obtaining an overtemperature deviation value with absolute temperature dimensions. Subsequently, this embodiment divides the overtemperature deviation value by the preset protection threshold to obtain a dimensionless ratio, and multiplies this ratio by one hundred to convert it into a percentage format. This calculated percentage characteristic is determined as the temperature risk coefficient. Conversely, when the highest temperature value is less than or equal to the preset protection threshold, it indicates that the current heat accumulation state has not triggered overheating conditions, and this embodiment directly assigns the temperature risk coefficient a value of zero.

[0060] In step S7, the optimized speed command is adjusted for safety based on the temperature risk coefficient to obtain the final scanning speed. The secondary adjustment power and the final scanning speed are then combined and sent out to obtain cleaning execution parameters, including: The speed increase ratio is obtained by multiplying the temperature risk coefficient with a preset gain ratio constant. Add the speed increase ratio to the value to obtain the safe speed increase coefficient; Extract the original scanning speed corresponding to the optimized speed command, and multiply the original scanning speed with the safety speed-up coefficient to obtain the final scanning speed; The secondary adjustment power and the final scanning speed are spliced ​​together according to a preset communication protocol format to obtain the cleaning execution parameters.

[0061] In one implementation, the temperature risk coefficient generated in the preceding steps is extracted. It should be noted that the engineering physics calibration method is used to determine the preset gain ratio constant in this embodiment. Specifically, this embodiment obtains a three-dimensional finite element heat conduction simulation model with the same material as the current substrate. Regarding the construction method of this three-dimensional finite element heat conduction simulation model, this embodiment uses a moving Gaussian surface heat source distribution model to characterize the dynamic energy feed state of the laser spot on the substrate surface; extracts the standard ambient room temperature value, sets the boundary conditions of the simulation model surface in contact with air to the natural convection heat transfer state at this room temperature, and loads the corresponding surface convection heat transfer coefficient; simultaneously, this embodiment performs high-density mesh generation in the laser-directly acting area and the adjacent heat-affected zone, and performs low-density mesh generation in the substrate edge area far from the heat source, generating the complete three-dimensional finite element heat conduction simulation model based on the above-mentioned locally refined mesh generation strategy. In this embodiment, an increasing scan velocity sequence is input into the three-dimensional finite element heat conduction simulation model under different initial overtemperature states. The decay sequence of heat input per unit area is calculated and extracted. A least squares method is used to perform linear regression fitting between the decay sequence and the corresponding temperature drop on the substrate surface. The absolute value of the slope reflecting the sensitivity of the negative correlation between velocity and temperature in the fitting function is extracted and objectively set as the preset gain ratio constant. This embodiment then performs a scalar multiplication operation between the temperature risk coefficient and the preset gain ratio constant to calculate the velocity increase ratio corresponding to the acceleration action of the equipment motion system.

[0062] It is worth noting that this embodiment performs algebraic addition, adding the speed increase ratio to the numerical value to calculate the dimensionless safety speed-up coefficient. Subsequently, this embodiment parses and extracts the absolute speed value (without thermal safety correction) from the optimized speed command generated in the preceding steps, determining it as the original scan speed. This embodiment multiplies the original scan speed by the safety speed-up coefficient to calculate the final scan speed representing the current physical cycle.

[0063] It should be noted that the preset communication protocol format in this embodiment is established based on the underlying hardware bus standard. This embodiment extracts the standard register mapping table of the industrial control system, defines a data structure including frame header, device address, function code, data area, and checksum, and strictly sets it to the preset communication protocol format. This embodiment extracts the secondary adjustment power output in step S5 and the previously calculated final scanning speed, converts these two floating-point values ​​into hexadecimal byte streams according to the requirements of the data structure, and fills them into the power control word segment and speed control word segment of the data area respectively, performing a message concatenation operation. This embodiment determines the completed byte sequence after concatenation as the cleaning execution parameters and outputs it to the hardware interface execution unit of the laser power supply and motion control console.

[0064] It should be noted that the communication protocol used in this embodiment has a frame header of two bytes, fixed as hexadecimal AA and 55; a device address of one byte, with the laser address being 01 and the galvanometer address being 02; a function code of one byte, fixed as 10 when writing to multiple registers; a data area length of one byte, indicating the number of subsequent data bytes; a total of eight bytes in the data area, of which the first four bytes are the power control word (IEEE 754 single-precision floating-point number, in watts), and the last four bytes are the speed control word (single-precision floating-point number, in millimeters per second); and a checksum of one byte, which is the sum of the frame header, address, function code, data area length, and all bytes in the data area, and the lower eight bits are taken; for example, a message with a power of 108 watts and a speed of 5 millimeters per second is AA 55 0110 08 42 D8 00 00 40 A0 00 00 7F.

[0065] In summary, this invention utilizes deep learning algorithms to perform semantic segmentation on real-time acquired image data, achieving precise quantification of surface contaminant distribution mapping and solving the technical challenge of ambiguous physical state perception in traditional laser cleaning processes. By establishing a multi-dimensional compensation feedback mechanism between contaminant density, cleaning progress change rate, and laser output power, it achieves dynamic matching between laser energy input and actual operation progress, effectively solving the problem of incomplete cleaning or excessive stripping caused by uneven contaminant distribution. By introducing a safety acceleration intervention mechanism based on real-time surface temperature monitoring, a robust thermal safety protection closed loop is constructed, eliminating the risk of substrate thermal damage caused by local energy accumulation from a physical logic perspective. Ultimately, this invention achieves full-dimensional coordinated adjustment of laser power, scanning speed, and operation progress, significantly improving the consistency and automation level of laser cleaning operations while greatly enhancing the equipment's adaptability to complex working conditions and substrate protection capabilities.

[0066] Reference Figure 2 The second embodiment of the present invention provides an adaptive adjustment system for laser cleaning process parameters, comprising: The image segmentation module is used to acquire image data of the laser-cleaned surface, perform semantic segmentation processing on the image data, and obtain a contaminant distribution map. The total quantity assessment module is used to perform total quantity assessment processing based on a preset density threshold and the mapping of pollutant distribution to obtain the initial adjustment power; The variable speed matching module is used to acquire the preset scanning path planning data of the laser cleaning task, perform parameter estimation processing based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed, and perform variable speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command. The compensation feedback module is used to acquire real-time coverage data of the cleaning area, perform progress calculation processing on the real-time coverage data to obtain the progress change rate, and when the progress change rate is lower than the preset change rate threshold, perform compensation feedback processing based on the pollutant distribution mapping and the progress change rate to obtain the power compensation factor. The parameter correction module is used to perform parameter correction processing on the initial adjusted power according to the power compensation factor to obtain the secondary adjusted power; The deviation conversion module is used to acquire substrate surface temperature data, and perform deviation conversion processing based on the substrate surface temperature data and a preset protection threshold to obtain a temperature risk coefficient. The combined delivery module is used to adjust the speed of the optimized speed command according to the temperature risk coefficient to obtain the final scanning speed, and to combine the secondary adjustment power and the final scanning speed to obtain the cleaning execution parameters.

[0067] It should be noted that the laser cleaning process parameter adaptive adjustment system provided in this embodiment of the invention is used to execute all the process steps of the laser cleaning process parameter adaptive adjustment method in the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0068] 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.

[0069] 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 method for adaptive adjustment of laser cleaning process parameters, characterized in that, include: Image data of the laser-cleaned surface is acquired, and semantic segmentation processing is performed on the image data to obtain a contaminant distribution map; The initial adjustment power is obtained by performing a total amount assessment based on the preset density threshold and the mapping of pollutant distribution. The system acquires the preset scanning path planning data for the laser cleaning task, performs parameter estimation processing based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed, and performs speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command. Real-time coverage data of the cleaning area is acquired, and the real-time coverage data is processed to calculate the progress change rate. When the progress change rate is lower than the preset change rate threshold, compensation feedback processing is performed based on the pollutant distribution mapping and the progress change rate to obtain the power compensation factor. The initial adjusted power is modified according to the power compensation factor to obtain the secondary adjusted power; Acquire substrate surface temperature data, and perform deviation conversion processing based on the substrate surface temperature data and a preset protection threshold to obtain a temperature risk coefficient; Based on the temperature risk coefficient, the speed of the optimized speed command is adjusted for safety to obtain the final scanning speed. The secondary adjustment power and the final scanning speed are then combined and sent out to obtain the cleaning execution parameters.

2. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The step of performing semantic segmentation processing on the image data to obtain a pollutant distribution map includes: The image data is subjected to image normalization processing to obtain a preprocessed image; The preprocessed image is input into a pre-trained U-Net semantic segmentation model for feature inference and extraction to obtain the pollutant probability distribution features; The pixel proportion of the pollutant probability distribution features is calculated to obtain the pollutant coverage rate, and the pollutant probability distribution features are mapped into a two-dimensional heat map in combination with a preset spatial coordinate system. Pollutant concentration areas are extracted from the two-dimensional heat map, and the K-means algorithm is used to perform cluster analysis on the pollutant concentration areas to obtain the coordinates of the cluster center points. The two-dimensional heat map and the coordinates of the cluster center points are then bound together to obtain the pollutant distribution mapping.

3. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The step of performing a total quantity assessment based on a preset density threshold and the mapping of pollutant distribution to obtain the initial adjustment power includes: Extract the density values ​​of all pixels in the two-dimensional heatmap corresponding to the pollutant distribution map; The density values ​​of all the pixels are added together to obtain the total density. The total density is then divided by the total number of pixels to obtain the average residual density value. When the average residual density value is greater than the preset density threshold, the difference between the average residual density value and the preset density threshold is calculated to obtain the excess density deviation. The excess density deviation, the preset base power, and the preset adjustment coefficient are multiplied to obtain the power increment. The base power and the power increment are added to obtain the initial adjustment power.

4. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The step of performing parameter estimation based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed includes: The difference between the initial adjustment power and the preset initial power is calculated to obtain the power deviation; The speed increment is obtained by multiplying the power deviation by a preset power adjustment factor. The preset base scanning speed is extracted from the scanning path planning data, and the base scanning speed is summed with the speed increment to obtain the base running speed.

5. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The step of performing variable speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain an optimized speed command includes: Extract the set of pixels with a density value greater than a preset high-density threshold from the pollutant distribution map, and calculate the ratio of the area of ​​the pixel set to the total cleaning area to obtain the proportion of high-density areas. The high-density region ratio is multiplied by a preset speed reduction coefficient to obtain the speed reduction ratio; Subtract the speed reduction ratio from the value to obtain the safety multiplier. Multiply the base operating speed by the safety multiplier to obtain the optimized speed value. The optimized speed value is converted into a hardware drive control signal to obtain the optimized speed command.

6. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The process of calculating the progress of the real-time coverage data to obtain the progress change rate includes: Extract the current coverage area at the current moment from the real-time coverage data, and extract the historical coverage area at the previous moment; The difference between the current coverage area and the historical coverage area is calculated to obtain the area increment; The progress change rate is obtained by dividing the area increment by the time span parameter between the current time and the previous time.

7. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The step of performing compensation feedback processing based on the pollutant distribution mapping and the rate of progress change to obtain a power compensation factor, and then performing parameter correction processing on the initial adjustment power based on the power compensation factor to obtain a secondary adjustment power, includes: Extract the real-time pollutant density of the current scanned area from the pollutant distribution map; The real-time pollutant density is multiplied by a preset area weight to obtain a compensation molecule variable; Divide the compensation numerator variable by the rate of change of progress to obtain the power compensation factor; The power compensation factor is converted into the corresponding adjustment suggestion difference through multiplication, and the initial adjustment power is added to the adjustment suggestion difference to obtain the secondary adjustment power.

8. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The step of performing deviation conversion processing based on the substrate surface temperature data and a preset protection threshold to obtain a temperature risk coefficient includes: Extract the highest temperature value from the substrate surface temperature data, and determine whether the highest temperature value is greater than the preset protection threshold. When the highest temperature value is greater than the preset protection threshold, the difference between the highest temperature value and the preset protection threshold is calculated to obtain the over-temperature deviation value. The over-temperature deviation value is divided by the preset protection threshold and converted into a percentage format to obtain the temperature risk coefficient. When the highest temperature value is not greater than the preset protection threshold, the temperature risk coefficient is set to zero.

9. The adaptive adjustment method for laser cleaning process parameters according to claim 1, characterized in that, The process involves adjusting the optimized speed command based on the temperature risk coefficient to obtain the final scanning speed. The secondary adjustment power and the final scanning speed are then combined and sent out to obtain cleaning execution parameters, including: The speed increase ratio is obtained by multiplying the temperature risk coefficient with a preset gain ratio constant. Add the speed increase ratio to the value to obtain the safe speed increase coefficient; Extract the original scanning speed corresponding to the optimized speed command, and multiply the original scanning speed with the safety speed-up coefficient to obtain the final scanning speed; The secondary adjustment power and the final scanning speed are spliced ​​together according to a preset communication protocol format to obtain the cleaning execution parameters.

10. A laser cleaning process parameter adaptive adjustment system, characterized in that, include: The image segmentation module is used to acquire image data of the laser-cleaned surface, perform semantic segmentation processing on the image data, and obtain a contaminant distribution map. The total quantity assessment module is used to perform total quantity assessment processing based on a preset density threshold and the mapping of pollutant distribution to obtain the initial adjustment power; The variable speed matching module is used to acquire the preset scanning path planning data of the laser cleaning task, perform parameter estimation processing based on the initial adjustment power and the scanning path planning data to obtain the basic operating speed, and perform variable speed matching processing based on the pollutant distribution mapping and the basic operating speed to obtain the optimized speed command. The compensation feedback module is used to acquire real-time coverage data of the cleaning area, perform progress calculation processing on the real-time coverage data to obtain the progress change rate, and when the progress change rate is lower than the preset change rate threshold, perform compensation feedback processing based on the pollutant distribution mapping and the progress change rate to obtain the power compensation factor. The parameter correction module is used to perform parameter correction processing on the initial adjusted power according to the power compensation factor to obtain the secondary adjusted power; The deviation conversion module is used to acquire substrate surface temperature data, and perform deviation conversion processing based on the substrate surface temperature data and a preset protection threshold to obtain a temperature risk coefficient. The combined delivery module is used to adjust the speed of the optimized speed command according to the temperature risk coefficient to obtain the final scanning speed, and to combine the secondary adjustment power and the final scanning speed to obtain the cleaning execution parameters.

Images