A method for controlling the laser etching depth on tape surface by multimodal data fusion

Abstract

This invention relates to a multimodal data fusion method for controlling the laser etching depth of an adhesive tape surface in the field of information technology. The method includes: acquiring raw image data of the adhesive tape surface and generating a texture distribution map; extracting local roughness features based on the texture distribution map and determining material uniformity; calculating the laser energy absorption coefficient distribution based on the roughness features and constructing an etching depth prediction model; obtaining prediction results based on the etching depth prediction model and analyzing equipment power deviation; adjusting the focus position based on the equipment power deviation and generating a focus offset correction trajectory; adjusting laser scanning path parameters based on the focus offset correction trajectory and generating a dynamic energy output command; and driving the laser to form a consistent etching depth pattern using the dynamic energy output command and verifying the etching results.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for controlling the laser etching depth on the surface of tape through multimodal data fusion. Background Technology

[0002] In the laser engraving business scenario of QR code tape, the uneven texture of the tape surface leads to uneven distribution of laser energy absorption coefficient, which directly causes the etching depth deviation. This is first manifested as insufficient absorption in areas with excessively high local roughness gradients, resulting in blurred edges of the QR code pattern and a depth that is shallower than the target value.

[0003] Subsequently, this absorption unevenness, combined with real-time fluctuations in laser power, amplifies the energy density differences at texture boundaries, causing the deviation to extend from local to the entire pattern area, resulting in systematic inconsistency. Simultaneously, the focal position shifts due to minor substrate anomalies, introducing environmental interference gradients and worsening the roughness feature distribution. This leads to abnormal thermal distribution and a surge in acoustic emission signals. These multi-source physical signals are highly coupled, yet the dynamic interaction process cannot be captured by traditional single-frame image post-processing.

[0004] As a result, the root causes of defects such as power drift, focus deviation and substrate abnormalities are difficult to attribute in real time, the compensation strategy lacks a closed-loop mechanism, and the process parameters cannot be adaptively adjusted, ultimately resulting in low product consistency and a scrap rate of over 20%, which seriously restricts the efficiency of high-precision mass production. Summary of the Invention

[0005] This invention provides a method for controlling the laser etching depth on the surface of adhesive tape through multimodal data fusion, mainly comprising:

[0006] The process involves: acquiring raw image data of the tape surface and generating a texture distribution map; extracting local roughness features based on the texture distribution map and determining material uniformity; calculating the laser energy absorption coefficient distribution based on the roughness features and constructing an etching depth prediction model; obtaining prediction results based on the etching depth prediction model and analyzing equipment power deviation; adjusting the focus position based on the equipment power deviation and generating a focus offset correction trajectory; adjusting the laser scanning path parameters based on the focus offset correction trajectory and generating a dynamic energy output command; and driving the laser to form a consistent etching depth pattern and verifying the etching results using the dynamic energy output command. Furthermore, the step of acquiring the original image data of the tape surface and generating a texture distribution map includes: acquiring a first image of the tape surface using a high-resolution line scan camera, and simultaneously acquiring a second image using a high-speed camera; fusing the first image and the second image to obtain original image data; annotating texture regions in the original image data, and simultaneously collecting reflected light intensity data to generate light intensity recording data; using the light intensity recording data to assist in dividing texture boundaries, and generating a texture boundary map from the texture regions; generating a texture distribution map of the tape surface using the texture boundary map, ensuring that the texture distribution map reflects the texture feature distribution of the tape surface; if there are outliers in the light intensity recording data, correcting the texture boundary map to obtain an optimized texture distribution map; and using the optimized texture distribution map to provide basic data for subsequent roughness feature extraction. Furthermore, the step of extracting local roughness features and determining material uniformity based on the texture distribution map includes: extracting local roughness gradient distribution using a gradient operator based on the texture distribution map to generate a gradient distribution map; fusing thermal distribution information obtained from infrared thermal images into the gradient distribution map to calculate the uniformity statistical index of the texture region; if the uniformity statistical index is lower than a preset threshold, locating the non-uniform texture region; for the non-uniform texture region, integrating acoustic emission signal monitoring data through timestamp alignment to generate an acoustic-thermal coupling distribution; extracting local roughness enhancement features from the acoustic-thermal coupling distribution through threshold segmentation to generate a roughness feature overlay map; and determining the surface roughness feature distribution of the tape based on the roughness feature overlay map to provide a basis for subsequent laser energy absorption analysis.Furthermore, the step of calculating the laser energy absorption coefficient distribution and constructing an etching depth prediction model based on the roughness characteristics includes: obtaining the laser energy absorption coefficient distribution through reflection spectral analysis based on the roughness characteristic distribution, and generating an absorption coefficient distribution map; recording power fluctuation data through a laser power signal acquisition device for the absorption coefficient distribution map, and aligning the power fluctuation data with the absorption coefficient distribution map using timestamps to generate a fused data map; extracting local absorption peak regions from the fused data map; if the absorption rate of the peak region is lower than a preset threshold, adjusting the laser wavelength parameters to determine an optimized absorption parameter set; using the optimized absorption parameter set to obtain material thermal conductivity parameters from the fused data map, constructing a numerical simulation model, and generating a correlation curve between the absorption coefficient and the etching depth; and performing mapping calculations through the correlation curve to obtain the etching depth prediction result. Furthermore, the step of obtaining the prediction result and analyzing the equipment power deviation based on the etching depth prediction model includes: obtaining the target area depth deviation from the etching depth prediction result; if the depth deviation exceeds a preset threshold, acquiring a laser power signal through a real-time monitoring device; extracting the fluctuation component from the laser power signal and obtaining the fluctuation frequency spectrum using a spectrum analysis method; obtaining the roughness feature distribution based on the fluctuation frequency spectrum and calculating the correlation coefficient between the fluctuation frequency and the roughness feature using a correlation analysis method; if the correlation coefficient exceeds a preset range, analyzing the degree of deviation between the fluctuation component and the standard power curve using power drift identification technology, determining the equipment power deviation, and generating a deviation compensation value; obtaining the laser beam focus calibration parameters from the deviation compensation value and determining a material surface texture optimization and adjustment scheme. Furthermore, the step of adjusting the focal position and generating a focal offset correction trajectory by means of the equipment power deviation includes: calculating the surface height difference by means of the equipment power deviation and the original roughness distribution of the tape surface, and generating a tape roughness feature sequence; extracting the environmental interference gradient from the tape roughness feature sequence, and generating a fused roughness feature by adding and fusing the equipment power deviation; if the fused roughness feature exceeds a preset threshold, marking the interference area and generating a marked interference gradient map; performing an initial scan on the marked interference gradient map by means of a focal offset correction method to determine an initial focal position offset estimate, and iteratively updating the initial focal position offset estimate until the offset change is less than a preset convergence threshold to obtain an updated focal position offset estimate; identifying potential abnormal areas of the substrate based on historical defect data, adjusting the offset parameters, and generating a preliminary focal offset trajectory; and performing multi-point sampling correction on the preliminary focal offset trajectory by means of smoothing filtering to determine the focal offset correction trajectory.Furthermore, the step of adjusting the laser scanning path parameters and generating dynamic energy output commands based on the focus offset correction trajectory includes: calculating the focus offset based on the laser focus position data and generating a focus offset detection result; calculating the trajectory deviation value based on the focus offset detection result and obtaining a correction trajectory from the original scanning trajectory; adjusting the path parameters and scanning speed using the correction trajectory to generate an adjustment path; calculating the scanning spacing to speed ratio using the adjustment path, extracting the energy density distribution, and determining the compensation distribution calculation result; matching a pre-established compensation strategy library from the compensation distribution calculation result to generate a preliminary energy output command sequence; setting a power threshold using process parameters based on the preliminary energy output command sequence to generate a dynamic power control command; and optimizing the output command sequence by integrating the processing accuracy feedback data with the dynamic power control command to generate the final dynamic energy output command. Furthermore, the step of driving the laser to form a consistent etching depth pattern and verifying the etching result through the dynamic energy output command includes: obtaining the target etching position coordinates and tape tension value through the texture distribution map; performing synchronous control based on the tape tension value to determine the pulse timing and energy level, and generating an energy output command sequence; driving the laser with the energy output command sequence, modulating the beam intensity in real time according to the pulse timing, and forming an initial etching depth pattern at the corresponding position in the texture distribution map; obtaining an etching result image from the initial etching depth pattern through online defect detection; comparing the etching result image with a preset depth threshold to determine the result deviation verification value; extracting deviation features from the result deviation verification value; and if the deviation features exceed the preset threshold, triggering a defect type mapping relationship, determining the defect type, and generating a parameter cyclic adjustment set. Furthermore, the step of driving the laser to form a consistent etching depth pattern and verifying the etching result through the dynamic energy output command includes: compensating for the ambient temperature value of the texture distribution map by combining the parameter cyclic adjustment set; driving the laser through the updated energy output command sequence to form an optimized etching depth pattern; obtaining a verification image for the optimized etching depth pattern again through online quality defect detection; determining the final deviation value by comparing the verification image with a preset depth threshold; if the final deviation value still exceeds the preset threshold, adjusting the intensity gain and timing offset in the parameter cyclic adjustment set to generate a new energy output command sequence; and driving the laser through the new energy output command sequence to ensure that a consistent etching depth pattern is formed at the texture position on the tape surface.Furthermore, the step of extracting local roughness features and determining material uniformity based on the texture distribution map includes: extracting roughness gradient features of local regions from the texture distribution map to generate local gradient distribution data; fusing the local gradient distribution data with thermal distribution information from an infrared thermal image to generate a comprehensive distribution map; calculating a uniformity index of the texture region using the comprehensive distribution map; if the uniformity index is lower than a preset threshold, marking the non-uniform regions to generate a non-uniform region distribution map; integrating acoustic emission signal data through multi-modal data synchronous processing of the non-uniform region distribution map to generate an acoustic-thermal feature distribution; and extracting enhanced roughness features from the acoustic-thermal feature distribution to generate a final roughness feature distribution, providing data support for subsequent laser etching depth prediction.

[0007] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0008] This invention discloses a multimodal data fusion method for controlling the laser etching depth of adhesive tape surfaces. Addressing the problem of etching depth deviation caused by uneven tape surface texture, this method integrates image processing, thermal distribution analysis, acoustic emission signal monitoring, and laser power signal analysis to construct a complete logical chain from texture distribution to etching depth prediction. First, high-resolution image capture and reflected light intensity data are used to generate a texture distribution map and extract roughness features. Infrared thermal images and acoustic emission data are then fused to determine material uniformity. Subsequently, based on numerical simulation of the laser energy absorption coefficient and etching depth, the etching depth is predicted and power deviation is analyzed. To address the deviation problem, focus deviation correction and energy density compensation strategies are employed to dynamically adjust the laser output command, ultimately achieving a consistent etching depth pattern. Parameters are then iteratively optimized through online detection. This invention significantly improves the etching accuracy of adhesive tape surfaces, solves the depth deviation problem caused by uneven texture and equipment fluctuations, and achieves efficient and precise technical results. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a method for controlling the laser etching depth on a tape surface using multimodal data fusion, according to the present invention. Detailed embodiments.

[0010] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0011] like Figure 1 This embodiment of a method for controlling the laser etching depth on a tape surface using multimodal data fusion may specifically include:

[0012] Step S101: Using a high-resolution line scan camera combined with high-speed camera image capture technology, the original image data of the tape surface is acquired, and the material texture area is marked in the image. At the same time, the reflected light intensity data is collected and recorded to assist in the division of texture boundaries, and a texture distribution map of the tape surface is generated.

[0013] A first image of the tape surface is acquired using a high-resolution line scan camera, while a second image is acquired using a high-speed camera. The first and second images are then fused to obtain raw image data. Texture regions are labeled within the raw image data, and light intensity data is obtained by synchronously acquiring reflected light intensity data. This light intensity data is used to assist in texture boundary delineation, resulting in a texture boundary map from the texture regions. Finally, a texture distribution map of the tape surface is generated using the texture boundary map.

[0014] In one embodiment, the tape surface texture detection system uses a high-resolution line scan camera combined with a high-speed camera for image capture. The high-resolution line scan camera continuously scans one-dimensional line images of the moving surface of the tape, and the camera scanning rate is synchronized with the conveyor belt speed on the tape production line, achieving high-precision imaging along the length direction.

[0015] Specifically, the line scan camera works by exposing the sensor array line by line. As the tape passes through the field of view at a constant speed *v*, the exposure time *t* for each line satisfies *v*t = the line spacing, thus stitching together a complete two-dimensional image. This type of camera can achieve a resolution of over 5000 pixels per millimeter, suitable for capturing micron-level texture details of the tape. Simultaneously, a high-speed camera captures static snapshots at a rate of thousands of frames per second, supplementing the potential distortion in the width direction of the line scan. The exposure times of the two cameras are synchronized by an external trigger signal to ensure spatiotemporal alignment of the image data. On the tape production line, the system is fixedly installed above the conveyor belt, the tape width is 50 cm, the running speed is 10 m / min, and the original image data acquired has a resolution of 8000x6000 pixels. This combined approach overcomes the resolution loss problem of a single camera under high-speed movement. Based on the acquired original image data, further material texture region annotation is performed. The annotation process first preprocesses the image to grayscale to highlight the texture contrast of the tape surface.

[0016] Specifically, textured areas exhibit irregular fibrous or granular structures, creating a stark contrast with smooth substrate areas. Operators or semi-automatic software use polygon tools to select texture boundaries, for example, marking raised fibrous texture areas on the tape insulation layer.

[0017] Preferably, an edge detection algorithm, such as a Sobel filter based on gradient operators, is introduced to initially extract the texture edge contours. Then, the set of annotation points is manually fine-tuned to form a vector boundary file. This annotation ensures that the texture area accurately covers more than 80% of the actual texture distribution on the tape surface, avoiding the omission of fine textures.

[0018] It should be noted that the acquisition of reflected light intensity data and image capture are strictly synchronized. The reflected light intensity data is recorded by an integrated photoelectric sensor array, which records the distribution of diffuse reflection intensity of incident light on the tape surface. The photoelectric sensor is placed next to the line scan camera, emitting a uniform white light source to illuminate the tape. The sensor receives the reflected light and converts it into a voltage signal, quantizing the light intensity value to a range of 0 to 5 volts. High light intensity areas correspond to smooth substrates, while low light intensity areas correspond to rough textured areas, because textured surfaces scatter more light.

[0019] In one possible implementation, the sensor resolution matches the image pixels, with each pixel corresponding to a light intensity value, synchronized to the image coordinate system via a timestamp. This data recording assists in texture boundary delineation; for example, when image edges are blurred, abrupt changes in low light intensity gradients serve as boundary cues, improving delineation accuracy by over 95%. During tape movement, data is collected at a sampling rate of 1000 times per second, forming a light intensity sequence of equal length to the image. Further, the labeled texture regions are fused with the reflected light intensity data to refine boundaries. In the fusion step, the light intensity data is superimposed on the image as a weight map, and the texture boundaries are adjusted to light intensity gradient threshold lines; for example, the threshold is set to extend the area below 2 volts by 5 pixels. This auxiliary mechanism is particularly suitable for scenarios with contaminated tape surfaces or uneven lighting, ensuring boundary accuracy.

[0020] In one embodiment, for a PVC tape production line, the system acquires images of a 10-meter-long section of tape. The reflected light intensity shows an average intensity of 1.2 volts in the textured area and 3.5 volts in the substrate area. After annotation, the texture coverage reaches 92%, and the generated distribution map clearly shows the densely textured bands.

[0021] Specifically, the process of generating the tape surface texture distribution map integrates the above data. First, the marked areas are converted into binary masks, with textured areas assigned a value of 1 and non-textured areas assigned a value of 0. Then, a normalized light intensity map is overlaid, and the texture density value is calculated as the mask multiplied by (1 - light intensity / maximum light intensity). Finally, the distribution map is rendered using a heatmap, with the color gradually changing from blue (low density) to red (high density). This map visually reflects the uniformity of the tape texture and is used for quality inspection.

[0022] For example, in another embodiment, for PET substrate tape, the high-speed camera frame rate is increased to 5000 frames per second to capture transient texture vibrations. Light intensity data assists in defining dynamic boundaries, and the distribution map shows a texture distribution deviation of less than 2%, supporting online monitoring in mass production.

[0023] Understandably, this technical solution enables visualization of texture distribution in the tape manufacturing field, improving inspection efficiency. Through multimodal data fusion, the accuracy of texture boundary segmentation is significantly improved, making it suitable for production quality inspection of different tape types, such as cloth-based tapes or paper-based tapes.

[0024] Preferably, in system integration, data processing uses an embedded processor to output a distribution map in real time, with a processing time of less than 1 second per meter of tape, ensuring continuous and uninterrupted operation of the production line.

[0025] Step S102: Based on the surface texture distribution map of the tape, the local roughness gradient distribution is extracted using image processing technology, and the thermal distribution information obtained from the infrared thermal image is fused to determine the material uniformity of the texture area. If the uniformity is lower than a preset threshold, the acoustic emission signal monitoring data is integrated through multimodal data synchronous processing technology to determine the surface roughness feature distribution of the tape.

[0026] Based on the surface texture distribution map of the adhesive tape, the Sobel operator is used to extract the local roughness gradient distribution, resulting in the gradient distribution map. Thermal distribution information obtained from infrared thermal images is fused to the gradient distribution map, and a uniformity statistical index of the texture region is calculated. If the uniformity statistical index is lower than a preset threshold, a non-uniform texture region is identified. For the non-uniform texture region, acoustic emission signal monitoring data is integrated using timestamp alignment to obtain the acoustic-thermal coupling distribution. From the acoustic-thermal coupling distribution, local roughness enhancement features are extracted through threshold segmentation to generate a roughness feature overlay map. Based on the roughness feature overlay map, the surface roughness feature distribution of the adhesive tape is determined.

[0027] In one implementation, for the surface quality inspection of electrical insulating tape, a surface texture distribution map of the tape is first acquired. This texture map is obtained on the tape production line using a high-resolution industrial camera, recording the microscopic texture details of the tape surface, such as fiber distribution and indentations.

[0028] Specifically, the camera scans the moving tape at 100 frames per second to ensure the image covers the entire detection area. Based on the texture distribution map described above, image processing techniques are used to extract the local roughness gradient distribution. These techniques include grayscale preprocessing and gradient filtering.

[0029] Specifically, the texture map is converted into a grayscale image, and then the Sobel operator is applied to calculate the horizontal and vertical gradient magnitudes, forming a gradient distribution map. The Sobel operator detects edge intensity through convolution kernels; for example, a horizontal kernel emphasizes vertical edges, and a vertical kernel emphasizes horizontal edges, thereby quantifying local roughness changes. The gradient distribution map uses pixel values ​​as roughness gradient values, highlighting uneven texture areas on the tape surface. Through this step, the roughness gradient characteristics of each local area of ​​the tape are obtained, providing basic data for subsequent fusion. Further, thermal distribution information obtained from infrared thermal images is fused. Infrared thermal images are simultaneously acquired by an infrared thermal imager, recording the temperature field on the tape surface, such as differences in the material's thermal response after heating.

[0030] In one possible implementation, image registration is first performed to align the texture gradient map with the thermal image space, and then affine transformation is used to adjust the coordinates based on keypoint matching. The fusion process maps thermal distribution information, such as temperature gradient values, onto the texture gradient map through pixel-level overlay, forming a composite feature map.

[0031] For example, regions with high temperature gradients may correspond to areas of material inhomogeneity. The material uniformity of textured regions is determined based on the fused composite feature map.

[0032] Specifically, a region of interest, such as the central band of the tape width, is selected, and statistical indicators are calculated for the composite image, including the gradient mean and temperature variance. The uniformity index is defined as the weighted average of the inverse variance; if it is lower than a preset threshold, such as 0.8, it indicates that there are defects in the material.

[0033] It should be noted that the threshold is pre-calibrated based on the tape specifications; for example, the uniformity threshold for insulating tape is determined through laboratory testing. This determination allows for a preliminary assessment of the tape material's density.

[0034] In one embodiment, if the uniformity is below a preset threshold, acoustic emission signal monitoring data is integrated using multimodal data synchronization processing technology. The acoustic emission signal is acquired by an AE sensor array positioned above the tape to monitor elastic wave signals generated at microcracks or inhomogeneities. Multimodal synchronization processing first aligns image frames with the AE signal based on timestamps, for example, using the NTP protocol to ensure millisecond-level clock accuracy. After synchronization, AE features such as peak amplitude, duration, and frequency spectrum are extracted and mapped to the low-uniformity texture region.

[0035] Specifically, multimodal data synchronization processing techniques include feature-level fusion and spatiotemporal registration. After noise removal from the AE signal preprocessing, the energy distribution vector is calculated and mapped to image pixels.

[0036] For example, for pixels with low uniformity, the corresponding time window AE energy value is superimposed to form a roughness feature distribution map. This map is represented in three dimensions, with each point containing gradient value, temperature value, and AE energy, achieving a comprehensive characterization of the tape surface roughness. In a production line scenario, this process is executed in real time, supporting defect location.

[0037] For example, in a laboratory verification embodiment of electrical insulating tape, a non-uniform tape sample was first simulated. The gradient distribution was obtained through the texture extraction described above. After fusing the thermal image and determining that the uniformity was below a threshold, AE monitoring was triggered. The AE sensor captured the signal peak corresponding to the crack area. After synchronous fusion, the roughness feature distribution map clearly showed the defect outline. This embodiment verified the applicability of the technology in tape quality control.

[0038] Preferably, in the on-line inspection scenario of the tape production line, the system integrates the above modules to achieve continuous monitoring. After collecting texture and thermal images, it extracts and fuses them for judgment. If a trigger is triggered, it integrates AE data and outputs a roughness distribution report. In this way, the surface roughness characteristics of the tape can be accurately determined, supporting subsequent repair or rejection operations and ensuring the high reliability of the power tape.

[0039] Step S103: Calculate the laser energy absorption coefficient distribution based on the surface roughness characteristics of the tape, and combine the laser power signal acquisition data to construct the correspondence between the absorption coefficient and the etching depth through numerical simulation technology to obtain the etching depth prediction result.

[0040] Based on the surface roughness distribution of the adhesive tape, the laser energy absorption coefficient distribution is obtained using reflectance spectroscopy analysis, resulting in the absorption coefficient distribution map. For this absorption coefficient distribution map, power fluctuation data is recorded using a laser power signal acquisition device, and the power fluctuation data is aligned with the absorption coefficient distribution map using timestamps to obtain a fused data map. From the fused data map, local absorption peak regions are extracted. If the absorption rate of the peak region is lower than a preset threshold, the laser wavelength parameters are adjusted to determine an optimized absorption parameter set. Using the optimized absorption parameter set, material thermal conductivity parameters are obtained from the fused data map, a numerical simulation model is constructed, and a correlation curve between the absorption coefficient and the etching depth is generated. The roughness characteristic distribution is mapped and calculated using this correlation curve to obtain the etching depth prediction result.

[0041] In one implementation, the laser energy absorption coefficient distribution is first calculated for the surface roughness characteristic distribution of the tape. This distribution is based on micro-texture data of the tape surface, such as roughness feature maps collected on an electrical insulating tape production line.

[0042] Specifically, the laser energy absorption coefficient reflects a material's ability to absorb laser light and is affected by surface roughness; for example, rough areas may lead to increased energy scattering. The calculation process involves scanning the tape surface, using a laser sensor to measure the intensity difference between the incident and reflected laser light, and thus deriving the absorption coefficient value for each local area.

[0043] For example, a grid is divided along the width of the tape, and an absorption formula is applied to each grid point to obtain a coefficient distribution map. This step provides the basis for subsequent data integration, ensuring that the detection process focuses on the energy response differences at points of surface inhomogeneity. Further, data is acquired using laser power signals. This data is collected in real time using a power meter to record fluctuations in the laser source output power, such as maintaining the laser power within a fixed range during tape inspection.

[0044] It should be noted that the laser power signal acquisition data is used to calibrate the absorption coefficient distribution and avoid calculation errors caused by power instability.

[0045] Specifically, aligning the time series of the power signal with the absorption coefficient map, for example by using a synchronous clock to match the acquisition time, creates an enhanced distribution map. In the quality control of electrical insulation tape, this combination helps identify the impact of power variations on surface absorption, ensuring data accuracy.

[0046] In one possible implementation, the relationship between the absorption coefficient and the etching depth is established using numerical simulation techniques. Numerical simulation is a method used to predict the behavior of laser etching processes, such as reproducing the interaction between a laser and adhesive tape materials in a computer environment. In principle, this technique is based on the finite element analysis framework, simulating the propagation and absorption of laser energy under varying surface roughness.

[0047] Specifically, a digital model of the adhesive tape material is first established, including its surface roughness characteristics and absorption coefficient values. Then, laser power data is input for iterative calculations. During the process, the simulation engine progressively calculates the amount of material removed due to energy deposition, i.e., the etching depth.

[0048] For example, boundary conditions are set for regions with different absorption coefficients, simulation loops are run, and the mapping curve between depth values ​​and coefficients is output. This correspondence is represented in tabular or functional form, supporting quick lookup. In the scenario of electrical insulation tape inspection, this construction process can simulate laser processing on a production line, revealing the impact of roughness on etching uniformity. Through multiple iterations, model parameters, such as the absorption threshold, are optimized to ensure the reliability of the correspondence.

[0049] Preferably, in the tape testing laboratory embodiment, the above correspondence is applied to obtain the etching depth prediction result.

[0050] Specifically, the actual absorption coefficient distribution and power data are input into the simulation model to calculate the predicted depth value for each surface point.

[0051] For example, in regions with uneven roughness, the prediction results show that the depth deviation is greater than the standard value, indicating potential defects. The results are output in the form of a heatmap for easy visualization and analysis.

[0052] For example, in another implementation, when calculating the absorption coefficient distribution of electrical insulating tape on a continuous production line, multi-wavelength lasers can be integrated to cover different surface texture types. By incorporating power signals, numerical simulations further incorporate environmental factors such as temperature effects, building a more robust correlation and thus obtaining accurate etching depth predictions, supporting real-time quality adjustments.

[0053] Understandably, the logic of this process starts with roughness features, gradually incorporating data and simulations to ensure the consistency of the prediction results.

[0054] In one embodiment, if the absorption coefficient distribution exhibits high variability, a mesh refinement step is added to the simulation to improve the accuracy of depth prediction. Furthermore, in extended scenarios of tape surface inspection, such as batch quality verification, parallel simulation technology can be used to process large amounts of data samples and obtain comprehensive prediction results when constructing the correspondence. This approach enhances the applicability of the technology without changing the core inspection field.

[0055] Step S104: If the etching depth prediction result shows that the depth deviation of the target area exceeds the preset threshold, the fluctuation component is extracted from the real-time acquired laser power signal, and the correlation between the fluctuation and the roughness feature distribution is analyzed by power drift identification technology to determine the power deviation of the equipment.

[0056] The depth deviation of the target area is obtained from the etching depth prediction results. If the depth deviation exceeds a preset threshold, a laser power signal is acquired through a real-time monitoring device. A fluctuation component is extracted from the laser power signal, and a Fourier transform is used to obtain the fluctuation frequency spectrum. For the fluctuation frequency spectrum, the roughness feature distribution is obtained, and the Pearson correlation coefficient between the fluctuation frequency and the roughness feature is calculated using a correlation analysis method to determine the correlation coefficient. If the correlation coefficient exceeds a preset range, the degree of deviation between the fluctuation component and the standard power curve is analyzed using power drift identification technology to determine the equipment power deviation and obtain a deviation compensation value. Laser beam focus calibration parameters are obtained from the deviation compensation value to determine the material surface texture optimization adjustment, resulting in the adjusted equipment power deviation.

[0057] In one implementation, when the etching depth prediction result shows that the depth deviation of the target area exceeds a preset threshold, the system initiates a real-time monitoring and analysis process of the laser power signal.

[0058] Specifically, the process begins by acquiring real-time laser power signals using power sensors mounted on the laser equipment. These signals record power changes in a time-series format; for example, in semiconductor wafer etching scenarios, the sensor collects hundreds of power data points per second to capture minute fluctuations. Further, the fluctuation components are extracted from the acquired laser power signals. This can be achieved using digital signal processing methods, such as employing a high-pass filter to remove low-frequency trend components from the signal while retaining high-frequency fluctuation components.

[0059] In one possible implementation, the fluctuation component is defined as a sequence of deviations of the power signal relative to its mean.

[0060] For example, if the power signal sequence is P(t), then the fluctuation component F(t) = P(t) - average(P). This extracted fluctuation reflects the instability of the laser output. In the etching of precision optical components, this extraction helps identify instantaneous power jumps. Power drift identification techniques are used to analyze the correlation between these fluctuation components and the roughness feature distribution.

[0061] It should be noted that power drift identification technology is an analytical framework based on statistical models. It quantifies the degree of drift by calculating the statistical characteristics of the fluctuation sequence, such as variance or spectral density.

[0062] Specifically, the analysis first involves acquiring the roughness characteristic distribution of the target region. These characteristics can be obtained through atomic force microscopy, forming a roughness distribution vector, which includes parameters such as average roughness Ra and peak-to-valley difference. Then, the correlation between the fluctuation component and the roughness vector is calculated, and the Pearson correlation coefficient formula is used to evaluate the linear relationship between the two. If the correlation coefficient exceeds 0.7, it indicates that power fluctuations significantly affect the roughness distribution. When applied to solar panel etching scenarios, this analysis can reveal how power drift leads to surface inhomogeneity.

[0063] For example, when determining the power deviation of the equipment, the system calculates the deviation based on the above correlation analysis results.

[0064] For example, if correlation analysis shows that the fluctuation amplitude is positively correlated with the roughness deviation, the power deviation can be estimated using a linear regression model, simplified to D = k * Var(F), where k is an empirical coefficient and Var(F) is the fluctuation variance. This deviation is then used to calibrate laser equipment, ensuring that etching depth is controlled at the nanometer level in integrated circuit manufacturing.

[0065] Preferably, in another embodiment, a multi-channel laser system is considered, and the fluctuation extraction can be extended to the power signal of each channel, while the correlation analysis integrates the multi-channel data to improve the accuracy of the judgment.

[0066] For example, in MEMS device etching, this method can handle uneven power distribution under complex patterns.

[0067] Understandably, through the above steps, the system can adjust the laser parameters in real time to reduce defects caused by depth deviation.

[0068] In one embodiment, tests showed that when the deviation exceeded a threshold, this analysis method controlled the device power deviation within 5%, improving the stability of the etching process. Furthermore, to adapt to different etching materials, such as silicon-based or compound semiconductors, the roughness characteristic distribution parameters can be adjusted according to material properties. For example, peak density is emphasized for silicon, while surface smoothness is considered for gallium arsenide, making power deviation assessment more targeted.

[0069] In one possible implementation, power drift identification technology can also incorporate machine learning elements, such as using support vector machines to classify the correlation between fluctuation patterns and roughness anomalies. This can predict potential deviations and intervene in advance during the etching of high-density memory chips.

[0070] Step S105: By integrating the environmental interference gradient in the surface roughness feature distribution of the tape with the equipment power deviation, the focal position offset estimation is iteratively updated using the focal deviation correction method. At the same time, the potential substrate anomaly identification method is identified by combining the defect-dominant factor analysis technology to obtain the focal offset correction trajectory.

[0071] By collecting the equipment power deviation and the original roughness distribution of the tape surface, the surface height difference of each sampling point in the original roughness distribution is calculated to obtain the tape roughness feature sequence. The environmental interference gradient is extracted from the tape roughness feature sequence and fused with the equipment power deviation to generate a fused roughness feature. If the fused roughness feature exceeds a preset threshold, the interference area is marked, resulting in a marked interference gradient map. An initial scan of the marked interference gradient map is performed using a focus deviation correction method to determine the initial focus position offset estimate. This initial focus position offset estimate is iteratively updated until the offset change is less than a preset convergence threshold, resulting in an updated focus position offset estimate. Based on the defect-dominant factors extracted from historical defect data, potential abnormal areas of the substrate are identified from the updated focus position offset estimate. The offset parameters are adjusted for these potential abnormal areas to obtain a preliminary focus offset trajectory. Multi-point sampling correction is performed on the preliminary focus offset trajectory using smoothing filtering to determine the continuity of the corrected trajectory, resulting in a corrected focus offset trajectory.

[0072] In the tape production and testing system, the equipment power deviation refers to the difference between the actual output power of the laser source and the nominal power, which is obtained by real-time acquisition and calculation through a power sensor.

[0073] Specifically, the sensor monitors the light source current and voltage, and the deviation is defined as the percentage of actual power minus a set power, used to compensate for imaging signal attenuation. In one embodiment, the surface roughness feature distribution of the tape is acquired using a line laser scanner to generate a two-dimensional roughness height map, the feature distribution including mean, variance, and texture statistics. The environmental interference gradient originates from grayscale gradient changes caused by external light fluctuations or vibrations.

[0074] It should be noted that interference components are separated using a high-pass filter to form a gradient vector field. Further, the environmental interference gradient in the tape surface roughness feature distribution is fused using the device power deviation. Specifically, the power deviation is used as a weighting coefficient, multiplied by the interference gradient vector, and then added point-to-point with the roughness feature points to form a fused feature map.

[0075] For example, when the production line is running at high speed, the interference gradient weight is increased when the power deviation increases, achieving dynamic compensation. This fusion ensures that the roughness characteristics are not affected by the environment, improving detection accuracy. Based on the above fusion results, a focus deviation correction method is used to iteratively update the focus position offset estimate. Focus deviation refers to the distance error between the imaging plane and the tape surface, and the correction method is based on the least squares criterion.

[0076] Specifically, the initial offset estimate is derived from the sharpness function of the fused feature map, and then the iterative formula is that the new estimate is equal to the old estimate minus the step size of the gradient direction multiplied by the derivative of the bias, and this is repeated until convergence.

[0077] It should be noted that the step size is dynamically adjusted by the power deviation to avoid oscillation.

[0078] In one possible implementation, the number of iterations is set to 10, and the offset is rescanned and verified after each update.

[0079] Preferably, a method for identifying potential substrate anomalies is simultaneously employed, combining defect-dominant factor analysis technology. Defect-dominant factors refer to the main variables leading to surface anomalies, such as uneven substrate thickness or adhesive distribution. The analysis technique extracts the weights of dominant factors through principal component analysis, and then uses thresholds to determine substrate anomalies; for example, a thickness deviation exceeding 5 micrometers is considered a potential anomaly. This method outputs a list of anomaly locations, which is input into the correction iteration to adjust focus priority.

[0080] For example, on the coating line of a conveyor belt, the fusion feature map reveals roughness anomalies, and defect analysis identifies substrate tensile stress as the dominant factor, triggering a focus shift towards these anomalies. Through these steps, a focus shift correction trajectory is obtained—a continuous curve of a series of discrete focus points—used to drive the coke motor to follow the conveyor belt's movement. The trajectory is smoothly generated using spline interpolation to ensure real-time tracking.

[0081] In one embodiment, it is applied to an insulating tape quality inspection production line. The equipment power deviation is obtained from the laser feedback module, fused, and the focus is iteratively corrected. Defect analysis identifies abnormal fiber breakage in the substrate, and trajectory correction improves imaging clarity, suitable for scenarios with winding speeds up to 50 meters per minute. In another embodiment, it is used for pressure-sensitive tape surface inspection. Environmental interference gradients mainly originate from workshop ventilation and lighting. Through fusion compensation, iterative updates of offset estimation are combined with factors dominant to adhesive layer defects, such as abnormal bubble distribution, to generate a trajectory that supports synchronous adjustment of multi-channel scanning heads.

[0082] Understandably, this technical solution provides stable focus tracking when identifying abnormalities in the tape substrate, enabling the detection of surface defects without omission.

[0083] Step S106: Adjust the laser scanning path parameters according to the focus offset correction trajectory, extract the energy density compensation distribution from the correction path, generate a dynamic energy output command sequence through compensation strategy library matching technology, and optimize the output command using the process parameter automatic adjustment function.

[0084] The focus offset is calculated based on the laser focus position data to obtain the focus offset detection result. A trajectory deviation value is then calculated based on the focus offset detection result, and a corrected trajectory is obtained from the original scanning trajectory. The corrected trajectory is used to adjust path parameters and scanning speed to generate an adjusted path. The energy density distribution is extracted by calculating the ratio of scanning distance to speed using the adjusted path, and the compensation distribution calculation result is determined. A pre-established compensation strategy library containing energy compensation modes is matched against the compensation distribution calculation result. If the matching degree exceeds a preset threshold based on the distribution similarity, the corresponding strategy is selected, generating a preliminary energy output command sequence. A power threshold is set based on the preliminary energy output command sequence using process parameters, including layer thickness and material properties, to obtain a dynamic power control command. The dynamic power control command is then fused with processing accuracy feedback data obtained during the processing to optimize the output command sequence and generate the final dynamic energy output command.

[0085] In one implementation, the laser processing system first monitors the laser beam focus shift using an optical sensor. Focus shift refers to the deviation of the laser focus position from the ideal position on the workpiece surface, typically caused by mechanical vibration or thermal deformation.

[0086] Specifically, the system collects Z-axis displacement data, forms an offset curve, and corrects the scanning trajectory based on a spline interpolation algorithm.

[0087] For example, in a metal plate laser cutting scenario, the sensor samples the offset value every millisecond, adjusting the original straight trajectory into a compensation curve to ensure the focus always falls on the processing surface. This corrected trajectory not only smooths path curvature but also provides redundancy margins to avoid edge burns. Through this step, the system achieves trajectory deviation precision control of less than 0.1 mm, providing basic data for subsequent path parameter adjustments. Based on the above corrected trajectory, the laser scanning path parameters are further adjusted. Path parameters include scanning speed, acceleration, and overlap ratio.

[0088] Specifically, the system calculates a velocity mapping function based on the trajectory curvature. For example, when the radius of curvature is less than 5 mm, the velocity is reduced by 20% to maintain a constant heat input. At the same time, the overlap rate is adaptively adjusted to 15% to 25% based on the trajectory spacing.

[0089] In one possible implementation, this adjustment is achieved through a parameter lookup table, ensuring a smooth path transition. Extracting the energy density compensation distribution from the corrected path is a crucial step. The energy density compensation distribution refers to the non-uniform energy distribution per unit area caused by path deformation.

[0090] Specifically, the system integrates the laser power and dwell time along the corrected trajectory to generate a two-dimensional density heat map, where the high-density area corresponds to the dense part of the trajectory, and the low-density area requires compensation and energy enhancement.

[0091] It should be noted that the extraction process first discretizes the trajectory into a set of N points, and calculates the local density ρ = P / (v·w) at each point, where P is the power, v is the velocity, and w is the spot width. Then, the distribution is smoothed using Gaussian filtering to form a compensation matrix. This distribution quantifies the energy deviation; for example, in complex curved paths, the density can increase by 30% at corners, thus guiding subsequent compensation.

[0092] Preferably, this distribution is stored in vector field form, supporting fast lookup. This step allows the system to accurately capture processing unevenness, laying the data foundation for dynamic instruction generation. A dynamic energy output instruction sequence is generated using a compensation strategy library matching technique. The compensation strategy library is a pre-built database containing compensation rules for various typical distributions, such as power modulation templates based on historical processing data. The matching technique employs a cosine similarity algorithm, comparing the extracted compensation distribution with the strategy vectors in the library and selecting the strategy with the highest similarity.

[0093] Specifically, each strategy in the library includes a power sequence, pulse width, and frequency adjustment parameters. For example, the "attenuation strategy" for high-density corner areas gradually reduces the power by 15%. The matching process is divided into two stages: first, coarse matching of cluster labels, and second, fine vector comparison, with a threshold set above 0.85.

[0094] In one embodiment, for the helical path of laser cutting metal tubing, the system matches a "rotation compensation strategy" to generate a sequence of pulse commands, such as pulses with power gradually decreasing from 1000W to 800W. This technology ensures that the command sequence responds in real time to changes in distribution, achieving an energy uniformity improvement of over 20%. The core of this step lies in the library's scalability; users can add strategies based on new materials, supporting system self-learning and updates. Simultaneously, the output commands are optimized using an automatic process parameter adjustment function. Process parameters include material thickness, reflectivity, and ambient temperature.

[0095] Specifically, the system integrates a feedback module to read parameter sensor data in real time and adjusts the command sequence through a proportional-integral controller.

[0096] For example, when the thickness increases by 10%, the pulse width is automatically extended by 5% to compensate for insufficient penetration.

[0097] It's important to note that the adjustment function is based on a rule engine, prioritizing parameters as follows: thickness > reflectivity > temperature, forming a closed-loop optimization. In laser welding scenarios, this function can control weld width fluctuations within 0.05 mm. Furthermore, the optimized output command sequence directly drives the laser power module. Commands are issued in timestamp form, for example, a power value every 0.1 seconds, ensuring synchronous execution. Through this process, the system achieves end-to-end control from trajectory correction to command optimization. For example...

[0098] In one embodiment, the system is applied to laser cutting of stainless steel plates. First, it corrects the wavy trajectory caused by focus offset. Extracting a compensation distribution reveals an edge density that is 15% lower than expected. Matching a "boundary enhancement strategy" from the library, a power sequence of 1200W to 1500W is generated, and the pulse frequency is automatically adjusted based on a plate thickness of 2mm. In this implementation, the cut perpendicularity reaches 0.02mm, demonstrating the versatility of the solution. In another embodiment, the system is applied to laser welding of aluminum alloy tubing. After correcting the spiral trajectory, distribution extraction reveals a high-density area in the weld bead. Matching a "pulse modulation strategy," the system optimizes the command to adapt to high reflectivity characteristics and automatically adjusts the power upper limit. In this scenario, the weld porosity is reduced to below 1%, demonstrating the applicability of the technology in welding.

[0099] Understandably, these implementation methods are all limited to the field of laser processing, covering cutting and welding scenarios. Through multi-scenario verification, the technical solutions demonstrate flexibility and robustness.

[0100] In step S107, the laser is driven by a dynamic energy output command sequence to modulate the beam intensity in real time and form a consistent etching depth pattern at the corresponding position of the texture distribution map on the tape surface. At the same time, the etching result is verified by online quality defect detection technology. If a deviation is detected, the defect type mapping relationship analysis is triggered, and the parameters are adjusted cyclically.

[0101] The target etching location coordinates and tape tension value are obtained by mapping the tape texture distribution. Based on the tape tension value, synchronous control is performed to determine the pulse timing and energy level, resulting in an energy output command sequence. This energy output command sequence drives the laser, modulating the beam intensity in real time according to the pulse timing to form an initial etching depth pattern at the corresponding position on the tape texture distribution map. Online defect detection is used to obtain etching result images from the initial etching depth pattern. The images are compared to a preset depth threshold to determine a deviation verification value. Deviation features are extracted from the deviation verification value. If the deviation features exceed the preset threshold, a defect type mapping relationship is triggered. The defect type is determined by matching the deviation features to a preset mapping table, resulting in a parameter cyclic adjustment set, including intensity gain and timing offset. The energy output command sequence is updated by compensating for the environmental temperature value of the tape texture distribution map using this parameter cyclic adjustment set, driving the laser to form an optimized etching depth pattern, achieving a consistent etching depth pattern across the tape surface texture.

[0102] In one implementation, the laser is driven by a dynamic energy output command sequence to achieve real-time modulation of the beam intensity.

[0103] Specifically, the instruction sequence is generated based on a preset energy distribution model.

[0104] For example, the surface texture map of the adhesive tape is converted into a digital instruction set, where each instruction corresponds to an energy output value at a specific location. These instructions are sent by a control system, and the laser adjusts its output power to ensure that the beam intensity changes within milliseconds to match the texture pattern requirements. This method enables precise control during tape processing, avoiding inconsistent etching caused by uneven energy distribution. Furthermore, in the process of forming a consistent etching depth pattern at the corresponding locations on the adhesive tape surface texture map, the texture map is first acquired. This map, generated by an optical scanning device, contains the two-dimensional coordinates of the tape surface and the expected depth information. The laser applies a beam at the corresponding locations according to the instruction sequence to form the etching pattern.

[0105] For example, on an adhesive tape production line, a texture map might represent the microscopic patterns of the bonding area. Laser etching ensures uniform depth within a specified range, such as 0.1 to 0.5 millimeters, thereby improving the tape's adhesion and durability. This pattern formation relies on the laser's positioning accuracy and energy modulation to ensure consistent etching depth across all locations.

[0106] It should be noted that the online defect detection technology is used to verify the etching results. This technology uses a high-resolution camera or laser scanner to capture real-time images of the etched tape surface and compares them with a preset texture distribution map. The specific process includes image preprocessing, such as noise reduction and edge extraction, followed by calculation of the depth deviation value. If the deviation exceeds a threshold, for example, a depth difference greater than 0.05 mm, it is determined to be a defect. This online detection is integrated into the processing line, enabling continuous monitoring during tape movement and improving production efficiency.

[0107] In one possible implementation, if a deviation is detected, a defect type mapping analysis is triggered. This analysis is based on a predefined mapping database.

[0108] For example, deviation types can be classified as excessive etching due to excessive energy or pattern misalignment caused by positioning shift. Mapping relationships are established using historical data, involving the analysis of the correspondence between deviation patterns and parameters, such as a functional model of the relationship between energy output values ​​and depth.

[0109] Specifically, the analysis process first extracts deviation features, such as location coordinates and depth differences, and then queries the database to match defect types, thereby identifying the root cause. This mapping helps to quickly locate the source of the problem, ensuring stable product quality in the tape processing industry.

[0110] Preferably, the cyclic adjustment parameters are performed after the defect type mapping relationship analysis.

[0111] For example, to address defects caused by excessive etching, the system automatically reduces the energy output value in the corresponding instruction sequence and re-drives the laser for local correction. This adjustment process forms a closed loop, for example, by iteratively optimizing parameters until the detection results meet the standards. This cycle can be repeated multiple times until the deviation is eliminated. In tape manufacturing scenarios, this mechanism reduces scrap rates and enables continuous production.

[0112] For example, in a practical application of tape surface etching, assuming the texture distribution map indicates a grid-like pattern and the initial modulated beam intensity of the laser is at a medium level, if uneven grid depths are found after inspection, and mapping analysis shows this is caused by energy fluctuations, then the timing parameters of the command sequence are adjusted to make the next etching more precise. This implementation demonstrates the versatility of the technology and is applicable to processing tapes of varying thicknesses.

[0113] Understandably, the integration of the above processes ensures the precision and efficiency of the etching.

[0114] In one embodiment, the entire system includes a control module, a laser module, and a detection module. The control module generates a sequence of instructions, the laser module performs modulation, and the detection module verifies the results and triggers adjustments. This modular design facilitates scalable application on tape production lines. Furthermore, the detailed process of defect type mapping analysis involves multi-step matching. First, deviation data, such as differences in grayscale values ​​in an image representing depth variations, is collected. Then, this data is input into a mapping model, which classifies based on a rule base.

[0115] For example, if the deviation is concentrated in the edge area, it is mapped as a positioning defect. After analysis, adjustment signals are generated, such as modifying the amplitude of the energy output command. This detailed analysis ensures the accuracy of the feedback and reduces manual intervention in tape processing. In another implementation, the cyclic adjustment parameters can be combined with real-time feedback sensors, such as using an infrared sensor to monitor the etching temperature, to assist in deviation detection. If an abnormal temperature causes a depth deviation, the system confirms the type through mapping analysis and cyclically optimizes the laser power parameters. This approach expands the applicability of the technology, such as processing tapes of different materials.

[0116] For example, in the continuous processing of adhesive tape rolls, the texture distribution map is dynamically updated, and the laser modulates the beam accordingly to form a consistent pattern. Upon detecting a deviation, the mapping analysis responds quickly, and cyclical adjustments ensure overall quality. This implementation highlights the real-time nature and reliability of the technology.

[0117] The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the above-described embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for controlling the laser etching depth on a tape surface using multimodal data fusion, characterized in that, include: The process involves: acquiring raw image data of the tape surface and generating a texture distribution map; acquiring a first image of the tape surface using a high-resolution line scan camera and a second image using a high-speed camera, fusing the first and second images to obtain raw image data; annotating texture regions in the raw image data and simultaneously collecting reflected light intensity data to generate light intensity recording data; using the light intensity recording data to help delineate texture boundaries and generating a texture boundary map from the texture regions; generating a texture distribution map of the tape surface using the texture boundary map to ensure that the texture distribution map reflects the texture feature distribution of the tape surface; if there are outliers in the light intensity recording data, correcting the texture boundary map to obtain an optimized texture distribution map; using the optimized texture distribution map to provide basic data for subsequent roughness feature extraction; extracting local roughness features and determining material uniformity based on the texture distribution map; calculating the laser energy absorption coefficient distribution based on the roughness features and constructing an etching depth prediction model; obtaining prediction results based on the etching depth prediction model and analyzing equipment power deviation; adjusting the focus position based on the equipment power deviation and generating a focus offset correction trajectory; adjusting the laser scanning path parameters based on the focus offset correction trajectory and generating dynamic energy output commands. The laser is driven by the dynamic energy output command to form a consistent etching depth pattern and verify the etching results.

2. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of extracting local roughness features and determining material uniformity based on the texture distribution map includes: extracting local roughness gradient distribution using a gradient operator based on the texture distribution map to generate a gradient distribution map; fusing thermal distribution information obtained from infrared thermal images into the gradient distribution map to calculate the uniformity statistical index of the texture region; if the uniformity statistical index is lower than a preset threshold, locating non-uniform texture regions; for the non-uniform texture regions, integrating acoustic emission signal monitoring data through timestamp alignment to generate an acoustic-thermal coupling distribution; extracting local roughness enhancement features from the acoustic-thermal coupling distribution through threshold segmentation to generate a roughness feature overlay map; and determining the surface roughness feature distribution of the tape based on the roughness feature overlay map to provide a basis for subsequent laser energy absorption analysis.

3. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of calculating the laser energy absorption coefficient distribution and constructing an etching depth prediction model based on the roughness characteristics includes: obtaining the laser energy absorption coefficient distribution through reflection spectral analysis based on the roughness characteristic distribution, and generating an absorption coefficient distribution map; recording power fluctuation data through a laser power signal acquisition device for the absorption coefficient distribution map, and aligning the power fluctuation data with the absorption coefficient distribution map using timestamps to generate a fused data map; extracting local absorption peak regions from the fused data map; if the absorption rate of the peak region is lower than a preset threshold, adjusting the laser wavelength parameters to determine an optimized absorption parameter set; using the optimized absorption parameter set to obtain material thermal conductivity parameters from the fused data map, constructing a numerical simulation model, and generating a correlation curve between the absorption coefficient and the etching depth; and performing mapping calculations through the correlation curve to obtain the etching depth prediction result.

4. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of obtaining prediction results and analyzing equipment power deviation based on the etching depth prediction model includes: obtaining the target area depth deviation from the etching depth prediction results; if the depth deviation exceeds a preset threshold, acquiring laser power signals through a real-time monitoring device; extracting fluctuation components from the laser power signals and obtaining the fluctuation frequency spectrum using a spectrum analysis method; obtaining the roughness feature distribution based on the fluctuation frequency spectrum and calculating the correlation coefficient between the fluctuation frequency and the roughness features using a correlation analysis method; if the correlation coefficient exceeds a preset range, analyzing the degree of deviation between the fluctuation components and the standard power curve using power drift identification technology, determining the equipment power deviation, and generating a deviation compensation value; obtaining laser beam focus calibration parameters from the deviation compensation value and determining a material surface texture optimization and adjustment scheme.

5. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of adjusting the focal position and generating a focal offset correction trajectory by means of the equipment power deviation includes: calculating the surface height difference by means of the equipment power deviation and the original roughness distribution of the tape surface, and generating a tape roughness feature sequence; extracting the environmental interference gradient from the tape roughness feature sequence, and generating a fused roughness feature by adding and fusing the equipment power deviation; if the fused roughness feature exceeds a preset threshold, marking the interference area and generating a marked interference gradient map; performing an initial scan of the marked interference gradient map by means of a focal offset correction method to determine an initial focal position offset estimate, and iteratively updating the initial focal position offset estimate until the offset change is less than a preset convergence threshold to obtain an updated focal position offset estimate; identifying potential abnormal areas of the substrate based on historical defect data, adjusting the offset parameters, and generating a preliminary focal offset trajectory; and performing multi-point sampling correction on the preliminary focal offset trajectory by means of smoothing filtering to determine the focal offset correction trajectory.

6. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of adjusting the laser scanning path parameters and generating dynamic energy output commands based on the focus offset correction trajectory includes: calculating the focus offset based on the laser focus position data and generating a focus offset detection result; calculating the trajectory deviation value based on the focus offset detection result and obtaining a correction trajectory from the original scanning trajectory; adjusting the path parameters and scanning speed using the correction trajectory to generate an adjustment path; calculating the scanning spacing to speed ratio using the adjustment path, extracting the energy density distribution, and determining the compensation distribution calculation result; matching a pre-established compensation strategy library from the compensation distribution calculation result to generate a preliminary energy output command sequence; setting a power threshold using process parameters based on the preliminary energy output command sequence to generate a dynamic power control command; and optimizing the output command sequence by integrating the processing accuracy feedback data with the dynamic power control command to generate the final dynamic energy output command.

7. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of driving the laser to form a consistent etching depth pattern and verifying the etching result through the dynamic energy output command includes: obtaining the target etching position coordinates and tape tension value through the texture distribution map; performing synchronous control based on the tape tension value to determine the pulse timing and energy level, and generating an energy output command sequence; driving the laser with the energy output command sequence, modulating the beam intensity in real time according to the pulse timing, and forming an initial etching depth pattern at the corresponding position on the texture distribution map; obtaining an etching result image from the initial etching depth pattern through online defect detection; comparing the etching result image with a preset depth threshold to determine the result deviation verification value; extracting deviation features from the result deviation verification value; and if the deviation features exceed the preset threshold, triggering a defect type mapping relationship, determining the defect type, and generating a parameter cyclic adjustment set.

8. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of driving the laser to form a consistent etching depth pattern and verifying the etching result by the dynamic energy output command includes: compensating for the ambient temperature value of the texture distribution map by combining the parameter cyclic adjustment set; driving the laser with the updated energy output command sequence to form an optimized etching depth pattern; obtaining a verification image for the optimized etching depth pattern again through online quality defect detection; determining the final deviation value by comparing the verification image with a preset depth threshold; if the final deviation value still exceeds the preset threshold, adjusting the intensity gain and timing offset in the parameter cyclic adjustment set to generate a new energy output command sequence; and driving the laser with the new energy output command sequence to ensure that a consistent etching depth pattern is formed at the texture position on the tape surface.

9. The method for surface texture processing and laser etching depth control of adhesive tape as described in claim 1, characterized in that, The step of extracting local roughness features and determining material uniformity based on the texture distribution map includes: extracting roughness gradient features of local regions from the texture distribution map to generate local gradient distribution data; fusing the local gradient distribution data with thermal distribution information from an infrared thermal image to generate a comprehensive distribution map; calculating the uniformity index of the texture region using the comprehensive distribution map; if the uniformity index is lower than a preset threshold, marking the non-uniform regions to generate a non-uniform region distribution map; integrating acoustic emission signal data through multi-modal data synchronous processing of the non-uniform region distribution map to generate an acoustic-thermal feature distribution; and extracting enhanced roughness features from the acoustic-thermal feature distribution to generate a final roughness feature distribution, providing data support for subsequent laser etching depth prediction.

Images