A vacuum film stripping efficiency evaluation method based on intelligent algorithm

Abstract

This application provides a method for evaluating vacuum demolding efficiency based on intelligent algorithms, comprising: comprehensively probing the surface of a plastic substrate, acquiring surface roughness parameters and texture directionality characteristic data, determining the specific patterns of microscopic peak-valley distribution and texture density distribution, and obtaining a texture feature distribution map of the substrate surface; determining, based on the spatial distribution parameters of adhesive strength, whether there are areas of uneven adhesive strength caused by excessively high proportion of local defects; if areas of uneven adhesive strength are detected, determining an optimal combination of peeling speed influencing parameters and peeling force fluctuation values ​​during the demolding process by combining surface energy differences and texture density distribution data; analyzing the influence of residual adhesive ratio and fragmentation accumulation effect on demolding quality, and determining whether the degree of fragmentation exceeds an acceptable range.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for evaluating vacuum demolding efficiency based on intelligent algorithms. Background Technology

[0002] In modern manufacturing, surface treatment technology for plastic products is a crucial area, directly impacting product quality and market competitiveness. Especially in the film removal process, improving efficiency and quality not only affects production costs but also the final product's appearance and functionality, making it a critical issue that cannot be ignored in the industry. With the introduction of intelligent technologies, research in this field is gradually moving towards greater efficiency and precision. However, current mainstream methods for evaluating film removal efficiency still have significant limitations. These methods often rely too heavily on fixed standards and single testing methods, making it difficult to adapt to the changing material properties and process conditions in complex production environments. Especially when dealing with plastic substrates with different surface textures, the evaluation results are often distorted. This is because fixed standards cannot reflect the actual impact of surface microstructure differences on film removal behavior and lack responsiveness to dynamic changes. This mismatch between fixed standards and dynamic process requirements leads to insufficient precision in quality control during production, thus affecting the optimization of subsequent processes. In-depth analysis of the challenges in this field reveals, through statistical analysis of different batches of production data and laboratory comparison verification, that the surface texture characteristics of the plastic substrate are the core factor affecting film adhesion and removal effectiveness. Differences in surface texture directly alter the adhesive strength distribution between the film and the substrate. Especially during demolding, delamination may occur, making the gradient change in adhesive strength unpredictable. This is because differences in surface texture depth and density lead to uneven stress distribution at the microscopic level, resulting in significant differences in the failure threshold of different regions under external force. This unevenness further exacerbates film fragmentation, making the evaluation criteria for demolding quality complex and unstable. Accurately assessing and adjusting quality standards in dynamic changes has become a pressing technical challenge. Therefore, how to dynamically adjust the grading criteria for demolding quality using intelligent methods to adapt to the impact of adhesive strength gradient distribution and fragmentation caused by changes in surface texture is a problem that needs to be solved. Summary of the Invention

[0003] This invention provides a method for evaluating vacuum demolding efficiency based on intelligent algorithms, mainly including:

[0004] The surface of a plastic substrate is probed to obtain surface roughness parameters and texture directionality feature data, determine the specific patterns of microscopic peak and valley distribution and texture density distribution, and obtain a texture feature distribution map of the substrate surface.

[0005] Based on the texture feature distribution map, local defect ratio and surface energy difference data are extracted, the influence of texture periodic changes on film bonding strength is analyzed, and the spatial distribution parameters of bonding strength on the substrate surface are determined.

[0006] Based on the spatial distribution parameters of the bond strength, determine whether there are areas of uneven bond strength.

[0007] If areas with uneven adhesive strength are detected, the optimal combination of peel speed influencing parameters and peel force fluctuation values ​​during the demolding process is determined by combining surface energy difference and texture density distribution data.

[0008] Based on the optimized combination of the peeling speed influence parameters and peeling force fluctuation values, the layer peeling frequency and peeling position distribution during the decoction process are monitored in real time to generate the distribution characteristics of fragment size range and fragmentation boundary features.

[0009] Analyze the impact of residual adhesive ratio and fragmentation accumulation effect on demolding quality, and determine whether the degree of fragmentation exceeds the acceptable range;

[0010] If the degree of fragmentation exceeds the acceptable range, adjust the demolding control precision parameters and determine the demolding process execution plan;

[0011] Based on the aforementioned demolding process execution plan, a quality prediction calibration is performed to obtain demolding quality evaluation indicators.

[0012] Furthermore, the process of probing the surface of the plastic substrate, acquiring surface roughness parameters and texture directionality feature data, determining the specific patterns of microscopic peak-valley distribution and texture density distribution, and obtaining a texture feature distribution map of the substrate surface includes:

[0013] The process involves collecting surface height values ​​and location coordinates of the substrate, calculating the height difference between adjacent points, and generating topographic data containing height and gradient distributions. Based on this topographic data, the arithmetic mean roughness and root mean square roughness are calculated, spatial frequency components are extracted, the dominant texture direction is determined, and a roughness parameter matrix and texture direction angle map are generated. Using the roughness parameter matrix and texture direction angle map, the peak-valley arrangement direction is identified, the peak-valley spacing and depth distribution are calculated, the number of texture direction lines is counted, and peak-valley distribution features and texture density values ​​are generated. Finally, based on the peak-valley distribution features and texture density values, a texture classification database is matched to determine the texture category, and a texture feature distribution map annotating the texture type and density is generated.

[0014] Furthermore, based on the texture feature distribution map, the local defect ratio and surface energy difference data are extracted, the influence of texture periodic changes on the film adhesion strength is analyzed, and the spatial distribution parameters of the adhesion strength on the substrate surface are determined, including:

[0015] Based on the texture feature distribution map, regions deviating from the texture type are identified as local defects. The proportion of defect points is statistically analyzed, surface energy values ​​are calculated, and a local defect proportion matrix and surface energy distribution data are generated. Based on the local defect proportion matrix and the surface energy distribution data, texture period parameters are extracted, the contact area ratio is calculated, and the initial bonding strength value is determined. Based on the initial bonding strength value and the surface energy difference, the bonding strength is corrected, and spatial distribution parameters of bonding strength are generated.

[0016] Furthermore, determining whether there are regions of uneven bond strength based on the spatial distribution parameters of the bond strength includes:

[0017] Based on the spatial distribution parameters of the bonding strength and the texture feature distribution map, the load per unit area at the peak point is calculated, and an interface stress distribution map is generated. Based on the interface stress distribution map, the stress standard deviation and uniformity coefficient are calculated, stress concentration points are identified, and a stress uniformity coefficient matrix and stress concentration area markers are generated. Based on the stress uniformity coefficient matrix and the stress concentration area markers, the location and range of the bonding strength unevenness area are determined.

[0018] Furthermore, if uneven adhesion strength is detected, the optimal combination of peel speed influencing parameters and peel force fluctuation values ​​during the demolding process is determined by combining surface energy difference and texture density distribution data, including:

[0019] Based on the uneven bonding strength region, the process parameter adjustment library is queried to extract process parameters that match the surface energy difference and texture density distribution, generating a preliminary process parameter set; based on the preliminary process parameter set, the theoretical value of peel force and speed adjustment coefficient are calculated to generate a demolding process correction scheme; based on the demolding process correction scheme, the peel speed and peel force are optimized, and the combination of peel speed influencing parameters and peel force fluctuation values ​​is determined.

[0020] Furthermore, the step of optimizing the combination of the peeling speed influence parameters and peeling force fluctuation values, and real-time monitoring of the delamination frequency and peeling location distribution during the decoction process to generate the distribution characteristics of fragment size range and fragmentation boundary features includes:

[0021] Based on the combination of the peeling speed influence parameters and the peeling force fluctuation value, vibration signals are collected, and the distribution of layer peeling frequency and peeling location is statistically analyzed. Based on the vibration signals, spectral features are extracted, the fragment generation time is identified, and the initial fragment size is calculated. Based on the initial fragment size, signal abrupt change points are analyzed, boundary flatness is determined, and fragment size range and fragmentation boundary features are generated.

[0022] Furthermore, the analysis of the impact of residual adhesion ratio and fragmentation accumulation effect on demolding quality, and the determination of whether the degree of fragmentation exceeds the acceptable range, includes:

[0023] Identify fragmented boundary contours and extract fragment geometric parameters; calculate the density of residual bonding points based on the fragment geometric parameters and detect the bonding ratio; calculate the rate of change of cumulative fragmentation effect based on the bonding ratio and determine whether the degree of fragmentation exceeds the acceptable range.

[0024] Furthermore, if the degree of fragmentation exceeds an acceptable range, the demolding control precision parameters are adjusted, and the demolding process execution plan is determined, including:

[0025] Based on the degree of fragmentation, a process control strategy matching the distribution of peeling positions is extracted, peeling force fluctuation characteristics are identified, and peeling control accuracy parameters are generated. Based on the peeling control accuracy parameters, the peeling speed and path are adjusted to generate a peeling process execution plan.

[0026] Furthermore, the step of performing quality prediction calibration based on the demolding process execution plan to obtain demolding quality evaluation indicators includes:

[0027] Based on the demolding process execution plan, the fragmentation accumulation effect and the residual adhesion ratio are matched to calculate the quality characteristic parameters and determine the quality level; based on the quality level, the demolding quality grading standard database is updated, the relationship between peeling speed and quality prediction is calibrated, and demolding quality evaluation indicators are generated.

[0028] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0029] This invention discloses a method for evaluating vacuum demolding efficiency based on intelligent algorithms. It acquires the surface roughness and texture directionality characteristics of the substrate through high-precision laser scanning, and determines the microscopic peak-valley distribution and texture density distribution patterns using a pre-set database. The influence of periodic texture variations on film adhesion strength is analyzed. For detected areas of uneven adhesion strength, this invention dynamically queries a process parameter adjustment library to obtain demolding process correction schemes, monitors the delamination frequency and location distribution during demolding in real time, and analyzes the impact of fragmentation degree on demolding quality. When the fragmentation degree exceeds an acceptable range, this invention extracts an optimization scheme from a pre-set strategy library to determine the final demolding process execution scheme. This invention effectively improves demolding quality and production efficiency through precise analysis of the surface characteristics of the plastic substrate and dynamic optimization of the demolding process. Attached Figure Description

[0030] Figure 1 This is a flowchart of a vacuum demolding efficiency evaluation method based on intelligent algorithms according to the present invention. Detailed Implementation

[0031] 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.

[0032] like Figure 1 This embodiment of a vacuum demolding efficiency evaluation method based on intelligent algorithms may specifically include:

[0033] S101. Comprehensively probe the surface of the plastic substrate, obtain surface roughness parameters and texture directionality characteristic data, determine the specific patterns of microscopic peak and valley distribution and texture density distribution, and obtain a texture feature distribution map of the substrate surface.

[0034] A laser scanner is used to scan the surface of a plastic substrate point by point at a preset sampling interval, recording the height value and position coordinates of each sampling point. The local gradient is calculated based on the height difference between adjacent sampling points, yielding raw topographic data containing both height and gradient distributions. For this raw topographic data, the arithmetic mean roughness Ra and root mean square roughness Rq of each region are calculated using a moving window method. Simultaneously, the spatial frequency components of the surface are extracted using Fourier transform, and the dominant direction of the texture is determined based on the direction of the dominant frequency component in the spectrum, resulting in a roughness parameter matrix and a texture direction angle map. Based on the Ra and Rq values ​​in the roughness parameter matrix, peaks exceeding the average height and valleys below the average height are identified. The arrangement direction of the peaks and valleys is determined using the texture direction angle map, and the peak-valley spacing and depth distribution are calculated. Simultaneously, the number of lines along the texture direction per unit area is counted, yielding peak-valley distribution characteristics and texture density values. The peak-valley distribution characteristics and texture density values ​​are matched with a pre-established classification database containing standard texture patterns. The closest texture category is determined by calculating the Euclidean distance between the feature to be tested and each standard pattern in the database. Based on the matching results, a substrate surface texture feature distribution map labeled with texture type, density level and orientation is generated.

[0035] Specifically, laser scanning technology plays a crucial role in the inspection of plastic substrate surfaces.

[0036] Specifically, laser scanners acquire surface height information by emitting a laser beam and receiving reflected signals. During the scanning process, the laser beam illuminates the substrate surface point by point at a set sampling interval. When the laser encounters minute undulations on the surface, the time difference in the reflected light accurately reflects the height value at that point. This non-contact measurement method avoids the deformation effects that traditional contact measurements may cause to soft plastic surfaces.

[0037] In one embodiment, acquiring the raw topographic data involves multi-dimensional information collection. In addition to recording the height value of each sampling point, its precise position coordinates on the substrate surface also need to be recorded. The calculation of the local gradient is achieved by comparing the height differences between adjacent sampling points. This gradient information reflects the surface's tilt and trend, providing fundamental data support for subsequent texture orientation analysis.

[0038] It should be noted that the moving window method has unique advantages in calculating roughness parameters. This method calculates the arithmetic mean roughness Ra and root mean square roughness Rq region by region by sliding a fixed-size window across the surface data. Ra reflects the arithmetic mean of the deviation of the surface profile from the mean line, while Rq provides the root mean square statistic of the deviation. The combination of the two can comprehensively characterize the surface roughness. The application of Fourier transform converts the surface topography data in the spatial domain to the frequency domain. By analyzing the dominant frequency components and their directions in the spectrum, the dominant direction of the texture can be accurately identified, which is particularly important for plastic products with regular textures.

[0039] For example, the peak-valley identification process is based on the numerical distribution in the roughness parameter matrix. Peaks are identified by setting a threshold above the average line, and valleys are identified by setting a threshold below the average line. Combining this with the texture direction angle map, the arrangement pattern of these peaks and valleys can be determined. The calculation of the peak-valley spacing reflects the periodicity of the texture, while the depth distribution reflects the severity of surface undulations. The texture density value is obtained by counting the number of lines along the texture direction per unit area; this parameter is directly related to the optical properties and tactile characteristics of the material.

[0040] Preferably, the texture classification database is established based on test data from a large number of standard samples. The database stores feature parameters of various typical texture patterns, including peak-valley distribution patterns, texture density ranges, and directional features. During the matching process, the calculation of Euclidean distance comprehensively considers the differences in multiple feature dimensions. By calculating the multidimensional spatial distance between the test sample and each standard pattern in the database, the most similar texture category can be found. The generated texture feature distribution map not only labels the texture type and density level but also visually displays the spatial distribution of texture directionality through color coding or contour lines.

[0041] S102. Based on the texture feature distribution map, extract the local defect ratio and surface energy difference data, analyze the influence of texture periodic changes on the film bonding strength, and determine the spatial distribution parameters of bonding strength on the substrate surface.

[0042] Based on the peak-valley distribution data and texture type annotations in the texture feature distribution map, regions deviating from the labeled texture type features by more than a preset threshold are identified as local defects. The ratio of defect points in each grid cell to the total number of sampling points is calculated. Simultaneously, the surface free energy of each region is calculated as the product of surface roughness Ra and Rq values ​​to obtain the surface energy value, resulting in a local defect proportion matrix and surface energy distribution data. Using the local defect proportion matrix and surface energy distribution data, the periodicity parameters of the texture are extracted through Fourier transform. The distance between adjacent peaks is calculated to determine the texture period length. The ratio of the actual contact area to the theoretical plane area is calculated based on the peak-valley height difference within the period, resulting in texture periodicity parameters and contact area ratio data. For the texture periodicity parameters and contact area ratio data, if the contact area ratio is lower than a preset threshold, the bonding strength of that region is determined to be low. The initial bonding strength value of each region is calculated by multiplying the contact area ratio by the material's inherent bonding coefficient. The initial bonding strength value is corrected by the difference in surface energy values ​​of adjacent regions, determining the spatial distribution parameters of bonding strength on the substrate surface.

[0043] Specifically, texture feature distribution maps serve as the fundamental data source for substrate surface analysis, containing rich information on surface morphology.

[0044] In one embodiment, local defect identification is based on the degree of deviation between the texture type label and the actual texture features. When the texture feature value of a certain area deviates from the standard value by more than a preset threshold, the area is marked as a defect area. The defect ratio is calculated through meshing, dividing the substrate surface into several mesh units, and statistically analyzing the proportion of defect points in each unit to the total sampling points, forming a local defect ratio matrix.

[0045] It should be noted that the calculation of surface energy involves the relationship between surface roughness and the surface free energy of the material. Surface free energy reflects the degree of unsaturation of molecules on the material surface, while roughness increases the actual surface area; the product of the two can characterize the total surface energy of the region. Specifically, the surface energy value of each region is obtained by multiplying the combined value of the arithmetic mean roughness Ra and the root mean square roughness Rq by the material's intrinsic surface free energy constant. This calculation method considers the influence of surface microstructure on energy distribution.

[0046] Specifically, the extraction of texture periodicity parameters is achieved through frequency domain analysis using Fourier transform. The Fourier transform converts texture information from the spatial domain to the frequency domain, with the dominant frequency component corresponding to the main period of the texture. The measurement of the distance between adjacent peaks is based on spectral analysis results, obtaining the spatial period length by identifying the reciprocal of the dominant frequency. The calculation of the contact area ratio is more complex, requiring consideration of the impact of peak-valley height differences on actual contact. When two surfaces are in contact, only the peaks truly contact, while the valleys form gaps; the actual contact area is much smaller than the theoretical planar area.

[0047] In one embodiment, the assessment of bond strength is based on the contact area ratio. A contact area ratio below a threshold indicates fewer actual contact points, resulting in a corresponding decrease in bond strength. The inherent adhesion coefficient of the material is a material property parameter determined through standard testing, reflecting the material's adhesive ability under ideal contact conditions. The initial bond strength value is obtained by multiplying the contact area ratio by the inherent adhesion coefficient; this linear relationship has been verified in engineering practice.

[0048] Preferably, the introduction of a surface energy gradient provides a correction mechanism for adhesive strength. The difference in surface energy between adjacent regions reflects the non-uniformity of energy distribution; there is a tendency for energy transfer from high-energy regions to low-energy regions, affecting the stability of interfacial adhesion. By calculating the energy difference between adjacent grid cells and correcting the initial adhesive strength value, the actual adhesive effect can be more accurately reflected. The determined spatial distribution parameters of adhesive strength include the adhesive strength value and its spatial coordinates for each grid cell, forming a complete adhesive performance distribution map.

[0049] S103. Based on the spatial distribution parameters of the bonding strength, determine whether there are areas of uneven bonding strength caused by an excessively high proportion of local defects.

[0050] Based on the strength values ​​of each grid cell in the spatial distribution parameters of bond strength, and combined with the peak and valley heights and peak positions in the aforementioned texture feature distribution map, the unit area load borne by each peak is calculated. The interface stress distribution value of each region is obtained by multiplying the peak density by the load, thus constructing an interface stress distribution map that includes stress magnitude and direction. Using the interface stress distribution map, the standard deviation of stress values ​​between adjacent grid cells is calculated. The stress uniformity coefficient is determined based on the ratio of the standard deviation to the average stress. Simultaneously, regions with a defect ratio exceeding a preset threshold are identified from the aforementioned local defect ratio matrix. Points within these regions where the stress value exceeds a preset multiple of the average value are identified as stress concentration points, resulting in a stress uniformity coefficient matrix and stress concentration region markers. For the stress uniformity coefficient matrix and stress concentration region markers, if the stress uniformity coefficient exceeds a preset threshold range, it is determined that there is a region with uneven bond strength. By comparing the overlap between the stress concentration region and the high defect ratio region, the specific location and range of the uneven bond strength region caused by the excessively high local defect ratio are determined.

[0051] Specifically, the interface stress analysis mechanism is constructed based on the peak and valley distribution characteristics at the micro level.

[0052] In one embodiment, when the film contacts the substrate, actual contact occurs only at the peak locations, forming discrete support points. The load borne by each peak depends on the peak distribution density and height difference around it. The calculation of the load per unit area needs to consider the effective load-bearing area of ​​the peaks; typically, the peaks are not ideal point contacts but have a certain contact area. The stress value of each peak is obtained by dividing the total load by the number of peaks actually in contact, and then by the average contact area of ​​a single peak.

[0053] It should be noted that peak density varies across different regions. Regions with higher density experience relatively smaller loads on individual peaks, while regions with lower density exhibit stress concentration. The interfacial stress distribution value is calculated by multiplying the peak density by the unit load; this method reflects the spatial distribution characteristics of stress. The constructed interfacial stress distribution map not only contains information on the magnitude of stress but also represents the direction of stress in vector form, providing a complete data foundation for subsequent uniformity analysis.

[0054] Specifically, determining the stress uniformity coefficient involves the application of statistical methods. The standard deviation reflects the spatial dispersion of stress values, while the mean stress represents the overall stress level. The ratio of the two can dimensionlessly characterize the uniformity of stress; a smaller ratio indicates more uniform stress distribution. Adjacent mesh cells are typically selected using an eight-neighbor or four-neighbor approach to ensure that local stress variations are captured.

[0055] In one embodiment, stress concentration is determined based on a statistical threshold method. When the stress value in a region exceeds a specific multiple of the overall average, that region is marked as a stress concentration region. This multiple is typically determined based on the material's load-bearing characteristics and safety factor, and is generally set between 1.5 and 2 times. The local defect proportion matrix provides spatial distribution information of defects; regions with a high defect proportion are often accompanied by abnormal peak distribution, leading to changes in the stress transmission path.

[0056] Preferably, the overlap analysis is achieved using a spatial overlay method. By overlaying the binarized images of stress concentration areas with those of areas with a high defect ratio, the proportion of the overlapping area to the total abnormal area is calculated. High overlap indicates that defects are the main cause of stress concentration, and establishing this causal relationship provides a clear direction for subsequent quality improvements. The location of areas with uneven bond strength is precisely determined using grid coordinates, and the range is determined through connected component analysis, forming a complete non-uniformity assessment result. This analysis method can accurately identify weak bond areas caused by surface defects, providing a quantitative basis for targeted surface treatment and process optimization.

[0057] S104. If areas with uneven bonding strength are detected, the optimal combination of peel speed influencing parameters and peel force fluctuation values ​​during the demolding process is determined by combining surface energy difference and texture density distribution data.

[0058] If an area with uneven adhesive strength is detected, based on the location coordinates and range information of that area, process parameter records matching the surface energy and texture density values ​​of the area are retrieved from a pre-established process parameter adjustment library. The corresponding demolding process parameter combination, including initial peel speed, speed change rate, and peel angle data, is extracted to obtain a preliminary process parameter set. Using this preliminary process parameter set, combined with the surface energy difference distribution of the uneven adhesive strength area, the theoretical peel force required at different locations is calculated. An adjustment coefficient for the peel speed is determined based on the ratio of texture density to reference density. The actual peel parameters for each location are obtained by multiplying the theoretical peel force value by the adjustment coefficient, forming a targeted demolding process correction scheme. For this demolding process correction scheme, the peel force fluctuation range is calculated using the standard deviation of the theoretical peel force value. Speed ​​influence parameters are determined based on the correspondence between high surface energy areas requiring low peel speeds. A genetic algorithm is used to optimize the combination of peel speed and peel force, determining the optimal combination of peel speed influence parameters and peel force fluctuation values ​​during the demolding process.

[0059] Specifically, the construction of the process parameter adjustment library is based on the accumulation of a large amount of demolding experimental data.

[0060] In one embodiment, the database employs a multidimensional index structure, using surface energy and texture density values ​​as primary search keys. Upon detecting areas of uneven adhesive strength, the database can quickly match process parameter records under similar conditions by inputting the characteristic parameters of that area. These records contain complete parameter combinations from successful demolding cases, such as initial peel speed typically ranging from 0.5 to 5 millimeters per second, the rate of change of speed reflecting acceleration or deceleration characteristics during the peeling process, and the peel angle affecting the direction of peel force decomposition.

[0061] It should be noted that the calculation of the theoretical peel force involves fundamental principles of materials mechanics and interface chemistry. The differential distribution of surface energy directly affects the interfacial bonding strength; higher energy regions exhibit stronger intermolecular forces, requiring greater peel force to achieve separation. During the calculation, according to Young's equation and interfacial tension theory, the peel force is positively correlated with surface energy. The influence of texture density is reflected in the change in contact area; higher density results in more actual contact points, thus increasing the required peel force.

[0062] Specifically, the adjustment coefficient is determined using a ratio method. The reference density is usually chosen as the average texture density of the material surface. When the texture density of a certain area is higher than the reference value, the adjustment coefficient is greater than 1, indicating that the peeling speed needs to be reduced to avoid film tearing; conversely, when the texture density is lower than the reference value, the peeling speed can be appropriately increased to improve efficiency. This dynamic adjustment mechanism ensures a smooth demolding process.

[0063] In one embodiment, determining the range of peel force fluctuations is crucial to the quality of film removal. The standard deviation calculation reflects the degree of spatial variation in peel force; excessive fluctuations can lead to uneven stress on the film, potentially causing localized tearing or residue. The relationship between surface energy and peel speed is based on the principle of energy conservation. Regions with high surface energy have tighter interfacial bonding and require slower peel speeds to ensure sufficient energy transfer and stress release, preventing sudden breakage.

[0064] Preferably, a genetic algorithm plays a crucial role in the optimization process. This algorithm encodes peeling speed and peeling force as chromosomes, and through simulating natural selection and genetic mutation processes, searches for the optimal parameter combination across multiple generations of evolution. The fitness function comprehensively considers multiple objectives such as peeling integrity, efficiency, and energy consumption, continuously improving the quality of the solution through crossover and mutation operations. The determined optimal combination not only includes specific parameter values ​​at each location but also provides control curves for parameter changes over time and location, achieving precise control of the peeling process. This targeted optimization scheme based on actual surface characteristics significantly improves the peeling success rate and product quality consistency.

[0065] S105. Based on the optimized combination of the peeling speed influence parameters and peeling force fluctuation values, monitor the layer peeling frequency and peeling position distribution during the decoction process in real time, and generate the distribution characteristics of fragment size range and fragmentation boundary features.

[0066] Based on an optimized combination of parameters affecting peeling speed and peeling force fluctuation, the peeling action is controlled according to the optimized parameters during the peeling process. Vibration signals at the peeling interface are collected in real time using a high-frequency vibration sensor. When the vibration signal frequency exceeds a preset threshold, it is recorded as a delamination peeling event. The number of delaminations per unit time is counted to obtain the delamination peeling frequency. Simultaneously, the coordinate position of each event is recorded, forming peeling position distribution data. Using the peeling position distribution data and vibration signal characteristics, the spectral characteristics of the vibration signal are extracted using Fast Fourier Transform. Based on the principle that a sudden increase in the amplitude of high-frequency components indicates fragment breakage, the fragment generation time is identified. The initial size of the fragment is calculated by combining the time interval and position distance between adjacent peeling events, obtaining preliminary fragment size data including timestamps and spatial coordinates. For the preliminary fragment size data, envelope analysis of the vibration signal is used to identify the fragment boundary positions corresponding to signal abrupt changes. The smoothness of the boundary is determined based on the rate of change of signal intensity. If the rate of change exceeds a preset threshold, it is marked as a serrated boundary; otherwise, it is marked as a smooth boundary. The distribution of fragment quantity in different size ranges is statistically analyzed to obtain the distribution characteristics of fragment size range and fragmentation boundary features.

[0067] Specifically, high-frequency vibration sensors play a key role in membrane removal monitoring.

[0068] In one embodiment, the sensor typically employs a piezoelectric or capacitive structure, capable of capturing minute vibrations at the kilohertz level. When the film separates from the substrate interface, the accumulated stress is released instantaneously, generating a characteristic high-frequency vibration signal. This vibration signal typically ranges in frequency from 1 to 10 kilohertz, significantly higher than the low-frequency vibrations of a normal peeling process. By setting an appropriate frequency threshold, the timing of the delamination event can be accurately identified.

[0069] It should be noted that the statistics of delamination frequency have significant implications for process guidance. An excessively high frequency indicates an unstable delamination process, potentially due to improper parameter settings. The delamination location distribution data is recorded using a coordinate mapping method, marking the location of each delamination event on a two-dimensional coordinate system on the substrate surface. This spatial distribution information can reveal areas of concentrated delamination, which are typically associated with surface defects or areas of weak adhesion.

[0070] Specifically, the application of Fast Fourier Transform (FFT) in vibration signal analysis is based on the principle of time-frequency conversion. The original time-domain vibration signal contains a superposition of various frequency components, which is decomposed into sinusoidal wave components of different frequencies through Fourier transform. When fragments are generated, the fracture process excites vibrations in specific frequency bands, manifested as a sudden increase in amplitude at the corresponding frequency in the frequency spectrum. The physical mechanism of this sudden increase phenomenon lies in the fact that the elastic energy released by the material fracture propagates in the form of vibration waves, and its frequency characteristics are inversely proportional to the fragment size; smaller fragments produce higher frequency vibrations.

[0071] In one embodiment, fragment size calculation incorporates information from both temporal and spatial dimensions. The time interval between adjacent peeling events reflects the advance velocity of the peeling front, while the location distance characterizes the spatial span of the peeling region. By dividing the spatial distance by the time interval, the local peeling velocity is obtained. Combined with the material properties of that region, the initial fragment size can be estimated. This method takes into account the dynamic characteristics of the peeling process and reflects reality more accurately than static measurements.

[0072] Preferably, envelope analysis of the vibration signal provides a basis for determining boundary characteristics. The envelope reflects the trend of signal amplitude change, and the rate of change of signal intensity is obtained by calculating the derivative of the envelope. A sharp rate of change corresponds to an irregular fracture process, forming a serrated boundary; while a gradual rate of change indicates a more uniform fracture process, producing a smooth boundary. This boundary characteristic is closely related to the brittleness of the material and the peeling speed. The obtained fragment distribution characteristics include statistical histograms of size ranges and classification results of boundary types, providing a quantitative basis for evaluating the demolding quality and optimizing process parameters. By analyzing the proportion of fragments in different size ranges, it is possible to determine whether the demolding process generates too many small fragments, avoiding adverse effects on subsequent processes.

[0073] S106. Analyze the impact of residual adhesion ratio and fragmentation accumulation effect on demolding quality, and determine whether the degree of fragmentation exceeds the acceptable range.

[0074] By combining the distribution characteristics of fragment size range and fragmentation boundary features, fragmentation boundary contour data is identified. Geometric parameters of the fragments, including length, width, and area, are extracted based on the contour morphology. If the fragment size exceeds a preset threshold, the area is marked as a target fragment region. Based on the spatial distribution of the target fragment region, the density distribution characteristics of residual adhesive points at the fragment boundary are calculated, the bonding state of the bonding interface is detected, and the bonding ratio data is obtained to determine the correlation between the bonding ratio and fragmentation boundary features. The bonding ratio data is processed through time series analysis, and the rate of change of the cumulative fragmentation effect is calculated using a sliding window method. The degree of delamination quality degradation is determined based on the rate of change. If the rate of change of the cumulative effect exceeds a preset threshold, a quality risk warning is triggered. Based on the risk level output by the quality risk warning, the degree of fragmentation is quantitatively assessed. A preset risk assessment threshold is used to determine whether the current fragmentation state is within an acceptable range, thus obtaining the delamination quality assessment result.

[0075] Specifically, the application of image processing technology in stripping surface scanning is based on computer vision principles, using a CCD camera or laser scanner to acquire two-dimensional image data of the stripping surface.

[0076] In one possible implementation, an industrial camera with a resolution of 2048×2048 pixels is used to perform a full-coverage scan of the demolded surface with a pixel accuracy of 0.1 mm, obtaining a digital image containing complete surface information. This high-precision scan can capture minute surface defects and boundary changes, providing a reliable data foundation for subsequent fragmentation analysis. The core of the edge detection algorithm lies in identifying regions of abrupt changes in pixel brightness in the image; these regions typically correspond to the boundary contours of fragments.

[0077] Specifically, the Canny edge detection algorithm identifies the boundary line between the fragment and the background by calculating the gradient magnitude and direction of the image.

[0078] For example, when an irregular fragment with a length of 15 mm and a width of 8 mm exists on the demolded surface, the algorithm can accurately extract the closed contour of the fragment and calculate its area as 120 square millimeters. The preset range threshold is set based on historical data statistics; when the fragment area exceeds 100 square millimeters or the aspect ratio exceeds 3:1, it is marked as a target fragment area. Statistical methods play a key role in calculating the density distribution of residual adhesive points, and the degree of aggregation of adhesive points is evaluated through spatial point pattern analysis technology.

[0079] In one embodiment, the system divides the target fragment region into 5×5 grid cells, counts the number of bonding points within each cell, and calculates an average density of 3.2 bonding points per square millimeter. This gridded statistical method can effectively identify the spatial distribution pattern of bonding points and discover locally clustered or sparse areas. Contact angle measurement technology assesses the bonding state by observing the wetting behavior of droplets at the bonding interface; the contact angle value directly reflects the bonding strength of the interface.

[0080] It should be noted that a contact angle less than 90 degrees indicates good wettability and strong adhesion, while a contact angle greater than 120 degrees indicates a tendency for separation at the bonded interface. By measuring the changes in contact angle at different locations, the system can quantify the bonding ratio. For example, in a 10×10 mm test area, if 85% of the measurement points have a contact angle less than 90 degrees, it indicates that the bonding ratio in that area is 85%. When processing the bonding ratio data through time series analysis, the moving average method and trend analysis method are used to identify the patterns of data change.

[0081] Specifically, the bonding ratio data for each hour within a continuous 24-hour period is collected to construct a time series curve, and the rate of change between adjacent time points is calculated using the sliding window method.

[0082] For example, when the bonding ratio drops from 85% in the 10th hour to 78% in the 11th hour, the change rate is -7% / hour. This negative change rate indicates that the cumulative effect of fragmentation is intensifying. The degree of deterioration in demolding quality is determined based on continuous monitoring of the cumulative effect change rate. When the change rate exceeds a preset threshold of -5% / hour for three consecutive hours, the system triggers a quality risk warning. This dynamic monitoring mechanism can promptly detect quality deterioration trends and avoid serious demolding failures. The threshold comparison method achieves a quantitative assessment of the degree of fragmentation by setting multi-level risk assessment standards.

[0083] In one possible implementation, three risk levels are set: low risk corresponds to a fragmentation degree of less than 20%, medium risk corresponds to 20%-40%, and high risk corresponds to greater than 40%. When the aforementioned analysis results show that the fragmentation degree reaches 35%, it is judged as a medium risk state, which is still within the acceptable range, and a "qualified" delamination quality assessment result is output.

[0084] S107. If the fragmentation level exceeds the acceptable range, extract the optimized scheme for the peeling position distribution and peeling force fluctuation value from the preset process control strategy library, obtain the adjusted demolding control accuracy parameters, and determine the demolding process execution scheme. If the fragmentation level does not exceed the acceptable range, continue with the subsequent quality assessment process.

[0085] If the fragmentation level exceeds the acceptable range, candidate solutions matching the peeling position distribution are extracted from the preset process control strategy library. The peeling force fluctuation value is characterized using pattern matching, and the most suitable process adjustment direction is determined based on the fluctuation amplitude and frequency characteristics, obtaining the corresponding control precision parameters. Based on these control precision parameters, the peeling speed and applied pressure are adjusted, and the movement trajectory of the peeling path is corrected using a position control device. The adjusted process parameter combination is obtained, and the demolding process execution plan is determined. The demolding operation is performed using the demolding process execution plan. Force sensor data is used to monitor the force feedback during the demolding process, obtaining new fragmentation level values. If the fragmentation level still exceeds the acceptable range, the process control strategy library is called again for parameter rematching. Based on the detection results of the fragmentation level values, if the fragmentation level does not exceed the acceptable range, the correspondence between process parameters and quality indicators is saved, updated strategy library data is obtained, and subsequent quality assessment and data update processes continue.

[0086] Specifically, the process control strategy library, as a pre-set database, contains a large number of process parameter combinations from historical successful cases. These parameter combinations are categorized and stored according to different stripping location distribution patterns.

[0087] In one possible implementation, the strategy library establishes an index system based on the spatial distribution characteristics of the peeling locations. When a peeling pattern concentrated in the edge region is detected, the system automatically matches the corresponding low-speed peeling scheme, while for peeling patterns in the central region, a high-precision positioning scheme is matched. This classification and storage mechanism ensures the targetedness and accuracy of scheme selection. The pattern matching method is based on the feature vector comparison principle, converting the currently detected peeling force fluctuation value into a feature vector and calculating its similarity with the standard patterns in the strategy library.

[0088] Specifically, when the peeling force fluctuations exhibit periodic oscillation characteristics, the system identifies a fluctuation pattern with an amplitude of 2.5 Newtons and a frequency of 5 Hz, and uses Euclidean distance calculations to find the most similar historical successful cases. This precise matching process avoids blind parameter adjustments and improves the success rate of process optimization. The acquisition of control precision parameters relies on the analysis and extraction process of the matching results; these parameters include precise values ​​for peeling speed, applied pressure, and motion trajectory.

[0089] In one embodiment, for a specific fluctuation pattern detected, the system extracts a combination of control parameters: a peeling speed of 0.8 mm / s and an applied pressure of 3.2 N / s. These precise parameters provide a reliable data foundation for subsequent equipment adjustments. The parameter calibration method achieves real-time adjustment of the equipment's operating state through a closed-loop control mechanism, ensuring consistency between the actual executed parameters and the target parameters.

[0090] It should be noted that the correction process employs a PID control algorithm, which eliminates parameter deviations and maintains stable operation through the coordinated action of proportional, integral, and derivative components. When the target peeling speed is 0.8 mm / s while the actual speed is 0.75 mm / s, the controller automatically increases the drive torque to achieve speed correction. The position control device achieves precise control of the motion trajectory through a stepper motor and a precision guide rail system, ensuring that the peeling process proceeds along the predetermined path.

[0091] In one possible implementation, the device employs micro-stepping control technology, subdividing the step angle to 0.01 degrees to achieve sub-millimeter-level positioning accuracy. This high-precision control avoids path deviations during the peeling process and reduces additional fragmentation caused by improper trajectories. A force sensor continuously monitors changes in applied force during the demolding operation, acquiring detailed force curves through high-frequency sampling.

[0092] Specifically, the sensor records force changes at a sampling frequency of 1000 Hz. When a sudden increase or abnormal fluctuation in force is detected, it immediately feeds back to the control device for corresponding adjustments. This real-time monitoring mechanism ensures the stability and controllability of the demolding process. The parameter rematching process reassesses the applicability of the current process scheme based on the new fragmentation level values.

[0093] In one embodiment, if the fragmentation level after the initial adjustment still reaches 45% and exceeds the acceptable threshold of 40%, the system automatically initiates a secondary matching process to select a more conservative, low-speed stripping scheme from the strategy library. This iterative optimization mechanism ensures the reliability of the final result. The data recording device stores the correspondence between process parameters and quality results in a structured storage method, providing data support for the continuous updating of the strategy library.

[0094] It should be noted that the recorded content includes input parameters, intermediate process data, and final quality indicators, forming a complete data chain. After statistical analysis, this historical data can identify new success patterns and supplement the strategy base, enabling the knowledge base to self-improve and expand.

[0095] S108. Based on the demolding process execution plan, and combined with the fragmentation accumulation effect and residual adhesion ratio data, perform quality prediction calibration on the influence of delamination frequency and delamination speed to obtain demolding quality evaluation indicators that meet the specific production scenario.

[0096] According to the demolding process execution plan, a data fusion method is used to correlate and match fragmented cumulative effect data with residual adhesion ratio data, calculate the correlation coefficient between the cumulative effect and the adhesion ratio, obtain comprehensive quality characteristic parameters, and determine the quality level of the current production batch. Based on the quality level, the new comprehensive quality characteristic parameters are written into the demolding quality grading standard database. The database's historical information is maintained through a data recording device, and an updated database containing the new quality standards is obtained to determine the influence weight of the delamination frequency on the quality level. Based on the influence weight, a linear regression method is used to establish a functional relationship between delamination speed and quality prediction results. Residual analysis is used to calibrate the prediction accuracy, obtain the calibrated quality prediction parameters, and determine the prediction accuracy range applicable to the current production scenario. Based on the prediction accuracy range, a weighted average method is used to adjust the parameter ratios for quality assessment. A unified-dimensional assessment index system is obtained through standardized calculations to determine demolding quality assessment indicators that meet the requirements of the specific production scenario, resulting in the final quality assessment standard.

[0097] Specifically, the data fusion method is based on the principle of multi-source information comprehensive processing. It achieves the organic combination of quality information from different dimensions by spatiotemporally aligning fragmented cumulative effect data with residual adhesion ratio data and extracting features.

[0098] In one possible implementation, when fragmented cumulative effect data shows a cumulative index of 0.75 for a certain region, and the residual adhesion ratio at the corresponding location is 65%, the fusion algorithm calculates a comprehensive quality index of 0.68 for that region using a weighted average. This fusion process eliminates the limitations of a single indicator and provides a more comprehensive basis for quality evaluation. Correlation analysis using statistical methods, through the calculation of the Pearson correlation coefficient, reveals the intrinsic relationship between the cumulative effect and the adhesion ratio.

[0099] Specifically, after collecting data from 100 sample points, the correlation coefficient was calculated to be -0.82, indicating a strong negative correlation between the cumulative effect and the adhesion ratio. This finding confirms the technical principle that the higher the degree of fragmentation, the weaker the residual adhesion, providing a theoretical basis for the establishment of a quality prediction model. The process of obtaining comprehensive quality characteristic parameters involves dimensionality reduction and feature compression of multidimensional data, and the extraction of the most representative quality features through principal component analysis.

[0100] In one embodiment, the original 8-dimensional quality data, after dimensionality reduction, retains only the first three principal components, which explain 89% of the data variance, forming a comprehensive feature parameter that includes a cumulative effect weight of 0.45, an adhesion ratio weight of 0.35, and a surface morphology weight of 0.20. This feature compression significantly simplifies the subsequent quality grading process. Database update operations ensure the effective integration of the new quality standards through structured storage and index reconstruction.

[0101] It should be noted that the update process adopts a transaction processing mechanism. When new comprehensive quality characteristic parameters are written to the database, the system synchronously updates the relevant index tables and relationships.

[0102] For example, when a new record with a quality grade of "Good" and feature parameters [0.68, 0.65, 0.72] is added, the database automatically adjusts the grading threshold boundaries to ensure the continuity and consistency of the classification criteria. The determination of the influencing weights is based on statistical analysis of historical data and machine learning training results, quantifying the contribution of different operating parameters to the final quality.

[0103] In one possible implementation, analysis of historical data from 1000 delamination operations revealed that the delamination frequency had a weighting of 0.38 on the quality grade, the delamination speed had a weighting of 0.42, and the combined weighting of other factors was 0.20. This quantitative weighting provides a scientific basis for the precise adjustment of process parameters. The linear regression method, by establishing a mathematical relationship between delamination speed and quality prediction results, achieves quantitative control of process parameters.

[0104] Specifically, when the regression equation Y = 0.85X + 0.12 is established, where Y represents the predicted quality index and X represents the standardized peeling speed, the coefficient of determination R² reaches 0.91, indicating that the model has good predictive accuracy. This quantitative relationship allows operators to accurately set the peeling speed parameters according to the target quality requirements. Residual analysis identifies and corrects systematic errors by examining the distribution of deviations between predicted and actual values.

[0105] In one embodiment, analysis of the residuals of 200 samples revealed that 85% of the prediction errors were controlled within ±5%, while the remaining 15% of larger errors were mainly concentrated under special operating conditions. Based on this analysis, the system established a specific correction factor for these special operating conditions, improving the overall prediction accuracy to 92%. The weighted average method plays a crucial role in adjusting the allocation of quality assessment parameters, achieving adaptive optimization of the assessment criteria through dynamic weight allocation.

[0106] It should be noted that the weight adjustment is based on feedback from real-time production data. When the importance of a certain parameter changes, the system automatically recalculates the weight allocation.

[0107] For example, in high-precision production scenarios, the weight of surface quality is adjusted from the standard 0.30 to 0.45, while in high-efficiency production scenarios, the weight of production speed is increased accordingly to ensure that the evaluation criteria match the actual production needs.

[0108] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for evaluating the efficiency of vacuum debonding based on intelligent algorithm, characterized in that, The method includes: The surface of a plastic substrate is probed to obtain surface roughness parameters and texture directionality characteristic data. The specific patterns of microscopic peak-valley distribution and texture density distribution are determined, resulting in a texture feature distribution map of the substrate surface. Based on this texture feature distribution map, local defect ratio and surface energy difference data are extracted to analyze the influence of periodic texture changes on film adhesion strength, determining the spatial distribution parameters of adhesion strength on the substrate surface. For these spatial distribution parameters, it is determined whether there are regions of uneven adhesion strength. If uneven adhesion strength regions are detected, the surface energy difference and texture density distribution data are combined to determine the bonding strength during the demolding process. The peeling speed influence parameter and peeling force fluctuation value are optimized and combined. Based on the optimized combination of peeling speed influence parameter and peeling force fluctuation value, the delamination peeling frequency and peeling position distribution during the demolding process are monitored in real time, generating the distribution characteristics of fragment size range and fragmentation boundary features. The influence of residual adhesion ratio and fragmentation accumulation effect on demolding quality is analyzed to determine whether the degree of fragmentation exceeds the acceptable range. If the degree of fragmentation exceeds the acceptable range, the demolding control precision parameter is adjusted to determine the demolding process execution plan. Based on the demolding process execution plan, quality prediction calibration is performed to obtain demolding quality evaluation index.

2. The method for evaluating the efficiency of vacuum debonding based on intelligent algorithm according to claim 1, characterized in that, The process involves probing the surface of the plastic substrate, acquiring surface roughness parameters and texture directionality feature data, determining the specific patterns of microscopic peak-valley distribution and texture density distribution, and obtaining a texture feature distribution map of the substrate surface, including: The process involves collecting surface height values ​​and location coordinates of the substrate, calculating the height difference between adjacent points, and generating topographic data containing height and gradient distributions. Based on this topographic data, the arithmetic mean roughness and root mean square roughness are calculated, spatial frequency components are extracted, the dominant texture direction is determined, and a roughness parameter matrix and texture direction angle map are generated. Using the roughness parameter matrix and texture direction angle map, the peak-valley arrangement direction is identified, the peak-valley spacing and depth distribution are calculated, the number of texture direction lines is counted, and peak-valley distribution features and texture density values ​​are generated. Finally, based on the peak-valley distribution features and texture density values, a texture classification database is matched to determine the texture category, and a texture feature distribution map annotating the texture type and density is generated.

3. The method for evaluating the efficiency of vacuum debonding based on intelligent algorithm according to claim 1, characterized in that, The step involves extracting local defect ratio and surface energy difference data based on the texture feature distribution map, analyzing the influence of texture periodic changes on film adhesion strength, and determining the spatial distribution parameters of adhesion strength on the substrate surface, including: Based on the texture feature distribution map, regions deviating from the texture type are identified as local defects. The proportion of defect points is statistically analyzed, surface energy values ​​are calculated, and a local defect proportion matrix and surface energy distribution data are generated. Based on the local defect proportion matrix and the surface energy distribution data, texture period parameters are extracted, the contact area ratio is calculated, and the initial bonding strength value is determined. Based on the initial bonding strength value and the surface energy difference, the bonding strength is corrected, and spatial distribution parameters of bonding strength are generated.

4. The vacuum stripping efficiency evaluation method based on intelligent algorithm according to claim 1, characterized in that, The determination of whether there are areas of uneven bond strength based on the spatial distribution parameters of the bond strength includes: Based on the spatial distribution parameters of the bonding strength and the texture feature distribution map, the load per unit area at the peak point is calculated, and an interface stress distribution map is generated. Based on the interface stress distribution map, the stress standard deviation and uniformity coefficient are calculated, stress concentration points are identified, and a stress uniformity coefficient matrix and stress concentration area markers are generated. Based on the stress uniformity coefficient matrix and the stress concentration area markers, the location and range of the bonding strength unevenness area are determined.

5. The method for evaluating the efficiency of vacuum debonding based on intelligent algorithm according to claim 1, characterized in that, If areas of uneven adhesive strength are detected, then by combining surface energy differences and texture density distribution data, an optimal combination of parameters affecting peel speed and peel force fluctuation values ​​during the demolding process is determined, including: Based on the uneven bonding strength region, the process parameter adjustment library is queried to extract process parameters that match the surface energy difference and texture density distribution, generating a preliminary process parameter set; based on the preliminary process parameter set, the theoretical value of peel force and speed adjustment coefficient are calculated to generate a demolding process correction scheme; based on the demolding process correction scheme, the peel speed and peel force are optimized, and the combination of peel speed influencing parameters and peel force fluctuation values ​​is determined.

6. The method for evaluating the efficiency of vacuum debonding based on intelligent algorithm according to claim 1, characterized in that, The process of optimizing the combination of peeling speed influence parameters and peeling force fluctuation values, and real-time monitoring of the delamination frequency and peeling location distribution during the decoction process to generate the distribution characteristics of fragment size range and fragmentation boundary features includes: Based on the combination of the peeling speed influence parameters and the peeling force fluctuation value, vibration signals are collected, and the distribution of layer peeling frequency and peeling location is statistically analyzed. Based on the vibration signals, spectral features are extracted, the fragment generation time is identified, and the initial fragment size is calculated. Based on the initial fragment size, signal abrupt change points are analyzed, boundary flatness is determined, and fragment size range and fragmentation boundary features are generated.

7. The method for evaluating the efficiency of vacuum debonding based on intelligent algorithm according to claim 1, characterized in that, The analysis of the impact of residual adhesion ratio and cumulative fragmentation on demolding quality, and the determination of whether the degree of fragmentation exceeds the acceptable range, includes: Identify fragmented boundary contours and extract fragment geometric parameters; calculate the density of residual bonding points based on the fragment geometric parameters and detect the bonding ratio; calculate the rate of change of cumulative fragmentation effect based on the bonding ratio and determine whether the degree of fragmentation exceeds the acceptable range.

8. The method for evaluating the efficiency of vacuum debonding based on intelligent algorithm according to claim 1, characterized in that, If the degree of fragmentation exceeds an acceptable range, the demolding control precision parameters are adjusted, and the demolding process execution plan is determined, including: Based on the degree of fragmentation, a process control strategy matching the distribution of peeling positions is extracted, peeling force fluctuation characteristics are identified, and peeling control accuracy parameters are generated. Based on the peeling control accuracy parameters, the peeling speed and path are adjusted to generate a peeling process execution plan.

9. The method for evaluating vacuum demolding efficiency based on intelligent algorithms according to claim 1, characterized in that, The step of performing quality prediction calibration according to the demolding process execution plan to obtain demolding quality evaluation indicators includes: Based on the demolding process execution plan, the fragmentation accumulation effect and the residual adhesion ratio are matched to calculate the quality characteristic parameters and determine the quality level; based on the quality level, the demolding quality grading standard database is updated, the relationship between peeling speed and quality prediction is calibrated, and demolding quality evaluation indicators are generated.

Images