Method for detecting uniformity of pvc film coating based on spectral analysis

Abstract

The application discloses a PVC film coating uniformity detection method based on spectral analysis, and relates to the technical field of optical detection, and comprises the following steps: S001, a controllable incident light illumination system with multi-angle rotation function is used to continuously angle-scan the surface of a polyvinyl chloride (PVC) film, reflection spectrum intensity data in a corresponding waveband range is collected according to different incident angles, and a spectrum data matrix containing interference characteristics is generated; S002, continuous wavelet transform processing is performed on the spectrum data matrix, the spectrum response signal is subjected to multi-scale local decomposition, periodic interference characteristic signals are extracted and separated, and a purified spectrum matrix after interference reduction is obtained. Through multi-angle spectrum acquisition and interference characteristic stripping, the application accurately identifies abnormal areas of the coating thickness, adjusts and controls spraying parameters in combination with an intensity index, and constructs a detection-compensation-retest closed loop mechanism, so that interference misjudgment is effectively avoided, and the accuracy and control efficiency of the PVC film coating uniformity detection are improved.

Description

Technical Field

[0001] This invention relates to the field of optical inspection technology, and more specifically to a method for detecting the uniformity of PVC film coatings based on spectral analysis. Background Technology

[0002] Spectral analysis-based PVC film coating uniformity testing is a non-contact method that utilizes spectral analysis to detect the thickness, color consistency, or compositional distribution of coatings on the surface of polyvinyl chloride (PVC) films. Specifically, when light shines on the PVC film surface, areas of different thicknesses or uneven coatings will exhibit differences in the reflected or transmitted spectral characteristics. This method collects and analyzes this spectral data to identify the spectral response intensity or wavelength shift of the coating at different locations, thereby determining the coating's uniformity. Compared to traditional physical contact measurement methods, spectral analysis offers advantages such as fast detection speed, high accuracy, and non-destructive testing, making it widely applicable for real-time monitoring and defect identification of coating quality in industrial online inspections. This method is particularly suitable for quality control in the production process of PVC functional film products that are sensitive to color differences and require precise coating thickness.

[0003] The existing technology has the following shortcomings:

[0004] In the process of detecting the spectral uniformity of PVC film coatings, the coating surface often exhibits micron-level periodic structures caused by processes such as calendering, stretching, or heat treatment. These structures include fine embossing, directional stretching textures, or microwave-like interference patterns. Furthermore, if the film material is a multi-layered composite structure, multi-interface reflection interference may also exist. Under specific incident angles and wavelengths of light, these microstructures or interface features are prone to triggering multiple interference effects, leading to periodic enhancement or attenuation of the spectral signal when the detection system acquires spectral data. Because these interference signals exhibit high repeatability and surface regularity in the spectral dimension, they are easily misjudged by the system algorithm as having uniform coating spectral characteristics, thus masking actual coating non-uniformity issues such as minute thickness fluctuations, formulation deviations, or uneven interface distribution. The existence of such pseudo-uniformity interference directly interferes with the system's accurate judgment of coating uniformity, causing serious missed detections. Especially in scenarios with extremely high requirements for coating uniformity, such as functional films and dielectric films, this can lead to performance failure, product scrapping, or even downstream application accidents, becoming one of the key hidden dangers affecting the reliability of detection and the stability of product quality.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method for detecting the uniformity of PVC film coatings based on spectral analysis. By acquiring spectra from multiple angles and stripping interference features, it can accurately identify areas with abnormal coating thickness. Combined with intensity index to control spraying parameters, and constructing a closed-loop mechanism of detection-compensation-re-inspection, it can effectively avoid interference misjudgment and improve the accuracy and control efficiency of PVC film coating uniformity detection, thereby solving the problems in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting the uniformity of PVC film coatings based on spectral analysis, comprising the following steps:

[0008] S001 uses a controllable incident light system with multi-angle rotation function to continuously scan the surface of polyvinyl chloride film at different angles, and collects the reflection spectral intensity data in the corresponding band range according to different incident angles to generate a spectral data matrix containing interference characteristics.

[0009] S002, perform continuous wavelet transform processing on the spectral data matrix, decompose the spectral response signal into a multi-scale local decomposition, extract and separate the periodic interference feature signals, and obtain the purified spectral matrix after interference reduction.

[0010] S003, based on the purification spectral matrix, a support vector machine classification model is constructed to classify the spectral feature vectors of each detection location, and to identify and locate the spatial regions where there are abnormal fluctuations in coating thickness.

[0011] S004, extract the spatial boundary information and spectral response gradient parameters of the thickness anomaly area, establish a coating thickness deviation intensity calculation model, and output the coating deviation intensity index of the corresponding detection area;

[0012] S005, Based on the coating deviation strength index, construct dynamic coating control rules to control the moving speed of the spray head and the flow rate of the coating liquid in the coating equipment, and perform local compensation operations for areas with abnormal thickness.

[0013] S006 After completing the local compensation operation, S001 to S005 are re-executed on the compensation area to update the spectral data matrix and repeat the interference identification, anomaly extraction and coating control process to construct an integrated closed-loop mechanism for the detection and control of polyvinyl chloride film coating uniformity.

[0014] Preferably, step S001 includes:

[0015] By configuring an incident light irradiation device with controllable rotation angle function, the surface of polyvinyl chloride film is continuously scanned from 0 degrees to 45 degrees, and reflectance spectrum data is collected every 5 degrees.

[0016] At each incident angle, the light is illuminated by a white LED array light source with a color temperature of 5500K to 6000K and a wavelength range of 400 to 1000 nanometers, and a parallel beam is output by an optical path shaping device with a collimating lens group.

[0017] The reflected signal is acquired simultaneously using a fiber-coupled linear array CCD spectrometer with a spectral response range of 380 to 1100 nanometers and a resolution better than 0.8 nanometers. The reflection receiving direction is symmetrically configured with the incident direction.

[0018] All collected data were normalized for reflectance, removed for background trends, and classified by number. Finally, a two-dimensional spectral data matrix was constructed, with the row dimension being the incident angle, the column dimension being the sampling wavelength, and the matrix unit being the reflection intensity value.

[0019] Preferably, step S002 includes:

[0020] Extract the original spectral data matrix containing multiple incident angles and multiple wavelength sampling points, and resample and align the spectral curves on the wavelength axis through interpolation reconstruction.

[0021] For each spectral curve, a continuous wavelet transform with MexicanHat as the mother wavelet function is performed to generate a sequence of wavelet coefficients at multiple scales.

[0022] Within a specified range of scales from 8 to 32, regions of continuous high-amplitude periodic fluctuations are identified, and their wavelet coefficients are attenuated using a soft-threshold compression method.

[0023] The inverse transform operation is performed based on the adjusted wavelet coefficients to reconstruct the purified spectral curves at each incident angle, and then combined to generate a purified spectral data matrix.

[0024] Preferably, step S003 includes:

[0025] Extract the spectral response curve of each detection point in the purification spectral matrix in the wavelength range of 400 nm to 1000 nm, and normalize each curve.

[0026] Feature values, including band slope, reflectance jump amplitude, and curvature continuity, are extracted from each normalized spectral curve to construct a standardized feature sequence;

[0027] Based on the pre-established physical response classification rules, the spectral characteristics of all detection points are compared and judged, and the classification result corresponding to each detection point is output.

[0028] Based on all classification results, a two-dimensional spatial positioning map of the membrane surface is constructed, the boundary positions of abnormal areas are marked and isolated interference points are eliminated, and the spatial positioning of thickness abnormal areas is completed.

[0029] Preferably, step S004 includes:

[0030] Spatial aggregation processing is performed on the detection points that are identified as anomalies to extract the minimum rectangular boundary of each connected anomaly region, and the boundary coordinates are smoothed by a five-point weighted average method.

[0031] Extract the purification spectral curves of each detection point in the 550 nm to 800 nm band within each abnormal region, and calculate the maximum slope, reflectance difference, and reflectance offset between specified wavelengths respectively.

[0032] Select the 10% and 10% of the detection points with the largest and smallest changes in each region, calculate the difference between the average values ​​of the three indicators, and multiply it by the square root of the region area to obtain the coating deviation strength index.

[0033] Output the corresponding numbers, boundary coordinates, and coating deviation strength indices for all abnormal areas, and draw two-dimensional spatial distribution maps of different strength levels.

[0034] Preferably, step S005 includes:

[0035] Based on the coating deviation strength index of each test area, it is divided into five levels, and the adjustment ratio of the spray head moving speed and the coating flow rate is set accordingly.

[0036] The spatial boundary position of the abnormal area is converted into the time coordinate of the membrane material transportation process to generate the control time axis for spraying compensation.

[0037] During the compensation period, the spray head is controlled to adjust its moving speed according to a set ratio, and the flow rate of the coating liquid is increased simultaneously.

[0038] After the spraying is completed, the spectral reflectance data of the coated area is re-detected, and the control parameters for the next round are adjusted based on the compensation effect.

[0039] Preferably, during the compensation period, the spray head is controlled to adjust its moving speed according to a set ratio, and the coating flow rate is increased simultaneously. The specific steps are as follows:

[0040] During the compensation period, the spray head is controlled by a linear motor to operate within the target speed range, and the speed adjustment accuracy is controlled within ±0.5 mm per second.

[0041] Meanwhile, the output flow rate of the coating liquid is adjusted by a combination of an electronically controlled pump and a pressure regulating valve, with a flow response time of less than 100 milliseconds, to ensure that the compensation area completes the synchronous adjustment of the spray head speed and the coating liquid flow rate at the moment of entry.

[0042] Preferably, step S006 includes:

[0043] After local compensation is completed, multi-angle illumination scanning is performed again to collect reflectance spectral intensity data of the compensation area in the 400 nm to 1000 nm band.

[0044] The newly acquired spectral data matrix is ​​subjected to interference recognition and purification processing to obtain an updated purified spectral matrix;

[0045] Based on the purification spectral matrix, abnormal areas are re-identified to determine whether there are still locations with insufficient compensation.

[0046] For areas with insufficient compensation, the deviation intensity index is recalculated and the compensation parameters are adjusted. The control steps are repeated to achieve closed-loop regulation.

[0047] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0048] This invention effectively incorporates joint information from the spectral and angular dimensions through multi-angle reflectance spectral acquisition, overcoming the limitations of traditional single-angle detection which is susceptible to interference-induced errors. It combines continuous wavelet transform to extract and remove periodic interference signals, significantly purifying spectral data and providing an accurate baseline for subsequent identification. A classification model based on the purified data is introduced to accurately locate thickness anomaly regions, and an intensity index is constructed through spatial boundary identification and gradient extraction to quantify the degree of deviation. This is then linked to the coating execution process, enabling dynamic adjustment of the spray head speed and coating flow rate. Finally, a complete closed-loop mechanism is constructed through re-detection and feedback of the compensation area. The overall process forms a closed-loop path from spectral data acquisition, interference suppression, anomaly identification, compensation control to effect re-verification, achieving multi-dimensional perception and dynamic correction of coating thickness fluctuations. This effectively avoids the risk of missed detection caused by interference pseudo-uniformity, significantly improving the detection accuracy and production control capabilities of PVC film coating consistency. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0050] Figure 1 This is a flowchart of the method for detecting the uniformity of PVC film coating based on spectral analysis according to the present invention. Detailed Implementation

[0051] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0052] This invention provides, for example Figure 1 The method for detecting the uniformity of PVC film coatings based on spectral analysis, as shown, includes the following steps:

[0053] S001 uses a controllable incident light system with multi-angle rotation function to continuously scan the surface of polyvinyl chloride film at different angles, and collects the reflection spectral intensity data in the corresponding band range according to different incident angles to generate a spectral data matrix containing interference characteristics.

[0054] To achieve high-precision testing of the uniformity of polyvinyl chloride film coatings, the following specific steps are implemented:

[0055] A multi-angle reflectance spectrum acquisition of the membrane surface is performed using an incident light irradiation device with controllable rotation angle. This device employs a white light source covering the visible to near-infrared band. The white light source is a hyperspectral continuous LED array with a color temperature stable between 5500K and 6000K and a wavelength range of 400 to 1000 nanometers. To ensure the beam is focused and parallel, a collimation device with a spherical focusing lens group is installed in front of the white light source to shape the emitted scattered light before projecting it onto the membrane surface. The membrane is fixed on an electrically controlled rotating platform that can rotate gradually from the vertical direction (0 degrees) to a 45-degree angle with the membrane surface in 0.5-degree increments, maintaining a stationary position for more than one second at each set angle for stable illumination and signal acquisition. This angle scanning process is achieved through a stepper motor and a position encoder, with a system feedback control accuracy of ±0.1 degrees, ensuring high repeatability of light incidence at different angles.

[0056] While the membrane surface is illuminated at each angle, a receiving unit positioned in the reflection direction collects the spectral signal reflected back from the membrane. This receiving unit employs a fiber-coupled high-sensitivity linear CCD spectrometer with a spectral response range of 380 to 1100 nanometers and a resolution better than 0.8 nanometers, enabling efficient recording of the reflection intensity changes of incident light at various wavelengths. During acquisition, the detection angle of the receiving unit maintains a reflection symmetry with the incident angle; that is, when the incident light is at 15 degrees, the reflection receiving device also collects the signal at the same angle in the opposite direction to the membrane. The spectrometer's input is equipped with a light-gathering fiber optic probe and an anti-stray light filter to improve the spectral purity of the measurement and suppress ambient light interference. Each spectral data acquisition includes one dark current recording, one background sampling, and one actual reflection signal recording, and the integration time is automatically adjusted to adapt to the reflection intensity changes at different incident angles. Finally, a complete reflection spectrum curve consisting of wavelength and corresponding reflected light intensity is output for each angular position.

[0057] Multiple spectral curves obtained from the aforementioned angles were numbered and categorized, and then uniformly input into the data processing unit. Data processing employed a two-dimensional matrix format to organize all reflectance spectral information. The row dimension of the matrix represented various incident angles, such as 0 degrees, 5 degrees, 10 degrees up to 45 degrees, while the column dimension represented various sampling wavelengths, such as 400 nm, 401 nm up to 1000 nm. The corresponding cell stored the reflected light intensity value at that wavelength and angle. Before building the matrix, all spectral curves underwent standardized preprocessing operations, including reflectance normalization of the reference curve measured using a standard white board to eliminate systematic errors caused by fluctuations in light source intensity; and smoothing out background trend lines in the curves using polynomial fitting to enhance the contrast of local variation features such as spectral peaks and valleys. The final constructed spectral data matrix can completely record the subtle differences in the reflectance characteristics of the membrane material under multi-angle illumination, providing a highly reliable data foundation for subsequent coating uniformity analysis.

[0058] After constructing the spectral data matrix, observing the response trend of the reflection spectrum with the incident angle can provide a preliminary assessment of whether interference features caused by the physical structure exist on the film surface. When the PVC film surface has fine periodic embossed structures, directional stretching stripes, or optical interface contrasts formed by multilayer composites, its reflection spectrum often exhibits significant periodic enhancement or attenuation with angle changes in certain specific bands. These phenomena are not caused by the coating thickness or composition itself, but rather by interference spectral signals caused by the microstructure of the film surface and interface reflection. The presence of such interference signals manifests as a regular stripe distribution in the spectral image, with strong regularity and high repeatability, easily masking the true thickness variations or uneven distribution. Therefore, it is crucial to highlight these band regions with periodic fluctuation characteristics in the spectral data matrix and retain the original data matrix as an important data basis for identifying interference signals and separating the differences in the actual coating in subsequent steps.

[0059] The core function of this step is to provide a complete spectral response data foundation from multiple angles and wavelengths for subsequent coating uniformity assessment, while identifying and retaining interference characteristic signals that may affect detection accuracy. In the actual manufacturing process of polyvinyl chloride (PVC) films, surface microstructures or multi-interface reflective layers are often introduced due to processes such as calendering, stretching, and thermal lamination. These structures cause complex optical interference effects at different incident angles and wavelengths, resulting in periodic fluctuations or abnormal enhancement / attenuation of the film's reflectance spectrum. If a fixed or single-angle spectral acquisition is used, angular coupling can easily lead to some interference signals being misinterpreted as the true coating response, resulting in misjudgments or missed detections. Therefore, by configuring an illumination device with a controllable incident angle to perform continuous angular scanning of the film surface, the spectral reflectance characteristics of the same detection area at different incident angles can be obtained, establishing a two-dimensional spectral data matrix containing both angular and wavelength dimensions. This not only enriches the spectral data dimensions but also exposes the interference-induced variation modes at multiple angles, thus facilitating accurate separation of optical interference from true coating characteristics. In addition, this data matrix can also serve as the input data source for key processes such as subsequent uniformity analysis, thickness fluctuation detection, and deviation area identification. It is a fundamental step in realizing non-contact, high-precision, and interference-resistant coating detection, and has significant advantages such as data integrity, response sensitivity, and result traceability.

[0060] S002, perform continuous wavelet transform processing on the spectral data matrix, decompose the spectral response signal into a multi-scale local decomposition, extract and separate the periodic interference feature signals, and obtain the purified spectral matrix after interference reduction.

[0061] To extract interference signals caused by factors such as periodic microstructure and interlayer interface reflections from spectral data collected at different incident angles of polyvinyl chloride (PVC) films, and thus retain spectral information that truly reflects coating thickness and composition variations, precise signal separation and purification are required on the constructed spectral data matrix. Therefore, this step employs a multi-scale spectral response analysis method based on continuous wavelet transform to specifically identify and remove interference components at the signal composition level, obtaining a purified spectral matrix with clear structure and accurate characteristics. The specific steps are as follows:

[0062] The original spectral data matrix containing multiple incident angles and multiple wavelength sampling points is extracted. Each row represents a specific incident angle (e.g., 0°, 5°, 10°, 15°, 20°, 25°, 30°, 35°, 40°, 45°), and each column represents a specific wavelength point (601 wavelength points ranging from 400 nm to 1000 nm, spaced 1 nm apart). To ensure consistency and high-precision analysis in subsequent processing, all curves need to be resampled and aligned along the wavelength axis to ensure that each spectral curve has the same wavelength coordinate axis distribution. This step achieves normalization of the spectral data through interpolation reconstruction, eliminating lateral bias caused by uneven instrument sampling intervals.

[0063] After unified sampling, continuous wavelet transform analysis was performed on each spectral curve. The mother wavelet function used was MexicanHat (i.e., second-order Gaussian derivative wavelet), which has strong local abrupt change detection and edge response capabilities, and is suitable for detecting periodic high-frequency components in the spectrum caused by interference. In practice, each spectral curve was continuously transformed at 20 scale levels, starting from a minimum scale of 2 corresponding to wavelength resolution and reaching a maximum of 512 to cover the entire spectral range. During the transformation, a corresponding wavelet coefficient sequence was output at each scale level, representing the intensity and structural distribution of spectral changes at that scale. All wavelet coefficients constituted a two-dimensional wavelet spectrum, from which it could be observed that periodic interference fluctuations typically concentrated in the mid-frequency region between scales 8 and 32.

[0064] For the aforementioned wavelet spectra, interference signals are identified and extracted using a fixed-scale window. Specifically, statistical detection is performed on all wavelet coefficients at six designated scales: 8, 12, 16, 20, 24, 28, and 32. When more than 10 consecutive wavelength points at a certain scale exhibit high-amplitude periodic fluctuations, with adjacent peaks exhibiting relatively equal intervals, and their average amplitude exceeding twice the average value of the corresponding curve at that scale, the signal segment is identified as the dominant interference region. After identifying the dominant interference region, a soft-threshold attenuation strategy is used to reduce the amplitude of its wavelet coefficients by 80%. This significantly weakens the interference signal while preserving the trend background, rather than abruptly eliminating it, thus avoiding disruption of the overall spectral morphology and the continuity of the coating's effective response.

[0065] After suppressing the interference signals at all wavelet scale levels, the inverse transform of the spectral curves was performed using only the wavelet results with high confidence coefficients to reconstruct the purified spectral response curves at each angle. The purified curves at all angles were then recombined to restore a two-dimensional purified spectral data matrix that is completely consistent with the original data structure. In this matrix, the main periodic interference features of each data point have been eliminated, while the true optical responses closely related to the coating's physical thickness and material composition, such as absorption peaks, edge slopes, and reflection depressions, are fully preserved. This provides a high-quality data foundation for subsequent calculations of coating thickness anomaly localization and deviation intensity index.

[0066] The main function of this step is to separate and remove periodic interference signals caused by micron-level periodic structures, multilayer composite interfaces, or orientation textures on the surface of the PVC film from the original spectral data matrix. This allows for the extraction of more accurate spectral information reflecting coating thickness variations, component distribution differences, and uniformity anomalies. During spectral detection, the multiple interference effects caused by the aforementioned physical structures result in periodic increases or decreases in the spectral response curve within a specific wavelength range. These signals are highly regular and repeatable, easily misinterpreted by traditional analysis methods as normal and consistent spectral characteristics, thus masking minor thickness deviations or uneven coating in local areas. By applying continuous wavelet transform to the spectral data matrix, the spectral signal can be locally decomposed at multiple scale levels, extracting high-frequency, low-frequency, and mid-frequency components. The distribution at each scale level identifies the dominant periodic interference features, and soft threshold suppression or filtering effectively reduces these interference components, allowing the reconstructed purified spectrum to more accurately represent the physical characteristics of the coating on the film surface. This step is crucial for achieving high-precision, non-contact, and interference-resistant coating quality inspection. It not only improves the reliability of spectral signal analysis but also provides an accurate data foundation for subsequent abnormal area identification, deviation intensity calculation, and compensation control. It effectively avoids misjudgment and missed detection caused by interference pseudo-signals, and improves the stability and intelligence level of the entire inspection process.

[0067] S003, based on the purification spectral matrix, a support vector machine classification model is constructed to classify the spectral feature vectors of each detection location, and to identify and locate the spatial regions where there are abnormal fluctuations in coating thickness.

[0068] To identify regions of abnormal coating thickness on the surface of polyvinyl chloride (PVC) films, after reducing the interference signal of the original spectral data and obtaining a purified spectral matrix, the purified spectral information is further used to classify each detection location, thereby determining which areas on the film surface exhibit abnormal thickness fluctuations. This classification process uses the spectral response of each location in the purified spectral matrix as input. By constructing a classification model based on differences in physical characteristics, it outputs the spatial assignment status of each detection point, laying the foundation for subsequent coating anomaly localization and boundary extraction. The specific steps are as follows:

[0069] Spectral response data for each detection point in the wavelength range of 400 nm to 1000 nm were extracted from the purified spectral matrix. Each spectral curve contains 601 data points, corresponding to 601 uniformly spaced wavelength sampling positions. To ensure that data from different detection points are compared on the same numerical scale, each spectral curve is first normalized according to its own maximum and mean values, so that all reflectance intensity values ​​fall uniformly between 0 and 1, while maintaining the overall shape. This normalization process helps to eliminate absolute intensity shifts caused by differences in local illumination conditions or differences in the background reflectance of the membrane material, improving the robustness and stability of overall identification.

[0070] For all normalized spectral curves, the most significantly changing feature values ​​are extracted and used as input features for classification at each detection location. The selected features include, but are not limited to, the following: first, the slope of reflectance changes at the 550 nm, 700 nm, and 850 nm wavelengths; second, the degree of reflectance jump at the boundary between the visible and near-infrared regions; and third, the degree of symmetry and curvature continuity of the overall spectral curve. These features are represented numerically and combined in a fixed order to form a feature sequence with practical physical meaning, ensuring that each detection point has a unified and clear input dimension in subsequent judgments.

[0071] A pre-constructed physical response classification rule is used to determine the characteristics of all detection locations. This rule establishes judgment conditions based on the known differences in spectral reflectance morphology between normal coating areas and abnormal thickness areas. For example, if the reflectance slope of a detection point at a wavelength of 700 nm is less than a preset threshold, and its overall curve shows discontinuous jumps in the visible light region, then there may be a sudden change in thickness or an abnormal optical composition at that location. For each detection point, its spectral characteristics are matched with the above conditions item by item, and the region is marked according to the matching results. Each point is determined to be either a "normal region" or an "abnormal region," and the results are recorded in a binary format.

[0072] "Pre-constructed physical response classification rules" refer to a set of spectral characteristic judgment criteria used to distinguish between "normal coating areas" and "areas with abnormal thickness" based on a systematic analysis and experimental verification of the spectral response characteristics of a large number of known film samples, before formal testing of PVC film coatings. These classification rules do not rely on abstract mathematical algorithms or machine learning models, but rather stem from a physical understanding of the actual spectral response behavior under different film layer states. Specifically, the construction of these rules is based on the following aspects:

[0073] By comparing the spectral curves of multiple normal samples and multiple samples with abnormal thickness at different wavelengths (such as 550 nm, 700 nm, and 850 nm), it was found that the thickness abnormal region often exhibits characteristic changes such as a decrease in slope, a sudden change in reflectance, a concave curve, or an anomalous peak position shift.

[0074] By further combining the mapping relationship between coating thickness and spectral absorption, the critical range of reflectivity decrease or jump at a specific wavelength was determined;

[0075] Experiments confirmed that these characteristics are universal across different batches and compositions of PVC membrane materials, demonstrating sufficient distinguishing ability. Therefore, the constructed classification rules are essentially a set of explicit physical criteria based on spectral morphology changes, such as "a region is considered abnormal if its reflectance decreases by more than 12% at a wavelength of 700 nm and the rate of change of slope in adjacent bands exceeds a certain threshold." These rules have a clear structure, can be directly programmed, and possess advantages such as high stability, high interpretability, and wide adaptability, making them suitable for rapid and accurate preliminary classification of membrane material uniformity in industrial testing processes.

[0076] Based on the above classification results, all detection points are reconstructed into a distribution map corresponding to the membrane structure according to their two-dimensional spatial positions. In this distribution map, all detection points identified as "abnormal areas" are marked with clear coordinates, and further spatial continuity analysis is performed. This determines which abnormal points form contiguous areas and which are isolated interference signals. Isolated points with an area less than 2 square millimeters that cannot form a continuous distribution are removed; for areas where the distance between adjacent points is too far, interpolation is used to fill in the missing boundary parts, ensuring that each abnormal area has a closed boundary and a clear outline. The final output is a complete spatial positioning map that accurately shows all areas on the membrane surface with abnormal thickness fluctuations, and the corresponding positional accuracy can be controlled within millimeters.

[0077] This step, based on the purified spectral matrix that has already eliminated interference signals, further refines the spectral response data of each detection location on the PVC film, performing fine classification and spatial localization to identify specific areas where the coating exhibits abnormal thickness fluctuations. This process is a crucial step in the coating uniformity detection workflow, its core objective being to effectively transform subtle differences in the spectrum into spatial divisions of "normal" and "abnormal" regions, providing accurate input for subsequent thickness deviation quantification analysis and spraying compensation control. The purified spectral matrix, during preprocessing, effectively eliminates periodic interference signals caused by the microstructure of the film surface or interface reflections, ensuring that the spectral curve of each detection point more accurately reflects the physical properties of the coating itself. Based on this, by extracting physically meaningful feature parameters such as reflectance slope, local extrema, and spectral symmetry from each purified spectral curve, a spectral feature vector reflecting the coating thickness variation trend is constructed. Using these feature vectors as input, combined with a predefined classification model, each detection point is individually assessed to identify whether significant thickness anomalies exist. After classification, the corresponding spatial distribution map is constructed using the two-dimensional spatial coordinates of the detection points, enabling visual localization and contour determination of abnormal regions. This step not only enables high-resolution difference identification in the spectral signal dimension, but also allows for precise marking of abnormal areas in the spatial dimension, effectively improving the sensitivity and accuracy of detection. It is a crucial supporting element for realizing a high-precision non-contact coating quality inspection closed-loop control system.

[0078] S004, extract the spatial boundary information and spectral response gradient parameters of the thickness anomaly area, establish a coating thickness deviation intensity calculation model, and output the coating deviation intensity index of the corresponding detection area;

[0079] To achieve accurate quantitative analysis of non-uniform PVC film coating thickness regions, after spatial identification of abnormal areas, it is necessary to further extract the boundary location and internal spectral reflectance variation characteristics of each abnormal region. Based on this, a numerical index reflecting the degree of coating thickness fluctuation in that region, namely the coating deviation intensity index, needs to be constructed. This index can be used to determine the severity of different abnormal regions, providing a clear reference for subsequent spraying compensation. The specific steps are as follows:

[0080] Spatial boundaries are extracted for detection points marked as anomalies. Based on the coordinates of each detection point on the two-dimensional plane of the membrane material, connected detection points are merged into several independent anomaly regions. Each anomaly region is represented by its smallest covered rectangle boundary. The coordinates of the top-left and bottom-right corners of the rectangle are recorded, and the corresponding area is calculated. To improve the accuracy of the boundaries, the coordinates of all boundary points are processed by a five-point weighted average according to their spatial order, eliminating abrupt edges and outliers. This makes the boundaries of each region smoother and more continuous, avoiding deviations in area or location judgments due to boundary irregularities in subsequent processing.

[0081] Within each anomalous region, key characteristic parameters of the purification spectral response were extracted from all detection points. The spectral response curve for each detection point covered the wavelength range of 400 nm to 1000 nm, with the 550 nm to 800 nm range selected as the primary analysis range. Three indicators were extracted: first, within this wavelength range, the segment with the fastest changing slope on the curve was identified and its maximum slope value was recorded to characterize abrupt changes in spectral morphology; second, the reflectance difference between the maximum peak and minimum trough within this segment was calculated to reflect the local fluctuation amplitude; and third, the reflectance difference at wavelengths of 650 nm and 900 nm was calculated to measure the overall trend shift. These three indicators collectively reflect the spectral discontinuities and absorption differences that may be caused by thickness variations, possessing clear physical significance and statistical comparative value.

[0082] For all detection points within each abnormal region, numerical processing is performed according to the three indicators mentioned above. First, the 10% of data points with the largest and smallest variations are identified from all detection points within the region. Their average values ​​are calculated, and the difference between these averages is taken as the "gradient difference" for that region. This gradient difference represents the extreme amplitude of spectral variation within the region, indirectly reflecting the instability of the coating thickness. Next, the gradient difference is multiplied by the square root of the area to calculate the final result of the coating deviation intensity index. This index is expressed as a dimensionless real number; the larger the value, the more uneven the coating and the more severe the thickness fluctuations in that region.

[0083] Each abnormal region's ID, spatial coordinate boundaries, and corresponding coating deviation intensity index are output and mapped to form a two-dimensional spatial distribution map. This map uses different colors to represent different intensity levels; for example, green represents low-intensity areas, yellow represents medium-intensity areas, and red represents high-intensity areas. This allows operators to quickly understand the degree of fluctuation in each abnormal region and clearly identify locations requiring focused intervention. In this way, coating non-uniformity is transformed from a qualitative assessment into a quantifiable and gradable numerical indicator, facilitating precise matching of subsequent spraying control actions to the actual problem location and significantly improving the targeting and efficiency of compensation operations.

[0084] This step aims to further quantitatively analyze the identified coating thickness anomalies to assess the severity and physical differences of various anomalies, thus providing a precise basis for subsequent local coating compensation and quality control decisions. The previous steps only classified and located the uneven coating areas, constituting preliminary identification. This step focuses on quantifying these anomalies by extracting their spatial boundary range and spectral reflectance variation characteristics to calculate a numerical index that accurately reflects the degree of thickness fluctuation, known as the coating deviation intensity index. Specifically, it first requires accurately defining the spatial location and boundary range of each anomaly on the film surface. Through coordinate extraction, connectivity assessment, and edge smoothing, a clear and closed spatial description is formed. Then, within this region, based on the spectral response curve of each detection point, thickness-sensitive characteristic parameters are extracted, such as the reflectance difference between specific wavelengths, the degree of slope change in the spectral curve, and the fluctuation amplitude of absorption peaks or depressions. These physically strong indicators can reveal the severity of coating thickness variations in this region. Next, based on the spatial distribution trends of these indicators, a value representing the intensity of thickness fluctuations within the region is calculated—the coating deviation intensity index. This index not only reflects the degree of spectral difference but also incorporates the actual area of ​​the region, making it both representative of the intensity of local anomalies and comparable in an engineering sense. Finally, the index results for each region are categorized, sorted, and visualized, providing crucial reference for subsequent decisions on whether to implement compensation control, the priority of compensation control, and the control parameters. Therefore, this step plays a vital role in the entire detection process, serving as a key link from "problem detection" to "precise response."

[0085] S005, Based on the coating deviation strength index, construct dynamic coating control rules to control the moving speed of the spray head and the flow rate of the coating liquid in the coating equipment, and perform local compensation operations for areas with abnormal thickness.

[0086] To effectively control areas with abnormal PVC film coating thickness, it is necessary to obtain the coating deviation strength index for each detection area, formulate corresponding control rules based on this index, and transmit control commands to the key actuators of the spraying equipment. This allows for dynamic adjustment of the spray head's movement speed and coating flow rate, thereby completing the local compensation operation for areas with abnormal thickness. The specific steps are as follows:

[0087] A one-to-one correspondence is established between the coating deviation intensity index and control commands. The deviation intensity index is a numerical expression of the degree of coating thickness fluctuation, divided into five levels according to its value: Level 1 (0.0-1.9), representing slight thickness fluctuation; Level 2 (2.0-3.9), representing relatively obvious but not drastic fluctuation; Level 3 (4.0-5.9), representing moderate thickness anomalies; Level 4 (6.0-7.9), representing a large fluctuation range; and Level 5 (8.0 and above), representing severe non-uniformity. Specific control parameters are defined for different levels: Level 1 areas perform standard spraying without compensation; Level 2 areas reduce the spray head movement speed by 10% and increase the coating flow rate by 8%; Level 3 areas reduce the speed by 15% and increase the flow rate by 15%; Level 4 areas reduce the speed by 25% and increase the flow rate by 20%; and Level 5 areas reduce the speed by 40% and increase the flow rate by 30%. This parameter table forms the basis of the dynamic control rules, ensuring that areas with different fluctuation levels receive corresponding local compensation.

[0088] Based on the spatial coordinate matching between the deviation intensity index and the membrane material conveying direction, a control time axis is established within the coating path. The start and end positions of each abnormal area on the membrane surface are determined using extracted boundary coordinates, and then converted into entry and exit points in the coating time by combining the membrane material's running speed on the conveyor belt. The control system generates an execution sequence for spray head adjustment based on these time points. For example, if a level 5 abnormal area is distributed from 1250 mm to 1320 mm, and the conveyor belt speed is 500 mm per minute, the control system will activate the compensation command between 150 seconds and 158.4 seconds after the membrane enters the system. During this period, the spray head moving speed is automatically adjusted to 60% of the standard speed, and the coating liquid flow rate is simultaneously increased to 130% of the normal flow rate to ensure that the area receives sufficient compensation coverage.

[0089] During actual operation, the spraying equipment receives control commands in real time and dynamically adjusts the spray head's movement trajectory and flow output. The spray head movement is driven by a linear motor, with continuous position feedback via a high-precision encoder, ensuring that the speed change accuracy does not exceed ±0.5 mm / s. The coating output is controlled by a combination of an electronically controlled pump and a pressure regulating valve, with a response time of less than 100 milliseconds, allowing for a sudden increase in output flow when the spray head enters an abnormal zone. Simultaneously, the opening and closing time of the spray head nozzles is adjusted synchronously to prevent over-spraying or under-spraying at the boundaries. All execution parameters are written to a traceable database and visualized in real time on the control panel display, facilitating operator monitoring of the control effect at any time.

[0090] After the membrane coating is completed, the spectral reflectance of the coated area is re-detected to check whether the thickness within the compensation area meets the set target. If the thickness of some areas is still below the standard lower limit, the system records the area number and deviation value. In the next batch of membrane processing, the control parameter of the area with that number is increased by one level to achieve continuous improvement. This feedback mechanism enables the control strategy to learn and automatically adapt to external factors such as batch differences in materials, nozzle aging, and changes in coating viscosity, thereby building a truly stable and continuously optimized dynamic control process.

[0091] This step transforms the "coating deviation intensity index" obtained from the preliminary detection and analysis into specific process control instructions, thereby achieving precise coating compensation for areas with abnormal thickness on the PVC film surface. This process is the core closed-loop link in the entire coating uniformity detection and control system. Its goal is not only to identify non-uniformity issues but also to dynamically respond and correct them on-site based on the severity of the problem. Through this step, a set of corresponding spraying control rules can be constructed based on the deviation intensity index level of each abnormal area, including parameters such as the spray head's moving speed, nozzle opening duration, and coating liquid flow rate adjustment ratio. When the detection system identifies a region where the coating is significantly thinner and the deviation index is at a high level, the system will instruct the spray head to reduce its moving speed when passing through that region and simultaneously increase the coating liquid output, so that the local area receives additional material coverage while maintaining coating uniformity, thereby gradually correcting the problem of insufficient local thickness. This control process does not rely on a fixed template but is based on real-time data and differentiated processing according to the deviation levels of different regions, possessing high flexibility and adaptability. By implementing this step, a shift from "static spraying" to "data-driven dynamic spraying" can be achieved, significantly improving the accuracy and response efficiency of coating consistency control, reducing material waste during the coating process, enhancing the overall quality stability and yield of membrane products, and providing reliable manufacturing assurance for high-end applications such as functional membranes and dielectric membranes that have strict thickness requirements.

[0092] S006 After completing the local compensation operation, S001 to S005 are re-executed in the compensation area to update the spectral data matrix and repeat the interference identification, anomaly extraction and coating control process to construct an integrated closed-loop mechanism for the detection and control of polyvinyl chloride film coating uniformity.

[0093] To ensure that areas with abnormal PVC film coating thickness reach the expected uniformity standard after compensation, and to verify the compensation effect and achieve dynamic continuous optimization, after completing the local compensation coating operation, the entire detection and control process must be re-executed for the compensated areas. This involves re-acquiring multi-angle spectral data, updating the spectral data matrix, identifying potential interference features, extracting abnormal areas, and executing a new round of coating control. This establishes a truly closed-loop quality control mechanism of detection-judgment-compensation-re-detection, enabling dynamic tracking and correction throughout the entire process. The specific steps are as follows:

[0094] After the local compensation operation is completed, the membrane material is passed through the spectral acquisition area again, and the original compensation area is re-scanned with multi-angle illumination using the same acquisition method as the initial test. The specific process includes: using a controllable incident light device with high-precision rotation function, scanning at 5-degree intervals from the vertical incident angle to 45 degrees, acquiring reflectance spectral intensity data in the 400 nm to 1000 nm wavelength band at each angle. Background correction and dark current removal are performed simultaneously during the acquisition process to ensure the accuracy and absence of clutter interference for each spectral curve. This results in a second round of spectral data matrix for the compensation area, providing updated data for subsequent assessments.

[0095] The newly acquired spectral data matrix undergoes further interference feature identification and purification. Specifically, this involves analyzing the spectral response curve for each incident angle, extracting high-frequency oscillation components at the local scale using wavelet decomposition, and identifying spectral interference signals with significant periodicity. The identified interference components are then weakened using a soft thresholding process, followed by inverse transform reconstruction of the spectral curves to obtain a purified spectral matrix with interference components removed. This step ensures that the current detection is no longer affected by any new optical structures or surface textures that may be introduced after compensation, guaranteeing a true reflection of the coating's bulk characteristics.

[0096] Based on the response changes at each detection point in the purification spectral matrix, a new spectral feature analysis is performed on each detection location, and key parameters such as reflectance slope, absorption difference, and band jump degree are extracted to determine whether thickness anomalies still exist in that area. Combining the spatial location of each detection point, a coating thickness anomaly distribution map is reconstructed. If, within the compensated area, some locations still exhibit spectral changes exceeding a set threshold, the compensation at that location is deemed insufficient, and it must be reclassified as an anomaly area and reinstated as a control target; otherwise, the compensation is considered effective, marked as a "compliant area," and no further coating is required. This process involves point-by-point judgment and clear classification, ensuring accurate feedback for each compensation action.

[0097] For locations identified as "insufficiently compensated" in this round of inspection, the system automatically invokes the coating deviation strength calculation rules and upgrades the compensation parameters based on the new round of index results. It then resets the spray head's movement speed and coating output ratio, and performs the compensation operation again for that area in the next round of spraying. Simultaneously, for areas confirmed to meet standards, the control process excludes them from subsequent spraying plans to avoid overspraying or excessive thickness. Through this continuous cycle of inspection-feedback-re-control, the compensation results are continuously optimized until the thickness of all inspected areas reaches the set standard. The entire process forms a high-precision, high-response, and end-to-end closed-loop quality control mechanism, achieving a complete closed-loop adjustment path from spectral detection, coating anomaly identification, local compensation execution, to compensation effect review and readjustment. This significantly improves the stability and control efficiency of PVC film coating uniformity, making it particularly suitable for continuous manufacturing scenarios of functional film materials requiring highly consistent film thickness.

[0098] This step elevates the process of detecting and compensating for the uniformity of PVC film coatings from a single-execution mode to a dynamic, continuously iterative, and optimized closed-loop control mechanism, thereby achieving the closed-loop control objective of the entire process quality management. Previous steps identified areas of abnormal coating thickness, calculated deviation intensity, and performed compensation operations. However, in actual industrial production, local compensation may fail to achieve the expected results due to factors such as execution deviations, differences in material flowability, equipment inertial response, or edge effects. Therefore, a single compensation cannot be considered sufficient for acceptance. Through this step, after completing a local compensation operation, the system re-executes the entire process from incident light angle scanning (S001), spectral data acquisition and purification (S002), spectral classification judgment (S003), spatial anomaly location (S004), to spray control rule execution (S005). This specifically targets the compensation area with "re-inspection—re-judgment—re-compensation," not only confirming the effectiveness of the compensation but also allowing for precise secondary or even multiple compensations for areas still experiencing problems. This ensures that coating adjustments are truly based on feedback-driven logic. This step tightly couples the detection data with the control actions, continuously narrowing the thickness fluctuation range through repeated closed-loop updates of the spectral data matrix and dynamic adjustments to the compensation behavior until all compensated areas reach the set thickness consistency standard. This closed-loop mechanism enables the system to continuously learn and self-correct, significantly improving the control accuracy and response sensitivity of coating uniformity. It is particularly suitable for the preparation of functional film materials in high-volume continuous production scenarios where the requirements for handling local defects are extremely high.

[0099] The aforementioned method for detecting the uniformity of PVC film coatings based on spectral analysis significantly improves the accuracy of identifying and compensating for true non-uniformity issues in coatings under the influence of microstructure interference, demonstrating significant engineering application value and technological breakthrough. Specifically, this method effectively incorporates joint information from the spectral and angular dimensions through multi-angle reflectance spectral acquisition, overcoming the limitations of traditional single-angle detection which is susceptible to interference-induced errors. Combined with continuous wavelet transform, it extracts and removes periodic interference signals, significantly purifying spectral data and providing an accurate baseline for subsequent identification. A classification model based on the purified data is introduced to accurately locate thickness anomaly regions, and an intensity index is constructed through spatial boundary identification and gradient extraction to quantify the degree of deviation. This is then linked to the coating execution process, enabling dynamic adjustment of the spray head speed and coating flow rate. Finally, a complete closed-loop mechanism is constructed through re-detection and feedback of the compensation area. The overall process forms a closed-loop path from spectral data acquisition, interference suppression, anomaly identification, compensation control to effect re-verification, achieving multi-dimensional perception and dynamic correction of coating thickness fluctuations. This effectively avoids the risk of missed detection caused by pseudo-uniformity due to interference, significantly improving the detection accuracy and production control capabilities of PVC film coating uniformity.

[0100] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for detecting the uniformity of PVC film coatings based on spectral analysis, characterized in that, Includes the following steps: S001 uses a controllable incident light system with multi-angle rotation function to continuously scan the surface of polyvinyl chloride film at different angles, and collects the reflection spectral intensity data in the corresponding band range according to different incident angles to generate a spectral data matrix containing interference characteristics. S002, perform continuous wavelet transform processing on the spectral data matrix, decompose the spectral response signal into a multi-scale local decomposition, extract and separate the periodic interference feature signals, and obtain the purified spectral matrix after interference reduction. Step S002 includes: Extract the original spectral data matrix containing multiple incident angles and multiple wavelength sampling points, and resample and align the spectral curves on the wavelength axis through interpolation reconstruction. For each spectral curve, a continuous wavelet transform with MexicanHat as the mother wavelet function is performed to generate a sequence of wavelet coefficients at multiple scales. Within a specified range of scales from 8 to 32, regions of continuous high-amplitude periodic fluctuations are identified, and their wavelet coefficients are attenuated using a soft-threshold compression method. Based on the adjusted wavelet coefficients, an inverse transform operation is performed to reconstruct the purified spectral curves at each incident angle, and the curves are combined to generate a purified spectral data matrix. S003, based on the purification spectral matrix, a support vector machine classification model is constructed to classify the spectral feature vectors of each detection location, and to identify and locate the spatial regions where there are abnormal fluctuations in coating thickness. S004, extract the spatial boundary information and spectral response gradient parameters of the thickness anomaly area, establish a coating thickness deviation intensity calculation model, and output the coating deviation intensity index of the corresponding detection area; Step S004 includes: Spatial aggregation processing is performed on the detection points that are identified as anomalies to extract the minimum rectangular boundary of each connected anomaly region, and the boundary coordinates are smoothed by a five-point weighted average method. Extract the purification spectral curves of each detection point in the 550 nm to 800 nm band within each abnormal region, and calculate the maximum slope, reflectance difference, and reflectance offset between specified wavelengths respectively. Select the 10% and 10% of the detection points with the largest and smallest changes in each region, calculate the difference between the average values ​​of the three indicators, and multiply it by the square root of the region area to obtain the coating deviation strength index. Output the corresponding numbers and boundary coordinates of all abnormal areas with the coating deviation strength index, and draw a two-dimensional spatial distribution map of different strength levels; S005, Based on the coating deviation strength index, construct dynamic coating control rules to control the moving speed of the spray head and the flow rate of the coating liquid in the coating equipment, and perform local compensation operations for areas with abnormal thickness. S006 After completing the local compensation operation, S001 to S005 are re-executed on the compensation area to update the spectral data matrix and repeat the interference identification, anomaly extraction and coating control process to construct an integrated closed-loop mechanism for the detection and control of polyvinyl chloride film coating uniformity.

2. The method for detecting the uniformity of PVC film coating based on spectral analysis according to claim 1, characterized in that, Step S001 includes: By configuring an incident light irradiation device with controllable rotation angle function, the surface of polyvinyl chloride film is continuously scanned from 0 degrees to 45 degrees, and reflectance spectrum data is collected every 5 degrees. At each incident angle, the light is illuminated by a white LED array light source with a color temperature of 5500K to 6000K and a wavelength range of 400 to 1000 nanometers, and a parallel beam is output by an optical path shaping device with a collimating lens group. The reflected signal is acquired simultaneously using a fiber-coupled linear array CCD spectrometer with a spectral response range of 380 to 1100 nanometers and a resolution better than 0.8 nanometers. The reflection receiving direction is symmetrically configured with the incident direction. All collected data were normalized for reflectance, removed for background trends, and classified by number. Finally, a two-dimensional spectral data matrix was constructed, with the row dimension being the incident angle, the column dimension being the sampling wavelength, and the matrix unit being the reflection intensity value.

3. The method for detecting the uniformity of PVC film coating based on spectral analysis according to claim 1, characterized in that, Step S003 includes: Extract the spectral response curve of each detection point in the purification spectral matrix in the wavelength range of 400 nm to 1000 nm, and normalize each curve. Feature values, including band slope, reflectance jump amplitude, and curvature continuity, are extracted from each normalized spectral curve to construct a standardized feature sequence; Based on the pre-established physical response classification rules, the spectral characteristics of all detection points are compared and judged, and the classification result corresponding to each detection point is output. Based on all classification results, a two-dimensional spatial positioning map of the membrane surface is constructed, the boundary positions of abnormal areas are marked and isolated interference points are eliminated, and the spatial positioning of thickness abnormal areas is completed.

4. The method for detecting the uniformity of PVC film coating based on spectral analysis according to claim 1, characterized in that, Step S005 includes: Based on the coating deviation strength index of each test area, it is divided into five levels, and the adjustment ratio of the spray head moving speed and the coating flow rate is set accordingly. The spatial boundary position of the abnormal area is converted into the time coordinate of the membrane material transportation process to generate the control time axis for spraying compensation. During the compensation period, the spray head is controlled to adjust its moving speed according to a set ratio, and the flow rate of the coating liquid is increased simultaneously. After the spraying is completed, the spectral reflectance data of the coated area is re-detected, and the control parameters for the next round are adjusted based on the compensation effect.

5. The method for detecting the uniformity of PVC film coating based on spectral analysis according to claim 4, characterized in that, The specific steps for controlling the spray head to adjust its moving speed according to a set ratio and simultaneously increasing the coating flow rate during the compensation period are as follows: During the compensation period, the spray head is controlled by a linear motor to operate within the target speed range, and the speed adjustment accuracy is controlled within ±0.5 mm per second. Meanwhile, the output flow rate of the coating liquid is adjusted by a combination of an electronically controlled pump and a pressure regulating valve, with a flow response time of less than 100 milliseconds, to ensure that the compensation area completes the synchronous adjustment of the spray head speed and the coating liquid flow rate at the moment of entry.

6. The method for detecting the uniformity of PVC film coating based on spectral analysis according to claim 1, characterized in that, Step S006 includes: After local compensation is completed, multi-angle illumination scanning is performed again to collect reflectance spectral intensity data of the compensation area in the 400 nm to 1000 nm band. The newly acquired spectral data matrix is ​​subjected to interference recognition and purification processing to obtain an updated purified spectral matrix; Based on the purification spectral matrix, abnormal areas are re-identified to determine whether there are still locations with insufficient compensation. For areas with insufficient compensation, the deviation intensity index is recalculated and the compensation parameters are adjusted. The control steps are repeated to achieve closed-loop regulation.

Images