Method for controlling the separation and regeneration performance of cleanroom filter contaminants

Abstract

This invention relates to a method for controlling the separation and regeneration performance of cleanroom filter media in the field of information technology. The method includes: acquiring pollutant distribution data of the filter media; extracting feature vectors and classifying pollutant clusters based on the pollutant distribution data; calculating density and particle size distribution from the feature vectors to determine pollutant type classification; using a separation priority sequence to guide a laser-assisted stripping device to perform layer-by-layer scanning and removal of the filter media surface; acquiring real-time resistance change data during the removal process to determine the filter media cleanliness index; if the filter media cleanliness index does not reach a preset threshold, injecting a neutral solvent through an ultrasonic impregnation module to perform secondary dissolution of residual pollutants, obtaining updated filter media state data after dissolution; running a wind resistance simulation model based on the updated filter media state data; extracting ventilation parameters from the simulation results to determine the filter media regeneration performance compliance level.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for controlling the performance of pollutant separation and regeneration of cleanroom filter media. Background Technology

[0002] High-efficiency dust removal and filtration devices for cleanrooms play a crucial role in air purification in high-precision manufacturing fields such as semiconductors, optics, and pharmaceuticals. Their filter cartridges or bags directly determine the cleanliness level of the production environment and the stable operation of the equipment. Because these devices are constantly exposed to high-concentration dust, a large amount of pollutants quickly accumulates on the surface and internal pores of the filter media, leading to a sharp increase in ventilation resistance. This, in turn, affects system airflow and energy consumption, becoming a consumable issue that must be addressed periodically on the production line. Traditional cleaning methods mainly rely on physical beating, simple air backflushing, or ordinary water washing. These methods are clearly insufficient when dealing with complex dust. Dust is often not a single type but a mixture of various particle sizes, densities, and properties. After cleaning, the surface may appear clean, but a stubborn adhesion layer remains inside, forming a difficult-to-remove "mud cake" structure that permanently increases the filter media resistance.

[0003] Meanwhile, the cleaning intensity is difficult to control precisely, easily pulling or breaking filter media fibers, leading to a significant decrease in filtration efficiency and a drastically shortened service life. More importantly, there is a lack of reliable performance testing methods after cleaning; the extent to which the filter media has recovered relies entirely on experience, making it difficult for users to determine whether it can still be safely reused. This forces them to prematurely discard expensive filter media, resulting in resource waste and high costs. The complexity of filter media contamination further amplifies the cleaning difficulty. Dust deposition inside the filter media is not uniform; heavy metal particles tend to embed deeply, while light organic matter or oily substances easily form clumps on the surface. Different contaminants can also adhere to each other and secondary agglomerate, making it difficult for a single cleaning method to effectively remove all types of deposits simultaneously. If these contaminants are not specifically distinguished and preliminarily separated before cleaning, the subsequent cleaning process can easily result in some contaminants being dispersed but then redeposited in other areas, leading to uneven overall cleanliness and incomplete performance recovery.

[0004] Therefore, the key issue in achieving efficient regeneration and reliable reuse of filter media lies in how to accurately identify the type and distribution of pollutants based on the actual pollution characteristics of the filter media during the professional regeneration and cleaning process, and on this basis, effectively separate and remove dust of different properties, while ensuring that the filter media structure is not damaged throughout the cleaning process and that the final performance is objectively verified. Summary of the Invention

[0005] This invention provides a method for controlling the pollutant separation and regeneration performance of cleanroom filter media, mainly comprising:

[0006] The process involves acquiring pollutant distribution data for the filter media, extracting feature vectors based on this data, classifying pollutant clusters, calculating density and particle size distribution from the feature vectors, and determining pollutant type classification. If the proportion of heavy particles in the pollutant type classification exceeds a preset threshold, a vibration separation module is activated to initially loosen deeply embedded particles, acquiring particle detachment trajectory data after loosening to obtain a preliminary separated particle set. For the preliminary separated particle set, bonding strength data from sensor feedback is fused, and a decision tree algorithm is used to construct a separation path. Light organic blocks are selected from the bonding strength data to determine a priority sequence for block separation. This priority sequence guides a laser-assisted stripping device to perform layer-by-layer scanning and removal of the filter media surface, acquiring real-time resistance change data during the removal process to determine the filter media cleanliness index. If the filter media cleanliness index does not reach a preset threshold, a neutral solvent is injected through an ultrasonic wetting module to perform secondary dissolution of residual pollutants, obtaining updated filter media state data after dissolution. Based on the updated filter media state data, a wind resistance simulation model is run, and ventilation parameters are extracted from the simulation results to determine the filter media regeneration performance compliance level. Furthermore, the acquisition of pollutant distribution data of the filter material includes: acquiring surface and internal multi-layer cross-sectional data of the filter material using an optical scanner to obtain an original scanned image; using image processing technology to filter and remove noise points and enhance contrast in the original scanned image to obtain a processed image; extracting pollutant features from the processed image and determining the pollutant location coordinates through pixel grayscale differences; generating a pollutant distribution map based on the pollutant location coordinates and mapping pollutant concentration values ​​through coordinate point density. Furthermore, the extraction of feature vectors and segmentation of pollutant clusters based on the pollutant distribution data, calculating density and particle size distribution from the feature vectors, and determining pollutant type classification includes: acquiring the grayscale value of each pixel in the filter material pollutant distribution map and calculating texture statistics within a local window; constructing a feature vector characterizing the pollutant region based on the grayscale value and texture statistics; using an image segmentation algorithm, setting a region segmentation threshold based on the grayscale contrast information in the feature vectors, segmenting different pollutant cluster regions in the distribution map, and performing segmentation on each independent connected cluster region. The clusters are labeled with numbers; from the images of the labeled cluster regions, the pixel set contained in each numbered region is extracted, and the equivalent circle diameter and region area of ​​the pixel set are calculated as morphological parameters. At the same time, the average value of the corresponding feature in the feature vector of the pixel set is calculated as a statistical parameter. Based on the morphological parameters and statistical parameters of each cluster, if its equivalent circle diameter and average gray value fall within the preset particulate matter criterion range, the cluster is determined to be a particulate pollutant type. If its region area is large and its average texture statistics are high, the cluster is determined to be a fiber pollutant type, and a list of clusters with pollutant type labels is obtained.Furthermore, if the proportion of heavy particles in the pollutant type classification is higher than a preset threshold, the vibration separation module is activated to initially loosen the deeply embedded particles, obtain the detachment trajectory data of the loosened particles, and obtain a preliminary separated particle set. This includes: for the pollutant type classification, obtaining the proportion data of heavy particles; if the proportion is higher than a preset threshold, activating the vibration separation module to perform initial loosening treatment on the deeply embedded particles using low-frequency pulse waves to obtain a loosened particle group; from the loosened particle group, using a trajectory capture device to record the particle detachment trajectory, obtaining a trajectory data set, and determining the motion parameters of the detached particles; based on the trajectory data set, classifying the types of detached particles, judging the distribution characteristics of the initially separated particles by comparing the motion parameters with preset type standards, and obtaining a separated particle subset; for the separated particle subset, integrating the classification proportion analysis and loosening data processing, constructing a preliminary separated particle set by calculating the proportion of particle types within the subset and fusing the trajectory data, and obtaining a particle set with type labels. Furthermore, the separation path is constructed using a decision tree algorithm based on the bonding strength data fed back by the sensor for the initially separated particle set. Lightweight organic matter blocks are then selected from the bonding strength data to determine the separation priority sequence of the bonded blocks. This includes: acquiring the initially separated particle set and its associated type label set; synchronously receiving the bonding strength data stream from the contact force sensor; aligning and splicing the type label, mass attribute, historical detachment trajectory parameters of each particle with the real-time acquired bonding strength value to construct a particle-strength feature fusion layer; from the particle-strength feature fusion layer, selecting all data records marked as lightweight organic matter based on the type label; extracting the bonding strength value, equivalent particle diameter, and trajectory terminal velocity of these records to form a lightweight organic matter feature subset, which serves as the input feature matrix of the decision tree model; and processing the input feature matrix using a preset CART decision tree algorithm, with minimizing classification confusion as the criterion. A split is generated to create a separation decision tree. Each leaf node of the separation decision tree corresponds to a separation action suggestion. The path from the root node to the leaf node constitutes a separation path graph. The separation path graph is traversed to identify all leaf nodes with bonding strength values ​​lower than a preset strength threshold. The lightweight organic blocks associated with these nodes are aggregated into a target set of blocks to be separated. For each block in the target set of blocks to be separated, its bonding strength value is obtained. The energy consumption required for separation is estimated based on the mass of the block and a preset separation force function. At the same time, the reciprocal of the average distance between the block and other blocks in the set in the original spatial coordinates of the initial separation particle set is calculated as the spatial aggregation degree. A weighted formula is used to fuse the bonding strength value, the estimated separation energy consumption, and the spatial aggregation degree to obtain the comprehensive priority parameter of the block. All blocks in the target set of blocks to be separated are sorted in descending order according to their comprehensive priority parameters to obtain the separation priority sequence of lightweight organic bonded blocks.Furthermore, the step of guiding the laser-assisted stripping device to perform layer-by-layer scanning and removal of the filter material surface through the separation priority sequence, acquiring real-time resistance change data during the removal process, and determining the filter material cleanliness index includes: acquiring the separation priority sequence, which guides the laser-assisted stripping device to generate a stripping path plan for the filter material surface; the device initiating layer-by-layer scanning and removal of the target area according to the planned path and preset initial laser energy control parameters; during the scanning process, a surface resistance sensing data stream is synchronously acquired through a high-precision force sensor integrated on the laser head; the acquired surface resistance sensing data stream is subjected to real-time resistance signal filtering processing to remove high-frequency noise and equipment vibration interference, resulting in a smoothed real-time resistance change curve; and the curve is compared with the resistance obtained by pre-calibration on a clean filter material sample. The baseline calibration value is compared; if the value of the real-time resistance change curve in a certain scanning area is continuously higher than the resistance baseline calibration value and exceeds a preset threshold, it is determined that there is residual contamination in that area, triggering an abnormal resistance warning. At the same time, the contamination load in that area is estimated based on the magnitude and duration of the resistance exceeding the baseline, and the cleaning progress of the completed scanning areas is calculated cumulatively. For the area that triggers the abnormal resistance warning, the laser energy control parameters and real-time layer depth monitoring data during the scanning of that area are retrieved. Combined with the abnormal resistance peak value, a preset stress-strain relationship model is used to assess the potential filter material damage risk level. The cleaning progress calculation data and the filter material damage risk level data of all scanning areas are integrated, and a comprehensive cleanliness index characterizing the overall state of the filter material is calculated according to a preset weighted fusion formula. Furthermore, if the cleanliness index of the filter material does not reach a preset threshold, a neutral solvent is injected through the ultrasonic wetting module to perform a secondary dissolution of the residual contaminants, obtaining updated filter material status data after dissolution. This includes: acquiring the comprehensive cleanliness index that does not reach the threshold; triggering the start command of the ultrasonic wetting module; the module injecting neutral solvent into the target filter material area according to a preset solvent type and initial injection flow rate; activating the transducer of the ultrasonic wetting module using preset ultrasonic frequency and power parameters; the transducer generating a cavitation effect in the solvent environment; the cavitation effect acting on the residual contaminants within the filter material pores; simultaneously collecting turbidity and pH change data of the solvent environment through an optical turbidity sensor and a pH sensor integrated in the wetting chamber; the change data reflecting the dissolution process of the contaminants; determining that the dissolution reaction is complete based on the time point when the turbidity change data tends to stabilize and the time when the pH change data returns to the neutral range; and obtaining a sample image of the solvent environment at this time as the updated filter material status data after dissolution.Furthermore, the step of running a wind resistance simulation model based on the filter media state update data, extracting ventilation parameters from the simulation results, and determining the filter media regeneration performance compliance level includes: acquiring a microstructure image of the filter media from the dissolved filter media state update data; using an image segmentation algorithm to identify the pore contours and fiber distribution in the image; constructing a three-dimensional digital geometric model of the filter media based on the identification results; performing unstructured mesh generation on the three-dimensional digital geometric model; obtaining a mesh cell size and quality report; adjusting the mesh density according to the report until the simulation accuracy requirements are met; simultaneously setting the simulation inlet to a constant flow velocity condition, the outlet to a pressure outlet condition, and the filter media wall surface to a no-slip boundary condition; importing the mesh model with the set boundary conditions into a computational fluid dynamics solver, selecting a laminar or turbulent flow model, setting air as the working medium, and starting the solver. Numerical iterative calculations are performed until the flow field residual curve converges to below a preset tolerance, resulting in stable flow field distribution data. From the stable flow field distribution data, the static pressure values ​​of the upstream and downstream sections flowing through the digital model of the filter material are extracted. The pressure difference parameter is calculated by subtracting the upstream static pressure from the downstream static pressure. Simultaneously, the air velocity of multiple sections inside the model is extracted, and the velocity distribution parameter is statistically obtained. The pressure difference parameter and the velocity distribution parameter together constitute a ventilation parameter set. The pressure difference parameter in the ventilation parameter set is compared with a preset clean filter material reference pressure difference. The pressure difference recovery rate index is calculated by dividing the pressure difference parameter by the reference pressure difference. At the same time, the cosine similarity between the velocity distribution parameter and the reference velocity distribution is calculated. If the pressure difference recovery rate is greater than or equal to a preset threshold and the distribution similarity is greater than or equal to another preset threshold, the filter material regeneration performance is judged to meet the standard, and a regeneration qualification assessment result is generated. Furthermore, after determining the pollutant type classification based on the morphological and statistical parameters of each cluster, the method further includes: for the cluster list with type labels, calculating the normalized distance between the morphological parameters of each cluster and the corresponding type criterion threshold, wherein the normalized distance is obtained by subtracting the threshold from the parameter value and then dividing by the threshold standard deviation; determining the classification confidence based on the normalized distance value, and marking the classification results with a distance value greater than a preset threshold as low confidence, thereby obtaining a final classification result table containing pollutant type and classification confidence identifier.

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

[0008] To address the challenges of complex contaminant distribution, difficult separation, and inaccurate regeneration performance assessment in business scenarios involving filter media, this invention utilizes optical scanning and image processing technologies to acquire contaminant distribution maps and extract feature vectors. Combined with image segmentation and decision tree algorithms, it accurately classifies contaminant types and prioritizes their separation. For cases where the proportion of heavy particles exceeds the standard, low-frequency pulse waves are activated to loosen the particles, and a laser-assisted peeling device removes contaminants layer by layer. Simultaneously, resistance changes are monitored in real time to assess cleanliness. If standards are not met, ultrasonic impregnation solvents are introduced for secondary dissolution. Finally, a wind resistance simulation model is used to evaluate regeneration performance. This invention, through multi-module collaborative operation, achieves end-to-end optimization from accurate contaminant identification to efficient separation and performance assessment, significantly improving filter media cleaning efficiency and regeneration effectiveness. Attached Figure Description

[0009] Figure 1 This is a flowchart of a method for controlling the pollutant separation and regeneration performance of cleanroom filter media according to the present invention;

[0010] Figure 2 This is a schematic diagram of the framework of S102 in the cleanroom filter material pollutant separation and regeneration performance control method of the present invention;

[0011] Figure 3 This is a schematic diagram of S103 in a method for controlling the separation and regeneration performance of cleanroom filter media according to the present invention. Detailed Implementation

[0012] 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.

[0013] like Figures 1-3 This embodiment of a method for controlling the pollutant separation and regeneration performance of cleanroom filter media may specifically include:

[0014] S101. Collect data on the surface and internal multi-layer cross-sections of the filter material using an optical scanner, and use image processing technology to denoise and enhance the collected data to obtain a pollutant distribution map of the filter material.

[0015] An optical scanner is used to acquire surface and internal multi-layer cross-sectional data of the filter material, resulting in a raw scan image. Image processing techniques are then used to filter and remove noise points and enhance contrast in this raw scan image, yielding a processed image. Pollutant features are extracted from the processed image, and their location coordinates are determined by pixel grayscale differences. A pollutant distribution map is generated based on these coordinates, and pollutant concentration values ​​are mapped using the density of coordinate points.

[0016] In one implementation, data on the surface and internal multi-layer cross-sections of filter media are acquired using an optical scanner. First, a suitable scanning device must be selected, such as an optical coherence tomography (OCT) scanner. This device utilizes a low-coherence light source to generate interference signals, enabling high-resolution, non-destructive imaging. Filter media typically refers to air filtration materials, such as the multi-layered fiber structure used in industrial dust removal or automotive air filters. During acquisition, the filter media is placed on the scanner's sample stage, and the beam focus is adjusted to penetrate layer by layer from the surface. The working principle of an OCT scanner is based on a Michelson interferometer. The light beam emitted from the light source is divided into a reference arm and a sample arm. The light from the sample arm interferes with the light reflected back from each layer of the filter media. The cross-sectional image is constructed by detecting the intensity of the interference signal. This method can acquire multi-layer data of the filter media without physically cutting it, thus maintaining the integrity of the material.

[0017] Specifically, the scanning resolution is set to the micrometer level, for example, 10 micrometers lateral resolution and 5 micrometers axial resolution, to capture the fine distribution of pollutant particles. The acquired data includes surface images and multiple internal cross-sectional layers, for example, 10 cross-sections, each spaced 50 micrometers apart. This results in a series of two-dimensional or three-dimensional image datasets for subsequent processing. Furthermore, the scanning parameters can be adjusted to suit different filter media types and operational scenarios.

[0018] In one possible implementation, for thicker filter media, such as industrial filters, the scanning depth is increased to several millimeters, and a longer wavelength light source, such as near-infrared light, is used to improve penetration.

[0019] It's important to note that the data acquisition process using optical scanners emphasizes real-time performance. For example, integrating scanning modules into filter media production lines enables continuous monitoring of contaminant accumulation. This acquisition method begins with filter media positioning, automatically scanning the surface, and then progressively penetrating deeper layers. Data from each layer is converted into digital signals by photodetectors and stored as image files such as TIFF, ensuring data integrity. This detailed acquisition process provides high-precision foundational data for contaminant analysis, supporting practical applications in filter media quality control. Denoising and enhancing the acquired data using image processing techniques are crucial subsequent steps.

[0020] Specifically, the first step is denoising. Median filtering is used to remove random noise from the image, such as salt and pepper noise, which may originate from sensor interference from the scanner. The principle of median filtering is to sort the values ​​of neighboring pixels around a given pixel and replace the original pixel value with the median, thus smoothing the image without losing edge details. In the context of filter media contaminant distribution, this denoising helps to highlight the difference between contaminant particles and filter fibers. Next, enhancement processing is performed, such as using histogram equalization to improve image contrast. This method redistributes pixel grayscale values, making the image brightness distribution more uniform and facilitating the identification of low-contrast contaminant areas. The entire processing is implemented using image processing software, such as open-source tools, with the input being the acquired raw image dataset and the output being the processed, clear image.

[0021] In one embodiment, wavelet transform denoising can be applied to the acquired multi-layer cross-sectional data. This method decomposes the image into different frequency sub-bands, suppressing high-frequency noise sub-bands while retaining low-frequency signals, thus more effectively handling the complex textures within the filter material. The practical application of wavelet transform is in filter material contaminant detection, where it can distinguish between fine particles and fiber structures, avoiding misjudgments. Enhancement processing can also be combined with edge detection algorithms, such as the Canny operator, to detect contaminant edges. This works by calculating image gradients and applying dual thresholds to extract strong edges. This combined processing ensures improved data quality and supports accurate contaminant distribution analysis. In practical applications, such as in filter material testing laboratories, the processed images are superimposed into a three-dimensional model to observe the permeation of contaminants between multiple layers, thereby evaluating the filtration efficiency of the filter material.

[0022] For example, in another implementation, denoising and enhancement can be optimized for specific contaminant types, such as dust particles, using adaptive filters to adjust filtering parameters based on local image statistical characteristics to achieve more accurate noise removal. This flexibility demonstrates the versatility of the technical solution in the field of filter media testing, adapting to the data processing needs of different production environments. The filter media contaminant distribution map is generated from the aforementioned processed data.

[0023] Specifically, the denoised and enhanced image is used for contaminant identification, for example, by using threshold segmentation methods to mark areas with pixel values ​​higher than a certain threshold as contaminants. This threshold is determined based on the image's grayscale histogram. The distribution map is presented as a heatmap, with color depth representing contaminant concentration; for example, red areas represent high-concentration areas. This chart helps visualize the distribution patterns of contaminants on and inside the filter media, supporting business decisions such as when to replace the filter media.

[0024] In one embodiment, the contaminant distribution map can be further quantified, for example, by calculating the percentage of contaminant coverage area for each layer and deriving a numerical report based on the size of the marked areas in the integral image. This quantification provides data support in filter media maintenance operations, enabling predictive maintenance. Furthermore, by combining various scenarios, such as in filter media aging tests, the data collection and processing process can be repeated to generate time-series distribution maps and observe contaminant accumulation trends. This approach ensures the comprehensiveness of the technical solution without exceeding the scope of filter media contaminant analysis.

[0025] Understandably, the entire process, from data collection to distribution map generation, forms a closed loop, which can effectively improve the accuracy of filter material performance evaluation.

[0026] For example, in industrial filter media applications, this distribution map is used for quality inspection to identify defective areas, thereby improving the manufacturing process.

[0027] S102. Extract feature vectors based on the pollutant distribution map of the filter media, use image segmentation algorithms to divide pollutant clusters in different regions, calculate density and particle size distribution from the feature vectors, and determine the pollutant type classification.

[0028] From the pollutant distribution map of the filter material, the grayscale value of each pixel is obtained, and the texture statistics within a local window are calculated. The texture statistics are obtained by calculating the standard deviation and entropy of the grayscale values ​​within the window. Based on the grayscale values ​​and texture statistics, a feature vector representing the pollutant region is constructed. Using an image segmentation algorithm, based on the grayscale contrast information in the feature vector, a region segmentation threshold is set to divide the different pollutant cluster regions in the distribution map, and each independent and connected cluster region is marked and numbered. From the image of the marked and numbered cluster regions, the pixel set contained in each numbered region is extracted, and the equivalent circle diameter and region area of ​​the pixel set are calculated as morphological parameters. The equivalent circle diameter is obtained by multiplying the region area by four and then dividing by pi and taking the square root. At the same time, the average value of the corresponding feature in the feature vector of the pixel set is calculated as a statistical parameter. Based on the morphological and statistical parameters of each cluster, if its equivalent circle diameter and average gray value fall within the preset particulate matter criterion range, the cluster is determined to be a particulate pollutant type. If its area is large and its average texture statistics are high, the cluster is determined to be a fibrous pollutant type, thus obtaining a list of clusters labeled with pollutant types. For the cluster list labeled with types, the normalized distance between the morphological parameters of each cluster and the corresponding type criterion threshold is calculated. The normalized distance is obtained by subtracting the threshold from the parameter value and then dividing by the threshold standard deviation. The classification confidence is determined based on the normalized distance value. Classification results with distance values ​​greater than the preset threshold are marked as low confidence, resulting in a final classification result table containing pollutant type and classification confidence identifiers.

[0029] In one implementation, feature vectors are extracted based on the pollutant distribution map of the filter media. First, key attributes need to be identified from the processed image.

[0030] Specifically, the feature vector includes the grayscale value, texture features, and shape parameters of the contaminant region. For example, the concentration level is quantified by calculating the average grayscale value of the pixels in the region. This extraction process is widely used in filter media quality control, providing a data foundation for subsequent analysis. Filter media contaminant distribution maps are typically presented in two-dimensional or three-dimensional form. During extraction, the image is converted into vector form, with each dimension corresponding to a feature. For example, shape invariance can be calculated using moment invariants to adapt to different filter media types, such as fiber filters. In this way, the feature vector ensures that it captures the core information of contaminant distribution, supporting the practical needs of filter media performance evaluation. Furthermore, image segmentation algorithms are used to divide contaminant clusters in different regions, based on the extracted feature vectors. Here, image segmentation algorithms refer to methods such as K-means clustering, which iteratively calculates the distance between pixels to group similar regions into clusters.

[0031] It's important to note that K-means clustering treats image pixels as data points, initially selecting K center points, then assigning points to the nearest center based on Euclidean distance, and updating the centers until convergence. In the operational process of filter media contaminant analysis, such as for air filter media, this segmentation helps separate surface dust clusters from internal deposits, achieving independent quantification of contaminant regions. One possible implementation sets K to 3 to 5, adjusted according to the filter media thickness, to divide into low, medium, and high-density clusters, thereby identifying potential clogging areas during the maintenance of industrial dust collector filters. This algorithm emphasizes parameter adaptability; for example, a preset iteration count of 50 ensures segmentation accuracy without introducing excessive computational burden.

[0032] For example, calculating density and particle size distribution from eigenvectors involves statistical analysis steps.

[0033] Specifically, density calculation refers to the ratio of the number of contaminant pixels within each cluster region; for example, dividing the total number of pixels in the cluster by the area of ​​the region yields the density value. Particle size distribution is achieved by measuring the length of the particle boundary and fitting a diameter distribution curve, based on morphological operations such as expansion and corrosion to refine the particle edges. In the business scenario of filter media testing, this calculation process begins with the segmented clusters. First, the boundary coordinates of each cluster are extracted, and then the minimum bounding circle method is applied to estimate the average particle size and standard deviation.

[0034] For example, in the quality inspection of filter media production, this calculation is performed on the collected distribution map to quantify the distribution ratio of pollutant particle sizes from 5 micrometers to 50 micrometers, supporting business decisions to assess the filtration efficiency of the filter media.

[0035] Understandably, this calculation avoids physical measurements, ensures non-destructive analysis, and allows for comparison of density changes between different batches of filter media.

[0036] Preferably, the pollutant type classification is determined by applying classification rules from the calculated density and particle size distribution.

[0037] Specifically, classification can be based on threshold judgments. For example, if the density is higher than a preset value and the particle size distribution is biased towards small particles, it is classified as dust; conversely, if the distribution shows large particle clusters, it may be fiber residue.

[0038] It should be noted that this classification process forms the basis for decision-making in filter media maintenance, such as triggering the cleaning process after identifying the type of dust.

[0039] In one embodiment, a simple decision tree model is introduced. This model uses density and particle size as input nodes and outputs type labels based on branching conditions such as "density > 0.3," thereby achieving automated classification in industrial filter media applications. This method enhances the versatility of the technical solution, adapting it to filter media scenarios with different pollutant sources while maintaining the objectivity of the analysis. Furthermore, in another embodiment, the entire process can be integrated into a filter media testing system. After extracting feature vectors from the distribution map, these vectors are directly input into a segmentation algorithm to calculate density and particle size distributions, ultimately classifying the pollutant types. This integration enables continuous monitoring in operations, such as real-time feedback of classification results in the filter media production line, supporting timely adjustments to manufacturing parameters. The logical connection of the above steps ensures the completeness of the filter media pollutant analysis.

[0040] S103. If the proportion of heavy particles in the pollutant type classification is higher than the preset threshold, the vibration separation module activates a low-frequency pulse wave to loosen the deeply embedded particles initially, obtains the particle detachment trajectory data after loosening, and obtains the preliminary separated particle set.

[0041] Based on pollutant type classification, the proportion of heavy particles is obtained. If the proportion exceeds a preset threshold, the vibration separation module is activated, and low-frequency pulse waves are used to initially loosen deeply embedded particles, resulting in a loosened particle group. From this loosened particle group, a trajectory capture device records the particle detachment trajectory, obtaining a trajectory data set to determine the motion parameters of the detached particles. Based on the trajectory data set, the detached particle types are classified, and the distribution characteristics of the initially separated particles are determined by comparing the motion parameters with preset type standards, resulting in a separated particle subset. For this separated particle subset, the classification proportion analysis and loosening data processing are integrated. By calculating the proportion of particle types within the subset and fusing the trajectory data, a preliminary separated particle set is constructed, obtaining a particle set with type labels.

[0042] In one implementation, if the proportion of heavy particles in the pollutant type classification is higher than a preset threshold, which is usually set to 0.4 to 0.6 and adjusted according to the specific application scenario of the filter material, such as an air filter, this judgment process is based on previously calculated density and particle size distribution data to ensure the reliability of the classification results.

[0043] Specifically, heavy particles refer to contaminants with a density greater than a preset value, such as metal debris or mineral particles. In filter media contaminant analysis, the proportion is calculated by statistically analyzing the ratio of particle types within clusters. For example, particle identification is performed on segmented areas, the number of heavy particles is counted, and the result is divided by the total number of particles to obtain the proportion value. This method is applied in filter media quality control to help identify potential clogging risks and support the triggering of subsequent separation operations. Furthermore, a vibration separation module activates low-frequency pulse waves to initially loosen deeply embedded particles. This module is a device integrated into the filter media testing system, including a vibration generator and a waveform controller.

[0044] It should be noted that the principle of low-frequency pulse waves is to generate periodic vibration waves with a frequency between 10 and 50 Hz, which are transmitted through the surface of the filter media to the interior, acting on the binding force of the embedded particles and weakening their adhesion to the filter media fibers. In filter media maintenance, such as for industrial dust collector filters, this activation process first detects the filter media thickness, and then sets the pulse intensity according to the thickness. For example, when the thickness is greater than 2 mm, a higher amplitude waveform is used to ensure loosening without damaging the filter media structure.

[0045] Understandably, the implementation of this module emphasizes parameter adaptability, such as preset pulse duration of 30 seconds to 1 minute, to achieve non-destructive treatment of deep particles and support the long-term use of filter media.

[0046] For example, acquiring data on the trajectory of loosened particles involves using an optical sensor or a high-speed camera to capture the movement path of the particles.

[0047] Specifically, after vibration, the particles begin to detach from the filter media surface. The trajectory data, including the particle's displacement vector, velocity, and direction, is obtained through image sequence analysis, such as acquiring 10 images per second to track the coordinate changes of the particles from the starting point to the detachment point. In the business scenario of filter media contaminant separation, this data acquisition helps quantify separation efficiency, such as calculating the average detachment velocity, and supports the evaluation of the vibration module's effectiveness.

[0048] Preferably, the preliminary separated particle set is obtained by collecting the detached particles and performing preliminary classification.

[0049] Specifically, a separated particle set refers to a group of loosened particles selected from trajectory data, such as by capturing particles using a screen or airflow collection device, and then classifying them according to particle size and density to form an initial set.

[0050] In one possible implementation, this process is integrated into the filter media production line, enabling real-time separation, such as separating heavy particles before filter media winding to reduce the risk of subsequent contamination. In another embodiment, the entire separation process can be automated based on classification results; for example, when the proportion exceeds a threshold, the system automatically activates a vibration module, records trajectory data, and outputs a particle set report. This integration provides a data foundation for filter media performance evaluation, ensuring the continuity of the separation process.

[0051] It should be noted that the activation principle of the vibration separation module is based on the mechanical resonance effect. Low-frequency pulse waves propagate through the filter media, generating shear force on deeply embedded particles, gradually reducing the friction between the particles and the filter media. In filter media testing, such as the treatment of fiber filters, this loosening step first calibrates the waveform frequency to match the particle embedding depth. If the depth exceeds 50 micrometers, the frequency is adjusted to a lower frequency to ensure that the particles gradually detach without causing secondary deposition.

[0052] Understandably, the application of this principle emphasizes business adaptability. In different filter media types, such as multi-layer composite filter media, the optimal pulse parameters are determined through pre-experiments to achieve efficient separation.

[0053] For example, the process of acquiring particle detachment trajectory data includes data preprocessing and path fitting steps.

[0054] Specifically, sensors are used to collect raw trajectory points, and then smoothing filters are applied to remove noise, fitting a continuous trajectory curve, for example, estimating the particle motion equation based on the least squares method. In the business environment of filter media maintenance, this data is used to calculate the separation success rate; if the trajectory length is greater than a preset value, it is considered a valid separation, supporting the optimization of vibration parameters.

[0055] In one embodiment, the formation of the initial separated particle set may further include density screening, such as gravity separation of the collected particles to further classify the heavy particle subset. In filter media quality inspection operations, this particle set provides samples for subsequent analysis, ensuring improved filtration capacity of the filter media.

[0056] S104. Based on the bonding strength data fed back by the fusion sensor of the initially separated particles, a decision tree algorithm is used to construct a separation path, and lightweight organic blocks are screened from the bonding strength data to determine the separation priority sequence of the bonded blocks.

[0057] The initial set of separated particles and its associated type label set are acquired. Adhesion strength data streams from contact force sensors are received synchronously. The type label, mass attribute, historical detachment trajectory parameters of each particle are aligned and stitched with the real-time acquired adhesion strength values ​​to construct a particle-strength feature fusion layer. From the particle-strength feature fusion layer, all data records labeled as lightweight organic matter are filtered out based on the type labels. The adhesion strength values, equivalent particle diameters, and terminal velocities of these records are extracted to form a lightweight organic matter feature subset, which serves as the input feature matrix for the decision tree model. The input feature matrix is ​​processed using a preset CART decision tree algorithm. Node splitting is performed based on minimizing classification confusion to generate a separation decision tree. Each leaf node of the separation decision tree corresponds to a separation action suggestion. The path from the root node to the leaf node constitutes a separation path graph. The separation path graph is traversed to identify all leaf nodes with adhesion strength values ​​below a preset strength threshold. The lightweight organic matter blocks associated with these nodes are aggregated into a target set of blocks to be separated. For each block in the target set of blocks to be separated, its bonding strength value is obtained. Based on the block's mass and a preset separation force function, the energy consumption required for separation is estimated by multiplying the mass by a preset force coefficient in the function. Simultaneously, the reciprocal of the average distance between this block and other blocks in the set within the initial separation particle set in the original spatial coordinates is calculated as the spatial aggregation degree. A weighted formula is used to fuse these three indicators by multiplying the bonding strength value by a first weight, the estimated separation energy consumption by a second weight, and the spatial aggregation degree by a third weight, and then summing the results to obtain the block's comprehensive priority parameter. All blocks in the target set of blocks to be separated are arranged in descending order according to their comprehensive priority parameter, with the block with the highest parameter placed at the beginning of the sequence, thus obtaining the separation priority sequence for the lightweight organic bonded blocks.

[0058] In one implementation, the process integrates data in filter media contaminant analysis operations based on the adhesion strength data fed back by the initial separation particle set fusion sensor.

[0059] Specifically, the initial separation particle set is the group of particles obtained from the previous vibration separation step. The adhesion strength data fed back by the sensors is used to measure the bonding force between the particles and the filter media fibers through mechanical sensors, such as using miniature force sensors to detect the tensile strength value of the embedded particles. In filter media quality control operations, this fusion first aligns the particle size and density data of the particle set with the adhesion strength data, and associates the attributes of each particle through a data matching algorithm to ensure that the fusion result reflects the actual adhesion state.

[0060] Understandably, this convergence emphasizes business continuity. For example, in industrial filter media testing systems, the convergence process handles sensor outputs in real time, supporting the accuracy of subsequent decisions. Furthermore, the acquisition of bond strength data is based on a sensor array arranged on the filter media surface. The principle is to measure the particle response by applying minute force pulses, generating a distribution of strength values.

[0061] For example, in the filter media maintenance business scenario, sensor feedback includes average bond strength and coefficient of variation. During fusion, a weighted average method is used to incorporate these data into the particle set attributes.

[0062] It should be noted that the principle of this data fusion is to improve the targeting of particle separation. When applied in the filter material production line business, the sensor sensitivity is first calibrated, and then the intensity data is used as an extended feature of the particle set to achieve a comprehensive description.

[0063] Preferably, a decision tree algorithm is used to construct the separation path, which is used to generate an optimized path in the filter media contaminant separation operation.

[0064] Specifically, the decision tree algorithm is a classification model based on a tree structure. Its construction process includes root node selection and branch splitting. For the fused particle set data, the algorithm splits nodes using the adhesion strength as the main feature.

[0065] In one possible implementation, the algorithm first learns rules from a training dataset derived from historical filter media test data, including particle samples with varying bonding strengths. It selects the optimal split point through information gain calculation; for example, splitting into a high-priority path occurs when the bonding strength exceeds a preset threshold. Furthermore, the principle behind constructing the separation path is to recursively partition the data space until leaf nodes represent specific separation actions. In filter media testing, this path construction supports automated separation processes; for example, the path can define a separation sequence starting from highly bonded particles to ensure efficiency. In filter media performance evaluation, the algorithm training process involves cross-validation to verify the robustness of the paths, enabling adaptation to complex pollutant distributions.

[0066] For example, screening lightweight organic blocks from bond strength data involves a threshold filtering step.

[0067] Specifically, lightweight organic blocks refer to organic pollutants with low density, such as fiber residues or plant debris. The screening process classifies particles with a binding strength below a certain value in the fused data. In filter media maintenance operations, this screening first extracts a histogram of intensity distribution, and then applies a threshold, such as an intensity less than 0.5 Newtons per square millimeter, to mark them as lightweight blocks, supporting subsequent priority determination.

[0068] In one embodiment, the priority sequence for separating the adhesive blocks is determined by the path output by the decision tree.

[0069] Specifically, sequence generation is based on the comparison of selected lightweight organic blocks with other particles, with higher priority blocks being separated first, such as blocks with lower bonding strength, to reduce damage to the filter media.

[0070] It should be noted that the principle behind this judgment lies in risk assessment. The sequence considers particle size and type. When applied in filter media quality inspection, the generated sequence serves as an operational guide to ensure an orderly separation process. Furthermore, the construction details of the decision tree algorithm include feature selection and tree depth control. In filter media contaminant analysis, feature selection prioritizes bonding strength and particle density, while the Gini index is used to assess splitting quality.

[0071] For example, in industrial dust removal filter media scenarios, the algorithm sets a maximum depth of 5 layers when constructing the path to avoid overfitting. The path branches from the root node; for instance, the first layer groups materials into high and low intensity groups based on strength thresholds, and the second layer incorporates density-based lightweight blocks. This construction process improves separation efficiency in business operations because the path provides a clear priority sequence, supporting long-term maintenance of the filter media.

[0072] Understandably, the process of fusing bond strength data may also include noise filtering steps, such as using median filtering to process sensor data, ensuring that a subset of data is selected for generating high-quality preview images. Furthermore, in filter media testing systems, this filtered, fused data is used as input to decision trees, improving the reliability of path construction. In filter media production line operations, noise-filtered data supports real-time determination of separation priorities.

[0073] In one embodiment, the entire process is integrated into a filter media performance evaluation system, which first fuses the data, then constructs a decision tree path, from which lightweight blocks are screened and sequences are generated.

[0074] For example, when processing composite filter media, path adjustment parameters are used to match different layer structures to achieve versatility.

[0075] For example, after screening lightweight organic blocks, the priority sequence is determined by considering business factors such as the number and distribution of blocks. In filter media testing, the sequence output is in list form, which is convenient for operators to execute.

[0076] It should be noted that this technical solution provides a data-driven separation strategy in filter media maintenance, ensuring the continuity and efficiency of the separation process through decision tree paths.

[0077] S105. Guided by the separation priority sequence, the laser-assisted stripping device performs layer-by-layer scanning and removal on the surface of the filter material, obtains real-time resistance change data during the removal process, and determines the cleanliness index of the filter material.

[0078] The separation priority sequence is obtained, which guides the laser-assisted stripping device to generate a stripping path plan for the filter material surface. Based on the planned path and preset initial laser energy control parameters, the device initiates a layer-by-layer scanning removal of the target area. During the scanning process, a high-precision force sensor integrated on the laser head simultaneously acquires surface resistance sensing data streams. The acquired surface resistance sensing data streams undergo real-time resistance signal filtering to remove high-frequency noise and equipment vibration interference, resulting in a smoothed real-time resistance change curve. This curve is compared with a pre-calibrated resistance baseline value obtained on a clean filter material sample. If the value of the real-time resistance change curve in a certain scanning area is consistently higher than the resistance baseline calibration value and exceeds a preset threshold, it is determined that there is residual contamination in that area, triggering an abnormal resistance warning. Simultaneously, based on the magnitude and duration of the resistance exceeding the baseline, the contamination load in that area is estimated by multiplying the magnitude by a time coefficient, and the cleaning progress of the completed scanning areas is calculated cumulatively. For the area triggering the abnormal resistance warning, the laser energy control parameters and real-time layer depth monitoring data during the scanning of that area are retrieved. Combined with the abnormal resistance peak value, a preset stress-strain relationship model is used. This model is based on the ratio of stress to strain values ​​in materials mechanics. The inputs are the laser energy control parameters, real-time layer depth monitoring data, and abnormal resistance peak value; the output is a risk level value, thereby assessing the potential filter media damage risk level. The cleaning progress calculation data and filter media damage risk level data from all scanned areas are integrated. According to a preset weighted fusion formula, the cleaning progress calculation data is multiplied by a first weight, and the filter media damage risk level data is multiplied by a second weight, then summed to obtain a comprehensive cleanliness index characterizing the overall state of the filter media.

[0079] In one implementation, precise control is achieved in filter media contaminant separation operations by using a separation priority sequence to guide the laser-assisted stripping device to perform layer-by-layer scanning and removal of the filter media surface.

[0080] Specifically, the separation priority sequence is a sequence generated from a previous decision tree algorithm, containing the separation order of the glued blocks, which guides the operation of the laser device. The laser-assisted stripping device includes a laser emitter and a scanning mechanism that adjusts the laser intensity and scanning path by receiving sequence data.

[0081] For example, in filter media quality control, the device first applies laser pulses to high-priority agglomerated areas to achieve layer-by-layer peeling. The scan starts from the filter media surface and proceeds downwards layer by layer, with the thickness of each layer based on particle depth information defined in the sequence. This guidance ensures an orderly separation process, avoiding excessive damage to the filter media fibers. Furthermore, the principle of layer-by-layer scanning removal lies in the selective destruction of adhesive forces by laser energy without affecting the substrate structure. In filter media maintenance scenarios, the device is equipped with optical sensors to monitor the scanning progress and supports real-time adjustments.

[0082] For example, the step of acquiring real-time resistance change data during the removal process is implemented by integrating sensors in filter media testing operations.

[0083] Specifically, the resistance change data refers to the mechanical feedback generated when particles detach from the filter material surface during laser ablation. A force sensor array is arranged below the device to measure the resistance value during the removal of each layer.

[0084] It should be noted that this data acquisition is based on the principle of continuous sampling, with a sampling frequency set to 10 times per second to capture the resistance peak and attenuation curve.

[0085] In one possible implementation, data is recorded in real time via a wireless transmission module to form a time-series dataset. For example, in an industrial filter material testing system, the change in resistance reflects the dynamic distribution of particle bonding strength, supporting the accuracy of subsequent analysis.

[0086] Understandably, this real-time acquisition process emphasizes business continuity, ensuring data integrity for cleanliness assessment. Furthermore, the process of determining filter media cleanliness indicators is calculated based on resistance change data in the filter media performance evaluation process.

[0087] Specifically, the cleanliness index is obtained by analyzing the statistical characteristics of the resistance change curve, such as calculating the average resistance value and the coefficient of variation. If the average value is lower than a preset threshold, such as 0.2 Newtons, the index is marked as high cleanliness.

[0088] In one embodiment, the determination involves an integral method to process the area of ​​a curve and generate a quantified score, which is applied in the filter material production line. The output of the index is a numerical report, which facilitates quality inspection.

[0089] Preferably, the determination of indicators also incorporates historical data comparison to achieve trend analysis. In filter media maintenance, the entire process is integrated into an automated system. First, a priority sequence is applied to initiate laser scanning, then resistance data is collected, and finally, cleanliness indicators are calculated.

[0090] For example, when processing multi-layer filter media, the scanning depth is adjusted sequentially, data acquisition covers all layers, and indicators reflect the overall cleanliness status. This integration supports improved operational efficiency by optimizing separation strategies through data-driven approaches.

[0091] Specifically, the layer-by-layer scanning of the laser-assisted stripping device demonstrates versatility in filter media contaminant analysis. For example, the device can sequentially guide the switching of laser wavelengths to achieve adaptive removal for different types of contaminants.

[0092] It should be noted that the scan path generates a grid pattern based on the priority in the sequence to ensure coverage of all target areas.

[0093] In one embodiment, the processing of real-time resistance change data includes a filtering step, using a smoothing algorithm to remove noise and improve data quality. In filter media quality control operations, this processed data is directly input into the cleanliness calculation module. Furthermore, the principle behind determining the cleanliness index lies in quantifying the level of remaining contaminants, for example, through resistance decay rate calculation, providing feedback for maintenance decisions.

[0094] For example, in the filter material testing business scenario, sequence-guided laser removal combined with data acquisition forms a closed-loop control, and the output of indicators supports the system's adaptive adjustment.

[0095] Understandably, when this technical solution is applied in filter material production line operations, it emphasizes real-time performance and precision, using indicators to monitor production quality.

[0096] In one possible implementation, the integration of the device with the sequence also includes a feedback loop that pauses the scan and re-evaluates priorities if the resistance data is abnormal, ensuring process safety.

[0097] S106. If the cleanliness index of the filter material does not reach the preset threshold, a neutral solvent is injected through the ultrasonic impregnation module to dissolve the residual pollutants a second time, and the updated status data of the filter material after dissolution is obtained.

[0098] The system acquires the comprehensive cleanliness index that fails to reach the threshold, triggering the start command of the ultrasonic wetting module. The module injects neutral solvent into the target filter media area according to a preset solvent type and initial injection flow rate. Using preset ultrasonic frequency and power parameters, the transducer of the ultrasonic wetting module is activated. The transducer generates a cavitation effect in the solvent environment, which acts on residual contaminants within the filter media pores. Through an optical turbidity sensor and a pH sensor integrated into the wetting chamber, the system simultaneously collects turbidity and pH change data of the solvent environment, reflecting the dissolution process of the contaminants. Based on the time point when the turbidity change data tends to stabilize and the time when the pH change data returns to the neutral range, the dissolution reaction is determined to be complete, and an image of the solvent environment sample at this moment is acquired as the updated status data of the dissolved filter media.

[0099] In filter media maintenance, when the system determines that the cleanliness index has not reached the preset threshold, it will automatically trigger a secondary processing procedure.

[0100] Specifically, the preset threshold is a value pre-set based on the filter media type and usage standards. For example, for high-efficiency air filters, this threshold might be set to a cleanliness score of 85. If the calculated index is lower than this value, it is determined that the initial laser ablation failed to completely remove deep or stubborn contaminants. Based on the above determination, the control system transmits instructions to the ultrasonic impregnation module. This module is a functional unit integrated into the processing station, mainly including a solvent storage tank, a precision infusion pump, an ultrasonic generator, and an impregnation chamber. In one embodiment, the operating frequency of the ultrasonic generator is set to 40 kHz, which can generate a cavitation effect in the solvent while avoiding structural damage to the filter media substrate. Furthermore, the process of injecting neutral solvent through the ultrasonic impregnation module is precisely controlled.

[0101] Specifically, a neutral solvent refers to a low surface tension liquid with a pH value close to 7 that does not chemically react with the filter media fibers.

[0102] For example, in business scenarios dealing with oily contaminants, specific emulsified neutral cleaning agents can be used. The infusion pump delivers the solvent to the wetting chamber at a constant flow rate, based on the area and degree of contamination of the filter media, and sprays it evenly onto the surface of the filter media.

[0103] It should be noted that the principle of ultrasonic impregnation lies in using high-frequency mechanical vibration to generate tiny bubbles in the solvent and cause them to collapse instantly, producing strong localized impact force and micro-jet. This force can penetrate into the gaps between filter media fibers and the interior of contaminants, effectively weakening the adhesion between contaminants and fibers, and promoting the diffusion and dissolution of the solvent.

[0104] In one possible implementation, the wetting process lasts for a preset period, such as 120 seconds, to ensure that the solvent is fully utilized.

[0105] For example, the process of obtaining updated data on the state of the dissolved filter material is achieved through a variety of sensing methods.

[0106] In one embodiment, an optical transmittance sensor is integrated inside the immersion chamber to monitor changes in light intensity through the filter media in real time. As contaminants dissolve and disperse, the transmittance of the filter media increases, and the change in the slope of the transmittance curve collected by the sensor can be used as status update data. In another embodiment, after secondary processing, the system initiates a simplified resistance measurement step again, using a light-contact probe to scan the surface of the filter media to obtain the distribution of residual adhesion resistance after dissolution. This distribution data is compared with the data before processing to generate a status update report.

[0107] Understandably, this state update data is used for closed-loop feedback.

[0108] For example, in the filter media regeneration process, if updated data indicates that the cleanliness standard has been met, the process ends and the filter media enters the drying stage; if the standard is still not met, the system may adjust the solvent formula or extend the ultrasonic treatment time, or mark the filter media as requiring special treatment. The entire secondary dissolution process is completed under automated control, ensuring the continuity of filter media maintenance services and the consistency of treatment results.

[0109] S107. Run the air resistance simulation model based on the updated filter media status data, extract ventilation parameters from the simulation results, and determine the level of filter media regeneration performance compliance.

[0110] The microstructure image of the filter media is obtained from the updated state data of the dissolved filter media. An image segmentation algorithm is used to identify the pore contours and fiber distribution in the image. Based on the identification results, a three-dimensional digital geometric model of the filter media is constructed, which serves as the physical basis for wind resistance simulation. The three-dimensional digital geometric model is then divided into an unstructured mesh, and a mesh cell size and quality report is obtained. The mesh density is adjusted according to the report until the simulation accuracy requirements are met. Simultaneously, the simulation inlet is set to a constant flow velocity condition, the outlet to a pressure outlet condition, and the filter media wall to a no-slip boundary condition. The mesh model with the boundary conditions set is imported into a computational fluid dynamics solver. A laminar or turbulent flow model is selected. The laminar flow model determines the flow state based on the Reynolds number and sets the viscosity parameters. The turbulent flow model uses the k-epsilon equation to calculate turbulent kinetic energy and dissipation rate. Air is set as the working medium, and the solver is started for numerical iterative calculations until the flow field residual curve converges below the preset tolerance, obtaining stable flow field distribution data. From the stable flow field distribution data, the static pressure values ​​of the upstream and downstream sections flowing through the digital model of the filter media are extracted. The pressure difference parameter is calculated by subtracting the upstream static pressure from the downstream static pressure. Simultaneously, the air velocity at multiple sections inside the model is extracted, and the velocity distribution parameter is statistically obtained. The pressure difference parameter and the velocity distribution parameter together constitute a ventilation parameter set. The pressure difference parameter in the ventilation parameter set is compared with the preset clean filter media reference pressure difference. The pressure difference recovery rate index is calculated by dividing the pressure difference parameter by the reference pressure difference. At the same time, the cosine similarity between the velocity distribution parameter and the reference velocity distribution is calculated. The cosine similarity is obtained by dividing the vector inner product by the modulus product. If the pressure difference recovery rate is greater than or equal to a preset threshold and the distribution similarity is greater than or equal to another preset threshold, the filter media regeneration performance is judged to meet the standard, and a regeneration qualification assessment result is generated.

[0111] In one implementation, the process of running a wind resistance simulation model based on updated data of filter media status first involves data input and model initialization.

[0112] Specifically, the filter media status update data includes the change in transmittance and the distribution of residual adhesion resistance after dissolution, which are derived from the aforementioned secondary dissolution treatment. The wind resistance simulation model is a numerical simulation tool based on fluid dynamics principles, used to predict the resistance performance of filter media in airflow environments.

[0113] It should be noted that this model uses the finite volume method to discretize the microstructure of the filter media, treating it as a porous medium, and adjusts the porosity and surface roughness parameters based on state update data. In this way, the model can simulate the pressure loss and velocity distribution as airflow passes through the filter media, thereby generating simulation results. This process is implemented in an automated system for filter media maintenance.

[0114] For example, the process runs on a workstation integrating processing units, ensuring real-time data import. Furthermore, the step of extracting ventilation parameters from simulation results is designed as automated extraction logic. These ventilation parameters primarily include drag coefficient, ventilation efficiency, and differential pressure, which reflect the regeneration performance of the filter media.

[0115] For example.

[0116] In one possible implementation, the system extracts the drag coefficient from the output matrix of the simulation results. This coefficient is obtained by calculating the pressure difference between the airflow inlet and outlet and dividing it by the square of the flow velocity.

[0117] It should be noted that the extraction process uses a script algorithm to scan the simulated data grid, identify parameter values ​​at key nodes, and generate a parameter report. This extraction helps quantify the ventilation capacity of the filter media.

[0118] For example, for industrial air filter media, the ventilation efficiency parameter can represent the airflow per unit area. In filter media regeneration operations, if the ventilation parameters show a differential pressure value below a preset threshold, it indicates that the filter media has been restored to a good condition.

[0119] Preferably, the operation of determining the level of filter media regeneration performance is based on the comparison of extracted ventilation parameters with standard thresholds.

[0120] In one embodiment, the compliance level is divided into three levels: high, medium, and low, corresponding to ventilation efficiencies of greater than 90%, 70-90%, and less than 70%, respectively.

[0121] Specifically, the system inputs the extracted parameters into the judgment module, which uses a threshold comparison algorithm.

[0122] For example, if the drag coefficient is less than 0.5 and the ventilation efficiency is higher than 85%, it is considered to meet the standard.

[0123] In one embodiment, for high-efficiency particulate air filter media, the evaluation process considers environmental factors, such as incorporating humidity variables into the simulation to assess performance stability under humid conditions. This evaluation ensures the reliability of the filter media in actual use and avoids immediate failure after regeneration.

[0124] Understandably, the entire process follows a closed-loop feedback loop, from data input to judgment and output. This applies to the continuous operation of filter media maintenance.

[0125] For example, when simulation results show that ventilation parameters are not up to standard, the system can trigger additional processing cycles, such as repeating the immersion step. This design improves business continuity. In one implementation, model runtime is controlled within 30 seconds, and simulation is accelerated through parallel computing to ensure efficient evaluation.

[0126] For example, in scenarios involving filter media for chemical pollutants, updated status data may show high residual resistance. The model simulation will correspondingly increase the predicted wind resistance value, thereby guiding the judgment module to output a low compliance level report. This embodiment demonstrates the application of the technical solution in the treatment of persistent pollutants. Furthermore, in another embodiment, ventilation parameter extraction can be combined with historical data trend analysis.

[0127] For example, the performance recovery rate can be calculated by comparing the current simulation results with the initial filter media parameters. If the recovery rate exceeds 80%, the target is considered met. This method is suitable for long-term maintenance operations, providing a quantitative basis.

[0128] It should be noted that the principle of the wind resistance simulation model lies in simulating airflow dynamics. The specific process includes mesh generation, boundary condition setting, and iterative solution. By dividing the filter media into thousands of micro-elements, the model calculates the flow velocity and pressure of each micro-element, and summarizes them to form the overall simulation results. This detailed simulation ensures the accuracy of the judgment and plays a key role in the evaluation of filter media regeneration performance.

[0129] For example, for fiber filter media, multiple parameter weights can be introduced when determining compliance levels. For instance, the drag coefficient could be weighted at 0.6, and the ventilation efficiency at 0.4, to calculate a comprehensive score. If the score is higher than a threshold, the performance meets the standard. This approach enhances the flexibility of the judgment. In one possible implementation, after the system outputs the judgment result, it generates a performance report for business records. This report includes ventilation parameter charts for easy operator review.

[0130] Understandably, this technical solution achieves an objective assessment of filter media regeneration performance through the integration of simulation and judgment modules, ensuring consistent filter media quality during maintenance operations.

[0131] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A method for controlling the pollutant separation and regeneration performance of cleanroom filter media, characterized in that, include: The process involves: acquiring pollutant distribution data of the filter media; extracting feature vectors and classifying pollutant clusters based on the data; calculating density and particle size distribution from the feature vectors to determine pollutant type classification; if the proportion of heavy particles in the pollutant type classification exceeds a preset threshold, activating the vibration separation module to initially loosen deeply embedded particles, acquiring particle detachment trajectory data after loosening, and obtaining a preliminary separated particle set; using the adhesion strength data fed back by the sensor to construct a separation path based on the preliminary separated particle set, filtering light organic blocks from the adhesion strength data, and determining the separation priority sequence of the adhesive blocks; guiding the laser-assisted stripping device to perform layer-by-layer scanning and removal of the filter media surface using the separation priority sequence, acquiring real-time resistance change data during the removal process, and determining the filter media cleanliness index; if the filter media cleanliness index does not reach the preset threshold, injecting neutral solvent through the ultrasonic wetting module to perform secondary dissolution of residual pollutants, obtaining updated filter media state data after dissolution; running a wind resistance simulation model based on the updated filter media state data, extracting ventilation parameters from the simulation results, and determining the filter media regeneration performance compliance level.

2. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, The process of acquiring pollutant distribution data for the filter media includes: acquiring surface and internal multi-layer cross-sectional data of the filter media using an optical scanner to obtain an original scanned image; using image processing techniques to filter and remove noise points and enhance contrast in the original scanned image to obtain a processed image; extracting pollutant features from the processed image and determining the pollutant location coordinates based on pixel grayscale differences; generating a pollutant distribution map based on the pollutant location coordinates and mapping pollutant concentration values ​​through coordinate point density.

3. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, The step of extracting feature vectors and dividing pollutant clusters based on the pollutant distribution data, calculating density and particle size distribution from the feature vectors, and determining pollutant type classification includes: obtaining the grayscale value of each pixel from the pollutant distribution map of the filter material, and calculating the texture statistics within a local window; constructing a feature vector characterizing the pollutant region based on the grayscale value and texture statistics; using an image segmentation algorithm, setting a region segmentation threshold based on the grayscale contrast information in the feature vector, dividing different pollutant cluster regions in the distribution map, and marking and numbering each independent and connected cluster region; extracting the pixel set contained in each numbered region from the image of the marked and numbered cluster regions, calculating the equivalent circle diameter and region area of ​​the pixel set as morphological parameters, and simultaneously calculating the average value of the corresponding feature in the feature vector as a statistical parameter; Based on the morphological and statistical parameters of each cluster, if its equivalent circle diameter and average gray value fall within the preset particulate matter criterion range, the cluster is determined to be a particulate pollutant type; if its area is large and its average texture statistics are high, the cluster is determined to be a fiber pollutant type, thus obtaining a list of clusters labeled with pollutant type.

4. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, If the proportion of heavy particles in the pollutant type classification is higher than a preset threshold, the vibration separation module is activated to initially loosen the deeply embedded particles, obtain the detachment trajectory data of the loosened particles, and obtain a preliminary separated particle set. This includes: For each pollutant type classification, obtaining the proportion of heavy particles; if the proportion is higher than a preset threshold, activating the vibration separation module to perform initial loosening treatment on the deeply embedded particles using low-frequency pulse waves to obtain a loosened particle group; From the loosened particle group, using a trajectory capture device to record the particle detachment trajectory, obtaining a trajectory data set, and determining the motion parameters of the detached particles; Based on the trajectory data set, classifying the types of detached particles, and judging the distribution characteristics of the initially separated particles by comparing the motion parameters with preset type standards to obtain a separated particle subset; For the separated particle subset, integrating the classification proportion analysis and loosening data processing, calculating the proportion of particle types within the subset and fusing the trajectory data to construct a preliminary separated particle set, obtaining a particle set with type labels.

5. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, The process involves fusing the adhesion strength data from the sensor feedback of the initially separated particle set, constructing a separation path using a decision tree algorithm, filtering lightweight organic matter blocks from the adhesion strength data, and determining the separation priority sequence of the adhered blocks. This includes: acquiring the initially separated particle set and its associated type label set; synchronously receiving the adhesion strength data stream from the contact force sensor; aligning and splicing the type label, mass attribute, historical detachment trajectory parameters of each particle with the real-time acquired adhesion strength value to construct a particle-strength feature fusion layer; filtering all data records labeled as lightweight organic matter from the particle-strength feature fusion layer based on the type label; extracting the adhesion strength value, equivalent particle diameter, and trajectory terminal velocity of these records to form a lightweight organic matter feature subset, which serves as the input feature matrix for the decision tree model; and processing the input feature matrix using a preset CART decision tree algorithm, performing node segmentation based on minimizing classification confusion. A separation decision tree is generated, where each leaf node corresponds to a separation action suggestion. The path from the root node to the leaf node constitutes a separation path graph. The separation path graph is traversed to identify all leaf nodes with bonding strength values ​​lower than a preset strength threshold. The lightweight organic blocks associated with these nodes are aggregated into a target set of blocks to be separated. For each block in the target set of blocks to be separated, its bonding strength value is obtained. The energy consumption required for separation is estimated based on the mass of the block and a preset separation force function. At the same time, the reciprocal of the average distance between the block and other blocks in the set in the original spatial coordinates of the initial separation particle set is calculated as the spatial aggregation degree. A weighted formula is used to fuse the bonding strength value, the estimated separation energy consumption, and the spatial aggregation degree to obtain the comprehensive priority parameter of the block. All blocks in the target set of blocks to be separated are sorted in descending order according to their comprehensive priority parameters to obtain the separation priority sequence of lightweight organic bonded blocks.

6. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, The process of using the separation priority sequence to guide the laser-assisted stripping device to perform layer-by-layer scanning and removal of the filter material surface, acquiring real-time resistance change data during the removal process, and determining the filter material cleanliness index includes: acquiring the separation priority sequence, which guides the laser-assisted stripping device to generate a stripping path plan for the filter material surface; the device initiating layer-by-layer scanning and removal of the target area according to the planned path and preset initial laser energy control parameters; during the scanning process, a high-precision force sensor integrated on the laser head synchronously acquires surface resistance sensing data streams; the acquired surface resistance sensing data streams are subjected to real-time resistance signal filtering processing to remove high-frequency noise and equipment vibration interference, resulting in a smoothed real-time resistance change curve; and the curve is compared with a resistance baseline pre-calibrated on a clean filter material sample. The values ​​are compared with the baseline values. If the real-time resistance change curve in a certain scanning area is consistently higher than the baseline resistance value and exceeds a preset threshold, it is determined that there is residual contamination in that area, triggering an abnormal resistance warning. At the same time, the contamination load in that area is estimated based on the magnitude and duration of the resistance exceeding the baseline, and the cleaning progress of the completed scanning areas is calculated cumulatively. For the area that triggers the abnormal resistance warning, the laser energy control parameters and real-time layer depth monitoring data during the scanning of that area are retrieved. Combined with the abnormal resistance peak value, a preset stress-strain relationship model is used to assess the potential filter material damage risk level. The cleaning progress calculation data and the filter material damage risk level data of all scanning areas are integrated, and a comprehensive cleanliness index characterizing the overall state of the filter material is calculated according to a preset weighted fusion formula.

7. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, If the cleanliness index of the filter material does not reach the preset threshold, a neutral solvent is injected through the ultrasonic wetting module to perform a secondary dissolution of the residual contaminants, obtaining updated filter material status data after dissolution. This includes: acquiring the comprehensive cleanliness index that does not reach the threshold; triggering the start command of the ultrasonic wetting module; the module injecting neutral solvent into the target filter material area according to the preset solvent type and initial injection flow rate; activating the transducer of the ultrasonic wetting module using preset ultrasonic frequency and power parameters; the transducer generating a cavitation effect in the solvent environment; the cavitation effect acting on the residual contaminants in the pores of the filter material; simultaneously collecting turbidity change data and pH change data of the solvent environment through an optical turbidity sensor and a pH sensor integrated in the wetting chamber; the change data reflecting the dissolution process of the contaminants; determining that the dissolution reaction is complete based on the time point when the turbidity change data tends to stabilize and the time when the pH change data returns to the neutral range, and obtaining a sample image of the solvent environment at this time as the updated filter material status data after dissolution.

8. The method for controlling the pollutant separation and regeneration performance of cleanroom filter media as described in claim 1, characterized in that, The step of running a wind resistance simulation model based on the filter media state update data, extracting ventilation parameters from the simulation results, and determining the filter media regeneration performance compliance level includes: acquiring a microstructure image of the filter media from the dissolved filter media state update data, using an image segmentation algorithm to identify the pore contours and fiber distribution in the image, and constructing a three-dimensional digital geometric model of the filter media based on the identification results; performing unstructured mesh generation on the three-dimensional digital geometric model, obtaining a mesh cell size and quality report, adjusting the mesh density according to the report until the simulation accuracy requirements are met, and simultaneously setting the simulation inlet to a constant flow velocity condition, the outlet to a pressure outlet condition, and the filter media wall surface to a no-slip boundary condition; Import the mesh model with pre-defined boundary conditions into the computational fluid dynamics solver, select a laminar or turbulent flow model, set air as the working medium, and start the solver to perform numerical iterative calculations until the flow field residual curve converges to below the preset tolerance, obtaining stable flow field distribution data. From the stable flow field distribution data, extract the static pressure values ​​of the upstream and downstream sections flowing through the digital model of the filter material, and calculate the pressure difference parameter by subtracting the upstream static pressure from the downstream static pressure. At the same time, extract the air velocity of multiple sections inside the model and statistically obtain the velocity distribution parameter. The pressure difference parameter and the velocity distribution parameter together constitute the ventilation parameter set. Compare the pressure difference parameter in the ventilation parameter set with the preset clean filter material reference pressure difference, and calculate the pressure difference recovery rate index by dividing the pressure difference parameter by the reference pressure difference. At the same time, calculate the cosine similarity between the velocity distribution parameter and the reference velocity distribution. If the pressure difference recovery rate is greater than or equal to a preset threshold and the distribution similarity is greater than or equal to another preset threshold, the filter material regeneration performance is judged to meet the standard, and a regeneration qualification assessment result is generated.

Images