Image sensor-based non-destructive testing system for uniformity of fluorine material modification layer

Abstract

The present application relates to fluorine material modification layer nondestructive testing technical field, especially fluorine material modification layer uniformity nondestructive testing system based on image sensor, including: the microtopography data and subsurface structure data before and after exposure are combined, the physical barrier degradation rate distribution diagram of each detection area of fluorine material modification layer is generated, thereby the consistent degree of physical uniformity degradation is analyzed;Surface chemical force distribution data and active drug molecule adsorption characteristics before and after exposure are combined, and the adsorption hot spot site is extracted and positioned;The passivation distribution diagram of the anti-adsorption function of each detection area of the fluorine material modification layer is generated, thereby the consistent degree of chemical uniformity passivation is analyzed;And the resistant degradation uniformity of the fluorine material modification layer of the pharmaceutical equipment in the high-purity drug solvent environment is evaluated;Thus, the differentiated process control scheme is executed, the precision of coating risk assessment is improved, and the long-term service reliability of the repaired coating and the safety of drug purity are ensured.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for fluorine-modified layers, and more particularly to a nondestructive testing system for the uniformity of fluorine-modified layers based on an image sensor. Background Technology

[0002] The high-purity environment in pharmaceutical manufacturing is a core production scenario for ensuring drug safety and efficacy. The integrity of the coating on the inner wall of the equipment directly affects the purity level and batch consistency of the drug. Fluorine-modified layers, with their excellent chemical inertness, low surface energy, and anti-adsorption properties, are widely used in pharmaceutical reactors, storage tanks, and piping systems to prevent direct contact between drugs and metal substrates, thus preventing the dissolution of metal ions and drug molecule residues. During long-term service, fluorine-modified layers are not only subject to continuous chemical erosion by high-purity drug solvents, gradually leading to the deactivation of surface functional groups and passivation of anti-adsorption properties; moreover, the osmotic pressure of drug molecules and the swelling effect of solvents can cause the initial micropores, scratches, and other physical defects inside the coating to expand geometrically, forming through cracks or even coating peeling. If the coupled evolution trend of physical barrier degradation and chemical passivation is not detected in time, it may lead to serious quality problems such as excessive drug adsorption, cross-contamination, and even deviations in the content of active ingredients, resulting not only in economic losses from batch scrapping but also threatening patient medication safety.

[0003] However, existing technologies for testing fluorine-modified layers do not spatially correlate two fundamentally different degradation modes: physical barrier degradation (such as defect propagation and new crack formation) and chemical functional passivation (such as increased adsorption hotspots and altered adhesion distribution). For example, when microcracks appear in a certain area of ​​the coating, if the decrease in anti-adsorption capacity caused by the simultaneous oxidation of surface functional groups in that area is not considered, it is difficult to accurately determine whether that area has become a high-risk failure coupling region, thus underestimating the risk of overall degradation tolerance uniformity. Furthermore, existing technologies lack a comprehensive consideration of the cross-correlation analysis between the spatial distribution of physical degradation rates and the spatial distribution of chemical passivation. This makes it difficult for the test results to accurately reflect the degree of degradation consistency in different areas of the coating under high-purity solvent environments, and fails to provide spatial positioning basis for differentiated process control and targeted repair.

[0004] To address these issues, this application presents a non-destructive testing system for the uniformity of fluorine-modified layers based on an image sensor. Summary of the Invention

[0005] The purpose of this invention is to provide a non-destructive testing system for the uniformity of fluorine-modified layers based on image sensors. By simultaneously acquiring data on the microstructure, subsurface structure, and surface chemical forces distribution of the fluorine-modified layer before and after exposure, the system quantifies the spatial distribution of physical defect propagation and new defects, extracts the characteristics of adsorption hotspot site changes, and generates a physical barrier degradation rate distribution map and an anti-adsorption passivation distribution map. Furthermore, spatial superposition and cross-correlation analysis are performed to identify failure coupling regions, construct an environmental tolerance degradation uniformity evaluation dimension, and achieve non-destructive assessment of coating uniformity and differentiated process control.

[0006] This invention is implemented as follows: In a first aspect, the present invention provides a non-destructive testing system for the uniformity of fluorine-modified layers based on image sensors, including a data acquisition module, a physical degradation monitoring module, a chemical passivation monitoring module, a degradation tolerance uniformity assessment module, and a process control module. Among them, the data acquisition module is used to simultaneously collect micromorphological data, subsurface structure data and surface chemical force distribution data of multiple detection areas of the fluorine material modified layer of pharmaceutical equipment before and after exposure to a high-purity drug solvent environment; The physical degradation monitoring module is used to combine microscopic morphology data and subsurface structure data before and after exposure to quantify the geometric expansion of initial physical defects and the spatial distribution of newly formed defects under the action of drug solvent penetration; and based on the geometric expansion and the spatial distribution of newly formed defects, to generate a physical barrier degradation rate distribution map of each detection area of ​​the fluorinated material modified layer, thereby analyzing the degree of uniformity degradation of physical homogeneity. The chemical passivation monitoring module is used to extract and locate adsorption hotspots by combining the surface chemical force distribution data before and after exposure with the adsorption characteristics of active drug molecules. By comparing the changes in hotspot density and distribution range before and after exposure, an anti-adsorption functional passivation distribution map of each detection area of ​​the fluorinated material modified layer is generated, thereby analyzing the degree of chemical uniformity passivation consistency. The degradation tolerance uniformity assessment module is used to perform spatial overlay and cross-correlation analysis on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map based on the degree of physical uniformity degradation and the degree of chemical uniformity passivation to identify failure coupling regions; and to evaluate the degradation tolerance uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment based on the failure coupling regions. The process control module is used to execute differentiated process control schemes based on the evaluation results of the tolerance to degradation uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment.

[0007] As a preferred embodiment of the present invention, the step of generating a physical barrier degradation rate distribution map of each detection area of ​​the fluorine-modified layer based on the geometric expansion amount and the spatial distribution of newly formed defects includes the following specific steps: S21. Optical coherence tomography (OCT) cross-sectional images of the same detection area before and after exposure in the subsurface structure data, and high-resolution optical microscopic morphology images in the microscopic morphology data, are registered with submicron precision. The OCT cross-sectional images of the same detection area before and after exposure are used as the registration objects. The image acquired before exposure is defined as the comparison image, and the image acquired after exposure is defined as the detection image. A feature matching algorithm based on feature point extraction and grid motion statistics is used to achieve feature matching between the detection image and the comparison image. The homography matrix is ​​calculated, and the images are aligned through perspective transformation. Pixel-level difference operations are performed on the aligned images to obtain a difference image, highlighting the areas that changed before and after exposure. S22. For each defect region identified in the differential image, determine whether it existed before exposure; if the corresponding defect region existed before exposure, it is defined as an existing defect, and the existing defect area expansion rate, existing defect depth expansion amount and existing defect volume expansion rate are calculated. S23. If the corresponding defect area does not exist before exposure and only appears after exposure, it is defined as a new defect. Then, the number of pixels of the corresponding defect area in the differential image after exposure is multiplied by the actual area corresponding to a single pixel to obtain the area of ​​the new defect; the depth value calculated based on the axial pixel resolution and the centroid position of the corresponding defect is obtained as the depth of the new defect; and the number of voxels of the corresponding defect area is multiplied by the actual volume corresponding to a single voxel to obtain the volume of the new defect. S24. Divide the number of newly generated defects by the total area of ​​the detection region to obtain the surface density of newly generated defects. S25. For each detection area, handle it according to the type of defect in the corresponding detection area; When there are existing defects in the corresponding detection area, the existing defect area expansion rate, existing defect depth expansion amount and existing defect volume expansion rate are weighted and summed to obtain the existing defect degradation component of the corresponding detection area. When a new defect exists in the corresponding detection area, the area, depth, and volume of the new defect are extracted. The area, depth, and volume of the new defect are normalized. The normalized area, depth, and volume of the new defect are then weighted and summed to obtain the degradation component of the new defect in the corresponding detection area. S26. The existing defect degradation component and the newly generated defect degradation component are weighted and summed to obtain the degradation trend state of the corresponding detection area; S27. Spatial interpolation is performed on the degradation trend status of all detection areas to generate the degradation rate distribution map of the physical barrier.

[0008] In a preferred embodiment of the present invention, the analysis of the degree of physical homogeneity degradation includes the following specific steps: S28. For each detection area, construct a physical degradation feature vector, which includes: the existing defect area expansion rate, existing defect depth expansion amount, existing defect volume expansion rate, and new defect surface density of the corresponding detection area; if there are no existing defects in the corresponding detection area, then the existing defect area expansion rate, existing defect depth expansion amount, and existing defect volume expansion rate are set to 0; if there are no new defects in the corresponding detection area, then the new defect surface density is set to 0. S29. The physical degradation feature vector, along with the local differential image region and the local optical coherence tomography (OCT) cross-sectional image region corresponding to the detection region, are input into a pre-trained convolutional neural network model. The convolutional neural network model extracts defect morphology features and cross-sectional structure features from the local differential image region and the local OCT cross-sectional image region through convolutional layers and pooling layers, respectively. The extracted defect morphology features and cross-sectional structure features are then concatenated and fused with the physical degradation feature vector, thereby outputting the degree of physical uniformity degradation of the corresponding detection region through a fully connected layer. For each detection region of the historical detection samples, the proportion of the cumulative area of ​​all penetrating defects in the corresponding detection region after exposure to the area of ​​the corresponding region is calculated and used as the training label of the convolutional neural network model. S210. Calculate the standard deviation and mean of the physical uniformity degradation degree of all detection areas, and subtract the ratio of the standard deviation and mean of the physical uniformity degradation degree of all detection areas from 1 to obtain the uniformity degradation degree of the physical uniformity of the fluorine material modified layer of the pharmaceutical equipment.

[0009] In a preferred embodiment of the present invention, the step of generating an anti-adsorption passivation distribution map of each detection area of ​​the fluorine-modified layer by comparing the changes in hotspot density and distribution range before and after exposure specifically includes the following steps: S31. Using a chemical force microscope, functionalized probes with the same functional groups as typical active drug molecules are modified on the surface. Force-distance curve measurements are performed at grid points in each detection area before and after exposure to obtain the adhesion force value of each pixel in the surface chemical force distribution data and generate an adhesion force distribution map of the entire surface. S32. Based on the adsorption characteristics of active drug molecules, an adsorption hotspot determination threshold is set. Pixels whose adhesion force value exceeds the adsorption hotspot determination threshold before exposure are determined as existing adsorption hotspot sites before exposure. Adsorption hotspot sites that did not exist before exposure but newly appear after exposure are determined as newly formed adsorption hotspot sites. The adsorption hotspot determination threshold is obtained by adding the average adhesion force before exposure to the product of the sensitivity coefficient and the standard deviation of the adhesion force before exposure. S33. Count the number of adsorption hotspots before and after exposure in each detection area, divide the number of adsorption hotspots in the corresponding detection area by the area of ​​the corresponding detection area to obtain the hotspot density of the corresponding detection area; obtain the hotspot density before exposure and the hotspot density after exposure in the corresponding detection area; divide the difference between the hotspot density after exposure and the hotspot density before exposure in the corresponding detection area by the hotspot density before exposure to obtain the hotspot density increase rate of the corresponding detection area. S34. The absolute value of the difference between the nearest neighbor index of the hotspot distribution before and after exposure in the corresponding detection area is taken as the change in the spatial clustering of hotspots in the corresponding detection area. S35. The hot spot density increase rate and hot spot spatial aggregation change in the corresponding detection area are weighted and summed to obtain the chemical passivation degree of the corresponding detection area. S36. Spatially interpolate the degree of chemical passivation in all detection areas to generate the passivation distribution map of the anti-adsorption function.

[0010] As a preferred embodiment of the present invention, the degree of uniformity of analytical chemical homogeneity passivation includes the following specific steps: S37. For each detection area, construct a chemical passivation feature vector, which includes: the hot spot density increase rate, hot spot spatial aggregation change, average adhesion force value after exposure, and standard deviation of adhesion force value for the corresponding detection area; if there are no adsorption hot spot sites in the corresponding detection area, then the hot spot density increase rate and hot spot spatial aggregation change are assigned to 0. S38. The chemical passivation feature vector and the local region of the adhesion distribution map of the corresponding detection area are input into the pre-trained gradient boosting decision tree model; wherein, the chemical passivation feature vector serves as the structured feature input branch of the gradient boosting decision tree model, and the local region of the adhesion distribution map serves as the image feature input branch of the gradient boosting decision tree model; the gradient boosting decision tree model extracts the adhesion spatial texture features from the local region of the adhesion distribution map through its internal convolutional feature extraction layer, concatenates and fuses the adhesion spatial texture features with the chemical passivation feature vector, performs weighted decision-making on the fused features through ensemble learning, and outputs the degree of chemical uniformity passivation of the corresponding detection area; In this process, for each detection region of the historical detection sample, the proportion of the cumulative area of ​​the adsorption hotspot sites in the corresponding region after exposure to the area of ​​the corresponding region is calculated, and this proportion is used as the training label for the gradient boosting decision tree model. S39. Calculate the standard deviation and mean of the chemical uniformity passivation degree of all detection areas, and subtract the ratio of the standard deviation to the mean of the chemical uniformity passivation degree of all detection areas from 1 to obtain the chemical uniformity passivation consistency of the fluorine material modified layer of the pharmaceutical equipment.

[0011] As a preferred embodiment of the present invention, the step of spatially superimposing and cross-correlation analysis of the physical barrier degradation rate distribution map and the anti-adsorption functional passivation distribution map to identify failure coupling regions; and evaluating the degradation tolerance uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment based on the failure coupling regions; includes the following specific steps: S41. Align the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map on spatial coordinates so that pixels at the same spatial location correspond to the same detection area in the two distribution maps. After alignment, assign each detection area the degradation trend state in the physical barrier degradation rate distribution map and the chemical passivation degree in the anti-adsorption function passivation distribution map. Combine the degradation trend state and chemical passivation degree of each detection area into a two-dimensional vector. Arrange the two-dimensional vectors of all detection areas according to their spatial positions to form a physical-chemical two-dimensional degradation correlation matrix. S42. Perform a two-dimensional normalized cross-correlation analysis on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map, and calculate the cross-correlation coefficient; the two-dimensional normalized cross-correlation analysis is to treat the two distribution maps as two two-dimensional arrays of the same size and calculate the normalized cross-correlation coefficient between the two two-dimensional arrays. S43. Preset physical degradation threshold and chemical passivation threshold, extract the degradation trend state and chemical passivation degree of each detection area from the physical and chemical dual-dimensional degradation correlation matrix, and mark the detection areas with degradation trend state greater than physical degradation threshold and chemical passivation degree greater than chemical passivation threshold as failure units; merge spatially adjacent failure units to form connected failure coupling regions; divide the sum of the areas of all failure coupling regions by the total area of ​​the detection areas to obtain the area ratio of the coupling region of the fluorine material modification layer of the pharmaceutical equipment; S44. Based on steps S41-S43, the coupling strength of the fluorine-modified layer of the pharmaceutical equipment is calculated. S45. Input the physical uniformity degradation consistency, chemical uniformity passivation consistency, and coupling strength of the fluorine material modified layer of the pharmaceutical equipment into a pre-trained random forest regression model. The corresponding random forest regression model is trained with simulated experimental data and outputs the tolerance degradation uniformity of the fluorine material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment.

[0012] As a preferred embodiment of the present invention, the differentiated process control scheme based on the evaluation results of the uniformity of degradation tolerance of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment includes the following specific contents: S51. Obtain the evaluation results of the resistance to degradation uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment; S52. A preset tolerance to degradation uniformity threshold is set. When the tolerance to degradation uniformity assessment result of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment is greater than or equal to the tolerance to degradation uniformity threshold, the current operating parameters are kept unchanged, and the next tolerance to degradation uniformity test is performed according to the regular testing cycle. When the tolerance to degradation uniformity assessment result of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment is less than the tolerance to degradation uniformity threshold, targeted coating repair or local replacement is performed on the failure coupling area, and a second test is performed on the repaired or replaced area. After confirming that the tolerance to degradation uniformity is greater than or equal to the tolerance to degradation uniformity threshold, the equipment is put back into operation.

[0013] Secondly, the present invention provides a non-destructive testing method for the uniformity of fluorine-modified layers based on image sensors, comprising the following specific steps: S1. Simultaneously collect micromorphological data, subsurface structure data, and surface chemical force distribution data for multiple detection areas of the fluorine-modified layer of pharmaceutical equipment before and after exposure to a high-purity drug solvent environment; S2. Combining microscopic morphology data and subsurface structure data before and after exposure, the geometric expansion of initial physical defects and the spatial distribution of newly formed defects under the action of drug solvent penetration are quantified; and based on the geometric expansion and the spatial distribution of newly formed defects, a physical barrier degradation rate distribution map of each detection area of ​​the fluorine material modified layer is generated, thereby analyzing the degree of uniformity degradation of physical homogeneity. S3. Combining the surface chemical interaction force distribution data before and after exposure with the adsorption characteristics of active drug molecules, adsorption hot spots are extracted and located; by comparing the changes in hot spot density and distribution range before and after exposure, an anti-adsorption function passivation distribution map of each detection area of ​​the fluorine material modified layer is generated, thereby analyzing the degree of chemical homogeneity passivation consistency. S4. Based on the degree of physical homogeneity degradation and chemical homogeneity passivation, spatial overlay and cross-correlation analysis are performed on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map to identify failure coupling regions; and based on the failure coupling regions, the tolerance of the fluorine material modified layer of pharmaceutical equipment to degradation homogeneity in a high-purity drug solvent environment is evaluated. S5. Based on the evaluation results of the tolerance to degradation uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment, implement differentiated process control schemes.

[0014] Thirdly, the present invention provides an electronic device, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes a non-destructive testing method for the uniformity of fluorine-modified layers based on an image sensor by calling the computer program stored in the memory.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. This invention achieves correlation analysis between physical barrier degradation and chemical functional passivation by simultaneously collecting physical morphology and chemical force data and performing spatial registration, thereby improving the accuracy of coating risk assessment; 2. This invention constructs a dual-dimensional evaluation index of physical homogeneity and chemical homogeneity by jointly quantifying the expansion of existing defects and the emergence of new defects, as well as dynamically monitoring the spatial aggregation of adsorption hotspots. This ensures the comprehensiveness of tolerance homogeneity assessment and reduces quality accidents caused by missed detection of local degradation. 3. Based on the spatial positioning results of the failure coupling region, this invention performs differentiated targeted repair and thresholding process control, which can avoid the waste of resources caused by global processing, reduce maintenance costs, and at the same time ensure the long-term service reliability of the repaired coating and the safety of drug purity. Attached Figure Description

[0016] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the overall process of the non-destructive testing method for the uniformity of fluorine-modified layers based on image sensors according to the present invention. Figure 2 This is an analytical flowchart of step S2 in the non-destructive testing method for the uniformity of fluorine-modified layers based on image sensors according to the present invention. Figure 3 This is an analytical flowchart of step S3 in the non-destructive testing method for the uniformity of fluorine-modified layers based on image sensors according to the present invention. Figure 4 This is a schematic diagram of the non-destructive testing system for the uniformity of fluorine-modified layers based on an image sensor according to the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0018] Example 1 In the high-purity environment of pharmaceutical manufacturing, fluorinated modified layers serve as a core protective barrier, and their resistance to degradation uniformity directly impacts drug purity and equipment operational safety. Existing detection technologies for coating degradation are mainly divided into two branches: physical detection and chemical detection. In physical detection, conventional methods such as optical microscopy, scanning electron microscopy, and optical coherence tomography (OCT) can acquire information on the microscopic morphology and subsurface structure of the coating surface, used to identify physical defects such as cracks, scratches, and peeling. In chemical detection, commonly used techniques include X-ray photoelectron spectroscopy, infrared spectroscopy, and chemical force microscopy, used to analyze changes in surface functional groups, chemical composition, and adsorption characteristics of drug molecules. However, a common limitation of existing technologies is that physical and chemical detection are independent and do not generate any spatial correlation information. Therefore, they cannot answer the crucial evaluation question of whether the degradation rate is consistent across different regions of the coating in high-purity pharmaceutical settings. Specifically, physical detection only focuses on the geometric expansion of defects, while chemical detection only focuses on the passivation of functional groups; the two cannot synergistically reveal the spatial coupling relationship between physical barrier degradation and chemical functional passivation. Furthermore, existing methods lack a quantitative evaluation dimension for degradation uniformity and cannot identify failure coupling regions—dangerous areas where physical degradation and chemical passivation occur simultaneously and exacerbate each other. This leads to process control schemes often adopting a global treatment strategy, either resulting in over-repair and waste, or the failure to detect local failures in time, leading to drug contamination risks. To address these shortcomings, this embodiment, for the first time, spatially superimposes and cross-correlation analyzes two fundamentally different degradation modes: physical barrier degradation and chemical functional passivation, identifying failure coupling regions. Based on this, an evaluation dimension for environmental tolerance degradation uniformity, which has not been addressed in existing technologies, is constructed.

[0019] Therefore, this embodiment provides a non-destructive testing method for the uniformity of fluorine-modified layers based on image sensors, such as... Figure 1 As shown, the specific steps include the following: S1. Simultaneously collect micromorphological data, subsurface structure data, and surface chemical force distribution data for multiple detection areas of the fluorine-modified layer of pharmaceutical equipment before and after exposure to a high-purity drug solvent environment; The core advantage of this step is that by simultaneously acquiring multi-data of physical morphology and chemical interaction within the same detection area and at the same spatial resolution, it lays the data foundation for subsequent physical-chemical spatial correlation analysis and avoids the spatial misalignment problem caused by the separation of physical and chemical detection in traditional methods.

[0020] Specifically, this step requires collecting three types of data for each detection area before exposure (i.e., the initial state of the coating) and after exposure (the state after being exposed to a high-purity drug solvent for a certain period of time).

[0021] Microscopic morphology data were acquired using high-resolution optical microscopy or scanning electron microscopy to record the surface morphology of the coating from the micrometer to submicrometer scale, including initial defects such as surface roughness, scratches, and pits, as well as newly formed cracks and peeling areas after exposure. Subsurface structure data were acquired using optical coherence tomography (OCT), which utilizes the low-coherence interference principle of near-infrared light to non-destructively penetrate the fluorine-modified layer to the substrate interface, acquiring cross-sectional structural images in the depth direction. The axial resolution can reach 1-2 micrometers, and the lateral resolution is about 3-5 micrometers, thus clearly distinguishing subsurface defects such as pores, delamination, and bubbles within the coating.

[0022] Surface chemical force distribution data were acquired using chemical force microscopy (CFM). Functionalized probes with the same functional groups as typical active drug molecules were selected (e.g., for drug molecules containing carboxyl groups, the probe surface was modified with carboxyl functional groups). Force-distance curve measurements were performed at grid points in each detection region (grid spacing was set to 5 micrometers to match the lateral resolution of OCT) to obtain the adhesion force value between the probe and the coating surface. This adhesion force value is derived from the combined contribution of hydrogen bonds, van der Waals forces and Coulomb interactions, and directly reflects the adsorption affinity of the surface for drug molecules.

[0023] S2. Combining microscopic morphology data and subsurface structure data before and after exposure, the geometric expansion of initial physical defects and the spatial distribution of newly formed defects under the action of drug solvent penetration are quantified; and based on the geometric expansion and the spatial distribution of newly formed defects, a physical barrier degradation rate distribution map of each detection area of ​​the fluorine material modified layer is generated, thereby analyzing the degree of uniformity degradation of physical homogeneity. S3. Combining the surface chemical interaction force distribution data before and after exposure with the adsorption characteristics of active drug molecules, adsorption hot spots are extracted and located; by comparing the changes in hot spot density and distribution range before and after exposure, an anti-adsorption function passivation distribution map of each detection area of ​​the fluorine material modified layer is generated, thereby analyzing the degree of chemical homogeneity passivation consistency. S4. Based on the degree of physical homogeneity degradation and chemical homogeneity passivation, spatial overlay and cross-correlation analysis are performed on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map to identify failure coupling regions; and based on the failure coupling regions, the tolerance of the fluorine material modified layer of pharmaceutical equipment to degradation homogeneity in a high-purity drug solvent environment is evaluated. S5. Based on the evaluation results of the tolerance to degradation uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment, implement differentiated process control schemes.

[0024] In this embodiment, for the first time, the propagation behavior of existing defects and the initiation behavior of new defects are incorporated into a unified quantitative framework. A degradation rate distribution map is generated through spatial interpolation, thereby assessing the uniformity of physical degradation across the entire coating surface. This solves the problem that traditional methods can only provide average degradation indices and cannot locate locally accelerated degradation regions. To achieve the above objectives, such as... Figure 2 As shown, step S2, which generates a physical barrier degradation rate distribution map for each detection area of ​​the fluorine-modified layer based on the geometric expansion and spatial distribution of newly formed defects, includes the following specific steps: S21. Optical coherence tomography (OCT) cross-sectional images of the same detection area before and after exposure from the subsurface structure data, and high-resolution optical microscopic morphology images from the microscopic morphology data, are registered with submicron precision. The OCT cross-sectional images of the same detection area before and after exposure are used as the registration objects. The image acquired before exposure is defined as the comparison image, and the image acquired after exposure is defined as the detection image. A feature matching algorithm based on feature point extraction and grid motion statistics is used to achieve feature matching between the detection image and the comparison image. In a specific embodiment, the registration algorithm uses a feature point extraction and grid motion statistics (GMS) matching algorithm based on accelerated robust feature extraction (SURF). The Hessian threshold for feature point extraction is set to 400, and the feature descriptor dimension is 64. In GMS matching, the grid size is set to 20×20 pixels, and the motion statistics consistency threshold is set to 0.85. The homography matrix is ​​calculated, and the images are aligned using perspective transformation. Further, in this embodiment, bilinear interpolation can be used to fill non-integer coordinate pixel values ​​after the transformation. Pixel-level difference operations are performed on the aligned image to obtain a difference image, highlighting the areas that changed before and after exposure. S22. For each defect region identified in the differential image, determine whether it existed before exposure; if the corresponding defect region existed before exposure, it is defined as an existing defect, and the existing defect area expansion rate, existing defect depth expansion amount and existing defect volume expansion rate are calculated. In this process, the corresponding defect area is located in the image before exposure, the number of pixels contained in the corresponding area is counted, and the number of pixels is multiplied by the actual area corresponding to a single pixel to obtain the area of ​​the defect before exposure. In the exposed image, the same area corresponding to the defect is located by spatial registration, the number of pixels contained in the corresponding area is counted, and the number of pixels is multiplied by the actual area corresponding to a single pixel (obtained by the optical system calibration, for example, each pixel corresponds to 0.18 micrometers × 0.18 micrometers under a 40x microscope objective) to obtain the area of ​​the exposed defect. Divide the difference between the defect area after exposure and the defect area before exposure by the defect area before exposure to obtain the expansion rate of the existing defect area. Locate the corresponding defect in the optical coherence tomography cross-sectional image before exposure, calculate the depth value of each pixel in the defect area, and take the depth corresponding to the centroid position of the depth value (axial distance from the coating surface to the bottom of the defect) as the defect depth before exposure. In the exposed optical coherence tomography cross-sectional image, the same defect is located by spatial registration, and the depth corresponding to the centroid position of the depth value of each pixel in the defect area is calculated as the depth of the defect after exposure. Subtracting the pre-exposure defect depth from the post-exposure defect depth yields the existing defect depth expansion. Three-dimensional reconstruction is performed on the optical coherence tomography cross-sectional image before exposure. The corresponding defect region of the three-dimensional reconstructed optical coherence tomography cross-sectional image before exposure is segmented. In a specific embodiment, the reconstruction algorithm adopts the moving cube method, the voxel size is set to a cube of 1 micrometer, the number of voxels contained in the corresponding defect is counted, and the number of voxels is multiplied by the actual volume corresponding to a single voxel to obtain the volume of the defect before exposure. Three-dimensional reconstruction was performed on the exposed optical coherence tomography cross-sectional image. The same defect was located by spatial registration. The number of voxels contained in the corresponding defect was counted and multiplied by the actual volume of a single voxel to obtain the volume of the exposed defect. The difference between the defect volume after exposure and the defect volume before exposure is divided by the defect volume before exposure to obtain the existing defect volume expansion rate. S23. If the corresponding defect area does not exist before exposure and only appears after exposure, it is defined as a new defect. Then, the number of pixels of the corresponding defect area in the differential image after exposure is multiplied by the actual area corresponding to a single pixel to obtain the area of ​​the new defect; the depth value calculated based on the axial pixel resolution and the centroid position of the corresponding defect is obtained as the depth of the new defect; and the number of voxels of the corresponding defect area is multiplied by the actual volume corresponding to a single voxel to obtain the volume of the new defect. S24. Divide the number of newly generated defects by the total area of ​​the detection region to obtain the surface density of newly generated defects. S25. For each detection area, handle it according to the type of defect in the corresponding detection area; When there are existing defects in the corresponding detection area, the existing defect area expansion rate, existing defect depth expansion amount and existing defect volume expansion rate are weighted and summed to obtain the existing defect degradation component of the corresponding detection area. When a newly formed defect exists within the corresponding detection area, its area, depth, and volume are extracted. These parameters are then normalized. A weighted sum of the normalized parameters is then calculated to obtain the degradation component of the newly formed defect within the corresponding detection area. In a specific embodiment, the normalization process uses a minimum-maximum normalization method: first, the minimum and maximum values ​​of the newly formed defect area are calculated across all detection areas; similarly, the minimum and maximum values ​​for depth and volume are calculated. Then, for each region with a newly formed defect, its original area, depth, and volume values ​​are subtracted from their respective minimum values, and then divided by the difference between their maximum and minimum values, thus uniformly transforming the three indicators to a range of 0 to 1. If a detection area has no newly formed defects, its area, depth, and volume are all considered as 0 and included in the statistics.

[0025] S26. The existing defect degradation component and the newly generated defect degradation component are weighted and summed to obtain the degradation trend state of the corresponding detection area; thus, the contribution of existing defects and newly generated defects to the degradation state can be dynamically adjusted according to the density of newly generated defects.

[0026] S27. Spatial interpolation is performed on the degradation trend states of all detection areas to generate the degradation rate distribution map of the physical barrier. In a specific embodiment, the Kriging interpolation method can be used to generate the degradation rate distribution map of the physical barrier. The variogram model of the Kriging interpolation is selected as a spherical model, the nugget constant is set to 0, the sill value is taken as the sample variance of the degradation trend state, and the range is set according to the median of the distance between detection areas.

[0027] In this embodiment, as Figure 2 As shown, step S2 analyzes the degree of physical homogeneity degradation uniformity, including the following specific steps: S28. For each detection area, construct a physical degradation feature vector, which includes: the existing defect area expansion rate, existing defect depth expansion amount, existing defect volume expansion rate, and new defect surface density of the corresponding detection area; if there are no existing defects in the corresponding detection area, then the existing defect area expansion rate, existing defect depth expansion amount, and existing defect volume expansion rate are set to 0; if there are no new defects in the corresponding detection area, then the new defect surface density is set to 0. S29. The physical degradation feature vector, along with the local differential image region and the local optical coherence tomography (OCT) cross-sectional image region corresponding to the detection region, are input into a pre-trained convolutional neural network model. The convolutional neural network model extracts defect morphology features and cross-sectional structure features from the local differential image region and the local OCT cross-sectional image region through convolutional layers and pooling layers, respectively. The extracted defect morphology features and cross-sectional structure features are then concatenated and fused with the physical degradation feature vector, thereby outputting the physical uniformity degradation degree of the corresponding detection region through a fully connected layer. For each detection region of the historical detection samples, the proportion of the cumulative area of ​​all penetrating defects in the corresponding detection region after exposure to the area of ​​the corresponding region is calculated as the training label of the convolutional neural network model. A penetrating defect refers to a defect that extends from the coating surface to the substrate interface, determined by the OCT cross-sectional image. For example, in this embodiment, a penetrating defect is defined as a defect that extends from the coating surface to the substrate interface, determined layer by layer by the OCT cross-sectional image (if there are defects in more than 10 consecutive layers and they extend from the top layer to the bottom layer, they are determined to be penetrating). The training labels of the convolutional neural network model range from 0 to 1. 0 indicates that there are no penetrating defects in the corresponding detection area and the physical barrier is completely intact. 1 indicates that the cumulative area of ​​penetrating defects in the corresponding detection area has reached or exceeded the area of ​​the corresponding detection area and the physical barrier has completely failed. In a specific embodiment, the specific architecture of the convolutional neural network is as follows: the input layer receives two image branches (each 32×32×1) and a feature vector (1×4).

[0028] Image processing branch: The first convolutional layer uses 64 3×3 convolutional kernels with a stride of 1 and the same padding, followed by a batch normalization layer and ReLU activation; the second convolutional layer uses 128 3×3 convolutional kernels with a stride of 1 and the same padding, followed by batch normalization and ReLU activation; then a 2×2 max pooling layer; the third convolutional layer uses 256 3×3 convolutional kernels with a stride of 1 and the same padding; another 2×2 max pooling layer; finally, it is flattened into a fully connected layer, outputting 512-dimensional features. The two image branches share the same structure but do not share weights.

[0029] The 512-dimensional features output from the two image branches are concatenated with the original feature vector (4-dimensional) to obtain a 1028-dimensional fused feature, which is then input into three fully connected layers: the first layer has 512 neurons with a dropout rate of 0.5; the second layer has 256 neurons with a dropout rate of 0.3; and the output layer has 1 neuron, which uses a sigmoid activation to output the physical uniformity degradation degree (range 0-1). In a specific implementation, the training set of the convolutional neural network contains 5000 historical detection region samples, the validation set contains 1000 samples, and the test set contains 1000 samples. Training uses the Adam optimizer with an initial learning rate of 0.001, which decays to 0.9 times the original rate every 10 epochs. The batch size is set to 32, the loss function is mean squared error, and an early stopping mechanism is used (the model stops if the validation loss does not decrease for 15 consecutive epochs). The model's mean absolute error on the test set is 0.037, and the coefficient of determination is 0.92.

[0030] S210. Calculate the standard deviation and mean of the physical uniformity degradation degree of all test areas. Subtract the ratio of the standard deviation and mean of the physical uniformity degradation degree of all test areas from 1 to obtain the physical uniformity degradation consistency of the fluorine material modified layer of pharmaceutical equipment. The closer the value is to 1, the more consistent the physical degradation rate of each area is, and the more uniform the physical barrier performance of the coating is.

[0031] In this embodiment, the quantitative measurement of surface chemical forces is combined with spatial distribution characteristics. The degree of chemical passivation is quantified using two complementary indicators: the hotspot density increase rate and the change in aggregation degree. This avoids the shortcomings of traditional methods that rely solely on changes in average adhesion force while ignoring spatial heterogeneity. Figure 3 As shown, step S3 generates an anti-adsorption passivation distribution map of each detection area of ​​the fluorine-modified layer by comparing the changes in hotspot density and distribution range before and after exposure; specifically, it includes the following steps: S31. Chemical force microscopy (CFM) was employed, using functionalized probes with surface-modified functional groups identical to those of typical active drug molecules. The CFM probes were prepared by surface modification of commercial silicon nitride probes using an aminosilane coupling agent, grafting with the same functional group as that of typical active drug molecules (e.g., atorvastatin calcium)—dihydroxyheptanoic acid. The spring constant of the probe was calibrated to 0.1-0.5 N / m using a thermal noise method. Force-distance curve measurements were performed at grid points in each detection region before and after exposure (grid spacing consistent with the OCT resolution of 5 micrometers in S1). In a specific embodiment, the curve measurement parameters were: approach velocity 2 μm / s, retraction velocity 2 μm / s, trigger force set to 2 nN, Z-axis scan range 2 μm, and 512 data points collected per point. This allows us to obtain the adhesion force value of each pixel in the surface chemical interaction force distribution data (the adhesion force value is extracted from the shrinkage segment of the force-distance curve, i.e., the maximum pulling force required for the probe to detach from the coating surface). The adhesion force values ​​of all grid points are arranged according to spatial coordinates to generate an adhesion force distribution map of the entire surface. The adhesion force value originates from the hydrogen bonds, van der Waals forces, and Coulomb interactions between the probe and surface functional groups, reflecting the adsorption affinity of the surface for drug molecules. S32. Based on the adsorption characteristics of active drug molecules, an adsorption hotspot determination threshold is set. Pixels whose adhesion force value exceeds the adsorption hotspot determination threshold before exposure are determined as existing adsorption hotspot sites before exposure. Adsorption hotspot sites that did not exist before exposure but newly appear after exposure are determined as newly formed adsorption hotspot sites. The adsorption hotspot determination threshold is obtained by adding the average adhesion force before exposure to the product of the sensitivity coefficient and the standard deviation of the adhesion force before exposure. The sensitivity coefficient is determined as follows: confirmed adsorption hotspots are obtained from multiple historical test samples. The adhesion force values ​​of these adsorption hotspots before exposure are extracted. The ratio of the difference between the adhesion force value of the confirmed adsorption hotspot before exposure and the mean adhesion force of all adsorption hotspots before exposure to the standard deviation of the adhesion force of all adsorption hotspots before exposure is calculated. The minimum value of the corresponding ratio in all historical samples is taken and rounded up as the sensitivity coefficient to ensure that all known adsorption hotspots can be correctly detected. S33. Count the number of adsorption hotspots before and after exposure in each detection area, divide the number of adsorption hotspots in the corresponding detection area by the area of ​​the corresponding detection area to obtain the hotspot density of the corresponding detection area; obtain the hotspot density before exposure and the hotspot density after exposure in the corresponding detection area; divide the difference between the hotspot density after exposure and the hotspot density before exposure in the corresponding detection area by the hotspot density before exposure to obtain the hotspot density increase rate of the corresponding detection area. S34. The absolute value of the difference between the nearest neighbor index of the hotspot distribution before and after exposure in the corresponding detection area is used as the change in the spatial clustering degree of the hotspots in the corresponding detection area, which is used to reflect the magnitude of the change in the spatial clustering pattern of hotspots before and after exposure. Among them, the nearest neighbor index (NNI) is used to describe the spatial distribution characteristics of hotspots. The nearest neighbor index is equal to the observed average nearest neighbor distance divided by the expected average nearest neighbor distance, where the expected average nearest neighbor distance is the product of the square root of the ratio of the detection area area to the number of hotspots and 0.5.

[0032] S35. The hot spot density increase rate and hot spot spatial aggregation change in the corresponding detection area are weighted and summed to obtain the chemical passivation degree of the corresponding detection area. S36. Spatially interpolate the degree of chemical passivation in all detection areas to generate the passivation distribution map of the anti-adsorption function.

[0033] In this embodiment, as Figure 3 As shown, step S3 analyzes the uniformity of chemical passivation, including the following specific steps: S37. For each detection area, construct a chemical passivation feature vector, which includes: the hot spot density increase rate, hot spot spatial aggregation change, average adhesion force value after exposure, and standard deviation of adhesion force value for the corresponding detection area; if there are no adsorption hot spot sites in the corresponding detection area, then the hot spot density increase rate and hot spot spatial aggregation change are assigned to 0. S38. The chemical passivation feature vector and the local region of the adhesion distribution map of the corresponding detection area are input into the pre-trained gradient boosting decision tree model; wherein, the chemical passivation feature vector serves as the structured feature input branch of the gradient boosting decision tree model, and the local region of the adhesion distribution map serves as the image feature input branch of the gradient boosting decision tree model; the gradient boosting decision tree model extracts the adhesion spatial texture features from the local region of the adhesion distribution map through its internal convolutional feature extraction layer, concatenates and fuses the adhesion spatial texture features with the chemical passivation feature vector, performs weighted decision-making on the fused features through ensemble learning, and outputs the degree of chemical uniformity passivation of the corresponding detection area; In a specific embodiment, the gradient boosting decision tree model adopts a dual-branch architecture: the image branch uses a lightweight convolutional neural network to extract the adhesion spatial texture features. This CNN contains two convolutional layers (the first layer has 32 3×3 convolutional kernels, and the second layer has 64 3×3 convolutional kernels) and a 2×2 global average pooling layer, which outputs a 128-dimensional feature vector.

[0034] The structured feature branch directly receives the 4-dimensional chemical passivation feature vector.

[0035] The 128-dimensional image features and 4-dimensional structured features are concatenated to form a 132-dimensional fused feature, which is then input into the XGBoost regressor. The XGBoost hyperparameters are set as follows: number of trees: 300; maximum depth: 6; learning rate: 0.05; subsample ratio: 0.8; column sampling ratio: 0.8; L2 regularization coefficient: 1.0; L1 regularization coefficient: 0.5.

[0036] Specifically, for each detection area of ​​historical test samples, the proportion of the cumulative area of ​​adsorption hotspots in the corresponding area after exposure to the area of ​​the corresponding area is calculated and used as the training label of the gradient boosting decision tree model; the value range of the training label of the gradient boosting decision tree model is from 0 to 1, where 0 indicates that there are no adsorption hotspots in the corresponding detection area and the anti-adsorption function is completely intact, and 1 indicates that the cumulative area of ​​adsorption hotspots in the corresponding detection area has reached or exceeded the area of ​​the corresponding detection area and the anti-adsorption function is completely lost. In a specific embodiment, the gradient boosting decision tree model is trained using five-fold cross-validation with 20 early stopping rounds, and the evaluation metric is root mean square error. For example, the model in this embodiment has a mean absolute error of 0.028 and a coefficient of determination of 0.94 on the test set.

[0037] S39. Calculate the standard deviation and mean of the chemical uniformity passivation degree of all detection areas. Subtract the ratio of the standard deviation to the mean of the chemical uniformity passivation degree of all detection areas from 1 to obtain the chemical uniformity passivation consistency of the fluorine material modified layer of pharmaceutical equipment. This can reflect the spatial uniformity of the anti-adsorption function degradation of the entire coating surface. The closer the value is to 1, the more consistent the passivation rate of each area.

[0038] In this embodiment, for the first time, two fundamentally different degradation modes—physical barrier degradation and chemical passivation—are spatially superimposed and cross-correlation analyzed to identify failure coupling regions. These regions are dangerous areas where physical degradation and chemical passivation occur simultaneously and exacerbate each other. Based on this, an evaluation dimension of environmental tolerance degradation uniformity—not addressed in existing technologies—is constructed. Step S4 involves spatially superimposing and cross-correlation analyzing the physical barrier degradation rate distribution map and the anti-adsorption functional passivation distribution map to identify failure coupling regions. Based on these failure coupling regions, the environmental tolerance degradation uniformity of the fluorinated material modified layer in pharmaceutical equipment under high-purity drug solvent environments is evaluated. This includes the following specific steps: S41. Align the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map in spatial coordinates, so that pixels at the same spatial location correspond to the same detection area in the two distribution maps. In this embodiment, since both distribution maps are generated based on the same detection area grid and have a unified coordinate system, they can be directly aligned by pixel coordinate correspondence. After alignment, assign each detection area both the degradation trend state in the physical barrier degradation rate distribution map and the chemical passivation degree in the anti-adsorption function passivation distribution map. Combine the degradation trend state and chemical passivation degree of each detection area into a two-dimensional vector. Arrange the two-dimensional vectors of all detection areas according to their spatial positions to form a physical-chemical two-dimensional degradation correlation matrix. S42. Perform a two-dimensional normalized cross-correlation analysis on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map, and calculate the cross-correlation coefficient; the two-dimensional normalized cross-correlation analysis is to treat the two distribution maps as two two-dimensional arrays of the same size and calculate the normalized cross-correlation coefficient between the two two-dimensional arrays. For example, in this embodiment, the calculation process of two-dimensional normalized cross-correlation analysis is to treat the two distribution maps as two two-dimensional arrays of the same size and calculate the degree of linear correlation between the two arrays pixel by pixel.

[0039] The following embodiment uses the physical barrier degradation rate distribution map (denoted as image A) and the anti-adsorption function passivation distribution map (denoted as image B) as examples to illustrate the specific calculation steps: A1. Assume that the size of both distribution maps is M rows × N columns, and the total number of pixels is M × N; where the pixel value of image A at coordinate (m,n) is denoted as A(m,n), and the pixel value of image B at coordinate (m,n) is denoted as B(m,n), where m is any term from 1 to M, and n is any term from 1 to N; A2. Calculate the arithmetic mean of all pixel values ​​in image A and image B respectively: ; ; In the formula, Let A be the pixel mean of image A. Let be the average pixel value of image B; A3. For each spatial coordinate position in the two distribution maps, calculate the deviation between the pixel value at each spatial coordinate position and its mean. Multiply the two deviations together, sum them over all spatial positions, and take the mean to obtain the covariance of the two distribution maps. ; In the formula, This is the covariance of two distribution maps; it describes the spatial consistency of the deviations of pixel values ​​from their respective means. If the deviations of the two distribution maps at the same spatial location are in the same direction (both above or below the mean), the product is positive; if the deviations are in opposite directions, the product is negative. A4. Calculate the variance of image A and image B respectively: ; ; In the formula, D(A) is the variance of image A, and D(B) is the variance of image B; A5. Divide the covariance in A3 by the square roots of the two variances in A4 to obtain the normalized cross-correlation coefficient r: ; In the formula, r is the normalized cross-correlation coefficient; S43. Preset physical degradation threshold and chemical passivation threshold, extract the degradation trend state and chemical passivation degree of each detection area from the physical and chemical dual-dimensional degradation correlation matrix, and mark the detection areas with degradation trend state greater than physical degradation threshold and chemical passivation degree greater than chemical passivation threshold as failure units; merge spatially adjacent failure units to form connected failure coupling regions; divide the sum of the areas of all failure coupling regions by the total area of ​​the detection areas to obtain the area ratio of the coupling region of the fluorine material modification layer of the pharmaceutical equipment; S44. Based on steps S41-S43, the coupling strength of the fluorine-modified layer of the pharmaceutical equipment is calculated. The formula for calculating the coupling strength of the fluorine-modified layer in the pharmaceutical equipment is as follows: ; In the formula, CSI represents the coupling strength of the fluorine-modified layer in pharmaceutical equipment, r is the normalized cross-correlation coefficient, Rc is the area ratio of the coupling region, upc is the arithmetic mean of the degradation trend states of all detection areas within the failure coupling region, upa is the arithmetic mean of the degradation trend states of all detection areas, ucc is the arithmetic mean of the chemical passivation degree of all detection areas within the failure coupling region, and uca is the arithmetic mean of the chemical passivation degree of all detection areas. When the arithmetic mean of the degradation trend states or the arithmetic mean of the chemical passivation degree of all detection areas is zero, it indicates that the coating has not degraded, and the coupling strength is assigned a value of 0. S45. The physical uniformity degradation consistency, chemical uniformity passivation consistency, and coupling strength of the fluorinated material modified layer of the pharmaceutical equipment are input into a pre-trained random forest regression model. The random forest regression model is trained with simulated experimental data and outputs the degradation tolerance uniformity of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment. The training labels of the random forest regression model are obtained in the following way: For historical test samples, the area ratio of their failure coupling regions is calculated, and the corresponding failure coupling region area ratio is used as the true value of degradation tolerance uniformity, where 0 indicates no failure coupling region and highly uniform degradation, and 1 indicates that the entire test area is a failure coupling region and extremely non-uniform degradation. At the same time, the precise spatial coordinates and area range of all failure coupling regions are automatically located and output.

[0040] In a specific embodiment, the training method for the random forest regression model is as follows: 800 historical detection samples are collected, each sample containing three-dimensional features (physical homogeneity degradation consistency, chemical homogeneity passivation consistency, and coupling strength) and manually labeled tolerance degradation homogeneity ground truth values. The ground truth labeling method is as follows: the area ratio of the failure coupling region in each sample is calculated, and this ratio is directly used as the tolerance degradation homogeneity ground truth value (0% corresponds to 0, 100% corresponds to 1).

[0041] In this embodiment, the hyperparameters of the random forest are set as follows: the number of trees is 200, the maximum depth is 15, the minimum number of sample splits is 5, the minimum number of leaf node samples is 2, the maximum number of features is "sqrt" (i.e., when sqrt(3)=1.73, it is rounded down to 2), the out-of-bag score is used for unbiased estimation, and the random seed is set to 42 to ensure repeatability.

[0042] In this embodiment, feature importance analysis after model training revealed that coupling strength contributed the most (approximately 0.55), followed by physical homogeneity degradation uniformity (approximately 0.30), and then chemical homogeneity passivation uniformity (approximately 0.15). This aligns with engineering intuition, namely, that the coupling effect between physical and chemical processes is the core factor determining the tolerance to degradation uniformity. In this embodiment, the model's root mean square error on the test set was 0.064, the mean absolute error was 0.051, and the coefficient of determination was 0.89. Simultaneously, the model can output the precise spatial coordinates and area range of all failure coupling regions, providing targeted localization information for subsequent differentiated process control.

[0043] In this embodiment, step S5, based on the evaluation results of the degradation tolerance uniformity of the fluorine-modified layer of the pharmaceutical equipment in a high-purity drug solvent environment, implements a differentiated process control scheme, including the following specific contents: S51. Obtain the evaluation results of the resistance to degradation uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment; S52. A preset tolerance to degradation uniformity threshold is set. When the tolerance to degradation uniformity assessment result of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment is greater than or equal to the tolerance to degradation uniformity threshold, the current operating parameters are kept unchanged, and the next tolerance to degradation uniformity test is performed according to the regular testing cycle. When the tolerance to degradation uniformity assessment result of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment is less than the tolerance to degradation uniformity threshold, targeted coating repair or local replacement is performed on the failure coupling area, and a second test is performed on the repaired or replaced area. After confirming that the tolerance to degradation uniformity is greater than or equal to the tolerance to degradation uniformity threshold, the equipment is put back into operation.

[0044] In this embodiment, the unspecified methods for determining the weights and thresholds are as follows: Microscopic morphology data, subsurface structure data, and surface chemical force distribution data of the fluorine-modified layer are acquired from 2000 sets of historical test samples. Physical degradation indicators such as the area expansion rate, depth expansion amount, volume expansion rate, and surface density of newly formed defects are calculated. Simultaneously, chemical passivation indicators such as the hotspot density increase rate and the change in spatial aggregation are calculated. The cumulative area ratio of penetrating defects after exposure is used as the training label for the degree of physical homogeneity degradation, and the cumulative area ratio of adsorbed hotspots after exposure is used as the training label for the degree of chemical homogeneity passivation. The 2000 sets of data were imported into SPSS or MATLAB fitting toolbox. A combination of multivariate logistic regression and grid search was used, with the optimization objective of maximizing the coefficient of determination. The weight combinations and threshold combinations were traversed, and the weight and threshold values ​​corresponding to the highest coefficient of determination were output. In a specific embodiment, the weights of the existing defect area expansion rate, existing defect depth expansion amount, and existing defect volume expansion rate in the existing defect degradation component were 0.35, 0.40, and 0.25, respectively. The weights of the new defect area, new defect depth, and new defect volume in the new defect degradation component were 0.3, 0.4, and 0.3, respectively. The weights of the chemical passivation degree were 0.7 (density increase rate) and 0.3 (aggregation degree change amount). The physical degradation threshold was 0.45, and the chemical passivation threshold was 0.35.

[0045] Example 2 like Figure 4 As shown, this embodiment provides a non-destructive testing system for the uniformity of fluorine-modified layers based on image sensors, including a data acquisition module, a physical degradation monitoring module, a chemical passivation monitoring module, a degradation tolerance uniformity assessment module, and a process control module. Among them, the data acquisition module is used to simultaneously collect micromorphological data, subsurface structure data and surface chemical force distribution data of multiple detection areas of the fluorine material modified layer of pharmaceutical equipment before and after exposure to a high-purity drug solvent environment; The physical degradation monitoring module is used to combine microscopic morphology data and subsurface structure data before and after exposure to quantify the geometric expansion of initial physical defects and the spatial distribution of newly formed defects under the action of drug solvent penetration; and based on the geometric expansion and the spatial distribution of newly formed defects, to generate a physical barrier degradation rate distribution map of each detection area of ​​the fluorinated material modified layer, thereby analyzing the degree of uniformity degradation of physical homogeneity. The chemical passivation monitoring module is used to extract and locate adsorption hotspots by combining the surface chemical force distribution data before and after exposure with the adsorption characteristics of active drug molecules. By comparing the changes in hotspot density and distribution range before and after exposure, an anti-adsorption functional passivation distribution map of each detection area of ​​the fluorinated material modified layer is generated, thereby analyzing the degree of chemical uniformity passivation consistency. The degradation tolerance uniformity assessment module is used to perform spatial overlay and cross-correlation analysis on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map based on the degree of physical uniformity degradation and the degree of chemical uniformity passivation to identify failure coupling regions; and to evaluate the degradation tolerance uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment based on the failure coupling regions. The process control module is used to execute differentiated process control schemes based on the evaluation results of the tolerance to degradation uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment.

[0046] The parameters and steps for implementing the corresponding functions of each unit module in the image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers of the present invention can be referred to the parameters and steps in the embodiments of the image sensor-based non-destructive testing method for the uniformity of fluorine-modified layers of fluorine materials described above, and will not be repeated here.

[0047] Example 3 An electronic device according to an embodiment of the present invention includes a processor and a memory. The memory stores a computer program that can be called by the processor. The processor executes a non-destructive testing method for the uniformity of fluorine-modified layers based on an image sensor by calling the computer program stored in the memory. It should be noted that all computer programs for the non-destructive testing method for the uniformity of fluorine-modified layers based on an image sensor are implemented using C language. The data acquisition module, physical degradation monitoring module, chemical passivation monitoring module, degradation tolerance uniformity assessment module, and process control module are all controlled by a remote server.

[0048] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with the corresponding embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0049] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A non-destructive testing system for the uniformity of fluorine-modified layers based on image sensors, characterized in that, It includes a data acquisition module, a physical degradation monitoring module, a chemical passivation monitoring module, a degradation tolerance uniformity assessment module, and a process control module; Among them, the data acquisition module is used to simultaneously collect micromorphological data, subsurface structure data and surface chemical force distribution data of multiple detection areas of the fluorine material modified layer of pharmaceutical equipment before and after exposure to a high-purity drug solvent environment; The physical degradation monitoring module is used to combine microscopic morphology data and subsurface structure data before and after exposure to quantify the geometric expansion of initial physical defects and the spatial distribution of newly formed defects under the action of drug solvent penetration; and based on the geometric expansion and the spatial distribution of newly formed defects, to generate a physical barrier degradation rate distribution map of each detection area of ​​the fluorinated material modified layer, thereby analyzing the degree of uniformity degradation of physical homogeneity. The chemical passivation monitoring module is used to extract and locate adsorption hotspots by combining the surface chemical force distribution data before and after exposure with the adsorption characteristics of active drug molecules. By comparing the changes in hotspot density and distribution range before and after exposure, an anti-adsorption functional passivation distribution map of each detection area of ​​the fluorinated material modified layer is generated, thereby analyzing the degree of chemical uniformity passivation consistency. The degradation tolerance uniformity assessment module is used to perform spatial overlay and cross-correlation analysis on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map based on the degree of physical uniformity degradation and the degree of chemical uniformity passivation to identify failure coupling regions; and to evaluate the degradation tolerance uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment based on the failure coupling regions. The process control module is used to execute differentiated process control schemes based on the evaluation results of the tolerance to degradation uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment.

2. The image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers according to claim 1, characterized in that, The process of generating a physical barrier degradation rate distribution map for each detection area of ​​the fluorine-modified layer based on the geometric expansion and spatial distribution of newly formed defects includes the following specific steps: S21. Optical coherence tomography (OCT) cross-sectional images of the same detection area before and after exposure in the subsurface structure data, and high-resolution optical microscopic morphology images in the microscopic morphology data, are registered with submicron precision. The OCT cross-sectional images of the same detection area before and after exposure are used as the registration objects. The image acquired before exposure is defined as the comparison image, and the image acquired after exposure is defined as the detection image. A feature matching algorithm based on feature point extraction and grid motion statistics is used to achieve feature matching between the detection image and the comparison image. The homography matrix is ​​calculated, and the images are aligned through perspective transformation. Pixel-level difference operations are performed on the aligned images to obtain a difference image, highlighting the areas that changed before and after exposure. S22. For each defect region identified in the differential image, determine whether it existed before exposure; if the corresponding defect region existed before exposure, it is defined as an existing defect, and the existing defect area expansion rate, existing defect depth expansion amount and existing defect volume expansion rate are calculated. S23. If the corresponding defect area does not exist before exposure and only appears after exposure, it is defined as a new defect. Then, the number of pixels of the corresponding defect area in the differential image after exposure is multiplied by the actual area corresponding to a single pixel to obtain the area of ​​the new defect; the depth value calculated based on the axial pixel resolution and the centroid position of the corresponding defect is obtained as the depth of the new defect; and the number of voxels of the corresponding defect area is multiplied by the actual volume corresponding to a single voxel to obtain the volume of the new defect. S24. Divide the number of newly generated defects by the total area of ​​the detection region to obtain the surface density of newly generated defects. S25. For each detection area, handle it according to the type of defect in the corresponding detection area; When there are existing defects in the corresponding detection area, the existing defect area expansion rate, existing defect depth expansion amount and existing defect volume expansion rate are weighted and summed to obtain the existing defect degradation component of the corresponding detection area. When a new defect exists in the corresponding detection area, the area, depth, and volume of the new defect are extracted. The area, depth, and volume of the new defect are normalized. The normalized area, depth, and volume of the new defect are then weighted and summed to obtain the degradation component of the new defect in the corresponding detection area. S26. The existing defect degradation component and the newly generated defect degradation component are weighted and summed to obtain the degradation trend state of the corresponding detection area; S27. Spatial interpolation is performed on the degradation trend status of all detection areas to generate the degradation rate distribution map of the physical barrier.

3. The image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers according to claim 2, characterized in that, The analysis of the degree of physical homogeneity degradation includes the following specific steps: S28. For each detection area, construct a physical degradation feature vector, which includes: the existing defect area expansion rate, existing defect depth expansion amount, existing defect volume expansion rate, and new defect surface density of the corresponding detection area; if there are no existing defects in the corresponding detection area, then the existing defect area expansion rate, existing defect depth expansion amount, and existing defect volume expansion rate are set to 0; if there are no new defects in the corresponding detection area, then the new defect surface density is set to 0. S29. The physical degradation feature vector, along with the local differential image region and the local optical coherence tomography (OCT) cross-sectional image region corresponding to the detection region, are input into a pre-trained convolutional neural network model. The convolutional neural network model extracts defect morphology features and cross-sectional structure features from the local differential image region and the local OCT cross-sectional image region through convolutional layers and pooling layers, respectively. The extracted defect morphology features and cross-sectional structure features are then concatenated and fused with the physical degradation feature vector, thereby outputting the degree of physical uniformity degradation of the corresponding detection region through a fully connected layer. For each detection region of the historical detection samples, the proportion of the cumulative area of ​​all penetrating defects in the corresponding detection region after exposure to the area of ​​the corresponding region is calculated and used as the training label of the convolutional neural network model. S210. Calculate the standard deviation and mean of the physical uniformity degradation degree of all detection areas, and subtract the ratio of the standard deviation and mean of the physical uniformity degradation degree of all detection areas from 1 to obtain the uniformity degradation degree of the physical uniformity of the fluorine material modified layer of the pharmaceutical equipment.

4. The image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers according to claim 3, characterized in that, The process involves comparing the changes in hotspot density and distribution range before and after exposure to generate an anti-adsorption passivation distribution map of each detection area in the fluorine-modified layer; specifically, it includes the following steps: S31. Using a chemical force microscope, functionalized probes with the same functional groups as typical active drug molecules are modified on the surface. Force-distance curve measurements are performed at grid points in each detection area before and after exposure to obtain the adhesion force value of each pixel in the surface chemical force distribution data and generate an adhesion force distribution map of the entire surface. S32. Based on the adsorption characteristics of active drug molecules, an adsorption hotspot determination threshold is set. Pixels whose adhesion force value exceeds the adsorption hotspot determination threshold before exposure are determined as existing adsorption hotspot sites before exposure. Adsorption hotspot sites that did not exist before exposure but newly appear after exposure are determined as newly formed adsorption hotspot sites. The adsorption hotspot determination threshold is obtained by adding the average adhesion force before exposure to the product of the sensitivity coefficient and the standard deviation of the adhesion force before exposure. S33. Count the number of adsorption hotspots before and after exposure in each detection area, divide the number of adsorption hotspots in the corresponding detection area by the area of ​​the corresponding detection area to obtain the hotspot density of the corresponding detection area; obtain the hotspot density before exposure and the hotspot density after exposure in the corresponding detection area; divide the difference between the hotspot density after exposure and the hotspot density before exposure in the corresponding detection area by the hotspot density before exposure to obtain the hotspot density increase rate of the corresponding detection area. S34. The absolute value of the difference between the nearest neighbor index of the hotspot distribution before and after exposure in the corresponding detection area is taken as the change in the spatial clustering of hotspots in the corresponding detection area. S35. The hot spot density increase rate and hot spot spatial aggregation change in the corresponding detection area are weighted and summed to obtain the chemical passivation degree of the corresponding detection area. S36. Spatially interpolate the degree of chemical passivation in all detection areas to generate the passivation distribution map of the anti-adsorption function.

5. The image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers according to claim 4, characterized in that, The analysis of the uniformity of passivation in chemical homogeneity includes the following specific steps: S37. For each detection area, construct a chemical passivation feature vector, which includes: the hot spot density increase rate, hot spot spatial aggregation change, average adhesion force value after exposure, and standard deviation of adhesion force value for the corresponding detection area; if there are no adsorption hot spot sites in the corresponding detection area, then the hot spot density increase rate and hot spot spatial aggregation change are assigned to 0. S38. The chemical passivation feature vector and the local region of the adhesion distribution map of the corresponding detection area are input into the pre-trained gradient boosting decision tree model; wherein, the chemical passivation feature vector serves as the structured feature input branch of the gradient boosting decision tree model, and the local region of the adhesion distribution map serves as the image feature input branch of the gradient boosting decision tree model; the gradient boosting decision tree model extracts the adhesion spatial texture features from the local region of the adhesion distribution map through its internal convolutional feature extraction layer, concatenates and fuses the adhesion spatial texture features with the chemical passivation feature vector, performs weighted decision-making on the fused features through ensemble learning, and outputs the degree of chemical uniformity passivation of the corresponding detection area; In this process, for each detection region of the historical detection sample, the proportion of the cumulative area of ​​the adsorption hotspot sites in the corresponding region after exposure to the area of ​​the corresponding region is calculated, and this proportion is used as the training label for the gradient boosting decision tree model. S39. Calculate the standard deviation and mean of the chemical uniformity passivation degree of all detection areas, and subtract the ratio of the standard deviation to the mean of the chemical uniformity passivation degree of all detection areas from 1 to obtain the chemical uniformity passivation consistency of the fluorine material modified layer of the pharmaceutical equipment.

6. The image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers according to claim 5, characterized in that, The spatial overlay and cross-correlation analysis of the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map are performed to identify failure coupling regions. Furthermore, based on the failure coupling region, the degradation tolerance uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment was evaluated. The specific steps include the following: S41. Align the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map on the spatial coordinates so that the pixels at the same spatial location correspond to the same detection area in the two distribution maps. After alignment, each detection region is simultaneously assigned the corresponding region's degradation trend state in the physical barrier degradation rate distribution map and the chemical passivation degree in the anti-adsorption function passivation distribution map. The degradation trend state and chemical passivation degree of each detection region are combined into a two-dimensional vector. The two-dimensional vectors of all detection regions are arranged according to their spatial positions to form a physical-chemical two-dimensional degradation correlation matrix. S42. Perform two-dimensional normalized cross-correlation analysis on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map, and calculate the cross-correlation coefficient; The two-dimensional normalized cross-correlation analysis involves treating the two distribution maps as two two-dimensional arrays of the same size and calculating the normalized cross-correlation coefficient between these two two-dimensional arrays. S43. Preset physical degradation threshold and chemical passivation threshold, extract the degradation trend state and chemical passivation degree of each detection area from the physical and chemical dual-dimensional degradation correlation matrix, and mark the detection areas with degradation trend state greater than physical degradation threshold and chemical passivation degree greater than chemical passivation threshold as failure units. Spatially adjacent failure units are merged to form a connected failure coupling region; Divide the sum of the areas of all failed coupling regions by the total area of ​​the detection area to obtain the proportion of the coupling region area of ​​the fluorine material modified layer in the pharmaceutical equipment. S44. Based on steps S41-S43, the coupling strength of the fluorine-modified layer of the pharmaceutical equipment is calculated. S45. Input the physical uniformity degradation consistency, chemical uniformity passivation consistency, and coupling strength of the fluorine material modified layer of the pharmaceutical equipment into a pre-trained random forest regression model. The corresponding random forest regression model is trained with simulated experimental data and outputs the tolerance degradation uniformity of the fluorine material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment.

7. The image sensor-based non-destructive testing system for the uniformity of fluorine-modified layers according to claim 6, characterized in that, Based on the evaluation results of the degradation tolerance uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment, a differentiated process control scheme is implemented, including the following specific contents: S51. Obtain the evaluation results of the resistance to degradation uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment; S52. A preset tolerance to degradation uniformity threshold is set. When the tolerance to degradation uniformity assessment result of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment is greater than or equal to the tolerance to degradation uniformity threshold, the current operating parameters are kept unchanged, and the next tolerance to degradation uniformity test is performed according to the regular testing cycle. When the tolerance to degradation uniformity assessment result of the fluorinated material modified layer of the pharmaceutical equipment in a high-purity drug solvent environment is less than the tolerance to degradation uniformity threshold, targeted coating repair or local replacement is performed on the failure coupling area, and a second test is performed on the repaired or replaced area. After confirming that the tolerance to degradation uniformity is greater than or equal to the tolerance to degradation uniformity threshold, the equipment is put back into operation.

8. A non-destructive testing method for the uniformity of fluorine-modified layers based on an image sensor, applied to the non-destructive testing system for the uniformity of fluorine-modified layers based on an image sensor as described in any one of claims 1-7, characterized in that, The specific steps include the following: S1. Simultaneously collect micromorphological data, subsurface structure data, and surface chemical force distribution data for multiple detection areas of the fluorine-modified layer of pharmaceutical equipment before and after exposure to a high-purity drug solvent environment; S2. Combining microscopic morphology data and subsurface structure data before and after exposure, the geometric expansion of initial physical defects and the spatial distribution of newly formed defects under the action of drug solvent penetration are quantified; and based on the geometric expansion and the spatial distribution of newly formed defects, a physical barrier degradation rate distribution map of each detection area of ​​the fluorine material modified layer is generated, thereby analyzing the degree of uniformity degradation of physical homogeneity. S3. Combining the surface chemical interaction force distribution data before and after exposure with the adsorption characteristics of active drug molecules, adsorption hot spots are extracted and located; by comparing the changes in hot spot density and distribution range before and after exposure, an anti-adsorption function passivation distribution map of each detection area of ​​the fluorine material modified layer is generated, thereby analyzing the degree of chemical homogeneity passivation consistency. S4. Based on the degree of physical homogeneity degradation and chemical homogeneity passivation, spatial overlay and cross-correlation analysis are performed on the physical barrier degradation rate distribution map and the anti-adsorption function passivation distribution map to identify failure coupling regions. Furthermore, based on the failure coupling region, the degradation tolerance uniformity of the fluorinated material modified layer of pharmaceutical equipment in a high-purity drug solvent environment was evaluated. S5. Based on the evaluation results of the tolerance to degradation uniformity of the fluorine-modified layer of pharmaceutical equipment in a high-purity drug solvent environment, implement differentiated process control schemes.

9. An electronic device, comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; characterized in that the processor executes the image sensor-based non-destructive testing method for the uniformity of fluorine-modified layers as described in claim 8 by calling the computer program stored in the memory.

Images