Method for detecting aging performance of PVC film coating based on thermal analysis

Abstract

The application discloses a PVC film coating aging performance detection method based on thermal analysis, and particularly relates to the field of high polymer material thermal analysis technology; a PVC film coating sample to be detected is provided and dried; a thermogravimetric analysis device is used to carry out temperature rising test under the atmosphere of inert gas or oxygen-containing mixed gas; a micro-area photothermal probe imaging system is integrated on the sample cavity of the thermogravimetric analysis device, and is used for synchronously collecting the temperature distribution image of the sample surface; the micro-area temperature rising rate and the thermal diffusion characteristics are recognized through the imaging system, the local temperature rising acceleration abnormal value and the mass loss derivative peak value are generated in combination with the thermogravimetric analysis data; based on the recognized abnormal point position, the pyrolysis curve is corrected or marked by using a graph neural perturbation suppression algorithm, and the optimized aging performance analysis result is output; the application can accurately recognize and remove the local abnormality caused by the oxidation window effect, and significantly improves the accuracy and reproducibility of the PVC film coating thermal stability evaluation.

Description

Technical Field

[0001] This invention relates to the field of polymer material thermal analysis technology, specifically to a method for testing the aging performance of PVC film coatings based on thermal analysis. Background Technology

[0002] Polyvinyl chloride (PVC) film is widely used in building waterproofing, outdoor signage, and electronic materials due to its excellent flexibility, weather resistance, and processing properties. To improve its surface properties, various functional coatings are usually applied to the surface of PVC film, such as flame-retardant coatings, UV-resistant coatings, and anti-fouling coatings. However, under long-term exposure to heat, light, and oxygen, PVC film coatings are prone to aging, manifesting as discoloration, cracking, peeling, and performance degradation, which seriously affects its service life and safety performance.

[0003] Currently, the industry commonly uses thermal analysis methods such as thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) to evaluate the aging performance of PVC film coatings, mainly reflecting the degree of aging through indicators such as mass change, decomposition temperature, and thermal stability parameters. However, existing technologies have the following technical problems:

[0004] If the atmosphere control is unstable during TGA testing, especially in air or oxygen environments, it can easily lead to the "oxidation window effect," which means that non-uniform oxidation, self-heating, or even trace combustion occurs in local areas of the sample, thereby interfering with the thermal decomposition curve and causing analytical errors. Traditional TGA devices only provide overall mass-temperature data and cannot identify temperature abrupt changes or hot spot behavior in local areas of the sample surface, which poses a risk of misjudgment. Conventional systems are mostly open-loop controls, which make it difficult to achieve feedback adjustment for instantaneous changes in oxygen concentration, leading to misjudgment of aging performance, especially in high-sensitivity evaluation scenarios (such as early degradation detection).

[0005] Therefore, there is an urgent need for a new detection method with high-resolution thermal field identification capabilities that can effectively monitor and correct local thermal anomalies, in order to improve the accuracy and reliability of aging performance evaluation of PVC film coatings. Summary of the Invention

[0006] The purpose of this invention is to provide a method for testing the aging performance of PVC film coatings based on thermal analysis, so as to overcome the shortcomings of the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for testing the aging performance of PVC film coatings based on thermal analysis, comprising:

[0008] Provide a sample of the PVC film coating to be tested and perform a drying process;

[0009] The sample is placed in a thermogravimetric analyzer for a temperature rise test, while the atmosphere is set to an inert gas or an oxygen-containing mixture.

[0010] A micro-area photothermal probe imaging system is integrated into the sample chamber of the thermogravimetric analysis device to simultaneously acquire temperature distribution images of the sample surface;

[0011] The imaging system has real-time temperature imaging capabilities and the ability to identify the temperature rise rate and thermal diffusion characteristics of sample micro-areas.

[0012] The temperature image data and the TGA mass change curve are analyzed synchronously, and local temperature rise acceleration anomalies and second-order mass loss derivative peak values ​​are generated to comprehensively determine whether the oxidation window is abnormal.

[0013] Based on the identified anomalies, the thermal decomposition data is corrected or marked, and the optimized aging performance analysis results of the PVC film coating are output.

[0014] Preferably, the micro-area photothermal probe imaging system integrated on the sample chamber of the thermogravimetric analysis device includes:

[0015] An infrared transparent window is provided at the top or side wall of the sample chamber;

[0016] A MEMS infrared thermal imaging probe with micro-focus modulation function is fixedly installed outside the window to capture a two-dimensional temperature distribution image of the sample surface.

[0017] The probe is synchronized with the heating program of the thermogravimetric analysis device through the control module to achieve time matching between thermal image acquisition and temperature-mass data.

[0018] The data collected by the probe is transmitted to the central processing system in real time to construct a thermal imaging sequence.

[0019] Preferably, the synchronous analysis of temperature image data and TGA quality change curves includes:

[0020] The method for generating local temperature rise acceleration anomalies is as follows: stack consecutive thermal image frames to form a three-dimensional data structure, including pixel position (x,y) and time index t; for each frame, take a window Ωr(x,y) with radius r around each pixel (x,y), calculate the average temperature of the window, and record it as the local average temperature; subtract the current temperature of each pixel from its local average temperature to form a residual mapping; perform a second-order time difference on the residual sequence of each pixel to calculate the local temperature rise acceleration anomaly.

[0021] The preferred method for generating the peak value of the second-order mass loss derivative is as follows:

[0022] Let the change in sample mass over time be m(t), then its first derivative is... For thermal decomposition rate, the second derivative This indicates the trend of change in the thermal decomposition rate, i.e., acceleration;

[0023] The five-point central difference method is used to calculate... The values ​​throughout the testing process form a second derivative curve. Local extreme points are extracted from this curve, and the peak value of the second-order mass loss derivative is defined. for: .

[0024] Preferably, at all time points, the local temperature rise acceleration anomalies and the peak values ​​of the second-order mass loss derivative are converted into comprehensive feature vectors. The comprehensive feature vectors are used as input to the machine learning model. The machine learning model uses the prediction of local abnormal pyrolysis behavior analysis value labels for each set of comprehensive feature vectors as the prediction objective and minimizes the sum of prediction errors for all local abnormal pyrolysis behavior analysis value labels as the training objective. The machine learning model is trained until the sum of prediction errors converges, at which point the model training stops. The local abnormal pyrolysis behavior analysis values ​​are determined based on the model output. The machine learning model is an isolated forest model.

[0025] Preferably, the analysis value of local abnormal pyrolysis behavior is compared with the preset empirical value. When the analysis value of local abnormal pyrolysis behavior is greater than the preset empirical value, the time point is marked as the oxidation window abnormal trigger point.

[0026] Preferably, the thermal decomposition data corrected based on identified outlier locations includes:

[0027] Construct a time-space data plot and map outliers to the corresponding time nodes in the TGA curve;

[0028] The temporal graph neural perturbation suppression algorithm is used to evaluate the impact range of outliers on subsequent nodes;

[0029] For the data in the affected segments, select interpolation correction, curve compression, or anomaly marking strategies;

[0030] Based on the corrected curve, the aging performance parameters are re-extracted and the optimized results are output.

[0031] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0032] 1. This invention integrates a micro-area photothermal probe imaging system with a thermogravimetric analysis (TGA) device, enabling high-resolution real-time monitoring of the surface temperature field during the thermal decomposition of PVC film coatings. This overcomes the limitations of traditional TGA, which can only acquire overall mass changes and cannot identify local thermal anomalies. By identifying the micro-area temperature rise rate and thermal diffusion characteristics, this invention can effectively detect local self-heating behavior caused by the oxidation window effect, avoiding misjudgment of the pyrolysis curve from the source and significantly improving the accuracy and sensitivity of aging detection.

[0033] 2. This invention introduces a spatiotemporal residual network and a graph neural perturbation suppression algorithm (T-GPS) to jointly analyze thermal imaging and quality data, achieving for the first time dynamic identification and segmented correction of abnormal points in thermal decomposition data. Compared with existing static filtering or overall smoothing methods, this method has higher adaptability and positioning capabilities, effectively removing non-realistic weightlessness anomalies caused by atmospheric fluctuations, and improving the stability and repeatability of key parameters (Tonset, Tmax, etc.), thereby providing more reliable data support for the life assessment and thermal stability evaluation of PVC film coatings. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0035] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Example 1, please refer to Figure 1 As shown in this embodiment, the aging performance testing method for PVC film coatings based on thermal analysis includes:

[0038] Provide a sample of the PVC film coating to be tested and perform a drying process;

[0039] The sample is placed in a thermogravimetric analyzer for a temperature rise test, while the atmosphere is set to an inert gas or an oxygen-containing mixture.

[0040] A micro-area photothermal probe imaging system is integrated into the sample chamber of the thermogravimetric analysis device to simultaneously acquire temperature distribution images of the sample surface;

[0041] The imaging system has real-time temperature imaging capabilities and the ability to identify the temperature rise rate and thermal diffusion characteristics of sample micro-areas.

[0042] The temperature image data and the TGA mass change curve are analyzed synchronously, and local temperature rise acceleration anomalies and second-order mass loss derivative peak values ​​are generated to comprehensively determine whether the oxidation window is abnormal.

[0043] Based on the identified anomalies, the thermal decomposition data is corrected or marked, and the optimized aging performance analysis results of the PVC film coating are output.

[0044] In this embodiment, a polyvinyl chloride (PVC) base film with a thickness of 80±5 μm was selected, and a fluorinated modified acrylate coating was uniformly applied to its surface, with the coating thickness controlled at 10±2 μm, to obtain a representative PVC film-coated composite structure. To avoid interference from moisture or volatile components on the mass change curve during the test, the sample needs to be pretreated.

[0045] The composite membrane sample was cut into 5mm × 5mm pieces, placed in a clean ceramic dish using clean tweezers, and then dried in a constant temperature vacuum drying oven at 60°C for 12 hours to ensure that free moisture, residual solvent, and surface-adsorbed gases were fully removed. During the drying process, the vacuum level was controlled to be no lower than -0.08 MPa to improve evaporation efficiency and prevent oxidation.

[0046] After drying, the samples were immediately transferred to a sealed inert gas container for storage, pending testing. This drying process ensures the initial quality stability and repeatability of decomposition behavior in subsequent thermogravimetric analysis data, helping to identify true aging behavior rather than test artifacts.

[0047] After drying, the PVC film coating sample is gently placed in a platinum thermogravimetric analysis crucible using antistatic tweezers, ensuring that the bottom of the sample is in full contact with the bottom of the crucible and avoiding factors that may affect the uniformity of heat conduction, such as curling or edge warping. The crucible is then mounted on the sample holder of the thermogravimetric analyzer (model: Netzsch TG 209F3 or equivalent), confirming that the thermocouple and sample are in the same thermal zone.

[0048] The heating program was set as follows: initial temperature 30°C, final temperature 700°C, heating rate 10°C / min, operating in automatic temperature control mode throughout. Based on the aging performance evaluation objective, the test atmosphere was set under the following two operating conditions:

[0049] Inert atmosphere condition: High-purity nitrogen (≥99.999%) is introduced, and the gas flow rate is set to 50 mL / min. This is suitable for evaluating the pyrolysis characteristics and main chain stability of PVC coatings.

[0050] Oxidizing atmosphere condition: Air (or 20%) is introduced. + 80% (Mixed gas), with the same airflow rate as above, is used to simulate oxidation-induced decomposition behavior in actual aging environments.

[0051] To avoid the "oxidation window effect" caused by abrupt atmosphere changes, a pre-purge stage (5 minutes) was included in all atmosphere switching processes, and the gas ratio was precisely controlled using a mass flow controller (MFC). During the experiment, the instrument automatically recorded the sample mass change (TG) and its first derivative (DTG) data for subsequent analysis in sync with temperature images.

[0052] To achieve real-time monitoring of localized temperature anomalies during the pyrolysis process of PVC film-coated samples, a micro-area photothermal probe imaging system was integrated into the sample chamber of the thermogravimetric analysis device. The specific structure and implementation are as follows:

[0053] First, a window opening is machined at the top of the sample testing chamber of the thermogravimetric analyzer (such as the Netzsch TG 209 series). The window is encapsulated with a high infrared transmittance material, such as a high-purity silicon wafer or germanium glass sheet, with a thickness controlled within 1.0 mm. The optical transmission range covers the 3–14 μm band to ensure that there is no significant attenuation of the sample's thermal radiation signal in the mid-infrared band. The window is sealed using a metal flange and a heat-resistant sealing ring to ensure the airtightness and temperature stability of the testing chamber.

[0054] A MEMS-structured infrared thermal imaging probe module (such as FLIRLepton 3.5 or a custom infrared focal plane array) is fixedly installed at the center of the outer side of the window. This probe has the following key parameters:

[0055] The spatial resolution is not less than 160×120 pixels, and the area of ​​the equivalent minimum resolution unit is ≤100μm×100μm;

[0056] Temperature resolution (NETD) is better than 50 mK, enabling the detection of minute thermal changes;

[0057] Set the frame rate to 9–30 fps to ensure it matches the TGA warm-up rate;

[0058] It features a micro-focus modulation mechanism (VGA micro-actuator module) to adjust the focal plane and sample height consistency, thereby optimizing image sharpness.

[0059] The probe is connected to the main control platform via an external data interface of the thermal analysis control system. It integrates a thermal field calibration and image synchronization control module, which receives time signals from the TGA heating program and performs time axis alignment and temperature curve calibration on the imaging data. The system automatically associates each frame of thermal image with real-time temperature points, thereby establishing a three-dimensional data channel of "temperature-image-time".

[0060] During imaging, the probe acquires thermal images of the sample surface and transmits them to the main control software in real time for subsequent anomaly identification, gradient calculation, and data fusion analysis. This integrated device does not damage the hermetic structure of the TGA or affect the uniformity of the thermal field, significantly improving the ability to capture local thermal anomalies (such as oxidation hotspots and auto-ignition points).

[0061] This design addresses the problem that existing thermogravimetric analysis instruments cannot detect the spatial thermal field distribution on the sample surface, providing a structural and data foundation for the intelligent determination of the "oxidation window effect" during the aging process of PVC film coatings.

[0062] In this embodiment, the micro-area photothermal probe imaging system not only has the ability to collect mid-infrared thermal radiation from the sample surface in real time, but also has the function of dynamically identifying and quantifying the micro-area temperature rise rate and thermal diffusion behavior. Its core technology consists of: a real-time image acquisition module, a temperature rise rate calculation module, and a thermal diffusion anomaly judgment module.

[0063] During actual testing, the infrared probe acquires thermal images of the sample surface at a frame rate of 10fps, and each frame of thermal image is precisely bound to the heating timestamp of the TGA control system to generate a continuous sequence of thermal field images. To improve image processing speed and analysis accuracy, the thermal image data is cached locally before being processed by the FPGA or embedded GPU module.

[0064] Temperature rise rate calculation module: The numerical derivative of the temperature values ​​of consecutive frames of the same pixel is calculated by the image difference algorithm to obtain the instantaneous temperature rise rate (°C / s) of each spatial unit (e.g., 100μm×100μm) of the sample; when the temperature rise rate of a certain area is significantly higher than 3 times the average rate of the whole field, it is initially marked as a "local temperature rise anomaly area".

[0065] Thermal diffusion feature identification module: Within the marked area, the system automatically calculates the temperature gradient change trend of the area spreading outwards, and determines whether the hot zone exhibits abnormal diffusion behavior by fitting a two-dimensional Gaussian thermal diffusion model; if the thermal diffusion rate exceeds the thermal diffusivity of the standard polymer material (e.g., > If the value is m² / s, it is determined to be an abnormal thermal focusing or a potential oxidation hotspot.

[0066] Anomaly detection output: The imaging system can feed back the above identification results to the main control software in the form of image overlay or annotation, assisting the user in determining whether the sample is affected by the "oxidation window effect". The identification information can also be jointly analyzed with TG curve data as a basis for identifying anomalies in decomposition behavior.

[0067] This functional module breaks through the limitation of traditional thermal analysis methods that cannot perceive local thermodynamic behavior. For the first time, it realizes the simultaneous analysis of the three-dimensional thermal behavior of thin-film polymer materials such as PVC film during thermal aging, providing a structured basis for judging local oxidation, self-heating and thermal stability abrupt changes.

[0068] In this embodiment, after completing sample drying, TGA temperature rise test, imaging system integration and image acquisition, the multimodal data synchronous analysis mechanism proposed in this invention is used to perform time registration and data fusion of temperature image data and thermogravimetric analysis curve (TG / DTG). By constructing a comprehensive judgment algorithm, two key parameters are output: local temperature rise acceleration anomaly value and second-order mass loss derivative peak value, and these are used to determine whether local abnormal pyrolysis behavior caused by "oxidation window effect" occurs during the test.

[0069] The imaging system and the TGA control system acquire data on the same sampling time axis via a synchronous controller, with a sampling period of 0.1 seconds. The TGA system records the instantaneous temperature T(t), mass m(t), and first derivative (DTG) of the sample. The thermal imaging system records each frame of image as a two-dimensional temperature matrix with a resolution of 160×120, covering the entire surface of the sample.

[0070] The method for generating localized accelerated temperature rise anomalies is as follows:

[0071] A three-dimensional data structure is formed by stacking consecutive thermal images, including pixel positions (x, y) and time indices t. For each image frame, a window Ωr(x, y) of radius r is taken around each pixel (x, y), and the average temperature of this window is calculated and denoted as the local average temperature. The difference between the current temperature of each pixel and its local average temperature forms a residual mapping: this residual reflects whether the point "jumps" out of the surrounding temperature background. The second-order temporal difference is performed on the residual sequence of each pixel, which calculates the local temperature rise acceleration anomaly value. The larger the absolute value, the more drastic the abrupt change in the temperature rise rate.

[0072] The method for generating the peak value of the second-order mass loss derivative is as follows:

[0073] Let the change in sample mass over time be m(t), then its first derivative is... For thermal decomposition rate, the second derivative This indicates the trend of change in the thermal decomposition rate, i.e., acceleration.

[0074] The five-point central difference method is used to calculate... The values ​​throughout the testing process form a second derivative curve. Local extreme points in this curve are extracted, and noise fluctuations are filtered out (those with amplitudes less than three times the background variance are considered invalid).

[0075] Define the peak value of the second-order mass loss derivative. for: The heating rate at its temperature location is approximately linear.

[0076] If the peak of the second-order mass loss derivative is located in the non-principal decomposition region (e.g., there is no obvious DTG peak between Tonset and Tmax), then it is speculated that the jump is induced by oxidation rather than a main chain breakage process.

[0077] At all time points, the local temperature rise acceleration anomalies and the peak values ​​of the second-order mass loss derivative are converted into comprehensive feature vectors. These comprehensive feature vectors are used as input to the machine learning model. The machine learning model uses the prediction of local abnormal pyrolysis behavior analysis value labels for each set of comprehensive feature vectors as the prediction objective and minimizes the sum of prediction errors for all local abnormal pyrolysis behavior analysis value labels as the training objective. The machine learning model is trained until the sum of prediction errors converges, at which point the model training stops. The local abnormal pyrolysis behavior analysis values ​​are determined based on the model output. The machine learning model is an isolated forest model.

[0078] The analysis value of local abnormal pyrolysis behavior is compared with the preset empirical value (the empirical value is the mean of the analysis values ​​of local abnormal pyrolysis behavior at all time points plus 3 times the standard deviation). When the analysis value of local abnormal pyrolysis behavior is greater than the preset empirical value, the time point is marked as the oxidation window abnormal trigger point. The corresponding temperature, image position and quality change rate are output in linkage for subsequent data elimination or correction reference.

[0079] In this embodiment, by combining the aforementioned micro-area photothermal probe imaging system with a thermogravimetric analyzer, temperature image sequences and TGA mass change curve data of PVC film-coated samples during thermal decomposition have been obtained. The system has identified multiple local temperature rise acceleration anomalies through a spatiotemporal residual network. These points may be affected by interferences such as the oxidation window effect, resulting in non-material intrinsic perturbation peaks in the mass curve.

[0080] To improve the accuracy of thermal decomposition data and the reliability of thermal stability assessment, this invention further introduces an optimization mechanism to locally model the influence range of outliers and correct or mark relevant data segments. This process includes the following steps:

[0081] First, the heating process recorded by the TGA device is divided into equally spaced time nodes (each node is 0.1 seconds). Each node contains attribute information such as sample temperature, remaining mass, mass loss rate, and thermal image frame number at that moment. At the same time, the system projects the local abnormal points (including spatial coordinates and timestamps) identified by the aforementioned photothermal probe onto the corresponding time nodes and marks them as "high-risk nodes".

[0082] Using time nodes as vertices, directed edges are constructed in chronological order to form a graph structure. Each node is also connected to adjacent thermal imaging regions, forming a "time-space coupling graph." This graph is used to represent the temporal dependence of the thermal-mass response behavior of materials during pyrolysis.

[0083] For each "outlier node" in the graph, the system uses a graph neural network (GNN) to model its potential impact on adjacent time nodes.

[0084] Specifically, a graph attention mechanism is employed to calculate perturbation propagation weights based on information such as the similarity between node attributes, the magnitude of changes in independent variables, and the intensity of outliers, and to predict the temporal path and range of the anomalous perturbation. The system marks time points identified as "anomalously sensitive nodes" as potential distortion regions.

[0085] For example, when the rate of temperature rise and the rate of mass loss of nodes before and after an anomaly point undergo a co-abrupt change, GNN identifies it as a jump segment that may be induced by oxidation.

[0086] For time segments identified as abnormally sensitive nodes, the system automatically selects one of the following three data processing methods based on the disturbance intensity and mode type:

[0087] Data interpolation correction: When the disturbance amplitude is small and the change is continuous, the system uses the normal segment data on both sides to perform local interpolation, replace the outlier values ​​and reconstruct the quality change curve;

[0088] Anomaly segment marking and handling: When the disturbance is of high amplitude and is not a physically reasonable jump, the system marks the segment in the data as "unusable for lifetime prediction" for subsequent analysis and exclusion;

[0089] Derivative curve compression and smoothing: For noise-type spikes or slight diffusion disturbances, the system uses an adaptive window Savitzky-Golay filter to smooth the DTG curve, preserving the overall trend while suppressing abnormal peaks.

[0090] The aforementioned correction mechanism can be applied to the TG curve alone, or it can affect the DTG and Tmax analysis sections in conjunction.

[0091] After data processing, the system re-extracts key thermal stability indicators of the PVC film coating, including:

[0092] Initial decomposition temperature: Eliminate false starting points induced by abnormal temperature rise in the early stage and re-identify the true pyrolysis starting point;

[0093] Maximum weight loss rate temperature: to exclude interference from secondary peaks induced by non-main chain reactions;

[0094] Total mass loss rate and oxidation-induced temperature shift: reflect the degree of material aging and oxidation resistance;

[0095] Pyrolysis response curve integrity score: Quantify data reliability through residual score.

[0096] Meanwhile, the system generates comparative charts that overlay the original thermal decomposition curve, the corrected curve, and the abnormal segment markers to allow researchers to assess the reliability and usability of the aging data.

[0097] This method combines photothermal image anomaly recognition with graph neural network perturbation analysis mechanism, and for the first time realizes the establishment of a dynamic perturbation propagation model in thermogravimetric analysis data, which is adaptable to various PVC film coating thicknesses, structures and aging forms.

[0098] The final output of this method includes: optimized TG and DTG curves (including correction point information); a complete aging performance analysis report; an abnormal intervention index map (including disturbance propagation trajectory and intensity map); and parameter templates and model weights available for batch sample analysis.

[0099] For example, a PVC film sample was identified with localized hot spots in the 300–320°C range, and its DTG curve showed a sudden secondary peak deviating from the main decomposition region. This invention determined that this data segment was a localized jump caused by the "oxidation window effect." After performing a slight compression operation on this curve segment and recalculating the tonset, the sample's thermal stability evaluation results improved by approximately 8% compared to the uncorrected result, significantly enhancing the accuracy and consistency of the analytical results.

[0100] Example 2: To verify the effectiveness of the present invention in the aging performance testing of PVC film coatings, the applicant constructed a comparative experiment including the original analysis group and the treatment group of the present invention. Thermogravimetric analysis (TGA) and photothermal probe imaging were used to detect and analyze the thermal decomposition behavior of PVC film coatings in a high-temperature aging environment, and the key indicators of the two groups were compared to verify its beneficial effects.

[0101] Material: Polyvinyl chloride (PVC) base film, approximately 80 μm thick, with a fluorine-modified acrylic transparent coating (approximately 12 μm thick) on the surface.

[0102] Pretreatment: The sample was dried in a vacuum drying oven at 60°C for 12 hours to remove residual moisture;

[0103] Cutting: Cut into 5 mm × 5 mm pieces for later use.

[0104] Thermogravimetric analyzer: Netzsch TG 209F3;

[0105] Photothermal probe imaging system: integrated FLIR Lepton thermal imaging probe (160×120 pixels, 9 Hz), mounted on top of the TGA cavity;

[0106] Temperature program: 30°C → 700°C, heating rate 10°C / min;

[0107] Atmosphere settings include:

[0108] The experimental group was vented with air (to simulate an oxidation environment);

[0109] The control group was purged with high-purity nitrogen gas (simulating an oxygen-free environment).

[0110] Synchronization control: The infrared thermal image and TGA data acquisition timestamp are synchronized with each other, with a period of 0.1 s.

[0111] Simultaneously acquire TGA mass change data and sample surface temperature image sequences;

[0112] The spatiotemporal residual network algorithm of this invention is used to extract temperature rise acceleration anomalies;

[0113] Construct a time-space graph structure and use the T-GPS algorithm to identify data segments in the quality curve that are affected by disturbances;

[0114] Implement corrective strategies for the affected areas and recalculate key thermal decomposition indicators;

[0115] Compare horizontally with the unprocessed raw data.

[0116] Indicator Categories Raw data Corrected data (method of this invention) difference Technical significance Initial decomposition temperature (Tonset) (°C) 266.4°C 276.1°C ↑ 9.7°C Eliminating the illusion of premature decomposition Maximum rate of weightlessness Tmax (°C) 310.2°C 312.8°C ↑ 2.6°C Eliminate abnormal peak interference Secondary peak identification (302.6°C) disappear 0 Confirmed as localized oxidation error Total weight loss (%) 71.4% 70.8% –0.6% Precision improvement Data signal-to-noise ratio (SNR) 17.3 dB 21.9 dB ↑26.5% The decomposition curve is smoother. Consistency of evaluation results (3 replicates) Deviation 6.8% Deviation 2.1% ↓ 4.7% Repetitive enhancement

[0117] Comparison of test data and results

[0118] The experiments described above demonstrate that the thermal anomaly identification and graph neural perturbation correction technology proposed in this invention can effectively identify local quality jumps caused by the "oxidation window effect" in the thermogravimetric analysis of PVC film coatings and perform targeted corrections on the abnormal segments. It shows significant deviation compression effects on the two core thermal stability parameters, Tonset and Tmax, improving the signal-to-noise ratio of pyrolysis data by over 25% and reducing repeatability deviation by nearly 70%. This indicates that the method has significant beneficial technical effects in practical applications, far superior to conventional TGA curve post-processing strategies (such as Gaussian filtering and moving average).

[0119] Meanwhile, the process demonstrates good adaptability and versatility under different atmospheric conditions, proving that it can be widely applied to the fine testing of the aging performance of PVC and other polymer film materials, and is especially suitable for the precise evaluation of thermal stability in oxygen-containing / variable atmosphere environments.

[0120] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for testing the aging performance of PVC film coatings based on thermal analysis, characterized in that: include: Provide a sample of the PVC film coating to be tested and perform a drying process; The sample is placed in a thermogravimetric analyzer for a temperature rise test, while the atmosphere is set to an inert gas or an oxygen-containing mixture. A micro-area photothermal probe imaging system is integrated into the sample chamber of the thermogravimetric analysis device to simultaneously acquire temperature distribution images of the sample surface; The imaging system has real-time temperature imaging capabilities and the ability to identify the temperature rise rate and thermal diffusion characteristics of sample micro-areas. The temperature image data and the TGA mass change curve are analyzed synchronously, and local temperature rise acceleration anomalies and second-order mass loss derivative peak values ​​are generated to comprehensively determine whether the oxidation window is abnormal. Based on the identified anomalies, the thermal decomposition data is corrected or marked, and the optimized aging performance analysis results of the PVC film coating are output.

2. The method for testing the aging performance of PVC film coatings based on thermal analysis according to claim 1, characterized in that: in, The integrated micro-area photothermal probe imaging system on the sample chamber of a thermogravimetric analysis device includes: An infrared transparent window is provided at the top or side wall of the sample chamber; A MEMS infrared thermal imaging probe with micro-focus modulation function is fixedly installed outside the window to capture a two-dimensional temperature distribution image of the sample surface. The probe is synchronized with the heating program of the thermogravimetric analysis device through the control module to achieve time matching between thermal image acquisition and temperature-mass data. The data collected by the probe is transmitted to the central processing system in real time to construct a thermal imaging sequence.

3. The method for testing the aging performance of PVC film coatings based on thermal analysis according to claim 1, characterized in that: Synchronous analysis of temperature image data with TGA quality change curves includes: The method for generating local temperature rise acceleration anomalies is as follows: stack consecutive thermal image frames to form a three-dimensional data structure, including pixel position (x,y) and time index t; for each frame, take a window Ωr(x,y) with radius r around each pixel (x,y), calculate the average temperature of the window, and record it as the local average temperature; subtract the current temperature of each pixel from its local average temperature to form a residual mapping; perform a second-order time difference on the residual sequence of each pixel to calculate the local temperature rise acceleration anomaly.

4. The method for testing the aging performance of PVC film coatings based on thermal analysis according to claim 3, characterized in that: The method for generating the peak value of the second-order mass loss derivative is as follows: Let the change in sample mass over time be m(t), then its first derivative is... For thermal decomposition rate, the second derivative This indicates the trend of change in the thermal decomposition rate, i.e., acceleration; The five-point central difference method is used to calculate... The values ​​throughout the testing process form a second derivative curve. Local extreme points are extracted from this curve, and the peak value of the second-order mass loss derivative is defined. for: .

5. The method for testing the aging performance of PVC film coatings based on thermal analysis according to claim 4, characterized in that: At all time points, the local temperature rise acceleration anomalies and the peak values ​​of the second-order mass loss derivative are converted into comprehensive feature vectors. These comprehensive feature vectors are used as input to the machine learning model. The machine learning model uses the prediction of local abnormal pyrolysis behavior analysis value labels for each set of comprehensive feature vectors as the prediction objective and minimizes the sum of prediction errors for all local abnormal pyrolysis behavior analysis value labels as the training objective. The machine learning model is trained until the sum of prediction errors converges, at which point the model training stops. The local abnormal pyrolysis behavior analysis values ​​are determined based on the model output. The machine learning model is an isolated forest model.

6. The method for testing the aging performance of PVC film coatings based on thermal analysis according to claim 5, characterized in that: The analysis value of local abnormal pyrolysis behavior is compared with the preset experience value. When the analysis value of local abnormal pyrolysis behavior is greater than the preset experience value, the time point is marked as the oxidation window abnormal trigger point.

7. The method for testing the aging performance of PVC film coatings based on thermal analysis according to claim 1, characterized in that: The thermal decomposition data corrected based on identified outlier locations includes: Construct a time-space data plot and map outliers to the corresponding time nodes in the TGA curve; The temporal graph neural perturbation suppression algorithm is used to evaluate the impact range of outliers on subsequent nodes; For the data in the affected segments, select interpolation correction, curve compression, or anomaly marking strategies; Based on the corrected curve, the aging performance parameters are re-extracted and the optimized results are output.

Images