TFT-LCD residual life probability prediction method fusing physical characteristics and Gaussian process regression
By integrating physical features with Gaussian process regression, a multi-source feature dataset was constructed and the model parameters were updated online. This solved the problems of individual differences and lack of physical mechanisms in TFT-LCD lifetime prediction, realizing personalized and probabilistic lifetime prediction of devices and improving the interpretability and credibility of the prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTHWEST TECHNICAL ENGINEERING RESEARCH INSTITUTE OF CHINA SOUTH IND GROUP
- Filing Date
- 2025-12-04
- Publication Date
- 2026-04-28
AI Technical Summary
Existing technologies for TFT-LCD lifetime prediction ignore individual device differences and the randomness of usage conditions, lack physical mechanism support, resulting in insufficient interpretability and engineering credibility of prediction results, and failing to provide uncertainty measurement and highly reliable maintenance decisions.
This paper proposes a method that integrates physical features and Gaussian process regression. By constructing a multi-source feature dataset, a Gaussian process regression model is established. Variational Bayesian inference is used to achieve online updating of model parameters. Furthermore, Monte Carlo simulation is used to generate the probability density function of remaining lifetime, thus establishing a personalized and probabilistic prediction method for devices.
It enables personalized and probabilistic prediction of the remaining lifespan of TFT-LCDs, enhances the interpretability and generalization ability of the model, provides a scientific basis for equipment maintenance decisions, and improves the robustness and reliability of the prediction.
Smart Images
Figure CN121935879A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electronic device lifetime prediction and safety management technology. Specifically, it relates to a method for predicting the remaining lifetime probability of TFT-LCDs by integrating physical characteristics and Gaussian process regression. Background Technology
[0002] Lifetime prediction for optoelectronic devices (such as TFT-LCD, OLED, and micro-LED) is crucial for reliability management, preventative maintenance, and spare parts planning in electronic equipment. Among these, TFT-LCD, as a widely used flat panel display technology, directly impacts the lifespan of end products and user experience through its long-term performance. Therefore, achieving accurate and reliable prediction of its remaining lifespan is a key challenge for improving the reliability of display systems. Traditional lifetime prediction methods primarily rely on statistical extrapolation from accelerated life testing data, which has significant limitations: firstly, it ignores individual device differences and the randomness of actual operating conditions; secondly, it cannot provide a measure of the uncertainty of the prediction results; and finally, it lacks physical mechanism support, resulting in low prediction reliability in small sample sizes.
[0003] In recent years, machine learning-based prediction methods have been introduced to address nonlinear problems of performance degradation, but most of them rely on pure data-driven approaches and lack an understanding of the physical processes of device failure, resulting in insufficient interpretability and engineering credibility of the prediction results.
[0004] Specifically, existing methods have failed to establish an explicit correlation between performance degradation and internal physical damage characteristics, and the prediction models are mostly deterministic, which cannot quantify the prediction risks and are difficult to support highly reliable condition-based maintenance decisions. Summary of the Invention
[0005] The purpose of this invention is to overcome the shortcomings and deficiencies of the existing technology and provide a TFT-LCD remaining lifetime probability prediction method that integrates physical characteristics and Gaussian process regression. This method effectively realizes personalized and probabilistic prediction of the remaining lifetime of TFT-LCD, providing a scientific basis for equipment maintenance decisions.
[0006] The objective of this invention is achieved through the following technical solution: a TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression, comprising the following steps:
[0007] S1. Construct a multi-source feature dataset: obtain performance monitoring data through laboratory accelerated aging experiments and natural environment exposure experiments, extract structural response feature data through transient thermo-mechanical coupling simulation, and construct a multi-source feature dataset from the performance monitoring data, structural response feature data, and environmental stress data.
[0008] S2. Establish a Gaussian process regression degradation model: Construct a Gaussian process regression model with time and environmental stress as inputs and performance degradation as output, and initialize the hyperparameters of the model using the multi-source feature dataset.
[0009] S3. Model Training and Parameter Update: Variational Bayesian inference is used to achieve online updating of hyperparameters and uncertainty quantification of the Gaussian process regression degradation model;
[0010] S4. Performance degradation trajectory prediction: Based on the updated Gaussian process regression degradation model, the predicted distribution of performance degradation values at future time points is obtained;
[0011] S5. Remaining life probability assessment: Based on the predicted distribution, the probability density function of the remaining life and the prediction intervals at different confidence levels are generated through Monte Carlo simulation.
[0012] Preferably, in step S1, the process of extracting structural response feature data through transient thermo-mechanical coupling simulation specifically includes:
[0013] S11. Establish a three-dimensional finite element model of the TFT-LCD module in the COMSOL multiphysics simulation platform. The model includes the backlight module, light guide plate, glass substrate and key interfaces.
[0014] S12. Apply corresponding thermal loads and humidity diffusion boundary conditions to the surface of the finite element model based on the temperature and humidity history in the environmental stress data.
[0015] S13. Perform transient thermo-mechanical coupling simulations corresponding to accelerated aging or natural exposure experiments to obtain the structural response field;
[0016] S14. Extract structural response feature data from the structural response field. The structural response feature data includes the stress concentration coefficient of the LED solder joints of the backlight module, the peak value of the equivalent stress of the light guide plate frame, and the interface peeling energy release rate of the glass substrate edge.
[0017] Preferably, in step S1, the performance monitoring data includes brightness, uniformity, and chromaticity parameters, and the environmental stress data includes temperature, humidity, and time-series changes.
[0018] Preferably, in step S2, the Gaussian process regression degradation model is defined as a Gaussian process over the performance degradation function:
[0019] f~GP(m(·),k(·,·)),
[0020] Among them, the mean function m(·) adopts a nonlinear function that considers the cumulative effect of damage, and the covariance function k(·,·) adopts a composite form of the Matérn 3 / 2 kernel function and the white noise kernel function.
[0021] Preferably, in step S3, the variational Bayesian inference includes:
[0022] Based on new performance monitoring data, a variational inference method is used to update the posterior distribution of the model hyperparameter θ. Specifically, the variational lower bound is iteratively optimized to approximate the posterior probability distribution of the hyperparameter under all given performance monitoring data, thereby realizing the quantification of parameter updates and uncertainties.
[0023] Preferably, in step S4, the predicted distribution is a Gaussian distribution:
[0024] p(f * |X * ,D)=N(μ * ,∑ * ),
[0025] Where p represents the prediction distribution, N represents the Gaussian distribution, D represents the training dataset, and X... * f represents the test input. * This represents the predicted degradation output, μ. * To predict the mean, ∑ * To predict the covariance matrix, * represents any input index.
[0026] Preferably, step S5 includes:
[0027] S51. Extract multiple sets of trajectory samples from the predicted distribution;
[0028] S52. Calculate the time when each trajectory first reaches the preset failure threshold to obtain a time sample set;
[0029] S53. Based on the time sample set, calculate the probability distribution and confidence interval of the remaining lifespan.
[0030] Preferably, the expression for the Matérn 3 / 2 kernel function is:
[0031]
[0032] Where, k M(r) This represents the Matérn 3 / 2 covariance function value with distance r as input; σ 2 f Let be the signal variance; r represent the Euclidean distance between the two input vectors; l is the length scale; and the final form of the composite covariance function is k(x). i ,x j )=k M (x i ,x j )+σ 2 n δ ij, where k(x i ,x j ) represents the input vector x i With x j The total covariance between them; σ 2 n To observe the noise variance; δ ij This refers to the Kronecker delta function.
[0033] Preferably, the preset failure threshold is set to 40% to 70% of the initial brightness value.
[0034] Preferably, the test conditions for the laboratory accelerated aging experiment include combinations of different temperatures and different humidity levels, and the natural environment exposure experiment is conducted in a typical hot and humid climate environment.
[0035] The present invention has the following advantages and effects compared with the prior art:
[0036] (1) This invention provides a method for predicting the remaining lifetime of TFT-LCDs by integrating physical features and Gaussian process regression. The method first constructs a multi-source feature dataset, and then establishes a Gaussian process regression degradation model. The dataset incorporates structural response features extracted from transient thermo-mechanical coupling simulation, establishes an explicit correlation between macroscopic performance degradation and microscopic physical damage, enhances the interpretability of the model, and overcomes the defect of weak generalization ability of pure data-driven models when training data is limited. Then, variational Bayesian inference is used to realize the online update of model parameters, which can integrate monitoring data of individual devices to realize personalized remaining lifetime prediction. Finally, the probability density function and confidence interval of the remaining lifetime are generated through Monte Carlo simulation, which effectively realizes the probabilistic prediction of the remaining lifetime of TFT-LCDs, enhances the generalization ability of the model, provides a scientific basis for equipment maintenance decisions, and has important engineering application value.
[0037] (2) The Gaussian process regression degradation model constructed in this invention can simultaneously process multi-source information such as laboratory accelerated aging experimental data, natural environment exposure experimental data and field monitoring data, which solves the problem of traditional methods relying on a single data source and enhances the robustness of prediction conclusions. Attached Figure Description
[0038] Figure 1 This is a flowchart illustrating the TFT-LCD remaining lifetime probability prediction method that integrates physical characteristics and Gaussian process regression according to the present invention.
[0039] Figure 2 The diagram shows the Gaussian process regression model of the TFT-LCD device in different regions in the embodiment.
[0040] Figure 3This is a prediction diagram of the regional brightness degradation trajectory of the TFT-LCD device in the embodiment. Detailed Implementation
[0041] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.
[0042] like Figure 1 The diagram shows a flowchart of a TFT-LCD remaining lifetime probability prediction method that integrates physical characteristics and Gaussian process regression according to the present invention, including the following steps:
[0043] Step S1: Construct a multi-source feature dataset.
[0044] This step aims to construct a feature dataset that integrates multi-dimensional information, providing a data foundation for subsequent probabilistic modeling. Device performance degradation is the result of the combined effects of time, environmental stress, and internal physical state. This step obtains performance time-series data through accelerated aging experiments in the laboratory and natural environment exposure experiments to reflect the degradation results. Simultaneously, structural response feature data is extracted through transient thermo-mechanical coupling simulation to quantitatively characterize the internal damage state (i.e., degradation mechanism). The performance monitoring data (including brightness, uniformity, and chromaticity parameters), structural response feature data, and environmental stress data (including temperature, humidity, and time-series changes) constitute a multi-source feature dataset.
[0045] The laboratory accelerated aging tests included various combinations of temperature and humidity, such as 85℃ / 85%RH, 60℃ / 85%RH, and 85℃ / 30%RH, with test cycles covering multiple time points including 0h, 24h, 48h, 72h, and 96h. The natural environment exposure tests were conducted at a test station in a typical hot and humid environment region, with samples periodically sent back to the laboratory for performance testing and failure analysis.
[0046] The specific process of extracting structural response feature data through transient thermo-mechanical coupling simulation includes:
[0047] S11. Establish a three-dimensional finite element model of the TFT-LCD module in the COMSOL multiphysics simulation platform. The model includes the backlight module, light guide plate, glass substrate and key interfaces.
[0048] S12. Apply corresponding thermal loads and humidity diffusion boundary conditions to the surface of the finite element model based on the temperature and humidity history in the environmental stress data.
[0049] S13. Perform transient thermo-mechanical coupling simulations corresponding to accelerated aging or natural exposure experiments to obtain the structural response field;
[0050] S14. Extract structural response feature data from the structural response field. The structural response feature data includes the stress concentration coefficient of the LED solder joints of the backlight module, the peak value of the equivalent stress of the light guide plate frame, and the interface peeling energy release rate of the glass substrate edge.
[0051] Step S2: Establish a Gaussian process regression degradation model.
[0052] Specifically, this step aims to establish a non-parametric probabilistic model that can describe the uncertainty of the degradation process. It constructs a Gaussian process regression model with time and environmental stress as inputs and performance degradation as output. Utilizing the functional distribution characteristics of the Gaussian process, it expresses the stochastic process of performance degradation with time, stress, and other input variables. The Gaussian process regression degradation model is defined by a mean function and a covariance function. The mean function describes the overall trend of degradation, while the covariance function describes the smoothness, volatility, and observation noise of the degradation trajectory. This model learns a probability distribution from the data, providing a natural mathematical framework for subsequent Bayesian updates and probabilistic predictions.
[0053] Specifically, the Gaussian process regression degradation model is defined as a Gaussian process over the performance degradation function f:
[0054] f~GP(m(·),k(·,·)),
[0055] The mean function m(·) is a nonlinear function that considers the cumulative effect of damage, and the covariance function k(·,·) is a composite form of the Matérn 3 / 2 kernel function and the white noise kernel function. The expression for the Matérn 3 / 2 kernel function is as follows:
[0056]
[0057] Where, k M(r) This represents the Matérn 3 / 2 covariance function value with distance r as input; σ 2 f The signal variance is used to control the output scale of the function; r represents the Euclidean distance between the two input vectors, r = |x i -x j |;l represents the length scale, and the final form of the composite covariance function is k(x) i ,x j )=k M (x i ,x j )+σ 2 n δ ij , where k(x i ,x j ) represents the input vector x i With x jThe total covariance between them; σ 2 n The variance of the observed noise is used to characterize the level of random noise in the measurement data; δ ij This is the Kronecker delta function, which has a value of 1 when i = j and 0 otherwise. This model has good characterization ability for degraded data under different regions or conditions, such as... Figure 2 As shown, Lv represents brightness, time represents time, and Location represents the region.
[0058] Step S3: Model training and parameter update.
[0059] Specifically, this step aims to enable the model to dynamically adapt to the characteristics of individual devices and quantify the uncertainty of the model's own parameters. Variational Bayesian inference is employed to achieve online updating and uncertainty quantification of hyperparameters in a Gaussian process regression degradation model. Specifically, when new field monitoring data is obtained, the method treats model hyperparameters (such as the signal variance of the kernel function and the length scale) as random variables, and iteratively optimizes the posterior probability distribution of these parameters by maximizing the lower bound of evidence. Through iterative updates using the coordinate ascent method, the posterior distribution of the hyperparameters is finally obtained, realizing online learning of model parameters and quantitative description of uncertainty, laying the foundation for the credibility assessment of prediction results.
[0060] Step S4: Performance degradation trajectory prediction.
[0061] Based on the updated Gaussian process regression degradation model with parametric posterior distribution, this step makes a probabilistic prediction of performance degradation under future conditions, taking advantage of the conditional distribution properties of Gaussian processes.
[0062] For any set of future input conditions (future time point X) * The model provides the corresponding performance degradation value f. * The complete prediction distribution p(f) * |X * The distribution, D), is also a Gaussian distribution, i.e., p(f * |X * ,D)=N(μ * ,Σ * ), where D represents the training dataset, containing input and output observations of the historical degradation process, and X * f represents the test input, corresponding to the time point or degradation stage to be predicted. * This indicates that the test input is X. * The corresponding degradation prediction output, μ * To predict the mean, Σ represents the expected trend of the degradation trajectory. *To predict the covariance matrix, we quantify the prediction uncertainty and generate a specified confidence interval. * represents any input index. Based on this prediction distribution, the brightness degradation trajectory prediction is as follows: Figure 3 As shown.
[0063] Step S5: Remaining life probability assessment.
[0064] Specifically, this step transforms probabilistic performance prediction into probabilistic lifetime prediction. Based on the predicted distribution, a probability density function of the remaining lifetime and prediction intervals at different confidence levels are generated through Monte Carlo simulation, specifically including:
[0065] First, from the predicted distribution obtained in step S4, a large number of trajectory samples (e.g., N = 1000) are randomly selected. Then, for each sample trajectory, the time t at which it first reaches the preset failure threshold T is calculated. The failure threshold is set as a specific percentage of the initial brightness value, such as 40% to 70% of the initial brightness value, according to the TFT-LCD product standard or actual application requirements. The distribution characteristics of these time values are statistically analyzed, including the mean, standard deviation, and quantiles. Finally, the N first failure times {t1, t2, ..., t...} are statistically analyzed. N By analyzing the distribution of}, the probability density function of the remaining lifetime and the prediction intervals at different confidence levels (such as the 95% confidence interval) can be obtained. This method completely transmits and transforms the uncertainty of performance prediction into the uncertainty of remaining lifetime prediction.
[0066] The remaining lifetime probability prediction method of the present invention is implemented below using a 7-inch TFT-LCD module as an example.
[0067] S1. Construction of Multi-Source Feature Dataset: Accelerated aging data from the laboratory under 85℃ / 85%RH conditions for 96 hours (brightness and uniformity tested every 24 hours) and natural environmental exposure data from a certain experimental station for 6 months (performance parameters tested monthly) were collected. Simultaneously, physical features reflecting the internal damage state of the device were extracted using COMSOL transient thermo-mechanical coupling simulation, including the stress concentration factor of the LED solder joint (2.8), the peak equivalent stress of the light guide plate frame (45MPa), and the release rate of the peel energy at the glass substrate edge (0.8J / m²). 2 The input vector X is constructed by combining time, environmental stress (temperature, humidity) and the aforementioned physical characteristics, and the normalized brightness decay rate is used as the output target y to build a training dataset.
[0068] S2. Establishment of Gaussian Process Regression Model: A Gaussian process regression model is established, with its covariance function using a composite form of the Matérn 3 / 2 kernel function and a white noise kernel function; the hyperparameters θ0 are initialized as follows: σ 2 f =1.0, l=1.0, σ 2n =0.01}.
[0069] S3. Model Training and Parameter Update: A variational Bayesian inference method was employed. Based on the constructed dataset D, iterative optimization was performed 100 times to update the posterior distribution of the hyperparameters. Model validation showed that the predictions trained on the first 80% of the data covered 85% of the actual observations with a 95% confidence interval for the remaining 20% of the test data. The model's coefficient of determination R0 was [value missing]. 2 The value reached 0.92, with a root mean square error of 2.3%.
[0070] S4. Remaining Lifetime Probability Assessment: In the remaining lifetime assessment phase, a brightness failure threshold is set to 70% of the initial value. 1000 future performance degradation trajectories are generated from a trained Gaussian process posterior distribution using Monte Carlo simulation. The time it takes for each trajectory to first cross the failure threshold is calculated, resulting in the remaining lifetime sample set {RL}. i Let i = 1, 2, ..., 1000, and RL represent the degradation time. Statistical analysis of this sample set yields an average remaining lifetime of 1250 hours, with a 90% confidence interval of [980, 1520] hours.
[0071] In the final application recommendations, preventative maintenance plans are developed based on the prediction results (e.g., replacing components 200 hours before the predicted mean time to failure), significantly improving system reliability and avoiding sudden failures. This embodiment verifies that the present invention, through a deep integration of physical mechanisms and data-driven approaches, overcomes the shortcomings of traditional methods in terms of insufficient prediction accuracy under small sample conditions. It can provide individualized lifetime predictions with confidence intervals, offering a reliable technical means for the maintenance and quality monitoring management of TFT-LCDs.
[0072] Furthermore, the multi-source physical feature fusion probabilistic prediction method proposed in this invention is modular. The physical features can originate from dedicated simulation models of different devices (such as thin-film stress in OLEDs and thermal resistance in Mini LEDs), and the performance degradation indices can be adjusted accordingly. Therefore, this method can be effectively applied to the remaining lifetime prediction of devices such as OLEDs and silicon-based optoelectronics, providing a powerful probabilistic prediction tool for the reliability management of a wide range of optoelectronic devices.
[0073] The above embodiments are preferred embodiments of the present invention and are not intended to limit the present invention. Any changes or other equivalent substitutions made without departing from the technical solution of the present invention are included within the protection scope of the present invention.
Claims
1. A method for predicting the remaining lifetime probability of TFT-LCDs by integrating physical characteristics and Gaussian process regression, characterized in that, Including the following steps: S1. Construct a multi-source feature dataset: obtain performance monitoring data through laboratory accelerated aging experiments and natural environment exposure experiments, extract structural response feature data through transient thermo-mechanical coupling simulation, and construct a multi-source feature dataset from the performance monitoring data, structural response feature data, and environmental stress data. S2. Establish a Gaussian process regression degradation model: Construct a Gaussian process regression model with time and environmental stress as inputs and performance degradation as output, and initialize the hyperparameters of the model using the multi-source feature dataset. S3. Model Training and Parameter Update: Variational Bayesian inference is used to achieve online updating of hyperparameters and uncertainty quantification of the Gaussian process regression degradation model; S4. Performance degradation trajectory prediction: Based on the updated Gaussian process regression degradation model, the predicted distribution of performance degradation values at future time points is obtained; S5. Remaining life probability assessment: Based on the predicted distribution, the probability density function of the remaining life and the prediction intervals at different confidence levels are generated through Monte Carlo simulation.
2. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, In step S1, the specific implementation process of extracting structural response feature data through transient thermo-mechanical coupling simulation includes: S11. Establish a three-dimensional finite element model of the TFT-LCD module in the COMSOL multiphysics simulation platform. The model includes the backlight module, light guide plate, glass substrate and key interfaces. S12. Apply corresponding thermal loads and humidity diffusion boundary conditions to the surface of the finite element model based on the temperature and humidity history in the environmental stress data. S13. Perform transient thermo-mechanical coupling simulations corresponding to accelerated aging or natural exposure experiments to obtain the structural response field; S14. Extract structural response feature data from the structural response field. The structural response feature data includes the stress concentration coefficient of the LED solder joints of the backlight module, the peak value of the equivalent stress of the light guide plate frame, and the interface peeling energy release rate of the glass substrate edge.
3. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, In step S1, the performance monitoring data includes brightness, uniformity, and chromaticity parameters, and the environmental stress data includes temperature, humidity, and time series change history.
4. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, In step S2, the Gaussian process regression degradation model is defined as a Gaussian process over the performance degradation function f: f~GP(m(·),k(·,·)), Among them, the mean function m(·) adopts a nonlinear function that considers the cumulative effect of damage, and the covariance function k(·,·) adopts a composite form of the Matérn 3 / 2 kernel function and the white noise kernel function.
5. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, In step S3, the variational Bayesian inference includes: updating the posterior distribution of the model hyperparameters based on new performance monitoring data using a variational inference method; specifically, iteratively optimizing the variational lower bound to approximate the posterior probability distribution of the hyperparameters given all performance monitoring data, thereby realizing parameter updates and quantification of uncertainty.
6. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, In step S4, the predicted distribution is a Gaussian distribution: p(f * |X * ,D)=N(μ * ,S * ), Where p represents the prediction distribution, N represents the Gaussian distribution, D represents the training dataset, and X... * f represents the test input. * This represents the predicted degradation output, μ. * To predict the mean, Σ * To predict the covariance matrix, * represents any input index.
7. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, Step S5 includes: S51. Extract multiple sets of trajectory samples from the predicted distribution; S52. Calculate the time when each trajectory first reaches the preset failure threshold to obtain a time sample set; S53. Based on the time sample set, calculate the probability distribution and confidence interval of the remaining lifespan.
8. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 4, characterized in that, The expression for the Matérn 3 / 2 kernel function is: Where, k M(r) This represents the Matérn 3 / 2 covariance function value with distance r as input; σ 2 f Let be the signal variance; r represent the Euclidean distance between the two input vectors; l is the length scale; and the final form of the composite covariance function is k(x). i ,x j )=k M (x i ,x j )+σ 2 n δ ij , where k(x i ,x j ) represents the input vector x i With x j The total covariance between them; σ 2 n To observe the noise variance; δ ij This refers to the Kronecker delta function.
9. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 7, characterized in that, The preset failure threshold is set to 40% to 70% of the initial brightness value.
10. The TFT-LCD remaining lifetime probability prediction method integrating physical characteristics and Gaussian process regression according to claim 1, characterized in that, The laboratory accelerated aging test conditions include combinations of different temperatures and humidity levels, and the natural environment exposure test is conducted in a typical hot and humid climate environment.