Device and method for testing thermal deformation of filter film

Abstract

The invention relates to the technical field of filter film testing, and discloses filter film thermal deformation testing equipment and a testing method. The method comprises the following steps: based on a test environment, carrying out a temperature cycle test on the filter film, and collecting a triple data set at the same time; performing image preprocessing and displacement field construction on the triple data set, and decomposing a displacement field into a linear combination of a standard deformation mode to obtain a thermal deformation response function; constructing a hierarchical probability graph model based on a thermal deformation response function, and extracting thermal deformation feature structural representation through variational inference; and performing deviation compensation on the structural representation of the thermal deformation characteristics to obtain environment self-adaptive thermal deformation characteristics, and calculating a thermal deformation reliability evaluation result comprising a thermal expansion coefficient matrix, a thermal deformation anisotropy index and thermal ripple strength. And comprehensive evaluation and reliability analysis of the thermal deformation characteristic of the filter film are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of filter film testing, and particularly to a filter film thermal deformation testing device and a testing method. Background Art

[0002] The filter film thermal deformation recognition technology has become an important physical layer testing method for evaluating the reliability of optical systems by utilizing the material property differences in the film layer structure. However, in the actual application environment, complex and variable testing conditions such as temperature gradients and humidity fluctuations seriously affect the accuracy and consistency of measurement results, resulting in significant deviations in the data collected under different environments and making it difficult to form a standardized evaluation system.

[0003] Traditional filter film thermal deformation testing methods aim to be carried out under constant environmental conditions and cannot effectively cope with environmental changes in actual application scenarios. Especially in extreme environments, the measurement accuracy will drop significantly. When the existing technology images and analyzes the deformation characteristics of the film layer, it is often limited to simple deformation modes with good continuity and recognizability, and lacks effective identification and quantification methods for phenomena such as non-uniform deformation, corrugated deformation, and thermal hysteresis in complex multi-layer filter films, making it difficult for the test results to truly reflect the performance characteristics of the filter film in the actual working environment. Summary of the Invention

[0004] The present invention provides a filter film thermal deformation testing device and a testing method, which can accurately capture the thermal deformation characteristics of the filter film and realize the comprehensive evaluation and reliability analysis of the thermal deformation characteristics of the filter film.

[0005] In a first aspect, the present invention provides a filter film thermal deformation testing method, which includes: Based on the test environment, perform a temperature cycle test on the filter film, and simultaneously collect a triple dataset including an environmental feature vector, a surface image of the filter film, and deformation displacement data; Perform image preprocessing and displacement field construction on the triple dataset, and decompose the displacement field into a linear combination of canonical deformation modes to obtain a thermal deformation response function; Based on the thermal deformation response function, construct a hierarchical probabilistic graphical model, and extract a structured representation of the thermal deformation characteristics through variational inference; Perform deviation compensation on the structured representation of the thermal deformation characteristics to obtain environment-adaptive thermal deformation characteristics, and calculate a thermal deformation reliability evaluation result including a thermal expansion coefficient matrix, a thermal deformation anisotropy index, and a thermal corrugation intensity.

[0006] In a second aspect, the present invention provides a filter film thermal deformation testing device, which includes: A temperature cycling test module is used to perform temperature cycling tests on the filter film based on the test environment, and simultaneously collect a triple dataset including environmental feature vectors, surface images of the filter film, and deformation displacement data; A decomposition module is used to perform image preprocessing and displacement field construction on the triple dataset, and decompose the displacement field into a linear combination of canonical deformation modes to obtain a thermal deformation response function; An extraction module is used to construct a hierarchical probabilistic graphical model based on the thermal deformation response function, and extract a structured representation of thermal deformation features through variational inference; A calculation module is used to perform bias compensation on the structured representation of thermal deformation features to obtain environment-adaptive thermal deformation features, and calculate a thermal deformation reliability evaluation result including a thermal expansion coefficient matrix, a thermal deformation anisotropy index, and a thermal ripple intensity.

[0007] In the technical solution provided by the present invention, by establishing a multi-environment parameter monitoring system and environmental feature vectors, and combining an environmental impact function and a compensation matrix, the present invention can perform accurate measurements under different temperature, humidity, and air pressure conditions, effectively reducing the measurement bias caused by environmental changes, and making the measurement results consistent and comparable in a changing environment. By using image preprocessing techniques such as adaptive histogram equalization, double-threshold segmentation, and sub-pixel edge detection, and combining the orthogonal polynomial decomposition method to represent the displacement field as a linear combination of canonical deformation modes, the thermal deformation characteristics of the filter film can be accurately captured, and the measurement accuracy reaches the nanometer level. By constructing a three-layer hierarchical probabilistic graphical model through the structural variational Bayesian method and integrating the generalized approximate message passing technique into variational inference, the one-dimensional continuity of in-plane deformation and the two-dimensional continuity of inter-layer deformation recognition are effectively improved, and the extraction ability of thermal deformation features of complex multi-layer filter films is enhanced. The environmental impact function implemented based on the weighted radial basis function network, combined with the compensation matrix constructed based on the deformation mode stability index, realizes intelligent compensation for the measurement bias caused by environmental changes, significantly improving the accuracy and repeatability of the measurement results. By constructing a characteristic index system including a thermal expansion coefficient matrix, a thermal deformation anisotropy index, a thermal ripple intensity, and a thermal deformation hysteresis index, and combining a multi-dimensional anomaly detection algorithm based on the modified Mahalanobis distance and a risk assessment model based on failure mode analysis, a comprehensive evaluation and reliability analysis of the thermal deformation characteristics of the filter film are realized. Description of the Drawings

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0009] Figure 1It is a schematic flowchart of the method for testing the thermal deformation of the filter film provided by the embodiment of the present application; Figure 2 It is a schematic block diagram of the structure of the device for testing the thermal deformation of the filter film provided by the embodiment of the present application. Detailed implementation manners

[0010] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0011] The flowchart shown in the accompanying drawings is only an example illustration, and does not necessarily include all contents and operations / steps, nor does it necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined, or partially merged. Therefore, the actual execution order may be changed based on the actual situation.

[0012] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification of the present application and the appended claims, unless otherwise clearly specified in the context, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0013] It should be further understood that the term " / and / " used in the specification of the present application and the appended claims refers to any combination and all possible combinations of one or more of the related listed items, and includes these combinations.

[0014] Next, some embodiments of the present application will be described in detail in conjunction with the accompanying drawings. Without conflict, the features in the following embodiments and the embodiments can be combined with each other.

[0015] Please refer to Figure 1 , Figure 1 It is a schematic flowchart of the method for testing the thermal deformation of the filter film provided by the embodiment of the present application. As Figure 1 shown, the method for testing the thermal deformation of the filter film provided by the embodiment of the present application includes steps S100 to S400.

[0016] Step S100: Based on the test environment, perform a temperature cycle test on the filter film, and simultaneously collect a triple data set including an environmental feature vector, a surface image of the filter film, and deformation displacement data; It can be understood that the execution entity of the present invention can be a filter film thermal deformation test device, or a terminal or a server, and no specific limitation is made here. In the embodiments of the present invention, the server is taken as an example of the execution entity for illustration.

[0017] Specifically, a multi-environment parameter monitoring system is constructed. The system includes a temperature sensor array, a humidity sensor array, a barometric pressure sensor, and a data acquisition unit with synchronous acquisition and processing capabilities. The temperature sensor array consists of nine thermocouples and is evenly distributed inside the test cavity in a 3×3 matrix structure. The temperature measurement accuracy of each sensor reaches ±0.1 °C, and the sampling frequency is 5 Hz; the humidity sensor array is composed of four capacitive sensors, which are respectively arranged at the four corners of the cavity, with an accuracy of ±0.5%RH and a sampling frequency of 2 Hz; the barometric pressure sensor adopts a piezoresistive design, with a measurement accuracy of ±0.1 kPa and a sampling frequency of 1 Hz. The data of all sensors are collected in real time through a data acquisition unit with a 32-bit processor as the core and an internal 16-bit analog-to-digital converter, and are processed through an internal digital filtering module. The filtering module uses an eighth-order Butterworth low-pass filter, and its cut-off frequency is set to one-fourth of the sampling frequency of each sensor to effectively suppress high-frequency noise and retain the effective signal components. After filtering, unit conversion and calibration are performed on the obtained environmental parameters. Among them, the temperature data is converted to Kelvin, the humidity data is converted to relative humidity percentage, and the barometric pressure data is converted to Pascal. After unit conversion, piecewise linear interpolation is used for calibration. This method constructs 10, 8, and 5 calibration points for temperature, humidity, and barometric pressure respectively to ensure the linear accuracy of the data within the entire measurement range. Based on the calibrated standard physical quantity data above, the spatial distribution characteristics inside the test cavity are calculated, including the temperature gradient vector , , and the humidity gradient vector , , , which are used to reflect the spatial non-uniformity of the environmental field; at the same time, the environmental stability index ESI is calculated. This index is defined as the ratio between the maximum fluctuation amount of each type of environmental parameter within 30 consecutive seconds and the corresponding measurement accuracy requirement, so as to be used to quantify the stability degree of the current test environment. When the environmental stability index ESI is less than the set target threshold (set to 0.5), it is determined that the environment is in a stable state, and the environmental feature vector at the current moment is automatically recorded. This vector includes the current temperature T, relative humidity RH, barometric pressure P, temperature spatial gradient , humidity spatial gradient and the environmental stability index ESI. On this basis, a preset temperature cycle test scheme is executed. This scheme includes multiple temperature control points, and typical settings include 25°C, 45°C, 65°C, 5°C, -15°C, etc., covering the full temperature range of the filter film during operation or storage. After reaching stability at each temperature point, that is, the temperature fluctuation is less than ±0.03°C and the stable duration exceeds 300 seconds, the system synchronously triggers the high-precision imaging module and the micro-displacement measurement unit to collect data. The imaging module consists of a scientific-grade CCD camera with a resolution of 2048×2048 pixels and a pixel size of 5.5μm and a high-precision macro lens. Its imaging resolution reaches 1μm / pixel and is used to obtain high-resolution images of the filter film surface; at the same time, the micro-displacement measurement unit consists of a laser interferometer and a nano-level piezoelectric displacement stage. The former has a resolution of 0.1nm, and the latter has a displacement control resolution of 1nm, which is used to accurately measure the surface micro-deformation of the filter film under the action of thermal load. During the entire test process, at each temperature point, 10 groups of image and displacement data are continuously collected, and the sampling interval is 30 seconds to reduce the influence of random disturbances and improve the statistical credibility of the data. At each temperature point, a triple dataset ST={ET, IT, DT} containing the environmental feature vector ET, the surface image IT, and the deformation displacement data DT is constructed. This dataset covers multiple temperature cycle points and is used for subsequent thermal deformation modeling and analysis.

[0018] Establish a filter film test environment that meets the experimental requirements. This test environment has basic characteristics such as high temperature control accuracy, strong spatial uniformity, and fast response speed. A precision temperature control device composed of a four-region independent PID control strategy is used as the core of temperature regulation. This device can achieve a temperature regulation accuracy of ±0.05 °C, and its heating and cooling rates can be flexibly adjusted in the range of 0.1 °C / min to 5 °C / min, and it has a temperature uniformity error performance better than 0.1 °C, thus ensuring the consistency and stability of the temperature field in the entire test chamber. At the same time, in this thermal deformation test system, an imaging system and a micro-displacement measurement unit are introduced. The imaging system consists of a scientific-grade CCD camera with 2048×2048 pixels, a pixel size of 5.5μm×5.5μm, a quantum efficiency greater than 95%, and a dynamic range exceeding 85dB. A macro lens with a focal length of 100mm is used in combination and the working distance is set to 150mm to achieve an image resolution of 1μm / pixel; while the micro-displacement measurement unit consists of a laser interferometer with nanometer resolution and a piezoelectric displacement stage working together. The former has a displacement resolution of 0.1nm and a measurement range of 10mm, and the latter has a control resolution of 1nm and a stroke range of 100μm, which can perform high-precision, nanometer-level real-time measurement of the small thermal deformation of the filter film. After the test system is built, the filter film sample to be tested is installed on a special sample bracket. This bracket is made of a material with a low coefficient of thermal expansion (such as quartz glass, whose coefficient of thermal expansion α < 5.5×10 -7( / °C) to avoid interference of thermal deformation of the bracket itself on the measurement results. To ensure the repeatability of measurement and the imaging clarity, the five-point positioning method is used to accurately install the sample, so as to ensure that the effective surface of the filter film is strictly perpendicular to the optical axis of the imaging system, forming a stable and reliable test in-place state. In this test in-place state, according to the preset temperature cycle plan, the filter film is gradually heated and cooled through the temperature control device. The test temperature sequence needs to cover the full temperature range encountered in its actual application. For example, the set typical temperature nodes include multiple key points such as 25°C → 45°C → 65°C → 45°C → 25°C → 5°C → -15°C → 5°C → 25°C, etc. The setting of each temperature point combines the feedback mechanism of the temperature control device to continuously monitor the current temperature fluctuation situation. Only when the temperature fluctuation is less than ±0.03°C and remains stable for more than 300 seconds can it be confirmed that the current temperature point has entered a stable state. In the temperature stable state, the system automatically triggers the synchronous start of the imaging system and the micro-displacement measurement unit, and respectively collects the high-resolution surface image of the filter film and its corresponding thermal deformation displacement data at this temperature point. To ensure the stability of the measurement results and error control, at each temperature point, 10 groups of images and displacement data are continuously collected, and the sampling interval is 30 seconds, so as to effectively reduce the influence of accidental noise and system drift on the data quality. During the whole test process, the imaging system provides two-dimensional visualization information to reflect the surface deformation morphology of the filter film under thermal loading conditions, while the micro-displacement unit provides high-precision quantitative data for reconstructing the two-dimensional or three-dimensional displacement field to form a quantitative description of the thermal deformation. At the same time, while collecting the images and displacement data, the environmental feature vector ET corresponding to the current temperature point is automatically associated. This feature vector includes the current environmental temperature T, relative humidity RH, air pressure P, temperature gradient vector , humidity gradient vector and the environmental stability index ESI. This vector is recorded by the pre-set multi-environment parameter monitoring system under the stable determination conditions. Therefore, the finally constructed triple dataset form is ST={IT, DT, ET}, where IT is the surface image of the filter film at this temperature point, DT is the corresponding displacement data, and ET is the environmental feature information at the same moment.

[0019] Step S200: Perform image preprocessing and displacement field construction on the triple dataset, and decompose the displacement field into a linear combination of canonical deformation modes to obtain the thermal deformation response function; Specifically, the surface images of the filter film in the triple data set are processed to extract the structural feature information for the construction of the displacement field. To enhance the local contrast of the original image and suppress the interference caused by the uneven gray distribution of the image, the image is processed by adaptive histogram equalization. During this process, the contrast limit factor is set to 8.0, and the grid size used is 1 / 8 of the image size, so as to ensure that the enhanced image can maintain details while avoiding over-enhancement. After the enhancement process, the overall contrast of the image is improved, and the gray difference between the film layer boundary and the background area is significantly amplified, which is convenient for subsequent segmentation operations. The double-threshold segmentation operation is performed on the enhanced image to extract the boundary region features of the filter film. This operation separates the foreground and the background by setting a low threshold and a high threshold, where the low threshold is set to 0.8 times the average gray value of the image, and the high threshold is 1.2 times, forming a mask image containing the film layer boundary region. After obtaining the boundary region, to improve the continuity and geometric smoothness of the edge data, edge smoothing is performed on it. A 5×5 Gaussian kernel is used to perform convolution operation on the boundary image, and the standard deviation σ of the Gaussian kernel is set to 1.2, so as to filter out high-frequency noise and edge artifacts on the premise of ensuring the integrity of the boundary features. After the smoothing process, sub-pixel level edge detection is performed on the boundary image to extract fine edge features with a resolution of 0.1 pixel. This process uses an edge sub-pixel localization algorithm based on gradient interpolation or fitting to ensure that the accuracy of the subsequent displacement field construction reaches the micron or even sub-micron level. The high-precision edge features are phase-correlation registered with the reference image of the filter film collected at the reference temperature (usually room temperature 25°C). The phase-correlation method is an image registration algorithm based on Fourier transform, which can achieve sub-pixel level image alignment accuracy and has rotation and scaling invariance. During the registration process, the translation vector of the current image relative to the reference image is determined by calculating the phase-correlation matrix in the frequency domain and extracting the position of the main peak, so as to achieve precise registration between the two. The registration result is used as the spatial reference for the subsequent displacement field construction to detect the deformation information generated by the image under the action of thermal load. Based on this registration result, the thermal deformation displacement field Φ(x, y) of the filter film is constructed, where x and y are the surface coordinates of the filter film, and Φ(x, y) represents the two-dimensional displacement vector of this position relative to the reference state under thermal excitation conditions. To perform structural decomposition and quantitative analysis on this displacement field, the orthogonal polynomial decomposition method is used to represent it as a linear combination of a set of canonical deformation modes. This canonical deformation mode library contains several typical thermal response forms of the film layer, such as uniform expansion mode, non-uniform radial expansion mode, tangential shear mode, periodic ripple mode and multiple high-order distortion modes, etc. Each mode has orthogonality and physical interpretability, which helps to separate different types of deformation mechanisms. The complete deformation mode spectrum is obtained by solving each αi coefficient through the least squares fitting algorithm.After obtaining the above deformation mode spectrum and combining it with the environmental feature vectors in the corresponding triple dataset, that is, establishing the thermal deformation response function of the filter film under this specific environmental state. This function is used to describe the specific deformation mode composition of the thermal load response characteristics of the filter film under specific environmental conditions, and it is the most core function expression in the entire thermal deformation modeling process. This thermal deformation response function can reflect the causal relationship between different environmental states and the deformation of the filter film, and can provide a quantitative basis for subsequent Bayesian structure modeling, stability analysis, and environmental adaptability compensation. By interpolating and integrating the response functions at different temperature points and different environmental feature vectors, a global thermal deformation model across temperature intervals is constructed.

[0020] Step S300: Construct a hierarchical probabilistic graphical model based on the thermal deformation response function, and extract the structured representation of the thermal deformation features through variational inference; Specifically, based on the pre-built thermal deformation response function, analyze the key variables it contains, and organize them into a multi-level probability structure to form a hierarchical probabilistic graphical model with statistical modeling capabilities. Perform principal component analysis on the environmental feature vector in the thermal deformation response function to extract the most representative environmental parameter dimensions as the environmental condition layer parameter θE in the model. This parameter can compress the original high-dimensional environmental features, reduce modeling redundancy, and improve inference efficiency. At the same time, the deformation mode coefficients in the deformation mode spectrum of the thermal deformation response function form a complete set of deformation mode parameters as the deformation mode layer parameters of the model, which are used to describe the specific response intensities of different deformation modes under the current environment. The two-dimensional thermal deformation displacement field Φ(x, y) constructed by the filter film at different temperature points is used as the observation data layer, which reflects the dependence relationship between the real physical response and the upper-layer latent factors. Based on these three types of parameters, a hierarchical probabilistic graphical model is constructed. Among them, the environmental condition layer determines the deformation mode distribution, and the deformation mode layer further determines the distribution of the observed displacement field. The conditional dependence between the three is represented by directed edges to construct a Bayesian graph structure. Based on this graph structure, define the probability relationships between the layers. Given the environmental condition θE and the deformation mode parameter θM, define the conditional probability distribution P(Φ|θM, θE) of the observed displacement field. This distribution characterizes the possibility of the actual displacement response of the filter film under specific thermal environments and deformation mechanisms. Define the prior distribution P(θM|θE), which reflects the prior uncertainty of the deformation mode parameters under specific environmental constraints. Usually, a Gaussian distribution form is selected to balance analyticity and adaptability. At the same time, define the marginal distribution P(θE) of the environmental parameters, which is used to characterize the probability distribution of various environmental conditions in practical applications. Through this step, a complete probability model including the prior structure, the observation model, and the dependence relationship of latent variables is formed. Since the above posterior distribution P(θM, θE|Φ) is computationally complex and has no analytical form, the variational inference method is used for approximate solution. Introduce the variational distribution Q(θM, θE) ≈ P(θM, θE|Φ), and assume that it can be factorized as Q(θM)·Q(θE), thus significantly simplifying the inference complexity of the joint distribution. In the model initialization stage, both Q(θM) and Q(θE) are set as multivariate Gaussian distributions, with (μM, ΣM) and (μE, ΣE) as the initial mean and covariance matrices respectively. By minimizing the Kullback-Leibler divergence (KL divergence) between the variational distribution and the true posterior distribution as the optimization goal, the variational expectation-maximization algorithm is used to perform parameter updates. During this process, the mean and covariance parameters of the variational distribution are continuously iteratively optimized until the convergence condition is satisfied (such as the ELBO change amount is less than 10 -6), the optimized variational distribution Q*(θM,θE) after convergence is obtained. This distribution is an approximate expression of the original posterior, which is used to reconstruct the potential thermal deformation mechanism statistically. To improve the approximation quality and inference accuracy of the model, the generalized approximate message passing technique is integrated into the variational inference process. The joint distribution structure is represented by a factor graph, and local message update operations are performed on the graph structure to enhance the expression ability of the dependence relationship between parameters and the inference accuracy. In the optimized variational distribution Q*(θM,θE), modal analysis is performed on the parameter space of θM to extract a set of modal deformation parameter sets that dominate the deformation characteristics of the filter film. Based on this set of modal deformation parameters, a stability analysis index is introduced to measure the response fluctuation of each deformation mode under environmental perturbations. The specific calculation method is the norm of the derivative of ηk with respect to θE, that is , the lower the stability index, the less sensitive the deformation mode is to environmental changes, and the higher the structural stability and physical reliability. The extracted set of modal deformation parameters η and its corresponding set of stability indices SI together constitute the structured representation S={η,SI} of the thermal deformation characteristics of the filter film.

[0021] Step S400: Perform deviation compensation on the structured representation of the thermal deformation characteristics to obtain the environment-adaptive thermal deformation characteristics, and calculate the thermal deformation reliability evaluation results including the thermal expansion coefficient matrix, the thermal deformation anisotropy index, and the thermal ripple intensity.

[0022] Specifically, based on the structured representation of thermal deformation characteristics, that is, the combined information including the main modal deformation parameters and the stability indices of each mode, and combined with the deformation deviation data measured under multiple different environmental conditions, an environmental impact function is constructed to characterize the degree of influence of environmental changes on the measurement results. The establishment of this function depends on the construction and training of a weighted radial basis function network. In the network, the input is a six-dimensional environmental feature vector, and the output is the predicted deviation of each modal deformation parameter. The hidden layer of the network connects the input and output through multiple non-linear kernel functions, and by virtue of its non-linear fitting ability, captures the complex mapping relationship between environmental disturbances and measurement errors. The network is trained using an objective function with a regularization term to effectively suppress the risk of overfitting while ensuring prediction accuracy. After training, the environmental feature vector collected under the current test environment is input into the trained network model to obtain the measurement deviation that occurs under this environmental condition. In order to effectively correct the predicted deviation, an environmental compensation matrix is constructed according to the stability indices corresponding to each mode in the structured representation of thermal deformation characteristics. This compensation matrix assigns different compensation weights to different modes. Modes with high stability have stronger resistance to environmental disturbances and play a dominant role in the compensation process, while modes with poor stability are assigned lower weights to prevent the spread and amplification of errors. According to this weight system, the environmental-related deviation values are combined with the compensation matrix to perform vector-level correction processing on the original deformation characteristics, resulting in a set of deformation parameters after deviation elimination, that is, the corrected deformation characteristics. An environmental adaptability calibration curve set is constructed based on the environmental response historical data of the filter film. This curve set is established by fitting the response data of the filter film under multiple known environmental conditions. Each curve corresponds to a modal deformation parameter, and a mapping relationship is constructed with the environmental parameters as the independent variable and the deformation response as the dependent variable. To adapt to the non-linear characteristics of environmental behavior, the calibration curve is represented in the form of a piecewise polynomial, and piecewise boundaries are set at the key environmental state turning points, so that each segment of the curve has sufficient fitting ability and maintains a certain mathematical smoothness. Taking the currently corrected modal deformation characteristics as the input, and passing through these calibration curves in sequence for mapping, the output is the adaptive thermal deformation characteristics with high environmental robustness, thus effectively realizing the decoupling between the deformation characteristics and the test environment and enhancing its cross-environment stability. Calculate multiple thermal deformation performance indicators of the filter film under the current environmental conditions based on the environmentally adaptive thermal deformation characteristics. Construct a coefficient of thermal expansion matrix, which is used to characterize the thermal response intensity and shear characteristics of the filter film in different directions. The diagonal components represent the linear expansion behavior in the principal axis direction, while the non-diagonal components reflect the in-plane shear response caused by temperature changes; extract the thermal deformation anisotropy index, which is used to quantify the degree of expansion difference of the filter film in different directions. The closer the value is to zero, the better the isotropy of the film layer; calculate the thermal ripple intensity index, which is used to measure whether there are periodic structural undulations on the surface of the filter film under thermal excitation. This index is an important basis for judging the optical interference stability of the film layer.Summarize the above indicators in the current environment and compare their fluctuation ranges in historical data to conduct a complete assessment of the thermal deformation characteristics of the filter film. Considering factors such as its environmental adaptability, deformation stability, and abnormal fluctuation degree, form a reliability assessment result of thermal deformation. This result is used to determine whether the current filter film sample meets the system design requirements.

[0023] In the environmentally adaptive thermal deformation characteristics, clarify the physical meanings represented by each main mode. Especially for the linear expansion - like modes, the amplitude change directly reflects the thermal expansion degree of the filter film in different directions. Therefore, a two - dimensional coefficient of thermal expansion matrix is constructed using the response amplitudes of these mode parameters along the principal axis directions. This matrix can systematically describe the planar expansion behavior of the filter film under heating conditions and reveal whether there are phenomena such as stress concentration, anisotropic strain, or shear coupling in different directions, forming a basic analysis result of thermal expansion characteristics. Calculate the thermal deformation anisotropy index based on the coefficient of thermal expansion matrix. This index reflects whether the filter film has a direction - dependent structural feature by quantifying the differences in expansion capabilities in different directions in the matrix. If the value of this index is relatively high, it means that there is obvious asymmetric expansion of the film layer in different principal axis directions, which may lead to problems such as in - plane optical path deviation, structural warping, and even thickness non - uniformity. If the index is relatively low, it indicates that the material response is more isotropic, which is beneficial to maintaining the structural stability and wavefront consistency of the optical system. At the same time, process the identified ripple patterns in the environmentally adaptive thermal deformation characteristics, extract the response amplitudes of the ripple mode and the uniform expansion mode, and obtain the thermal ripple intensity index by calculating their ratio. This index can be used to evaluate whether the surface of the filter film is prone to generating periodic undulating structures under thermal excitation, and then judge whether its surface shape control ability meets the standard. If the ripple intensity is large, it indicates that there are problems such as local stress release, interface delamination, or material thickness fluctuation in the film layer structure, thus increasing the risk of optical distortion. On the basis of completing the above index calculations, considering the response consistency of the filter film during the temperature cycle, analyze the difference in modal responses between the heating path and the cooling path, and calculate the root - mean - square of the deformation values between them to obtain the thermal deformation hysteresis index. The larger this index is, the more obvious the hysteresis characteristics of the response behavior of the filter film during heating and cooling are, which are caused by reversible stress hysteresis, visco - elastic properties, or material phase changes in the internal structure, and special attention needs to be paid to its impact on system stability. Combine the above four types of indexes, namely thermal expansion characteristics, direction dependence, ripple intensity, and hysteresis index, to form a high - dimensional feature vector, and apply a multi - dimensional anomaly detection algorithm to evaluate the position of the current filter film sample in this index space to determine whether it deviates from the normal range in the historical sample set. This process is realized based on the statistical distribution distance metric. If a data point is too far from the central distribution in the index space, it is determined as an abnormal sample, and further attribution analysis of its abnormal cause is carried out. After completing the multi - dimensional evaluation of the thermal deformation behavior of the filter film, quantify its risk level into a reliability evaluation result, introduce the failure mode and effect analysis method, that is, the FMEA mechanism, to construct a thermal deformation reliability evaluation model of the filter film. In this model, define a set of typical failure modes during the thermal deformation of the filter film, such as uniform over - expansion, non - uniform expansion, strong corrugation, interface delamination, modal drift, or hysteresis asymmetry. Each failure mode corresponds to a different risk type.For each failure mode, evaluate its severity, occurrence probability, and detectability in sequence, and use the product of the three as the risk priority number to represent the impact degree of the failure on the overall system. After all risk priority numbers are calculated, obtain their average or total, and then compare it with the reference value under the full score system to obtain the comprehensive thermal deformation reliability index of the filter film. The higher this index is, the higher the thermal stability and use safety of the sample are under the current environment and historical comparison conditions. Combine the aforementioned anomaly detection results with the comprehensive index output by the reliability evaluation model to form a complete thermal deformation reliability evaluation result report.

[0024] In the embodiments of the present invention, by establishing a multi-environment parameter monitoring system and environmental characteristic vectors, and combining an environmental impact function and a compensation matrix, the present invention can perform accurate measurements under different temperature, humidity, and air pressure conditions, effectively reducing the measurement deviation caused by environmental changes, and making the measurement results consistent and comparable in a changing environment. By using image preprocessing techniques such as adaptive histogram equalization, double-threshold segmentation, and sub-pixel edge detection, and combining the orthogonal polynomial decomposition method to represent the displacement field as a linear combination of canonical deformation modes, the thermal deformation characteristics of the filter film can be accurately captured, and the measurement accuracy reaches the nanometer level. By constructing a three-layer hierarchical probabilistic graphical model using the structural variational Bayesian method and integrating the generalized approximate message passing technique into variational inference, the one-dimensional continuity of in-plane deformation and the two-dimensional continuity recognition of inter-layer deformation are effectively improved, and the ability to extract the thermal deformation characteristics of complex multi-layer filter films is enhanced. The environmental impact function implemented based on the weighted radial basis function network, combined with the compensation matrix constructed based on the deformation mode stability index, realizes the intelligent compensation of the measurement deviation caused by environmental changes, significantly improving the accuracy and repeatability of the measurement results. By constructing a characteristic index system including the thermal expansion coefficient matrix, thermal deformation anisotropy index, thermal ripple intensity, and thermal deformation hysteresis index, and combining the multi-dimensional anomaly detection algorithm based on the modified Mahalanobis distance and the risk assessment model based on failure mode analysis, the comprehensive evaluation and reliability analysis of the thermal deformation characteristics of the filter film are realized.

[0025] In a specific embodiment, the process of executing step S100 may specifically include the following steps: Construct a multi-environment parameter monitoring system using a temperature sensor array, a humidity sensor array, a barometric pressure sensor, and a data acquisition unit, and collect the original data set through the multi-environment parameter monitoring system; Perform digital filtering processing on the original data set to obtain the filtered environmental parameter data, and calibrate the filtered environmental parameter data to obtain the standard physical quantity data; Calculate the spatial distribution parameters of the test environment based on the standard physical quantity data to obtain the environmental spatial distribution characteristics, and calculate the ratio of the maximum fluctuation amount of the standard physical quantity data to the measurement accuracy requirement based on the environmental spatial distribution characteristics to obtain the environmental stability index; When the environmental stability index is less than the preset target value, record the environmental feature vector including the current temperature, relative humidity, air pressure, temperature gradient vector, humidity gradient vector, and environmental stability index; Conduct a temperature cycle test on the filter film, and simultaneously collect the surface image and deformation displacement data of the filter film to construct a triple dataset including the environmental feature vector, the surface image of the filter film, and the deformation displacement data.

[0026] Specifically, multi-dimensional real-time monitoring of the test environment is achieved through reasonable configuration of the sensor module. The system includes a temperature sensor array, a humidity sensor array, a barometric pressure sensor, and a data acquisition unit integrating data acquisition and processing functions. Among them, the temperature sensor array is composed of thermocouples with an accuracy of plus or minus 0.1 degree Celsius, and no less than nine sensor nodes are evenly distributed in the test cavity in a three-by-three matrix manner to achieve multi-point spatial coverage of the temperature field; the humidity sensor array uses capacitive highly sensitive components and is configured at the four corners of the cavity to monitor the water vapor distribution state, with an accuracy reaching plus or minus 0.5% relative humidity; barometric pressure measurement uses a piezoresistive barometric pressure sensor, and the accuracy is controlled within plus or minus 0.1 kPa to ensure the reliability of measurement even when the external barometric pressure fluctuates; all the above sensors are connected to the central data acquisition unit by wired or bus means. This unit is based on a 32-bit high-performance microcontroller and integrates a high-precision 16-bit analog-to-digital converter module, supporting high-frequency synchronous acquisition of various analog signals, and at the same time having the capabilities of data buffering, filtering, and preliminary processing. When the original sensor data acquisition is completed, digital filtering processing is performed on the environmental original data to eliminate the interference of non-ideal factors such as high-frequency noise, transient disturbances, and system jitters. An eighth-order Butterworth low-pass filter is used as the core of the digital filtering algorithm, and the cut-off frequency of the filter corresponding to each type of sensor is set to one-fourth of the sampling frequency of this type. For example, if the temperature sensor samples at 5 Hz, the corresponding cut-off frequency is 1.25 Hz, so as to fully suppress high-frequency disturbances while retaining the effective change trend. After filtering processing, a smooth and continuous time series of environmental parameters is obtained, and through a pre-established calibration curve, the filtered original readings are converted into international standard physical quantity units. For example, temperature is converted to Kelvin, humidity remains as a percentage, and the barometric pressure unit is converted to Pascal. This calibration process uses the piecewise linear interpolation method, and multiple calibration points are set within the sensor response range. Ten points are set for temperature, eight points for humidity, and five points for barometric pressure, so as to ensure sufficient conversion accuracy and linear response ability of different physical parameters in different range intervals. When stable standard physical quantity data is obtained, to judge the spatial uniformity and stability of the current environment, based on the spatial distribution coordinates of the sensors, spatial gradient calculations are performed on the temperature and humidity data. By analyzing the change rates of temperature and humidity in three-dimensional space, a temperature gradient vector and a humidity gradient vector are constructed to describe the directionality and severity of the temperature and humidity distributions in space, and are incorporated into the subsequent environmental evaluation system as part of the spatial distribution characteristics. At the same time, to quantify the stability degree of the environment changing over time, an environmental stability index is calculated. The calculation method of this index is the ratio between the maximum value of the changes of various environmental physical quantities and their corresponding measurement accuracies within a 30-second time window in the past. For example, whether the temperature change exceeds plus or minus 0.1 degree Celsius, whether the humidity exceeds plus or minus 0.5%, etc.Taking this ratio as the judgment basis, when the stability indices of all environmental parameters are lower than the empirical threshold of 0.5, it is determined that the test environment has entered a stable state. Under the condition that the environment is confirmed to be in a stable state, the environmental feature vector at the current time point is automatically recorded. This vector includes real-time temperature, relative humidity, atmospheric pressure, temperature spatial gradient, humidity spatial gradient, and environmental stability index, constituting the environmental state description set. The preset temperature cycling scheme is started to perform thermal excitation treatment on the filter film sample at multiple temperature points. This process is realized by a precision temperature control device to ensure that each temperature point maintains a sufficient long time under a stable and uniform thermal field to complete thermal equilibrium. After each temperature point reaches the set temperature and passes the verification of the environmental stability index, the system starts the imaging module and the displacement measurement module. The former acquires the surface image of the filter film to capture the surface contour and structural changes after thermal deformation, and the latter records the in-plane or out-of-plane micro-displacement data for accurately reconstructing the thermal deformation displacement field. At each temperature point, the current environmental feature vector, the surface image of the filter film, and the corresponding displacement data are combined to form a triple dataset.

[0027] In a specific embodiment, the process of performing the temperature cycling test on the filter film, simultaneously acquiring the surface image of the filter film and the deformation displacement data, and constructing a triple dataset including the environmental feature vector, the surface image of the filter film, and the deformation displacement data may specifically include the following steps: Establish a test environment for the filter film, and use a temperature control device, an imaging system, and a micro-displacement measurement unit to construct a filter film thermal deformation test system; Fix the filter film sample to be tested on the sample bracket, and ensure that the surface of the filter film is perpendicular to the optical axis of the imaging system through the five-point positioning method to obtain the test in-place state; Based on the test environment and the temperature control device, set the test temperature sequence for the filter film in the test in-place state, and monitor the temperature fluctuations at each temperature point in the test temperature sequence to obtain the temperature stable state; Trigger the imaging system and the micro-displacement measurement unit in the filter film thermal deformation test system in the temperature stable state to obtain the surface image of the filter film and the deformation displacement data at each temperature point; Combine the surface image of the filter film, the deformation displacement data, and the environmental feature vector to form a triple dataset.

[0028] Specifically, a thermal deformation test system for the filter film is established by precisely controlling the thermal environment, accurately collecting image data and nano-scale displacement information. A test chamber with good thermal field uniformity, controllability and sealability is selected, and the core of the system is built on a multi-zone temperature control device with independent temperature control capabilities. The temperature control device adopts a four-channel PID control strategy to achieve independent heating or cooling of different zones, with a control accuracy higher than ±0.05 °C, a fluctuation not exceeding 0.03 °C after the temperature stabilizes, and a temperature field uniformity error controlled within 0.1 °C, providing precise, stable and repeatable thermal excitation conditions for the filter film. At the same time, to ensure the spatial resolution and time response ability of deformation monitoring, a high-precision imaging system and a micro-displacement measurement unit are integrated into the test system. The imaging system uses a scientific-grade CCD camera with a resolution of not less than 2048×2048 pixels, a pixel size controlled within 5.5 microns, a dynamic range exceeding 85 dB, and a quantum efficiency higher than 95%. It is used in conjunction with an industrial macro lens with a focal length of 100 mm, and the working distance is set at about 150 mm to achieve an imaging accuracy of 1 micron per pixel, so as to clearly capture the minute undulations and edge contour changes on the surface of the filter film. The micro-displacement measurement unit includes a combined system of a laser interferometer and a nano-piezoelectric displacement stage. The laser interferometer is used to measure the out-of-plane displacement of the filter film, with a resolution of not less than 0.1 nm and a measurement range of up to 10 mm; the piezoelectric displacement stage is used for fine adjustment and compensation operations of the sample position, with a resolution of 1 nm and a stroke range covering 100 microns, ensuring that the true deformation trajectory of the film layer is captured at the microscopic level. These two types of measurement devices achieve coordinated actions through a synchronous control mechanism, so as to obtain surface shape information and displacement data at the same moment, and can cooperate with subsequent image processing and displacement field reconstruction tasks. In terms of sample preparation, the filter film to be tested is placed on a specially customized sample holder. The material of the holder is selected as a stable substrate with an extremely low coefficient of thermal expansion, such as quartz glass or low-expansion ceramic, which has strong thermal stability and can effectively avoid interference with the film layer caused by the deformation of the holder itself. The filter film is fixed in the center of the holder through a special mechanical clamping mechanism, and the five-point positioning method is used to ensure that its surface is strictly perpendicular to the optical axis of the imaging system. This positioning process is completed through two horizontal fine-tuning knobs, two tilt adjustment screws and a limit block, and a laser alignment instrument is used for spatial attitude confirmation, with the positioning error controlled within 0.01 degrees, so as to ensure that the imaging system always obtains a complete, clear and angle-consistent film layer image under different temperature conditions, avoiding data distortion caused by image distortion or centroid shift. After the sample is in place and the test system is initialized, a temperature cycle test plan is set according to the thermal application range of the filter film to be tested. The plan covers the normal working temperature, the extreme environmental temperature and multiple key nodes in between, and ensures that it includes heating and cooling paths.The temperature sequence is set to rise from 25 degrees Celsius to 45 degrees Celsius, then further to 65 degrees Celsius, then drop back to 25 degrees Celsius and continue to drop to 5 degrees Celsius and -15 degrees Celsius, and finally return to room temperature to form a complete thermal cycle process. During the heating or cooling process of each temperature point, the temperature curve feedback by the temperature control device needs to be monitored in real time, and automatic fluctuation detection is carried out through the embedded stability judgment module. Only when the temperature fluctuation range continues to be less than 0.03 degrees Celsius within 300 seconds, the system determines that the temperature point reaches a stable state and starts the subsequent image acquisition and displacement measurement tasks. When the temperature stable state is confirmed to meet the conditions, the test system automatically synchronously triggers the imaging system and the micro-displacement measurement unit to collect the surface image of the current filter film and the deformation displacement data respectively. The image is used for subsequent boundary extraction, morphological registration and displacement field calculation, while the displacement data provides the absolute value reference of the actual deformation amount and is used to calibrate the amplitude of the image deformation information. At each temperature point, the system continuously collects 10 groups of images and displacement data, and the sampling time interval is set to 30 seconds. The random error is removed by averaging multiple groups of data, and the data stability and measurement credibility are improved. To ensure data integrity and environmental matching, when each group of images and displacement data is collected, the environmental feature vector at the current moment is synchronously recorded. This feature vector consists of six parts: temperature, relative humidity, atmospheric pressure, temperature gradient, humidity gradient and environmental stability index, and is provided by the multi-environment monitoring system deployed in the test environment, including real-time data such as thermocouple arrays, capacitive humidity sensors and barometric pressure sensors. The surface image of the filter film, the deformation displacement data and the environmental feature vector corresponding to the same temperature point are integrated to form a data triple with a unified structure and rich content. Each triple contains a complete response record of the filter film under a certain temperature and environmental state. As the test temperature sequence is gradually executed, this triple data set will continue to expand to form a high-dimensional data matrix covering all thermal load paths.

[0029] In a specific embodiment, the process of executing step S200 may specifically include the following steps: Perform adaptive histogram equalization processing on the surface image of the filter film in the triple data set to obtain an enhanced image, and perform double-threshold segmentation on the enhanced image to obtain the film layer boundary features; Perform edge smoothing processing and sub-pixel edge detection on the film layer boundary features to obtain fine edge features; Perform phase correlation registration on the fine edge features and the reference image at the reference temperature to obtain a registration result; Construct a thermal deformation displacement field based on the registration result, and use the orthogonal polynomial decomposition method to represent the displacement field as a linear combination of M canonical deformation modes to obtain a deformation mode spectrum; Establish a thermal deformation response function based on the environmental feature vector and the deformation mode spectrum in the triple data set.

[0030] Specifically, for the surface images of the filter film collected at each temperature point from the triple dataset, the adaptive histogram equalization method is used to enhance them. This method divides the image into several local grid regions and performs contrast enhancement separately in each region to solve the problems of uneven overall image gray level and low local contrast. In the specific implementation process, a contrast limit factor is set to avoid edge artifacts caused by over-enhancement. At the same time, an appropriate grid size is selected according to the image size, and a typical setting is one-eighth of the original image resolution. After the adaptive histogram equalization process, the difference between the bright and dark regions of the image is more obvious, the boundary region features are more prominent, and the overall image hierarchy is clearer. The enhanced image is subjected to double-threshold segmentation, that is, two gray level ranges of a low threshold and a high threshold are set simultaneously, and the image pixels are divided into foreground, background, and edge-uncertain regions. The low threshold is set to 80% of the global gray level mean of the image, and the high threshold is set to 120%. The pixels between the two thresholds are judged for attribution through connected regions and strong boundaries. This processing method can effectively remove the low-contrast background information in the image, while highlighting the gray level jump region of the film layer contour and its edge, thereby obtaining a preliminary feature map of the film layer boundary. To improve the continuity, smoothness, and geometric accuracy of the boundary, edge smoothing is performed on the above binary boundary map, and a Gaussian filter kernel is used for convolution operation. In this process, the window size of the Gaussian kernel is set to 5×5, and the standard deviation is controlled at about 1.2 to smooth the fine artifacts and edge jitters while maximizing the retention of the boundary structure. After edge smoothing, sub-pixel edge detection is performed to improve the positioning accuracy of the boundary points from integer pixels to one decimal place or even higher. This processing is achieved by means of image gradient fitting or edge contour interpolation. For example, by fitting the gray level distribution surface model around the boundary, the sub-pixel position of the edge points in space is accurately calculated to obtain a high-precision set of boundary points. The fine edge features in this image are registered with the edge features of the reference image at the pre-collected reference temperature. The registration process uses the phase correlation algorithm, which is based on the image frequency domain transformation. By performing the fast Fourier transform on the source image and the target image respectively and calculating their phase differences, and then obtaining the translation vector between the images through the inverse transform, the spatial alignment of the images at the sub-pixel level is achieved. Since this method is insensitive to brightness changes and the registration accuracy can be better than 0.5 pixels, it is suitable for the front and back image alignment tasks of the thermal deformation field of the filter film. Based on the above registration results, the thermal deformation displacement of the filter film at different temperature points relative to the reference state is calculated. By comparing the spatial differences of the corresponding points between the registered image and the reference image, a two-dimensional thermal deformation displacement field is constructed, where each pixel point corresponds to a spatial displacement vector, reflecting the translation amount or deformation direction of the point under the thermal load.To analyze the cause characteristics and structural distribution law of the deformation field, it is expressed as a linear superposition of several typical deformation modes. Therefore, the orthogonal polynomial decomposition method is introduced to represent the entire displacement field as a linear combination of a set of canonical deformation modes. The set of canonical deformation modes is predefined by the system, including the uniform expansion mode, the radially non-uniform expansion mode, the shear mode, the corrugation mode, and multiple high-order complex modes. Each mode has a clear physical meaning and spatial distribution function, and they are mathematically orthogonal to each other. In actual operations, the superposition coefficients of each canonical mode are determined by the least squares fitting method, and a deformation mode spectrum is formed, that is, which basic modes the deformation field of the filter film consists of at each temperature point, and the contribution ratio of each mode to the total deformation, so as to quantitatively express the modal response structure driven by the current thermal environment. The obtained deformation mode spectrum is jointly modeled with the environmental feature vectors recorded in the corresponding triple dataset to establish a thermal deformation response function with the environmental state as the input and the modal response as the output. The mathematical form of this function is an explicit expression or an interpolation model, and its goal is to predict or trace back the thermal response behavior of the filter film under a certain state through environmental parameters, and serve as the basic data for subsequent reliability assessment, structural calibration, and environmental adaptability analysis.

[0031] In a specific embodiment, the process of executing step S300 may specifically include the following steps: Extract the principal component analysis result of the environmental feature vectors in the thermal deformation response function as the environmental condition layer parameters, take the set of deformation mode coefficients of the deformation mode spectrum in the thermal deformation response function as the deformation mode layer parameters, take the thermal deformation displacement field as the observation data layer, and construct a hierarchical probabilistic graphical model according to the environmental condition layer parameters, the deformation mode layer parameters, and the observation data layer; Define the conditional probability according to the hierarchical probabilistic graphical model to represent the probability of observing the preset thermal deformation displacement field given the environmental conditions and deformation modes, define the prior distribution to represent the distribution of the deformation mode parameters under the preset environmental conditions, and define the marginal distribution to represent the distribution of the environmental conditions to obtain a complete probability model; Introduce a variational distribution into the complete probability model, obtain the initial variational distribution by initializing the variational parameters, and execute the variational expectation maximization algorithm on the initial variational distribution to obtain an optimized variational distribution; Integrate the generalized approximate message passing technique into the optimized variational distribution to extract the set of modal deformation parameters; Calculate the stability index of each target deformation mode under different environmental conditions based on the set of modal deformation parameters, and combine the set of modal deformation parameters and its stability index to obtain a structured representation of the thermal deformation characteristics.

[0032] Specifically, the thermal deformation response function establishes a clear mapping relationship between the thermal deformation behavior observed on the surface of the filter film at each test temperature point and the corresponding environmental state, which includes two core data structures: one is the set of feature vectors describing the environmental conditions, and the other is the deformation mode spectrum describing the structural response of thermal deformation. To simplify the high-dimensional environmental parameter space and reduce the interference of redundant information on subsequent model inference, principal component analysis is performed on all environmental feature vectors in the thermal deformation response function, that is, the multi-dimensional environmental parameter matrix including temperature, humidity, air pressure and their gradient components is projected into the principal component space. By retaining the first few principal components with high variance contribution rates, the high-dimensional environmental variables are compressed into a parameter set with physical interpretability and low-dimensional expression. This principal component set constitutes the parameter representation of the environmental condition layer in the probabilistic graphical model. The coefficient set of each canonical deformation mode in the deformation mode spectrum extracted from the thermal deformation response function is used as the parameter representation of the deformation mode layer. These coefficients reflect the response intensity of various standard deformation forms exhibited by the filter film under different thermal environments and are the direct feature expressions of the structural thermal response. The thermal deformation displacement field, as the observation data layer in the graphical model, records the spatial offset of each pixel point under the action of thermal load in the form of a two-dimensional field and has directly measurable spatial deformation information. By incorporating the three types of variables, namely the environmental parameters after principal component analysis, the deformation mode spectrum coefficients, and the actual measured displacement field, into a unified hierarchical graphical model, a top-down causal structure is established, where the environmental layer affects the deformation mode layer, and the deformation mode layer further determines the distribution of the actually measured deformation field in the observation layer, thus completing the structural modeling of the entire thermal deformation process. After completing the construction of the graphical model, a probability distribution is defined for the dependence relationships between its nodes and levels to form a complete probabilistic graphical model. Among them, the conditional probability function of the observation layer is used to describe the possibility of observing a specific deformation field given the known environmental conditions and deformation mode parameters. This distribution is derived based on historical sample error statistics; the prior distribution is used to express the possible value range and distribution form of the deformation mode parameters under specific environmental states, reflecting the statistical prior information of the environmental conditions on the structural response law; at the same time, the marginal distribution is defined, that is, the distribution characteristics of the environmental parameters themselves in the entire sample space, which is convenient for deriving the complete joint distribution and marginal inference relationship. Through the clear setting of these three types of distributions, a structurally complete and hierarchically clear probabilistic modeling framework is formed for the subsequent posterior inference process. Since there are multiple high-dimensional variables involved in the above model and the form of the posterior distribution is complex, the direct calculation is difficult. Therefore, the variational inference method is introduced to approximately solve it. During this process, an initial variational distribution is set, and its parameters are adjusted to gradually approximate the true posterior distribution. The initial variational distribution adopts a factorized structure, that is, it is assumed that the joint variational distribution between the parameters of the environmental layer and the deformation mode layer can be split into two independent distributions, and optimization is performed for different parameter sets respectively.In actual calculations, by setting the initial mean and covariance parameters, the initial point of the variational expectation-maximization algorithm is formed, and with the maximization of the lower bound of the model evidence as the optimization objective, the variational parameters of each variable are iteratively updated until the change between two consecutive iterations is less than a preset convergence threshold, thereby obtaining the final optimized variational distribution as an approximation of the true posterior distribution. After obtaining the optimized variational distribution, the generalized approximate message passing technique is introduced, representing the entire graphical model as a factor graph structure, and passing marginal distribution information between the nodes in the graph to improve the inference accuracy and convergence speed through local updates and global coordination mechanisms. Based on this optimized variational structure, the most representative set of modal deformation parameters is extracted from the deformation mode layer parameters, that is, a set of deformation modes that stably exist, have a high probability density, and low environmental sensitivity, usually with a small number but sufficient to represent the typical deformation behavior of the filter film under various thermal environments. Based on this set of modal deformation parameters, combined with the perturbation input of the environmental condition variables, the response intensity of each modal parameter to environmental changes is calculated, thereby deriving the stability index. This index is measured in the form of derivative norm, response fluctuation amplitude, or sensitivity index, reflecting the reliability and consistency of a specific mode in the face of a complex thermal environment. Combining these modal deformation parameters with the corresponding stability indices forms a structured representation of the thermal deformation characteristics.

[0033] In a specific embodiment, the process of executing step S400 may specifically include the following steps: Construct an environmental impact function based on the structured representation of the thermal deformation characteristics and the corresponding measurement deviation data, and perform the construction of a weighted radial basis function network according to the environmental impact function to obtain an environmental impact function network model; Train the environmental impact function network model to obtain a trained environmental impact function, and input the environmental feature vector of the current test environment into the trained environmental impact function to calculate the predicted deviation value caused by environmental changes, obtaining the measurement deviation related to the environment; Construct an environmental compensation matrix based on the stability index in the structured representation of the thermal deformation characteristics, and correct the original deformation characteristics by applying the measurement deviation related to the environment and the environmental compensation matrix to obtain the corrected deformation characteristics; Construct a set of environmental adaptability calibration curves based on the environmental response historical data of the filter film, apply the set of environmental adaptability calibration curves to the corrected deformation characteristics, and obtain the environmentally adaptive thermal deformation characteristics through piecewise polynomial function mapping; Calculate the thermal expansion coefficient matrix, the thermal deformation anisotropy index, and the thermal ripple intensity based on the environmentally adaptive thermal deformation characteristics, and evaluate the thermal deformation characteristics of the filter film to obtain the thermal deformation reliability evaluation result.

[0034] Specifically, an environmental impact function is constructed based on the structured representation of thermal deformation characteristics and the corresponding measurement deviation data to characterize the non-linear mapping relationship from the environmental feature vector to the observation error. Its input is a six-dimensional parameter set including temperature, relative humidity, air pressure, and the corresponding gradients and stability indices at the test moment, and the output is the environmental-related measurement deviation of each main mode deformation parameter. Since this relationship is highly dimensional and non-linear, it is difficult to express it with an explicit function. Therefore, a function network model with strong fitting ability is used to complete this mapping construction. To this end, a neural network structure with weighted radial basis functions as the core is constructed. This network consists of an input layer, a hidden layer, and an output layer. The input layer accepts the six-dimensional environmental feature vector as input. The hidden layer is composed of multiple nodes with the form of Gaussian kernel functions. Each node calculates the distance between the current input and the preset center and outputs a response value accordingly. At the same time, adjustable weight parameters are set after each node to implement the weighting mechanism, thereby enhancing the generalization ability of the network and the ability to adapt to data densities in different regions. The dimension of the output layer is the same as the number of main mode deformation parameters and is used to output the prediction deviation caused by the environment. The initialization of the network parameters is combined with the distribution law of environmental samples to ensure that the center positions can cover the main distribution areas of the environmental feature space, and overfitting is prevented through the regularization constraint of the weights. The collected training sample set is used for model training. Each sample contains a set of environmental feature vectors and the deviation between the measured and theoretical deformations in this environment. During the training process, a loss function based on the L1 norm is adopted to enhance the robustness of the model to local anomalies, and at the same time, the regularization term is combined to suppress the network complexity. The training uses an iterative optimization algorithm with the learning rate dynamically adjusted, and the optimization process continues until the loss function converges stably, obtaining an environmental impact function network model that can be used in practical applications. After training is completed, the environmental feature vector under the current test conditions is input into this network model to obtain the predicted deviation values corresponding to each deformation mode under this specific environmental condition. These deviation values form the measurement error vector under the current environment. To enable the compensation strategy to have differential capabilities, an environmental compensation matrix is established in combination with the stability indices of each mode provided in the structured representation of thermal deformation characteristics. This matrix is a diagonal matrix, and its diagonal elements are weight coefficients generated according to the mapping of the stability indices. The mapping relationship should satisfy that the higher the stability of the mode, the greater the corresponding compensation weight, so as to ensure the dominant position of reliable modes in compensation. Multiply this compensation matrix by the predicted deviation vector and subtract this compensation amount from the original thermal deformation characteristic parameters to obtain the corrected deformation characteristics. To improve the adaptability of the deformation parameters to wider environmental changes after the preliminary correction is completed, the construction of an environmental adaptability calibration curve set is introduced. This curve set is derived from the modal response change trends of the filter film under different environmental parameters during the historical sampling process. Each curve corresponds to a modal parameter, with a specific environmental dimension as the independent variable and the amplitude of the compensated deformation mode under this environmental condition as the dependent variable.Due to the non-linear mutation of the relationship between the environment and the response in different intervals, the curve fitting adopts a piecewise polynomial strategy. Each segment uses a third-order or fourth-order polynomial for fitting, and smooth connection points are set between each segment to ensure the continuity and differentiability of the overall curve. After the curve construction is completed, the current test environment characteristics are input into the corresponding calibration curve model, and the compensated modal deformation parameters are adjusted according to the fitting function to obtain the environment-adaptive thermal deformation characteristics, which are manifested as stable, accurate, and sustainable tracking modal response results under environmental perturbation conditions. Taking this environment-adaptive thermal deformation characteristic as the input, the key thermal deformation performance indicators of the filter film are systematically calculated. Calculate the thermal expansion coefficient matrix, which is used to describe the thermal expansion ability of the filter film in the principal axis and non-principal axis directions. The diagonal terms represent linear expansion, while the non-diagonal terms reflect the shear deformation tendency of the film layer under thermal load. According to the symmetry and directional differences of the thermal expansion matrix, calculate the thermal deformation anisotropy index, which quantifies the degree of inconsistency of the thermal response of the filter film in different directions and is used to judge whether there is a risk of thermal stress concentration or structural asymmetry in a complex thermal environment. Analyze the thermal ripple intensity by extracting the relative amplitude of the ripple pattern in the deformation characteristics and calculating the ratio with the uniform expansion pattern to judge whether there are periodic surface undulations caused by thermal drive in the film layer. If the intensity is too high, it will affect the imaging accuracy of the optical system. The above indicators form a multi-dimensional evaluation result of the thermal deformation performance. By comparing its historical distribution, control limits, and tolerance intervals, it is judged whether the current filter film sample has excellent thermal stability, and the final thermal deformation reliability evaluation result is generated accordingly.

[0035] In a specific embodiment, the process of performing the steps of calculating the thermal expansion coefficient matrix, the thermal deformation anisotropy index, and the thermal ripple intensity based on the environment-adaptive thermal deformation characteristics and evaluating the thermal deformation characteristics of the filter film to obtain the thermal deformation reliability evaluation result may specifically include the following steps: Construct an index system for thermal deformation characteristics based on the environment-adaptive thermal deformation characteristics, calculate the thermal expansion coefficient matrix, and obtain the thermal expansion characteristics; Calculate the thermal deformation anisotropy index based on the thermal expansion coefficient matrix to obtain a quantitative index of the direction dependence of the film layer thermal expansion; Calculate the ratio of the amplitude of the ripple pattern to the amplitude of the uniform expansion pattern in the environment-adaptive thermal deformation characteristics to obtain the thermal ripple intensity, and calculate the root mean square of the deformation difference during the heating process and the cooling process to obtain the thermal deformation hysteresis index; Perform multi-dimensional anomaly detection on the thermal expansion characteristics, the quantitative index of direction dependence, the thermal ripple intensity, and the thermal deformation hysteresis index to obtain the anomaly detection result; Construct a reliability evaluation model for the thermal deformation of the filter film based on the failure mode and effect analysis method, define the set of thermal deformation failure modes of the filter film, calculate the risk priority number for each failure mode in the set of thermal deformation failure modes, and obtain the comprehensive thermal deformation reliability index of the filter film; Generate the reliability evaluation result of the thermal deformation of the filter film based on the anomaly detection result and the comprehensive thermal deformation reliability index.

[0036] Specifically, a thermal deformation characteristic index system is constructed based on the environment-adaptive thermal deformation characteristics. This system includes multi-angle characterizations of the expansion trend, direction consistency, local disturbance, and thermal response consistency. Different modal response quantities are extracted from the environment-adaptive thermal deformation characteristics, and a structural description in matrix form is established based on these modes. During the construction of the thermal expansion characteristic index, the modal response values representing the expansion behavior of the main axis direction of the filter film are preferentially extracted, and a two-dimensional thermal expansion coefficient matrix is established according to the planar orthogonal coordinate system. The diagonal components of this matrix are used to reflect the thermal expansion and contraction capabilities of the film layer in the horizontal and vertical directions, while the off-diagonal components reflect the degree of shear response, thus completely describing the thermal expansion behavior of the overall structure of the film layer at multiple temperature points. To quantify the anisotropy of thermal response, the thermal deformation anisotropy index is calculated based on the thermal expansion coefficient matrix. This index forms a dimensionless quantification index by comparing the differences in thermal expansion capabilities in two main axis directions. The closer this index is to zero, the higher the anisotropy; the larger the deviation value, the greater the risk of asymmetric structural deformation in the film layer when heated, resulting in overall displacement eccentricity or optical path deviation. To identify whether there is a periodic undulation behavior in the filter film under thermal drive, the modal amplitudes of the ripple mode and the uniform expansion mode are extracted from the thermal deformation characteristics, and the ratio between them is calculated. This ratio is the thermal ripple intensity, which is an important manifestation of local disturbance on the film surface, especially obvious under multiple thermal cycles or high-speed heating and cooling conditions. If this ratio remains high, it means that there is local residual stress release in the material, or there are defect structures such as uneven thickness or coating discreteness on the surface, which causes substantial interference to the surface shape accuracy in the optical system. To evaluate the response consistency of the filter film to the temperature path direction, the deformation modal parameters at the same temperature points in the heating path and the cooling path are respectively extracted, and the root mean square calculation of the differences of the same mode during the lifting and lowering processes is performed to obtain the thermal deformation hysteresis index. This index represents whether there are phenomena such as response hysteresis, asymmetric recovery, or untimely release of internal stress in the filter film during the thermal excitation process. The larger its value, the more obvious the potential structural memory effect or thermal fatigue signs in the film layer. After the above four types of index extractions are completed, they are combined in the form of a multi-dimensional feature vector, and the state of the current sample is evaluated through an anomaly detection method. This anomaly detection is realized by the statistical distribution distance method. Based on a large number of historical normal sample data, the mean vector and covariance matrix of the multi-dimensional thermal deformation index are established, and then the current sample index is substituted into the distance function model to calculate its Mahalanobis distance in the index space. If this distance exceeds the set threshold, it is determined to be abnormal. At the same time, the reasons for the anomaly are identified by sorting the deviation degrees in different dimensions, whether it comes from inconsistent expansion, local corrugation, or response lag, to form a preliminary anomaly recognition result.In order to combine the above abnormal analysis results with the engineering reliability evaluation criteria, the failure mode and effect analysis method is introduced to construct a reliability evaluation model for the thermal deformation of the filter film. This model defines a set of typical failure modes during the thermal deformation process, such as excessive uniform expansion, significant non-uniform expansion, strong corrugation response, hysteresis greater than the safety threshold, or irreversible mode drift, etc. And severity, occurrence probability, and detectability scoring criteria are set for each failure mode. Then, the product of the three scores is used to obtain the risk priority number for each mode. The risk priority numbers of multiple failure modes are integrated into a comprehensive thermal deformation risk value through averaging or weighted averaging. The higher this value, the more risk factors the current filter film sample exhibits in the thermal environment, the greater the uncertainty of the system performance, and the lower the service life or accuracy guarantee ability. And this value constitutes the comprehensive thermal deformation reliability index of the filter film. The comprehensive reliability index is cross-validated with the results of the previous multi-dimensional anomaly detection, and combined with the historical stable sample interval to judge the thermal stability level and failure probability level of the current sample, generating a reliability evaluation report for the thermal deformation of the filter film. The report lists multiple evaluation results such as whether the thermal expansion behavior is within the control range, whether the directionality is unbalanced, whether the corrugation exceeds the limit, whether the thermal hysteresis is abnormal, and whether the failure modes are superimposed, supplemented by a risk level diagram and a recommended threshold reference, so as to provide quantifiable decision-making support for engineering use, batch quality control, or process optimization, and provide a necessary technical feedback channel for the next process iteration, material replacement, or structural adjustment.

[0037] Please refer to Figure 2 , Figure 2 which is a schematic block diagram of the structure of the filter film thermal deformation test device 200 provided by the embodiment of the present application. As Figure 2 shown, the filter film thermal deformation test device 200 includes: A temperature cycle test module 210, configured to perform a temperature cycle test on the filter film based on the test environment, and simultaneously collect a triple dataset including an environmental feature vector, a filter film surface image, and deformation displacement data; A decomposition module 220, configured to perform image preprocessing and displacement field construction on the triple dataset, and decompose the displacement field into a linear combination of canonical deformation modes to obtain a thermal deformation response function; An extraction module 230, configured to construct a hierarchical probabilistic graph model based on the thermal deformation response function, and extract a structured representation of the thermal deformation features through variational inference; A calculation module 240, configured to perform bias compensation on the structured representation of the thermal deformation features to obtain environment-adaptive thermal deformation features, and calculate a reliability evaluation result of the thermal deformation including a thermal expansion coefficient matrix, a thermal deformation anisotropy index, and a thermal corrugation strength.

[0038] Through the collaborative cooperation of the above-mentioned various components, by establishing a multi-environmental parameter monitoring system and environmental feature vectors, and combining the environmental impact function and compensation matrix, the present invention can perform accurate measurements under different temperature, humidity, and air pressure conditions, effectively reducing the measurement deviation caused by environmental changes, and making the measurement results consistent and comparable in a variable environment. By using image preprocessing techniques such as adaptive histogram equalization, double-threshold segmentation, and sub-pixel edge detection, and combining the orthogonal polynomial decomposition method to represent the displacement field as a linear combination of canonical deformation modes, the thermal deformation characteristics of the filter film can be accurately captured, and the measurement accuracy reaches the nanometer level. By constructing a three-layer hierarchical probabilistic graph model using the structural variational Bayesian method and integrating the generalized approximate message passing technique into variational inference, the one-dimensional continuity of in-plane deformation and the two-dimensional continuity recognition of inter-layer deformation are effectively improved, and the extraction ability of the thermal deformation characteristics of complex multi-layer filter films is enhanced. The environmental impact function implemented based on the weighted radial basis function network, combined with the compensation matrix constructed based on the deformation mode stability index, realizes the intelligent compensation for the measurement deviation caused by environmental changes, significantly improving the accuracy and repeatability of the measurement results. By constructing a characteristic index system including the thermal expansion coefficient matrix, thermal deformation anisotropy index, thermal ripple intensity, and thermal deformation hysteresis index, and combining the multi-dimensional anomaly detection algorithm based on the modified Mahalanobis distance and the risk assessment model based on failure mode analysis, the comprehensive evaluation and reliability analysis of the thermal deformation characteristics of the filter film are realized.

[0039] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0040] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0041] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for testing thermal deformation of a filter film, characterized in that: include: Based on the test environment, the filter film is subjected to a temperature cycle test, and a triple data set including the environmental feature vector, the filter film surface image and the deformation displacement data is collected; Performing image preprocessing and displacement field construction on the triplet data set, and decomposing the displacement field into a linear combination of canonical deformation modes to obtain a thermal deformation response function; A hierarchical probabilistic graphical model is constructed based on the thermal deformation response function, and a structured representation of thermal deformation features is extracted through variational inference; Deviation compensation is performed on the structured representation of the thermal deformation characteristics to obtain environmentally adaptive thermal deformation characteristics, and thermal deformation reliability evaluation results including a thermal expansion coefficient matrix, a thermal deformation anisotropy index, and thermal corrugation strength are calculated.

2. The method for testing thermal deformation of a filter film according to claim 1, characterized in that: Based on the test environment, the filter film is subjected to a temperature cycle test, and a triple data set including an environmental feature vector, a filter film surface image, and deformation displacement data is collected, including: A multi-environmental parameter monitoring system is constructed by using a temperature sensor array, a humidity sensor array, an air pressure sensor and a data acquisition unit, and an original data set is collected by the multi-environmental parameter monitoring system; Performing digital filtering on the original data set to obtain filtered environmental parameter data, and calibrating the filtered environmental parameter data to obtain standard physical quantity data; Calculating the spatial distribution parameters of the test environment based on the standard physical quantity data to obtain the environmental spatial distribution characteristics, and calculating the ratio of the maximum fluctuation of the standard physical quantity data to the measurement accuracy requirement based on the environmental spatial distribution characteristics to obtain the environmental stability index; When the environmental stability index is less than a preset target value, recording an environmental feature vector including current temperature, relative humidity, air pressure, temperature gradient vector, humidity gradient vector and environmental stability index; The filter film is subjected to a temperature cycle test, and a surface image of the filter film and deformation displacement data are collected simultaneously to construct a triple data set including the environmental feature vector, the surface image of the filter film and the deformation displacement data.

3. The method for testing thermal deformation of a filter film according to claim 2, characterized in that: The temperature cycle test is performed on the filter film, and the surface image and deformation displacement data of the filter film are collected at the same time, and a triplet data set including the environmental feature vector, the surface image of the filter film and the deformation displacement data is constructed, including: Establishing a test environment for the filter film, and constructing a filter film thermal deformation test system using a temperature control device, an imaging system, and a micro-displacement measurement unit; Fix the filter film sample to be tested on the sample holder, and use the five-point positioning method to ensure that the filter film surface is perpendicular to the optical axis of the imaging system to obtain the test position state; Setting a test temperature sequence for the filter film in the test-in-place state based on the test environment and the temperature control device, and monitoring the temperature fluctuation of each temperature point in the test temperature sequence to obtain a temperature stable state; Triggering the imaging system and the micro-displacement measurement unit in the filter film thermal deformation test system under the temperature stable state to obtain the filter film surface image and deformation displacement data at each temperature point; The filter film surface image, the deformation displacement data and the environmental feature vector are combined to form a triplet data set.

4. The method for testing thermal deformation of a filter film according to claim 1, characterized in that: The image preprocessing and displacement field construction are performed on the triplet data set, and the displacement field is decomposed into a linear combination of the canonical deformation modes to obtain the thermal deformation response function, including: Performing adaptive histogram equalization processing on the filter film surface image in the triplet data set to obtain an enhanced image, and performing double threshold segmentation on the enhanced image to obtain film layer boundary features; Performing edge smoothing and sub-pixel edge detection on the boundary features of the film layer to obtain fine edge features; Performing phase correlation registration on the fine edge feature and the reference image at a reference temperature to obtain a registration result; Based on the registration result, a thermal deformation displacement field is constructed, and an orthogonal polynomial decomposition method is used to express the displacement field as a linear combination of M standard deformation modes to obtain a deformation mode spectrum; A thermal deformation response function is established based on the environmental feature vector in the triplet data set and the deformation mode spectrum.

5. The method for testing thermal deformation of a filter film according to claim 1, characterized in that: The step of constructing a hierarchical probability graph model based on the thermal deformation response function and extracting a structured representation of thermal deformation features through variational inference includes: Extracting principal component analysis results of environmental feature vectors in the thermal deformation response function as environmental condition layer parameters, taking a set of deformation mode coefficients of a deformation mode spectrum in the thermal deformation response function as deformation mode layer parameters, taking a thermal deformation displacement field as an observation data layer, and constructing a hierarchical probability graph model according to the environmental condition layer parameters, the deformation mode layer parameters and the observation data layer; According to the hierarchical probability graph model, a conditional probability is defined to represent the probability of observing a preset thermal deformation displacement field under given environmental conditions and deformation modes, a prior distribution is defined to represent the distribution of deformation mode parameters under preset environmental conditions, and a marginal distribution is defined to represent the distribution of environmental conditions, thereby obtaining a complete probability model; Introducing a variational distribution into the complete probability model, obtaining an initial variational distribution by initializing variational parameters, and performing a variational expectation maximization algorithm on the initial variational distribution to obtain an optimized variational distribution; integrating a generalized approximate message passing technique into the optimized variational distribution to extract a set of modal deformation parameters; The stability index of each target deformation mode under different environmental conditions is calculated based on the modal deformation parameter set, and the modal deformation parameter set and its stability index are combined to obtain a structured representation of thermal deformation characteristics.

6. The method for testing thermal deformation of a filter film according to claim 1, characterized in that: The method of performing deviation compensation on the structured representation of the thermal deformation feature to obtain the environment-adaptive thermal deformation feature and calculating the thermal deformation reliability evaluation result including the thermal expansion coefficient matrix, the thermal deformation anisotropy index and the thermal corrugation strength includes: Constructing an environmental impact function based on the structured representation of the thermal deformation characteristics and the corresponding measurement deviation data, and performing weighted radial basis function network construction according to the environmental impact function to obtain an environmental impact function network model; Training the environmental impact function network model to obtain a trained environmental impact function, inputting the environmental feature vector of the current test environment into the trained environmental impact function, calculating the prediction deviation value caused by the environmental change, and obtaining the environment-related measurement deviation; constructing an environmental compensation matrix based on the stability index in the structured representation of the thermal deformation feature, and applying the environment-related measurement deviation and the environmental compensation matrix to correct the original deformation feature to obtain a corrected deformation feature; An environmental adaptability calibration curve set is constructed based on the environmental response history data of the filter film, the environmental adaptability calibration curve set is applied to the corrected deformation characteristics, and the environmental adaptive thermal deformation characteristics are obtained by piecewise polynomial function mapping; Based on the environmental adaptive thermal deformation characteristics, the thermal expansion coefficient matrix, the thermal deformation anisotropy index and the thermal ripple strength are calculated, and the thermal deformation characteristics of the filter film are evaluated to obtain the thermal deformation reliability evaluation result.

7. The method for testing thermal deformation of a filter film according to claim 6, characterized in that: The thermal expansion coefficient matrix, thermal deformation anisotropy index and thermal ripple strength are calculated based on the environmental adaptive thermal deformation characteristics, and the thermal deformation characteristics of the filter film are evaluated to obtain the thermal deformation reliability evaluation result, including: Based on the environmental adaptive thermal deformation characteristics, a thermal deformation characteristic index system is constructed, a thermal expansion coefficient matrix is ​​calculated, and thermal expansion characteristics are obtained; Calculating the thermal deformation anisotropy index based on the thermal expansion coefficient matrix to obtain a directional dependence quantitative index of thermal expansion of the film layer; The ratio of the ripple mode amplitude to the uniform expansion mode amplitude in the environmental adaptive thermal deformation feature is calculated to obtain the thermal ripple strength, and the root mean square of the deformation difference between the heating process and the cooling process is calculated to obtain the thermal deformation hysteresis index; Performing multi-dimensional anomaly detection on the thermal expansion characteristics, the direction-dependent quantitative index, the thermal ripple intensity, and the thermal deformation hysteresis index to obtain an anomaly detection result; Based on the failure mode and effect analysis method, a thermal deformation reliability evaluation model of the filter film is constructed, a thermal deformation failure mode set of the filter film is defined, a risk priority number is calculated for each failure mode in the thermal deformation failure mode set, and a comprehensive thermal deformation reliability index of the filter film is obtained; A thermal deformation reliability evaluation result of the filter film is generated based on the abnormality detection result and the comprehensive thermal deformation reliability index.

8. A filter film thermal deformation testing device, characterized in that: Used to perform the filter film thermal deformation test method according to any one of claims 1 to 7, the filter film thermal deformation test equipment comprises: The temperature cycle test module is used to perform a temperature cycle test on the filter film based on the test environment, and simultaneously collect a triple data set including the environmental feature vector, the filter film surface image and the deformation displacement data; A decomposition module, used for performing image preprocessing and displacement field construction on the triplet data set, and decomposing the displacement field into a linear combination of canonical deformation modes to obtain a thermal deformation response function; An extraction module, used for constructing a hierarchical probabilistic graphical model based on the thermal deformation response function, and extracting a structured representation of thermal deformation features through variational inference; The calculation module is used to compensate the deviation of the structured representation of the thermal deformation characteristics, obtain the environment-adaptive thermal deformation characteristics, and calculate the thermal deformation reliability evaluation results including the thermal expansion coefficient matrix, the thermal deformation anisotropy index and the thermal corrugation strength.

Images