A method for monitoring adhesive curing based on optical path difference change

Abstract

The present invention discloses a glue curing monitoring method based on optical path difference change, which belongs to the field of glue curing monitoring, including: configuring a test scheme, establishing a test environment, and configuring environmental characteristics; placing a test sample in the configured test environment, and performing sample timing monitoring, and establishing a curing timing data set; establishing and updating phase data, and establishing timing curing data of the cured glue; establishing a predictive fitting model, and performing a fitting test, and establishing a fitting test result; judging whether the fitting test result meets a preset test threshold, and when it meets, completing the glue curing monitoring of the cured glue through the predictive fitting model. The present application solves the technical problem that the contact monitoring used in the existing glue curing monitoring requires puncturing the colloid, resulting in damage to the colloid, and achieves the technical effect of establishing a predictive fitting model through non-contact multi-point monitoring of the spectrometer and phase data analysis, thereby performing glue curing monitoring and realizing non-destructive monitoring of the glue curing process.

Description

Technical Field

[0001] The invention relates to the field of glue curing monitoring, and in particular to a glue curing monitoring method based on optical path difference change. Background Art

[0002] In industrial production, glue curing is widely used in bonding, sealing, coating, molding and other fields. In order to ensure the final performance and quality of the product, the glue curing process needs to be effectively monitored. In the prior art, the method of monitoring glue curing is contact-type, which requires the sensor or test device to be in direct contact with the colloid, or to puncture the colloid to obtain parameter changes during the glue curing process. However, the contact-type monitoring method will cause varying degrees of damage to the colloid, introduce additional defects, and even change the glue curing reaction process, ultimately affecting the performance and quality of the glue-cured product. Summary of the invention

[0003] The present application provides a glue curing monitoring method based on optical path difference change, aiming to solve the technical problem that the contact monitoring used in the existing glue curing monitoring requires puncturing the colloid, resulting in damage to the colloid.

[0004] The present application discloses a method for monitoring glue curing based on optical path difference change, comprising: configuring a test scheme, establishing a test environment based on the test scheme, and configuring environmental characteristics of the test environment, wherein the environmental characteristics include environmental temperature characteristics, environmental humidity characteristics, and light source characteristics; establishing a test sample of the curing glue, and evenly distributing N measurement points on each test sample, placing the test sample in the configured test environment, and performing sample timing monitoring of the test sample through a spectrometer to establish a curing timing data set; extracting phase data from the curing timing data set, and performing adaptive smoothing processing on the curing timing data set according to the measurement distribution and timing identification of the N measurement points to establish updated phase data; obtaining a package phase with the updated phase data, and calculating a phase difference change and an optical path difference change of the package phase, and establishing time-series curing data of the curing glue according to the calculation results; establishing a prediction fitting model through the time-series curing data and the test scheme, and calling additional verification data to perform a fitting test of the prediction fitting model to establish a fitting test result; judging whether the fitting test result meets a preset test threshold, and when the fitting test result meets the preset test threshold, completing the glue curing monitoring of the curing glue through the prediction fitting model.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0006] Due to the use of a configuration test plan, a test environment is established based on the test plan, and the environmental characteristics of the test environment are configured. The environmental characteristics include environmental temperature characteristics, environmental humidity characteristics, and light source characteristics. Through precise control of the test environment, it is ensured that the glue curing monitoring process is carried out under specific conditions, thereby improving the consistency and repeatability of the test results; a test sample for curing glue is established, and N measurement points are evenly distributed on each test sample. The test sample is placed in the configured test environment, and the sample timing of the test sample is monitored through a spectrometer to establish a curing timing data set. Through multi-point arrangement and timing monitoring, rich data reflecting the glue curing process are obtained, laying the foundation for subsequent analysis; the phase data in the curing timing data set is extracted, and the curing timing data set is adaptively smoothed according to the measurement distribution and timing identification of the N measurement points, and updated phase data is established. Through phase data extraction and adaptive smoothing, measurement noise is eliminated, and the phase change characteristics of the glue curing process are highlighted; the package phase is obtained by updating the phase data, and the phase difference change and optical path difference change of the package phase are measured. The invention relates to a method for calculating the curing performance of the cured adhesive by calculating the curing performance of the cured adhesive and establishing the time series curing data of the cured adhesive according to the calculation results. The evolution of the internal structure and performance of the colloid during the curing process of the adhesive is accurately characterized by quantitative analysis of the changes in phase difference and optical path difference. The predictive fitting model is established through the time series curing data and the test scheme, and the additional verification data is called to perform the fitting test of the predictive fitting model, and the fitting test results are established. By establishing and verifying the predictive fitting model, the prediction and extrapolation of the adhesive curing process are realized to guide the actual production. The method determines whether the fitting test result meets the preset test threshold. When the fitting test result meets the preset test threshold, the curing monitoring of the cured adhesive is completed through the predictive fitting model. The accuracy and reliability of the predictive fitting model are controlled by the threshold judgment of the fitting test result. The technical solution for ensuring the applicability of the adhesive curing monitoring solves the technical problem that the contact monitoring in the existing adhesive curing monitoring requires puncturing the colloid, which causes damage to the colloid. The technical effect of establishing the predictive fitting model through the non-contact multi-point monitoring of the spectrometer and the phase data analysis, thereby performing adhesive curing monitoring and realizing the non-destructive monitoring of the adhesive curing process is achieved.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic flow chart of a method for monitoring adhesive curing based on optical path difference change is provided for an embodiment of the present application;

[0009] Figure 2A flow chart of establishing and updating phase data in a method for monitoring adhesive curing based on optical path difference change is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0010] The overall idea of ​​the technical solution provided by this application is as follows:

[0011] The embodiment of the present application provides a method for monitoring glue curing based on the change of optical path difference. First, by configuring the test environment and test samples, a spectrometer is used to perform non-contact multi-point time series monitoring of the glue curing process to obtain a rich curing time series data set. Then, by performing phase data extraction and adaptive smoothing on the acquired time series data, and calculating the phase difference change and the optical path difference change, time series curing data that accurately reflects the glue curing process is established. On this basis, a prediction fitting model is constructed in combination with the test plan, and the model is fitted and tested by additional verification data, ultimately achieving accurate prediction and monitoring of the glue curing process.

[0012] Compared with the prior art, the present application adopts a non-contact spectrometer monitoring and phase data analysis method, which not only avoids damage to the colloid, but also can comprehensively and dynamically characterize the glue curing process, and realize non-destructive and high-precision online monitoring. At the same time, through the prediction fitting model, the subsequent progress of the glue curing process can be accurately predicted based on the monitoring data in the initial stage, providing a reliable basis for production control and optimization. In the process of glue curing monitoring, there is no need to contact or destroy the colloid, which greatly improves the non-destructiveness, accuracy and applicability of glue curing monitoring.

[0013] After introducing the basic principles of the present application, the non-limiting implementation methods of the present application will be specifically introduced below in conjunction with the drawings in the specification.

[0014] like Figure 1 As shown, the embodiment of the present application provides a method for monitoring glue curing based on optical path difference change, the method comprising:

[0015] S1: Configure a test plan, establish a test environment based on the test plan, and configure environmental characteristics of the test environment, wherein the environmental characteristics include environmental temperature characteristics, environmental humidity characteristics, and light source characteristics.

[0016] Specifically, to achieve non-contact glue curing monitoring, first, configure the corresponding test plan, which includes establishing a test environment and configuring the environmental characteristics of the test environment. The purpose of establishing a test environment is to simulate the real scene of glue curing in order to obtain accurate and reliable monitoring data. The test environment is built according to actual needs. For example, a small test platform can be built in the laboratory, or a test area can be directly established in the production workshop.

[0017] On the basis of establishing the test environment, configure the environmental characteristics of the test environment. The so-called environmental characteristics refer to various factors that affect the glue curing process, including environmental temperature characteristics, environmental humidity characteristics and light source characteristics. Among them, the environmental temperature characteristics refer to the temperature of the test environment. The glue curing process is sensitive to temperature, and different temperatures will affect the rate and degree of glue curing. Therefore, the temperature of the test environment is monitored and controlled to ensure that the glue curing process is carried out under relatively stable temperature conditions. During the test, the temperature sensor is set in the test environment to collect environmental temperature data in real time. The environmental humidity characteristics refer to the humidity of the test environment. Humidity is also an important factor affecting glue curing. Appropriate humidity helps the evaporation of water in the colloid, thereby accelerating the curing process. However, if the humidity is too high, it will inhibit the curing of the colloid. Therefore, during the test, the humidity sensor is used to collect humidity data and monitor and adjust the humidity of the test environment. The light source characteristics mainly refer to the light source used in the curing process, including parameters such as the type, power, irradiation distance and angle of the light source. Different light source characteristics will affect the efficiency of the colloid in absorbing light energy, and then affect the curing rate of the glue. During the test, the light source is selected according to the material properties of the glue and the curing requirements, and the various parameters of the light source are precisely controlled.

[0018] Through the configuration of the test plan and the construction of the test environment, the necessary basic conditions are provided for the subsequent glue curing monitoring. Comprehensive monitoring and control of the environmental characteristics of the test environment can effectively reduce the interference of external factors and improve the accuracy and reliability of the test data.

[0019] S2: Establish a test sample of the curing glue, and evenly distribute N measurement points on each test sample, place the test sample in a configured test environment, and use a spectrometer to monitor the sample timing of the test sample to establish a curing timing data set.

[0020] Specifically, after completing the configuration of the test plan and the construction of the test environment, establish the curing glue test sample for monitoring. First, determine the number and specifications of the test samples according to the test requirements, such as the size, thickness and other parameters of the samples. Then, prepare the samples according to the determined quantity and specifications. During the preparation process, strictly control the process parameters such as the ratio of the glue solution, stirring time, and degassing to ensure that the prepared samples have uniform and consistent performance.

[0021] In order to achieve accurate monitoring of the curing glue test sample, a certain number of measurement points are evenly distributed on each test sample. The number of measurement points is represented by N, which is determined according to the size of the test sample and the accuracy requirements of monitoring. Among them, the distribution of measurement points should be as uniform as possible, covering all areas of the test sample, so as to comprehensively collect the curing information of the sample.

[0022] After the preparation of the test sample and the arrangement of the measurement points are completed, the test sample is placed in the previously configured test environment under controlled temperature, humidity and light conditions. Then, the test sample is monitored in real time using a spectrometer. The spectrometer can collect the spectral signal of the test sample during the curing process, and obtain the timing information of the sample curing state by analyzing the changes in the spectral signal.

[0023] For each test sample, the spectrometer continuously collects data during the entire curing process. The time interval is set according to the requirements of the curing rate and monitoring accuracy, for example, data is collected every few seconds. The collected data includes the spectral information of each measurement point at different times, as well as the corresponding timestamp.

[0024] By continuously monitoring multiple test samples in time series, a complete curing time series data set is obtained. This data set contains the curing status information of different test samples at different time points, as well as the changes in environmental conditions, providing important basic data for subsequent data analysis and curing process modeling. During the entire monitoring process, the relative position between the spectrometer and the test sample is ensured to remain stable to avoid additional measurement errors due to position changes. At the same time, the spectrometer is calibrated regularly to ensure the accuracy and reliability of the measurement data.

[0025] By obtaining the spectral monitoring data of the curing adhesive at different time points, a complete curing time series data set is established, providing important data support for subsequent data analysis and optimization of the curing process.

[0026] S3: extracting the phase data in the solidified time series data set, and performing adaptive smoothing processing on the solidified time series data set according to the measurement distribution and time series identifiers of the N measurement points, and establishing updated phase data.

[0027] Specifically, first, phase data is extracted from the curing time series data set. The phase data reflects the phase changes of the curing glue sample at different times and is closely related to the curing state of the sample. By analyzing the changing trend of the phase data, the curing process of the curing glue sample can be understood. However, the original phase data may be affected by measurement noise and environmental interference, and there are certain fluctuations and instabilities. In order to obtain more stable and reliable phase characteristics, the phase data is smoothed.

[0028] Considering that there are N measurement points distributed on the test sample, each measurement point has corresponding phase data in the time series. In view of this situation, the spatial distribution information and time series characteristics of the measurement points are used to adaptively smooth the phase data. The basic idea of ​​adaptive smoothing is to comprehensively consider the spatial neighborhood information of the measurement point and the continuity of the time series, and perform weighted averaging on the phase data of each measurement point to reduce the influence of noise and fluctuations. Specifically, for each measurement point on the test sample, its spatial neighborhood range is determined, such as other measurement points within a certain radius centered on the measurement point. Then, the smoothed phase value of the measurement point is calculated based on the phase data and distance weight of the measurement point in the neighborhood. At the same time, for each measurement point, not only the neighborhood information at the current moment should be considered, but also the phase data at the previous and next moments should be combined to smooth the phase data using the time window. Through adaptive smoothing, high-frequency noise and abnormal fluctuations in the phase data are effectively removed, and a more stable and reliable phase feature sequence is obtained. The smoothed phase data is the updated phase data, which reflects the smoothed phase characteristics of the cured glue sample at each measurement point and time point.

[0029] By extracting stable and reliable phase features from the solidification time series data set, a data foundation is laid for subsequent phase analysis and solidification process modeling.

[0030] S4: obtaining the wrapping phase by updating the phase data, and calculating the phase difference change and the optical path difference change of the wrapping phase, and establishing the time sequence curing data of the curing glue according to the calculation results.

[0031] Specifically, first, the wrapping phase is obtained by using the updated phase data. The so-called wrapping phase means that due to the periodic limitation of phase measurement, when the actual phase change exceeds an integer multiple of a certain period, the measured phase value will undergo a periodic wrapping phenomenon and a phase jump will occur. In order to obtain continuous phase change information, the wrapping phase is unfolded. The purpose of phase unfolding is to remove the periodic jumps of the phase and restore a continuous phase change curve. Among them, phase unfolding methods include mass map method, least squares method, etc. Through phase unfolding, the continuous phase change data of the cured glue sample during the entire curing process can be obtained.

[0032] After obtaining the unfolded continuous phase data, calculate the phase difference change between adjacent time points. The phase difference reflects the phase change of the cured glue sample in a short period of time, which is closely related to the curing rate of the sample. By calculating the phase difference between each measurement point at consecutive time points, a series of phase difference data is obtained. Furthermore, according to the corresponding relationship between the optical path difference and the phase difference, the phase difference change is converted into the optical path difference change. The optical path difference refers to the difference in the distance that light propagates in the test sample, which is closely related to the refractive index and thickness of the test sample. During the glue curing process, the refractive index and thickness of the glue will change, causing a change in the optical path difference.

[0033] Through the above calculations and analysis, physical quantities reflecting the glue curing process are obtained, including changes in optical path difference, phase difference, etc., forming complete time-series curing data of the curing glue, reflecting the state changes of the test samples at different curing time points, and providing a data basis for monitoring the glue curing process.

[0034] S5: Establish a prediction fitting model through time-series solidification data and test scheme, call additional verification data to perform fitting test of the prediction fitting model, and establish fitting test results.

[0035] Specifically, the establishment of the prediction and fitting model comprehensively considers the factors of time series fixed data and test scheme. Time series fixed data provides the sample data required for modeling, including time series data such as optical path difference change and phase difference change. The test scheme provides the constraints and prior knowledge required for modeling, such as the temperature, humidity, and lighting conditions of the test environment. Among them, the prediction and fitting model can be implemented using a variety of mathematical methods and machine learning algorithms, such as polynomial fitting, neural networks, support vector machines, etc. By selecting the model structure and parameters, the time series fixed data can be effectively fitted and predicted.

[0036] In the process of establishing the prediction fitting model, the model's training data and validation data are divided. The training data is used for model parameter estimation and optimization, and the validation data is used to evaluate the generalization performance of the model. Through methods such as cross-validation, the optimal parameter settings of the model are obtained. In order to further evaluate the performance of the prediction fitting model, additional validation data is introduced to perform fitting tests on the model. Additional validation data is sample data independent of the training data and the original validation data, and is used to simulate unknown data in actual application scenarios.

[0037] During the fitting test, the additional verification data is input into the prediction fitting model to obtain the corresponding prediction results. Then, the prediction results are compared with the true values ​​of the additional verification data, and the corresponding performance indicators, such as mean square error, determination coefficient, etc., are calculated as the fitting test results, which reflect the prediction ability and generalization ability of the prediction fitting model on unknown data. By analyzing the fitting test results, the pros and cons of the prediction fitting model are comprehensively evaluated. If the fitting test results show that the prediction performance of the model is good and meets the needs of practical applications, the model can be considered to be effective and reliable. On the contrary, if the fitting test results are not ideal, the model is further improved and optimized, such as adjusting the model structure, increasing the amount of training data, introducing regularization terms, etc.

[0038] S6: Determine whether the fitting test result meets a preset test threshold. When the fitting test result meets the preset test threshold, the curing monitoring of the curing glue is completed through the prediction fitting model.

[0039] Specifically, the obtained fitting test results include various performance indicators, such as mean square error, determination coefficient, etc. In order to evaluate whether the prediction fitting model meets the requirements of practical application, the fitting test results are compared with the preset test threshold. The preset test threshold is a set of target values ​​or ranges of performance indicators set according to the actual needs and empirical knowledge of adhesive curing monitoring, reflecting the minimum performance level that the prediction fitting model needs to achieve in practical application, such as the prediction error does not exceed a certain value, the determination coefficient is greater than a certain value, etc.

[0040] When judging whether the fitting test results meet the preset test threshold, compare the actual value of each performance indicator with the corresponding threshold one by one. If all performance indicators meet the requirements of the preset test threshold, it means that the performance of the predictive fitting model has reached the standard of practical application and can be used to monitor the curing of the curing glue. When the performance of the predictive fitting model meets the preset test threshold, it is put into the actual glue curing monitoring task. During the monitoring process, by collecting the real-time curing data of the glue sample and inputting it into the predictive fitting model, the progress of the glue curing process is predicted and tracked in real time. The predictive fitting model outputs various curing process parameters, such as changes in optical path difference, phase difference, etc., as well as key nodes of the curing process, such as curing start time, curing end time, maximum shrinkage rate, etc., to achieve contactless glue curing monitoring.

[0041] Further, such as Figure 2 As shown, the embodiment of the present application also includes:

[0042] Acquire sample data of the test sample, extract sample features through the sample data, and establish a sample feature set;

[0043] Performing a credible evaluation of the measurement distribution of N measurement points through the sample feature set, and establishing a credible evaluation result;

[0044] Establishing a measurement distribution evaluation network based on the credible evaluation result, and performing a first adaptive smoothing process on the solidified time series data set through the measurement distribution evaluation network;

[0045] The updated phase data is established according to the first adaptive smoothing result.

[0046] In a preferred embodiment, in order to realize the adaptive smoothing of the solidification time series data set, first, the sample data of the test sample is obtained, including various attribute information of the test sample, such as the material type, size specification, preparation process, etc. of the sample. On the basis of obtaining the sample data, the sample feature extraction is performed. The sample feature refers to a set of quantitative indicators that can reflect the essential attributes and characteristics of the test sample. By analyzing the sample data, a series of relevant sample features are extracted for subsequent measurement distribution credibility evaluation and adaptive smoothing. Among them, the sample features can be extracted and defined from multiple aspects. For example, the position importance feature of the measurement point is considered to characterize the importance and representativeness of the measurement point in the overall structure of the sample; the structural features of the sample are considered, such as the shape, size, internal structure, etc. of the sample; the glue coating features of the sample are considered, such as the thickness distribution of the glue coating layer, surface roughness, etc. By comprehensively and systematically extracting various features of the sample, a complete sample feature set is established, which contains rich sample attribute information, and provides an important data basis for the subsequent measurement distribution credibility evaluation.

[0047] After obtaining the sample feature set, the credibility of the measurement distribution of N measurement points is evaluated. The credibility of the measurement distribution reflects the representativeness and importance of the measurement point in the overall structure of the sample, as well as the reliability of the measurement data. The purpose of evaluating the credibility of the measurement distribution is to identify the key measurement points that have a greater impact on the curing process, as well as the abnormal measurement points that may have measurement errors or interference. By evaluating and screening the measurement points, targeted data smoothing and correction are performed to improve the reliability and accuracy of the monitoring data. Based on the sample feature set, the credibility of the measurement distribution of N measurement points is evaluated, and a credible evaluation result is established. For example, for measurement points with high position importance, their measurement data have a greater impact on reflecting the changing trend of the overall curing process, so the credibility is relatively high; for measurement points at the edge of the sample or in irregular areas, their measurement data may be interfered by factors such as boundary effects, and the credibility is relatively low. By comprehensively analyzing the sample feature set and the credibility evaluation of the measurement distribution, a complete credible evaluation result is established, which reflects the credibility of the data of each measurement point and provides an important reference for subsequent adaptive smoothing processing.

[0048] After obtaining the measurement distribution credible evaluation results, a measurement distribution evaluation network is further established to perform adaptive smoothing on the solidified time series data set. The measurement distribution evaluation network realizes adaptive smoothing and correction of the measurement data by organizing and weighing the topological relationship and data correlation between the measurement points. In the process of establishing the measurement distribution evaluation network, the prior knowledge provided by the credible evaluation results is fully utilized. Specifically, the node weights and connection strengths in the network are set according to the credibility of the measurement points. For measurement points with high credibility, the corresponding network nodes should be given larger weights and play a leading role in the smoothing process; while for measurement points with low credibility, the corresponding network nodes should be given smaller weights to reduce their influence on the smoothing results. In addition, the measurement distribution evaluation network can also consider the spatial correlation and temporal correlation between measurement points. Spatial correlation characterizes the geometric position relationship and data similarity between different measurement points; temporal correlation reflects the data change trend and continuity of the same measurement point at different times. By characterizing and utilizing these correlations, data smoothing and feature extraction can be achieved more accurately and efficiently.

[0049] Then, based on the established measurement distribution evaluation network, the solidified time series data set is subjected to the first adaptive smoothing process. Specifically, through the information transmission and fusion between network nodes, data smoothing on local and global scales is achieved, the influence of measurement noise and outliers is suppressed, and the main change trends and characteristic patterns of the measurement data are extracted. In actual implementation, the measurement distribution evaluation network can be adopted in various forms, such as graph theory model, neural network, Gaussian process, etc. By selecting the network structure and parameters, the optimal smoothing effect can be achieved in different application scenarios. After the first adaptive smoothing process, the solidified time series data after preliminary processing is obtained. In order to further improve the data quality, the updated phase data is established on this basis. The updated phase data refers to the secondary processing and optimization of the phase data based on the result of the first smoothing process, combined with other prior knowledge and constraints. By updating the phase data, the residual measurement errors and inconsistencies are eliminated, the smoothness and continuity of the phase data are improved, and a high-quality data foundation is provided for the subsequent phase unwrapping and optical path difference calculation.

[0050] By extracting the sample feature set, evaluating the credibility of the measurement distribution, establishing a measurement distribution evaluation network, and performing the first adaptive smoothing process and establishing updated phase data, the quality and reliability of the curing monitoring data can be effectively improved, providing high-quality data support for subsequent phase unwrapping, optical path difference calculation and other analyses, thereby achieving more accurate and stable online monitoring of glue curing.

[0051] Furthermore, the embodiment of the present application also includes:

[0052] Performing feature normalization conversion on the sample feature set to establish a normalized conversion result;

[0053] Establishing a main component screening channel, wherein the main component screening channel includes a threshold screening sub-channel and an additional authentication sub-channel;

[0054] Inputting the normalized conversion result into the principal component screening channel, and determining the last key feature through the threshold screening sub-channel;

[0055] Synchronize the last key feature to the additional authentication subchannel, and perform additional authentication of neighborhood features of the last key feature through the additional authentication subchannel to generate an additional authentication result;

[0056] Key features are output according to the additional authentication results and the normalized conversion results, and a credible evaluation result is established with the key features.

[0057] In a feasible implementation, when performing a reliable evaluation of the measurement distribution, the sample feature set is first preprocessed. Due to the large differences in the dimensions and numerical ranges of different features, directly using the original eigenvalues ​​for evaluation will lead to imbalance and deviation in the results. Therefore, through feature normalization conversion, the values ​​of different features are mapped to the same scale to facilitate subsequent analysis and comparison. The specific method of feature normalization conversion can be selected according to actual needs and data characteristics. For example, the normalization method used includes maximum and minimum value normalization, zero mean normalization, decimal calibration normalization, etc. Through normalization, each eigenvalue in the sample feature set is uniformly mapped to a fixed interval such as [0, 1] or [-1, 1] to eliminate the dimensional influence between the features and maintain the relative size relationship of the eigenvalues. After the normalization conversion, a normalized sample feature set is obtained, which is the normalization conversion result. Each eigenvalue in the normalization conversion result is in the same numerical range, which is convenient for subsequent feature screening and evaluation.

[0058] After obtaining the normalized conversion result, the most important and critical feature subset for the credibility evaluation of the measurement distribution is screened out from the numerous sample features, which is called the principal component feature. In order to achieve efficient screening of the principal component features, a principal component screening channel is established. The principal component screening channel consists of two parallel sub-channels, namely the threshold screening sub-channel and the additional authentication sub-channel. The two sub-channels screen and evaluate the normalized conversion results from different angles to obtain comprehensive principal component features. The threshold screening sub-channel directly screens out the features that meet the conditions from the normalized conversion results according to the preset threshold conditions. Specifically, one or more threshold parameters are set, such as the upper and lower limits of the feature value, the variance contribution rate, etc. By comparing the normalized conversion result with the threshold condition, the key features that have a greater impact on the credibility evaluation of the measurement distribution are quickly located and extracted. On the basis of threshold screening, the additional authentication sub-channel further considers the mutual relationship and constraints between features to optimize and supplement the screening results. Specifically, domain knowledge and empirical rules, such as the correlation between features and physical constraints, are introduced to conduct secondary evaluation and authentication of the features obtained by threshold screening. Through additional authentication, redundant or invalid features are eliminated, and missing key features are supplemented to improve the comprehensiveness and accuracy of the principal component features.

[0059] After the main component screening channel is established, the normalized transformation result is used as input and firstly screened through the threshold screening sub-channel. The threshold screening sub-channel quickly locates and extracts the key feature subsets that have a greater impact on the credibility evaluation of the measurement distribution according to the preset threshold conditions. The screened feature subsets are sorted according to the degree of importance, and the feature with the lowest importance is called the last key feature. The last key feature refers to one or several features with the lowest importance score in the feature subset screened by the threshold screening sub-channel. Although these features are relatively low in importance in the current screening results, they may still contain some potential effective information and need further analysis and verification. After the last key feature is determined through the threshold screening sub-channel, it is synchronously transferred to the additional authentication sub-channel for further analysis and optimization. The purpose of the additional authentication sub-channel is to conduct a secondary evaluation and supplement of the last key feature by introducing the mutual relationship and constraint conditions between the features, so as to generate a more comprehensive and accurate additional authentication result. The additional authentication sub-channel is used to perform additional authentication of the neighborhood features of the last key feature. Neighborhood features refer to other features that are adjacent to or related to the last key feature in the sample feature space. These features have certain semantic or physical correlations with the last key feature and contain potential effective information. The neighborhood feature additional authentication is centered on the last key feature, and searches and identifies important features associated with it in the feature subspace around it. By analyzing the relationship between the last key feature and the neighborhood features, such as correlation, dependency, causality, etc., the key features missed by the threshold screening are found, or the importance of the last key feature is corrected and supplemented. Among them, the specific implementation of the neighborhood feature additional authentication can adopt a variety of methods, such as correlation analysis, cluster analysis, principal component analysis, etc. By calculating the correlation coefficient, distance measurement and other indicators between the last key feature and other features, the similarity and dependency between the features can be quantitatively characterized.

[0060] After obtaining the additional authentication results of the additional authentication subchannel, it is necessary to conduct a comprehensive analysis with the normalized transformation results, and finally output a set of key features as the basis for the credibility evaluation of the measurement distribution. Specifically, first, the feature subset obtained by the threshold screening subchannel screening is extracted from the normalized transformation results as the candidate set of key features. Then, according to the authentication results of the additional authentication subchannel, the features in the candidate set are further screened and optimized. By analyzing the neighborhood features and constraint relationships in the additional authentication results, the features in the candidate set are ranked and selected in terms of importance. For features that are strongly correlated with multiple neighborhood features and meet important constraints, their importance scores will be increased accordingly; while for features that are weakly correlated with neighborhood features and do not meet constraints, their importance scores will be reduced accordingly. According to the importance scores of the features, several features with the highest scores are selected from the candidate set as the final key features output. These key features combine the evaluation information of threshold screening and additional authentication, and can reflect the credibility and key influencing factors of the measurement distribution to the greatest extent. Based on the key features, the credibility evaluation results of the measurement distribution are further established, which describes the credibility and importance ranking of the data of each measurement point. By analyzing the credible evaluation results, the key measurement points that have a greater impact on the curing process and the abnormal measurement points that may have measurement errors or interference are identified, providing an important reference for the subsequent adaptive smoothing processing.

[0061] Furthermore, the embodiment of the present application also includes:

[0062] Performing data timing alignment of the first adaptive smoothing processing result according to the timing identifier;

[0063] Adaptive filtering under long and short timing is performed by aligning the data timing, as follows:

[0064] ;

[0065] ;

[0066] ;

[0067] in, For the The comprehensive filtering phase value at the moment, is the impact weight of the short-term time series, Characterization The short-term smoothed phase value at time , Characterization The long-term smoothed phase value at time , Characterizes the length of the short-term time window, is the length of the long-term time window, For the The data credibility score at each moment, Characterization The phase value after phase unwrapping at the time;

[0068] The phase data is updated based on the adaptive filtering under long and short time sequences.

[0069] In a preferred embodiment, the updated phase data is established by further processing the result of the first adaptive smoothing process.

[0070] Since the first adaptive smoothing process involves data from different measurement points, and there is a certain deviation or inconsistency in the data collection time of each measurement point, data timing alignment is required. The purpose of data timing alignment is to adjust the smoothing results of different measurement points to the same time base to ensure the synchronization of phase data in the time dimension. Data timing alignment is achieved through timing identification, which refers to the timestamp or time index associated with the data of each measurement point. By comparing the timing identifications of different measurement points, the relative position and sequence of each data in the time dimension are determined.

[0071] After completing the data timing alignment, the aligned phase data is further filtered to eliminate residual noise and interference and improve the smoothness and stability of the phase data. Among them, the adaptive filtering method under long and short time series is adopted. The basic idea of ​​long and short time series adaptive filtering is to comprehensively consider the changing characteristics of phase data at different time scales, and obtain a smoother and more stable phase estimate through weighted averaging.

[0072] Specifically, for the The phase value at the moment is first calculated by smoothing its values ​​in the short-term time window and the long-term time window, respectively, which are recorded as and The length of the short-term time window is , the length of the long-term time window is ,satisfy Less than .

[0073] Short-term smoothed phase value The calculation formula is:

[0074] ;

[0075] in, Indicates The phase value after phase unwrapping at the moment, Indicates Data credibility score at all times.

[0076] Long-term smoothed phase value The calculation formula is:

[0077] ;

[0078] Among them, the short-term smoothed phase value takes into account the changing trend of the phase data in the short term, while the long-term smoothed phase value takes into account the overall changing trend of the phase data in a longer time range.

[0079] After obtaining the short-term smoothed phase value and the long-term smoothed phase value, the first The comprehensive filtering phase value at time , the calculation formula is:

[0080] ;

[0081] in, is the impact weight of short-term time series, and its value range is [0, 1]. The value of determines the relative importance of the short-term change trend and the long-term change trend in the comprehensive filtering phase value. When it is larger, it means that more attention is paid to the short-term variation characteristics of the phase data; when When it is smaller, it means that more attention is paid to the long-term change characteristics of the phase data. Through long and short time series adaptive filtering, the phase data is smoothed and stabilized on different time scales, which effectively suppresses random noise and interference and improves the signal-to-noise ratio and reliability of the phase data.

[0082] After long and short time series adaptive filtering, a smoother and more stable phase data sequence is obtained, which is denoted as { }, where N is the total number of moments. The phase data sequence after filtering is the updated phase data. Compared with the first adaptive smoothing result, the updated phase data has higher smoothness, stability and reliability, can better reflect the real change trend of the phase during the glue curing process, effectively eliminates residual noise and interference, and improves the quality and consistency of the phase data.

[0083] Furthermore, the embodiment of the present application also includes:

[0084] Configure the calculation formula, which includes the calculation formula for optical path difference change and phase and the calculation formula for shrinkage strain and optical path difference change:

[0085] ∆L t = l c 4π unwrap{diff[ ϕ j k ， t ， ϕ j k ， t 0 ]} ;

[0086] ;

[0087] in, Characterization in curing time The optical path difference changes under is the central wavelength of the light source, Characterize the phase unwrapping, diff[] Characterize the phase difference calculation, Characterize the initial time, Indicates the curing time, Characterize the channel, Characterized in the Channel and Location time The phase value of Characterized in the Channel and Location time The phase value of Characterizes the shrinkage strain, is the thickness of the cured adhesive;

[0088] The calculation results are obtained according to the calculation formula to establish the time sequence curing data of the curing glue.

[0089] In a preferred embodiment, the phase difference change and the optical path difference change of the wrapping phase are calculated, and the time-series curing data of the curing glue are established according to the calculation results.

[0090] First, calculation formulas are configured, including a calculation formula for optical path difference change and phase and a calculation formula for contraction strain and optical path difference change, which are used to describe the relationship between optical path difference change and phase, and the relationship between contraction strain and optical path difference change, respectively.

[0091] The calculation formula of optical path difference change and phase is:

[0092] ∆L t = l c 4π unwrap{diff[ ϕ j k ， t ， ϕ j k ， t 0 ]} ;

[0093] in, Indicates the curing time The optical path difference changes under is the central wavelength of the light source, represents the phase unwrapping operation, diff[] represents the phase difference calculation, Indicates the initial time, Indicates the curing time, Indicates the channel, Indicated in Channel and Location time The phase value of Indicated in Channel and Location time The formula describes the quantitative relationship between the optical path difference change and the phase. Specifically, the optical path difference change is obtained by calculating the phase difference and the phase unwrapping operation. The phase difference calculation means that the curing time is calculated under the same channel and position. With initial time The phase unwrapping operation is to expand the phase difference result into a continuous phase domain, eliminating The continuous phase change during the curing process is obtained by phase difference calculation and phase unwrapping, and then multiplied by the wavelength factor , converted into the corresponding optical path difference change.

[0094] The calculation formula for shrinkage strain and optical path difference change is:

[0095] ;

[0096] in, represents the shrinkage strain, Indicates the curing time The optical path difference changes under is the thickness of the cured adhesive. This formula describes the quantitative relationship between shrinkage strain and optical path difference change. Shrinkage strain refers to the strain change caused by the volume shrinkage of the adhesive material during the adhesive curing process. According to the ratio of the optical path difference change to the adhesive thickness, the size of the shrinkage strain can be directly calculated.

[0097] By configuring the above two calculation formulas, a quantitative correlation between the optical path difference change, phase difference change and shrinkage strain is established, providing a theoretical basis for the subsequent calculation and analysis of the time-series curing data. After configuring the calculation formula, the wrapped phase data can be calculated step by step according to the formula definition to obtain the time-series data of the optical path difference change and shrinkage strain. Specifically, for each measurement channel and position, firstly, the phase difference value at different curing times is calculated according to the phase difference calculation formula; then, the phase difference value is phase-unwrapped to obtain a continuous phase change; then, according to the calculation formula of the optical path difference change and the phase, the phase change is converted into the corresponding optical path difference change; then, according to the calculation formula of the shrinkage strain and the optical path difference change, the corresponding shrinkage strain value is calculated. By traversing all measurement channels and positions and repeating the above calculation process at different curing times, the complete time-series curing data, including the optical path difference change data and the shrinkage strain data, can be obtained. Each data point of the time-series curing data corresponds to the optical path difference change value and the shrinkage strain value of a certain measurement channel and position at a certain curing moment. By analyzing and visualizing the time-series curing data, the dynamic changes of the optical path difference and shrinkage strain during the adhesive curing process can be intuitively presented, providing important data support for monitoring the adhesive curing process.

[0098] Furthermore, the embodiment of the present application also includes:

[0099] Perform sample richness evaluation of internal data and establish sample richness evaluation results;

[0100] Balanced data selection is performed on internal data according to the sample richness evaluation result, additional verification data is constructed based on the balanced data selection result, and the fitting test of the prediction fitting model is performed through the additional verification data.

[0101] In a preferred embodiment, sample richness evaluation of internal data and balanced data selection are introduced to complete the construction and fitting test of additional verification data.

[0102] Before conducting the fitting test of the predictive fitting model, first, evaluate the sample richness of the internal data. Internal data refers to the original data set used to train and optimize the predictive fitting model, which comes from the actual glue curing monitoring experiment and data collection. The purpose of sample richness evaluation is to quantitatively evaluate the coverage and representativeness of internal data in the sample space. The ideal internal data set should cover the different process conditions, material formulations, environmental factors and other influencing factors of the glue curing process as comprehensively as possible, and be evenly distributed within the value range of each influencing factor to ensure the generalization performance and robustness of the predictive fitting model. Sample richness evaluation is achieved through a variety of quantitative indicators, such as sample density, sample distribution entropy, sample diversity index, etc. These indicators characterize the distribution characteristics of internal data in the sample space from different angles, reflecting the richness and representativeness of the sample. Among them, the sample density index is measured by calculating the number of samples in the unit sample space volume. The higher the density, the higher the coverage of the samples in the area; the sample distribution entropy index is measured by calculating the distribution probability of samples in different value intervals. The larger the entropy value, the more uniform the distribution of samples in each interval; the sample diversity index is measured by calculating the types of values ​​of samples in different influencing factors. The higher the index, the richer the combination of influencing factors covered by the sample. By comprehensively evaluating multiple sample richness indicators, a comprehensive internal data sample richness evaluation result is obtained, which describes the coverage of internal data in the sample space, the uniformity of distribution, and the richness of influencing factors, providing an important reference for the subsequent balanced data selection and additional verification data construction.

[0103] After obtaining the sample richness evaluation results of the internal data, the internal data is further balanced to construct additional validation data. The additional validation data is an additional data set used to evaluate the generalization performance of the predictive fitting model. The purpose of balanced data selection is to select a subset of samples with balanced and reasonable distribution from the internal data, so that its coverage and representativeness in the sample space are equivalent to the entire internal data set, while not over-repeating or missing certain specific sample areas. Balanced data selection can be achieved using a variety of strategies and algorithms, such as stratified sampling, cluster sampling, and synthetic sampling. Balanced data selection groups, divides, and reorganizes the internal data based on the sample richness evaluation results to obtain a balanced data subset. Among them, the stratified sampling strategy divides the internal data into several mutually exclusive subsets according to the value range of the influencing factors, and then extracts samples from each subset in proportion to ensure that the distribution of the selected samples on each influencing factor is consistent with the internal data; the cluster sampling strategy divides the internal data into several clusters with high similarity through a clustering algorithm, and then extracts representative samples from each cluster to cover the main distribution pattern of the internal data; the synthetic sampling strategy generates new samples based on the internal data through methods such as artificial synthesis or interpolation to make up for the sparse or missing samples in some areas of the internal data. Through balanced data selection, a subset with balanced and reasonably distributed samples is obtained from the internal data, which not only retains the main characteristics and distribution laws of the internal data, but also avoids the problem of excessive concentration or sparseness of samples, and improves the representativeness and effectiveness of the additional verification data. After obtaining the balanced data subset, it can be used as additional verification data for fitting test of the prediction fitting model. Through the fitting test of the additional verification data, the actual performance and scope of application of the prediction fitting model are comprehensively evaluated. If the fitting test results show that the model can predict the additional validation data well, it means that the model has good generalization ability and can provide reliable prediction results in practical applications. On the contrary, if the fitting test results are not ideal, it is necessary to further optimize the model structure, parameters or training algorithm to improve the fitting performance and robustness of the model.

[0104] Furthermore, the embodiment of the present application also includes:

[0105] If the sample richness evaluation result fails to meet the expected richness threshold, a call request is generated;

[0106] Encrypted data is called from a trusted third party through the calling requirement, data screening is performed based on the encrypted data and internal data, and additional verification data is established based on the data screening result.

[0107] In a feasible implementation, after evaluating the sample richness of the internal data, the evaluation result is compared with the expected richness threshold. The expected richness threshold is a quantitative indicator for measuring the adequacy of internal data, which is set according to the complexity of the glue curing monitoring problem and the requirements for model performance. If the sample richness evaluation result is lower than the expected richness threshold, it means that the current internal data has insufficient coverage and representativeness in the sample space, and it is difficult to support the training and verification of high-quality predictive fitting models. In this case, relying solely on internal data to construct additional verification data will affect the evaluation and optimization of model performance. Therefore, when the sample richness of the internal data does not meet the requirements, a call request is generated. A call request is a signal or instruction to request additional data support from an external data source, which contains attribute information such as the type, format, and quantity of the required data.

[0108] After the call demand is generated, a data call request is automatically sent to an external trusted third party according to the content of the demand. A trusted third party refers to a data provider with good reputation and security in the process of data sharing and exchange, such as a professional glue curing data service platform, industry association, scientific research institution, etc. After receiving the data call request, the external trusted third party searches and extracts the relevant glue curing monitoring data in its database according to the content of the request. In order to protect the privacy and security of the data, the provided data is usually encrypted to form an encrypted data packet. The encrypted data packet contains the characteristic information of the original monitoring data, but hides the specific value and source to ensure the confidentiality of the data during transmission and use. After receiving the encrypted data packet returned by the trusted third party, it is merged with the internal data to form an expanded data set. In order to select the most valuable and representative samples from the expanded data set and construct high-quality additional verification data, the merged data is further screened and optimized, so that a subset with balanced samples, reasonable distribution and rich information is selected from the expanded data set as additional verification data.

[0109] Furthermore, the embodiment of the present application also includes:

[0110] Establishing a verification node, and verifying the monitoring results of the prediction fitting model through the verification node;

[0111] Verifying and generating verification feedback according to the monitoring results, and conducting a verification credibility evaluation on the verification feedback;

[0112] Adaptive optimization management of the prediction fitting model is performed by verifying the credible evaluation results.

[0113] In a feasible implementation, in the process of using the predictive fitting model to monitor the adhesive curing process, in order to ensure the accuracy and reliability of the model output results, the monitoring results of the model are continuously verified and evaluated.

[0114] First, a verification node is established to realize dynamic verification of the monitoring results of the prediction fitting model. The verification node refers to the measurement point set in the glue curing production line for comparing and verifying the model prediction results. By collecting the actual glue curing process parameters and quality data at the verification node and comparing them with the output results of the prediction fitting model, the accuracy and degree of deviation of the model prediction are evaluated. Specifically, the error indicators between the model prediction value and the actual measured value of the verification node are calculated, such as absolute error, relative error, mean square error, etc., to quantitatively characterize the model prediction performance.

[0115] By comparing the output of the prediction fitting model with the actual process data at the verification node, verification feedback of the monitoring results is generated. Verification feedback is a comprehensive evaluation of the model prediction performance, reflecting the effectiveness and limitations of the model in practical applications. The content of verification feedback includes many aspects of information, such as statistical indicators of model prediction error, distribution characteristics of prediction deviation, evaluation of model generalization ability, etc. By analyzing the verification feedback, problems and deficiencies of the model, such as overfitting, underfitting, and prediction inaccuracy under specific working conditions, are found, providing an important basis for subsequent model optimization. In order to ensure the reliability and effectiveness of the verification feedback, the verification feedback is verified and trusted. Verification and trusted evaluation refers to the analysis and evaluation of the quality and credibility of the verification feedback to identify possible abnormal or misleading information. Verification and trusted evaluation is carried out from multiple dimensions, such as the reliability of the data source, the representativeness of the verification node, and the rationality of the comparison method. By comprehensively considering various influencing factors, the confidence level and uncertainty of the verification feedback are evaluated to provide a more reliable reference for subsequent model optimization decisions. Specifically, the verification credibility evaluation adopts methods such as expert experience judgment, statistical hypothesis testing, confidence interval estimation, etc., and automatically scores and screens the credibility of verification feedback by setting evaluation indicators and thresholds, identifies high-quality and highly credible feedback information, and obtains verification credibility evaluation results to guide the iterative optimization of the model.

[0116] After obtaining credible evaluation results of verification feedback, the prediction fitting model is adaptively optimized and managed using high-quality feedback information. The adaptive optimization of the model refers to dynamically adjusting the structure, parameters and training methods of the model according to the feedback data in the actual application, so that it can better adapt to the changes and development of the production process. The specific implementation of model adaptive optimization can adopt a variety of strategies and algorithms, such as incremental learning, online learning, active learning, etc., so as to use new verification data to partially update and fine-tune the model without completely retraining the model, thereby improving the real-time and adaptability of the model. For example, when the verification feedback shows that the model has a large deviation in certain specific working conditions or quality indicators, relevant supplementary data can be collected in a targeted manner to optimize the corresponding part of the model; when the verification feedback reveals a downward trend in the overall prediction performance of the model, the comprehensive retraining and upgrading of the model can be triggered in a timely manner, and new algorithms and features can be introduced to improve the robustness and generalization ability of the model.

[0117] By setting verification nodes, generating and evaluating verification feedback, and dynamically optimizing and managing the model based on reliable feedback information, the applicability, reliability and robustness of the predictive fitting model in actual glue curing monitoring can be effectively improved, achieving continuous evolution and improvement of model performance.

[0118] In summary, the adhesive curing monitoring method based on optical path difference change provided in the embodiment of the present application has the following technical effects:

[0119] Configure the test plan, establish the test environment based on the test plan, and configure the environmental characteristics of the test environment. The environmental characteristics include environmental temperature characteristics, environmental humidity characteristics, and light source characteristics to ensure that the glue curing monitoring is carried out under specific conditions, eliminate the interference of environmental factors, and improve the consistency and repeatability of the test results. Establish a test sample for curing glue, and evenly distribute N measurement points on each test sample. Place the test sample in the configured test environment, and use the spectrometer to perform sample timing monitoring of the test sample to establish a curing timing data set. The multi-point arrangement can comprehensively collect the curing information on the surface and inside of the colloid, and the timing monitoring records the dynamic evolution of the glue curing process. Extract the phase data in the curing timing data set, and perform adaptive smoothing of the curing timing data set based on the measurement distribution and timing identification of the N measurement points to establish updated phase data. The phase data contains the change information of the optical path difference during the glue curing process. Through smoothing, the random noise in the measurement process can be effectively eliminated, and the phase change characteristics of glue curing can be highlighted. The package phase is obtained by updating the phase data, and the phase difference change and optical path difference change of the package phase are calculated. The time-series curing data of the curing glue are established according to the calculation results. Through the analysis of the optical path difference change, the key performance indicators such as the curing degree and shrinkage strain of the colloid are accurately evaluated. The prediction fitting model is established through the time-series curing data and the test plan, and the additional verification data is called to perform the fitting test of the prediction fitting model, and the fitting test results are established. Based on the monitoring data in the initial stage, the subsequent progress of the glue curing is predicted, providing a basis for production control. It is judged whether the fitting test results meet the preset test threshold. When the fitting test results meet the preset test threshold, the glue curing monitoring of the curing glue is completed through the prediction fitting model. When the model accuracy reaches the predetermined requirements, it can be applied to the online monitoring and prediction of glue curing in actual production, guiding production optimization and quality control, and realizing non-destructive, accurate and comprehensive online monitoring and prediction of the glue curing process.

[0120] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0121] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these changes and variations.

Claims

1. A method for monitoring adhesive curing based on optical path difference change, characterized in that: The method comprises: Configure a test plan, establish a test environment based on the test plan, and configure environmental characteristics of the test environment, wherein the environmental characteristics include environmental temperature characteristics, environmental humidity characteristics, and light source characteristics; Establish a test sample of the curing glue, and evenly distribute N measurement points on each test sample. Place the test sample in a configured test environment, and use a spectrometer to monitor the sample timing of the test sample to establish a curing timing data set. Extract the phase data in the solidified time series data set, and perform adaptive smoothing processing on the solidified time series data set according to the measurement distribution and time series identification of N measurement points to establish updated phase data; The wrapping phase is obtained by updating the phase data, and the phase difference change and the optical path difference change of the wrapping phase are calculated, and the time sequence curing data of the curing glue is established according to the calculation results; Establish a prediction fitting model through time series solidification data and test scheme, call additional verification data to perform fitting test of the prediction fitting model, and establish fitting test results; Determining whether the fitting test result meets a preset test threshold, and when the fitting test result meets the preset test threshold, completing the glue curing monitoring of the curing glue through the prediction fitting model; Adaptive smoothing of the solidified timing data set is performed according to the measurement distribution of N measurement points and the timing test results to establish updated phase data, which also includes: Acquire sample data of the test sample, extract sample features through the sample data, and establish a sample feature set; Performing a credible evaluation of the measurement distribution of N measurement points through the sample feature set, and establishing a credible evaluation result; Establishing a measurement distribution evaluation network based on the credible evaluation result, and performing a first adaptive smoothing process on the solidified time series data set through the measurement distribution evaluation network; Establishing updated phase data according to the first adaptive smoothing processing result; The reliable evaluation of the measurement distribution of N measurement points is performed by using the sample feature set to establish a reliable evaluation result, and further includes: Performing feature normalization conversion on the sample feature set to establish a normalized conversion result; Establishing a main component screening channel, wherein the main component screening channel includes a threshold screening sub-channel and an additional authentication sub-channel; Inputting the normalized conversion result into the principal component screening channel, and determining the last key feature through the threshold screening sub-channel; Synchronize the last key feature to the additional authentication subchannel, and perform additional authentication of neighborhood features of the last key feature through the additional authentication subchannel to generate an additional authentication result; Outputting key features according to the additional authentication results and the normalized conversion results, and establishing a credible evaluation result with the key features; Establishing updated phase data according to the first adaptive smoothing result also includes: Performing data timing alignment of the first adaptive smoothing processing result according to the timing identifier; Adaptive filtering under long and short timing is performed by aligning the data timing, as follows: ； ； ； in, For the The comprehensive filtering phase value at the moment, is the impact weight of the short-term time series, Characterization The short-term smoothed phase value at time , Characterization The long-term smoothed phase value at time , Characterizes the length of the short-term time window, is the length of the long-term time window, For the The data credibility score at each moment, Characterization The phase value after phase unwrapping at the time; Establishing and updating phase data based on adaptive filtering under long and short time sequences; Calculate the phase difference change and optical path difference change of the package phase, and establish the time sequence curing data of the curing glue according to the calculation results, which also includes: Configure the calculation formula, which includes the calculation formula for optical path difference change and phase and the calculation formula for shrinkage strain and optical path difference change: ； ； in, Characterization in curing time The optical path difference changes under is the central wavelength of the light source, Characterize the phase unwrapping, Characterize the phase difference calculation, Characterize the initial time, Indicates the curing time, Characterize the channel, Characterized in the Channel and Location time The phase value of Characterized in the Channel and Location time The phase value of Characterizes the shrinkage strain, is the thickness of the cured adhesive; The calculation results are obtained according to the calculation formula to establish the time sequence curing data of the curing glue.

2. The adhesive curing monitoring method based on optical path difference change according to claim 1, characterized in that: The calling of the additional verification data to perform a fitting test on the prediction fitting model and establishing a fitting test result also includes: Perform sample richness evaluation of internal data and establish sample richness evaluation results; Balanced data selection is performed on internal data according to the sample richness evaluation result, additional verification data is constructed based on the balanced data selection result, and the fitting test of the prediction fitting model is performed through the additional verification data.

3. The adhesive curing monitoring method based on optical path difference change according to claim 2, characterized in that: The method further comprises: If the sample richness evaluation result fails to meet the expected richness threshold, a call request is generated; Encrypted data is called from a trusted third party through the calling requirement, data screening is performed based on the encrypted data and internal data, and additional verification data is established based on the data screening result.

4. The adhesive curing monitoring method based on optical path difference change according to claim 1, characterized in that: The method further comprises: Establishing a verification node, and verifying the monitoring results of the prediction fitting model through the verification node; Verifying and generating verification feedback according to the monitoring results, and conducting a verification credibility evaluation on the verification feedback; Adaptive optimization management of the prediction fitting model is performed by verifying the credible evaluation results.

Images