CN120703003A

CN120703003A - A glass production line cold end stress detection system and method

Info

Publication number: CN120703003A
Application number: CN202510871131.5A
Authority: CN
Inventors: 郭建
Original assignee: Nantong Xinzhou Glass Co ltd
Current assignee: Nantong Xinzhou Glass Co ltd
Priority date: 2025-06-26
Filing date: 2025-06-26
Publication date: 2025-09-26

Abstract

The present invention proposes a cold-end stress detection system and method for a glass production line. This method belongs to the field of glass production detection technology. The method comprises: collecting stress-related spectral data from the cold-end region of the glass production line using a multispectral stress sensing array, preprocessing the collected spectral data to obtain a preprocessed spectral dataset; extracting spectral features from the preprocessed spectral dataset to construct a spectral feature vector set; and extracting the spatiotemporal features of the glass cold-end stress using a spatiotemporal correlation analysis algorithm based on the spectral feature vector set to obtain a stress spatiotemporal feature matrix. By collecting spectral data using a multispectral stress sensing array, the impact of traditional contact detection on glass products and production lines is avoided, thereby improving detection accuracy and real-time performance.

Description

Cold end stress detection system and method for glass production line

Technical Field

The invention provides a cold end stress detection system and method for a glass production line, and belongs to the technical field of glass production detection.

Background

In the glass production process, the cold end link is critical to the quality of the glass. Stress can be generated in the cooling process of the glass, and if the stress is unevenly distributed or excessively large, the problems of cracking, breaking and the like of the glass can be caused, so that the quality and the production efficiency of the product are seriously affected. At present, the existing glass cold end stress detection method has the problems that the detection precision is not high, the stress distribution situation cannot be comprehensively reflected in real time, the stress abnormal region is difficult to accurately locate, and the like, and the requirements of modern glass production on high quality and high efficiency detection are difficult to meet. Therefore, the development of the high-efficiency and accurate cold end stress detection method for the glass production line has important practical significance.

Disclosure of Invention

The invention provides a cold end stress detection system and method for a glass production line, which are used for solving the problems mentioned in the background art:

the invention provides a cold end stress detection method of a glass production line, which is characterized by comprising the following steps of:

S1, acquiring stress related spectrum data of a cold end area of a glass production line through a multispectral stress sensing array, and preprocessing the acquired spectrum data to obtain a preprocessed spectrum data set;

s2, extracting spectral features of the preprocessed spectral data set to construct a spectral feature vector set, and extracting space-time features of the cold end stress of the glass by using a space-time correlation analysis algorithm based on the spectral feature vector set to obtain a stress space-time feature matrix;

S3, collecting historical glass cold end stress abnormality case data, performing feature cluster analysis on the stress space-time feature matrix to obtain a plurality of stress feature clusters, combining the historical glass cold end stress abnormality case data, performing abnormal feature matching on each stress feature cluster, screening out feature clusters with potential stress abnormality, and extracting stress abnormality feature vectors of the feature clusters;

S4, constructing a stress abnormality mode feature library based on historical glass cold end stress abnormality case data, and carrying out stress abnormality mode identification by using a mode matching algorithm according to the stress abnormality feature vector and the stress abnormality mode feature library to obtain a candidate stress abnormality mode list;

S5, collecting factor data related to stress at the cold end of the glass production line, and constructing a stress related factor data set, establishing a stress abnormality probability prediction model by using a machine learning algorithm according to historical glass cold end stress abnormality case data and the stress related factor data set to obtain an initial stress abnormality probability vector;

s6, probability sorting is conducted on the candidate stress abnormal mode list according to the stress abnormal probability distribution data to obtain a stress abnormal mode sorting list, equipment layout data and glass transmission path data of the glass production line are obtained, stress abnormal area positioning analysis is conducted on the basis of the stress abnormal mode sorting list, the equipment layout data and the glass transmission path data to obtain stress abnormal area positioning result data, stress abnormal reason analysis is conducted according to the stress abnormal area positioning result data and the stress abnormal mode feature library, and a cold end stress detection report of the glass production line is generated.

The invention provides a cold end stress detection system of a glass production line, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the memory, wherein the processor executes the program to realize the cold end stress detection method of the glass production line.

The method has the beneficial effects that the multispectral stress sensing array is used for collecting the spectrum data, so that the influence of the traditional contact detection on glass products and production lines is avoided, and the detection precision and the real-time performance are improved.

By combining spectral feature extraction, space-time correlation analysis and feature clustering algorithm, abnormal modes in the stress distribution of the cold end of the glass can be automatically identified, and the intelligent judgment capability of the detection system is improved.

By analyzing and modeling the historical abnormal cases, an abnormal pattern library containing spectral characteristics, stress distribution characteristics and defect information is formed, a reliable basis is provided for subsequent pattern matching, and the learning and adaptation capacity of the system is enhanced.

By means of a pattern matching algorithm and similarity calculation, whether the current stress state is matched with a known abnormal pattern or not can be effectively identified, and the similarity degree of the current stress state is quantized, so that the accuracy and the robustness of abnormality identification are improved.

By introducing influencing factors such as process parameters and environmental parameters, a stress anomaly probability prediction model is established by combining a machine learning algorithm, and dynamic adjustment can be performed according to real-time data, so that the scientificity and practicability of prediction are improved.

By combining equipment layout and glass transmission path data, a specific region where stress abnormality occurs can be accurately positioned, and cause analysis is performed based on a feature library, so that quick response and fault investigation are facilitated.

The finally output stress detection report not only comprises abnormal information, but also can provide cause analysis and region positioning, and provides powerful data support for production process adjustment, equipment maintenance and quality control.

Drawings

FIG. 1 is a diagram of the steps of the method of the present invention;

FIG. 2 is a diagram illustrating step S2 of FIG. 1 according to the present invention.

Detailed Description

The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.

In one embodiment of the present invention, as shown in fig. 1, a method for detecting cold end stress of a glass production line, the method comprises:

The technical scheme has the working principle that the multispectral stress sensing array is utilized to collect spectrum data of a cold end area of a glass production line, the data are closely related to stress states of glass, and the collected original spectrum data are preprocessed to eliminate noise, enhance signals and the like, so that a preprocessed spectrum data set with higher quality is obtained, and a reliable data base is provided for subsequent analysis.

And extracting the space-time characteristics of the cold-end stress of the glass from the spectrum characteristic vector set by utilizing a space-time correlation analysis algorithm to form a stress space-time characteristic matrix. This step helps to understand the distribution and variation law of stresses in time and space.

Collecting historical glass cold end stress abnormality case data, including spectrum characteristics, stress distribution conditions and corresponding glass defect information when stress is abnormal, providing reference for abnormal characteristic matching, and performing characteristic cluster analysis on a stress space-time characteristic matrix to obtain a plurality of stress characteristic clusters. And carrying out abnormal feature matching on each cluster by combining with the historical case data, screening out feature clusters with potential stress abnormality, and extracting stress abnormality feature vectors of the clusters.

The method comprises the steps of establishing a stress abnormal mode feature library based on historical case data, wherein the library comprises spectrum features, stress distribution features and glass defect features of different stress abnormal modes, identifying a candidate stress abnormal mode list according to the stress abnormal feature vector and the stress abnormal mode feature library by using a mode matching algorithm, and carrying out similarity calculation on each mode in the candidate mode list to generate a stress abnormal similarity vector for evaluating the matching degree of each mode and actual stress abnormality.

The method comprises the steps of collecting process parameter data, environment parameter data and other stress-related factor data at the cold end of a glass production line, constructing a stress-related factor data set, utilizing a machine learning algorithm to establish a stress abnormality probability prediction model according to historical case data and the stress-related factor data set to obtain an initial stress abnormality probability vector, and dynamically correcting the initial probability vector according to the stress-related factor data set to obtain more accurate stress abnormality probability distribution data.

The method comprises the steps of carrying out probability sorting on a candidate stress abnormal mode list according to stress abnormal probability distribution data to obtain a stress abnormal mode sorting list, conveniently processing high-probability abnormal modes preferentially, obtaining equipment layout data and glass transmission path data of a glass production line, carrying out stress abnormal region positioning analysis by combining the stress abnormal mode sorting list to determine the specific position where stress abnormality occurs, carrying out stress abnormality cause analysis according to stress abnormal region positioning result data and a stress abnormal mode feature library to generate a cold end stress detection report of the glass production line, and providing decision support for maintenance and optimization of the production line.

The technical scheme has the advantages that stress data of a cold end area of a glass production line can be captured and analyzed more accurately through the multispectral stress sensing array and the space-time correlation analysis algorithm, and the condition of missing detection or misjudgment caused by the defects of a traditional detection method is reduced.

Through machine learning and a pattern matching algorithm, the stress abnormal pattern is automatically identified, the dependence on manual inspection and judgment is obviously reduced, and the detection efficiency is improved.

By combining the historical data and the stress association factor data, the stress abnormality possibly occurring at the cold end of the glass can be predicted in advance by establishing a stress abnormality probability prediction model, and preventive measures can be taken in advance, so that glass defects and shutdown risks in the production process are reduced.

By accurately positioning the stress abnormal region and analyzing the abnormal reason, the method is beneficial to improving the production process and optimizing the equipment layout and improving the stability of the glass production line, thereby ensuring the produced glass to have more reliable quality.

By timely identifying and positioning the stress abnormality, measures can be taken in early stage to avoid production of unqualified glass, and defect rate and production loss are reduced.

According to the method, influence factors such as environmental parameter data are comprehensively considered, the stress anomaly probability prediction model can be dynamically adjusted, and production changes under different environmental conditions can be well adapted.

By carrying out cluster analysis and pattern recognition on the stress characteristics, data support can be provided for subsequent process optimization, and a more scientific production adjustment strategy is formulated based on historical cases and actual production conditions.

In one embodiment of the present invention, the S1 includes:

s11, uniformly arranging a multispectral stress sensing array in a key area of a cold end of a glass production line, calibrating the arranged multispectral stress sensing array, measuring spectral responses of the multispectral stress sensing array under different stress conditions, and establishing a calibration curve between the spectral responses and the stresses;

S12, in the glass running process, the multispectral stress sensing array acquires stress related spectrum data of the glass surface in real time, and the acquired spectrum data is transmitted to the data processing system in real time for preliminary storage and management;

s13, denoising the collected original spectrum data, normalizing the denoised original spectrum data, smoothing the normalized spectrum data, and obtaining a preprocessed spectrum data set.

The technical scheme has the working principle that multispectral stress sensing arrays are uniformly arranged in key areas of the cold end of the glass production line, such as a cooling roller way, a conveying belt and the like. These areas are where the glass stress changes are significant, and placement of the sensor array ensures accurate capture of the stress changes on the glass surface. And calibrating the arranged multispectral stress sensing array to ensure the accuracy of the measurement result. The calibration process includes measuring the spectral response of the sensing array under different stress conditions and establishing a calibration curve between the spectral response and the stress. This curve will serve as a basis for the subsequent spectral data and stress conversion.

The glass line is started and glass is passed through the cold end zone at normal production speed. In the running process of the glass, the multispectral stress sensing array can collect stress related spectrum data of the surface of the glass in real time, and the collected spectrum data can be transmitted to the data processing system in real time for preliminary storage and management. This step ensures timeliness and integrity of the data, providing a basis for subsequent data processing and analysis.

And denoising the acquired original spectrum data to eliminate noise and interference in the data. The denoising processing can improve the signal-to-noise ratio of the data, so that the subsequent analysis is more accurate, and the original spectrum data after the denoising processing is normalized. The normalization processing can unify the data with different orders to the same scale, so that the subsequent comparison and analysis are convenient.

And smoothing the normalized spectrum data to further eliminate tiny fluctuation and noise in the data. The smoothing process can make the spectrum data smoother, and improve the stability and reliability of the data, and the preprocessing spectrum data set is obtained after the processing steps. The data set contains spectral data subjected to denoising, normalization and smoothing treatment, and provides a high-quality data base for subsequent feature extraction and stress analysis.

The technical scheme has the advantages that the multispectral stress sensing arrays are uniformly arranged in the key areas of the cold end, and the multispectral stress sensing arrays are calibrated, so that the accurate and stable spectral response under different stress conditions is ensured, and high-quality stress data are provided.

Through denoising, normalization and smoothing of spectrum data, interference of environmental noise, equipment errors and other factors is effectively reduced, so that a final preprocessed spectrum data set is more accurate, and subsequent analysis is facilitated.

In the glass production process, spectral data are collected and transmitted to a data processing system in real time, so that continuous monitoring of the stress condition of a cold end area of the glass is ensured, and possible anomalies are found in time.

According to the method, the stress state of the cold end of the glass is monitored in real time, the data acquisition is ensured not to interfere with the production flow, accurate detection is facilitated under the condition that the production efficiency is not affected, and the production process is optimized.

The automatic acquisition and data processing process of the multispectral stress sensor array reduces the complexity of manual operation, improves the automation level of the production line, and reduces the risk of manual operation.

The collected spectrum data is transmitted in real time and stored in the data processing system, a systematic data management and storage mode is established, subsequent data analysis and tracing are facilitated, and reliable historical data support is provided.

Through the standard data preprocessing step, the reliability of the data in the subsequent spectral feature extraction and space-time feature analysis is ensured, and the accuracy of the final stress detection result is improved.

In one embodiment of the present invention, as shown in fig. 2, the step S2 includes:

S21, extracting spectrum characteristics related to stress from the preprocessed spectrum data set, adopting a characteristic selection algorithm to perform dimension reduction treatment on the extracted spectrum characteristics, and constructing a spectrum characteristic vector set;

s22, processing the optical characteristic vector set through a space-time correlation analysis algorithm, and extracting space-time characteristics of stress of the cold end of the glass;

S23, organizing and storing the extracted space-time features according to a preset format, and constructing a stress space-time feature matrix, wherein rows of the matrix represent different time points or space positions, columns represent different space-time features, and elements in the matrix represent space-time feature values corresponding to the time and space positions.

The working principle of the technical scheme is that the spectrum characteristics related to the stress are extracted from the preprocessed spectrum data set. These features include, but are not limited to, spectral peak, peak-to-valley position, spectral bandwidth, spectral integrated intensity, and the like. These features can reflect different aspects of the glass surface stress and are important basis for subsequent stress analysis. And adopting a feature selection algorithm to perform dimension reduction treatment on the extracted spectral features. Since the original spectral features may contain a large amount of redundant information, the redundant features can be removed through dimension reduction processing, and features which are most critical to stress analysis are reserved, so that the efficiency and accuracy of subsequent analysis are improved. And organizing the spectrum characteristics after the dimension reduction treatment according to a certain format to construct a spectrum characteristic vector set. This set of vectors will serve as the basis for subsequent spatio-temporal correlation analysis.

The set of spectral feature vectors is processed by a spatio-temporal correlation analysis algorithm, such as a spatio-temporal clustering algorithm, a spatio-temporal Markov model, or the like. The algorithms can mine the correlation of spectral features in time and space, thereby extracting the space-time features of the stress of the cold end of the glass. The extracted space-time characteristics can reflect the distribution rule and the change trend of the stress in time and space. These features are important for understanding the dynamic behavior of the cold end stress of the glass, predicting stress anomalies, and optimizing the production process.

And organizing and storing the extracted space-time characteristics according to a preset format. This step ensures the ordering and accessibility of the spatio-temporal features, facilitating subsequent analysis and processing. And constructing a stress space-time feature matrix based on the organized and stored space-time features. In this matrix, rows represent different time points or spatial positions, columns represent different spatio-temporal features, and elements in the matrix represent spatio-temporal feature values corresponding to the time and spatial positions. This matrix provides important data support for subsequent stress anomaly pattern recognition, probability prediction, and region localization.

The technical scheme has the advantages that by extracting the spectrum characteristics related to the stress, the stress information of the cold end area of the glass can be captured more accurately, and more representative data can be provided for subsequent analysis.

The feature selection algorithm is adopted to perform dimension reduction processing on the extracted spectral features, redundant data is reduced, the data processing is more efficient, and meanwhile, the problems of calculation burden and overfitting caused by high-dimension data are avoided.

By means of a space-time correlation analysis algorithm, space-time characteristics of the cold-end stress of the glass can be extracted more effectively, and distribution rules and variation trends of the stress in time and space are analyzed comprehensively.

By extracting and analyzing the space-time characteristics, the dynamic change and distribution characteristics of the cold end stress of the glass can be deeply known, and the identification of potential production problems and anomalies is facilitated.

By adopting an automatic space-time correlation analysis algorithm, the dependence on manual intervention is reduced, so that the whole data processing flow is more intelligent and efficient, and the influence of human factors on results is reduced.

The extracted space-time features are organized and stored according to a preset format, a stress space-time feature matrix is constructed, subsequent data query, analysis and decision are facilitated, and the systematicness and operability of data management are improved.

By constructing the space-time feature matrix, accurate data support can be provided for subsequent stress anomaly prediction and process optimization, and more scientific and targeted production adjustment strategies can be formulated.

In one embodiment of the present invention, the step S23 includes:

Dividing the space-time characteristics into time dimension characteristics and space dimension characteristics according to the physical meaning and analysis requirements characterized by the space-time characteristics, and defining identifiers for each type of characteristics;

For the space-time characteristic values of different dimensions and value ranges, adopting a unified standardized method to process, and mapping the characteristic values to specific numerical intervals;

The method comprises the steps of constructing a time-space index system, distributing a unique index number for each time point according to a time sequence in a time dimension, distributing a corresponding index identifier for each space position according to the actual space layout of a glass cold end in a space dimension;

Dynamically planning the dimension of the stress space-time feature matrix; determining the number of rows of the matrix according to the time span of data acquisition and the time resolution requirement of analysis in the time dimension;

Filling the normalized space-time characteristic values to the corresponding positions of the matrix one by one according to a time-space index system, estimating and filling the data missing condition by adopting an interpolation method, and forming a complete stress space-time characteristic matrix after filling all the characteristic values;

Performing quality check on the constructed stress space-time feature matrix, and correcting and optimizing based on a check result;

and storing the stress space-time characteristic matrix subjected to quality check and optimization into a specified database or file system, and simultaneously recording key information of the matrix to form a complete metadata record.

The working principle of the technical scheme is that the space-time characteristics are divided into time dimension characteristics and space dimension characteristics according to physical meanings represented by the space-time characteristics (such as the change rate of stress along with time reflects stress dynamic characteristics, and the space distribution gradient reflects the non-uniformity of stress in space) and analysis requirements (such as the periodic fluctuation or local aggregation of stress is concerned). An identifier is defined for each type of feature, so that different features can be accurately identified and referenced in the matrix construction, data processing and analysis processes.

Different empty features may have different dimensions and value ranges, and direct matrix construction and analysis using the original feature values may result in inaccurate or difficult comparison of results. And processing the characteristic values by adopting a unified standardization method (such as mapping the characteristic values to the [0,1] interval by maximum and minimum standardization), eliminating the influence of the difference of dimension and value range, and enabling different characteristics to have comparability in a matrix.

In the time dimension, a unique index number is allocated to each time point according to the time sequence, so that each time point is ensured to have a unique determined position in the matrix, and the stress characteristics can be analyzed and tracked on the time sequence conveniently. According to the actual space layout of the glass cold end, corresponding index marks are distributed for each space position in a grid division, region coding and other modes, so that the space position can be accurately positioned in a matrix, and the distribution and the change of stress in different space regions can be conveniently analyzed. The index system is used as a positioning basis of rows and columns of the matrix, so that the space-time characteristic values can be accurately placed at corresponding positions of the matrix, and the corresponding relation between the matrix structure and actual space-time characteristic data is ensured.

And determining the number of rows of the matrix according to the time span of data acquisition and the time resolution requirement of analysis. The time span determines the time range covered by the matrix, and the time resolution requirement affects the accuracy of the representation of each time point in the matrix. And determining the column number of the matrix according to the space division fineness of the cold end of the glass. The finer the space division, the more the number of columns of the matrix, and the more detailed the distribution of stress in space can be reflected. A certain expansion space is reserved to cope with the data volume possibly increased in the future or the new space-time characteristic analysis requirement, and the adaptability and the flexibility of the matrix are improved. And filling the normalized space-time characteristic values into corresponding positions of the matrix one by one according to a time-space index system, so that the matrix completely represents the characteristics of the stress at different time and space positions. For the possible data missing condition of some time points or space positions, a reasonable interpolation method (such as linear interpolation, spline interpolation and the like) is adopted for estimating and filling, so that the integrity and continuity of matrix data are ensured, and the influence of the data missing on the subsequent analysis result is avoided.

And (3) performing quality check on the constructed stress space-time characteristic matrix, and checking whether abnormal values (such as values exceeding a reasonable range) exist in the matrix. And eliminating or replacing the abnormal value to ensure the accuracy and reliability of the matrix data. Checking whether the data have inconsistent conditions (such as contradiction between the characteristic values of a time point or a space position) and tracing and adjusting the inconsistent conditions so as to ensure the logic property and consistency of the data. And a high-efficiency storage format (such as a sparse matrix storage format when a large number of zero elements exist in a matrix) and a compression algorithm are adopted, so that the occupation of storage space is reduced, the data reading and writing efficiency is improved, and the subsequent storage, management and analysis of matrix data are facilitated.

And storing the stress space-time characteristic matrix subjected to quality check and optimization into a specified database or file system, so as to ensure the safety and accessibility of data. And recording key information such as construction time, data sources, processing parameters and the like of the matrix to form a complete metadata record. The metadata record is beneficial to understanding, tracing and reusing the matrix data later, and a user can conveniently know the construction process of the matrix and related background information.

The technical scheme has the advantages that the space-time characteristics are divided into the time dimension characteristics and the space dimension characteristics, and the identifiers are defined for each type of characteristics, so that the performances of stress changes in different dimensions can be analyzed and understood more clearly, and the interpretability of data is enhanced.

The space-time characteristic values of different dimensions and value ranges are processed by adopting a unified standardized method, so that the unified dimension of the characteristic values is ensured, deviation caused by dimensional differences among different characteristics is effectively avoided, and the consistency of data analysis is improved.

By constructing a complete time-space index system, accurate positioning of space-time characteristic values is ensured, dislocation or confusion of data in a matrix is avoided, and the systematicness and accuracy of data management are improved.

When the space-time feature matrix dimension is dynamically planned, the fineness of time span and space division is considered, an expansion space is reserved, and the system is ensured to be capable of coping with the increase of future data quantity or new analysis requirements, so that the flexibility and the future adaptability of the system are improved.

And for the time points or the space positions with data missing, a reasonable interpolation method is adopted for estimating and filling, so that the influence of missing data on the quality of the time-critical feature matrix is effectively reduced, and the integrity and the continuity of the data are ensured.

The constructed stress space-time feature matrix is subjected to quality check, so that the problem of abnormal value or inconsistent data can be found and corrected in time, the accuracy and the reliability of the data are improved, and the validity of an analysis result is ensured.

By optimizing the matrix storage structure and adopting a sparse matrix storage format and a compression algorithm, the memory space occupation is reduced, the data reading and writing efficiency is improved, and the performance and the response speed of large-scale data processing are further improved.

Storing the optimized stress space-time characteristic matrix into a specified database or file system, and recording key information such as matrix construction time, data sources and processing parameters to form a complete metadata record, thereby ensuring traceability and manageability of data.

In one embodiment of the present invention, the S3 includes:

S31, collecting stress abnormal case data which appear at the cold end of a glass production line in a period of time in the past, and establishing a historical glass cold end stress abnormal case database;

S32, performing feature cluster analysis on the stress space-time feature matrix by adopting a clustering algorithm;

S33, combining cold end stress abnormal case data of the historical glass, carrying out abnormal feature matching on each stress feature cluster, comparing the features of the cluster with the abnormal features in the historical case, and calculating similarity, wherein the similarity is obtained through the following formula:

Wherein, the cluster eigenvector of A is [ [ ,,...,The historical abnormal characteristic vector of B is [ [,,...,N represents a feature dimension (e.g., a dimension of time, position, spectral feature, etc.);

s34, setting a similarity threshold value, screening out a characteristic cluster with potential stress abnormality, and judging whether a stress abnormality condition exists for the cluster with similarity exceeding the threshold value;

s35, extracting stress abnormal feature vectors from the screened feature clusters, and selecting the most representative features from the clusters to construct the stress abnormal feature vectors.

The working principle of the technical scheme is that stress abnormality case data which appear at the cold end of a glass production line in a past period of time are collected and are important bases for subsequent analysis. The case data comprise basic information such as time, position, glass specification, batch and the like of occurrence of stress abnormality, and key data such as spectrum characteristics, stress distribution condition, corresponding glass defect information and the like when the stress abnormality occurs, so that various characteristics and influences of the stress abnormality are comprehensively reflected. And establishing a historical glass cold end stress anomaly case database, and carrying out centralized storage and management on the collected case data, so that subsequent inquiry, calling and analysis are convenient. The collected historical case data is sorted and classified according to the type, severity and the like of stress abnormality. The classification mode is beneficial to quickly positioning stress abnormality cases of specific types or severity in subsequent analysis, and improves analysis efficiency and pertinence.

And performing feature cluster analysis on the stress space-time feature matrix by adopting a clustering algorithm. The clustering algorithm can aggregate stress spatio-temporal data with similar characteristics to form different clusters, so that the internal mode and structure of the stress spatio-temporal characteristics are revealed. And determining the clustering number and the similarity measurement method according to the characteristics and the distribution rule of the stress space-time characteristics. The selection of the number of clusters needs to comprehensively consider the complexity and analysis requirement of the data, and a similarity measurement method determines how to measure the similarity between different data points. The quality of the clusters is assessed using cluster effectiveness indicators, such as profile coefficients, davies-Bouldin index, etc. The indexes can reflect the advantages and disadvantages of the clustering result from different angles and help to select the optimal clustering scheme. And analyzing the characteristics and distribution conditions of each cluster, and knowing the characteristics of different stress space-time characteristic modes. Through deep analysis of the cluster, potential rules and abnormal modes of stress space-time characteristics can be found, and basis is provided for subsequent abnormal characteristic matching.

And carrying out abnormal feature matching on each stress feature cluster by combining the cold end stress abnormal case data of the historical glass. Comparing the characteristics of the cluster with the abnormal characteristics in the historical cases, and calculating the similarity. This step is intended to determine if the characteristics of the current cluster are similar to the abnormal characteristics in the historical case, and thus determine if a potential stress anomaly exists. The similarity calculation is a key link of abnormal feature matching, and the similarity between the cluster features and the historical abnormal features can be accurately measured by a reasonable similarity measurement method, so that a basis is provided for subsequent screening and judgment. And setting a similarity threshold value for screening out the feature cluster with potential stress abnormality. The setting of the similarity threshold value needs to be adjusted according to actual conditions and analysis requirements, so that the real potential abnormal cluster can be screened out, and erroneous judgment is avoided. And judging whether a stress abnormality exists for the cluster with the similarity exceeding the threshold value. Based on the result of similarity matching, the step carries out preliminary abnormality judgment on the cluster, and provides a direction for subsequent extraction and analysis of abnormal feature vectors.

And extracting stress abnormal feature vectors from the screened feature clusters, and selecting the most representative features from the clusters. These representative features can summarize the main features and anomaly patterns of the cluster, providing key information for subsequent stress anomaly pattern recognition and analysis. The selected representative features are constructed into stress anomaly feature vectors that will serve as important inputs for subsequent stress anomaly pattern recognition and analysis. By constructing the feature vector, the complex stress space-time feature information can be simplified into a form convenient to process and analyze, and the analysis efficiency and accuracy are improved.

The technical scheme has the advantages that by collecting and arranging the historical glass cold end stress abnormality case data, a perfect database is established, high-quality reference data is provided for subsequent abnormality analysis, and the stress abnormality occurrence mode can be accurately identified.

By sorting and classifying the historical case data and dividing the historical case data according to the type and severity of stress abnormality, the data processing flow is simplified, and the subsequent analysis work is more systematic and efficient.

The clustering algorithm is adopted to perform the clustering analysis of the stress space-time feature matrix, so that the distribution rule of different stress features can be effectively identified, the reliability and the accuracy of a clustering result are ensured by using the clustering effectiveness index, and the analysis depth and accuracy are improved.

By carrying out abnormal feature matching on each stress feature cluster and comparing the stress feature clusters with the historical cases, potential stress abnormal modes can be found rapidly, and the sensitivity and reliability of abnormal detection are improved.

The similarity threshold is set to screen out the feature cluster of potential stress abnormality, and through reasonable threshold setting, the risks of misjudgment and missed judgment can be effectively reduced, and the accuracy and practicality of abnormality detection are improved.

By extracting stress abnormal feature vectors from the screened feature clusters and selecting the most representative features, the abnormal features can be accurately extracted, and accurate basis is provided for subsequent abnormal diagnosis and early warning.

The scheme can cope with new types of anomalies possibly added in the future or new analysis demands, and along with the increase of data volume and the optimization of algorithms, the system has stronger expansibility and flexibility, and can adapt to production environments with different scales and complexity.

Through automatic clustering and abnormal feature matching processes, the requirement of manual intervention is reduced, the working efficiency is improved, the occurrence probability of human errors is reduced, and therefore the stability and the production efficiency of the production line are ensured.

Through effectively identifying potential stress abnormality and predicting abnormal characteristics, preventive measures can be taken in advance, and major faults are avoided, so that the reliability and the production quality of the whole production line are improved.

In one embodiment of the present invention, the S4 includes:

s41, constructing a stress abnormal mode feature library based on historical glass cold end stress abnormal case data, carrying out standardized processing on data in the feature library, and unifying the dimensions and the ranges of different features;

s42, selecting a pattern matching algorithm, and optimally setting parameters of the algorithm according to the characteristics and distribution rules of stress abnormal characteristics;

S43, according to the stress abnormal feature vector and the stress abnormal pattern feature library, performing stress abnormal pattern recognition by using a selected pattern matching algorithm;

S44, obtaining a candidate stress abnormal mode list according to the similarity calculation result, and carrying out similarity calculation on each mode in the candidate stress abnormal mode list to generate a stress abnormal similarity vector.

The working principle of the technical scheme is that a stress abnormality mode feature library is constructed based on historical glass cold end stress abnormality case data. The characteristic library comprehensively covers the spectrum characteristics, stress distribution characteristics and glass defect characteristics corresponding to different stress abnormal modes. These features describe the manifestation of stress anomalies from different angles, providing a rich information basis for subsequent pattern recognition. Because different features in the feature library may have different dimensions and ranges, direct use of the raw data for matching and analysis may lead to inaccurate results. Therefore, the data in the feature library is standardized, the dimensions and the ranges of different features are unified, so that the different features have comparability in the subsequent pattern recognition, and the accuracy and the reliability of the recognition are improved.

And selecting a proper pattern matching algorithm, such as cosine similarity matching, euclidean distance matching and the like, according to the requirement of stress abnormal pattern recognition. The algorithm can measure the similarity degree between two vectors from different angles, and is suitable for matching the stress abnormality characteristic vectors. And optimally setting parameters of the selected algorithm according to the characteristics and distribution rules of the stress abnormal characteristics. Different parameter settings can influence the performance and the recognition effect of the algorithm, the algorithm can be better adapted to the characteristics of the stress abnormal characteristics by optimizing parameters, and the recognition accuracy and efficiency are improved. And training the algorithm by using the stress abnormality feature vectors and the corresponding stress abnormality modes in the historical case data. In the training process, parameters of the algorithm are continuously adjusted, so that the algorithm can more accurately identify stress abnormal modes in historical cases, and generalization capability and identification accuracy of the algorithm are improved.

And according to the stress abnormal feature vector and the stress abnormal pattern feature library, performing stress abnormal pattern recognition by using a selected pattern matching algorithm. And matching the stress abnormal feature vector with each mode in the feature library, and measuring the proximity degree between the feature vector and each mode by calculating the similarity. The similarity calculation is a core link of pattern recognition, and the similarity degree between the feature vector and the pattern can be accurately reflected by a reasonable similarity measurement method, so that a basis is provided for subsequent pattern screening and decision making.

And obtaining a candidate stress abnormal mode list according to the similarity calculation result. The modes in the list are ranked from high to low according to the similarity, and the mode with higher similarity is more likely to be the stress abnormality mode corresponding to the current stress state. The sequencing mode is beneficial to quickly positioning the most possible stress abnormal mode, and improves the efficiency of fault diagnosis. And carrying out similarity calculation on each mode in the candidate stress abnormal mode list to generate a stress abnormal similarity vector. The elements in the vector represent similarity values of the current stress state and each candidate stress abnormal mode, and the similarity values can intuitively reflect the proximity degree between the current stress state and each candidate mode, so that an important quantification basis is provided for subsequent fault diagnosis and decision.

The technical scheme has the advantages that by constructing the stress abnormal mode feature library, the system can more comprehensively store and organize the spectrum features, the stress distribution features and the glass defect features of different stress abnormal modes, and provide detailed and standardized data support for mode identification, so that the accuracy of mode identification is greatly improved.

By carrying out standardized processing on the feature library data, the dimension and the range of different features are unified, the interference among the different features is effectively reduced, the pattern matching is more concise and efficient, and the processing difficulty caused by data inconsistency is avoided.

By selecting a proper pattern matching algorithm (such as cosine similarity and Euclidean distance) and optimizing according to the characteristics and distribution rules of stress abnormal characteristics, different types of stress abnormal patterns can be better adapted, and the flexibility and the robustness of pattern recognition are improved.

By optimizing algorithm parameters and training by using historical case data, the system can complete stress abnormal pattern recognition in a shorter time, so that the processing efficiency is remarkably improved, and the response delay to real-time data is reduced.

By generating a candidate stress abnormal mode list according to the similarity calculation result and sorting according to the similarity, the most similar modes can be preferentially matched, so that the risk of false identification is reduced, and the identified abnormal modes are ensured to be more consistent with the current stress state.

By generating the stress abnormal similarity vector, an accurate basis is provided for subsequent fault diagnosis and decision making, accurate judgment and timely response to an abnormal mode in the decision making process are ensured, and the reliability and effectiveness of the decision making process are improved.

By accumulating abundant historical data in the feature library and continuously optimizing the pattern matching algorithm, the system can continuously evolve, adapt to new production scenes and complex changes, and improve the long-term maintainability and adaptability of the system.

Through the automatic stress abnormal mode identification process, the dependence on manual judgment is reduced, the working pressure of operators is lightened, the occurrence rate of human errors is reduced, and the stability of the production line is further improved.

Through effective pattern matching and similarity calculation, the system can predict and warn potential stress abnormality conditions in advance, thereby helping to take preventive measures in time, reducing production line downtime and improving the running efficiency and the product quality of the production line.

In one embodiment of the present invention, the step S43 includes:

The method comprises the steps of carrying out feature vector pre-alignment operation on a stress abnormal feature vector and each mode feature vector in a feature library, rearranging the current stress abnormal feature vector according to the definition and sequence of features in the feature library to ensure that the feature sequence of the stress abnormal feature vector and the feature sequence of the feature library are consistent;

the stress abnormal feature vector after prealignment and the mode feature vector in the feature library are split in a multi-dimension mode according to the feature type;

Respectively carrying out similarity calculation on the split feature vectors of each dimension by adopting a selected pattern matching algorithm;

Fusing the similarity values obtained by dimension calculation according to pre-assigned weights to obtain comprehensive similarity values of stress abnormal feature vectors and modes in a feature library, comprehensively evaluating the fused similarity values, and primarily screening a plurality of candidate modes 20% before similarity;

And comparing the preliminarily screened candidate modes with a dynamic threshold value, and further screening candidate modes with similarity higher than the threshold value to be used as a final candidate stress abnormal mode set.

The working principle of the technical scheme is that the stress abnormal feature vector is pre-aligned with each mode feature vector in the feature library, and the current stress abnormal feature vector is rearranged according to the definition and the sequence of the features in the feature library, so that the feature sequence of the stress abnormal feature vector and the feature sequence of the feature library are consistent. This operation is to ensure that elements at the same position in different feature vectors represent the same feature meaning in subsequent similarity calculation, and avoid calculation errors caused by inconsistent feature sequences. For features that exist in the feature library but the current feature vector is missing, a specific filling strategy (such as filling the average or median of the feature in the historical data) is adopted to supplement. The existence of the missing value can influence the accuracy of similarity calculation, and the integrity of the feature vector can be restored as much as possible through a reasonable filling strategy, so that the influence of the missing value on a result is reduced.

And carrying out multidimensional splitting on the prealigned stress abnormal feature vector and the mode feature vector in the feature library according to the feature type (spectrum feature, stress distribution feature and glass defect feature). Different feature types have different roles and meanings in the stress abnormal pattern recognition, and features with different dimensions can be respectively analyzed and processed through splitting, so that the pertinence and the accuracy of the recognition are improved. For the features of different dimensions, corresponding weights are allocated according to the importance of the features in the stress anomaly pattern recognition. The weight assignment may be determined based on expert experience, historical data analysis, or machine learning algorithms (e.g., entropy weighting). For example, spectral features may be more critical for the identification of certain stress anomaly patterns, and may be given higher weight, while glass defect features may have less impact in certain situations, and may be given relatively lower weight. The reasonable weight distribution can highlight the role of important features in similarity calculation, and the reliability of the identification result is improved. And (3) adopting a selected pattern matching algorithm (such as cosine similarity matching, euclidean distance matching and the like) to respectively calculate the similarity of the split feature vectors of each dimension. For the dimension of the spectrum feature, due to the continuity and complexity of the spectrum data, the spectrum feature can be subjected to further data processing (such as Fourier transform to extract the frequency domain feature) and then the similarity is calculated so as to better capture the key information in the spectrum feature.

For the dimension of the stress distribution characteristics, the distribution characteristics of the stress in space are considered, and a similarity measurement method based on the space distribution, such as calculating the space correlation coefficient of the stress distribution, can be adopted to accurately measure the similarity between the stress distribution characteristics. For the characteristic dimension of the glass defect, a proper similarity calculation mode is adopted according to the type and degree of the defect, such as calculating the similarity of codes after the type of the defect is coded, or different similarity thresholds are set according to the degree of the defect. And obtaining similarity values under each dimension through dimension similarity calculation, and providing a basis for subsequent comprehensive evaluation.

And fusing the similarity values obtained by dimension calculation according to pre-assigned weights to obtain the comprehensive similarity values of the stress abnormal feature vector and each mode in the feature library. In the fusion process, different fusion strategies such as a weighted average method, a weighted geometric average method and the like can be adopted, and the most suitable method is selected according to actual conditions. Meanwhile, in consideration of the correlation among the similarity of each dimension, the similarity vector can be subjected to dimension reduction processing by adopting methods such as Principal Component Analysis (PCA) and the like, and the main information is extracted and then fused, so that the influence of redundant information on the result is reduced, and the accuracy and the stability of the fusion result are improved. And comprehensively evaluating the fused similarity value, and primarily screening a plurality of candidate modes with 20% of the similarity. The purpose of the preliminary screening is to narrow the candidate range and improve the efficiency of the subsequent analysis.

And dynamically adjusting the similarity threshold according to the similarity distribution conditions of different stress abnormal modes in the historical data and the characteristics (such as stress magnitude, change trend and the like) of the current stress state. The dynamic adjustment of the threshold value can enable the screening process to be more flexible and accurate, and meets the stress abnormal pattern recognition requirements under different conditions. And comparing the preliminarily screened candidate modes with a dynamic threshold value, and further screening candidate modes with similarity higher than the threshold value to be used as a final candidate stress abnormal mode set. The final candidate stress abnormality mode set can more accurately reflect the stress abnormality mode possibly corresponding to the current stress state, and provides powerful support for subsequent fault diagnosis and decision.

The technical scheme has the advantages that the consistency of the current stress abnormal feature vector and the pattern feature vector in the feature library in the feature sequence is ensured by carrying out the pre-alignment operation of the feature vector, and the matching error caused by inconsistent sequence is eliminated, so that the pattern matching precision is greatly improved.

The special filling strategy (such as filling missing features by using an average value or a median in the historical data) is adopted, so that pattern recognition failure or false recognition caused by feature missing is avoided, and the integrity of the feature data and the recognition stability are ensured.

By splitting the feature vector into multiple dimensions such as spectrum features, stress distribution features and glass defect features, the similarity of each dimension is calculated, and the contribution of each feature dimension to the stress anomaly mode can be captured more carefully, so that the accuracy of overall identification is improved.

By distributing weights for the features of different dimensions and dynamically adjusting the weights by combining expert experience or a machine learning algorithm (such as an entropy weight method), the influence of the features on the pattern recognition can be flexibly adjusted according to different characteristics of stress abnormal patterns, and the recognition precision is further improved.

The similarity is calculated respectively for the features with different dimensions, and then the weighted fusion is carried out, so that the complexity of high-dimensional data processing is effectively reduced, the interference of redundant information is reduced, and the high efficiency of the pattern recognition process is ensured.

By adopting a fusion strategy such as a weighted average method or a weighted geometric average method and the like to fuse the similarity of the sub-dimensionality, the influence of each dimensionality can be comprehensively considered, a more comprehensive and reasonable similarity value can be obtained, and the reliability of pattern recognition is improved.

The similarity vector is subjected to dimension reduction processing by adopting a dimension reduction method such as Principal Component Analysis (PCA), so that the interference of redundant information is reduced, the adaptability of the system to different stress states is further enhanced, and the flexibility of pattern recognition is improved.

Through comprehensive evaluation of the fused similarity values, candidate modes with high similarity can be accurately screened, the probability of false screening is reduced, and the screened candidate modes are ensured to be more in line with the current stress state.

Through dynamic adjustment of the similarity threshold value, the screening standard can be flexibly adjusted according to the characteristics of the current stress state, so that the mode screening is more targeted, and the accuracy and reliability of mode identification are improved.

By primarily screening the first 20% of candidate modes on the basis of similarity distribution, the range of the candidate modes can be rapidly narrowed, the efficiency of subsequent diagnosis and decision making is improved, and unnecessary calculation and analysis are reduced.

In one embodiment of the present invention, the step S5 includes:

S51, collecting stress-related factor data at the cold end of the glass production line, and establishing a stress-related factor data acquisition system to acquire and store the factor data in real time;

S52, establishing a stress anomaly probability prediction model by utilizing a machine learning algorithm according to the cold end stress anomaly case data of the historical glass and the stress association factor data set;

S53, preprocessing the data and performing characteristic engineering, integrating stress related factor data and stress abnormality case data, and extracting characteristics which have important influence on stress abnormality prediction;

S54, dividing the processed data into a training set and a testing set, training a machine learning model by using the training set, adjusting parameters of the model, and evaluating and verifying the performance of the model by using the testing set;

S55, dynamically correcting the initial stress anomaly probability vector according to the real-time change of the stress association factor data set to obtain stress anomaly probability distribution data.

The working principle of the technical scheme is that the stress-related factor data such as process parameter data and environment parameter data at the cold end of the glass production line are collected. The stress state of the glass is directly influenced by technological parameters such as cooling speed, temperature gradient, glass thickness and the like, and the stress of the glass is also influenced by environmental parameters such as temperature, humidity, air pressure and the like. The stress-related factor data acquisition system is built, and the real-time acquisition and storage of the data are realized. By means of real-time acquisition, the latest data of the cold end of the glass production line can be acquired timely to reflect the current stress association factor state, and the stored data provides historical data support for subsequent data analysis and model training.

And establishing a stress anomaly probability prediction model by using a machine learning algorithm according to the historical glass cold end stress anomaly case data and the stress correlation factor data set. The historical case data comprises occurrence conditions of stress abnormality and corresponding associated factor data, and potential relations between the occurrence conditions and the associated factor data can be mined through a machine learning algorithm, so that a model capable of predicting stress abnormality probability is constructed. The machine learning algorithm has powerful data processing and pattern recognition capabilities and can learn complex rules and patterns from a large amount of data. Different machine learning algorithms are suitable for different types of data and problems, and the prediction performance of the model can be improved by selecting a proper algorithm.

And integrating the stress related factor data and the stress abnormality case data so as to perform unified analysis and processing. The integrated data contains stress abnormality occurrence or non-occurrence and corresponding associated factor information, and a complete data set is provided for subsequent feature extraction and model training. And (3) cleaning the data, removing noise and abnormal values in the data, and ensuring the quality of the data. Meanwhile, the missing values are processed, such as mean filling, median filling or model-based prediction filling, so that influence of the missing values on model training is avoided. Feature encoding is performed to convert the classification variables into numerical variables so that the machine learning algorithm can process the numerical variables. The method has the advantages that the characteristics which have important influence on stress abnormality prediction are extracted, the most representative characteristics are screened out through a characteristic selection method, the dimensionality and redundant information of data are reduced, and the training efficiency and the prediction performance of the model are improved. The processed data is divided into a training set and a testing set. The training set is used for training the model to enable the model to learn rules and modes in the data, and the testing set is used for evaluating and verifying the performance of the model and checking the performance of the model on the unseen data.

And training the machine learning model by using the training set, and adjusting parameters of the model so that the model can better fit training data. By continuously adjusting parameters, the performance of the model is optimized, and the prediction accuracy of the model is improved. And evaluating and verifying the performance of the model by using the test set, and selecting the model with the optimal performance as a stress anomaly probability prediction model. The evaluation indexes can comprise accuracy, recall rate, F1 value and the like, and the appropriate evaluation indexes are selected according to actual requirements. And obtaining an initial stress anomaly probability vector according to the current stress correlation factor dataset by using the trained stress anomaly probability prediction model. The elements in the vector represent initial probabilities of occurrence of different regions or different stress anomaly modes, which reflect the magnitude of the probability of occurrence of different stress anomaly modes under the current stress-related factors.

And dynamically correcting the initial stress anomaly probability vector according to the real-time change of the stress association factor data set. The stress-related factor data is constantly changing, and these changes may affect the occurrence probability of stress anomalies. The stress abnormality probability vector is timely adjusted by monitoring the change of stress related factor data in real time, so that the current stress abnormality probability distribution situation can be more accurately reflected. And obtaining stress anomaly probability distribution data after dynamic correction. The data can provide real-time stress abnormality early warning information for operators of the glass production line, help the operators to find potential stress abnormality problems in time, and adopt corresponding measures to adjust and optimize the stress abnormality problems so as to ensure the quality and stability of glass production.

The technical scheme has the advantages that the process parameter and environment parameter data of the cold end of the glass production line are collected, and the stress related factor data acquisition system is established, so that the real-time acquisition and storage of the data are realized, the comprehensiveness and timeliness of the data are ensured, and more accurate input data are provided for the subsequent stress anomaly probability prediction.

By using the historical glass cold end stress abnormality case data and the stress association factor data set and combining a machine learning algorithm, a stress abnormality probability prediction model is established, errors possibly brought by a traditional experience judgment method are effectively reduced, and the reliability and accuracy of prediction are improved.

By means of data preprocessing and feature engineering, stress associated factor data and historical abnormal case data are integrated, operations such as cleaning, missing value processing and feature encoding are performed, features which have important influences on stress abnormal prediction are effectively extracted, processing capacity on complex data is improved, and input quality of a model is improved.

The model is trained by the training set and the testing set by dividing the processed data into the training set and the testing set, and the model performance is verified by the testing set, so that the model parameters can be dynamically adjusted, and the model with optimal performance is selected, thereby improving the adaptability of the model and ensuring the prediction accuracy of the model under different stress states.

The performance of the machine learning model is evaluated by using methods such as cross verification and the like, and the consistency of the model in the training data and the test data is ensured, so that the risk of overfitting of the model is effectively reduced, the model has stronger generalization capability, and the model can adapt to more unknown data scenes.

According to the change of the real-time stress association factor data set, the initial stress anomaly probability vector is dynamically corrected, so that more real-time and accurate stress anomaly probability distribution data are obtained, and the capability of the system for quickly responding to the change and adjusting the prediction result in the actual production process is improved.

The stress anomaly probability vector is dynamically adjusted, real-time data change is combined, and timely adjustment can be made according to the actual condition of a production line, so that the flexibility of a prediction model is improved, and the early warning of the stress anomaly mode can accurately reflect the current production state.

The stress anomaly probability prediction result is automatically generated through the machine learning model, so that the frequency and the dependence of human intervention are reduced, subjective errors caused by manual judgment are reduced, and the automation degree and the working efficiency of the system are improved.

By combining machine learning and real-time data analysis, more scientific and reliable data support can be provided for production decision making, the decision making capability of enterprises in stress anomaly prediction and production adjustment is enhanced, and the stability and efficiency of the whole production line are improved.

Because of the real-time collection and dynamic correction of the data, the prediction model can timely respond to the occurrence of stress abnormality, provides quicker early warning and response time for a production line, and reduces the influence of potential stress abnormality on glass quality and production efficiency in the production process.

In one embodiment of the present invention, the step S6 includes:

s61, probability ranking is carried out on the candidate stress abnormal mode list according to the stress abnormal probability distribution data, and the stress abnormal probability is ranked from high to low to obtain a stress abnormal mode ranking list;

s62, acquiring equipment layout data and glass transmission path data of a glass production line, and carrying out stress abnormal region positioning analysis based on the stress abnormal mode sorting list, the equipment layout data and the glass transmission path data;

S63, analyzing the reasons of the stress abnormality according to the positioning result data of the stress abnormality area and the stress abnormality mode feature library, and generating a cold end stress detection report of the glass production line according to the analysis result.

The working principle of the technical scheme is that probability ordering is carried out on the candidate stress abnormality mode list according to stress abnormality probability distribution data. The stress abnormality probability distribution data reflects the probability of occurrence of different stress abnormality modes under the condition of the current stress association factors, the stress abnormality probabilities are ordered from high to low, which stress abnormality modes are more likely to occur can be intuitively displayed, a stress abnormality mode ordering list is obtained, candidate modes are ordered according to the high and low of the stress abnormality probabilities, a priority order is provided for subsequent stress abnormality region positioning and cause analysis, and operators can pay attention to the stress abnormality modes with higher probability.

Equipment layout data and glass transmission path data of a glass production line are acquired. The equipment layout data comprises the positions and parameters of cooling equipment, conveying equipment, detecting equipment and the like, the information can reflect the physical structure and equipment configuration of the production line, and the glass transmission path data comprises the information of the movement track, speed and the like of glass on the production line, so that the flow condition of the glass in the production process can be known. And carrying out stress abnormal region positioning analysis based on the stress abnormal mode ordered list, the equipment layout data and the glass transmission path data. By analyzing the relationship between the stress abnormality pattern and the equipment layout and glass transmission path, the specific area where the stress abnormality may occur is determined. For example, certain stress anomalies may be associated with the location of a particular cooling device or conveying device, or with a particular stage in the glass during conveyance. The method can adopt methods such as space analysis and path tracking to carry out positioning analysis, the space analysis can intuitively display the space relation between the stress abnormal mode and the equipment position, and the path tracking can help to determine the area possibly affected by stress in the transmission process of the glass. And determining a specific area where stress abnormality may occur, and providing a clear target position for subsequent analysis and processing of the cause of the stress abnormality.

And analyzing the cause of the stress abnormality according to the positioning result data of the stress abnormality region and the stress abnormality mode feature library. And analyzing possible reasons for causing stress abnormality by combining the factors such as the running state of equipment, the process parameter setting, the environmental conditions and the like. For example, equipment failure may lead to uneven cooling and thus abnormal stress, unreasonable process parameter settings, such as too fast or too slow cooling speed, may also affect the stress state of the glass, and changes in environmental conditions, such as fluctuations in temperature and humidity, may also affect the stress of the glass. By comprehensively analyzing the factors, the root cause of the stress abnormality can be found out. And generating a cold end stress detection report of the glass production line according to the analysis result. The report should include the stress anomaly pattern ordered list, the stress anomaly region location results, the stress anomaly cause analysis, and corresponding processing advice. The stress abnormality mode sequencing list can enable operators to know occurrence probability of different stress abnormality modes, a stress abnormality area positioning result determines specific positions where stress abnormality is likely to occur, stress abnormality cause analysis reveals root causes of the stress abnormality, and processing suggestions provide specific solutions and measures for the operators. The detection report has the characteristics of clarity, accuracy and comprehensiveness, provides decision basis for maintenance and management of the glass production line, helps operators to take measures in time, eliminates stress abnormality, and ensures quality and stability of glass production.

The technical scheme has the advantages that the candidate stress abnormality modes can be accurately identified by probability sequencing according to the stress abnormality probability distribution data, so that the accuracy and pertinence of stress abnormality prediction are effectively improved.

By sequencing the stress anomaly probability from high to low, a stress anomaly mode sequencing list is formed, the screening process of the anomaly modes is simplified, the complexity of manual judgment is reduced, and the working efficiency is improved.

By combining equipment layout data, glass transmission path data and a stress abnormal mode sequencing list, the stress abnormal area is subjected to positioning analysis by adopting methods such as space analysis and path tracking, so that the accurate identification and positioning of the abnormal area are ensured, and more accurate data support is provided for subsequent processing.

Through the stress abnormality region positioning analysis, a potential stress abnormality occurrence region can be quickly found, fault investigation and production line adjustment can be timely carried out, the response speed of the production line is obviously improved, and the downtime of equipment faults is reduced.

By carrying out cause analysis by combining the factors such as the running state of equipment, the process parameter setting, the environmental conditions and the like, the possible cause of the stress abnormality can be obtained in a short time, the time waste of the traditional manual diagnosis is avoided, and the analysis efficiency is improved.

By combining multiple factors (such as equipment state, environmental parameters, process settings and the like), the comprehensive stress abnormality cause analysis is performed, the comprehensiveness and scientificity of an analysis result are enhanced, and the one-sided analysis of a single factor is avoided.

By generating the cold end stress detection report containing stress abnormal mode sequencing, stress abnormal region positioning results and stress abnormal cause analysis, clear, accurate and comprehensive decision basis is provided for production line maintenance and management, and the reliability of decision is improved.

By generating a detailed stress detection report, potential stress abnormality problems can be identified in advance, references are provided for formulating preventive and treatment measures, and the risk of faults of the production line in the running process is reduced.

The stress abnormal modes are sequenced through a machine learning algorithm, and the automation degree of the whole system is remarkably enhanced by combining with the automatic regional positioning analysis and the reason analysis, so that the dependence of manual operation is reduced, and the intelligent level of a production line is improved.

By comprehensively analyzing the stress abnormality mode, the equipment layout and the transmission path, the precision and the response capability of the fault early warning system are enhanced, the early warning information can be timely obtained by the production line before abnormality occurs, and potential production interruption is avoided.

In one embodiment of the present invention, the step S62 includes:

Constructing a device-stress abnormality mode association matrix based on the analysis result, wherein rows of the matrix represent different stress abnormality modes, columns represent different devices and parameters thereof, and matrix elements represent association strength between the device parameters and the stress abnormality modes;

the method comprises the steps of dividing a motion track of glass on a production line into a plurality of key path segments according to glass transmission path data, evaluating the occurrence probability of stress abnormality modes in the path segments for each path segment, establishing a path segment-stress abnormality probability evaluation model, and calculating the occurrence probability of various stress abnormality modes in each path segment by combining a stress abnormality mode sorting list and a device-stress abnormality mode association matrix;

Constructing a space topology model of the glass production line by using equipment layout data, and defining the space position relation and connection relation among the equipment and the circulation direction of the glass among the equipment;

Simulating a motion process of the glass on a production line by adopting a path tracking algorithm, and simulating a propagation process of stress in the glass by combining a stress abnormal mode feature library;

The method comprises the steps of integrating a correlation analysis result of equipment and stress abnormal modes, a segmentation and stress influence evaluation result of a glass transmission path, a space topological relation analysis result and a path tracking and stress propagation simulation result, adopting a weighted comprehensive evaluation method, and distributing corresponding weights for each analysis result according to the importance and reliability of different analysis results;

by comprehensive calculation, specific areas where stress anomalies may occur are determined and prioritized.

The working principle of the technical scheme is that potential association between each stress abnormality mode and different equipment is deeply analyzed. For example, for stress anomaly patterns due to cooling non-uniformity, key parameters focused on cooling equipment (e.g., cooling air knives, cooling water tanks, etc.) include position, angle, wind speed/water flow speed adjustment range, etc. Because these parameters directly affect the cooling effect, if the setting is improper, uneven cooling is easily caused, and thus abnormal stress is generated. For the stress abnormality caused by uneven stress in the conveying process, parameters such as roller way distance, roller way material, transmission speed stability and the like of conveying equipment are related, and the stress condition of glass in the conveying process can be influenced by the changes of the parameters, so that the stress abnormality is caused. Based on the analysis result, a device-stress anomaly mode association matrix is constructed. The rows of the matrix represent different stress anomaly patterns, the columns represent different devices and their parameters, and the matrix elements represent the strength of the association between the device parameters and the stress anomaly patterns. The matrix form can intuitively display the association relation between the equipment parameters and the stress abnormal mode, and provides basic data support for subsequent analysis and positioning.

And dividing the movement track of the glass on the production line into a plurality of key path segments according to the glass transmission path data. For example, from the time when the glass enters the cooling region, the glass is divided into a cooling stage, a transition stage, a detection stage, and the like. Different path sections have different process characteristics and stress conditions, and have different influences on occurrence of stress abnormality. For each path segment, the probability of a stress anomaly pattern occurring at that path segment is evaluated. The factors such as the residence time of the glass in the path section, the speed change, the stress condition and the like are considered. For example, in the cooling stage, the glass is cooled at a relatively low speed, and if the cooling is uneven, the path section has a high probability of occurrence of stress abnormality, and in the inspection stage, if the inspection apparatus applies an improper force to the glass, the stress abnormality may be caused. And combining the stress anomaly mode ordered list and the equipment-stress anomaly mode associated matrix, establishing a path segment-stress anomaly probability evaluation model, and calculating the probability of each path segment in various stress anomaly modes so as to determine the potential distribution situation of stress anomalies on the glass transmission path.

And constructing a space topology model of the glass production line by using the equipment layout data, and determining the space position relationship and the connection relationship among the equipment and the circulation direction of the glass among the equipment. The model can intuitively display the spatial layout of the production line and the flow path of the glass, and provides a basis for analyzing the relationship between stress abnormality and equipment spatial distribution. The stress anomaly pattern is spatially projected and analyzed for its correlation with the spatial distribution of the device. For example, if a stress anomaly pattern spatially characterizes a distribution around a cooling device, it may be initially determined that the device may have a problem that may lead to stress anomalies. And carrying out cluster analysis on the distribution of different stress abnormality modes in space through a spatial clustering algorithm (such as a DBSCAN algorithm), identifying a high-incidence area or equipment cluster with stress abnormality, and further reducing the area range where the stress abnormality possibly occurs.

And a path tracking algorithm is adopted to simulate the movement process of the glass on a production line, and the actions of equipment, received force, temperature and the like of the glass at different time points are recorded. By simulating the movement process of the glass, the stress condition and the temperature change of the glass in the whole production process can be known, and detailed data are provided for analyzing the generation of stress abnormality. And simulating the propagation process of stress in the glass by combining the stress abnormal mode feature library. For example, regarding stress abnormality due to a temperature gradient, propagation path and strength change of stress from a high temperature region to a low temperature region are simulated according to heat conduction characteristics and temperature change conditions of glass. By simulation, the specific distribution position of stress abnormality on glass and the area range possibly affected are determined, so that more accurate information is provided for positioning the stress abnormality area.

And fusing the device and stress abnormal mode relevance analysis result, the glass transmission path segmentation and stress influence evaluation result, the space topological relation analysis result and the path tracking and stress propagation simulation result. And (3) adopting a weighted comprehensive evaluation method, and distributing corresponding weights for each analysis result according to the importance and the reliability of different analysis results. For example, the device and stress anomaly pattern correlation analysis results may have higher reliability and may be given a greater weight, while the path tracking and stress propagation simulation results may be given a lesser weight when the data is incomplete. By comprehensive calculation, specific areas where stress anomalies may occur are determined and prioritized. The region with a high priority indicates that stress abnormality is highly likely to occur and has a serious influence, and is required to be preferentially examined and processed. The priority ordering can help operators reasonably allocate resources and improve the efficiency of fault detection and processing.

The technical scheme has the advantages that by analyzing potential correlations between each stress abnormality mode and different devices and parameters thereof and constructing a device-stress abnormality mode correlation matrix, key devices and parameters causing stress abnormality can be more accurately identified, and the accuracy of abnormality mode diagnosis is improved.

The method comprises the steps of dividing the path segments of the production line according to the glass transmission path data, evaluating the probability of occurrence of stress abnormality of each path segment, reducing prediction errors caused by incomplete path analysis, and improving the reliability of abnormality occurrence prediction.

By constructing a spatial topology model of the glass production line and carrying out cluster analysis on the spatial distribution of different stress abnormality modes by combining a spatial clustering algorithm, a high-incidence area or equipment cluster with abnormal stress can be effectively identified, and the comprehensiveness and the precision of the spatial analysis of the production line are enhanced.

The motion process of the glass on the production line is simulated through the path tracking algorithm, and the propagation process of stress on the glass can be more accurately simulated by combining the stress abnormality mode feature library, so that the prediction accuracy of the occurrence position and the occurrence range of stress abnormality is improved.

Different analysis results are weighted through a weighted comprehensive evaluation method, different weights are distributed according to the importance and the reliability of the analysis results, the interference of data noise can be effectively reduced, and the effectiveness and the reliability of comprehensive analysis are improved.

By comprehensively considering the association analysis result of the equipment and the stress abnormality mode, the path segment stress abnormality probability evaluation, the space topology analysis result and the path tracking simulation result, the possible area of the stress abnormality can be more accurately determined, and the high-risk area can be timely examined and processed according to the priority ranking.

By sequencing the priority of the specific areas with the possibility of stress abnormality, the method can help the maintenance personnel of the production line to treat the areas with higher risks preferentially, reduce the influence of faults on the production line and improve the maintenance accuracy and timeliness of the production line.

By comprehensively analyzing a plurality of factors, clear priority ordering of stress abnormal areas is provided, so that the decision cost of maintenance personnel in fault processing can be reduced, and the efficiency of fault detection and processing is improved.

By combining various analysis methods (relevance analysis, path segment evaluation, space analysis, simulation analysis and the like), the intelligent degree and the prediction capability of the whole system are enhanced, the human intervention is reduced, and the automation level of the system is improved.

By analyzing the stress abnormal areas in detail and combining with priority ordering, the production line can identify potential fault areas in advance and perform preventive maintenance before faults occur, so that the occurrence frequency of the faults of the production line is reduced, and the reliability of management is improved.

In one embodiment of the invention, a cold end stress detection system for a glass production line comprises a memory, a processor and a computer program stored on the memory and capable of running on the memory, wherein the processor executes the program to realize the cold end stress detection method for the glass production line.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims

1.A method for detecting cold end stress of a glass production line, the method comprising:

2. The method for detecting cold end stress of glass production line according to claim 1, wherein S1 comprises:

3. The method for detecting cold end stress of glass production line according to claim 1, wherein S2 comprises:

4. A method for detecting cold end stress of a glass production line according to claim 3, wherein S23 comprises:

Filling the normalized space-time characteristic values to the corresponding positions of the matrix one by one according to a time-space index system, estimating and filling the data missing condition, and forming a complete stress space-time characteristic matrix after filling all the characteristic values;

5. The method for detecting cold end stress of glass production line according to claim 1, wherein the step S3 comprises:

s33, combining cold end stress abnormal case data of the historical glass, and carrying out abnormal feature matching on each stress feature cluster;

6. The method for detecting cold end stress of glass production line according to claim 1, wherein S4 comprises:

7. The method for detecting cold end stress of glass production line according to claim 1, wherein S5 comprises:

8. The method for detecting cold end stress of glass production line according to claim 1, wherein S6 comprises:

9. The method for detecting cold end stress of glass production line according to claim 8, wherein S62 comprises:

And determining specific areas where stress abnormality occurs through comprehensive calculation, and prioritizing the areas.

10. A cold end stress detection system for a glass production line, comprising a memory, a processor and a computer program stored on said memory and operable on said memory, said processor executing said program to implement a cold end stress detection method for a glass production line as claimed in any one of claims 1 to 9.