Printing machine top project establishment design method based on multivariate data

Abstract

The invention provides a multivariate data-based printing machine top project approval design method, which relates to the technical field of printing equipment, and comprises the following steps of: acquiring printing machine operation data comprising printing quality parameters, equipment working condition parameters and environment parameters; performing feature analysis and parameter extraction through an adaptive fuzzy clustering algorithm, dynamically adjusting a sample classification membership degree, identifying a key parameter combination influencing the printing quality, and constructing a correlation matrix; constructing a dynamic Bayesian probability network model based on a parameter mapping relation, constructing a core network structure through an information entropy value and information gain, and establishing probability association between a printing quality parameter and an equipment working condition parameter; and determining the optimal parameter configuration of the function modules of the printing machine by utilizing a bidirectional constraint dynamic programming algorithm and combining the coupling weight and the state transition equation among the function modules, and generating a top design scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of printing equipment, and particularly to a top-level project design method for a printing press based on multi-source data. Background Art

[0002] With the rapid development of the printing industry, the market's demand for personalized customization of printing equipment is increasing day by day. The top-level design project establishment of a printing press needs to consider multiple factors such as the technical parameters of existing equipment and the specific needs of users at the same time. At present, when printing press manufacturers carry out new product project establishment, they mainly rely on experience for parameter selection and demand weight allocation, lacking a scientific and systematic decision-making support method.

[0003] However, the existing technology still has deficiencies. Traditional parameter analysis methods are difficult to effectively handle the complex correlation relationships between multi-source heterogeneous data of printing presses and cannot accurately identify the key parameter combinations affecting printing quality; most of the existing quality prediction models adopt static modeling methods and do not fully consider the dynamic coupling characteristics between parameters, resulting in low prediction accuracy; the parameter configuration optimization of printing press functional modules usually adopts one-way constraint methods and does not comprehensively consider the mutual influence between modules and the requirements of quality objectives, making it difficult to obtain a globally optimal design scheme.

[0004] In summary, the present invention aims to solve the above technical problems and proposes a top-level project design method for a printing press based on multi-source data analysis. Through an adaptive fuzzy clustering algorithm, the operation data of the printing press is analyzed for features and key parameters are identified. A dynamic association model between parameters is established by combining a dynamic Bayesian probability network, and a two-way constraint dynamic programming algorithm is used to optimize the allocation of parameter weights, realizing a scientific weight configuration of equipment parameters and user requirements. It not only maintains the ability to deeply mine multi-source heterogeneous data of printing presses but also enhances the adaptability of the top-level design scheme to the actual needs of users, thereby improving the reliability and practical value of the top-level design of printing presses. Summary of the Invention

[0005] An embodiment of the present invention provides a top-level project design method for a printing press based on multi-source data, which can solve the problems in the existing technology.

[0006] A top-level project design method for a printing press based on multi-source data is provided, including: Obtain the operation data of the printing press, where the operation data of the printing press includes printing quality parameters, equipment working condition parameters, and environmental parameters; Use an adaptive fuzzy clustering algorithm to perform feature analysis and parameter extraction on the operation data of the printing press, and identify the key parameter combinations affecting printing quality by determining the number of clusters and dynamically adjusting the membership degree of sample classification; Construct a correlation matrix between the key parameter combinations; Construct a dynamic Bayesian probability network model according to the parameter mapping relationship in the correlation matrix, construct the core network structure by calculating the information entropy value and information gain, optimize the conditional probability distribution using the logarithmic likelihood expectation value, establish a probability association between the printing quality parameters and the equipment condition parameters, and generate a printing quality prediction result; Based on the correlation matrix and the printing quality prediction result, use the bidirectional constraint dynamic programming algorithm to construct a state transition equation through the coupling weights between functional modules, and combine the bidirectional constraints to determine the optimal parameter configuration of the printing press functional modules, and generate a top-level design scheme for the printing press.

[0007] In an alternative embodiment, Adopt an adaptive fuzzy clustering algorithm to perform feature analysis and parameter extraction on the printing press operation data. By determining the number of clusters and dynamically adjusting the membership degree of sample classification, identify the key parameter combinations that affect printing quality, including: Segment the printing press operation data by the sliding window method, and calculate and determine the feature vectors based on each window; Calculate the minimum description length value of the feature vectors to determine the target cluster number interval. Calculate the fuzzy partition coefficient value and the partition entropy value within the target cluster number interval, select the cluster number corresponding to the largest fuzzy partition coefficient value and the smallest partition entropy value, and determine the final number of clusters; Construct an initial value of the sample classification membership degree, calculate the distribution parameters of each dimension of the feature vectors, and obtain the sample classification membership degree value through dynamic adjustment according to the distribution parameters; Calculate the cluster center coordinates according to the sample classification membership degree value, calculate the spatial position relationship strength between samples as a penalty factor to construct an objective function; Iteratively update the sample classification membership degree value and the cluster center coordinates according to the final number of clusters, and minimize the objective function to determine the clustering result; Based on the clustering result, analyze the feature patterns of each cluster center, calculate the contribution degree of each feature in distinguishing different categories using the statistical discrimination ratio, and select the feature combination with the statistical discrimination ratio greater than the preset threshold as the key parameter combination for printing quality; Input the key parameter combination into the support vector machine classifier for cross-validation to verify the effectiveness of the key parameter combination.

[0008] In an alternative embodiment, Obtaining the sample classification membership degree value through dynamic adjustment according to the distribution parameters includes: By calculating the mean Euclidean distance between a sample point and its nearest neighbor sample points, a local density index is obtained. Combining the standard deviation of the distance sequence from the sample point to the cluster center to calculate the distance deviation value, determining the local reliability value, and generating an adaptive weight factor. According to whether the local reliability value is lower than a threshold, an exponential decay coefficient is introduced to adjust the sample classification membership value, specifically including: For each sample point, select multiple nearest neighbor sample points, calculate the Euclidean distance between the sample point and the nearest neighbor sample points, and take the mean of the Euclidean distances as the local density index of the sample point; calculate the Euclidean distance from each sample point to each cluster center to obtain a distance sequence, and calculate the standard deviation of the distance sequence as the distance deviation value of the sample point; divide the local density index by the distance deviation value to obtain the local reliability value of the sample point; According to the local reliability value, combined with a preset adjustment coefficient, generate an adaptive weight factor; Multiply the adaptive weight factor by the initial value of the sample classification membership to obtain the adjusted value of the sample classification membership; When the local reliability value is lower than the preset reliability threshold, multiply the exponential decay coefficient by the adjusted value of the sample classification membership to obtain the sample classification membership value; when the local reliability value is not lower than the reliability threshold, take the adjusted value of the sample classification membership as the sample classification membership value; where the exponential decay coefficient specifically refers to: ; Wherein, D represents the exponential decay coefficient, represents the reliability decay rate, T r represents the reliability threshold, R l represents the local reliability value.

[0009] In an alternative embodiment, Using the statistical discrimination ratio to calculate the contribution of each feature in distinguishing different classes, and selecting the feature combination with the statistical discrimination ratio greater than the preset threshold as the key parameter combination for print quality includes: Calculate the between-class scatter matrix and the within-class scatter matrix, and determine the statistical discrimination ratio sequence of the features; at the same time, calculate the mutual information matrix between the features to judge the marked redundant features; through a progressive strategy, starting from the feature with the maximum statistical discrimination ratio, and gradually constructing the feature combination by iteratively calculating the statistical discrimination ratio gain value, specifically including: Determine the between-class scatter matrix by calculating the difference between the means of different class samples, and determine the within-class scatter matrix by calculating the difference between the samples within the same class and the class mean of the corresponding class; based on the between-class scatter matrix and the within-class scatter matrix, calculate the statistical discrimination ratio of each feature to obtain the feature statistical discrimination ratio sequence; Calculate the mutual information value between features to obtain a feature mutual information matrix. When the mutual information value is greater than the first preset threshold, mark the features with a statistical discrimination ratio less than the second preset threshold in the corresponding feature pair as redundant features; select the feature with the largest statistical discrimination ratio and not marked as a redundant feature from the feature statistical discrimination ratio sequence as the initial key feature; For the features that are not marked as redundant features and not selected as the initial key features, sort them in descending order of the statistical discrimination ratio to obtain a candidate feature sequence; add the features in the candidate feature sequence to the existing key feature combination one by one, calculate the combined statistical discrimination ratio before addition and the combined statistical discrimination ratio after addition respectively, and obtain the statistical discrimination ratio gain value; When the statistical discrimination ratio gain value is greater than the third preset threshold, add the corresponding candidate feature to the key feature combination; Repeat the execution until all the features in the candidate feature sequence are processed to obtain the final key feature combination.

[0010] In an alternative embodiment, Calculate the between-class scatter matrix and the within-class scatter matrix, and determine that the feature statistical discrimination ratio sequence includes: Calculate the number of samples in each category of the multi-class sample data, and perform normalization processing to obtain a category weight vector, and construct a category weight matrix; Calculate the category mean vector of each category of samples in the multi-class sample data, and the overall mean vector; calculate the difference between the category mean vector and the overall mean vector, and construct a category mean difference matrix; multiply the category mean difference matrix by the category weight matrix to obtain the between-class scatter matrix; Calculate the difference between each category of samples in the multi-class sample data and the corresponding category mean vector, and construct a sample difference matrix; multiply the sample difference matrix by the category weight matrix to obtain the within-class scatter matrix; Extract the diagonal elements of the between-class scatter matrix and the diagonal elements of the within-class scatter matrix, and determine the between-class scatter vector and the within-class scatter vector respectively; Calculate the ratio of the between-class scatter vector to the within-class scatter vector to obtain the feature statistical discrimination ratio sequence.

[0011] In an alternative embodiment, According to the parameter mapping relationship in the correlation matrix, construct a dynamic Bayesian probability network model, build a core network structure by calculating the information entropy value and the information gain, optimize the conditional probability distribution using the logarithmic likelihood expectation value, establish a probability association between the printing quality parameters and the equipment working condition parameters, and generate a printing quality prediction result including: Based on the dynamic Bayesian probability network, construct the initial network structure through the correlation matrix, use the information entropy value to calculate the information gain, screen the parameter nodes to construct the core network structure, and realize parameter prediction through the optimization of the logarithmic likelihood expectation value and the real-time calculation of the information gain, specifically including: Establish the connection between parameter pairs according to the correlation coefficients of parameter pairs in the correlation matrix; determine the direction of the connection based on the time series characteristics of the parameter pairs, and construct the initial network structure; Calculate the information entropy value of each parameter node in the initial network structure, construct an evaluation index of parameter importance based on the information entropy value, and calculate the information gain of each parameter node on the prediction result; sort the parameter nodes according to the information gain, select the parameter nodes with the information gain greater than the dynamic threshold to construct the core network structure, obtain the corresponding parent node set for the parameter nodes in the core network structure, and construct the conditional probability distribution based on the parent node set; Calculate the logarithmic likelihood expectation value based on the core network structure and the conditional probability distribution, and maximize the logarithmic likelihood expectation value through iterative optimization to obtain the optimized conditional probability distribution; Calculate the information gain of the parameter nodes at the current sampling moment, update the parameter composition of the core network structure according to the change amount of the information gain, and construct a new core network structure; based on the new core network structure and the corresponding conditional probability distribution, calculate the parameter prediction probability at the next sampling moment; Select the parameter value corresponding to the maximum parameter prediction probability as the printing quality prediction result.

[0012] In an alternative embodiment, Based on the correlation matrix and the printing quality prediction result, use the bidirectional constraint dynamic programming algorithm to construct a state transition equation through the coupling weights between functional modules, and combine the bidirectional constraints to determine the optimal parameter configuration of the printing press functional modules, and generate the top-level design scheme of the printing press, including: Calculate the coupling weights between functional modules according to the correlation matrix, and the coupling weights represent the degree of mutual influence between different functional modules; Construct a state transition equation according to the coupling weights, multiply the parameter state of the previous functional module by the corresponding coupling weight and sum through the state transition equation to obtain the parameter state of the current functional module; Establish bidirectional constraint conditions according to the printing quality prediction result, and the bidirectional constraint conditions include forward constraint conditions and backward constraint conditions. The forward constraint conditions limit the adjustment range of the parameters corresponding to the current functional module according to the parameter state of the previous functional module; the backward constraint conditions ensure that the subsequent functional modules reach the preset quality target according to the printing quality prediction result; Traverse each functional module forward according to the forward constraint condition, calculate the local optimal solution and record the state transition path to obtain the state transition sequence; according to the backward constraint condition, start from the last functional module and trace back the state transition sequence in reverse to determine the optimal parameter configuration of the functional module. Input the optimal parameter configuration into the printing quality prediction model for verification. When the printing quality prediction result obtained by verification does not meet the requirements, adjust the bidirectional constraint condition and re-execute the parameter configuration optimization.

[0013] In the embodiment of the present invention, by collecting multi-dimensional data such as printing quality, equipment working conditions, and environment, and adopting an adaptive fuzzy clustering algorithm to achieve dynamic sample classification and key parameter combination recognition, the data feature extraction is more accurate; using cluster analysis to determine the core parameters affecting printing quality and constructing a correlation matrix can reveal the internal mapping relationship between parameters, helping designers intuitively understand the interaction and influence mechanism among multiple factors, and effectively reducing design uncertainty; constructing a dynamic Bayesian probability network model, optimizing the conditional probability distribution through information entropy, information gain, and logarithmic likelihood expectation value, realizing the probability association between printing quality and equipment working condition parameters, thereby accurately predicting printing quality, improving the system prediction ability and response speed; based on the correlation matrix and prediction results, using the bidirectional constraint dynamic programming algorithm to construct a state transition equation, determining the optimal parameter configuration from the perspective of the coupling weight of functional modules, realizing the overall coordination and optimization of each functional module of the printing machine, and finally generating a scientific and reasonable top-level design scheme, reducing the trial-and-error cost and improving the design efficiency. The present invention integrates multi-dimensional information in a data-driven manner and adopts advanced algorithm means, which helps to achieve intelligent decision-making and system optimization under complex working conditions, thereby improving the overall performance and product quality of the printing machine and promoting the research and development and design of printing equipment to a higher level. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic flowchart of the top-level project design method for a printing machine based on multi-source data according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0016] The technical solution of the present invention will be described in detail below with specific embodiments. The following several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0017] Figure 1 FIG. is a schematic flow chart of the top-level project design method of the multi-source data printing press according to the embodiment of the present invention. As Figure 1 shown, the method includes: Obtain the printing press operation data, which includes printing quality parameters, equipment working condition parameters, and environmental parameters; The printing press operation data is obtained in real time through the sensor acquisition system installed on the printing press. Among them, the printing quality parameters include registration accuracy, ink color density, dot gain value, printing contrast, ink layer thickness, and graphic clarity; the equipment working condition parameters include printing speed, paper tension, impression pressure, ink fountain opening, paper feed wheel speed, and transmission gear clearance; the environmental parameters include environmental temperature, environmental humidity, dust concentration, and air flow rate.

[0018] Specifically, the acquisition of printing quality parameters is achieved through the following methods: Install a CCD camera at the paper output end of the printing press to collect printing images, and calculate the registration accuracy through image processing algorithms; Use a spectrophotometer to measure the ink color density and printing contrast of the printed matter; Measure the dot size with a microscope to calculate the dot gain value; Use an ink layer thickness gauge to measure the ink layer thickness; Evaluate the graphic clarity through image analysis software.

[0019] The acquisition of equipment working condition parameters is achieved through the following methods: Use an optical encoder to measure the printing speed and paper feed wheel speed; Detect the paper tension with a tension sensor; Monitor the impression pressure through a pressure sensor; Measure the ink fountain opening with a displacement sensor; Use a clearance gauge to measure the transmission gear clearance.

[0020] The acquisition of environmental parameters is achieved through the following methods: Install temperature and humidity sensors around the printing press to monitor the environmental temperature and humidity in real time; Use a dust concentration detector to measure the dust content in the air; Measure the air flow rate of the printing environment with an anemometer.

[0021] All the collected data is transmitted to the data acquisition server through the industrial Ethernet and stored according to a unified time tag. The data sampling period is 100 ms, and the data is normalized and outliers are removed every 1 hour to generate a standardized data set for subsequent analysis.

[0022] The collected data is stored in XML format. The specific data structure includes fields such as data acquisition timestamp, parameter type identifier, parameter value, data unit, and acquisition device number. By establishing a parameter index table, fast retrieval and correlation analysis of different types of parameters are realized.

[0023] To ensure the reliability of the data, the system also establishes a data quality assessment mechanism: regularly calibrate the sensors; set the range limit of parameter changes to judge outliers; adopt a data redundancy backup strategy; establish a sensor fault diagnosis and data compensation mechanism.

[0024] In an optional embodiment, an adaptive fuzzy clustering algorithm is used to perform feature analysis and parameter extraction on the operating data of the printing press. By determining the number of clusters and dynamically adjusting the membership degree of sample classification, the key parameter combinations affecting printing quality are identified, including: Segment the operating data of the printing press by the sliding window method, and calculate and determine the feature vectors based on each window; Calculate the minimum description length value of the feature vectors to determine the target cluster number interval. Calculate the fuzzy partition coefficient value and the partition entropy value within the target cluster number interval, select the cluster number corresponding to the largest fuzzy partition coefficient value and the smallest partition entropy value to determine the final number of clusters; Construct the initial value of the sample classification membership degree, calculate the distribution parameters of each dimension of the feature vectors, and obtain the sample classification membership degree value through dynamic adjustment according to the distribution parameters; Calculate the cluster center coordinates according to the sample classification membership degree value, calculate the strength of the spatial position relationship between samples as a penalty factor to construct an objective function; Iteratively update the sample classification membership degree value and the cluster center coordinates according to the final number of clusters, and minimize the objective function to determine the clustering result; Analyze the feature patterns of each cluster center based on the clustering result, calculate the contribution degree of each feature in distinguishing different categories using the statistical discrimination ratio, and select the feature combination with the statistical discrimination ratio greater than the preset threshold as the key parameter combination for printing quality; Input the key parameter combination into a support vector machine classifier for cross-validation to verify the effectiveness of the key parameter combination.

[0025] In a specific implementation manner, The first step: data segmentation and feature extraction; segment the operating data of the printing press by the sliding window method. Specifically, set the window length to 1 hour and the sliding step to 10 minutes. Calculate the statistical features of the data within each window, including mean, standard deviation, peak value, skewness, kurtosis, etc., to form feature vectors.

[0026] Step 2: Determine the optimal number of clusters; First, calculate the minimum description length value of the feature vector to preliminarily determine that the target number of cluster intervals is 2 - 10 classes. Within this interval, calculate the fuzzy partition coefficient value and the partition entropy value for different numbers of clusters one by one. The larger the fuzzy partition coefficient value, the clearer the clustering result, and the smaller the partition entropy value, the more stable the clustering result. Through comparative analysis, select the number of clusters corresponding to the largest fuzzy partition coefficient value and the smallest partition entropy value as the final number of clusters.

[0027] Step 3: Initialize and dynamically adjust the membership degree; Construct the initial value of the sample classification membership degree through a random method, and calculate distribution parameters such as the mean and variance of each dimension of the feature vector. Based on the distance relationship between the sample point and its nearest neighbor sample point, calculate the local density index. Combine the distance distribution of the sample point to each cluster center to dynamically adjust the sample classification membership degree value.

[0028] Step 4: Construct and optimize the objective function; Calculate the cluster center coordinates of each category based on the sample classification membership degree value. By calculating the Euclidean distance between sample points, determine the spatial position relationship strength between samples, and use it as a penalty factor to construct the objective function. The objective function comprehensively considers the distance between the sample and the cluster center and the spatial relationship between samples.

[0029] Step 5: Iteratively optimize the clustering result; According to the determined final number of clusters, alternately update the sample classification membership degree value and the cluster center coordinates. By minimizing the objective function, continuously optimize the clustering result until the convergence condition is reached or the maximum number of iterations is reached.

[0030] Step 6: Feature contribution analysis; Analyze the feature patterns of each cluster center and calculate the statistical discrimination ratio of each feature when distinguishing different categories. The larger the statistical discrimination ratio, the greater the contribution of the feature to category discrimination. Select the feature combination with a statistical discrimination ratio greater than 0.8 as the key parameter combination for printing quality.

[0031] Step 7: Validation of effectiveness; Input the identified key parameter combination into the support vector machine classifier and use the 10-fold cross-validation method to verify its effectiveness. Evaluate the reliability of the key parameter combination by calculating indicators such as classification accuracy, precision, and recall.

[0032] Exemplarily, taking the operation data of a printing press as an example, 480 samples are obtained after segmentation by a sliding window. Through calculation, the optimal number of clusters is found to be 4. By dynamically adjusting the membership degree of sample classification, the final clustering results show that: the first category mainly represents the normal printing state, the second category reflects the registration deviation state, the third category corresponds to the ink color abnormality state, and the fourth category represents the comprehensive quality abnormality state. Through statistical discriminant ratio analysis, 8 key parameters such as ink fountain opening, impression pressure, paper tension, and environmental humidity are identified. These parameters are input into a support vector machine classifier for cross-validation, and the classification accuracy reaches 92.5%, verifying the effectiveness of the key parameter combination. These key parameters are then used to construct a correlation matrix, providing a basis for subsequent quality prediction and parameter optimization.

[0033] In this embodiment, the operation data of the printing press is carefully segmented by the sliding window method, and key statistical features (such as mean, standard deviation, peak value, etc.) are extracted to effectively capture the time-varying characteristics of the data, providing high-quality feature data for subsequent clustering and analysis; through the optimization analysis of the fuzzy partition coefficient and partition entropy value, the optimal number of clusters is accurately determined, making the clustering results clearer and more stable, and improving the accuracy and interpretability of data classification; by dynamically adjusting the membership degree of sample classification, combining the local density and the distance relationship between the sample and the cluster center, the clustering effect is optimized, enhancing the adaptability and classification accuracy of the clustering model to different data types; through feature contribution degree analysis and support vector machine verification, the key parameter combination that has the greatest impact on printing quality is successfully identified, and its high accuracy in actual classification is verified, providing a reliable basis for quality prediction and parameter optimization.

[0034] In an alternative embodiment, obtaining the sample classification membership degree value by dynamic adjustment according to the distribution parameter includes: By calculating the mean Euclidean distance between the sample point and its nearest neighbor sample points, a local density index is obtained. Combining the standard deviation of the distance sequence from the sample point to the cluster center to calculate the distance deviation value, determining the local reliability value, and generating an adaptive weight factor. According to whether the local reliability value is lower than the threshold, an exponential decay coefficient is introduced to adjust the sample classification membership degree value, specifically including: Select multiple nearest neighbor sample points for each sample point, calculate the Euclidean distance between the sample point and the nearest neighbor sample points, and take the mean of the Euclidean distances as the local density index of the sample point; calculate the Euclidean distance from each sample point to each cluster center to obtain a distance sequence, and calculate the standard deviation of the distance sequence as the distance deviation value of the sample point; divide the local density index by the distance deviation value to obtain the local reliability value of the sample point; Generate an adaptive weight factor according to the local reliability value in combination with a preset adjustment coefficient; Multiply the adaptive weight factor by the initial value of the sample classification membership degree to obtain the adjusted value of the sample classification membership degree; When the local reliability value is lower than the preset reliability threshold, multiply the exponential decay coefficient by the adjusted value of the sample classification membership degree to obtain the sample classification membership degree value; when the local reliability value is not lower than the reliability threshold, use the adjusted value of the sample classification membership degree as the sample classification membership degree value; where the exponential decay coefficient specifically refers to: ; where, D represents the exponential decay coefficient, represents the reliability decay rate, T r represents the reliability threshold, R l represents the local reliability value.

[0035] In a specific implementation, The first step: Calculate the local density index; for each sample point, select the 5 nearest neighboring sample points as its nearest neighbor sample points. Calculate the Euclidean distance from the current sample point to these 5 nearest neighbor sample points, and take the average of these distance values as the local density index of the sample point. The smaller the local density index, the denser the sample distribution in this area.

[0036] The second step: Calculate the distance deviation value; calculate the Euclidean distance from each sample point to all cluster centers to obtain a set of distance sequences. Calculate the standard deviation of this distance sequence as the distance deviation value of the sample point. The larger the distance deviation value, the more obvious the difference in the distances from the sample point to each cluster center.

[0037] The third step: Determine the local reliability value; divide the local density index of the sample point by its distance deviation value to obtain the local reliability value. This value reflects the distribution characteristics of the sample point in the local area and the reliability degree of the class attribution. The higher the local reliability value, the more reliable the class attribution of the sample point.

[0038] The fourth step: Generate the adaptive weight factor; generate the adaptive weight factor according to the local reliability value of the sample point, in combination with the preset adjustment coefficient of 0.8. The adjustment coefficient is used to control the adjustment amplitude of the weight factor to prevent excessive adjustment.

[0039] The fifth step: Calculate the adjusted membership degree value; multiply the adaptive weight factor by the initial membership degree value of the sample to obtain the adjusted membership degree value. This step realizes the preliminary adjustment based on local features.

[0040] Step 6: Introduce exponential decay adjustment; set the reliability threshold to 0.6 and the reliability decay rate to 0.5. When the local reliability value of a sample point is lower than the threshold 0.6, calculate the exponential decay coefficient and multiply it by the membership adjustment value to obtain the final membership value. When the local reliability value is not lower than the threshold, directly use the membership adjustment value as the final membership value.

[0041] Exemplarily, take the data of a sample point in a certain printing process as an example. This sample point represents the printing state at a certain moment, and its feature vector includes parameters such as ink color density and registration accuracy. Through calculation, find the 5 nearest neighboring sample points, and calculate the average Euclidean distance to be 0.15 as the local density index. Calculate the distance sequence from this sample point to 4 cluster centers as [0.8, 0.6, 0.9, 1.2], and obtain the distance standard deviation 0.25 as the distance deviation value.

[0042] Calculate the local reliability value to be 0.15 / 0.25 = 0.6. Since this value is equal to the reliability threshold 0.6, use the adjustment coefficient 0.8 to generate the adaptive weight factor 0.75. Assume the initial membership value of this sample point to a certain category is 0.8, then the membership adjustment value is 0.8×0.75 = 0.6. Since the local reliability value is equal to the threshold, directly take 0.6 as the final membership value, indicating that the belonging degree of this sample point to this category is 0.6. This dynamic adjustment ensures that the clustering result can better reflect the local distribution characteristics of the data.

[0043] In this embodiment, by calculating the average Euclidean distance between the sample point and its neighboring sample points and the distance standard deviation to the cluster centers, the local density and distribution differences of the sample points can be accurately evaluated. This provides a basis for the dynamic adjustment of the sample classification membership, ensuring that the clustering analysis can better adapt to the local structure of the data; combining the local reliability value and the adjustment coefficient to generate an adaptive weight factor to dynamically adjust the membership of the sample point. This adaptive mechanism effectively improves the accuracy of sample classification, avoids excessive or insufficient classification adjustments, and ensures the consistency between the clustering result and the actual data distribution; when the local reliability value is lower than the preset threshold, introduce the exponential decay coefficient to decay and adjust the sample membership, further optimizing the clustering effect. This mechanism can effectively handle sample points with high uncertainty, ensuring the stability and accuracy of clustering; by combining the local reliability, weight factor and exponential decay mechanism, dynamically adjust the membership value of each sample point, making the clustering result more refined and reliable. This dynamic adjustment ensures that the clustering model can accurately reflect the belonging degree of each sample, improving the flexibility and accuracy of the overall clustering algorithm.

[0044] In an alternative embodiment, the contribution degree of each feature in distinguishing different categories is calculated by using the statistical discrimination ratio, and the feature combination with the statistical discrimination ratio greater than the preset threshold is selected as the key parameter combination of printing quality, including: Calculate the between-class scatter matrix and the within-class scatter matrix, and determine the statistical discrimination ratio sequence of the features; at the same time, calculate the mutual information matrix between the features to judge the marked redundant features; through a progressive strategy, starting from the feature with the largest statistical discrimination ratio, calculate the statistical discrimination ratio gain value through iterative calculation, and gradually construct the feature combination, specifically including: Determine the between-class scatter matrix by calculating the difference between the sample means of different categories, and determine the within-class scatter matrix by calculating the difference between the samples within the same category and the category mean of the corresponding category; based on the between-class scatter matrix and the within-class scatter matrix, calculate the statistical discrimination ratio of each feature to obtain the feature statistical discrimination ratio sequence; Calculate the mutual information value between the features to obtain the feature mutual information matrix. When the mutual information value is greater than the first preset threshold, mark the feature with the statistical discrimination ratio less than the second preset threshold in the corresponding feature pair as a redundant feature; select the feature with the largest statistical discrimination ratio and not marked as a redundant feature from the feature statistical discrimination ratio sequence as the initial key feature; Sort the features that are not marked as redundant features and not selected as the initial key features in descending order of the statistical discrimination ratio to obtain a candidate feature sequence; add the features in the candidate feature sequence to the existing key feature combination one by one, and calculate the combined statistical discrimination ratio before and after the addition respectively to obtain the statistical discrimination ratio gain value; When the statistical discrimination ratio gain value is greater than the third preset threshold, add the corresponding candidate feature to the key feature combination; Repeat until all the features in the candidate feature sequence are processed to obtain the final key feature combination.

[0045] In a specific implementation manner, The first step: Calculate the between-class scatter and the within-class scatter; first calculate the between-class scatter matrix: for each feature, calculate the mean of the samples of different categories, and then calculate the degree of difference between these category means. Specifically, the between-class scatter matrix is obtained by calculating the sum of the squared deviations of the category means from the overall mean.

[0046] Then calculate the within-class scatter matrix: for each feature, calculate the sum of the squared deviations between each sample within the same category and the category mean to obtain the within-class scatter matrix. Based on these two matrices, calculate the statistical discrimination ratio of each feature to form the feature statistical discrimination ratio sequence.

[0047] Step 2: Identify redundant features; calculate the mutual information values between all pairs of features to construct a feature mutual information matrix. Set the first preset threshold to 0.8 and the second preset threshold to 0.6. When the mutual information value of a pair of features is greater than 0.8, compare the statistical discrimination ratios of this pair of features, and mark the feature with a statistical discrimination ratio less than 0.6 as a redundant feature.

[0048] Step 3: Select the initial key features; from the sequence of feature statistical discrimination ratios, select the feature with the largest statistical discrimination ratio that has not been marked as redundant as the initial key feature. This feature will serve as the starting point for constructing the feature combination.

[0049] Step 4: Construct a candidate feature sequence; sort all the features that have not been marked as redundant and have not been selected as the initial key feature in descending order of their statistical discrimination ratios to form a candidate feature sequence.

[0050] Step 5: Iteratively construct the feature combination; add the features in the candidate feature sequence to the existing key feature combination one by one. For each added feature, calculate the combined statistical discrimination ratio before and after the addition to obtain the statistical discrimination ratio gain value. Set the third preset threshold to 0.1. When the gain value is greater than 0.1, include the candidate feature in the key feature combination.

[0051] Step 6: Complete feature selection; repeat the process of Step 5 until all candidate features have been processed, and finally obtain the complete key feature combination.

[0052] Exemplarily, taking the operating data of a printing press as an example, the initial features include: 12 feature parameters such as ink color density, registration accuracy, printing speed, paper tension, impression pressure, ink fountain opening, ambient temperature, and ambient humidity.

[0053] By calculating the between-class scatter and within-class scatter, the sequence of feature statistical discrimination ratios is obtained: ink color density (0.85), registration accuracy (0.82), impression pressure (0.78), paper tension (0.75), ink fountain opening (0.72), printing speed (0.68), ambient humidity (0.65), ambient temperature (0.58), etc.

[0054] After calculating the feature mutual information matrix, it is found that the mutual information value between ambient temperature and ambient humidity is 0.85, which is greater than the first preset threshold of 0.8. Since the statistical discrimination ratio of ambient temperature (0.58) is less than the second preset threshold of 0.6, it is marked as a redundant feature.

[0055] Select the ink color density (0.85) with the largest statistical discrimination ratio as the initial key feature. The remaining features that have not been marked as redundant are arranged in descending order of their statistical discrimination ratios to form a candidate feature sequence.

[0056] Through iterative calculations, when registering accuracy is added, the combined statistical discrimination ratio gain value is 0.15; when stamping pressure is added, the gain value is 0.12; when paper tension is added, the gain value is 0.11; when ink fountain opening is added, the gain value is 0.08. Therefore, the final combination of key features includes: ink density, registering accuracy, stamping pressure, and paper tension. These parameters will be used for subsequent printing quality control and optimization.

[0057] In this embodiment, by calculating the between-class scatter and within-class scatter, and combining the statistical discrimination ratio, the contribution of each feature in distinguishing different classes is effectively measured. This ensures that during the feature selection process, features that are most powerful for discriminating printing quality can be preferentially selected, thereby improving the accuracy of subsequent analysis and prediction; by calculating the mutual information matrix between features and setting a threshold, redundant features can be automatically identified and removed. This reduces the correlation between features, improves the independence of the feature combination, and ensures that the selected key features are not interfered by redundant information, thereby enhancing the reliability and performance of the model; adopting a progressive strategy, by iteratively calculating the statistical discrimination ratio gain value, features that are most contributive to printing quality control are gradually selected. This method ensures that the finally selected feature combination has the maximum discrimination ability in distinguishing different classes, thereby enhancing the effect of quality prediction; finally, the key feature combination obtained through iterative screening (such as ink density, registering accuracy, etc.) effectively reflects the core parameters affecting printing quality. These parameters can provide a scientific basis for subsequent quality control and optimization, and improve the stability of the printing process and product quality.

[0058] In an alternative embodiment, calculating the between-class scatter matrix and within-class scatter matrix, and determining the statistical discrimination ratio sequence of features includes: Calculating the number of samples in each category of multi-class sample data, and performing normalization processing to obtain a category weight vector, and constructing a category weight matrix; Calculating the category mean vector of each category of samples in the multi-class sample data, and the overall mean vector; calculating the difference between the category mean vector and the overall mean vector, and constructing a category mean difference matrix; multiplying the category mean difference matrix by the category weight matrix to obtain the between-class scatter matrix; Calculating the difference between each category of samples in the multi-class sample data and the corresponding category mean vector, and constructing a sample difference matrix; multiplying the sample difference matrix by the category weight matrix to obtain the within-class scatter matrix; Extracting the diagonal elements of the between-class scatter matrix and the diagonal elements of the within-class scatter matrix, respectively determining the between-class scatter vector and the within-class scatter vector; Calculating the ratio of the between-class scatter vector to the within-class scatter vector to obtain the statistical discrimination ratio sequence of features.

[0059] In a specific embodiment, Step 1: Construct a class weight matrix; count the number of samples in each class, divide the number of samples in each class by the total number of samples to obtain the class weights. For example, for sample data of 4 classes, calculate the proportion of samples in each class to form a class weight vector. Then expand the class weight vector into a diagonal matrix to construct the class weight matrix.

[0060] Step 2: Calculate the between-class scatter matrix; first calculate the class mean vector of the samples in each class, and at the same time calculate the overall mean vector of all samples. For each class, calculate the difference between its class mean vector and the overall mean vector to form a class mean difference matrix. Multiply the class mean difference matrix by the class weight matrix to obtain the between-class scatter matrix.

[0061] Step 3: Calculate the within-class scatter matrix; for the samples within each class, calculate the difference between them and the corresponding class mean vector to construct a sample difference matrix. Multiply the sample difference matrix by the class weight matrix to obtain the within-class scatter matrix.

[0062] Step 4: Extract the scatter vectors; extract the diagonal elements from the between-class scatter matrix to form the between-class scatter vector. Extract the diagonal elements from the within-class scatter matrix to form the within-class scatter vector. These two vectors respectively reflect the performance of each feature in class discrimination and within-class aggregation.

[0063] Step 5: Calculate the statistical discrimination ratio; divide the between-class scatter vector by the within-class scatter vector to obtain the statistical discrimination ratio sequence of the features. This sequence reflects the contribution degree of each feature in distinguishing different classes.

[0064] Exemplarily, taking the operation data of a printing press as an example, it includes 4 classes (normal state, registration deviation state, ink color abnormality state, comprehensive quality abnormality state), and a total of 12 feature parameters.

[0065] The number of samples in each class is statistically obtained as 200, 150, 100, and 50 respectively, and the total number of samples is 500.

[0066] The calculated class weight vector is [0.4, 0.3, 0.2, 0.1].

[0067] Construct a 4×4 class weight diagonal matrix.

[0068] Taking the ink density feature as an example: the mean of class 1 is 1.5, the mean of class 2 is 1.3, the mean of class 3 is 1.8, the mean of class 4 is 1.2; the overall mean is 1.45; calculate the difference between the mean of each class and the overall mean: [0.05, -0.15, 0.35, -0.25]; the between-class scatter value of this feature is obtained as 0.068 through matrix operations.

[0069] Calculate the difference between each sample within a category and the category mean: The sum of squared differences of samples in Category 1 is 0.02, in Category 2 is 0.03, in Category 3 is 0.04, and in Category 4 is 0.02. Through matrix operations, the within-class scatter value of this feature is obtained as 0.028.

[0070] For the ink color density feature, the between-class scatter is 0.068 and the within-class scatter is 0.028. The calculated statistical discrimination ratio is 0.068 / 0.028 = 2.43.

[0071] Calculate the statistical discrimination ratios of all features in this way. Finally, the sequence of feature statistical discrimination ratios is obtained as: ink color density (2.43), registration accuracy (2.15), stamping pressure (1.89), paper tension (1.76), etc. This sequence intuitively shows the importance of each feature for category discrimination.

[0072] In this embodiment, by counting the number of samples in each category and performing normalization processing, a category weight matrix is constructed, which provides a basis for subsequent between-class and within-class scatter calculations; this method effectively quantifies the relative importance of different categories and accurately calculates the between-class and within-class scatters through matrix operations, ensuring the accuracy and effectiveness of data processing; by calculating the statistical discrimination ratio of features, the contribution degree of each feature in distinguishing different categories can be quantified; the sequence of feature statistical discrimination ratios provides a clear ranking to help identify the features most influential for category discrimination, thereby optimizing feature selection and improving the effect of the classification model; through the ratio of between-class and within-class scatters, it can be clearly identified which features have strong discrimination ability for category discrimination and which features are weak; combined with mutual information analysis, redundant features can be further removed to improve the independence and discrimination ability of feature combinations, ensuring that the selected features maximize the improvement of classification accuracy; by calculating the statistical discrimination ratio of each feature, the most discriminative features can be selected according to the ratio ranking to form an optimized feature combination. This process improves the accuracy of printing quality prediction and control and provides a strong feature basis for subsequent model optimization and quality assessment.

[0073] Construct a correlation matrix between the described key parameter combinations; In a specific implementation, preprocess the identified key parameter combination data. First, perform data normalization to convert parameters with different dimensions to a unified scale. Then, detect and process outliers, and use the moving median method to fill in missing values. Finally, perform a stationarity test on the data and perform differencing processing if necessary to eliminate the trend effect.

[0074] The Pearson correlation coefficient method is used to calculate the linear correlation relationship between parameters. For each pair of parameters, their covariance is calculated and divided by the product of the standard deviations to obtain the correlation coefficient. The value range of the correlation coefficient is between -1 and 1, where 1 represents a perfect positive correlation, -1 represents a perfect negative correlation, and 0 represents no linear correlation.

[0075] The Spearman rank correlation coefficient method is used to evaluate the degree of non - linear correlation between parameters. First, the parameter values are converted into rank sequences, and then the correlation coefficient between the rank sequences is calculated. This can capture the monotonic non - linear relationship between parameters.

[0076] By sliding the time window, the correlation coefficients between parameters at different time delays are calculated. The maximum time lag is set to 10 sampling periods, and the correlation coefficients at different time delays are calculated step by step. The time delay value with the strongest correlation is selected as the final result.

[0077] With other parameters kept unchanged, the partial correlation coefficient between any two parameters is calculated. This can exclude the influence of other parameters and reflect the true degree of association between the two parameters.

[0078] The linear correlation coefficient, non - linear correlation coefficient, time - delay correlation, and conditional correlation are combined with weights. The linear correlation weight is set to 0.4, the non - linear correlation weight is 0.3, the time - delay correlation weight is 0.2, and the conditional correlation weight is 0.1. The comprehensive correlation coefficient is obtained through weighted summation.

[0079] The calculated comprehensive correlation coefficients are organized in matrix form. The rows and columns of the matrix correspond to the respective key parameters, and the matrix elements represent the comprehensive degree of correlation between parameters. The diagonal elements are 1, indicating the perfect correlation of a parameter with itself.

[0080] A significance test is performed on each element in the correlation matrix. The significance level is set to 0.05, and the P - value of each correlation coefficient is calculated. The non - significant correlation coefficients are set to 0, and the significant correlation results are retained.

[0081] The correlation matrix is sparsified. The correlation coefficient threshold is set to 0.3, and the correlation coefficients below the threshold are set to 0 to highlight the important correlation relationships. At the same time, the symmetry of the matrix is ensured, that is, the correlation of parameter A with parameter B should be equal to the correlation of parameter B with parameter A.

[0082] The final correlation matrix is converted into a heatmap form, with different shades of color representing the degree of correlation. Positive correlations are represented by the red color system, negative correlations are represented by the blue color system, and the stronger the correlation, the darker the color. This visual representation intuitively shows the strength of the association between parameters.

[0083] In this embodiment, through data normalization, outlier detection and processing, missing value filling, and stationarity testing, the quality and consistency of the input data are ensured, laying a solid foundation for subsequent correlation analysis. Through these steps, interference factors in the data are eliminated, and the accuracy of the analysis is improved. Multiple methods such as Pearson correlation coefficient, Spearman rank correlation coefficient, time-delay correlation, and conditional correlation coefficient are used to comprehensively evaluate the linear and non-linear relationships between parameters. Through sliding time window and partial correlation analysis, the true associations between parameters at different time delays and conditions are considered, providing a more comprehensive and refined analysis of parameter relationships. By weighted combination of different types of correlations (linear, non-linear, time-delay, conditional), a comprehensive correlation coefficient matrix is obtained. This weighted combination enables reasonable integration of different types of relationships, more accurately reflecting the comprehensive correlation degree between parameters and ensuring the comprehensiveness and effectiveness of the analysis. Through significance testing, sparsification processing, and finally heatmap visualization of the correlation matrix, intuitive and easy-to-understand analysis results of parameter associations are provided. The heatmap form effectively presents the strength of the correlation between parameters, helping decision-makers quickly identify important association relationships and providing a scientific basis for printing quality control.

[0084] In an alternative embodiment, according to the parameter mapping relationship in the correlation matrix, a dynamic Bayesian probability network model is constructed. By calculating the information entropy value and information gain, the core network structure is constructed, and the conditional probability distribution is optimized using the logarithmic likelihood expectation value. The printing quality parameters and the equipment condition parameters are probabilistically associated to generate a printing quality prediction result, including: Based on the dynamic Bayesian probability network, an initial network structure is constructed through the correlation matrix. Using the information entropy value, the information gain is calculated, and parameter nodes are screened to construct the core network structure. Parameter prediction is achieved through iterative optimization of the logarithmic likelihood expectation value and real-time calculation of the information gain, specifically including: Based on the correlation coefficients of parameter pairs in the correlation matrix, connections are established between parameter pairs; based on the time-series characteristics of the parameter pairs, the direction of the connections is determined to construct an initial network structure; Calculate the information entropy value of each parameter node in the initial network structure, construct an evaluation index of parameter importance based on the information entropy value, and calculate the information gain of each parameter node for the prediction result; sort the parameter nodes according to the information gain, select the parameter nodes with information gain greater than the dynamic threshold to construct the core network structure, obtain the corresponding parent node set for the parameter nodes in the core network structure, and construct the conditional probability distribution based on the parent node set; Calculate the logarithmic likelihood expectation value based on the core network structure and the conditional probability distribution, and maximize the logarithmic likelihood expectation value through iterative optimization to obtain the optimized conditional probability distribution; Calculate the information gain of the parameter nodes at the current sampling moment, update the parameter composition of the core network structure according to the change amount of the information gain, and construct a new core network structure; calculate the parameter prediction probability at the next sampling moment based on the new core network structure and the corresponding conditional probability distribution; Select the parameter value corresponding to the maximum parameter prediction probability as the printing quality prediction result.

[0085] In a specific embodiment, Based on the correlation coefficients in the correlation matrix, determine the connection relationships between parameters. Set the correlation coefficient threshold to 0.5, and establish a connection when the correlation coefficient of a parameter pair is greater than the threshold. By analyzing the time-series data of the parameters, determine the order of parameter changes, and accordingly determine the direction of the connection to construct a directed network structure.

[0086] For each parameter node in the network, statistically analyze the probability distribution of its historical data and calculate the information entropy value. Construct a parameter importance index based on the information entropy value. Evaluate its contribution to the prediction result by calculating the information gain of each parameter node for the prediction target. Set the dynamic threshold to 1.2 times the mean value of the information gain, and filter out the parameter nodes with information gain greater than this threshold.

[0087] Form the core network structure with the filtered high-information-gain parameter nodes. For each node, determine its parent node set, that is, the upstream nodes that have a direct impact on this node. Based on the statistical analysis of historical data, establish a conditional probability distribution table for each node relative to its parent node set.

[0088] Use the expectation maximization algorithm to optimize the conditional probability distribution. Calculate the logarithmic likelihood expectation value under the current network structure, and maximize the logarithmic likelihood expectation value by iteratively adjusting the conditional probability parameters. Set the maximum number of iterations to 100 times and the convergence threshold to 0.001.

[0089] At each sampling moment, recalculate the information gain of the parameter nodes. When the change amount of the information gain exceeds a preset threshold (such as 20%), update the parameter composition of the core network structure. Calculate the conditional probability distribution for the newly added nodes and prune the relevant connections of the removed nodes.

[0090] Based on the updated core network structure and the optimized conditional probability distribution, calculate the probabilities of possible values of each parameter at the next sampling moment. Select the combination of parameter values with the maximum probability as the prediction result.

[0091] Exemplarily, take the printing press registration system as an example: The correlation matrix shows that the correlation coefficient between paper tension and registration accuracy is 0.75, the correlation coefficient between printing speed and registration accuracy is 0.68, and the correlation coefficient between impression pressure and registration accuracy is 0.55. Time series analysis shows that the changes in paper tension and printing speed precede the change in registration accuracy, and a connection from these two parameters to registration accuracy is established.

[0092] The information entropy values of each parameter node are calculated as follows: paper tension: 0.85, printing speed: 0.78, impression pressure: 0.62, registration accuracy: 0.92; The information gain for predicting registration accuracy is calculated as follows: paper tension: 0.45, printing speed: 0.38, impression pressure: 0.25, and the dynamic threshold is set to 0.35; Paper tension and printing speed are selected as the core network nodes; the set of parent nodes of registration accuracy is {paper tension, printing speed}; a conditional probability table is established, such as: P(registration accuracy = high | tension = normal, speed = moderate) = 0.85; P(registration accuracy = medium | tension = high, speed = fast) = 0.65.

[0093] The initial log-likelihood expectation value is -856.3. After 87 iterations of optimization, the log-likelihood expectation value is improved to -782.5. The optimized conditional probabilities are as follows: P(registration accuracy = high | tension = normal, speed = moderate) = 0.92.

[0094] The calculation results of the information gain at the new sampling moment are as follows: paper tension: 0.48 (an increase of 6.7%), printing speed: 0.42 (an increase of 10.5%), impression pressure: 0.36 (an increase of 44%). Since the change in the information gain of impression pressure exceeds the threshold, it is added to the core network structure.

[0095] The probability distribution of predicting the registration accuracy at the next moment is as follows: high accuracy: 0.85, medium accuracy: 0.12, low accuracy: 0.03; "high accuracy" with the highest probability is selected as the prediction result.

[0096] In this embodiment, by calculating the correlation coefficient and analyzing the time-series data among parameters, a directional parameter network was successfully constructed. This network structure can reflect the relationships among different parameters and their change order, providing more accurate parameter-dependency analysis and a clear framework for subsequent prediction and optimization. By calculating the information entropy value and information gain, core parameter nodes that contribute significantly to the prediction target were successfully screened out. Dynamically adjusting the core network structure ensures that the network nodes always contain the most predictive parameters, which improves the flexibility and prediction accuracy of the network model. Using the expectation-maximization algorithm to optimize the conditional probability distribution, the conditional probability of the model was optimized by iteratively calculating the maximized expected value of the log-likelihood, making the prediction of each node more accurate. This optimization process improves the convergence speed and stability of the model and enhances the model's adaptability to data changes. By calculating the change in information gain in real time and dynamically updating the core network structure, the network model is ensured to adapt to parameter changes at any time. This mechanism enables the model to make the most accurate prediction at each sampling moment, providing reliable prediction results and effectively improving the accuracy of parameter adjustment and quality control in practical applications.

[0097] In an alternative embodiment, based on the correlation matrix and the printing quality prediction result, using the bidirectional constraint dynamic programming algorithm, a state transition equation is constructed through the coupling weights between functional modules, and the optimal parameter configuration of the printing press functional modules is determined by combining the bidirectional constraints, and the top-level design scheme of the printing press includes: Calculate the coupling weights between functional modules according to the correlation matrix, and the coupling weights characterize the degree of mutual influence between different functional modules; Construct a state transition equation according to the coupling weights, multiply the parameter state of the previous functional module by the corresponding coupling weight and sum through the state transition equation to obtain the parameter state of the current functional module; Establish bidirectional constraint conditions according to the printing quality prediction result, and the bidirectional constraint conditions include a forward constraint condition and a backward constraint condition. The forward constraint condition limits the adjustment range of the parameters corresponding to the current functional module according to the parameter state of the previous functional module; the backward constraint condition ensures that the subsequent functional modules reach the preset quality target according to the printing quality prediction result; According to the forward constraint condition, traverse each functional module in the forward direction, calculate the local optimal solution and record the state transition path to obtain a state transition sequence; according to the backward constraint condition, start from the last functional module and trace back the state transition sequence in the reverse direction to determine the optimal parameter configuration of the functional module; Input the optimal parameter configuration into the printing quality prediction model for verification. When the printing quality prediction result obtained by verification does not meet the requirements, adjust the bidirectional constraint conditions and re-execute the parameter configuration optimization.

[0098] In a specific embodiment, the correlation degree between functional modules is analyzed based on the correlation matrix. For each pair of functional modules, the correlation coefficients between the parameter pairs they contain are statistically calculated, and the average value is taken as the coupling weight between the modules. The larger the coupling weight, the stronger the mutual influence between the two functional modules.

[0099] The parameter state of each functional module is represented as a state vector. According to the coupling weight, a relationship between the current functional module state and the previous functional module state is established. The state vector of the current module is obtained by multiplying the state vector of the previous module by the corresponding coupling weight and summing them up.

[0100] Set forward constraint conditions: According to the actual parameter state of the previous functional module, determine the adjustable range of the parameters of the current module to avoid parameter adjustment exceeding the physical limits of the device.

[0101] Set backward constraint conditions: According to the printing quality prediction result, set a quality target threshold for the subsequent functional modules to ensure that the final printing quality meets the requirements.

[0102] Starting from the first functional module, traverse backward in sequence. For each module, enumerate the possible parameter states within the forward constraint range, calculate the minimum cost to reach that state, and record the optimal state transition path. Save the arrival path for each possible state.

[0103] Starting from the last functional module, according to the backward constraint conditions, select the final state that meets the quality requirements. Trace back forward along the recorded state transition path to determine the optimal parameter configuration for each functional module.

[0104] Input the obtained optimal parameter configuration into the printing quality prediction model to verify the expected effect. If the prediction result does not meet the requirements, adjust the parameter range of the two-way constraint conditions and re-execute the optimization process.

[0105] Exemplarily, taking the paper feeding system, impression system, ink supply system, and paper delivery system of a four-color printing press as an example: Calculate the coupling weights between functional modules: paper feeding - impression, 0.75; paper feeding - ink supply, 0.45; impression - ink supply, 0.65; impression - paper delivery, 0.70; ink supply - paper delivery, 0.55.

[0106] The state vector of the paper feeding system includes: paper tension, paper feeding speed; the state vector of the impression system includes: impression pressure, printing speed; example of the state transition equation: impression system state = 0.75 × paper feeding system state.

[0107] Forward constraint conditions: paper feeding speed range 20 - 60 m / min; paper tension range 150 - 300 N; impression pressure range 2000 - 4000 N; Backward constraint conditions: registration accuracy error ≤ 0.1 mm; color difference ΔE ≤ 3; printing density error ≤ 0.05; Possible states of the paper feeding system: State 1, speed 30 m / min, tension 200 N; State 2, speed 40 m / min, tension 250 N; State 3, speed 50 m / min, tension 280 N; Corresponding optimal states of the impression system: State 1 corresponds to a pressure of 2500 N and a speed of 28 m / min; State 2 corresponds to a pressure of 3000 N and a speed of 38 m / min; State 3 corresponds to a pressure of 3500 N and a speed of 48 m / min.

[0108] Starting from the paper feeding system and tracing backward, select the state that meets the registration accuracy requirements: the speed of the paper feeding system is 38 m / min, the ink fountain opening of the ink supply system is 45%, the pressure of the impression system is 3000 N, the speed is 38 m / min, the speed of the paper delivery system is 40 m / min, and the tension is 250 N.

[0109] Input the above parameter configuration into the prediction model: predict the registration accuracy error of 0.08 mm, predict the color difference ΔE: 2.8; predict the printing density error of 0.04.

[0110] Verify that the result meets the quality requirements and confirm it as the final parameter configuration scheme.

[0111] In this embodiment, by calculating the coupling weights between functional modules, the mutual influence intensity between different functional modules can be quantified. This analysis helps to identify the interdependencies of key modules, provides a reliable basis for subsequent parameter adjustment, and optimizes the overall performance of the system. Set forward and backward constraint conditions to ensure that the parameter adjustment of each functional module is balanced between physical limitations and quality requirements. The forward constraint controls the adjustable range of parameters and avoids adjustments beyond the device's capabilities, while the backward constraint ensures that the final printing quality meets the requirements, improving the controllability and stability of the system. By recording the optimal state transition paths of each module and performing backtracking, the optimal parameter configuration of each functional module can be determined. This method considers the minimum-cost path during the optimization process, ensuring that the parameter configuration of each module of the system is optimal and meets the quality objectives. By inputting the optimized optimal parameter configuration into the printing quality prediction model for verification, it is ensured that the prediction results meet the quality requirements. This process guarantees the accuracy of the final parameter configuration and improves the quality control ability of the overall printing process.

[0112] The present invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for performing various aspects of the present invention.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A top-level project design method for a printing press based on multi-source data, characterized in that, Including: Obtain the operating data of the printing press, where the operating data of the printing press includes printing quality parameters, equipment condition parameters, and environmental parameters; Use an adaptive fuzzy clustering algorithm to perform feature analysis and parameter extraction on the operating data of the printing press. By determining the number of clusters and dynamically adjusting the membership degree of sample classification, identify the key parameter combinations that affect printing quality; Construct a correlation matrix between the key parameter combinations; According to the parameter mapping relationship in the correlation matrix, construct a dynamic Bayesian probability network model. Build the core network structure by calculating the information entropy value and information gain, optimize the conditional probability distribution using the logarithmic likelihood expectation value, establish a probability association between the printing quality parameters and the equipment condition parameters, and generate a printing quality prediction result; Based on the correlation matrix and the printing quality prediction result, use a two-way constrained dynamic programming algorithm to construct a state transition equation through the coupling weights between functional modules, and determine the optimal parameter configuration of the printing press functional modules by combining two-way constraints, generating a top-level design scheme for the printing press.

2. The method according to claim 1, characterized in that Using an adaptive fuzzy clustering algorithm to perform feature analysis and parameter extraction on the operating data of the printing press. By determining the number of clusters and dynamically adjusting the membership degree of sample classification, the key parameter combinations that affect printing quality include: Segment the operating data of the printing press by the sliding window method, and calculate and determine the feature vectors based on each window; Calculate the minimum description length value of the feature vectors to determine the target number of clusters interval. Calculate the fuzzy partition coefficient value and the partition entropy value within the target number of clusters interval, select the number of clusters corresponding to the largest fuzzy partition coefficient value and the smallest partition entropy value, and determine the final number of clusters; Construct an initial value of the sample classification membership degree, calculate the distribution parameters of each dimension of the feature vectors, and obtain the sample classification membership degree value through dynamic adjustment according to the distribution parameters; Calculate the cluster center coordinates according to the sample classification membership degree value, calculate the spatial position relationship strength between samples as a penalty factor to construct an objective function; Iteratively update the sample classification membership degree value and the cluster center coordinates according to the final number of clusters, and minimize the objective function to determine the clustering result; Based on the clustering result, analyze the characteristic patterns of each cluster center, calculate the contribution degree of each feature in distinguishing different categories using the statistical discrimination ratio, and select the feature combination with the statistical discrimination ratio greater than the preset threshold as the key parameter combination for printing quality; Input the key parameter combination into a support vector machine classifier for cross-validation to verify the effectiveness of the key parameter combination.

3. The method according to claim 2, wherein Obtaining the sample classification membership degree value through dynamic adjustment according to the distribution parameters includes: Obtain the local density index by calculating the mean Euclidean distance between the sample points and the nearest neighbor sample points, calculate the distance deviation value by combining the standard deviation of the distance sequence from the sample points to the cluster center, determine the local reliability value, and generate an adaptive weight factor. According to whether the local reliability value is lower than the threshold, introduce an exponential decay coefficient to adjust the sample classification membership degree value, specifically including: For each sample point, multiple nearest neighbor sample points are selected, the Euclidean distance between the sample point and the nearest neighbor sample points is calculated, and the mean value of the Euclidean distances is used as the local density index of the sample point; the Euclidean distance from each sample point to each cluster center is calculated to obtain a distance sequence, and the standard deviation of the distance sequence is calculated as the distance deviation value of the sample point; the local density index is divided by the distance deviation value to obtain the local reliability value of the sample point; According to the local reliability value, an adaptive weight factor is generated in combination with a preset adjustment coefficient; The adaptive weight factor is multiplied by the initial value of the sample classification membership degree to obtain the adjusted value of the sample classification membership degree; When the local reliability value is lower than the preset reliability threshold, the sample classification membership degree value is obtained by multiplying the exponential decay coefficient by the adjusted value of the sample classification membership degree; when the local reliability value is not lower than the reliability threshold, the adjusted value of the sample classification membership degree is used as the sample classification membership degree value; where the exponential decay coefficient specifically refers to: ； Among them, D represents the exponential decay coefficient, represents the reliability decay rate, T r represents the reliability threshold, R l represents the local reliability value.

4. The method according to claim 2, wherein The contribution degree of each feature in distinguishing different categories is calculated by using the statistical discrimination ratio, and the feature combination with the statistical discrimination ratio greater than the preset threshold is selected as the key parameter combination of the printing quality, including: The between-class scatter matrix and the within-class scatter matrix are calculated to determine the statistical discrimination ratio sequence of the features; at the same time, the mutual information matrix between the features is calculated to judge the redundant features; through a progressive strategy, starting from the feature with the largest statistical discrimination ratio, the statistical discrimination ratio gain value is calculated through iteration, and the feature combination is gradually constructed, specifically including: The between-class scatter matrix is determined by calculating the difference between the means of different category samples, and the within-class scatter matrix is determined by calculating the difference between the samples within the same category and the category mean of the corresponding category; based on the between-class scatter matrix and the within-class scatter matrix, the statistical discrimination ratio of each feature is calculated to obtain the feature statistical discrimination ratio sequence; The mutual information values between the features are calculated to obtain the feature mutual information matrix. When the mutual information value is greater than the first preset threshold, the feature with the statistical discrimination ratio less than the second preset threshold in the corresponding feature pair is marked as a redundant feature; the feature with the largest statistical discrimination ratio and not marked as a redundant feature is selected from the feature statistical discrimination ratio sequence as the initial key feature; The features that are not marked as redundant features and not selected as the initial key features are sorted in descending order of the statistical discrimination ratio to obtain a candidate feature sequence; each feature in the candidate feature sequence is added to the existing key feature combination one by one, and the combined statistical discrimination ratio before addition and the combined statistical discrimination ratio after addition are calculated respectively to obtain the statistical discrimination ratio gain value; When the statistical discrimination ratio gain value is greater than the third preset threshold, the corresponding candidate feature is added to the key feature combination; Repeat the execution until all the features in the candidate feature sequence are processed to obtain the final key feature combination.

5. The method according to claim 4, wherein Calculating the between-class scatter matrix and the within-class scatter matrix to determine the feature statistical discrimination ratio sequence includes: Calculating the number of samples in each category of the multi-category sample data, and performing normalization processing to obtain the category weight vector, and constructing the category weight matrix; Calculate the class mean vectors of each class of samples in the multi-class sample data, as well as the overall mean vector; calculate the difference between the class mean vector and the overall mean vector, and construct a class mean difference matrix; multiply the class mean difference matrix by the class weight matrix to obtain an inter-class scatter matrix; Calculate the difference between each class of samples in the multi-class sample data and the corresponding class mean vector, and construct a sample difference matrix; multiply the sample difference matrix by the class weight matrix to obtain an intra-class scatter matrix; Extract the diagonal elements of the inter-class scatter matrix and the diagonal elements of the intra-class scatter matrix, and respectively determine the inter-class scatter vector and the intra-class scatter vector; Calculate the ratio of the inter-class scatter vector to the intra-class scatter vector to obtain a statistical discriminant ratio sequence of features.

6. The method according to claim 1, characterized in that, According to the parameter mapping relationship in the correlation matrix, construct a dynamic Bayesian probability network model, construct a core network structure by calculating the information entropy value and information gain, optimize the conditional probability distribution using the logarithmic likelihood expectation value, and establish a probability association between the printing quality parameters and the equipment operating condition parameters to generate a printing quality prediction result, including: Based on the dynamic Bayesian probability network, construct an initial network structure through the correlation matrix, use the information entropy value, calculate the information gain, screen parameter nodes to construct a core network structure, and realize parameter prediction through the optimization of the logarithmic likelihood expectation value and the real-time calculation of the information gain, specifically including: Establish connections between parameter pairs according to the correlation coefficients of parameter pairs in the correlation matrix; determine the direction of the connection based on the temporal characteristics of the parameter pairs, and construct an initial network structure; Calculate the information entropy value of each parameter node in the initial network structure, construct an evaluation index of parameter importance based on the information entropy value, and calculate the information gain of each parameter node on the prediction result; sort the parameter nodes according to the information gain, select the parameter nodes with information gain greater than the dynamic threshold to construct a core network structure, obtain the corresponding parent node set for the parameter nodes in the core network structure, and construct a conditional probability distribution based on the parent node set; Calculate the logarithmic likelihood expectation value based on the core network structure and the conditional probability distribution, and maximize the logarithmic likelihood expectation value through iterative optimization to obtain an optimized conditional probability distribution; Calculate the information gain of the parameter node at the current sampling moment, update the parameter composition of the core network structure according to the change amount of the information gain, and construct a new core network structure; based on the new core network structure and the corresponding conditional probability distribution, calculate the parameter prediction probability at the next sampling moment; Select the parameter value corresponding to the maximum parameter prediction probability as the printing quality prediction result.

7. The method according to claim 1, wherein Based on the correlation matrix and the printing quality prediction result, use the bidirectional constraint dynamic programming algorithm to construct a state transition equation through the coupling weight between functional modules, and determine the optimal parameter configuration of the printing press functional module in combination with the bidirectional constraint to generate a top-level design scheme of the printing press, including: Calculate the coupling weight between functional modules according to the correlation matrix, and the coupling weight characterizes the degree of mutual influence between different functional modules; Construct a state transition equation according to the coupling weights, and multiply and sum the previous functional module parameter states by the corresponding coupling weights through the state transition equation to obtain the current functional module parameter states; Establish bidirectional constraint conditions according to the printing quality prediction results. The bidirectional constraint conditions include forward constraint conditions and backward constraint conditions. The forward constraint conditions limit the adjustment range of the parameters corresponding to the current functional module according to the previous functional module parameter states; the backward constraint conditions ensure that subsequent functional modules reach a preset quality target according to the printing quality prediction results; According to the forward constraint conditions, traverse each functional module in the forward direction, calculate the local optimal solution and record the state transition path to obtain a state transition sequence; according to the backward constraint conditions, starting from the last functional module, trace back the state transition sequence in the reverse direction to determine the optimal parameter configuration of the functional module; Input the optimal parameter configuration into the printing quality prediction model for verification. When the printing quality prediction result obtained by the verification does not meet the requirements, adjust the bidirectional constraint conditions and re-execute the parameter configuration optimization.

Images