Ink coating distribution monitoring system and method based on big data
Through big data technology and spatial anomaly attention prediction network model, the problems of low monitoring efficiency and deviation of traditional ink coatings are solved, and efficient and accurate detection and control of ink coatings are achieved to ensure the stability and quality of the production process .
Patent Information
- Application Number
- CN202411251935.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-09
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-09
AI Technical Summary
Traditional ink coating distribution monitoring methods rely on single variable evaluation and manual detection, resulting in low monitoring efficiency and deviations in quality results, and it is easy to ignore subtle anomalies.
Adopt a big data-based ink coating distribution monitoring system, obtain data through sensor integration equipment, combine historical evaluation data for feature extraction and spatial anomaly analysis, build a spatial anomaly attention prediction network model, and automatically detect and predict abnormal situations of ink coatings , and intelligent monitoring is carried out through the control parameter abnormality monitoring engine.
Multi-source variable evaluation is achieved, the ink coating monitoring efficiency and quality evaluation accuracy is improved, problems in production are discovered and corrected in a timely manner, and the ink coating stability and product consistency are ensured.
Smart Images

Figure CN118779643B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and in particular to an ink coating distribution monitoring system and method based on big data. Background Art
[0002] The intelligent ink coating distribution monitoring method collects and analyzes a large amount of coating data to achieve accurate monitoring of the quality and distribution of the ink coating, helping manufacturers to monitor the coating quality in the production process in real time, timely discover and correct coating defects, improve production efficiency and product quality, thereby reducing production costs and improving customer satisfaction. Ink coating distribution monitoring based on big data uses sensors and data acquisition technology to obtain data such as coating thickness and uniformity in real time, and combines big data analysis and machine learning algorithms to process and analyze these data to achieve accurate monitoring of the ink coating in the production process, automatically feedback coating quality problems, optimize production processes, improve product quality and production efficiency, and thus achieve intelligent production management and quality control. However, traditional ink coating distribution monitoring methods often evaluate ink coating monitoring through a single variable and rely on manual ink coating quality assessment, making the monitored ink coating quality effect poor and inefficient; through manual inspection and ink coating detection based on simple rules, it is easy to ignore subtle abnormalities, resulting in a large deviation between the feedback ink coating quality results and the actual results. Summary of the invention
[0003] Based on this, the present invention provides an ink coating distribution monitoring system and method based on big data to solve at least one of the above technical problems.
[0004] To achieve the above purpose, a method for monitoring ink coating distribution based on big data includes the following steps:
[0005] Step S1: using a sensor integrated device to perform ink coating distribution monitoring processing on a printed sample to obtain ink coating distribution monitoring data;
[0006] Step S2: acquiring historical ink coating evaluation data; performing ink coating feature evaluation type analysis based on the historical ink coating evaluation data to obtain ink coating feature evaluation type data; performing feature extraction on the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data;
[0007] Step S3: performing effective data analysis of local ink coating distribution monitoring characteristics based on the ink coating distribution monitoring characteristic data to generate effective local ink coating distribution monitoring characteristic data; designing effective ink coating distribution monitoring attention parameters based on the effective local ink coating distribution monitoring characteristic data; and constructing a spatial abnormal attention prediction network model for the ink coating based on the effective ink coating distribution monitoring attention parameters;
[0008] Step S4: transmitting the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis to generate ink coating spatial anomaly data; performing global ink coating stability analysis based on the ink coating spatial anomaly data to generate global ink coating stability data; transmitting the global ink coating stability data to the terminal to perform ink coating distribution monitoring anomaly feedback operation;
[0009] Step S5: Obtain ink coating control parameters; establish a mapping relationship between ink coating spatial anomaly and ink coating control parameters based on ink coating spatial anomaly data and ink coating control parameters to obtain an ink coating control parameter anomaly prediction model; configure an ink coating control parameter anomaly monitoring engine according to the ink coating control parameter anomaly prediction model, and perform ink coating control parameter intelligent monitoring operations through the ink coating control parameter anomaly monitoring engine.
[0010] The present invention monitors and processes the ink coating through a sensor integrated device to obtain ink coating distribution monitoring data, provides basic monitoring data, provides a real-time information basis for subsequent analysis and decision-making, helps to timely discover problems in production, and lays the foundation for data processing and analysis in subsequent steps. Historical ink coating evaluation data is obtained to analyze the comprehensive characteristics of the quality and characteristics of the ink coating, including past performance and problems under different conditions. Based on the historical ink coating evaluation data, the ink coating feature evaluation type analysis is performed to identify the characteristics and performance of the ink coating under different types, such as uniformity, thickness, coverage, etc. Feature extraction is performed on the ink coating distribution monitoring data according to the ink coating feature evaluation type data, and the features extracted from the evaluation data are applied to the ink coating distribution monitoring data to identify and screen the ink coating evaluation type with better feature performance. Based on the ink coating distribution monitoring feature data, effective data analysis of the local ink coating distribution monitoring feature is performed to determine the distribution and quality of the ink coating in a specific area, so as to locate potential problems or anomalies. The effective local ink coating distribution monitoring feature data reduces the amount of data processed and improves the running speed of data processing, so as to more effectively identify and respond to ink coating problems. According to the effective local ink coating distribution monitoring feature data, effective ink coating distribution monitoring attention parameters are designed, and the focus and priority of attention are adjusted based on the analysis results, so that the system can more accurately monitor and predict the spatial anomalies of the ink coating. Based on the effective ink coating distribution monitoring attention parameters, a spatial anomaly attention prediction network model of the ink coating is constructed to predict and respond to spatial anomalies that may affect production quality in real time, thereby supporting timely adjustment and control. The ink coating distribution monitoring feature data is transmitted to the spatial anomaly attention prediction network model for ink coating spatial anomaly analysis. By analyzing the real-time monitored ink coating data through the prediction network model, the existing spatial anomalies, such as thickness changes and uniformity problems of the ink coating, can be detected. According to the output of the prediction model, detailed ink coating spatial anomaly data is generated, including the location, type and severity of the anomaly, to help operators quickly identify and locate the problem. According to the ink coating spatial anomaly data, a global ink coating stability analysis is performed to evaluate the stability and quality level of the overall ink coating, providing a global perspective and decision-making basis for quality control of the production process. The output global stability data reflects the change trend and stability performance of the ink coating throughout the production cycle, providing an important reference for production management and optimization. The global ink coating stability data is transmitted to the terminal to perform the ink coating distribution monitoring abnormality feedback operation. By timely feedback of stability data, production parameters are adjusted or abnormalities are handled to ensure the stability and production quality of the ink coating. The ink coating control parameters used in the current production process are obtained, including key parameters such as coating speed, pressure, and temperature.Based on the ink coating spatial anomaly data and ink coating control parameters, a mapping relationship between the ink coating spatial anomaly and the ink coating control parameters is established, the relationship between the ink coating anomaly data and the control parameters is analyzed, and an anomaly model is established to help predict the type and degree of ink coating anomalies caused by specific control parameters, so as to obtain an ink coating control parameter anomaly prediction model. Through the output results of the ink coating control parameter anomaly prediction model, the control parameter changes that may cause ink coating anomalies are identified in advance, so as to take preventive measures to avoid quality problems in the production process. Configure the ink coating control parameter anomaly monitoring engine, integrate the anomaly monitoring engine into the control system, monitor the changes in control parameters in real time and compare them with the anomaly model, and issue alarms or adjust the control strategy in time. The ink coating control parameter anomaly monitoring engine performs intelligent monitoring of ink coating control parameters, automatically manages and adjusts control parameters to maintain the stability and quality of the ink coating, and improves production efficiency and product consistency.
[0011] This specification provides an ink coating distribution monitoring system based on big data, which is used to execute the ink coating distribution monitoring method based on big data. The ink coating distribution monitoring system based on big data includes:
[0012] The ink coating distribution monitoring module is used to perform ink coating distribution monitoring processing on the printed sample using the sensor integrated device to obtain ink coating distribution monitoring data;
[0013] The ink coating feature analysis module is used to obtain historical ink coating evaluation data; perform ink coating feature evaluation type analysis based on the historical ink coating evaluation data to obtain ink coating feature evaluation type data; perform feature extraction on the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data;
[0014] The spatial anomaly attention prediction network model building module is used to perform effective data analysis of local ink coating distribution monitoring characteristics based on ink coating distribution monitoring feature data, and generate effective local ink coating distribution monitoring feature data; design effective ink coating distribution monitoring attention parameters based on effective local ink coating distribution monitoring feature data; and build a spatial anomaly attention prediction network model for ink coating based on the effective ink coating distribution monitoring attention parameters;
[0015] The ink coating distribution monitoring anomaly feedback module is used to transmit the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis and generate ink coating spatial anomaly data; perform global ink coating stability analysis based on the ink coating spatial anomaly data and generate global ink coating stability data; transmit the global ink coating stability data to the terminal to perform ink coating distribution monitoring anomaly feedback operations;
[0016] The ink coating control parameter intelligent monitoring module is used to obtain the ink coating control parameters; based on the ink coating spatial anomaly data and the ink coating control parameters, a mapping relationship between the ink coating spatial anomaly and the ink coating control parameters is established to obtain an ink coating control parameter anomaly prediction model; according to the ink coating control parameter anomaly prediction model, an ink coating control parameter anomaly monitoring engine is configured, and the ink coating control parameter intelligent monitoring operation is performed through the ink coating control parameter anomaly monitoring engine.
[0017] The beneficial effect of the present application is that the ink coating distribution monitoring method based on big data of the present invention performs multi-source variable evaluation on ink coating monitoring, and automatically detects and predicts abnormal conditions of the ink coating through a self-designed spatial abnormal attention prediction network, thereby improving the efficiency of ink coating monitoring and the quality evaluation results. By establishing an abnormal prediction model and monitoring engine for ink coating control parameters, and providing a data-driven ink coating quality management method, it is possible to identify subtle abnormal conditions of the ink coating caused by the ink coating control parameters, so that the feedback ink coating quality results are more consistent with the actual results. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 A schematic diagram of the steps of a method for monitoring ink coating distribution based on big data according to the present invention;
[0019] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;
[0020] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.
[0021] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0022] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.
[0023] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.
[0024] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.
[0025] To achieve this, please refer to Figures 1 to 3 The present invention provides an ink coating distribution monitoring method based on big data, comprising the following steps:
[0026] Step S1: using a sensor integrated device to perform ink coating distribution monitoring processing on a printed sample to obtain ink coating distribution monitoring data;
[0027] Step S2: acquiring historical ink coating evaluation data; performing ink coating feature evaluation type analysis based on the historical ink coating evaluation data to obtain ink coating feature evaluation type data; performing feature extraction on the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data;
[0028] Step S3: performing effective data analysis of local ink coating distribution monitoring characteristics based on the ink coating distribution monitoring characteristic data to generate effective local ink coating distribution monitoring characteristic data; designing effective ink coating distribution monitoring attention parameters based on the effective local ink coating distribution monitoring characteristic data; and constructing a spatial abnormal attention prediction network model for the ink coating based on the effective ink coating distribution monitoring attention parameters;
[0029] Step S4: transmitting the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis to generate ink coating spatial anomaly data; performing global ink coating stability analysis based on the ink coating spatial anomaly data to generate global ink coating stability data; transmitting the global ink coating stability data to the terminal to perform ink coating distribution monitoring anomaly feedback operation;
[0030] Step S5: Obtain ink coating control parameters; establish a mapping relationship between ink coating spatial anomaly and ink coating control parameters based on ink coating spatial anomaly data and ink coating control parameters to obtain an ink coating control parameter anomaly prediction model; configure an ink coating control parameter anomaly monitoring engine according to the ink coating control parameter anomaly prediction model, and perform ink coating control parameter intelligent monitoring operations through the ink coating control parameter anomaly monitoring engine.
[0031] The present invention monitors and processes the ink coating through a sensor integrated device to obtain ink coating distribution monitoring data, provides basic monitoring data, provides a real-time information basis for subsequent analysis and decision-making, helps to timely discover problems in production, and lays the foundation for data processing and analysis in subsequent steps. Historical ink coating evaluation data is obtained to analyze the comprehensive characteristics of the ink coating quality and characteristics, including past performance and problems under different conditions. Based on the historical ink coating evaluation data, the ink coating feature evaluation type analysis is performed to identify the characteristics and performance of the ink coating under different types, such as uniformity, thickness, coverage, etc. Feature extraction is performed on the ink coating distribution monitoring data according to the ink coating feature evaluation type data, and the features extracted from the evaluation data are applied to the ink coating distribution monitoring data to identify and screen the ink coating evaluation type with better feature performance. Based on the ink coating distribution monitoring feature data, effective data analysis of the local ink coating distribution monitoring feature is performed to determine the distribution and quality of the ink coating in a specific area, so as to locate potential problems or anomalies. The effective local ink coating distribution monitoring feature data reduces the amount of data processed and improves the running speed of data processing, so as to more effectively identify and respond to ink coating problems. According to the effective local ink coating distribution monitoring feature data, effective ink coating distribution monitoring attention parameters are designed, and the focus and priority of attention are adjusted based on the analysis results, so that the system can more accurately monitor and predict the spatial anomalies of the ink coating. Based on the effective ink coating distribution monitoring attention parameters, a spatial anomaly attention prediction network model of the ink coating is constructed to predict and respond to spatial anomalies that may affect production quality in real time, thereby supporting timely adjustment and control. The ink coating distribution monitoring feature data is transmitted to the spatial anomaly attention prediction network model for ink coating spatial anomaly analysis. By analyzing the real-time monitored ink coating data through the prediction network model, the existing spatial anomalies, such as thickness changes and uniformity problems of the ink coating, can be detected. According to the output of the prediction model, detailed ink coating spatial anomaly data is generated, including the location, type and severity of the anomaly, to help operators quickly identify and locate the problem. According to the ink coating spatial anomaly data, a global ink coating stability analysis is performed to evaluate the stability and quality level of the overall ink coating, providing a global perspective and decision-making basis for quality control of the production process. The output global stability data reflects the change trend and stability performance of the ink coating throughout the production cycle, providing an important reference for production management and optimization. The global ink coating stability data is transmitted to the terminal to perform the ink coating distribution monitoring abnormality feedback operation. By timely feedback of stability data, production parameters are adjusted or abnormalities are handled to ensure the stability and production quality of the ink coating. The ink coating control parameters used in the current production process are obtained, including key parameters such as coating speed, pressure, and temperature.Based on the ink coating spatial anomaly data and ink coating control parameters, a mapping relationship between the ink coating spatial anomaly and the ink coating control parameters is established, the relationship between the ink coating anomaly data and the control parameters is analyzed, and an anomaly model is established to help predict the type and degree of ink coating anomalies caused by specific control parameters, so as to obtain an ink coating control parameter anomaly prediction model. Through the output results of the ink coating control parameter anomaly prediction model, the control parameter changes that may cause ink coating anomalies are identified in advance, so as to take preventive measures to avoid quality problems in the production process. Configure the ink coating control parameter anomaly monitoring engine, integrate the anomaly monitoring engine into the control system, monitor the changes in control parameters in real time and compare them with the anomaly model, and issue alarms or adjust the control strategy in time. The ink coating control parameter anomaly monitoring engine performs intelligent monitoring of ink coating control parameters, automatically manages and adjusts control parameters to maintain the stability and quality of the ink coating, and improves production efficiency and product consistency.
[0032] As an embodiment of the present invention, refer to Figure 1 The above is a schematic diagram of the steps of a method for monitoring ink coating distribution based on big data of the present invention. In this embodiment, the method for monitoring ink coating distribution based on big data includes the following steps:
[0033] Step S1: using a sensor integrated device to perform ink coating distribution monitoring processing on a printed sample to obtain ink coating distribution monitoring data;
[0034] In an embodiment of the present invention, a sensor integrated device is used to monitor the ink coating distribution of a printed sample. The sensor integrated device includes an optical sensor, a laser sensor or a pressure sensor, etc. These sensors can measure key parameters such as the thickness, reflection spectrum and coverage of the ink coating. The arrangement position of the sensor is selected according to the characteristics and size of the printed sample to ensure that the distribution of the ink coating can be fully and effectively monitored. The selected sensor is installed on the printing device to collect monitoring data of the ink coating in real time. The sensor collects data through continuous measurement or periodic scanning. The collected data may include information such as thickness changes, surface flatness, and color reflection spectrum of the ink coating, and locates the area collected by the sensor integrated device, and the ink coating data collected by the sensor integrated device and the ink coating area data collected by the sensor integrated device are matched, and the collected ink coating data reflects its area position to obtain ink coating distribution monitoring data.
[0035] Step S2: acquiring historical ink coating evaluation data; performing ink coating feature evaluation type analysis based on the historical ink coating evaluation data to obtain ink coating feature evaluation type data; performing feature extraction on the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data;
[0036] In an embodiment of the present invention, ink coating evaluation data is obtained from a historical record system in the printing production process. The measurement data including the thickness, color, uniformity, etc. of the printed sample exists in the form of time series or batch records. It is necessary to ensure the integrity and accuracy of the data. The required historical data is extracted from the production database using a database query language (such as SQL), and the time range of the data is ensured to cover a sufficiently long period for a comprehensive evaluation. Data mining and statistical analysis techniques, such as principal component analysis (PCA) and factor analysis, are used to analyze and model the historical ink coating evaluation data, identify different types of ink coating feature evaluations, such as thickness uniformity, color consistency, coating coverage, etc., and use data analysis software (such as pandas, numpy, scikit-learn and other libraries in Python) to process and analyze the data to identify and define the importance of different types of ink coating features to the ink coating quality evaluation. According to the type of ink coating feature evaluation, the corresponding features are extracted from the ink coating distribution monitoring data obtained in actual production, that is, the ink coating evaluation type with low correlation is eliminated, and the ink coating distribution monitoring feature data is generated, that is, the ink coating distribution monitoring data corresponding to the ink coating quality is included.
[0037] Step S3: performing effective data analysis of local ink coating distribution monitoring characteristics based on the ink coating distribution monitoring characteristic data to generate effective local ink coating distribution monitoring characteristic data; designing effective ink coating distribution monitoring attention parameters based on the effective local ink coating distribution monitoring characteristic data; and constructing a spatial abnormal attention prediction network model for the ink coating based on the effective ink coating distribution monitoring attention parameters;
[0038] In the embodiment of the present invention, the characteristic data of ink coating distribution monitoring is preprocessed, including removing outliers, filling missing data, standardizing or normalizing, ensuring data quality and consistency, and analyzing and processing the characteristic data of ink coating distribution monitoring according to the preset local monitoring area range, such as selecting the printed sample as the central node, the area range of the radius 30, extracting key local features, such as local thickness change, local variance of color distribution, etc. Use the data processing library (such as pandas), numerical calculation library (such as numpy), machine learning library (such as scikit-learn) in Python, etc., combined with appropriate algorithms and models for data analysis and effective data extraction, for example, divide the local area into cells, and remove abnormal cells through cluster analysis to screen and obtain effective data related to the ink coating quality, so as to obtain effective local ink coating distribution monitoring characteristic data. Based on the effective local ink coating distribution monitoring feature data, key attention parameters suitable for monitoring and feedback of ink coating anomalies are designed, such as monitoring window size, attention weight distribution, etc. The designed parameters are adjusted and verified to ensure that they can accurately capture and reflect the abnormal characteristics of the ink coating under different conditions. Data simulation, experimental design or model training methods are used to optimize and verify the parameters in combination with actual data. Based on the effective ink coating distribution monitoring attention parameters, a neural network model suitable for predicting the spatial anomaly of the ink coating is constructed to obtain a spatial anomaly attention prediction network model for the ink coating, and the effective local ink coating distribution monitoring feature data is transmitted to the spatial anomaly attention prediction network model, so that the model learns to identify effective ink coating distribution monitoring data, thereby being able to identify and locate abnormal ink coating distribution monitoring data.
[0039] Step S4: transmitting the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis to generate ink coating spatial anomaly data; performing global ink coating stability analysis based on the ink coating spatial anomaly data to generate global ink coating stability data; transmitting the global ink coating stability data to the terminal to perform ink coating distribution monitoring anomaly feedback operation;
[0040] In the embodiment of the present invention, the ink coating distribution monitoring characteristic data is transmitted to the spatial anomaly attention prediction network model for ink coating spatial anomaly analysis, and a data transmission interface is designed to ensure that data is obtained from a data source (which may be a real-time monitoring device or a database) and transmitted to the input port of the spatial anomaly attention prediction network model. After the model receives the data, a prediction analysis is performed to generate ink coating spatial anomaly data, which involves the forward propagation calculation of the model to identify and locate the abnormal conditions in the ink coating, including the degree of ink coating abnormality, abnormality type, abnormal location data, etc. The abnormal data obtained from the spatial anomaly attention prediction network model is integrated into the ink coating spatial anomaly data from a global perspective, and the ink coating spatial anomaly data is analyzed using statistical analysis or machine learning methods to evaluate the overall stability of the coating and generate global ink coating stability data, which reflects the stability of the coating in time and space. The generated global ink coating stability data is configured into a format and interface suitable for terminal execution, and an abnormal feedback operation based on the global ink coating stability data is designed and implemented, including an automated control system or manual intervention measures to correct coating abnormalities.
[0041] Step S5: Obtain ink coating control parameters; establish a mapping relationship between ink coating spatial anomaly and ink coating control parameters based on ink coating spatial anomaly data and ink coating control parameters to obtain an ink coating control parameter anomaly prediction model; configure an ink coating control parameter anomaly monitoring engine according to the ink coating control parameter anomaly prediction model, and perform ink coating control parameter intelligent monitoring operations through the ink coating control parameter anomaly monitoring engine.
[0042] In the embodiment of the present invention, the current ink coating control parameters are obtained from the coating production equipment or database, including coating speed, thickness setting, coating pressure, rough surface temperature, etc. Data analysis and data association are performed based on the ink coating spatial anomaly data and the ink coating control parameters, and a mapping model between the ink coating spatial anomaly and the ink coating control parameters is established using machine learning or statistical modeling technology to predict the impact of parameter changes on coating anomalies, and the model is trained and verified based on the data analysis and data association results to ensure that the model predicts the abnormal ink coating control parameters that cause poor ink coating quality on the new data, and obtains an ink coating control parameter anomaly prediction model. The established ink coating control parameter anomaly prediction model is configured as an ink coating control parameter anomaly monitoring engine to ensure that the changes in coating control parameters and their impact on coating anomalies can be monitored in real time, and the anomaly threshold and early warning mechanism of the ink coating control parameter anomaly monitoring engine are designed. When an anomaly is detected, the corresponding feedback and adjustment measures are triggered. The configured ink coating control parameter anomaly monitoring engine is connected to the printing production line to monitor the control parameter status in real time. When the monitoring engine detects an anomaly, it automatically triggers pre-set corrective measures, such as issuing an early warning to the operator, automatically adjusting relevant parameters, etc., and continuously monitors and optimizes the anomaly prediction model, continuously improves the monitoring strategy according to actual production conditions, and improves the intelligence level of ink coating control.
[0043] Preferably, step S1 comprises the following steps:
[0044] Step S11: using a sensor integrated device to perform ink coating monitoring on a printed sample;
[0045] Step S12: collecting ink coating monitoring nodes for the ink coating monitoring operation, generating ink coating monitoring node data, and establishing an ink coating distribution network according to the ink coating monitoring node data;
[0046] Step S13: collecting ink coating monitoring data for the ink coating monitoring operation to obtain ink coating monitoring data;
[0047] Step S14: transmitting the ink coating monitoring data to the ink coating distribution network to identify the location characteristics of the ink coating monitoring data, and generating ink coating distribution monitoring data.
[0048] The present invention utilizes sensor integrated equipment to perform ink coating monitoring operations on printed samples, monitors the formation process of the ink coating in real time through sensors, and timely captures key indicators such as the thickness, uniformity, and coverage of the ink coating to ensure printing quality and production consistency. The ink coating monitoring node is collected for the ink coating monitoring operation, and an ink coating distribution network is established based on the generated ink coating monitoring node data to form a dynamic, real-time updated ink coating distribution network, which helps to accurately understand and describe the spatial distribution characteristics of the ink coating, and provides a basis for subsequent analysis and control. The ink coating monitoring data is collected for the ink coating monitoring operation, and the collected monitoring data reflects the actual formation and quality status of the ink coating, providing real basic data for further analysis and processing. The ink coating monitoring data is transmitted to the ink coating distribution network for position feature identification of the ink coating monitoring data, and the position features of the monitoring data are identified in the distribution network, so that each monitoring data point can be accurately located and analyzed, which helps to quickly identify and respond to potential ink coating quality problems, and improve production efficiency and product quality stability.
[0049] In the embodiment of the present invention, a suitable sensor integrated device is selected, such as a laser displacement sensor, a high-speed camera, a spectral sensor, etc., and a suitable sensor type is selected according to the properties of the ink coating and the monitoring requirements. The sensor integrated device is installed on the printing device, and the position and angle of the sensor are reasonably arranged to ensure that the ink coating state can be fully and accurately monitored. The sensor integrated device is programmed to control the sampling frequency, scanning range and other parameters, and the ink coating of the printed sample is monitored in real time and dynamically. The collected ink coating monitoring data is recorded and stored through the sensor data acquisition system to ensure the integrity and traceability of the data. According to the layout and scanning range of the sensor integrated device, the printed sample is divided into multiple monitoring nodes, such as a grid distribution, and the spatial coordinate information of each monitoring node is recorded and associated with the corresponding ink coating monitoring data to generate ink coating monitoring node data. All ink coating monitoring node data are integrated to describe the spatial distribution characteristics of the ink coating, and the ink coating distribution network is constructed using data visualization technology. Acquire the ink coating monitoring data collected in real time by the sensor integration device, including parameters such as coating thickness, coating spectral reflectance, uniformity, etc., ensure the timestamp information of the monitoring data, facilitate the subsequent association with the monitoring node data, store the collected ink coating monitoring data in the database or file system, and establish a corresponding data management mechanism. Import the ink coating monitoring data into the ink coating distribution network model through the data interface, associate the monitoring data with the corresponding network node according to the spatial coordinate information of the monitoring node, identify the location characteristics of the monitoring data in the ink coating distribution, integrate the ink coating monitoring data with the identified location characteristics, generate ink coating distribution monitoring data, provide basic data for the analysis and application of subsequent steps, and use visualization technology to intuitively display the spatial distribution of the ink coating.
[0050] Preferably, step S2 comprises the following steps:
[0051] Step S21: Acquire historical ink coating evaluation data;
[0052] Step S22: extracting historical ink coating evaluation type data and historical ink coating evaluation target data of the historical ink coating evaluation data;
[0053] Step S23: performing type subset and target association processing of ink coating evaluation according to historical ink coating evaluation type data and historical ink coating evaluation target data, and generating evaluation type subset-target association data;
[0054] Step S24: performing assessment type-target correlation analysis based on the assessment type subset-target association data to generate assessment type-target correlation data;
[0055] Step S25: performing ink coating feature evaluation type analysis according to the evaluation type-target core correlation data to obtain ink coating feature evaluation type data;
[0056] Step S26: extracting features from the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data.
[0057] The present invention establishes a comprehensive understanding of the quality of ink coatings in different time periods by acquiring historical ink coating evaluation data, including various evaluation indicators and quality data of ink coatings in past printing operations, and provides a historical reference basis for subsequent analysis and improvement. Extracting historical ink coating evaluation type data and historical ink coating evaluation target data of historical ink coating evaluation data is helpful to analyze the performance and association of different evaluation types under different targets, and is used to optimize the quality control strategy of ink coatings. According to the historical ink coating evaluation type data and the historical ink coating evaluation target data, the type subset of ink coating evaluation is associated with the target, and by associating the evaluation type and the evaluation target, a detailed relationship model of ink coating evaluation is established, which helps to determine how to adjust and optimize the characteristics of ink coating according to specific targets under different evaluation types, thereby improving printing quality and production efficiency. Based on the evaluation type subset-target association data, an evaluation type-target correlation analysis is performed. By analyzing the correlation between the evaluation type and the target, the degree of their influence on the quality of the ink coating is deeply understood, which helps to optimize the evaluation strategy and formulate more accurate control measures to achieve a higher level of printing quality and stability. According to the evaluation type-target core correlation data, the ink coating feature evaluation type analysis is performed to determine the key ink coating feature evaluation type. These evaluation type data reflect the key features of the ink coating under different conditions, providing a basis for subsequent feature extraction and distribution monitoring. Feature extraction is performed on the ink coating distribution monitoring data based on the ink coating feature evaluation type data to extract key feature information from the real-time monitored ink coating data. These feature data can accurately describe the spatial distribution characteristics of the ink coating, help accurately identify and handle abnormal conditions of the ink coating, and improve production efficiency and product quality.
[0058] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation step flow chart of step S1 in the embodiment, step S1 in this embodiment includes:
[0059] Step S21: Acquire historical ink coating evaluation data;
[0060] In an embodiment of the present invention, ink coating evaluation data over a period of time in the past (e.g., 1 year or 5 years) is extracted from a production management system or a quality control database, including measurement data of indicators such as thickness, color, glossiness, and surface roughness of printed samples, as well as evaluated ink coating quality data, and corresponding timestamps and batch information.
[0061] Step S22: extracting historical ink coating evaluation type data and historical ink coating evaluation target data of the historical ink coating evaluation data;
[0062] In an embodiment of the present invention, various indicators recorded in the historical ink coating evaluation data are analyzed and classified into different evaluation types, including: thickness uniformity, color consistency (coating spectral reflectance information), surface smoothness, coverage, etc., to form historical ink coating evaluation type data. At the same time, the final quality evaluation results or production target indicators of the printed samples are extracted from the historical data as historical ink coating evaluation target data, ensuring that the evaluation type data and the evaluation target data can correspond to each other in the time / batch dimension, so as to prepare for subsequent correlation analysis.
[0063] Step S23: performing type subset and target association processing of ink coating evaluation according to historical ink coating evaluation type data and historical ink coating evaluation target data, and generating evaluation type subset-target association data;
[0064] In an embodiment of the present invention, data mining and machine learning methods are used to analyze the relationship between historical ink coating evaluation type data and evaluation target data. For example, correlation analysis, regression model and other methods are used to find out the key evaluation type subsets that have a greater impact on the final quality evaluation results. Through multivariate linear regression analysis, it is identified that factors such as thickness uniformity and color consistency have a higher contribution to the final quality target. These key evaluation type subsets are associated with the corresponding evaluation target data to form evaluation type subset-target association data, which reflects the degree of influence of different evaluation types on the final quality target and provides an important basis for subsequent feature extraction and model building.
[0065] Step S24: performing assessment type-target correlation analysis based on the assessment type subset-target association data to generate assessment type-target correlation data;
[0066] In an embodiment of the present invention, correlation analysis methods such as Pearson correlation coefficient, Spearman, etc. are used to deeply analyze the correlation strength between the evaluation type subset and the evaluation target, calculate the correlation coefficient between each evaluation type and the final quality target, and organize the results into evaluation type-target correlation data. Correlation analysis helps to identify the most critical evaluation type indicators, that is, those indicators that have the greatest impact on the final quality target. Further, principal component analysis or random forest methods are used to sort and screen the correlations to find the most important evaluation types. The generated evaluation type-target correlation data provides a basis for subsequent feature evaluation type analysis.
[0067] Step S25: performing ink coating feature evaluation type analysis according to the evaluation type-target core correlation data to obtain ink coating feature evaluation type data;
[0068] In an embodiment of the present invention, the importance of each evaluation type is further analyzed and screened based on the evaluation type-target correlation data, and expert knowledge or machine learning methods are used to comprehensively consider factors such as correlation strength, representativeness of evaluation indicators, and fit with actual production to determine the most critical ink coating feature evaluation type. For example, indicators such as thickness uniformity, color consistency, and surface gloss are identified as key feature evaluation types, and the determined ink coating feature evaluation types are organized into a data set to form ink coating feature evaluation type data, in preparation for subsequent feature extraction, and these feature evaluation types are encoded or standardized for subsequent calculations and applications.
[0069] Step S26: extracting features from the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data.
[0070] In the embodiment of the present invention, the ink coating distribution monitoring data is associated with the ink coating characteristic evaluation type data, and for each key characteristic evaluation type, the corresponding characteristic index is extracted from the ink coating distribution monitoring data, for example, for thickness uniformity, the variance or fluctuation degree of the local thickness is calculated; for color consistency, the ink color difference or chromaticity distribution is measured; for ink thickness, the ink thickness layer uniformity is analyzed, etc. The various extracted characteristic indexes are integrated into the ink coating distribution monitoring characteristic data as the input of the subsequent steps, and the characteristic data is pre-processed such as standardization, normalization or principal component analysis as needed, and quantized to the same data level to improve the effect of subsequent analysis, and finally generate the ink coating distribution monitoring characteristic data.
[0071] Preferably, step S24 includes the following steps:
[0072] Step S241: performing relevant characteristic factor analysis of the assessment type and the assessment target according to the assessment type subset-target association data to generate assessment type-target relevant characteristic factor data;
[0073] Step S242: constructing a core-related evaluation model of the evaluation type and the evaluation target based on the preset core-typical correlation analysis algorithm and the evaluation type-target related characteristic factor data to obtain an evaluation type-target core-related evaluation model;
[0074] Step S243: performing an assessment type-target correlation analysis on the assessment type subset-target associated data based on the assessment type-target core correlation assessment model to generate assessment type-target correlation data.
[0075] The present invention performs a related characteristic factor analysis of the evaluation type and the evaluation target based on the evaluation type subset-target association data, and determines the key characteristic factors that affect the evaluation results. These factors, such as specific quality indicators, process parameters or environmental conditions, help to deeply understand the key influencing factors in the evaluation process. Based on the preset nuclear typical association analysis algorithm and the evaluation type-target related characteristic factor data, a nuclear correlation evaluation model of the evaluation type and the evaluation target is constructed. The correlation algorithm of the kernel function more accurately reveals the correlation characteristics between the evaluation type and the target, providing strong support for subsequent analysis and prediction. Based on the evaluation type-target nuclear correlation evaluation model, the evaluation type-target correlation analysis is performed on the evaluation type subset-target association data, which helps to optimize the evaluation strategy, select highly relevant evaluation types, reduce the data type dimension, and improve the accuracy of data analysis, so that the evaluation process is more efficient and effective, while improving the management and control level of ink coating quality.
[0076] In an embodiment of the present invention, the relevant characteristic factor analysis of the evaluation type and the evaluation target is performed based on the evaluation type subset-target association data, the key relevant characteristic factors between each evaluation type and the evaluation target are identified, and the key characteristic factors that have a greater impact on the final quality target in each evaluation type indicator are found using correlation analysis and principal component analysis methods. For example, for the evaluation type of thickness uniformity, local thickness variance, maximum thickness difference, etc. are identified as key characteristic factors. For color consistency, average color difference value, color difference distribution standard deviation, etc. are identified as key characteristic factors. These key relevant characteristic factors are organized into a data set to form evaluation type-target related characteristic factor data, which provides a basis for the subsequent nuclear related evaluation model construction. Based on the preset nuclear typical correlation analysis algorithm and the evaluation type-target related characteristic factor data, the nuclear related evaluation model of the evaluation type and the evaluation target is constructed, and a suitable nuclear typical correlation analysis (KCCA) algorithm is selected as the preset analysis method, which effectively captures the complex nonlinear relationship between the evaluation type indicator and the quality target. The assessment type-target related characteristic factor data is used as the input of the KCCA algorithm to model the correlation between the assessment type and the assessment target. The KCCA algorithm will output a set of weight coefficients to characterize the contribution of each characteristic factor to the final prediction result. The analysis results of the KCCA algorithm are organized into an assessment type-target core-related assessment model to provide a calculation basis for subsequent correlation analysis. The constructed assessment type-target core-related assessment model is applied to the assessment type subset-target associated data. The core-related assessment model is used to calculate the core correlation coefficient between each assessment type indicator and the final quality target. The core correlation coefficient more accurately reflects the nonlinear correlation between the assessment type and the target. The calculated assessment type-target core correlation coefficient is organized into a data set to form assessment type-target correlation data, which is further used to identify the assessment type indicators that have the most critical impact on the final quality target.
[0077] Preferably, step S241 includes the following steps:
[0078] Based on the assessment type subset-target association data, an association volatility analysis is performed to generate assessment type subset-target association volatility data; based on the assessment type subset-target association volatility data, a correlation characteristic factor analysis of the assessment type and the assessment target is performed to generate assessment type-target correlation characteristic factor data.
[0079] The present invention performs correlation volatility analysis based on the evaluation type subset-target correlation data, identifies the change trend and periodic fluctuation in the evaluation process, understands the uncertainty and volatility of the evaluation results, and provides a basis for subsequent in-depth analysis and prediction. Based on the evaluation type subset-target correlation volatility data, the relevant characteristic factor analysis of the evaluation type and the evaluation target is performed to identify the key characteristic factors between the evaluation type and the target. These factors involve the fluctuation of quality indicators, changes in process parameters or other factors that affect the evaluation results, helping to deepen the understanding of the influencing factors in the evaluation process.
[0080] In an embodiment of the present invention, an associated volatility analysis is performed based on the assessment type subset-target associated data, the volatility relationship between each assessment type indicator and the final quality target is studied, and indicators such as the covariance between the assessment type indicator and the target indicator, the fluctuation range or standard deviation of the correlation coefficient, etc. are calculated to reflect the change of the correlation between them over time or batch. For example, analyze whether the correlation between thickness uniformity and the final qualified rate is stable in different batches of samples, or whether there is obvious fluctuation over time, and organize these associated volatility indicators into a data set to form the assessment type subset-target associated volatility data, which provides a basis for subsequent characteristic factor analysis. Based on the assessment type subset-target associated volatility data, the related characteristic factor analysis of the assessment type and the assessment target is performed, and the key related characteristic factors between each assessment type indicator and the final quality target are analyzed. The key characteristic factors that contribute more to the volatility of the target indicator are identified using correlation analysis, principal component analysis and other methods. For example, if the local fluctuation range of a certain evaluation type indicator is found to be the key factor causing the fluctuation of the final pass rate, or if the correlation change between certain evaluation indicators has an important impact on the target indicator, these key related characteristic factors can be organized into data sets to form evaluation type-target related characteristic factor data. These characteristic factor data not only reflect the correlation between the evaluation indicators and the targets, but also take into account the stability of the relationship between them, thus better characterizing the complexity of the actual production process.
[0081] Preferably, step S3 comprises the following steps:
[0082] Step S31: extracting the local selected interval data of the central node of the ink coating distribution monitoring characteristic data according to the preset local ink coating selection interval data, so as to obtain the local ink coating distribution monitoring characteristic data;
[0083] Step S32: performing local monitoring characteristic fluctuation deviation analysis according to the local ink coating distribution monitoring characteristic data to generate local monitoring characteristic fluctuation deviation data;
[0084] Step S33: Screening effective data of the local ink coating distribution monitoring characteristic data according to the local monitoring characteristic fluctuation deviation data to generate effective local ink coating distribution monitoring characteristic data;
[0085] Step S34: designing effective ink coating distribution monitoring attention parameters according to effective local ink coating distribution monitoring characteristic data;
[0086] Step S35: Construct a spatial abnormal attention prediction network model of the ink coating based on the effective ink coating distribution monitoring attention parameters.
[0087] The present invention extracts the local selected interval data of the central node of the ink coating distribution monitoring feature data according to the preset local ink coating selection interval data, accurately captures the distribution of the ink coating in the key area, helps to determine the key monitoring points of the ink coating, and then optimizes the monitoring strategy and resource allocation. By analyzing the fluctuation deviation of the local ink coating distribution monitoring feature data, the fluctuation and abnormal changes of the ink coating in the local area are identified, which helps to timely discover potential problems and take corresponding measures to adjust and optimize. According to the local monitoring characteristic fluctuation deviation data, the local ink coating distribution monitoring feature data is screened for effective data, the abnormal information in the data is filtered out, and the effective data with practical significance and value is retained, which is the basis for subsequent analysis and modeling. According to the effective local ink coating distribution monitoring feature data, the effective ink coating distribution monitoring attention parameters are designed, which can accurately define the monitoring system's attention level and monitoring strategy for the key features of the ink coating, and the optimization of these parameters can improve the sensitivity and accuracy of the monitoring system. Based on the effective ink coating distribution monitoring attention parameters, a spatial abnormal attention prediction network model is constructed, which can realize the prediction and identification of the ink coating spatial abnormality, help predict potential abnormal situations, and take measures in advance to adjust and intervene to ensure the quality and stability of the ink coating.
[0088] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 FIG. 1 is a flow chart of detailed implementation steps of step S3 in FIG. 1 . In this embodiment, step S3 includes:
[0089] Step S31: extracting the local selected interval data of the central node of the ink coating distribution monitoring characteristic data according to the preset local ink coating selection interval data, so as to obtain the local ink coating distribution monitoring characteristic data;
[0090] In the embodiment of the present invention, according to the actual production process and the characteristics of the printed sample, a suitable local selection interval is predetermined, such as a range with a radius of 30 cm at the center of the printed area. According to the preset local ink coating selection interval data, the ink coating distribution monitoring characteristic data in these preset local areas are extracted, and a sliding window method is used to ensure that each sample can obtain the corresponding local characteristic data, so as to obtain the local ink coating distribution monitoring characteristic data, and prepare for the subsequent fluctuation deviation analysis.
[0091] Step S32: performing local monitoring characteristic fluctuation deviation analysis according to the local ink coating distribution monitoring characteristic data to generate local monitoring characteristic fluctuation deviation data;
[0092] In an embodiment of the present invention, statistical characteristics of each characteristic indicator of the local ink coating distribution monitoring characteristic data in the local area are calculated, such as the mean, standard deviation, skewness coefficient, etc., and based on these statistical characteristics, the fluctuation range and deviation degree of the local characteristic indicators are analyzed, and a reasonable fluctuation deviation threshold is set for each characteristic indicator to identify abnormal areas where the fluctuation exceeds the normal range, and these fluctuation deviation indicators are organized into local monitoring characteristic fluctuation deviation data to provide support for subsequent effective data screening.
[0093] Step S33: Screening effective data of the local ink coating distribution monitoring characteristic data according to the local monitoring characteristic fluctuation deviation data to generate effective local ink coating distribution monitoring characteristic data;
[0094] In an embodiment of the present invention, effective data screening is performed on local ink coating distribution monitoring feature data based on local monitoring feature fluctuation deviation data. Feature indicator data whose fluctuation deviation exceeds the normal range are temporarily excluded and do not participate in subsequent analysis. Local feature indicator data with relatively stable fluctuations and higher credibility are retained to form effective local ink coating distribution monitoring feature data, thereby improving the reliability of subsequent analysis and modeling and avoiding interference from local abnormal data.
[0095] Step S34: designing effective ink coating distribution monitoring attention parameters according to effective local ink coating distribution monitoring characteristic data;
[0096] In an embodiment of the present invention, corresponding attention parameters are designed through effective local ink coating distribution monitoring feature data. The attention parameters reflect the importance of each local feature indicator to the final quality target, provide a weight basis for the subsequent prediction model, and adopt a heuristic method based on expert knowledge, or use machine learning technology such as attention mechanism to automatically learn appropriate attention parameters from the data. The designed attention parameters should be able to effectively highlight those key local feature indicators that have a greater impact on the quality target, that is, the designed effective ink coating distribution monitoring attention parameters reflect the ability to automatically analyze whether the quality of the ink coating meets the standards, by comparing the monitored ink coating with the effective ink coating. When the standards are met, it reflects that the monitored ink coating is also effective. When the standards are not met, the deviation between the monitored ink coating and the effective ink coating is calculated.
[0097] Step S35: Construct a spatial abnormal attention prediction network model of the ink coating based on the effective ink coating distribution monitoring attention parameters.
[0098] In the embodiment of the present invention, a U-Net or Encoder-Decoder structure based on a convolutional neural network (CNN) is used as the backbone network, and an attention mechanism such as Self-Attention or Channel-Wise Attention is added to the network to improve the model's perception of key features. The input of the network is effective local ink coating distribution monitoring feature data, and the output includes an abnormal probability map, an abnormal category map, and abnormal area coordinates. The effective ink coating distribution monitoring attention parameters are integrated into the network. The attention parameters can be used as feature channel weight factors to enhance those features that have a greater impact on the prediction results. The attention parameters can also be used as weight terms in the loss function to improve the learning effect of the model on key features. The three tasks of abnormal probability prediction, abnormal classification, and abnormal location are integrated into a unified network for end-to-end training, and a joint loss function such as weighted cross entropy loss, IoU Loss, etc. is designed. The learning objectives of the three tasks are optimized at the same time. The multi-task learning method is used to enable the network to fully extract features and achieve good performance in all three tasks. Use effective local ink coating distribution monitoring feature data as the training set, perform end-to-end training of the network, adopt appropriate optimization algorithms, such as Adam or RMSProp, and set reasonable hyperparameters, such as learning rate, batch size, etc., use the validation set to evaluate the model in real time, and adjust the network structure and hyperparameters according to the evaluation results until the performance meets the requirements. Finally, a spatial anomaly attention prediction network model for ink coating is constructed.
[0099] Preferably, step S35 includes the following steps:
[0100] Step S351: designing a spatial anomaly probability segmentation mapping relationship of the ink coating according to the effective ink coating distribution monitoring attention parameter to obtain a spatial anomaly probability segmentation head;
[0101] Step S352: performing weighted spatial abnormal attention channel analysis processing based on the evaluation type-target correlation data and the ink coating feature evaluation type data to generate weighted spatial abnormal attention channel data;
[0102] Step S353: designing a spatial anomaly classification mapping relationship of the ink coating according to the spatial anomaly probability segmentation head and the weighted spatial anomaly attention channel data to obtain a spatial anomaly classification head;
[0103] Step S354: designing a spatial anomaly positioning mapping relationship of the ink coating according to the spatial anomaly monitoring probability segmentation head and the spatial anomaly monitoring classification head to obtain a spatial anomaly positioning head;
[0104] Step S355: construct a spatial anomaly attention prediction network model for the ink coating based on the spatial anomaly probability segmentation head, the spatial anomaly classification head, and the spatial anomaly positioning head.
[0105] The present invention designs the spatial anomaly probability segmentation mapping relationship of the ink coating according to the effective ink coating distribution monitoring attention parameter, accurately divides the probability distribution area where the ink coating appears abnormal, helps to identify the area where the problem may occur in advance, and reduces the defective product rate in production. Based on the evaluation type-target correlation data and the ink coating feature evaluation type data, the weighted spatial anomaly attention channel analysis and processing are performed to determine the importance and influence of different types of anomalies in space, which helps to optimize the sensitivity and accuracy of anomaly detection. According to the spatial anomaly probability segmentation head and the weighted spatial anomaly attention channel data, the spatial anomaly classification mapping relationship is designed, which can accurately classify and identify different types of anomalies, help to quickly locate the anomaly, and take corresponding adjustment measures to ensure the stability and consistency of the ink coating quality. According to the spatial anomaly monitoring probability segmentation head and the spatial anomaly classification head, the spatial anomaly positioning mapping relationship of the ink coating is designed, and the specific location and range of each anomaly are accurately located, which helps to quickly respond to and handle abnormal situations and reduce downtime and resource waste in production. A spatial anomaly attention prediction network model for ink coating is constructed based on the spatial anomaly probability segmentation head, spatial anomaly classification head and spatial anomaly positioning head. It can comprehensively consider the probability, type and location of various abnormal situations, not only predict the probability of abnormality, but also provide detailed abnormality information to help the production team respond and adjust quickly.
[0106] In the embodiment of the present invention, the effective ink coating distribution is used to monitor the attention parameter, and the input end of a probability segmentation network is designed. The network adopts a semantic segmentation model architecture such as U-Net or DeepLab, and the input is the local ink coating feature data, and the output is the abnormal probability value of each pixel. During the training process, a loss function such as weighted cross entropy is used to increase the influence of the feature with a larger attention parameter on the final output. After the training is completed, the feature map of the middle layer of the model is output as a spatial abnormality probability segmentation head for subsequent abnormality detection. The correlation between each evaluation type indicator and the final quality target is analyzed by the evaluation type-target correlation data. Correspondingly, the ink coating feature evaluation type data is used to find out which feature indicators belong to each evaluation type to construct a spatial abnormal attention feature channel, and the evaluation type-target correlation is used as a weight factor to weight the spatial abnormal attention feature channel, enhance those feature channels that have a greater impact on the final quality, and organize the weighted feature channels into weighted spatial abnormal attention channel data to provide support for subsequent abnormal classification. According to the spatial abnormality probability segmentation head and the weighted spatial abnormality attention channel data, the spatial abnormality classification mapping relationship of the ink coating is designed to obtain the spatial abnormality classification head. Combine the spatial anomaly probability segmentation head with the weighted spatial anomaly attention channel data to design a classification-based deep learning model. The input is the weighted feature channel data, and the output is the anomaly category to which each pixel belongs. For example, a convolutional neural network or transformer architecture is used, and cross entropy loss is used for end-to-end training. After the training is completed, the feature map output of the middle layer of the model is taken out as the spatial anomaly classification head for subsequent anomaly type classification. Combine the spatial anomaly probability segmentation head and the spatial anomaly classification head to design a spatial anomaly localization module. This module can output the location coordinates and shape information of each detected anomaly area based on the idea of region proposal network (RPN) or instance segmentation. Using non-maximum suppression (NMS) and other technologies, the overlapping anomaly areas are merged to obtain the final anomaly localization result. The output feature map of the localization module is used as the spatial anomaly localization head for anomaly localization analysis. The spatial anomaly probability segmentation head, spatial anomaly classification head and spatial anomaly positioning head are combined into a unified deep learning network. The network adopts the encoder-decoder U-Net architecture and takes the local ink coating feature data as input. The output of the network includes: anomaly probability segmentation map, anomaly category classification map, and abnormal area positioning information. During the training process, a joint optimization method is adopted to simultaneously optimize the loss functions of the three output tasks. After the training is completed, an end-to-end spatial anomaly attention prediction network model is obtained, which can quickly and accurately identify abnormal areas in the ink coating.
[0107] Preferably, step S352 includes the following steps:
[0108] According to the ink coating feature evaluation type data, the spatial anomaly attention channel analysis of the ink coating is performed to generate spatial anomaly attention channel data; according to the evaluation type-target correlation data, the spatial anomaly monitoring attention channel data is weighted to generate weighted spatial anomaly attention channel data.
[0109] The present invention performs spatial anomaly attention channel analysis of the ink coating based on the ink coating feature assessment type data, identifies the characteristics and regularities of different types of anomalies in spatial distribution, helps determine the areas and types where anomalies may occur, and provides basic data support for subsequent anomaly detection and processing. The generated spatial anomaly attention channel data reflects the distribution of abnormal attention in different areas or locations, and can help distinguish abnormal areas. Using the assessment type-target correlation data to perform weighted processing on the spatial anomaly monitoring attention channel data and adjust the importance of different anomaly types in spatial anomaly detection, the sensitivity and accuracy of anomaly detection can be optimized, ensuring that key anomalies can be processed and responded to in a timely and effective manner.
[0110] In the embodiment of the present invention, the spatial anomaly attention channel analysis of the ink coating is performed according to the ink coating feature evaluation type data, and the corresponding spatial anomaly attention channel is designed one by one through each type of data in the ink coating feature evaluation type data, which is used to distinguish the anomalies identified by different ink coating feature evaluation types and generate spatial anomaly attention channel data. The evaluation type-target correlation data records the degree of correlation between each evaluation type indicator and the final quality target, and the evaluation type-target correlation is used as a weight factor to weight the spatial anomaly attention channel data, such as exponential weighting, logarithmic weighting, etc., to highlight the importance of different data types and generate weighted spatial anomaly attention channel data.
[0111] Preferably, step S5 comprises the following steps:
[0112] Step S51: obtaining ink coating control parameters;
[0113] Step S52: performing control parameter abnormality label identification processing on the ink coating control parameter according to the ink coating space abnormality data, and generating an abnormal label identification ink coating control parameter;
[0114] Step S53: establishing a mapping relationship between ink coating space anomalies and ink coating control parameters based on the supervised learning algorithm and the abnormal label identification ink coating control parameters, so as to obtain an ink coating control parameter abnormality prediction model;
[0115] Step S54: configuring an ink coating control parameter abnormality monitoring engine according to the ink coating control parameter abnormality prediction model, and executing the ink coating control parameter intelligent monitoring operation through the ink coating control parameter abnormality monitoring engine.
[0116] The present invention obtains ink coating control parameters, which is a key step in reflecting the stability of ink coating quality during the production process. The ink coating control parameters are processed with abnormal labels based on the ink coating spatial abnormal data, and the control parameters associated with the abnormalities are effectively marked to help distinguish between normal and abnormal operating conditions. A supervised learning algorithm is used to establish a prediction model based on the ink coating control parameters identified by the abnormal labels to predict the ideal setting or adjustment of the control parameters under different ink coating abnormal conditions to keep the ink coating quality within an acceptable range. An ink coating control parameter abnormality monitoring engine is configured, and the established abnormality prediction model is deployed to the real-time monitoring system to monitor the abnormal conditions of the ink coating control parameters in real time, and timely adjustment measures are taken to ensure stability and quality consistency during the production process.
[0117] In an embodiment of the present invention, ink coating control parameters are obtained, for example, various control parameters related to the ink coating on the production line are collected, such as temperature, pressure, flow rate, rotation speed, etc. These parameters are usually obtained by production equipment, automation system or manual recording, and the collected ink coating control parameters are organized into a structured data format for subsequent analysis and modeling. The control parameter abnormality label identification processing is performed on the ink coating control parameters according to the ink coating spatial abnormality data, and the ink coating distribution monitoring feature data is monitored in real time in combination with the constructed spatial abnormality attention prediction network model. When the model detects a spatial abnormality, the ink coating spatial abnormality data is obtained, and the ink coating control parameter data at the corresponding time point of the ink coating spatial abnormality data is obtained. These control parameter data are marked with abnormal labels to identify which parameter values have deviated under abnormal conditions, and generate abnormal labels to identify the ink coating control parameters. A mapping relationship between the ink coating spatial anomaly and the ink coating control parameters is established based on the supervised learning algorithm and the abnormal label identification of the ink coating control parameters to obtain an abnormal prediction model for the ink coating control parameters. The ink coating control parameter data set is identified by abnormal labels as training samples for supervised learning. Through supervised learning algorithms such as decision trees, random forests or neural networks, the mapping relationship between ink coating spatial anomalies and ink coating control parameters is established. When training the model, the ink coating control parameters are identified by abnormal labels as supervisory signals, and the model parameters are optimized so that they can accurately predict the control parameter anomalies, and finally the ink coating control parameter anomaly prediction model that can predict the ink coating control parameter anomalies is obtained. The ink coating control parameter anomaly prediction model is deployed to the production site as the core algorithm module of the anomaly monitoring engine. The control parameter data of the ink coating is collected in real time and input into the anomaly monitoring engine for prediction and analysis. When the engine detects that some control parameters are about to become abnormal, it will issue a warning signal in time to remind production management personnel to intervene. Through continuous monitoring and feedback, the performance of the anomaly monitoring engine is continuously optimized to improve the ability to prevent ink coating quality anomalies.
[0118] This specification provides an ink coating distribution monitoring system based on big data, which is used to execute the ink coating distribution monitoring method based on big data. The ink coating distribution monitoring system based on big data includes:
[0119] The ink coating distribution monitoring module is used to perform ink coating distribution monitoring processing on the printed sample using the sensor integrated device to obtain ink coating distribution monitoring data;
[0120] The ink coating feature analysis module is used to obtain historical ink coating evaluation data; perform ink coating feature evaluation type analysis based on the historical ink coating evaluation data to obtain ink coating feature evaluation type data; perform feature extraction on the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data;
[0121] The spatial anomaly attention prediction network model building module is used to perform effective data analysis of local ink coating distribution monitoring characteristics based on ink coating distribution monitoring feature data, and generate effective local ink coating distribution monitoring feature data; design effective ink coating distribution monitoring attention parameters based on effective local ink coating distribution monitoring feature data; and build a spatial anomaly attention prediction network model for ink coating based on the effective ink coating distribution monitoring attention parameters;
[0122] The ink coating distribution monitoring anomaly feedback module is used to transmit the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis and generate ink coating spatial anomaly data; perform global ink coating stability analysis based on the ink coating spatial anomaly data and generate global ink coating stability data; transmit the global ink coating stability data to the terminal to perform ink coating distribution monitoring anomaly feedback operations;
[0123] The ink coating control parameter intelligent monitoring module is used to obtain the ink coating control parameters; based on the ink coating spatial anomaly data and the ink coating control parameters, a mapping relationship between the ink coating spatial anomaly and the ink coating control parameters is established to obtain an ink coating control parameter anomaly prediction model; according to the ink coating control parameter anomaly prediction model, an ink coating control parameter anomaly monitoring engine is configured, and the ink coating control parameter intelligent monitoring operation is performed through the ink coating control parameter anomaly monitoring engine.
[0124] The beneficial effect of the present application is that the ink coating distribution monitoring method based on big data of the present invention performs multi-source variable evaluation on ink coating monitoring, and automatically detects and predicts abnormal conditions of the ink coating through a self-designed spatial abnormal attention prediction network, thereby improving the efficiency of ink coating monitoring and the quality evaluation results. By establishing an abnormal prediction model and monitoring engine for ink coating control parameters, and providing a data-driven ink coating quality management method, it is possible to identify subtle abnormal conditions of the ink coating caused by the ink coating control parameters, so that the feedback ink coating quality results are more consistent with the actual results.
[0125] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.
[0126] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.
Claims
1. A method for monitoring ink coating distribution based on big data, characterized in that: The following steps are involved: Step S1: using a sensor integrated device to perform ink coating distribution monitoring processing on a printed sample to obtain ink coating distribution monitoring data; Step S2: Obtaining historical ink coating evaluation data; Performing ink coating characteristic evaluation type analysis based on historical ink coating evaluation data to obtain ink coating characteristic evaluation type data; Extracting features of the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data; Step S3: performing effective data analysis of local ink coating distribution monitoring characteristics based on the ink coating distribution monitoring characteristic data to generate effective local ink coating distribution monitoring characteristic data; designing effective ink coating distribution monitoring attention parameters based on the effective local ink coating distribution monitoring characteristic data; and constructing a spatial abnormal attention prediction network model for the ink coating based on the effective ink coating distribution monitoring attention parameters; Step S4: transmitting the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis to generate ink coating spatial anomaly data; Perform global ink coating stability analysis based on ink coating spatial anomaly data to generate global ink coating stability data; Transmit global ink coating stability data to the terminal to perform ink coating distribution monitoring abnormal feedback operations; Step S5; Obtain ink coating control parameters; A mapping relationship between ink coating space anomaly and ink coating control parameters is established based on ink coating space anomaly data and ink coating control parameters to obtain an ink coating control parameter anomaly prediction model; An ink coating control parameter anomaly monitoring engine is configured according to the ink coating control parameter anomaly prediction model, and an ink coating control parameter intelligent monitoring operation is performed through the ink coating control parameter anomaly monitoring engine.
2. The ink coating distribution monitoring method based on big data according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: using a sensor integrated device to perform ink coating monitoring on a printed sample; Step S12: collecting ink coating monitoring nodes for the ink coating monitoring operation, generating ink coating monitoring node data, and establishing an ink coating distribution network according to the ink coating monitoring node data; Step S13: collecting ink coating monitoring data for the ink coating monitoring operation to obtain ink coating monitoring data; Step S14: transmitting the ink coating monitoring data to the ink coating distribution network to identify the location characteristics of the ink coating monitoring data, and generating ink coating distribution monitoring data.
3. The ink coating distribution monitoring method based on big data according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: Acquire historical ink coating evaluation data; Step S22: extracting historical ink coating evaluation type data and historical ink coating evaluation target data of the historical ink coating evaluation data; Step S23: performing type subset and target association processing of ink coating evaluation according to historical ink coating evaluation type data and historical ink coating evaluation target data, and generating evaluation type subset-target association data; Step S24: performing assessment type-target correlation analysis based on the assessment type subset-target association data to generate assessment type-target correlation data; Step S25: performing ink coating feature evaluation type analysis according to the evaluation type-target core correlation data to obtain ink coating feature evaluation type data; Step S26: extracting features from the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data.
4. The ink coating distribution monitoring method based on big data according to claim 3 is characterized in that: Step S24 includes the following steps: Step S241: performing relevant characteristic factor analysis of the assessment type and the assessment target according to the assessment type subset-target association data to generate assessment type-target relevant characteristic factor data; Step S242: constructing a core-related evaluation model of the evaluation type and the evaluation target based on the preset core-typical correlation analysis algorithm and the evaluation type-target related characteristic factor data to obtain an evaluation type-target core-related evaluation model; Step S243: performing an assessment type-target correlation analysis on the assessment type subset-target associated data based on the assessment type-target core correlation assessment model to generate assessment type-target correlation data.
5. The ink coating distribution monitoring method based on big data according to claim 4 is characterized in that: Step S241 includes the following steps: Based on the assessment type subset-target association data, an association volatility analysis is performed to generate assessment type subset-target association volatility data; based on the assessment type subset-target association volatility data, a correlation characteristic factor analysis of the assessment type and the assessment target is performed to generate assessment type-target correlation characteristic factor data.
6. The ink coating distribution monitoring method based on big data according to claim 3 is characterized in that: Step S3 includes the following steps: Step S31: extracting the local selected interval data of the central node of the ink coating distribution monitoring characteristic data according to the preset local ink coating selection interval data, so as to obtain the local ink coating distribution monitoring characteristic data; Step S32: performing local monitoring characteristic fluctuation deviation analysis according to the local ink coating distribution monitoring characteristic data to generate local monitoring characteristic fluctuation deviation data; Step S33: Screening effective data of the local ink coating distribution monitoring characteristic data according to the local monitoring characteristic fluctuation deviation data to generate effective local ink coating distribution monitoring characteristic data; Step S34: designing effective ink coating distribution monitoring attention parameters according to effective local ink coating distribution monitoring characteristic data; Step S35: Construct a spatial abnormal attention prediction network model of the ink coating based on the effective ink coating distribution monitoring attention parameters.
7. The ink coating distribution monitoring method based on big data according to claim 6 is characterized in that: Step S35 includes the following steps: Step S351: designing a spatial anomaly probability segmentation mapping relationship of the ink coating according to the effective ink coating distribution monitoring attention parameter to obtain a spatial anomaly probability segmentation head; Step S352: performing weighted spatial abnormal attention channel analysis processing based on the evaluation type-target correlation data and the ink coating feature evaluation type data to generate weighted spatial abnormal attention channel data; Step S353: designing a spatial anomaly classification mapping relationship of the ink coating according to the spatial anomaly probability segmentation head and the weighted spatial anomaly attention channel data to obtain a spatial anomaly classification head; Step S354: designing a spatial anomaly positioning mapping relationship of the ink coating according to the spatial anomaly monitoring probability segmentation head and the spatial anomaly monitoring classification head to obtain a spatial anomaly positioning head; Step S355: construct a spatial anomaly attention prediction network model for the ink coating based on the spatial anomaly probability segmentation head, the spatial anomaly classification head, and the spatial anomaly positioning head.
8. The ink coating distribution monitoring method based on big data according to claim 7 is characterized in that: Step S352 includes the following steps: According to the ink coating feature evaluation type data, the spatial anomaly attention channel analysis of the ink coating is performed to generate spatial anomaly attention channel data; according to the evaluation type-target correlation data, the spatial anomaly monitoring attention channel data is weighted to generate weighted spatial anomaly attention channel data.
9. The ink coating distribution monitoring method based on big data according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: obtaining ink coating control parameters; Step S52: performing control parameter abnormality label identification processing on the ink coating control parameter according to the ink coating space abnormality data, and generating an abnormal label identification ink coating control parameter; Step S53: establishing a mapping relationship between ink coating space anomalies and ink coating control parameters based on the supervised learning algorithm and the abnormal label identification ink coating control parameters, so as to obtain an ink coating control parameter abnormality prediction model; Step S54: configuring an ink coating control parameter abnormality monitoring engine according to the ink coating control parameter abnormality prediction model, and executing the ink coating control parameter intelligent monitoring operation through the ink coating control parameter abnormality monitoring engine.
10. An ink coating distribution monitoring system based on big data, characterized in that: Used to execute the ink coating distribution monitoring method based on big data as claimed in claim 1, the ink coating distribution monitoring system based on big data comprises: The ink coating distribution monitoring module is used to perform ink coating distribution monitoring processing on the printed sample using the sensor integrated device to obtain ink coating distribution monitoring data; The ink coating feature analysis module is used to obtain historical ink coating evaluation data; perform ink coating feature evaluation type analysis based on the historical ink coating evaluation data to obtain ink coating feature evaluation type data; perform feature extraction on the ink coating distribution monitoring data according to the ink coating feature evaluation type data to generate ink coating distribution monitoring feature data; The spatial anomaly attention prediction network model building module is used to perform effective data analysis of local ink coating distribution monitoring characteristics based on ink coating distribution monitoring feature data, and generate effective local ink coating distribution monitoring feature data; design effective ink coating distribution monitoring attention parameters based on effective local ink coating distribution monitoring feature data; and build a spatial anomaly attention prediction network model for ink coating based on the effective ink coating distribution monitoring attention parameters; The ink coating distribution monitoring anomaly feedback module is used to transmit the ink coating distribution monitoring feature data to the spatial anomaly attention prediction network model to perform ink coating spatial anomaly analysis and generate ink coating spatial anomaly data; perform global ink coating stability analysis based on the ink coating spatial anomaly data and generate global ink coating stability data; transmit the global ink coating stability data to the terminal to perform ink coating distribution monitoring anomaly feedback operations; The ink coating control parameter intelligent monitoring module is used to obtain the ink coating control parameters; based on the ink coating spatial anomaly data and the ink coating control parameters, a mapping relationship between the ink coating spatial anomaly and the ink coating control parameters is established to obtain an ink coating control parameter anomaly prediction model; according to the ink coating control parameter anomaly prediction model, an ink coating control parameter anomaly monitoring engine is configured, and the ink coating control parameter intelligent monitoring operation is performed through the ink coating control parameter anomaly monitoring engine.
