An industrial internet-based decorative paper production data processing method and system

Abstract

The present application relates to the technical field of industrial internet, and proposes a decoration paper production data processing method and system based on industrial internet. By acquiring local micro-environment information of the coating section, paper web micro-characteristics information before the coating section, and coating behavior information after the drying section, and aligning them with the time-space coordinate system of the production line, real-time correlation analysis is performed on different types of event markers aligned in time-space, event combinations meeting the preset correlation rules are identified, and potential defect event markers are triggered. By acquiring downstream online quality detection results, and according to the potential defect event markers and the quality detection results, the correlation between the event combinations and the coating quality defects is determined. Thus, effective diagnosis and optimization suggestions are provided for quality fluctuations in decoration paper production, thereby significantly improving the stability of product quality and production efficiency.

Description

Technical Field

[0001] This application relates to the field of industrial internet technology, and more specifically, to a method and system for processing decorative paper production data based on industrial internet. Background Technology

[0002] In decorative paper production, the industrial internet system introduced aims to stabilize and improve product quality by monitoring macroscopic process parameters. However, in practice, a contradictory phenomenon often occurs: the system displays everything as normal, but the product exhibits subtle fluctuations in coating uniformity or surface hardness, rendering existing data analysis models ineffective. In-depth investigation revealed that the root cause lies in the instantaneous fluctuations of the local microenvironment. For example, in critical drying areas after coating, due to their proximity to channels and ventilation openings, minute changes in temperature and humidity, lasting only a few seconds, can occur that are difficult for macroscopic sensors to capture. Such transient disturbances are sufficient to interfere with the coating curing process, leading to defects at the microscopic level.

[0003] Furthermore, the inherent microscopic inhomogeneities of the paper web (such as differences in porosity) interact in complex ways with the aforementioned microenvironmental changes, collectively triggering localized quality issues. Existing industrial internet systems, with their sensor layouts and data models designed for overall and average states, are unable to perceive these highly localized, transient fluctuations, nor can they correlate them with the microscopic characteristics of the paper web. Therefore, they exhibit insufficient diagnostic capabilities when facing complex quality problems caused by the interplay of multiple implicit factors. This highlights the need for higher-dimensional, more nuanced data sensing technologies in the pursuit of refined production.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] This application discloses a data processing method and system for decorative paper production based on the Industrial Internet, which aims to solve the technical problems in existing decorative paper production where product quality may still fluctuate unpredictably even when macroscopic process parameters are normal, and where existing data analysis methods are inadequate in the face of complex quality problems and cannot provide effective diagnosis and optimization suggestions.

[0006] The technical solution of this application is as follows:

[0007] Firstly, this application discloses a method for processing decorative paper production data based on the Industrial Internet, the method comprising:

[0008] Acquire local microenvironment information from the coating section, paper web microstructure information from the pre-coating section, and coating behavior information from the post-coating drying section;

[0009] The local microenvironment information, paper web microstructure information, coating behavior information are aligned with the production line spatiotemporal coordinate system, and corresponding event markers are generated based on preset dynamic thresholds. These event markers include microenvironment disturbance event markers, paper web microstructure abnormal event markers, and coating behavior abnormal event markers.

[0010] Real-time correlation analysis is performed on multiple event tags of different types that are aligned in time and space to identify event combinations that meet preset correlation rules and trigger potential defect event tags;

[0011] The downstream online quality inspection results are obtained, and the correlation between the event combination and coating quality defects is determined based on the paper web position information, the potential defect event markers, and the downstream online quality inspection results; the paper web position information includes longitudinal position and transverse position.

[0012] Furthermore, based on the above, real-time correlation analysis is performed on multiple event markers of different types aligned in time and space to identify event combinations that satisfy preset correlation rules and trigger potential defect event markers. This includes: assigning a time-space identifier based on timestamps and spatial coordinates to each event marker; aggregating multiple sets of event markers of different types that are adjacent in time and space based on the time-space identifiers using preset sliding time windows and spatial windows; determining whether the event marker set satisfies the preset correlation rules; the preset correlation rules include logical combinations consisting of abnormal conditions of local microenvironment information, abnormal conditions of paper web micro-characteristic information, and abnormal conditions of coating behavior information; the preset correlation rules include one or more types of logical combinations; in response to the fact that all events in the event marker set satisfy the preset correlation rules, it is determined that the event combination corresponding to the event marker set has a potential defect, and the potential defect event marker is triggered.

[0013] In some preferred embodiments, the preset association rules include condition judgment thresholds corresponding to abnormal local microenvironment information, abnormal paper web micro-characteristic information, and abnormal coating behavior information; determining that the event combination corresponding to the event mark set has a potential defect and triggering the potential defect event mark, further includes: determining the initial confidence level of the potential defect event mark based on the degree of deviation between each event parameter in the event mark set and the corresponding condition judgment threshold.

[0014] Furthermore, based on the potential defect event markers and downstream online quality inspection results, the correlation between the event combination and coating quality defects is determined, including: matching the spatiotemporal information of the potential defect event markers with the spatiotemporal information of the downstream online quality inspection results; if the match is successful, increasing the confidence score of the event combination; and confirming a causal relationship between the event combination and the coating quality defects when the confidence score exceeds a preset confidence threshold.

[0015] Based on the above, the method further includes: when the severity of quality defects shown in the downstream online quality inspection results deviates significantly from the expected defect characteristics in the causal relationship, determining that there is a missing link in the causal relationship, generating hypotheses about one or more missing links; and dynamically completing the causal relationship based on the missing link hypotheses.

[0016] As a technological improvement, the causal relationship is dynamically completed based on the missing link hypothesis, including: initiating real-time monitoring of one or more proxy signals corresponding to the missing link hypothesis; the proxy signal is used to indirectly reflect the state of the missing link; based on the proxy signal obtained by real-time monitoring, it is spatiotemporally correlated and compared with the production process events and quality defects collected in real time; when the change of the proxy signal is consistent with the expected state of the missing link hypothesis, and matches the production process events and quality defects in spatiotemporal terms, the causal relationship is dynamically completed.

[0017] In one implementation, hypotheses about one or more missing links are generated, including: when the quality defect is a decrease in the adhesion of the decorative paper exceeding a preset threshold, at least one of the following hypotheses is generated: the dynamic surface tension of the coating liquid increases; submicron-level voids are formed at the interface between the coating and the paper web.

[0018] Based on the above, the causal relationship is dynamically completed according to the missing link hypothesis, including: fusing multi-source information from real-time monitored agent signals, event combinations, and quality defects to generate multi-source fused information; when the multi-source fused information meets the preset matching logic, the causal chain containing the missing link hypothesis is confirmed; a confidence score is generated for the confirmed causal chain, and the confidence score is increased each time the preset matching logic is successfully met; when the confidence score reaches the preset confirmation threshold, the causal chain is marked as a reliable diagnostic basis, and the causal relationship is completed.

[0019] Based on the above, the method further includes: generating and outputting a visualization interface based on causal relationships; the visualization interface is used to display the sequence of related events that lead to coating quality defects in chronological order and spatial location; in a web-based interactive dashboard, macroscopic process parameter curves, micro-environmental disturbance events, paper web micro-abnormal events, coating behavior abnormal events and location markers of final quality defects are displayed synchronously in the form of a combination of time axis and paper web cross-sectional view.

[0020] Secondly, this application also discloses a decorative paper production data processing system based on the Industrial Internet. The system includes: an acquisition module for acquiring local microenvironment information from the coating section, paper web microstructure information from the pre-coating section, and coating behavior information from the post-coating drying section; an alignment and generation module for aligning the local microenvironment information, paper web microstructure information, and coating behavior information with the production line's spatiotemporal coordinate system, and generating corresponding event markers based on preset dynamic thresholds; these event markers include microenvironmental disturbance event markers, paper web microstructure abnormality event markers, and coating behavior abnormality event markers; a correlation analysis module for performing real-time correlation analysis on multiple spatiotemporally aligned event markers of different types to identify event combinations that satisfy preset correlation rules and trigger potential defect event markers; and a determination module for acquiring downstream online quality inspection results and determining the correlation between the event combinations and coating quality defects based on paper web position information, the potential defect event markers, and the downstream online quality inspection results; the paper web position information includes longitudinal and transverse positions.

[0021] Beneficial Effects: The decorative paper production data processing method based on the Industrial Internet disclosed in this application acquires local micro-environment information of the coating section, paper web micro-characteristic information of the pre-coating section, and coating behavior information of the post-coating drying section, and aligns these with the spatiotemporal coordinate system of the production line. This enables the capture of transient micro-changes that are difficult to detect by traditional macroscopic sensors. By generating micro-environmental disturbance event markers, paper web micro-abnormal event markers, and coating behavior abnormal event markers based on preset dynamic thresholds, this application can transform these subtle anomalies into identifiable events, overcoming the problem that existing systems cannot capture local, transient micro-disturbances. Furthermore, by performing real-time correlation analysis on these spatiotemporally aligned event markers of different types, this application identifies event combinations that meet preset correlation rules and triggers potential defect event markers. This allows this application to reveal quality problems caused by the interplay of multiple hidden factors, solving the limitation of existing systems in effectively correlating micro-changes with product quality problems. Ultimately, by obtaining downstream online quality inspection results and determining the correlation between event combinations and coating quality defects based on potential defect event marking and quality inspection results, this application can establish a precise causal chain, providing effective diagnosis and optimization suggestions for quality fluctuations in decorative paper production, thereby significantly improving product quality stability and production efficiency, and overcoming the shortcomings of existing data analysis methods in dealing with complex quality problems. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the steps of the decorative paper production data processing method based on the Industrial Internet disclosed in an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the structure of a decorative paper production data processing system based on the Industrial Internet disclosed in an embodiment of the present invention. Detailed Implementation

[0024] The implementation details of the technical solution in this embodiment are described in detail below:

[0025] Traditional data processing methods for decorative paper production struggle to address complex quality issues. Even when the system displays all macroscopic process parameters as ideal, unpredictable fluctuations in product quality can still occur, such as subtle deviations in coating uniformity or surface hardness. This renders existing data analysis methods inadequate, failing to provide effective diagnostic and optimization suggestions. Failure to address these issues will result in low production efficiency, unstable product quality, and increased production costs.

[0026] In response to this, firstly, this application proposes a method for processing decorative paper production data based on the Industrial Internet, such as... Figure 1 As shown, the method includes:

[0027] S101, acquire local microenvironment information from the coating section, paper web microstructure information from the pre-coating section, and coating behavior information from the post-coating drying section.

[0028] S102, the local microenvironment information, paper web microstructure information, coating behavior information are aligned with the production line spatiotemporal coordinate system, and corresponding event markers are generated based on preset dynamic thresholds; the event markers include microenvironment disturbance event markers, paper web microstructure abnormal event markers, and coating behavior abnormal event markers;

[0029] S103, performs real-time correlation analysis on multiple event markers of different types aligned in time and space to identify event combinations that meet preset correlation rules and trigger potential defect event markers;

[0030] S104, Obtain downstream online quality inspection results, and determine the correlation between the event combination and coating quality defects based on the paper web position information, the potential defect event markers, and the downstream online quality inspection results; the paper web position information includes longitudinal position and transverse position.

[0031] This application introduces a spatiotemporal alignment and correlation analysis mechanism for multi-source heterogeneous data, which can more precisely capture micro-events that may lead to quality defects in the production process. Combined with downstream quality inspection results, it establishes a causal relationship between events and defects, thereby effectively overcoming the limitations of existing technologies in diagnosing complex quality problems.

[0032] To better understand the technical solution proposed in this application, some key terms are explained first. "Local microenvironment information" refers to microscopic environmental parameters that affect the coating process within a localized area of ​​the coating section, such as temperature, humidity, airflow velocity, and dust concentration. This information is typically collected in real-time near the coating head using high-precision sensors. "Base paper microstructure information" refers to the physical and chemical properties of the base paper before coating, such as surface roughness, porosity, fiber distribution uniformity, surface tension, and hygroscopicity. These properties significantly influence the spreading and penetration behavior of the coating liquid on the base paper. "Coating behavior information" refers to the dynamic changes of the coating liquid after it forms a coating on the base paper, such as coating thickness, leveling properties, drying rate, and degree of curing. This information can be obtained through online spectrometers, infrared sensors, or machine vision systems. The "production line spatiotemporal coordinate system" is a unified reference framework used to map data collected from different sources and at different times to specific locations and times on the production line, ensuring the spatiotemporal consistency of the data. "Event tag" refers to an identifier automatically generated by the system when a monitored parameter exceeds a preset dynamic threshold, used to indicate potential anomalies in the production process. "Preset dynamic threshold" refers to a parameter threshold determined based on production process requirements and historical data analysis, which can be dynamically adjusted according to production conditions or product type. When a monitored value exceeds this threshold, it is considered an anomaly. "Event combination" refers to a set of multiple event tags of different types that are interrelated in time and space; the combined effect of these events may lead to specific coating quality defects. "Potential defect event tag" refers to a combination of events identified through correlation analysis that foreshadows potential quality defects. "Coating quality defects" refer to various problems that do not meet quality standards during the coating process of decorative paper, such as uneven coating, poor adhesion, surface spots, and blistering.

[0033] The method proposed in this application first requires acquiring local microenvironment information from the coating stage, microscopic property information of the base paper from the pre-coating stage, and coating behavior information from the post-coating drying stage. Acquiring this information is fundamental for achieving refined data analysis. For example, local microenvironment information can be collected in real time by deploying multiple high-precision temperature and humidity sensors, airflow sensors, and dust sensors near the coating head. Base paper microscopic property information can be continuously measured before the base paper enters the coating machine using an online scanning electron microscope, surface roughness meter, or near-infrared spectrometer. Coating behavior information can be obtained by setting up an online coating thickness sensor, infrared dryness sensor, or high-speed camera in the post-coating drying stage to acquire dynamic images of the coating surface. These sensors and devices will continuously collect data and transmit it to the data processing system.

[0034] Next, the local microenvironment information, base paper microstructure information, and coating behavior information are aligned with the production line's spatiotemporal coordinate system, and corresponding event markers are generated based on preset dynamic thresholds. For example, each data point collected by a sensor can be assigned a unique timestamp and corresponding production line location coordinates. When a local microenvironment parameter (such as temperature) exceeds its preset dynamic threshold for several consecutive seconds, the system generates a microenvironment disturbance event marker, which includes the time, location, abnormal parameter value, and duration of the event. Similarly, when the surface roughness of the base paper exceeds a preset dynamic threshold in a certain area, a base paper microstructure abnormality event marker is generated; when the coating drying rate is abnormal, a coating behavior abnormality event marker is generated. The generation of these event markers transforms massive amounts of raw data into more business-meaning discrete events, facilitating subsequent analysis.

[0035] Subsequently, real-time correlation analysis is performed on multiple event markers of different types aligned in time and space to identify event combinations that satisfy preset correlation rules and trigger potential defect event markers. For example, the system can continuously monitor whether microenvironmental disturbance event markers, base paper micro-anomaly event markers, and coating behavior anomaly event markers occur simultaneously within the same time period and production line area. If the preset correlation rule is defined as the simultaneous occurrence of "local temperature rise" and "abnormal base paper surface tension," followed by the event of "poor coating leveling," the system will identify this event combination and trigger a potential defect event marker, indicating that there may be coating leveling defects in that area. This real-time correlation analysis can capture complex causal relationships that are difficult to reveal with a single event.

[0036] Finally, the system obtains the downstream online quality inspection results and determines the correlation between the event combination and the coating quality defect based on the potential defect event markers and the downstream online quality inspection results. For example, when the system triggers a potential defect event marker, it waits for the batch of products to pass through the downstream online quality inspection equipment. If the online quality inspection results show that a coating leveling defect does exist in the spatiotemporal region corresponding to the potential defect event marker, then the system will confirm a strong correlation between the previously identified event combination (local temperature rise, abnormal surface tension of the base paper, poor coating leveling) and the coating quality defect. In this way, the system can verify and strengthen the diagnosis of potential defects and establish a reliable causal chain.

[0037] The overall working principle of this application lies in achieving early warning and accurate diagnosis of potential quality defects through in-depth mining and correlation analysis of multi-source heterogeneous data in the decorative paper production process. Traditional data processing methods often only focus on macroscopic process parameters, making it difficult to capture microscopic events that cause quality fluctuations. This application acquires local micro-environment information from the coating section, microscopic characteristics of the base paper from the pre-coating section, and coating behavior information from the post-coating drying section. It then aligns these multi-dimensional, multi-scale data in time and space and generates event markers based on preset dynamic thresholds, thereby transforming continuous production data into a discrete, analyzable event stream.

[0038] Furthermore, by performing real-time correlation analysis on these spatiotemporally aligned event markers, this application can identify event combinations that satisfy preset correlation rules. These combinations are often the underlying causes of specific coating quality defects. For example, a single local temperature increase may not be sufficient to cause a quality problem, but if it occurs simultaneously with multiple micro-events such as abnormal surface tension of the base paper or excessively rapid coating drying rate, it is highly likely to trigger defects such as decreased coating adhesion or surface unevenness. By triggering potential defect event markers, this application achieves early warning of potential quality problems.

[0039] Finally, this application compares and verifies these potential defect events with downstream online quality inspection results. This mechanism, which combines actual quality feedback, not only confirms the correlation between event combinations and coating quality defects, but also continuously optimizes and improves the preset correlation rules, making the diagnostic model more accurate and reliable. Therefore, this application can provide more instructive diagnostic information for the production process, helping production personnel to adjust process parameters in a timely manner to avoid or reduce the occurrence of quality defects.

[0040] Compared to existing technologies, this application has significant advantages and innovations. Traditional methods for processing data in decorative paper production mainly rely on monitoring and analyzing macroscopic process parameters, such as coating speed, oven temperature, and coating amount. When product quality defects occur, it is often difficult to find direct, microscopic causes from these macroscopic data, leading to low diagnostic efficiency or even an inability to diagnose. For example, even if all macroscopic parameters are within the normal range, the product may still exhibit subtle deviations in coating uniformity or surface hardness, making existing methods inadequate for handling complex quality problems.

[0041] The core innovation of this application lies in its ability to break through the limitations of macroscopic parameters and delve into the microscopic level of the production process. By acquiring information on the local microenvironment, the microscopic properties of the base paper, and the behavior of the coating, this application can capture minute changes and anomalies that are imperceptible by traditional methods. For example, airflow disturbances near the coating head, local fluctuations in the microporosity of the base paper surface, and abnormal micro-rheological behavior of the coating during the drying process are all key factors affecting the quality of the final product, but are difficult to effectively monitor using macroscopic parameters.

[0042] Furthermore, this application introduces a spatiotemporal alignment and real-time correlation analysis mechanism, enabling the integrated analysis of micro-events from different sources, time points, and spatial locations to identify potential event combinations. This ability to fuse and analyze multi-source heterogeneous data allows this application to establish a more comprehensive and refined causal chain, thereby more accurately predicting and diagnosing coating quality defects. By combining with downstream online quality inspection results, this application can not only verify the accuracy of the diagnosis but also continuously optimize the diagnostic model, forming a closed-loop intelligent diagnostic system. This complete chain from capturing micro-events to diagnosing macro-quality defects is not available in existing technologies, greatly enhancing the quality control and optimization capabilities of the decorative paper production process.

[0043] Specifically, the above-mentioned real-time correlation analysis of multiple event markers of different types aligned in time and space to identify event combinations that meet preset correlation rules and trigger potential defect event markers can be further refined into the following steps: assigning a time-space identifier based on timestamps and spatial coordinates to each event marker; based on the time-space identifier, using preset sliding time windows and spatial windows, aggregating multiple event marker sets of different types that are adjacent in time and space; determining whether the event marker set meets the preset correlation rules; the preset correlation rules include logical combinations composed of abnormal conditions of local microenvironment information, abnormal conditions of microstructure information of base paper, and abnormal conditions of coating behavior information; the preset correlation rules include one or more types of logical combinations; in response to the fact that all events in the event marker set meet the preset correlation rules, determining that the event combination corresponding to the event marker set has a potential defect, and triggering the potential defect event marker.

[0044] Assigning a spatiotemporal identifier based on timestamps and spatial coordinates to each event marker means attaching precise time information (e.g., the specific time the event occurred) and spatial information (e.g., the specific location of the event on the production line, such as the lateral position of the coating roller or the longitudinal position of the paper web) to the event marker when it is generated. These spatiotemporal identifiers ensure the accuracy and traceability of subsequent correlation analysis.

[0045] Furthermore, based on the aforementioned spatiotemporal identifiers, a preset sliding time window and spatial window are used to aggregate multiple sets of event markers of different types that are spatiotemporally adjacent. Specifically, the sliding time window defines a time period, such as 5 seconds, 10 seconds, or longer, for collecting all event markers occurring within that time period. The sliding spatial window defines a spatial range, such as a certain area in the transverse direction of the paper web or a certain length in the longitudinal direction, for collecting event markers occurring within that spatial range. By combining these two windows, the system can identify microenvironmental disturbance event markers, base paper microscopic anomaly event markers, and coating behavior anomaly event markers that coexist within a specific time period and a specific spatial region, and aggregate them into a single event marker set. This aggregation mechanism helps to discover potential connections between different types of anomalous events.

[0046] In practical applications, determining whether the event tag set satisfies the preset association rules refers to performing a logical evaluation on the aggregated event tag set. The preset association rules are a series of predefined condition combinations, consisting of abnormal conditions related to local microenvironment information, abnormal conditions related to the microstructure of the base paper, and abnormal conditions related to the coating behavior. For example, an association rule could be "when the local microenvironment temperature rises abnormally, the surface roughness of the base paper is abnormal, and the leveling property of the coating is abnormal." These abnormal conditions are themselves based on preset dynamic thresholds. The preset association rules can include one or more types of logical combinations, such as "A AND B," "A OR C," or "(A AND B) OR C," to cover various possible defect initiation patterns. Therefore, in response to events in the event tag set all satisfying the preset association rules, it is determined that the event combination corresponding to the event tag set has a potential defect, triggering a potential defect event tag. This means that once all events in a certain event tag set meet the preset logical combination conditions, the system considers these events to constitute a potential defect initiation and generates a potential defect event tag for subsequent quality inspection results to verify and associate.

[0047] This application's solution assigns precise spatiotemporal identifiers to each event marker and aggregates these event markers using preset sliding time and spatial windows, effectively capturing the co-occurrence of different types of abnormal events in both time and space dimensions. It is precisely this refined spatiotemporal alignment and aggregation that enables the system to identify potential defects caused by the combined effects of multiple abnormal events, which are difficult to detect in traditional single-dimensional monitoring. By comparing these aggregated event sets with preset association rules, the system can quickly determine whether there are potential triggers for coating quality defects based on logical combinations constructed from expert knowledge or historical data, thus triggering potential defect event marking before the defects actually occur or are detected downstream. This mechanism significantly improves the accuracy and timeliness of defect prediction.

[0048] Through the above technical solution, this application enables early and accurate identification of potential coating quality defects during the production of decorative paper. Compared to simple event tagging and coarse correlation, this solution, by introducing spatiotemporal identifiers and a sliding window aggregation mechanism, greatly improves the precision and accuracy of event correlation analysis, effectively avoiding misjudgments or omissions caused by spatiotemporal misalignment. Furthermore, the introduction of preset correlation rules allows the system to selectively identify high-risk event combinations based on known defect patterns or expert experience. This triggers early warnings before coating quality defects fully manifest, providing a valuable time window for timely intervention and adjustment in the production process, significantly improving the level of intelligent management of the production process and the stability of product quality.

[0049] In some embodiments described above, this application proposes real-time correlation analysis of multiple event markers of different types aligned in time and space to identify event combinations that meet preset correlation rules and trigger potential defect event markers. However, in practical applications, simply triggering potential defect event markers may not effectively distinguish the severity or probability of occurrence of different potential defects, resulting in a lack of prioritization criteria for subsequent processing. Failure to address this issue may lead to wasted resources or untimely responses to high-risk defects. Therefore, this application further proposes determining the initial confidence level of potential defect event markers based on the deviation between event parameters and conditional judgment thresholds, thereby providing a more refined basis for subsequent defect confirmation and intervention.

[0050] In some embodiments of this application, the aforementioned preset association rules include conditional judgment thresholds corresponding to abnormal local microenvironment information, abnormal microstructure information of the base paper, and abnormal coating behavior information. Specifically, these conditional judgment thresholds are critical values ​​used to define whether various types of information are in an abnormal state. For example, the upper or lower limits of local microenvironment parameters such as temperature, humidity, and airflow velocity in the coating section; the allowable fluctuation range of microstructure characteristics such as porosity and surface roughness of the base paper; and the qualification standards for coating behavior parameters such as coating thickness and uniformity. When the actual monitored event parameters exceed these thresholds, they are judged to be abnormal.

[0051] Further, determining that the event combination corresponding to the event tag set has a potential defect and triggering the potential defect event tag also includes determining the initial confidence level of the potential defect event tag based on the degree of deviation between each event parameter in the event tag set and the corresponding condition judgment threshold. Here, the degree of deviation can be understood as the difference between the actual value of the event parameter and the corresponding condition judgment threshold. For example, if the threshold for a temperature parameter is 25℃, and the actual monitored value is 28℃, then the deviation is 3℃. The larger the degree of deviation, the more severe the abnormality or the greater the deviation from the normal state. The initial confidence level is a quantitative assessment assigned to the potential defect event tag based on this degree of deviation, used to initially measure the probability or severity of the potential defect. In practical applications, the degree of deviation can be mapped to a confidence score between 0 and 1 using a preset function or model, for example, using a linear function, exponential function, or piecewise function. The purpose is to provide a preliminary risk assessment for subsequent defect diagnosis and processing, enabling the system to prioritize potential defects with higher confidence levels.

[0052] This application's solution introduces conditional thresholds and determines the initial confidence level of potential defect event markers based on the deviation of event parameters from these thresholds. This effectively solves the problem in traditional solutions where only potential defect event markers are triggered without distinguishing their severity or probability of occurrence. Specifically, when multiple event markers of different types align spatiotemporally and satisfy preset association rules, the system not only identifies potential defect event combinations but also, further, comprehensively assesses the overall risk of the event combination by quantifying the degree of anomaly of each constituent event (i.e., deviation from the threshold). This quantitative assessment allows the system to assign an initial confidence level to each potential defect event marker, thereby quickly identifying event combinations most likely to cause serious quality defects from massive amounts of production data, providing a more refined and reliable basis for subsequent decision-making and intervention.

[0053] Through the above technical solution, this application enables refined assessment and prioritization of potential defect events. By assigning an initial confidence level to each potential defect event, the system can distinguish the severity and urgency of different potential defects, avoiding the drawback of treating all potential defects indiscriminately. This not only helps production managers allocate resources more effectively and prioritize high-risk potential defects, thereby improving production efficiency and product quality, but also enables earlier detection and intervention of events that may lead to serious quality problems, significantly reducing losses caused by the failure to detect defects in a timely manner, and enhancing the level of intelligent and lean management throughout the decorative paper production process.

[0054] In some preferred embodiments, a specific example is given below. Assume that during the decorative paper production process, the preset association rules include the following conditional judgment thresholds: the local micro-environment temperature in the coating section should be within the range of 20℃ ± 2℃, the surface roughness of the base paper should be less than 5 micrometers, and the coating thickness should be within the range of 10 micrometers ± 1 micrometer. When the system monitors the following event combination in real time: local micro-environment temperature is 24℃ (deviation 2℃), base paper surface roughness is 6 micrometers (deviation 1 micrometer), and coating thickness is 11.5 micrometers (deviation 0.5 micrometers), the system will calculate the initial confidence level of the potential defect event marker based on these deviation levels. For example, a simple weighted average model or an evaluation model based on fuzzy logic can be set. If a temperature deviation of 2℃ is assigned a confidence contribution of 0.7, a surface roughness deviation of 1 micrometer is assigned a confidence contribution of 0.8, and a coating thickness deviation of 0.5 micrometers is assigned a confidence contribution of 0.6, then the initial confidence level of this event combination may be calculated as 0.75. If the deviation of the other event combination is small, such as a temperature deviation of 1°C, a surface roughness of 4.5 micrometers, and a coating thickness of 10.2 micrometers, its initial confidence level may be calculated as 0.4. In this way, the system can clearly identify that the first event combination has a higher potential defect risk, thereby prompting operators or automated systems to prioritize further analysis or intervention, effectively improving the accuracy and response efficiency of defect warnings.

[0055] In some embodiments described above in this application, by performing real-time correlation analysis on multiple event markers of different types aligned in time and space, and determining the initial confidence level of potential defect event markers based on the deviation of each event parameter in the event marker set from the corresponding conditional judgment threshold, the causal relationship between the event markers and the potential defect event markers can be determined. However, relying solely on the occurrence of process parameter anomalies and event combinations may not be able to completely and accurately determine the true causal relationship with the final coating quality defects, leading to the risk of false positives or false negatives, especially in complex and ever-changing production environments. Without introducing actual quality inspection results for verification, the confidence level of the determined potential defect event markers may be insufficient to support reliable defect diagnosis and process optimization. Therefore, this application further proposes a more accurate method for determining the correlation between event combinations and coating quality defects by matching potential defect event markers with downstream online quality inspection results and introducing a confidence score mechanism to more reliably confirm the causal relationship.

[0056] The above-mentioned determination of the correlation between event combinations and coating quality defects based on potential defect event markers and downstream online quality inspection results includes: matching the spatiotemporal information of the potential defect event markers with the spatiotemporal information of the downstream online quality inspection results; if the matching is successful, increasing the confidence score of the event combination; and confirming a causal relationship between the event combination and the coating quality defects when the confidence score exceeds a preset confidence threshold.

[0057] Specifically, matching the spatiotemporal information of potential defect event markers with the spatiotemporal information of downstream online quality inspection results refers to comparing the timestamps associated with potential defect event markers obtained from upstream process data analysis and their spatial locations on the paper web with the timestamps and corresponding locations on the paper web recorded by downstream online quality inspection equipment (e.g., online thickness gauges, surface defect detectors, etc.). This matching can be achieved by setting a reasonable time window and spatial window. For example, if a potential defect event marker occurs at a certain time point T and paper web position X, then if a corresponding quality defect is detected within the time window T ± Δt and the spatial window X ± Δx, the match is considered successful. The time window Δt and the spatial window Δx can be dynamically adjusted according to the production line speed, the response time of the inspection equipment, and the characteristics of defect propagation.

[0058] If a match is successful, the confidence score of the event combination is increased. This confidence score quantifies the probability or reliability that a particular combination of events leads to a specific coating quality defect. When a potential defect event marker matches an actual detected quality defect in both time and space, it indicates that the event combination is indeed associated with an actual quality problem, and therefore its confidence score should be increased. Confidence score increases can be achieved using various strategies, such as simply adding a fixed value or using a weighted increase based on the accuracy of the match (e.g., the smaller the time-space deviation, the greater the increase).

[0059] When the confidence score exceeds a preset confidence threshold, a causal relationship between the event combination and the coating quality defect is confirmed. The preset confidence threshold is a pre-defined value representing the minimum confidence level required to confirm a causal relationship. When the cumulative confidence score of the event combination reaches or exceeds this threshold, the system can confidently conclude that the event combination is the true cause of the specific coating quality defect. This confirmation process transforms a potential correlation into a clear causal relationship, providing a solid foundation for subsequent process optimization and fault diagnosis.

[0060] This application's solution addresses the inaccuracy of relying solely on process parameter anomalies to determine causality by incorporating downstream online quality inspection results to verify potential defect event markers. Specifically, when upstream process data analysis identifies potential defect event markers, these only represent abnormal combinations of process parameters, and not all anomalies may ultimately lead to product quality defects. By matching the spatiotemporal information of these potential defect event markers with actual downstream online quality inspection results, process anomalies that do not cause actual quality problems can be effectively filtered out, focusing on event combinations that truly affect product quality. The introduction of a confidence score mechanism makes this causal relationship confirmation process cumulative and quantifiable. Each successful spatiotemporal match increases the confidence score, indicating further verification of the association between the event combination and the quality defect. When the confidence score accumulates to a sufficiently high level and exceeds a preset confidence threshold, a high degree of certainty can be established that there is a causal relationship between the event combination and the coating quality defect, thus avoiding the risk of judgment based on a single event or insufficient evidence and significantly improving the accuracy and reliability of defect diagnosis.

[0061] Through the above technical solution, this application can more accurately and reliably identify the true causes of coating quality defects. By combining process anomalies with actual quality results, the false alarm rate is effectively reduced, and overreaction to irrelevant process fluctuations is avoided. Simultaneously, the introduction of a confidence score mechanism makes the causal relationship confirmation process more robust, enabling the gradual accumulation of evidence and ultimately leading to a highly credible diagnostic conclusion. This not only improves the efficiency and accuracy of defect diagnosis but also provides more precise guidance for the optimization and control of the production process, helping to reduce scrap rates and improve product quality stability.

[0062] In some preferred embodiments, it is assumed that during the decorative paper production process, the local microenvironment information of the coating section shows abnormal temperature fluctuations, the microstructure information of the base paper shows that the surface roughness of the base paper exceeds the standard in some areas, and the coating behavior information shows that the coating liquid spreads unevenly. These abnormal events are aligned in time and space and satisfy preset association rules, triggering a potential defect event marker, such as "risk of decreased coating adhesion". At this time, the system will assign an initial confidence level to the potential defect event marker. Subsequently, when the batch of decorative paper enters the downstream online quality inspection stage, the online inspection equipment detects that the adhesion of the decorative paper is indeed lower than the standard value at the time and space location corresponding to the above-mentioned potential defect event marker, and marks it as "adhesion defect". The system will match the time and space information of the potential defect event marker (e.g., occurrence time T1, paper web lateral position X1) with the time and space information of the downstream online quality inspection result (e.g., defect detection time T1+Δt, defect location X1). If the match is successful, the system will increase the confidence score of the event combination "risk of decreased coating adhesion". For example, the initial confidence level is 0.6, which increases to 0.75 after a successful match. As production continues, if similar event combinations occur repeatedly and each time successfully match with adhesion defects detected downstream online, the confidence score of this event combination will accumulate. When this accumulated confidence score exceeds a preset confidence threshold (e.g., 0.85), the system will confirm a clear causal relationship between the event combination of "abnormal local microenvironment temperature fluctuations, excessive base paper surface roughness, and uneven coating liquid spreading" and "decreased adhesion of decorative paper coating." Based on this established causal relationship, operators can then adjust coating temperature control strategies, optimize base paper production processes, or improve coating liquid formulations to fundamentally solve the problem of decreased adhesion.

[0063] In some embodiments described above, by performing real-time correlation analysis on multiple event markers of different types aligned in time and space, and combining this with downstream online quality inspection results, the correlation between the event combination and coating quality defects can be determined, thereby establishing a preliminary causal chain. However, in actual production processes, due to the complexity and variability of decorative paper production processes, the established causal relationship may not always be completely accurate or comprehensive. For example, when a certain defect characteristic is expected based on the identified event combination, the downstream online quality inspection results show a quality defect severity that deviates significantly from the expectation. This indicates that the current causal relationship may have unidentified intermediate links or deeper causes, which may lead to inaccurate or missed diagnoses.

[0064] In response, this application further proposes a method, which further includes: when the severity of quality defects shown in subsequent downstream online quality inspection results deviates significantly from the expected defect characteristics in the causal relationship, determining that there is a missing link in the causal relationship, generating hypotheses about one or more missing links; and dynamically completing the causal relationship based on the missing link hypotheses.

[0065] Specifically, the phrase "a significant deviation between the severity of quality defects shown in the downstream online quality inspection results and the expected defect characteristics in the causal relationship" means that, in the confirmed causal relationship between the event combination and the coating quality defect, the system predicts or anticipates a specific type of defect and its approximate severity based on the parameters of the event combination and a preset model. When the quality defect data actually obtained through downstream online quality inspection differs from this expectation in type, location, severity, or manifestation beyond a preset tolerance range, it is considered a significant deviation. For example, a slight surface roughness might be expected, but a severe decrease in adhesion is actually detected. "Determining a missing link in the causal relationship" means that once the aforementioned significant deviation is detected, the system infers that the currently known causal chain is insufficient to fully explain the observed quality defect, therefore there are one or more intermediate processes or influencing factors that have not been considered, i.e., "missing links." "Generating hypotheses about one or more missing links" means that, for the determined missing links, the system proposes a series of possible explanatory hypotheses based on a preset knowledge base, expert experience, or machine learning models. These hypotheses aim to fill gaps in the causal chain. For example, they might assume the existence of some unmonitored physicochemical change, the intervention of new environmental factors, or abnormal equipment condition. "Dynamically completing the causal relationship based on the hypotheses of missing links" means that the system actively adjusts or expands the original causal relationship model using the generated hypotheses of missing links. This typically involves initiating real-time monitoring of proxy signals related to the hypotheses, or conducting further data analysis and verification to confirm or refute these hypotheses, and ultimately integrating the verified missing links into the causal relationship to make it more complete and accurate.

[0066] This application's solution addresses the limitations of traditional causal models in complex and ever-changing production environments by introducing a self-assessment and dynamic correction mechanism for the integrity of causal relationships. When established causal relationships fail to fully explain observed quality defects, the system no longer simply ignores or misjudges them, but proactively identifies "missing links" in the causal chain. By generating hypotheses about these missing links and conducting further data collection and analysis based on these hypotheses, the system can gradually reveal hidden or unmonitored factors, thereby enabling the causal relationships to be dynamically completed and optimized. This mechanism makes the diagnostic model more adaptable and robust, capable of handling a wider range of defect types and more complex production scenarios.

[0067] Through the above technical solution, this application can significantly improve the accuracy and comprehensiveness of quality defect diagnosis in the decorative paper production process. It not only identifies known causal relationships, but more importantly, it provides a mechanism to discover and understand the underlying causes or intermediate steps leading to quality defects that were initially unidentified. This allows for more refined control and optimization of the production process, reduces production losses caused by misdiagnosis or missed diagnosis, and provides deeper insights for process improvement. Furthermore, this dynamic completion capability enables the diagnostic system to continuously learn and evolve, constantly improving its diagnostic capabilities as the understanding of the production process deepens.

[0068] In some preferred embodiments, it is assumed that a causal relationship has been established in the decorative paper production process: when the coating liquid temperature is consistently high and the coating speed is too fast, a slight "orange peel" defect is expected to appear on the surface of the decorative paper. However, in a certain production run, the system detects the combination of high coating liquid temperature and excessively fast coating speed, but downstream online quality inspection results show that the decorative paper exhibits a severe "adhesion reduction" defect, rather than the expected "orange peel" defect, and the severity is far greater than expected.

[0069] At this point, the system will determine that an existing causal relationship is missing a link. Based on a pre-set knowledge base or expert rules, the system may generate one or more of the following hypotheses regarding the missing link:

[0070] Increased dynamic surface tension of the coating liquid leads to uneven coating or poor wetting.

[0071] The coating and the base paper interface form submicron-level voids, which affect the adhesion.

[0072] The initial drying rate in the drying section was too fast, resulting in excessive internal stress in the coating.

[0073] Subsequently, the system dynamically completes the causal relationship based on these hypotheses. For example, if the hypothesis of "increased dynamic surface tension of the coating liquid" is generated, the system may initiate real-time monitoring of the dynamic surface tension of the coating liquid. If the monitoring data shows that in the same spatiotemporal region where a severe adhesion degradation defect occurs, the dynamic surface tension of the coating liquid does indeed show an abnormal increase, and this increase is highly consistent spatiotemporally with the combination of events such as high coating liquid temperature, excessively fast coating speed, and the final adhesion degradation defect, then the system will integrate the "increased dynamic surface tension of the coating liquid" link into the original causal chain, forming a more complete and accurate diagnostic model: "High coating liquid temperature + excessively fast coating speed + increased dynamic surface tension of the coating liquid → severe adhesion degradation." In this way, the system can discover and confirm new and deeper causes of defects.

[0074] This application further proposes a method for dynamically completing the above causal relationship based on the aforementioned missing link hypothesis, which includes: initiating real-time monitoring of one or more proxy signals corresponding to the missing link hypothesis; the proxy signals are used to indirectly reflect the state of the missing link; based on the proxy signal data obtained from real-time monitoring, the proxy signals are spatiotemporally correlated and compared with the production process events and the quality defects collected in real time; when the change of the proxy signals is consistent with the expected state of the missing link hypothesis and matches the production process events and the quality defects in spatiotemporal terms, the causal relationship is dynamically completed.

[0075] Specifically, "initiating real-time monitoring of one or more proxy signals corresponding to the proposed missing link hypothesis" refers to identifying and continuously acquiring data on physical or chemical parameters that can indirectly reflect the state of the missing link hypothesis. For example, if the missing link hypothesis is an increase in the dynamic surface tension of the coating liquid, the proxy signal could be the temperature, viscosity, or concentration of specific additives in the coating liquid, as changes in these parameters affect the surface tension. Monitoring the proxy signal can be achieved by installing corresponding sensors on the production line or utilizing the data interface of existing equipment, ensuring the real-time nature and accuracy of the data.

[0076] The phrase "based on the proxy signal data obtained from real-time monitoring, perform spatiotemporal correlation and comparison with the real-time collected production process events and the quality defects" can be understood as precisely aligning the timestamp and spatial location information of the proxy signal with previously acquired microenvironmental disturbance event markers, base paper microscopic anomaly event markers, coating behavior anomaly event markers, and quality defect information in downstream online quality inspection results. This alignment aims to identify whether changes in the proxy signal occur simultaneously with or sequentially with known production process anomalies and the final quality defects within the same or closely connected time periods and production line locations.

[0077] In practical applications, "dynamically completing the causal relationship when the change in the proxy signal is consistent with the expected state of the missing link hypothesis and matches the production process event and the quality defect in time and space" means that if the real-time monitoring data of the proxy signal shows a change trend or numerical range that matches the expected change trend or range of the missing link hypothesis, and this change highly matches the production process event that led to the potential defect and the final location of the quality defect in time and space, then the missing link hypothesis is considered to have been verified by data. Based on this, the missing link is formally incorporated into the original causal chain, thereby making the causal relationship more complete and accurate.

[0078] This application's solution effectively addresses the lack of direct verification methods in the aforementioned causal relationship completion process by introducing a real-time monitoring mechanism for proxy signals. Because many intermediate links are difficult to measure directly, indirect proxy signals are needed to reflect their state. By monitoring these proxy signals in real-time and continuously, and rigorously correlating and comparing them with known production process events and quality defects in time and space, strong data support can be provided for the missing link hypothesis. This data-verified dynamic completion method ensures that the establishment of causal relationships no longer relies solely on theoretical inference, but can be corrected and improved based on actual production data, thereby enhancing the accuracy and reliability of causal relationship diagnosis.

[0079] In some preferred embodiments, it is assumed that during the decorative paper production process, downstream online quality inspection results show that the adhesion of the decorative paper decreases by more than a preset threshold, and a missing link hypothesis of "increased dynamic surface tension of the coating liquid" is generated according to the above method. To dynamically complete this causal relationship, real-time monitoring of proxy signals such as coating liquid temperature, viscosity, and coating speed can be initiated. For example, temperature sensors and viscometers can be installed in the coating liquid supply pipeline, and coating speed data can be obtained from the coating machine control system. When real-time monitoring data shows that the coating liquid temperature suddenly increases and the viscosity decreases in a specific time period and coating area, while the coating speed fluctuates, these changes in proxy signals are consistent with the expected state of "increased dynamic surface tension of the coating liquid" (e.g., an increase in temperature may lead to a change in surface tension). Furthermore, if the anomalous changes in these proxy signals match in time and space with the previously identified "microenvironmental disturbance event markers" (e.g., heater failure) and the final "adhesion reduction" quality defect location markers, then the missing link hypothesis of "increased dynamic surface tension of the coating liquid" is confirmed as an effective component of the causal chain, thereby dynamically completing the complete causal relationship from heater failure to increased surface tension of the coating liquid to decreased adhesion.

[0080] In some embodiments described above, this application proposes that when the severity of quality defects shown in subsequent downstream online quality inspection results deviates significantly from the expected defect characteristics in the causal relationship, a missing link in the causal relationship is determined, and hypotheses about one or more missing links are generated. However, in practical applications, if the process of generating hypotheses about missing links is not refined, the scope of the hypotheses may be too broad or inaccurate, thereby affecting the efficiency and accuracy of subsequent causal relationship completion. Therefore, this application further proposes a specific method for generating hypotheses about one or more missing links, aiming to provide a more targeted diagnostic direction for specific quality defects.

[0081] Specifically, the generation of hypotheses about one or more missing links includes: when the quality defect is that the adhesion of the decorative paper decreases by more than a preset threshold, generating at least one of the following hypotheses: the dynamic surface tension of the coating liquid increases; submicron-level voids are formed at the interface between the coating and the base paper.

[0082] The quality defect is defined as a decrease in the adhesion of the decorative paper exceeding a preset threshold. This means that the adhesion data of the decorative paper obtained through downstream online quality testing is compared with the baseline adhesion value under normal production conditions. When the decrease exceeds a preset acceptable range, it is identified as a quality defect due to decreased adhesion. This preset threshold can be set based on product standards, customer requirements, or historical data analysis results.

[0083] When such adhesion degradation defects are identified, the system generates targeted hypotheses about the missing link. Specifically, one hypothesis is an increase in the dynamic surface tension of the coating liquid. The dynamic surface tension of the coating liquid is a key parameter affecting its wetting and spreading performance on the base paper surface. If it increases, the coating liquid may not be able to adequately wet the base paper, thus affecting the physical or chemical bonding between the coating and the base paper, ultimately resulting in decreased adhesion. Another hypothesis is the formation of submicron-level voids at the coating-base paper interface. These tiny voids may form during coating or drying, weakening the effective contact area between the coating and the base paper, reducing the interfacial bonding strength, and consequently leading to decreased adhesion. These hypotheses are based on common adhesion defect mechanisms in decorative paper production and have high specificity and diagnostic value.

[0084] This application's solution achieves intelligent and precise diagnostics by closely linking the generation of missing link hypotheses with specific quality defect types (such as decreased adhesion of decorative paper). When the system detects a specific quality defect, such as a decrease in the adhesion of decorative paper exceeding a preset threshold, it no longer blindly generates general hypotheses. Instead, based on a preset knowledge base or expert experience, it directly generates specific hypotheses highly correlated with the defect, such as increased dynamic surface tension of the coating liquid or the formation of submicron-level voids at the coating-substrate interface. This mechanism allows subsequent proxy signal monitoring and causal relationship completion to focus on the most likely potential causes, thereby avoiding the monitoring and analysis of a large number of irrelevant parameters and significantly improving the efficiency and accuracy of fault diagnosis. By providing these specific, physical mechanism-level hypotheses, this application's solution can reveal potential problems in the production process more deeply, providing more guiding information for process optimization and quality control.

[0085] Through the above technical solution, this application can provide highly focused and accurate hypotheses about missing links for specific coating quality defects, such as decreased adhesion of decorative paper. This significantly narrows the search space for potential causes, avoids ineffective diagnostic paths, and thus greatly improves the efficiency and accuracy of fault diagnosis. Furthermore, since the generated hypotheses are based on specific physical or chemical mechanisms, subsequent causal relationship completion and process optimization are more scientific and feasible, helping to fundamentally solve quality problems in production and improve the stability and reliability of product quality.

[0086] In some preferred embodiments, it is assumed that during the decorative paper production process, the downstream online quality inspection system continuously reports that the adhesion test results of a certain batch of decorative paper are significantly lower than the standard value, and the decrease exceeds a preset threshold. In this case, according to the above method, the system will immediately identify the specific quality defect of "decreased adhesion of decorative paper". Based on this, the system will automatically generate two main hypotheses regarding the missing link: the first hypothesis is "increased dynamic surface tension of the coating liquid," which may lead to poor wetting of the coating liquid on the base paper surface; the second hypothesis is "submicron-level voids form at the interface between the coating and the base paper," which may weaken the bonding strength between the coating and the base paper. These hypotheses can then guide the system to initiate real-time monitoring of relevant proxy signals. For example, for the hypothesis of "increased dynamic surface tension of the coating liquid," the system can initiate monitoring of the online surface tension sensor of the coating liquid; for the hypothesis of "submicron-level voids form at the interface between the coating and the base paper," the system can initiate refined monitoring of parameters such as the temperature and humidity curves of the drying section after coating, as well as the surface roughness of the base paper. In this way, the system can collect data in a targeted manner to verify or rule out these hypotheses, thereby efficiently completing the causal chain that leads to decreased adhesion.

[0087] This application further proposes a step to dynamically complete the above causal relationship based on the above-mentioned missing link hypothesis, including: fusing multi-source information by combining the real-time monitored proxy signal with the above-mentioned event combination and quality defects to generate multi-source fused information.

[0088] Specifically, multi-source information fusion refers to integrating and correlating proxy signal data from different sensors and data sources, combined event data from the production process, and quality defect data obtained from downstream online quality inspection within a unified spatiotemporal coordinate system. For example, data fusion algorithms such as Kalman filtering, Bayesian networks, or deep learning models can be used to process this heterogeneous data to extract more comprehensive and accurate features, thereby forming a comprehensive multi-source fused information. Its purpose is to overcome the limitations of a single data source and improve the accuracy and robustness of judging the state of missing links.

[0089] When the aforementioned multi-source fusion information satisfies the preset matching logic, the causal chain containing the aforementioned missing link hypothesis is confirmed to be valid. The matching logic requires that the changes in the aforementioned proxy signals, the aforementioned event combinations, and the aforementioned quality defects are aligned in both time and space. The matching logic can be understood as a series of predefined rules or models used to determine whether the multi-source fusion information supports a specific missing link hypothesis. Specifically, this requires that the changing trends of the proxy signals, the occurrence patterns of the event combinations, and the location and timing of the quality defects must maintain a high degree of consistency within a preset tolerance range. For example, if the missing link hypothesis is "increased dynamic surface tension of the coating liquid," the matching logic might require that the proxy signal (such as surface tension sensor data) shows an increasing trend, while specific microenvironmental disturbance events (such as coating liquid temperature fluctuations) and abnormal coating behavior events (such as poor coating wetting) occur in the same spatiotemporal region, and a corresponding quality defect (such as decreased adhesion) is detected downstream. The purpose is to ensure that the confirmed causal chain has high spatiotemporal consistency and logical rationality.

[0090] A confidence score is generated for the confirmed causal chain above, and the confidence score is increased each time the preset matching logic is successfully met;

[0091] In practical applications, the confidence score is a quantitative metric used to assess the reliability of a confirmed causal chain. Initially, a base confidence score can be assigned to each newly confirmed causal chain. Each time a new real-time data stream again satisfies the aforementioned matching logic and further supports the causal chain, the confidence score is increased accordingly. For example, this can be achieved through accumulation, weighted averaging, or methods based on statistical models. The aim is to gradually increase confidence in the causal chain through continuous validation and data accumulation, avoiding hasty judgments based on a single event or short-term observations.

[0092] When the confidence score reaches the preset confirmation threshold, the causal chain is marked as a reliable diagnostic basis, and the above causal relationship is completed.

[0093] The preset confirmation threshold is a pre-defined value representing the minimum confidence level required to confirm a causal chain. When the confidence score of a causal chain reaches or exceeds this threshold, it indicates that the causal chain has been sufficiently verified and can be considered a reliable diagnostic basis. At this point, the causal chain will be formally incorporated into the knowledge base to guide subsequent production process optimization and troubleshooting, ultimately completing the aforementioned causal relationship. The purpose is to establish a rigorous confirmation mechanism to ensure that only fully verified causal relationships are adopted, thereby improving the accuracy and practicality of the entire diagnostic system.

[0094] This application's solution effectively addresses the problem of inaccurate or unreliable causal chain confirmation in complex industrial environments, where relying solely on simple spatiotemporal correlations may lead to unreliable causal chain confirmation. Specifically, multi-source information fusion integrates data from different dimensions and granularities, providing a more comprehensive insight into missing links and reducing information blind spots. Building upon this, the pre-defined matching logic ensures high internal consistency of identified causal relationships through strict temporal and spatial alignment requirements for changes in proxy signals, event combinations, and quality defects. Furthermore, by generating and dynamically increasing a confidence score for each causal chain, this application's solution accumulates confidence in causal relationships through continuous data verification, avoiding the randomness that may arise from a single match. Only when the confidence score reaches a pre-defined confirmation threshold is the causal chain marked as a reliable diagnostic basis, providing a quantifiable and credible standard for the final confirmation of causal relationships, thereby significantly improving the accuracy and reliability of dynamically completing causal relationships.

[0095] Through the above technical solutions, this application enables more accurate and reliable confirmation of the causal chain of missing links. The application of multi-source information fusion allows the system to examine problems from a broader data perspective, improving its ability to identify potential causal relationships. The introduction of rigorous matching logic and confidence scores provides a quantitative and dynamic evaluation standard for confirming the causal chain, effectively avoiding misjudgments and enhancing the credibility of diagnostic results. Therefore, this application can more accurately identify the underlying causes of coating quality defects, providing a more solid and reliable basis for production process optimization and troubleshooting, thereby significantly improving the intelligent management level and product quality stability of the decorative paper production process.

[0096] In some preferred embodiments, it is assumed that during the decorative paper production process, downstream online quality inspection results show that the adhesion of the decorative paper decreases by more than a preset threshold, and according to the above method, a missing link hypothesis regarding "increased dynamic surface tension of the coating liquid" has been generated. In order to dynamically complete this causal relationship, the system will initiate real-time monitoring of the dynamic surface tension of the coating liquid (as a proxy signal).

[0097] Specifically, the system fuses multi-source information, including real-time monitored dynamic surface tension data of the coating liquid, previously identified combinations of events related to decreased adhesion (e.g., coating liquid temperature fluctuations, abnormal coating blade pressure), and adhesion-related defect data detected by downstream online quality control. For example, this data is input into a fusion model that analyzes their spatiotemporal correlation. The system then determines whether this multi-source fusion information satisfies a preset matching logic. This matching logic might require that the dynamic surface tension of the coating liquid continuously increases within a specific time period, and that the time and location of this increase are highly aligned spatiotemporally with coating liquid temperature fluctuations and abnormal coating blade pressure events, ultimately resulting in the detection of adhesion-related quality defects in the same paper web area. If the matching logic is met, the system generates an initial confidence score for the causal chain of "increased dynamic surface tension of the coating liquid leading to decreased adhesion." Subsequently, each time a new production batch or a new real-time data stream satisfies the same matching logic again, the confidence score of this causal chain is increased. For example, if the same pattern is observed and the matching logic is met in ten consecutive production runs, the confidence score will continue to accumulate. Eventually, when the confidence score of this causal chain reaches a preset confirmation threshold (e.g., 0.95), the system will formally confirm that "the increase in the dynamic surface tension of the coating liquid is a reliable cause of the decrease in the adhesion of the decorative paper," and mark it as a reliable diagnostic basis, thus successfully completing this causal relationship. Subsequently, when the adhesion decrease problem occurs again, the system can prioritize diagnosing and recommend checking the dynamic surface tension of the coating liquid, and take corresponding process adjustment measures.

[0098] In some of the embodiments described above in this application, methods are proposed to identify and confirm the causal relationships that lead to coating quality defects during the production process, and even to dynamically fill in missing links. However, in actual production environments, if these complex causal chains are presented only in the form of data or text, they may be difficult for operators or engineers to understand and apply quickly and intuitively, thus affecting the efficiency and timeliness of fault diagnosis.

[0099] In this regard, this application further proposes that the above method also includes: generating and outputting a visualization interface based on the above causal relationship; the visualization interface is used to display the sequence of related events that lead to the coating quality defects in chronological order and spatial location; in a web-based interactive dashboard, macroscopic process parameter curves, micro-environmental disturbance events, base paper micro-abnormal events, coating behavior abnormal events and the location markers of the final quality defects are displayed simultaneously in the form of a combination of time axis and paper web cross-sectional view.

[0100] Specifically, the visualization interface refers to a graphical user interface designed to present complex production data, event markers, and ultimately determined causal relationships to users in an intuitive and easy-to-understand manner. This interface is configured to display the sequence of related events leading to coating quality defects in chronological order and spatial location. This means that when a coating quality defect is identified, related microenvironmental disturbance events, base paper micro-anomalies, coating behavior anomalies, etc., will be arranged and displayed according to their chronological order and spatial location on the production line, thus clearly revealing the full picture of the event's development.

[0101] The visualization interface can be implemented as a web-based interactive dashboard. This dashboard allows users to access and operate it via a web browser, providing excellent accessibility and flexibility. In the dashboard, information is displayed synchronously in a combination of a timeline and a cross-sectional view of the paper web. The timeline shows the chronological sequence of events, while the cross-sectional view indicates the specific location of events or defects along the width of the paper web. This combination allows users to simultaneously observe the temporal and spatial distribution of events.

[0102] Furthermore, the dashboard is designed to simultaneously display macroscopic process parameter curves, micro-environmental disturbance events, base paper microscopic anomalies, coating behavior anomalies, and the location markers of final quality defects. Macroscopic process parameter curves may include coating speed, drying temperature, coating amount, etc., and the trends of these parameters can be compared and analyzed with anomalies and defects. Micro-environmental disturbance events, base paper microscopic anomalies, and coating behavior anomalies are marked with specific icons or colors on the time axis and paper web cross-sectional view, their location markers precisely corresponding to their time and spatial coordinates. The location markers of final quality defects clearly indicate the specific location of the defect on the finished paper, for example, by using color coding or specific symbols to indicate the defect type and severity.

[0103] This application's solution effectively addresses the problem of traditional data analysis results being difficult to quickly understand and apply by introducing a visual interface. Once complex causal chains are identified, they are transformed into intuitive graphical displays, allowing operators and engineers to clearly see all relevant events leading to specific coating quality defects, as well as the spatiotemporal evolution of these events during the production process. The timeline allows tracking the sequence and duration of events; the cross-sectional view of the paper web pinpoints the specific locations of defects and related events across the paper web width. This multi-dimensional, synchronous display enables users to quickly identify clusters of abnormal events or key time points, thereby rapidly pinpointing potential root causes. For example, when abnormal coating behavior events in a certain area are observed to highly overlap with microscopic anomalies in the base paper in time and space, and a quality defect is subsequently detected in that area, combined with macroscopic process parameter curves, it can be quickly inferred that the defect is caused by the interaction between the base paper characteristics and coating behavior, and may be related to fluctuations in a certain process parameter.

[0104] In some preferred embodiments, it is assumed that during the decorative paper production process, downstream online quality inspection results show a series of quality defects with decreased adhesion on the left edge of the paper web. Based on the diagnostic method of this application, the system has determined the causal relationship leading to the defect. At this point, the visualization interface will play its role. In a web-based interactive dashboard, the user will see a timeline that marks microenvironmental disturbance events (e.g., airflow fluctuations on the left side of the coating head), microscopic anomalies in the base paper (e.g., abnormally high porosity on the left edge of the base paper), and coating behavior anomalies (e.g., decreased wettability of the coating liquid on the left edge) detected in the left edge region of the paper web some time before the defect occurred. These events will be clearly displayed on the timeline with different colors or icons indicating their occurrence time. Simultaneously, on the cross-sectional view of the paper web, the locations of these events and the final adhesion-decreased defect will be precisely aligned with the left edge region of the paper web. In addition, the dashboard will also synchronously display relevant macroscopic process parameter curves, such as slight fluctuations in coating speed or drying temperature during the event. This intuitive display allows operators to quickly identify how, within a specific timeframe, abnormal base paper characteristics at the left edge of the paper web, combined with localized microenvironmental disturbances, led to decreased wettability of the coating solution, ultimately resulting in adhesion defects. This visual presentation transforms the diagnostic process from complex numerical and textual analysis into intuitive graphical recognition, significantly improving diagnostic efficiency and accuracy.

[0105] Secondly, this application also discloses a decorative paper production data processing system based on the Industrial Internet, such as... Figure 2 As shown, the system includes:

[0106] The acquisition module 201 is used to acquire local microenvironment information from the coating section, paper web microstructure information from the pre-coating section, and coating behavior information from the post-coating drying section.

[0107] The alignment and generation module 202 is used to align the local microenvironment information, paper web microstructure information, coating behavior information with the production line spatiotemporal coordinate system, and generate corresponding event markers based on preset dynamic thresholds; the event markers include microenvironment disturbance event markers, paper web microstructure abnormal event markers, and coating behavior abnormal event markers;

[0108] The correlation analysis module 203 is used to perform real-time correlation analysis on multiple event tags of different types that are aligned in time and space, in order to identify event combinations that meet preset correlation rules and trigger potential defect event tags.

[0109] The determination module 204 obtains the downstream online quality inspection results and determines the correlation between the event combination and coating quality defects based on the paper web position information, the potential defect event markers, and the downstream online quality inspection results.

[0110] The decorative paper production data processing system based on the Industrial Internet provided in this application can perform the method described in the first aspect.

[0111] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for processing decorative paper production data based on the Industrial Internet, characterized in that, The method includes: Acquire local microenvironment information from the coating section, paper web microstructure information from the pre-coating section, and coating behavior information from the post-coating drying section; The local microenvironment information, paper web microstructure information, and coating behavior information are aligned with the production line spatiotemporal coordinate system, and corresponding event markers are generated based on preset dynamic thresholds; the event markers include microenvironment disturbance event markers, paper web microstructure abnormal event markers, and coating behavior abnormal event markers. Real-time correlation analysis is performed on multiple event tags of different types that are aligned in time and space to identify event combinations that meet preset correlation rules and trigger potential defect event tags; The downstream online quality inspection results are obtained, and the correlation between the event combination and coating quality defects is determined based on the paper web position information, the potential defect event markers, and the downstream online quality inspection results; the paper web position information includes longitudinal position and transverse position.

2. The method for processing decorative paper production data based on the Industrial Internet according to claim 1, characterized in that, The real-time correlation analysis of multiple spatiotemporally aligned event markers of different types to identify event combinations that meet preset correlation rules and trigger potential defect event markers includes: Assign a spatiotemporal identifier based on timestamps and spatial coordinates to each event marker; Based on the spatiotemporal identifier, a set of event markers of different types that are adjacent in spatiotemporal space are aggregated using a preset sliding time window and spatial window. Determine whether the event tag set satisfies a preset association rule; the preset association rule includes a logical combination consisting of abnormal conditions of local microenvironment information, abnormal conditions of paper web micro-characteristic information, and abnormal conditions of coating behavior information; the preset association rule includes one or more types of logical combinations. In response to the fact that all events in the event tag set satisfy the preset association rules, it is determined that the event combination corresponding to the event tag set has a potential defect, and the potential defect event tag is triggered.

3. The method for processing decorative paper production data based on the Industrial Internet according to claim 2, characterized in that, The preset association rules include condition judgment thresholds corresponding to abnormal local microenvironment information, abnormal paper web microstructure information, and abnormal coating behavior information; determining that the event combination corresponding to the event tag set has a potential defect and triggering the potential defect event tag also includes: The initial confidence level of the potential defect event marker is determined based on the degree of deviation between each event parameter in the event marker set and the corresponding condition judgment threshold.

4. The decorative paper production data processing method based on the Industrial Internet according to claim 3, characterized in that, Based on the paper web location information, the potential defect event markers, and the downstream online quality inspection results, the correlation between the event combination and coating quality defects is determined, including: The spatiotemporal information of the potential defect event markers is matched with the spatiotemporal information of the downstream online quality inspection results; if the match is successful, the confidence score of the event combination is increased. When the confidence score exceeds a preset confidence threshold, a causal relationship is confirmed between the event combination and the coating quality defect.

5. The method for processing decorative paper production data based on the Industrial Internet according to claim 4, characterized in that, The method further includes: When the severity of quality defects shown in subsequent downstream online quality inspection results deviates significantly from the expected defect characteristics in the causal relationship, it is determined that there is a missing link in the causal relationship, and hypotheses about one or more missing links are generated. The causal relationship is dynamically completed based on the assumption of the missing link.

6. The method for processing decorative paper production data based on the Industrial Internet according to claim 5, characterized in that, Based on the assumption of missing links, the causal relationship is dynamically completed, including: Real-time monitoring of one or more proxy signals corresponding to the missing link hypothesis is initiated; the proxy signals are used to indirectly reflect the status of the missing link. Based on the proxy signal obtained from real-time monitoring, it is spatiotemporally correlated and compared with the real-time collected production process events and the quality defects. When the change in the proxy signal is consistent with the expected state of the missing link hypothesis, and matches the production process event and the quality defect in time and space, the causal relationship is dynamically completed.

7. The method for processing decorative paper production data based on the Industrial Internet according to claim 5, characterized in that, The generation of hypotheses about one or more missing links includes: when the quality defect is that the adhesion of the decorative paper decreases by more than a preset threshold, generating at least one of the following hypotheses: the dynamic surface tension of the coating liquid increases; submicron-level voids are formed at the interface between the coating and the paper web.

8. The method for processing decorative paper production data based on the Industrial Internet according to claim 6, characterized in that, Based on the assumption of missing links, the causal relationship is dynamically completed, including: The real-time monitored agent signals are combined with the event combination and quality defects to generate multi-source fused information. When the multi-source fusion information satisfies the preset matching logic, the causal chain containing the assumption of the missing link is confirmed to be valid; A confidence score is generated for the confirmed causal chain, and the confidence score is increased each time a preset matching logic is successfully met; When the confidence score reaches the preset confirmation threshold, the causal chain is marked as a reliable diagnostic basis, and the causal relationship is completed.

9. The method for processing decorative paper production data based on the Industrial Internet according to any one of claims 4-8, characterized in that, The method further includes: Based on the causal relationship, a visualization interface is generated and output; the visualization interface is used to display the sequence of related events that lead to the coating quality defects in chronological order and spatial location. The web-based interactive dashboard displays macroscopic process parameter curves, micro-environmental disturbance events, paper web micro-anomaly events, coating behavior anomaly events, and the location markers of final quality defects in a combined timeline and paper web cross-sectional view.

10. A decorative paper production data processing system based on the Industrial Internet, characterized in that, The system includes: The acquisition module is used to acquire local microenvironment information from the coating section, paper web microstructure information from the pre-coating section, and coating behavior information from the post-coating drying section. The alignment and generation module is used to align the local microenvironment information, paper web microstructure information, coating behavior information with the production line spatiotemporal coordinate system, and generate corresponding event markers based on preset dynamic thresholds; the event markers include microenvironment disturbance event markers, paper web microstructure abnormal event markers, and coating behavior abnormal event markers; The correlation analysis module is used to perform real-time correlation analysis on multiple event tags of different types that are aligned in time and space to identify event combinations that meet preset correlation rules and trigger potential defect event tags. The module determines the correlation between the event combination and coating quality defects based on the paper web position information, the potential defect event markers, and the downstream online quality inspection results; the paper web position information includes longitudinal position and transverse position.

Images