A Knowledge Graph-Based Quality Traceability Method and System for Coating Process Chain

Abstract

This application discloses a method and system for quality traceability of the coating process chain based on knowledge graphs, belonging to the field of big data processing technology for quality traceability. The specific implementation scheme is as follows: Dynamic process parameter curve feature analysis is performed to determine whether to adopt time-series curve feature optimization and coating timestamp consistency calibration; temporal causal determination of the coating graph is performed to determine whether to trigger graph temporal constraint supplementation and correlation completion; quality traceability reasoning for the coating process is performed; and after the reasoning is completed, the temporal logic qualification of the coating traceability path is verified. If the verification is qualified, the final coating quality defect traceability result is output; if the verification is unqualified, a notification of unqualified coating quality defect traceability result is sent. This helps to solve the problem of low accuracy of coating process chain quality traceability results.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology for quality traceability, and in particular to a method and system for quality traceability of the coating process chain based on knowledge graphs. Background Technology

[0002] In packaging, printing, and electronic component encapsulation, the lamination process is a crucial step in ensuring product surface performance (such as abrasion resistance and light transmittance). Its quality stability directly determines the final product's pass rate. However, traditional lamination process quality traceability requires manual cross-system verification, which is extremely inefficient. Knowledge graphs, on the other hand, can integrate multi-source data in a structured manner, clearly presenting the relationships between various production elements and providing reliable data and logical support for traceability. The main implementation process of existing technologies in the lamination process chain quality traceability involves: based on the operational data records of the entire lamination production process, obtaining structured data (such as operating parameters) of lamination equipment (such as automatic laminators), and lamination processing... Unstructured data from the equipment (such as maintenance reports and on-site records from quality inspectors) is used to generate an ontology framework, which is then combined with the domain rules of the coating process to clarify the types and association rules of core entities (such as raw materials and quality defects). Next, knowledge extraction is performed on both types of data (including structured and unstructured data) (such as directly extracting entities, attributes, and association information from structured data). Subsequently, the information extracted from the two types of data is fused using semantic similarity calculation methods (such as cosine similarity algorithms) to remove redundant information, align heterogeneous data, and generate triple data (such as entity-relationship-entity). Based on the triples, a knowledge graph for quality traceability of the coating process chain is constructed.

[0003] For example, the IoT-based lithium battery separator production quality traceability method and system with publication number CN121073266A includes: collecting historical and real-time separator production information; identifying key nodes in lithium battery separator production and assigning a unique quality and safety traceability code to each key node; constructing a quality anomaly knowledge graph by combining historical separator production information and quality and safety traceability codes; inputting real-time separator production information into the quality anomaly knowledge graph to obtain quality anomaly parameters; verifying the credibility of the quality anomaly parameters; determining the causes of quality anomalies using a root cause analysis model and generating compensation parameters based on the causes of quality anomalies; compensating the quality anomaly parameters based on the compensation parameters and performing quality traceability in the quality anomaly knowledge graph based on the compensated quality anomaly parameters.

[0004] During the coating quality inspection process, when a quality defect is detected in the finished coated product, the defect is defined as the current coating quality event. A community detection algorithm, such as the Louvain algorithm, is used to match the coating process chain quality traceability knowledge graph to identify the range of nodes directly or indirectly related to the current coating quality event (covering nodes related to equipment, materials, and operation records in the corresponding process). Then, real-time data from the identified range of related nodes is input into a pre-defined classification model (such as a random forest) to output the core correlation factors of the current coating quality defect. The process involves identifying key factors (such as excessive parameters of specific equipment or insufficient purity of a batch of materials). Using the knowledge graph node corresponding to this core factor as the starting point for reasoning, and the source of the coating process (such as raw material procurement or initial process settings) as the reasoning guide, reverse reasoning is performed in the knowledge graph (e.g., along the reasoning link of factor-process-source) to locate the root cause of the current coating quality defect. Finally, the traced root cause information and its impact range are integrated to generate a coating process association result containing node relationships and tracing paths (e.g., raw material procurement-adhesive warehousing-adhesive use-coating process-pressing defects).

[0005] The above-mentioned technology has at least the following technical problems:

[0006] In the process of quality traceability in the production chain, existing technologies assign traceability codes to key production nodes and construct an anomaly knowledge graph. After verifying the credibility of detected abnormal parameters and performing root cause analysis, compensation parameters are generated. Finally, quality traceability is completed in the knowledge graph based on the compensated parameters.

[0007] In the process of quality traceability in the lamination process chain, there is a complex nonlinear and temporal correlation between core lamination quality indicators (such as adhesive strength and light transmittance uniformity) and dynamic process parameters (such as lamination temperature, pressing pressure, film conveying speed, and adhesive coating amount). Existing technologies, when introducing continuous process parameters into a knowledge graph, only record instantaneous single-point values, failing to retain the shape characteristics of the curves (slope, peak value, phase, duration) and their sequential relationships. This results in the graph being unable to express the temporal causal relationship between the evolution of dynamic parameters and quality defects, potentially generating false traceability paths, and consequently affecting the lamination quality inspection process. When the community discovery algorithm initially locates the range of associated nodes, the lack of precise temporal constraints and causal edges in the knowledge graph of the coating process chain quality traceability may lead to overly broad identified associated subgraphs or omission of key abnormal nodes. Furthermore, when reasoning about the core associated factors of coating quality defects based on inaccurate associated subgraph data, the credibility of the output core associated factors is reduced. Ultimately, when performing reverse reasoning along the knowledge graph of coating process chain quality traceability, multiple false traceability paths that are logically reasonable in static terms but violate the actual production timeline may be generated, resulting in low accuracy of coating process chain quality traceability results. Summary of the Invention

[0008] This invention provides a knowledge graph-based method and system for quality traceability in the coating process chain, which can improve the accuracy of traceability results. The technical solution provided in this application is as follows:

[0009] Firstly, a knowledge graph-based method for tracing the quality of the coating process chain is provided. The specific implementation of this method is as follows: Dynamic process parameter curve feature analysis is performed to determine whether to implement time-series curve feature optimization and coating timestamp consistency calibration. If implemented, the analysis is checked for compliance after implementation. If compliance is achieved, a time-series causal determination of the coating graph is performed; if compliance is achieved, a failure notification is sent. If compliance is not achieved, a time-series causal determination of the coating graph is performed to determine whether to trigger graph time-series constraint supplementation and relational completion. If triggered, the time-series causal determination of the coating graph is checked for compliance after triggering. If compliance is achieved, coating process quality tracing reasoning is performed; if compliance is achieved, a graph completion failure notification is sent. If compliance is not triggered, coating process quality tracing reasoning is performed, and after the coating process quality tracing reasoning is completed, the compliance of the coating traceability path time-series logic is checked. If the check is successful, the final coating quality defect tracing result is output; if the check is unsuccessful, a coating quality defect tracing result failure notification is sent.

[0010] Secondly, a knowledge graph-based quality traceability system for the coating process chain is provided, including: a dynamic process parameter feature monitoring module for coating, a coating graph temporal causality monitoring module, and a coating quality traceability reasoning and source path verification module. The dynamic process parameter feature monitoring module analyzes the curve features of the dynamic process parameters to determine whether to perform temporal curve feature optimization and coating timestamp consistency calibration. If so, after the optimization, it checks whether the dynamic process parameter curve feature analysis is qualified. If qualified, it performs temporal causality determination of the coating graph; if unqualified, it sends a failure notification. The temporal causality monitoring module is used to determine whether to perform optimization or coating timestamp consistency calibration. The coating pattern timing causality determination is then performed to determine whether to trigger the timing constraint supplementation and correlation completion of the pattern. If triggered, after the trigger ends, it is determined whether the coating pattern timing causality determination is qualified. If qualified, the coating process quality traceability reasoning is performed. If unqualified, a pattern completion failure prompt is sent. The coating quality traceability reasoning and traceability path verification module is used to perform coating process quality traceability reasoning if not triggered, and to perform coating traceability path timing logic qualification verification after the coating process quality traceability reasoning is completed. If the verification is qualified, the final coating quality defect traceability result is output. If the verification is unqualified, a coating quality defect traceability result unqualified prompt is sent.

[0011] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0012] 1. By performing dynamic process parameter curve feature analysis for coating, it is possible to determine whether to implement time-series curve feature optimization and coating timestamp consistency calibration. This helps address the technical problem of existing coating quality traceability schemes failing to pre-verify the validity and temporal correlation of dynamic process parameter features, leading to distorted results in subsequent analyses. This ensures high integrity and temporal accuracy of coating data sources in subsequent analysis stages, improving the reliability of coating quality traceability. If not implemented, temporal causality determination of the coating spectrum is performed to determine whether to trigger spectrum temporal constraint supplementation and correlation completion. This helps address the issue of existing schemes directly using fixed coating spectra for causal determination, which is not adapted to the current production batch. The characteristics of the process data lead to the technical problem of poor adaptability of causal judgment. To improve the accuracy of causal judgment, the knowledge graph makes the results of causal judgment more consistent with the actual coating process fluctuations of the current production batch. If no trigger is triggered, the coating process quality traceability reasoning is performed. After the coating process quality traceability reasoning is completed, the temporal logic qualification of the coating traceability path is verified. This helps to solve the technical problem that the existing solution directly outputs the traceability result after completing the traceability reasoning without verifying the temporal rationality of the traceability path. This may result in the contradiction between the time sequence of the traceability path and the production process, leading to the unreliability of the traceability result. This provides the temporal logic level credibility guarantee for the traceability result of coating quality defects, ensuring the rigor of the traceability result.

[0013] 2. By acquiring dynamic coating process parameters including coating temperature, coating pressure, coating conveying speed, and adhesive coating amount sequences through targeted membrane pressure testing, this method helps address the shortcomings of existing coating process parameter acquisition methods that only select a single parameter (such as coating temperature). These four types of parameters correspond to four core process stages: coating heating and bonding, bonding tightness, conveying stability, and adhesive medium uniformity, comprehensively covering the key dimensions affecting coating quality and reducing blind spots in tracing quality defects caused by incomplete parameter acquisition. Based on the dynamic coating process parameters, the method extracts the curve shape features of the coating process parameters, including slope, peak, phase, and duration characteristics. This helps overcome the shortcomings of existing technologies that only extract instantaneous absolute values ​​or simple statistical values ​​(such as average values) of process parameters, transforming the continuous dynamic coating process parameter sequence into quantifiable features that reflect the parameter change patterns, thus capturing the dynamic coating process. The correlation signal between process fluctuations and coating core indicator quality defects is used to determine whether the completeness rate of time-series curve shape feature extraction is greater than the preset time-series feature completeness threshold. If so, coating feature time-series matching discrimination is performed; otherwise, time-series curve feature optimization is adopted. This helps to solve the shortcomings of existing technologies that do not verify the completeness of feature extraction and directly use invalid data for analysis. Time-series curve feature optimization can improve the completeness and fidelity of feature extraction. Coating feature time-series matching discrimination means determining whether the coating feature time-series matching compliance rate is greater than the preset feature time-series matching threshold. If so, coating pattern time-series causal determination is performed; otherwise, coating feature time-series cumulative discrimination is performed. This helps to solve the shortcomings of existing technologies that only judge time-series matching once and are prone to misjudging occasional small time deviations in production (such as equipment instantaneous delays) as time-series correlation failures, improve the accuracy of time-series correlation, and ensure the rigor of time-series logic in causal determination.

[0014] 3. When the traceability path covers more than the preset production subunit threshold for the coating production subunit entity nodes, and the timestamp references of different coating production subunit entity nodes have time deviations within the allowable range of timestamp completion (such as deviations in the standard timestamp references of each coating production subunit, which are within the fault tolerance range of timestamp completion for the coating production subunit), because Implementation 1 relies on a unified single timestamp reference to complete the full node traversal verification, the small timestamp reference deviations across coating production subunits will be amplified by the full node traversal, leading to misjudgments in the temporal logic qualification verification of the coating traceability path. Therefore, it is necessary to execute the second scheme for temporal logic qualification verification of the coating traceability path. From the traceability path of coating quality defects, the core temporal link nodes that ensure the integrity of the coating production process are extracted. The core temporal link nodes need to cover key... In this step, the time information corresponding to the extracted core timing link nodes is obtained, and the time sequence of each core timing link node is checked to see if it meets the preset core timing requirements. This helps to solve the misjudgment problem caused by amplifying small time deviations across sub-units due to full node traversal in Implementation 1, while reducing the computational load of full node traversal and improving verification efficiency. The number of node pairs in the core timing link that do not meet the preset core timing requirements is counted. If the number of node pairs that do not meet the preset core timing requirements is 0, the final coating quality defect tracing result is output based on the preset process rules and correlation relationships. Otherwise, a coating quality defect tracing result failure prompt is sent. This helps to solve the shortcomings of the existing technology in that the timing verification results are vague and lack clear judgment standards, ensuring that the coating quality tracing result has rigorous timing logic support and guaranteeing the reliability of coating quality traceability. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart summarizing the knowledge graph-based coating process chain quality traceability method provided in this embodiment of the invention.

[0017] Figure 2 This is a schematic diagram of the structure of the knowledge graph-based coating process chain quality traceability system provided in an embodiment of the present invention;

[0018] Figure 3 This invention provides a knowledge graph for tracing the quality of a coating process chain based on a knowledge graph, as described in this embodiment.

[0019] Figure 4This is an analysis diagram of the core correlation factors of coating quality defects based on the random forest model in the coating process chain quality traceability method based on knowledge graph provided in the embodiments of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0021] Embodiment 1 of this invention provides a knowledge graph-based method for quality traceability of the coating process chain. This method monitors the characteristics of dynamic process parameters during the coating process chain traceability. It performs dynamic process parameter curve feature analysis to capture the temporal fluctuation patterns of dynamic process parameters and verify the temporal matching effectiveness between the characteristics of dynamic process parameters and core quality indicators of the coating process. This analysis determines whether to implement temporal curve feature optimization and coating timestamp consistency calibration. If implemented, after completion, the method checks whether the dynamic process parameter curve feature analysis is qualified. If qualified, it performs a temporal causal determination of the coating graph to quantify the temporal causal correspondence between coating process parameter fluctuations and coating quality defects. If unqualified, it sends a failure alert. The temporal curve feature optimization is used to address issues in the coating process. To address the issue of insufficient completeness in extracting shape features from parametric curves, this paper ensures that the shape features (such as slope, peak value, phase, and duration) extracted from the dynamic coating process parameters possess high completeness and high fidelity. Coating timestamp consistency calibration is used to unify the time base of various data acquisition devices for coating, correcting timestamp deviations between the dynamic coating process parameters and core coating quality indicators, and ensuring the effectiveness of the temporal correlation between these parameters. Monitoring the characteristics of the dynamic coating process parameters helps control data quality from the source of the coating traceability process, reducing subsequent analysis distortions caused by missing dynamic coating process parameter features and temporal deviations. This provides highly reliable quantitative data support for the causal correlation analysis between coating quality defects and dynamic coating process parameters, reducing the cost of ineffective traceability.

[0022] If no action is taken, a temporal causal monitoring of the coating graph is performed to determine whether to trigger the supplementation of temporal constraints and the completion of correlation relationships. If triggered, the temporal causal monitoring of the coating graph is evaluated for compliance. If compliant, quality traceability reasoning for the coating process is performed to extract the root causes and propagation paths of coating quality defects. If non-compliant, a graph completion failure notification is sent. The supplementation of temporal constraints and the completion of correlation relationships are used to supplement missing temporal logical constraints in the coating graph and correct deviations in the correlation between dynamic coating process parameters and core coating quality indicators. Through temporal causal monitoring of the coating graph, the temporal causal correspondence between fluctuations in dynamic coating process parameters and quality defects can be accurately quantified, potential correlation deviations and logical deficiencies in the knowledge graph can be corrected, and the causal determination results can be made more consistent with actual production. This provides accurate correlation basis for subsequent quality traceability reasoning for coating processes and improves the accuracy of locating the root causes of coating quality defects.

[0023] If the coating quality traceability reasoning and tracing path verification are not triggered, the coating process quality traceability reasoning is performed. After the coating process quality traceability reasoning is completed, the temporal logic compliance of the coating quality defect tracing path is verified to ensure that the time sequence of the tracing path conforms to the actual coating production process. If the verification is successful, the final coating quality defect tracing result is output. If the verification fails, a failure message is sent. Through coating quality traceability reasoning and tracing path verification, it is helpful to trace the root cause, responsible link and abnormal propagation path of coating quality defects. At the same time, the temporal logic verification reduces the temporal sequence contradictions of the tracing path, ensuring that the final output tracing result has high accuracy.

[0024] Those skilled in the art should understand that the knowledge graph-based coating process chain quality traceability method provided in this application requires the construction and maintenance of a coating process knowledge and data fusion library to support the stable operation of the traceability system before implementation. The content of this fusion library comes from diverse sources, including basic setting parameters directly defined by industry experts in combination with coating process mechanisms, equipment operating characteristics, and quality control requirements, such as preset process thresholds and preset feature data volume thresholds. It also integrates reference datasets from historical production practices and quality verification, such as time-series samples of coating dynamic process parameters under different production batches, traceability path cases of typical quality defects, and correlation mapping data between process features and quality indicators. These data provide practical support for the rationality and applicability of static setting parameters. In terms of data management architecture, a hybrid storage scheme can be adopted, using a relational database to store structured setting parameters, node relationships, and quality inspection results, while using non-relational databases and time-series databases to efficiently store and quickly retrieve massive amounts of unstructured and semi-structured data such as coating process time-series data, knowledge graph node data, and feature extraction results.

[0025] As described above, monitoring of dynamic process parameters for coating, monitoring of coating patterns in a time-series causal manner, and verification of coating quality traceability reasoning and tracing paths help to build a system from coating data quality control to coating causal relationship determination. This ultimately solves the technical problems of insufficient data validity, ambiguous causal relationships, and low credibility of traceability results in existing traceability schemes. Specifically, monitoring of dynamic process parameters for coating provides high-quality, time-series consistent basic data for subsequent stages; monitoring of coating patterns in a time-series causal manner receives effective data from the front end and establishes accurate causal relationships, providing a clear analytical direction for traceability reasoning; and verification of coating quality traceability reasoning and tracing paths is the final guarantee, outputting traceability results based on accurate causal relationships and verifying their compliance. These three aspects are progressive and mutually supportive, ensuring that the coating process chain quality traceability system continuously adapts to the dynamic changes in production processes.

[0026] like Figure 1 The flowchart shown is an overview of the knowledge graph-based coating process chain quality traceability method. Figure 1 It is known that: Dynamic process parameter curve feature analysis is performed on the coating process, and the completeness rate of time-series curve shape feature extraction and the compliance rate of coating feature time-series matching are obtained. It is determined whether the completeness rate of time-series curve shape feature extraction is greater than the preset time-series feature completeness threshold. If so, coating feature time-series matching discrimination is performed; otherwise, time-series curve feature optimization is adopted. Coating feature time-series matching discrimination means determining whether the compliance rate of coating feature time-series matching is greater than the preset feature time-series matching threshold. If not, coating feature time-series cumulative discrimination is performed; otherwise, coating spectrum time-series causal determination is performed, and coating spectrum quality assessment parameters are obtained. It is determined whether the coating spectrum quality assessment parameters meet the coating spectrum quality adaptation conditions. If not, spectrum time-series constraint supplementation and correlation completion are triggered; otherwise, coating process quality traceability reasoning is performed. After the coating process quality traceability reasoning is completed, the coating traceability path time-series logic qualification verification is performed. If the verification is qualified, the final coating quality defect traceability result is output; otherwise, a coating quality defect traceability result unqualified prompt is sent.

[0027] Preferably, the specific process for analyzing the characteristics of the coating dynamic process parameter curve is as follows: Obtain the coating dynamic process parameters used to characterize the factors influencing the quality of the coating process; the coating dynamic process parameters include the coating temperature sequence collected by thermocouples, the coating pressing pressure sequence collected by the pressure transmitter of the pressing roller to reflect the tightness of the adhesion between the film and the substrate, the coating conveying speed sequence collected by the speed sensor of the conveying roller to reflect the film conveying stability, and the adhesive coating amount sequence collected by the flow sensor of the coating roller to reflect the uniformity of coating adhesion; the coating temperature value at each sampling point in the coating temperature sequence is represented by the arithmetic mean of the temperature values ​​synchronously collected by multiple temperature sensors (such as thermocouples) deployed on the surface of the coating heating roller or key areas of the film surface (such as left, middle, and right). For example, the coating temperature sequence is T... 1 =[(t p T p )], where p=1,2,3,...,n, p is the number of sampling points for the coating temperature, n represents the total number of sampling points for the coating temperature, t represents the sampling time, and T represents the coating temperature value corresponding to the sampling time.

[0028] Based on the dynamic process parameters of the coating process, the curve shape features of the coating process parameters, including slope features, peak features, phase features, and duration features, are extracted. The specific extraction process is as follows: Within a sliding window, the time-series data points corresponding to the dynamic process parameters of the coating process are linearly fitted using the least squares method, and the slope value of the fitted line is used as the slope feature. The slope features include the slope features of coating temperature, coating pressing pressure, coating conveying speed, and adhesive coating amount, which correspond to the change rate of various dynamic process parameters of the coating process, and can characterize the rising, falling, or stable change trend of the corresponding dynamic process parameters of the coating process. Based on the sliding window extreme value detection algorithm, the local maxima of the dynamic process parameters of the coating process are identified within the sliding window. The peak feature is defined as the local maxima of the coating dynamic process parameters, along with the peak offset. This peak feature captures instantaneous shocks or periodic fluctuations in the parameters. The peak offset is represented by the difference between the timestamps corresponding to the extreme points of the coating dynamic process parameters and the timestamps at the start of the sliding window. The correlation coefficient between the slope feature sequences of any two types of coating dynamic process parameters is obtained using correlation analysis methods (such as Pearson product moment and Spearman rank correlation analysis), and this correlation coefficient is used as the phase feature. The sampling duration for coating process parameter values ​​exceeding a preset process threshold is used as the duration feature. The preset process threshold includes a preset coating temperature threshold and a preset coating pressing pressure. Thresholds, preset coating conveying speed thresholds, and preset adhesive coating amount thresholds are preset by designated personnel. The sampling duration is represented by the sum of time intervals corresponding to sampling data points exceeding the preset process thresholds. The time interval represents the time difference between two consecutive sampling data points exceeding the preset process thresholds, calculated by subtracting the sampling timestamps of the two data points. The coating characteristics are matched and their integrity is verified based on the curve shape features of the coating process parameters. The specific process is as follows: The integrity rate of the time-series curve shape feature extraction is obtained to characterize the completeness of the extracted coating process parameters. The integrity rate of the time-series curve shape feature extraction is determined by the successfully extracted coating process parameters monitored by the coating data acquisition instrument. The result of the ratio calculation between the data volume corresponding to the curve shape feature and the preset feature data volume threshold is represented, where the preset feature data volume threshold is set in advance by preset personnel; the time-series matching compliance rate of the coating feature is obtained to characterize the time-series correlation between the curve shape feature of the coating process parameter and the quality index; the time-series matching compliance rate of the coating feature is represented by the ratio calculation between the qualified time-series data of the coating feature monitored by the coating data acquisition instrument and the total data volume of the curve shape feature of the coating process parameter; the qualified time-series data of the coating feature represents the total amount of data of the curve shape feature of the coating process parameter that meets the feature time-series deviation value of less than the preset time deviation threshold, where the preset time deviation threshold is represented by the average value of the feature time-series deviation value over a historical period;The characteristic timing deviation value is represented by the result of subtracting the timestamp of the center point of the sliding window corresponding to the shape feature of the film laminating process parameter curve from the timestamp of the detection of the corresponding film laminating core quality indicators (adhesive strength, light transmittance). The characteristic timing deviation value is used to quantify the degree of timestamp misalignment between the film laminating process parameter features and the film laminating core quality indicators, ensuring that the causal relationship analysis between the dynamic fluctuations of the film laminating process parameters and the film laminating quality defects has a rigorous timing logic support; the timestamp of the film laminating core quality indicator detection comes from the automatic recording of the online quality inspection equipment or offline sampling inspection equipment supporting the film laminating process. The adhesive strength is obtained by detecting the bonding force between the film and the substrate of the film laminating finished product through a tensile testing device, and the light transmittance is obtained by detecting the light transmittance consistency of the film laminating finished product through a light transmittance testing device.;

[0029] Based on the extraction completeness rate of the timing curve shape features, the integrity of the film laminating characteristic timing curve is judged. The specific process is as follows: Judge whether the extraction completeness rate of the timing curve shape features is greater than the preset timing feature integrity threshold. If so, the timing matching of the film laminating features is judged. Otherwise, it means that only the instantaneous single-point value is recorded during feature extraction, and the shape features of the curve are not retained, and no effective timing data can be provided for the subsequent map construction. Therefore, the timing curve feature optimization is adopted. Among them, the preset timing feature integrity threshold is represented by the average value of the extraction completeness rate of the timing curve shape features in the historical time period; the timing matching of the film laminating features judgment means judging whether the passing rate of the timing matching of the film laminating features is greater than the preset feature timing matching threshold. If so, it means that the timing correlation between the shape features of the film laminating process parameter curve and the film laminating core quality indicators is effective, and the corresponding shape features of the film laminating process parameter curve are marked as qualified film laminating timing feature data, and the timing causal judgment of the film laminating map is carried out based on the qualified film laminating timing feature data, and at the same time, the timing counter is reset to the initial value. Otherwise, the timing counter is incremented, the cumulative value of the timing counter is obtained, and the cumulative judgment of the film laminating feature timing is carried out based on the cumulative value of the timing counter. Among them, the preset feature timing matching threshold is represented by the average value of the passing rate of the film laminating feature timing matching in the historical time period; the increment operation means adding 1 to the value of the timing counter.

[0030] As described above, the analysis of the characteristics of dynamic process parameters in the coating process helps to systematically control the integrity and temporal consistency of coating process data from the source of the traceability process. It captures the temporal fluctuation patterns of core parameters such as coating temperature and pressing pressure, and fully extracts key shape features such as slope, peak value, phase, and duration. This can effectively reduce the distortion of subsequent analysis caused by missing coating data and temporal misalignment, reduce the risk of misjudgment and ineffective calculation costs in the coating quality traceability process, improve the temporal correlation accuracy between the characteristics of dynamic process parameters in the coating process and the core quality indicators of the coating process, and enhance the reliability of subsequent temporal causal determination of the graphs. This achieves high fidelity and high reliability of the data source for coating quality traceability, and provides solid data support for accurate tracing, root cause location, and process optimization of quality defects in the coating process chain.

[0031] Preferably, the specific process for the cumulative judgment of coating feature time sequence is as follows: It is determined whether the accumulated value of the time sequence counter is greater than a preset time sequence error threshold. If so, coating timestamp consistency calibration is performed. After the coating timestamp consistency calibration is completed, the time sequence counter value is reset to the initial value. Otherwise, a loop termination constraint judgment is triggered. The preset time sequence error threshold is used to define the tolerable number of consecutive coating feature time sequence matching deviations. When the number of consecutive deviations exceeds this threshold, it indicates that the time sequence correlation deviation has formed a stable trend. If calibration is not performed, it will continue to affect the time sequence correlation effectiveness between the coating dynamic process parameters and the coating core quality indicators. This threshold is set in advance by the preset personnel. The loop termination constraint judgment indicates the determination of the current coating... If the cumulative number of executions of the dynamic process parameter curve feature analysis exceeds the preset analysis loop threshold, it indicates that the coating feature time sequence matching compliance rate of the current coating process parameter curve feature is within the tolerable range and the state is stable. The corresponding coating process parameter curve shape feature is marked as qualified coating time sequence feature data, and the coating spectrum time sequence causality determination is performed based on the qualified coating time sequence feature data. Otherwise, the dynamic process parameter curve feature analysis of coating continues to be executed. The preset analysis loop threshold is used to limit the maximum number of executions of the dynamic process parameter curve feature analysis of coating to avoid falling into an infinite loop when the time sequence matching deviation continues to fail to meet the standard and the calibration condition is not triggered. It is preset by the personnel in advance.

[0032] As described above, the cumulative discrimination of coating feature time series helps to distinguish between single, tolerable minor deviations and continuous, stable time series anomalies in coating feature time series matching. This avoids misjudging occasional equipment delays or minor fluctuations in data acquisition during production as time series correlation failures, ensuring the rigor of time series correlation verification, preventing infinite loops, improving the accuracy, stability, and system operating efficiency of the time series correlation determination of coating dynamic process parameter features and core coating quality indicators, achieving refined control and dynamic adaptation of time series deviations, and providing a reliable data foundation that balances time series causality and efficiency for subsequent coating pattern time series causal determination. This promotes the intelligentization and efficiency of the coating quality traceability process.

[0033] Preferably, the specific process of time series curve feature optimization is as follows: The completeness rate of time series curve shape feature extraction and the smoothness of time series curve signal are input into a preset sliding window length adjustment mapping table for querying to obtain the sliding window length adjustment coefficient; the smoothness of the time series curve signal is represented by the root mean square value of the difference between adjacent sampling data points of the coating process parameters; the magnitude corresponding to the sliding window length adjustment coefficient is set as the adjustment step size, and the sliding window time span used for time series curve shape feature extraction is adjusted step by step along the direction of increasing sliding window duration (each time the sliding window time span is adjusted, the completeness rate of time series curve shape feature extraction is re-obtained; if the completeness rate of time series curve shape feature extraction is not greater than the preset time series feature completeness threshold, the adjusted sliding window time span is used as the initial value for the next adjustment, and the sliding window time span is continued to be adjusted in the direction of increasing sliding window duration), which helps to... To reduce the mismatch between window length and parameter fluctuation patterns caused by large-scale adjustments in a single instance, and to prevent feature redundancy or weakening of key fluctuation signals due to excessive window enlargement, the sliding window length is refined and optimized. The sliding window time length is less than a preset sliding window time threshold, which is pre-set by designated personnel. The completeness rate of time-series curve shape feature extraction is continuously monitored. If the completeness rate is greater than the preset time-series feature integrity threshold, the corresponding coating process parameter curve shape feature is marked as qualified coating time-series feature data, and the coating spectrum time-series causality is determined based on the qualified coating time-series feature data. Otherwise, time-series curve feature optimization continues. When the sliding window time length is not less than the preset sliding window time threshold, if the completeness rate of time-series curve shape feature extraction is still not greater than the preset time-series feature integrity threshold, a time-series curve feature optimization failure prompt is sent.

[0034] Researchers in this field should understand that the sliding window length adjustment mapping table used in this solution for performing time series curve feature optimization is calibrated and stored in a dedicated database in advance by technicians through multi-scenario verification, providing reliable support for the rapid acquisition of sliding window length adjustment coefficients. When the system performs time series curve feature extraction optimization, the appropriate adjustment coefficients can be accurately retrieved from the mapping table to ensure that the adjustment range of the sliding window time span is highly consistent with the time series fluctuation law of the coating process parameters.

[0035] Specifically, the mapping table is constructed based on a large amount of effective process data from historical coating production scenarios. It covers parameter combinations for different coating process parameter ranges, production batch characteristics, parameter fluctuation frequencies, etc. (including multi-dimensional combinations of time-series curve shape feature extraction completeness rate, time-series curve signal smoothness, and corresponding sliding window length adjustment coefficients). Each parameter combination is assigned a quantitative weight based on its adaptability to feature extraction effect (e.g., when the time-series curve shape feature extraction completeness rate is significantly lower than the preset threshold, a larger adjustment coefficient is matched to expand the sliding window coverage). At the same time, the actual effective values ​​of the adjustment coefficients in each historical scenario are recorded. Correlation analysis (such as Spearman correlation coefficients) is used to eliminate abnormal correlation data caused by instantaneous sensor drift (such as temperature sensor fluctuations) and sudden interference in the production environment (such as equipment vibration), while retaining statistically significant parameter correspondences to ensure the stability and adaptability of the mapping table.

[0036] As described above, time-series curve feature optimization helps to accurately adapt to the time-series fluctuation patterns of coating dynamic process parameters, effectively compensates for the feature loss problem caused by unreasonable initial window settings, improves the completeness, fidelity, and accuracy of extracting the curve shape features (slope, peak value, phase, and duration) of coating process parameters, avoids the weakening of key fluctuation signals or feature redundancy caused by excessive window adjustment, and realizes the dynamic adaptation of the sliding window length to the time-series distribution pattern of coating process parameters. It provides highly reliable quantitative feature support for the time-series correlation analysis and causal determination of the graph of coating dynamic process parameter features and coating core quality indicators.

[0037] Preferably, the specific process for calibrating the consistency of the coating timestamp is as follows: The standard time of the coating production control center is set as the unified time reference; based on the network time protocol, the deviation value between the unified time reference and the current time of each coating data acquisition device (such as temperature sensors for collecting coating temperature, sensor transmitters for collecting pressing pressure, flow meters for collecting adhesive coating amount, etc.) is obtained, and this deviation value is used as the coating characteristic time deviation value; the coating characteristic time deviation value is referenced to the unified time reference. A positive value indicates that the time sequence / quality index detection data of the coating process parameters collected by the corresponding acquisition device is ahead, and a negative value indicates that the time sequence / quality index detection data of the coating process parameters collected by the corresponding acquisition device is behind. The positive and negative definitions of the coating characteristic time deviation value can clearly quantify the time misalignment direction (lagging or ahead) between each coating data acquisition device and the unified time reference, providing a clear directional basis for unifying the time reference of multi-source coating data, thereby ensuring the time sequence consistency of coating process parameters and coating quality index detection data; regarding the coating process parameters... For time-series data, the timestamps of the coating process parameter time-series data are corrected based on their coating feature time deviation values. The specific process is as follows: First, obtain the coating feature time deviation value corresponding to the single coating process parameter time-series data that needs correction; superimpose the coating feature time deviation value onto the original timestamp of the coating process parameter time-series data to obtain the calibrated timestamp. Superposition means performing an algebraic sum operation between the original timestamp and the coating feature time deviation value; For the coating core quality indicator detection data, perform the same logic correction on its corresponding detection timestamp, that is, obtain the coating feature time deviation value corresponding to the coating core quality indicator detection data, and then superimpose the coating feature time deviation value onto the original detection timestamp of the coating core quality indicator detection data to obtain the calibrated timestamp; After the coating timestamp consistency calibration is completed, the coating feature time-series matching compliance rate is re-obtained. If the coating feature time-series matching compliance rate is greater than the preset feature time-series matching threshold, the coating spectrum time-series causality determination is performed; otherwise, a timestamp calibration failure prompt is sent.

[0038] As described above, the consistency calibration of coating timestamps helps to unify the time base of multiple sources of equipment, such as equipment for collecting dynamic coating process parameters and equipment for detecting core coating quality indicators. It corrects timestamp deviations between different devices (such as being ahead or behind), improves the accuracy of the time-series correlation between dynamic coating process parameters and core coating quality indicators, enhances the consistency of multi-source data and the compliance rate of time-series matching, reduces the interference of inconsistent time bases on the rigor of traceability logic, and achieves time-series collaborative alignment of cross-device data.

[0039] Preferably, the specific process for determining the temporal causality of the coating pattern is as follows: Obtain coating pattern quality assessment parameters, including the coating pattern temporal association missingness, used to characterize the completeness of the association between pattern entities and qualified coating temporal feature data, and the coating temporal constraint completeness rate, used to characterize the coverage of temporal attributes between pattern nodes; the coating pattern temporal association missingness is represented by the ratio of the number of unassociated entity nodes to the total number of target entity nodes, monitored by the coating data volume monitor; the number of unassociated entity nodes represents the entity nodes used to characterize dynamic process parameters (such as coating temperature sensor nodes, pressing pressure sensor nodes, film conveying speed detection nodes, adhesive coating amount measurement nodes, etc.) in the coating process chain quality traceability knowledge graph, and counts those nodes that are not associated with any qualified coating temporal feature data units. The number of nodes establishing links; the total number of target entity nodes represents the total number of entity nodes representing dynamic process parameters in the coating process chain quality traceability knowledge graph; the coating time sequence constraint completeness rate is represented by the ratio of the number of time-series attribute associations monitored by the coating data volume monitor to the total number of associations between target nodes; the number of time-series attribute associations represents the number of edges containing time-series attributes (such as timestamp range, time sequence identifier, etc.) by traversing the attribute fields of all causal relationship edges and process execution relationship edges in the coating process chain quality traceability knowledge graph; the total number of associations between target nodes represents the total number of all causal relationship edges and process execution relationship edges in the coating process chain quality traceability knowledge graph (including association edges with and without time-series attributes).

[0040] The system determines whether the coating graph quality assessment parameters meet the coating graph quality adaptation conditions. If they do, the system determines that the temporal correlation and causal logic adaptation of the coating graph meet the standards, and subsequent coating process quality traceability reasoning can be carried out based on the current graph. Otherwise, the system outputs a coating graph problem list and triggers the supplementation of graph temporal constraints and the completion of correlation relationships. The coating graph quality adaptation conditions indicate that the coating graph temporal correlation missing degree is less than a preset missing degree threshold, and the coating temporal constraint completeness rate is greater than a preset completeness rate threshold. The preset missing degree threshold is represented by the average of the coating graph temporal correlation missing degree over a historical time period, and the preset completeness rate threshold is represented by the average of the coating temporal constraint completeness rate over a historical time period. The coating graph problem list indicates a list of specific defects identified in the coating graph temporal causality determination stage that hinder the knowledge graph from accurately expressing temporal causal relationships. These include a list of dynamic process parameter entity nodes that have not been associated with qualified coating temporal feature data, and a list of causal relationship edges and process execution relationship edges that lack temporal attributes.

[0041] As described above, the temporal causal determination of the coating graph helps to establish a quantitative causal mapping relationship between the temporal characteristics of the coating dynamic process parameters and coating quality defects. This effectively compensates for the potential lack of temporal constraints and the deviation in the association of entity nodes in the coating graph, reduces the risk of defect root cause location deviation and misjudgment of traceability direction caused by fuzzy association logic and temporal misalignment, and enhances the dynamic adaptability of the knowledge graph to the actual working conditions of the current production batch. This provides a clear causal orientation and rigorous logical support for subsequent coating process quality traceability reasoning.

[0042] Preferably, the specific process for supplementing the temporal constraints and completing the correlation relationships in the coating map is as follows: Based on the coating map problem list, extract the specific objects to be supplemented in the coating map, including the set of dynamic process parameter entity nodes with unassociated feature data, and the set of causal relationship edges and process execution relationship edges with missing temporal attributes; extract the core information of qualified coating temporal feature data, including parameter identifiers, calibrated timestamps, time intervals, and associated production batch numbers, as the core data basis for coating map completion; for dynamic process parameter entity nodes in the problem list that have not been associated with qualified coating temporal feature data (such as unassociated coating temperature sensor nodes, adhesive coating amount metering nodes, etc.), based on qualified coating temporal features... The parameter identifiers in the data are associated with production batch number information to establish a correspondence between nodes and data units. Specifically, the process involves matching the parameter type attributes (such as coating temperature, adhesive coating amount, etc.) of the dynamic process parameter entity node with the parameter identifier field of the qualified coating time-series characteristic data. Simultaneously, the production batch number identifiers of both are matched to ensure a one-to-one association between the coating dynamic process parameter entity node and the corresponding qualified coating time-series characteristic data unit for the same parameter type and production batch. For example, there exists a coating dynamic process parameter entity node X, where X represents a coating temperature sensor entity node. The parameter type attribute of this node is explicitly marked as coating temperature (i.e., the coating dynamic process parameter category corresponding to node X is...). The data includes coating temperature monitoring data. Simultaneously, data unit Y exists within the qualified coating time-series characteristic data. The parameter identifier field of data unit Y explicitly states coating temperature (indicating that this data unit records time-series characteristic data related to coating temperature). Furthermore, the associated production batch number identifier for data unit Y is P (where P represents a unique identifier for a specific production batch, used to distinguish coating dynamic process data from different production cycles). By comparing the parameter type attribute of entity node X with the parameter identifier field of data unit Y, it is confirmed that both point to the same parameter type: coating temperature. Then, the associated production batch number of entity node X (this node is already associated with the identifier of production batch P in the knowledge graph) is compared with the associated production batch number of data unit Y. The batch number P confirms that the two belong to the same production batch. Based on the above dual matching results, a one-to-one association relationship is established between entity node X and data unit Y in the knowledge graph. That is, entity node X (coating temperature sensor node) corresponds only to data unit Y (coating temperature time-series feature data of the same batch), and data unit Y is also only associated with entity node X, ensuring the accurate binding of parameter nodes and corresponding time-series data. After the association is established, a prompt to add a time-series feature relationship edge is sent in the coating process chain quality traceability knowledge graph to complete the graph association relationship. The prompt to add a time-series feature relationship edge means sending a prompt to add a relationship edge with time-series features from the entity node to the qualified coating time-series feature data unit.

[0043] For causal relationship edges and process execution relationship edges lacking time sequence attributes, corresponding time information is extracted from qualified lamination time sequence feature data and added to the attribute fields of the relationship edges. Specifically, for causal relationship edges, the cause triggering timestamp (the calibrated timestamp corresponding to the abnormal parameter feature causing the relationship), the result generation timestamp (the calibrated timestamp corresponding to the quality indicator detection), and the time sequence identifier are added. For example, in the lamination process chain quality traceability knowledge graph, there is a causal relationship edge Z. The two nodes connected by this relationship edge are the lamination dynamic process parameter entity node M (representing the adhesive coating amount measurement node, corresponding to the adhesive coating amount in the lamination dynamic process parameter) and the lamination quality event node N (representing the surface bubble defect node of the lamination finished product, corresponding to the surface bubble in the quality defect). The core meaning of relationship edge Z is that the abnormal adhesive coating amount leads to surface bubbles in the lamination finished product. However, this relationship edge does not contain time sequence attribute information. The core time information corresponding to entity node M is extracted from qualified lamination time sequence feature data, whereby... The calibrated timestamp corresponding to the abnormal adhesive coating amount feature M is recorded as W1 (W1 represents the specific time point after timestamp consistency calibration, used to accurately reflect the occurrence time of the abnormal adhesive coating amount). Then, the core time information corresponding to the coating quality event node N is extracted, that is, the calibrated timestamp of the detection of surface bubble defects of the coated finished product corresponding to N is recorded as W2 (W2 represents the specific time point after timestamp consistency calibration, used to accurately reflect the detection time of surface bubble defects). Timestamp W1 is used as the cause trigger timestamp and timestamp W2 is used as the result generation timestamp to supplement the attribute field of the causal relationship edge Z. At the same time, the time sequence identifier F is added (F represents the time sequence logic identifier, its value is cause before result, clearly indicating that the occurrence time of the abnormal adhesive coating amount W1 is earlier than the detection time of surface bubble defects W2). The time sequence attribute supplementation of the causal relationship edge Z is completed, so that the relationship edge clearly defines the time sequence logic of cause trigger time, result generation time, and cause and result sequence.

[0044] For each process execution relationship edge, the execution start time interval, execution end time interval (determined based on the time range of the dynamic process parameter time series data associated with the corresponding process), and time sequence identifier (marking the order of process execution, such as "preheating process" before "coating process") are added. After the attribute fields of the relationship edge are added, it is ensured that the attribute fields of each target relationship edge contain time information to form a complete time sequence constraint. After the time sequence constraint of the graph is supplemented and the relationship is completed, a new coating process chain quality traceability knowledge graph is regenerated, and the coating graph quality assessment parameters are re-acquired. If the coating graph quality assessment parameters meet the coating graph quality adaptation conditions, coating process quality traceability reasoning is performed; otherwise, a graph completion failure prompt is sent.

[0045] As described above, by supplementing the knowledge graph with temporal constraints and completing the relationships, it is helpful to systematically compensate for the problems of missing entity node associations and insufficient temporal attributes of relation edges in the knowledge graph for quality traceability of the coating process chain. It can accurately establish a one-to-one association between the dynamic process parameter nodes of coating and the temporal feature data of qualified coating, clearly define the temporal logic of causal relationships and process execution relationships, improve the data integrity, node association accuracy and temporal logic rigor of the knowledge graph, and provide a structurally complete and logically consistent knowledge graph foundation for subsequent quality traceability reasoning of coating process and temporal logic verification of traceability paths.

[0046] Preferably, the specific process of quality traceability reasoning in the lamination process is as follows: When a quality defect is detected in the finished laminated product, the defect is defined as the current lamination quality event. From the quality traceability knowledge graph of the lamination process chain, with the lamination quality event node as the query center, a graph traversal algorithm (such as breadth-first search) is used to identify the range of nodes directly or indirectly related to the current lamination quality event (covering nodes such as equipment, materials, and operation records of the corresponding process), generating a quality defect relationship subgraph. The quality defect represents an abnormal state in which the finished laminated product does not meet the preset standard requirements in terms of core quality attributes. The core quality attributes cover the key control indicators of the lamination process (such as the adhesion strength between the film and the substrate, the light transmittance of the finished laminated product, and the surface smoothness). The requirements include no bubbles, wrinkles, or curling edges (requiring an overall appearance integrity). Preset standards include industry technical standards, internal enterprise quality control standards, or customer-customized quality requirements. Specifically, this manifests as adhesive strength exceeding a preset adhesive threshold, light transmittance deviation less than a preset light transmittance threshold, and the absence of obvious bubbles, wrinkles, or curling edges on the surface. These defects require specific tracing to pinpoint their root cause at the process, equipment, material, or operational level. Specifically, the preset adhesive threshold is represented by the average adhesive strength over a historical period, and the preset light transmittance threshold is represented by the average light transmittance deviation over a historical period. The appearance integrity requirements (such as the absence of bubbles, wrinkles, or curling edges) are assessed by capturing a full-surface image of the coated product using a high-definition industrial camera, followed by image preprocessing (such as Gaussian filtering for noise reduction). Features such as surface unevenness, warping, and localized reflective anomalies are extracted and compared with preset feature thresholds for bubbles, wrinkles, and edge curling (e.g., preset bubble diameter, preset wrinkle height, preset edge curling angle, etc.). If the detection results are consistent, it is determined that there are no obvious bubbles / wrinkles / edge curling on the coated surface. The preset bubble diameter is represented by the average bubble diameter over a historical time period, the preset wrinkle height by the average wrinkle height over a historical time period, and the preset edge curling angle by the average edge curling angle over a historical time period. The structured time-series feature data (e.g., slope feature value, peak feature value, phase feature value, duration value, etc.) associated with the quality defect relationship subgraph, as well as the attributes of the nodes themselves (e.g., material batch number), are then analyzed. Vectorization is used to construct a quality defect feature matrix. Vectorization involves converting non-numerical data in the quality defect relationship subgraph into computer-recognizable numerical vectors based on preset encoding rules (such as label encoding for unique identifiers like production batch numbers). Simultaneously, existing numerical data is standardized and regularized (e.g., Z-score standardization) to ultimately form feature vectors with uniform dimensions and standardized formats. These processed feature vectors are then concatenated in a preset order to form a one-dimensional numerical vector containing all core information. Each vector corresponds to a quality defect-related entity node or data unit. All vectors are combined to construct the quality defect feature matrix. The preset order is pre-set by a designated person.

[0047] The quality defect feature matrix is ​​input into a preset classification model (such as random forest), which outputs the core related factors of the current coating quality defect (such as insufficient purity of a batch of materials) and the corresponding confidence scores of the coating quality defect factors. The specific training process of the classification model (such as random forest) is as follows: First, collect quality defect case data from the historical coating production process. Each case needs to include a corresponding quality defect relationship subgraph, structured time-series feature data, node attribute information, and core related factor labels that have been manually labeled or actually verified. Second, using the same vectorization processing method as above, convert all relevant data from historical cases into a quality defect feature matrix and divide it into training and test sets according to a preset ratio. Then, initialize the basic parameters of the classification model (such as the number of decision trees) and input the training set into the model. The model is trained by employing an ensemble learning logic for decision trees. Feature selection is performed on the split nodes of each decision tree, constructing multiple independent decision trees based on key information in the feature matrix. Then, grid search or cross-validation methods are used to fine-tune the model's hyperparameters. The predictive performance of different hyperparameter combinations is validated using a test set (evaluation metrics include accuracy and recall), and the optimal hyperparameter combination is selected. Finally, the model with the optimal hyperparameter configuration is evaluated on the test set (evaluation metrics include accuracy and recall). If the evaluation results reach preset performance thresholds (e.g., accuracy greater than a preset accuracy threshold, recall greater than a preset recall threshold, etc.), the model training is complete and saved. If the thresholds are not reached, the process is repeated by adjusting hyperparameters or supplementing historical case data until the model performance meets the requirements. Preset ratios are used to ensure the consistency of the distribution between the training and test sets. These are pre-set by designated personnel. The preset accuracy threshold is represented by the average of preset accuracy thresholds over a historical time period, and the preset recall threshold is represented by the average recall rate over a historical time period.

[0048] If the confidence level of a coating quality defect factor is greater than a preset confidence threshold, the core associated factor of the corresponding coating quality defect is identified as a qualified coating quality defect associated factor. Based on the qualified coating quality defect associated factor, reverse inference of the coating process chain quality is performed. The preset confidence threshold is represented by the average confidence level of coating quality defect factors over a historical period. Reverse inference of the coating process chain quality is represented by the entity nodes corresponding to the knowledge graph of the coating process chain quality traceability corresponding to the qualified coating quality defect associated factor, as the starting point set for inference, starting from the source of the coating process (…). Using raw material procurement and initial process settings as the reasoning target, a reverse tracing algorithm (such as backtracking algorithm) is used in the knowledge graph to generate a traceability path for coating quality defects. This traceability path includes defect-related factor nodes, intermediate process / equipment / material nodes, source nodes, and the relationships between each node (such as raw material procurement - adhesive warehousing - adhesive use - coating process - pressing defects), as well as time sequence information (time stamps / time intervals of each link). The traceability path for coating quality defects is used to verify the time sequence logic of the coating traceability path.

[0049] As described above, quality traceability reasoning in the coating process helps to integrate quantitative data of the quality defect feature matrix, the correlation logic of the knowledge graph, and the time sequence constraint information. This allows for the accurate identification of the core related factors (such as material batch issues) and the entire chain propagation path of coating quality defects, reducing the inefficiency caused by blind tracing. At the same time, by combining entity node attributes and structured time sequence characteristics, it enables in-depth analysis from the phenomenon of coating quality defects to the root cause of coating defects, providing a feasible decision-making basis for adjusting dynamic coating process parameters, optimizing material control, and upgrading equipment maintenance.

[0050] Preferably, the sequential logic compliance verification of the coating traceability path involves traversing the timestamp information corresponding to all nodes in the generated traceability path, checking whether the time sequence of each node strictly conforms to the actual production process sequence of the coating process (e.g., the order of raw material procurement time → material inspection time → process execution time → parameter change time → defect generation time → defect detection time). The number of node pairs in the traceability path that do not conform to the actual production process sequence of the coating process is counted. If the number of node pairs that do not conform to the actual production process sequence of the coating process is 0, the verification is deemed qualified; otherwise, the verification is deemed unqualified. If the verification is qualified, the final coating quality defect traceability result (including traceability path and key abnormal nodes) is output based on the traceability path and key abnormal nodes confirmed by the sequential logic compliance verification of the coating traceability path, the preset process rules in the coating process chain quality traceability knowledge graph, and the associated relationships. If the verification is unqualified, a coating quality defect traceability result unqualified prompt is sent.

[0051] For example, a compliant traceability path confirmed through the sequential logic verification of the coating traceability path is Node G → Node H → Node I → Node J → Node K → Node M. The meanings of each node are as follows: Node G represents the raw material procurement node (recording the procurement of basic raw materials required for coating production); Node H represents the material inspection node (recording the quality inspection of procured raw materials); Node I represents the coating process execution node (recording the start-up and operation of the core coating production process); Node J represents the critical anomaly node (specifically, the adhesive coating amount parameter node, where the associated process parameters experience abnormal fluctuations during production); Node K represents the defect generation node (recording the stage where bubble defects appear on the surface of the coated finished product); and Node M represents the defect detection node (recording the stage where surface bubble defects are found during finished product inspection). Furthermore, the timestamps corresponding to each node strictly follow the sequential order of "raw material procurement → material inspection → process execution → parameter anomaly → defect generation → defect detection." Simultaneously, the coating process chain quality traceability knowledge... The pre-defined process rules in the knowledge graph clearly state that "abnormal adhesive coating amount parameters (exceeding the preset process threshold) are the core causal relationship leading to surface bubble defects in the coated finished product." The pre-defined association relationship clearly states that "there is a direct causal relationship between the adhesive coating amount parameter node (node ​​J) and the surface bubble defect generation node (node ​​K)." Based on the above compliant traceability path, the abnormal information of the key abnormal node J, and the corresponding pre-defined process rules and association relationships in the knowledge graph, the final output traceability result of the coating quality defect is as follows: The traceability path is: raw material procurement node (G) → material inspection node (H) → coating process execution node (I) → adhesive coating amount parameter abnormal node (J) → surface bubble defect generation node (K) → defect detection node (M). The key abnormal node is the adhesive coating amount parameter node (J). The root cause of the defect is that the adhesive coating amount parameter exceeds the preset process threshold. The association logic is that the adhesive coating amount parameter abnormality (node ​​G) directly leads to the surface bubble defect (node ​​K) through the causal association edge.

[0052] As described above, by verifying the time sequence logic of the traceability path for film coating, it is helpful to systematically identify contradictions in the time sequence of entity nodes in the traceability path, ensure that the time sequence of each link strictly matches the actual process of film coating production, reduce the risk of distorted traceability results and misjudgment of the root cause of defects caused by time sequence disorder, improve the time sequence compliance of the traceability path and the credibility of the traceability results, and realize the time sequence closed-loop control of the entire traceability of film coating quality defects.

[0053] Example 2, as an alternative to the timing logic qualification verification of the coating traceability path in Example 1, when the number of entity nodes of the coating production sub-unit in the traceability path is greater than the preset production sub-unit threshold, and there are entity nodes whose time deviation is within the allowed range of timestamp completion (such as the deviation of the standard time base of the coating production sub-unit, which is within the fault tolerance range of the timestamp completion of the coating production sub-unit), because Example 1 relies on a unified single time base to complete the full node traversal verification, the small time base deviation across coating production sub-units will be amplified by the traversal of the full number of nodes, leading to misjudgment in the timing logic qualification verification of the coating traceability path. Therefore, it is necessary to execute the second scheme for timing logic qualification verification of the coating traceability path. The specific process is as follows: From the traceability path of coating quality defects, extract the core timing link nodes that ensure the integrity of the coating production process. The core timing link nodes need to cover the key links (such as the key links of raw material use → process execution → process parameter change → quality defect generation → defect detection); obtain the time information corresponding to the extracted core timing link nodes (such as raw material use). Using timestamps, process execution timestamps, defect generation timestamps, etc., verify whether the time sequence of each core timing link node conforms to the preset core timing requirements (e.g., raw material usage time → process execution time → parameter change time → defect generation time → defect detection time); count the number of node pairs in the core timing link that do not conform to the preset core timing requirements. If the number of node pairs that do not conform to the preset core timing requirements is 0, then based on the compliant traceability path and key abnormal nodes confirmed by the timing logic of the coating traceability path, the preset process rules and relationships in the coating process chain quality traceability knowledge graph, the final coating quality defect traceability result (including traceability path and key abnormal nodes) is output. Otherwise, a coating quality defect traceability result non-compliant prompt is sent. Among them, the allowed range of timestamp completion represents the maximum acceptable time deviation range, which is set in advance by the preset personnel; the preset production sub-unit threshold represents the minimum number of coating production sub-units required to trigger the second scheme. That is, when the number of coating production sub-units covered by the traceability path exceeds this threshold, the impact of cross-sub-unit time base deviation needs to be considered, which is set in advance by the preset personnel.

[0054] As described above, the replacement scheme for the timing logic qualification verification of the coating traceability path can accurately adapt to the scenario of collaborative traceability of multiple production sub-units. It effectively reduces the verification misjudgment problem caused by the amplification of small deviations in the time base across sub-units by traversing all nodes in Example 1. It helps to focus on the core timing links of the coating production process for targeted verification, ignore irrelevant deviations of non-critical nodes, and is compatible with the time base differences of different production sub-units (within the allowable deviation range). This improves the adaptability, accuracy and efficiency of the timing logic qualification verification of the coating traceability path in complex production organization modes, and realizes fault-tolerant adaptation to differences in time bases across sub-units and precise control of core timing logic.

[0055] like Figure 3 The diagram shown illustrates the structure of a knowledge graph-based coating process chain quality traceability system. Figure 3It is known that the knowledge graph-based coating process chain quality traceability system includes: a coating dynamic process parameter feature monitoring module, a coating graph temporal causality monitoring module, and a coating quality traceability reasoning and source path verification module. The coating dynamic process parameter feature monitoring module performs curve feature analysis of the coating dynamic process parameters to determine whether to implement temporal curve feature optimization and coating timestamp consistency calibration. If implemented, it checks the performance of the dynamic process parameter curve feature analysis after implementation. If successful, it performs temporal causality determination of the coating graph; otherwise, it sends a failure alert. By monitoring the results of the coating dynamic process parameter curve feature analysis, it helps to build a quality control barrier for coating process data from the source, dynamically identify integrity gaps and temporal misalignments in the parameter feature extraction process, and provide high-fidelity, temporally consistent core data support for subsequent graph temporal causality analysis, reducing resource consumption for full-process traceability. The coating graph temporal causality monitoring module, if not implemented, performs temporal causality determination of the coating graph to determine whether to trigger graph temporal constraint supplementation. The system completes the relationship between the coating graph and related relationships. If triggered, after the trigger ends, it checks whether the temporal causal judgment of the coating graph is qualified. If qualified, it performs quality traceability reasoning for the coating process; if unqualified, it sends a graph completion failure prompt. By monitoring the temporal causal relationship of the coating graph, it helps to accurately locate logical defects in the knowledge graph, such as missing entity node associations and incomplete temporal attributes. This provides a graph foundation with rigorous temporal constraints and clear association rules for traceability reasoning. At the same time, the graph completion failure judgment avoids the transmission impact of logical defects on the reasoning results. The coating quality traceability reasoning and source path verification module is used to perform coating process quality traceability reasoning if not triggered. After the coating process quality traceability reasoning is completed, it performs temporal logic qualification verification of the coating traceability path. If the verification is qualified, it outputs the final coating quality defect source tracing result; if the verification is unqualified, it sends a coating quality defect source tracing result failure prompt. By monitoring the coating quality source path verification result, it helps to achieve terminal closed-loop verification of the traceability reasoning conclusion and filter out traceability paths that do not conform to the actual production process from the temporal logic dimension.

[0056] As described above, the coating dynamic process parameter feature monitoring module, the coating pattern time-series causal monitoring module, and the coating quality traceability reasoning and source path verification module help improve the accuracy, logical rigor, and scenario adaptability of coating process quality traceability. This effectively covers complex production scenarios such as single production sub-units and multi-production sub-unit collaboration, reducing the risk of traceability failure due to data defects, logical loopholes, and time-series chaos. Specifically, the modules present a rigorous logical relationship of hierarchical progression and mutual constraints. The qualified parameter feature data output by the coating dynamic process parameter feature monitoring module directly determines the analysis accuracy of the coating pattern time-series causal monitoring module. The coating pattern time-series causal monitoring module receives preceding data, and the qualified patterns it determines provide necessary logical support for the coating quality traceability reasoning and source path verification module. The coating quality traceability reasoning and source path verification module, as the terminal guarantee, performs reasoning and verification based on the output of the preceding modules. If the source path verification fails, a prompting mechanism is used to report potential problems in the preceding data processing or pattern construction stages.

[0057] like Figure 3 The knowledge graph-based method for quality traceability in the coating process chain shown is based on a knowledge graph. Figure 3 It can be seen that, with the quality defects in the lamination process as the core, the light blue circles correspond to operational status data such as lamination temperature gradient and protective film tension strain value, the light green circles represent fault characteristic data such as light transmittance deviation value and lamination bubble ratio, and the pink circles represent sub-defect categories such as light transmittance defects and bubble defects. Each node is connected by a link, forming a traceability path of operational status data → fault characteristic data → sub-defect classification → core quality defects (such as protective film tension strain value → substrate conveying synchronization rate → surface wrinkling density → peeling / wrinkling defects, pressing pressure fluctuation value + adhesive coating amount deviation → lamination bubble ratio → bubble defects). All links eventually converge to the quality defects in the lamination process at the center, intuitively presenting the cause transmission and correlation logic of defects in the lamination process.

[0058] like Figure 4 The diagram shown is an analysis of the core correlation factors of coating quality defects based on a random forest model in the knowledge graph-based coating process chain quality traceability method. Figure 4It can be seen that: taking the quality defect feature matrix as input, it is decomposed into feature subsets such as coating temperature gradient + working condition fluctuation, pressing pressure fluctuation + material deviation, etc., and then input into multiple base decision trees (decision tree 1 to decision tree n) of random forest for parallel training. Each tree outputs the coating defect-related factor matching result; then the single tree bias is eliminated by the majority voting / average ensemble strategy of random forest, and finally the core related factors of coating quality defects and the corresponding confidence of coating quality defect factors are output. This process is linked with the coating process quality traceability knowledge graph. With the help of the ensemble learning advantage of random forest, a precise defect traceability closed loop of feature quantification → multi-tree parallel matching → ensemble output is realized.

Claims

1. A knowledge graph-based method for quality traceability in the coating process chain, characterized in that, The method includes: Perform dynamic process parameter curve feature analysis on the coating process to determine whether to optimize the time-series curve features and calibrate the coating timestamp consistency. If so, after the optimization is completed, determine whether the analysis is qualified. If qualified, perform time-series causal determination of the coating pattern. If unqualified, send a failure prompt. The specific process for analyzing the characteristics of the dynamic process parameter curves for film coating is as follows: Obtain dynamic process parameters for film coating; The dynamic process parameters for coating include coating temperature sequence, coating pressing pressure sequence, coating conveying speed sequence, and adhesive coating amount sequence; Based on the dynamic process parameters of the coating process, the curve shape features of the coating process parameters, including slope features, peak features, phase features, and duration features, are extracted. The specific extraction process is as follows: Within the sliding window, the time series data points corresponding to the dynamic process parameters of the film coating are linearly fitted based on the least squares method, and the slope value of the fitted line is used as the slope feature. Based on the sliding window extreme value detection algorithm, the local maximum points of the coating dynamic process parameters within the sliding window are identified, and the values ​​of the coating dynamic process parameters and the peak offset corresponding to the local maximum points of the coating dynamic process parameters are used as peak features. The peak offset is represented by the result of the difference calculation between the timestamp corresponding to the extreme point of the dynamic process parameter of the coating and the timestamp of the starting point of the sliding window. The correlation coefficient between the slope characteristic sequences of any two types of coating dynamic process parameters is obtained based on the correlation analysis method, and this correlation coefficient is used as the phase feature. The sampling duration when the coating process parameter value exceeds the preset process threshold is used as the duration feature; The sampling duration is represented by the sum of the time intervals corresponding to the sampling data points that exceed the preset process threshold; Based on the shape characteristics of the coating process parameter curves, the timing matching and integrity verification of coating characteristics are performed. The specific process is as follows: The completeness rate of extracting the shape features of time-series curves; Obtain the temporal matching compliance rate of the coating feature; Based on the integrity rate of extraction from time-series curve shape features and the time-series matching success rate of coating features, time-series matching and integrity discrimination of coating characteristics are performed. The specific process is as follows: If the extraction completeness rate of the shape feature of the time series curve is greater than the preset time series feature completeness threshold, then perform the time series matching judgment of the overlay feature; otherwise, take the time series curve feature optimization. The coating feature time sequence matching discrimination means judging whether the coating feature time sequence matching compliance rate is greater than the preset feature time sequence matching threshold. If so, the corresponding coating process parameter curve shape feature is marked as qualified coating time sequence feature data, and the coating pattern time sequence causality is determined based on the qualified coating time sequence feature data. Otherwise, the coating feature time sequence accumulation discrimination is performed based on the time sequence counter accumulation value. The specific process of the temporal accumulation discrimination of the coating feature is as follows: Determine whether the accumulated value of the timing counter is greater than the preset timing error threshold. If so, perform overlay timestamp consistency calibration. After the overlay timestamp consistency calibration is completed, reset the timing counter value to the initial value. Otherwise, trigger the loop termination constraint judgment. The loop termination constraint judgment means determining whether the cumulative number of executions of the current coating dynamic process parameter curve feature analysis is greater than the preset analysis loop threshold. If so, the corresponding coating process parameter curve shape feature is marked as qualified coating time sequence feature data, and the coating pattern time sequence causality determination is performed based on the qualified coating time sequence feature data. Otherwise, the coating dynamic process parameter curve feature analysis continues to be executed. If no action is taken, the temporal causal determination of the coating pattern will be performed to determine whether to trigger the temporal constraint supplementation and correlation completion of the pattern. If triggered, after the trigger ends, it will be determined whether the temporal causal determination of the coating pattern is qualified. If qualified, the quality traceability reasoning of the coating process will be performed. If unqualified, a pattern completion failure prompt will be sent. If not triggered, the coating process quality traceability reasoning is performed. After the coating process quality traceability reasoning is completed, the coating traceability path timing logic qualification is checked. If the check is qualified, the final coating quality defect traceability result is output. If the check is unqualified, a coating quality defect traceability result unqualified prompt is sent.

2. The knowledge graph-based coating process chain quality traceability method as described in claim 1, characterized in that, The specific process of optimizing the time series curve features is as follows: The completeness of the extraction of the shape features of the time series curve and the smoothness of the time series curve signal are input into a preset sliding window length adjustment mapping table for querying to obtain the sliding window length adjustment coefficient; Set the magnitude of the sliding window length adjustment coefficient as the adjustment step size, and adjust the time span of the sliding window step by step along the direction of increasing sliding window duration; The system continuously monitors the completeness rate of time-series curve shape feature extraction. If the completeness rate of time-series curve shape feature extraction is greater than the preset time-series feature completeness threshold, the system performs time-series matching judgment of overlay features. Otherwise, the system continues to perform time-series curve feature optimization. If the completeness rate of time-series curve shape feature extraction is still not greater than the preset time-series feature completeness threshold when the sliding window time length is not less than the preset sliding window time threshold, the system sends a time-series curve feature optimization failure prompt.

3. The knowledge graph-based coating process chain quality traceability method as described in claim 1, characterized in that, The specific process for calibrating the consistency of the overlay timestamp is as follows: Set the standard time of the film coating production control center as a unified time benchmark; The deviation between the unified time reference and the current time of each film-covering data acquisition device is obtained based on the network time protocol, and this deviation is used as the characteristic time deviation value of film-covering. For the time series data of coating process parameters, the timestamp of the time series data of coating process parameters is corrected according to its coating characteristic time deviation value. The specific process is as follows: obtain the coating characteristic time deviation value corresponding to the single time series data of coating process parameters that needs to be corrected. The time deviation value of the coating feature is superimposed on the original timestamp of the coating process parameter time series data to obtain the calibrated timestamp; For the core quality index test data of the coating, the same logic is applied to the corresponding test timestamp to correct it. That is, the coating feature time deviation value corresponding to the core quality index test data of the coating is obtained, and then the coating feature time deviation value is superimposed on the original test timestamp of the core quality index test data of the coating to obtain the calibrated timestamp. After the coating timestamp consistency calibration is completed, the coating feature time sequence matching compliance rate is re-acquired. If the coating feature time sequence matching compliance rate is greater than the preset feature time sequence matching threshold, the coating spectrum time sequence causality determination is performed; otherwise, a timestamp calibration failure prompt is sent.

4. The knowledge graph-based coating process chain quality traceability method as described in claim 3, characterized in that, The specific process for determining the temporal causality of the coating pattern is as follows: Obtain coating map quality assessment parameters, including coating map temporal correlation missingness and coating temporal constraint completeness rate; Determine whether the coating pattern quality assessment parameters meet the coating pattern quality adaptation conditions. If they do, perform quality traceability reasoning for subsequent coating processes based on the current pattern. Otherwise, output a coating pattern problem list and trigger pattern temporal constraint supplementation and correlation relationship completion. The coating map quality adaptation condition indicates that the temporal correlation missing degree of the coating map is less than the preset missing degree threshold, and the coating temporal constraint completeness rate is greater than the preset completeness rate threshold. The coating pattern problem list includes a list of dynamic process parameter entity nodes that are not associated with qualified coating time sequence characteristic data, as well as a list of causal relationship edges and process execution relationship edges that lack time sequence attributes.

5. The knowledge graph-based coating process chain quality traceability method as described in claim 4, characterized in that, The specific process for supplementing the temporal constraints and completing the association relationships in the graph is as follows: Based on the list of issues in the coating map, the specific objects to be completed in the coating map are extracted, including the set of dynamic process parameter entity nodes without associated feature data, and the set of causal relationship edges and process execution relationship edges with missing time-series attributes. Extract the core information of qualified film coating time series characteristic data, including parameter identifier, calibrated timestamp, time interval, and associated production batch number; For dynamic process parameter entity nodes in the problem list that are not associated with qualified coating time sequence characteristic data, a correspondence between nodes and data units is established based on the parameter identifiers and associated production batch number information in the qualified coating time sequence characteristic data. The specific establishment process is as follows: Match the parameter type attributes of dynamic process parameter entity nodes with the parameter identifier field of qualified coating time sequence characteristic data; In the knowledge graph for quality traceability of the coating process chain, a prompt to add time-series feature relationship edges is sent to complete the graph association relationship; For causal relationship edges and process execution relationship edges lacking time sequence attributes, the corresponding time information is extracted from the qualified coating time sequence feature data and added to the attribute fields of the relationship edges, specifically: For causal edges, add cause trigger timestamp, result generation timestamp, and time sequence identifier; For the edges representing process execution relationships, add execution start time interval, execution end time interval, and timing sequence identifier; After the timing constraints and relationships of the knowledge graph are completed, a new knowledge graph for quality traceability of the coating process chain is regenerated, and the quality assessment parameters of the coating graph are re-acquired. If the quality assessment parameters of the coating graph meet the quality adaptation conditions of the coating graph, the quality traceability reasoning of the coating process is performed; otherwise, a knowledge graph completion failure prompt is sent.

6. The knowledge graph-based coating process chain quality traceability method as described in claim 5, characterized in that, The specific process of quality traceability reasoning for the coating process is as follows: When a quality defect is detected in the coated finished product, the defect is defined as the current coating quality event. From the coating process chain quality traceability knowledge graph, with the coating quality event node as the query center, the graph traversal algorithm is used to identify the range of nodes that are directly or indirectly related to the current coating quality event, and a quality defect relationship subgraph is generated. The structured time-series feature data associated with the quality defect relationship subgraph and the attributes of the nodes themselves are vectorized to construct a quality defect feature matrix; Input the quality defect feature matrix into the preset classification model, and output the core related factors of the current coating quality defect and the corresponding confidence level of the coating quality defect factors. If the confidence level of the coating quality defect factor is greater than the preset confidence threshold, then the core correlation factor of the corresponding coating quality defect is determined as the correlation factor of qualified coating quality defect, and the quality of the coating process chain is reversed based on the correlation factor of qualified coating quality defect. The reverse reasoning of the coating process chain means taking the entity nodes corresponding to the coating process chain quality traceability knowledge graph corresponding to the factors related to qualified coating quality defects as the starting point set of reasoning, taking the source of the coating process as the reasoning target direction, and performing reverse tracing in the knowledge graph based on the reverse reasoning algorithm to generate the source tracing path of coating quality defects. The timing logic of the traceability path for the coating quality defect is verified.

7. The knowledge graph-based coating process chain quality traceability method as described in claim 6, characterized in that, The validity check of the timing logic of the coating traceability path involves traversing the timestamp information of all nodes in the generated traceability path, checking whether the time sequence of each node strictly conforms to the actual production process sequence of the coating process, and counting the number of node pairs in the traceability path that do not conform to the actual production process sequence of the coating process. If the number of node pairs that do not conform to the actual production process sequence of the coating process is 0, the check is deemed valid; otherwise, the check is deemed invalid. If the verification is successful, the source of the coating quality defect will be output based on the preset process rules and correlation. Otherwise, a message indicating that the source of the coating quality defect is unqualified will be sent.

8. The knowledge graph-based coating process chain quality traceability method as described in claim 6, characterized in that, The specific process for verifying the sequential logic compliance of the coating traceability path is as follows: From the source tracing path of coating quality defects, extract the core time-series link nodes, which need to cover key links; Obtain the time information corresponding to the extracted core timing link nodes, and verify whether the time sequence of each core timing link node meets the preset core timing requirements; The system counts the number of node pairs in the core timing link that do not meet the preset core timing requirements. If the number of node pairs that do not meet the preset core timing requirements is 0, the system outputs the final traceability result of the coating quality defect based on the preset process rules and relationships. Otherwise, the system sends a prompt that the traceability result of the coating quality defect is unqualified.

9. A knowledge graph-based coating process chain quality traceability system, employing the knowledge graph-based coating process chain quality traceability method as described in any one of claims 1-8, characterized in that, include: The following modules are included: dynamic process parameter monitoring module for film coating, time-series causal monitoring module for film coating patterns, and traceability reasoning and source path verification module for film coating quality. The coating dynamic process parameter feature monitoring module is used to perform coating dynamic process parameter curve feature analysis to determine whether to take time-series curve feature optimization and coating timestamp consistency calibration. If taken, after the optimization is completed, it is judged whether the coating dynamic process parameter curve feature analysis is qualified. If qualified, the coating spectrum time-series causal determination is performed. If unqualified, a failure prompt is sent. The coating pattern temporal causality monitoring module is used to determine the coating pattern temporal causality if no action is taken, in order to determine whether to trigger the supplementation of pattern temporal constraints and the completion of correlation relationships. If triggered, after the trigger ends, it is determined whether the coating pattern temporal causality determination is qualified. If qualified, the coating process quality traceability reasoning is performed. If unqualified, a pattern completion failure prompt is sent. The coating quality traceability reasoning and traceability path verification module is used to perform coating process quality traceability reasoning if it is not triggered, and to perform coating traceability path timing logic qualification verification after the coating process quality traceability reasoning is completed. If the verification is qualified, the final coating quality defect traceability result is output. If the verification is unqualified, a coating quality defect traceability result unqualified prompt is sent.

Images