Process industry quality intelligent tracing and process optimization method and system

Abstract

The application provides a process industry quality intelligent tracing and process optimization method and system, which is applied to the technical field of industrial production quality control and process optimization, realizes rapid diagnosis of quality abnormalities and accurate positioning of process causes by constructing a full-link automatic tracing model from external feedback data to production process parameters. The traditional quality control method is converted into a structured analysis path by using a digital lean rule engine, completely eliminating the dependence on manual experience. At the same time, by introducing a self-learning process strategy library and a closed-loop feedback mechanism, the system can automatically update and optimize the strategy according to the implementation effect of the scheme, realize the transition of quality control from post-processing to prediction and prevention, and greatly improve the production quality management efficiency and process robustness of the process industry.

Description

Technical Field

[0001] This application relates to the field of industrial production quality control and process optimization technology, and in particular to a method and system for intelligent quality traceability and process optimization in the process industry. Background Technology

[0002] In process industries, production processes are characterized by continuity, significant dynamic fluctuations in process parameters, and a strong correlation between finished product quality and the process. Current quality management models primarily rely on manual experience and paper records. Data from each stage (such as customer complaints, finished product inspections, and process parameters) is isolated, lacking an effective link from customer feedback to the root cause of the process issues. When handling quality anomalies, frontline personnel often need to manually switch between multiple business systems and compare data. Due to the lack of standardized analysis paths and effective auxiliary tools, the same type of quality problem recurs, making it difficult to develop standardized process response strategies. This inefficient, subjective, and fragmented management approach prevents enterprises from achieving fully automated closed-loop management of the quality control process, significantly impacting the response speed and accuracy of quality problem resolution.

[0003] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0004] In view of the shortcomings of the prior art, this application provides a method and system for intelligent traceability and process optimization in the process industry, which aims to solve the problems of inefficient and repetitive quality control caused by the isolation of quality and process data and the lack of automated root cause traceability mechanism in the process industry.

[0005] Firstly, a method for intelligent quality traceability and process optimization in process industries, the method comprising the following steps: S1: Obtain external feedback data, wherein the external feedback data includes an associated index identifier; S2: Retrieve the corresponding quality inspection data based on the associated index identifier, and identify abnormal quality indicators from the quality inspection data; S3: Call the preset quality analysis rule base to perform logical analysis on the abnormal quality indicators and determine the logical analysis path of the quality problem; S4: Based on the logical analysis path, determine the target process segment corresponding to the logical analysis path, and use the production period corresponding to the associated index as the time window to retrieve the production process parameter data of the target process segment from the production process control system; S5: Perform time-series correlation analysis on the production process parameter data and the abnormal quality indicators to locate the root cause of the quality abnormality; S6: Match the root causes of the process with the process optimization strategy library and output the corresponding standardized adjustment scheme; S7: Collect quality inspection data after implementing the standardized adjustment scheme, and iteratively update the process optimization strategy library based on the quality improvement effect.

[0006] Furthermore, step S1 includes: S11: Receive unstructured customer complaint information and identify the customer complaint information as the external feedback data; S12: Extract the order number, complaint type, and product specifications from the customer complaint information, and determine the order number as the associated index identifier.

[0007] Furthermore, step S2 includes: S21: Using the associated index identifier as the query key, retrieve the finished product quality inspection report of the corresponding production order from the quality inspection system. The finished product quality inspection report contains the measured values ​​of multiple key quality indicators and the corresponding preset specification limits. S22: Compare the measured values ​​of each key quality indicator with the corresponding preset specification limits item by item, and calculate the pass rate of each quality indicator; S23: Mark quality indicators with a pass rate lower than a preset threshold and / or measured values ​​exceeding the preset specification limits as abnormal quality indicators.

[0008] Furthermore, the quality analysis rule base includes a general lean quality analysis rule engine, which is a configurable and pluggable digital component. Step S3 includes: S31: Call the preset quality analysis rule base to perform logical analysis on the abnormal quality indicators and determine the logical analysis path of the quality problem; wherein, the quality analysis rule base contains digital constraint rules transformed from lean quality tools and methods; the lean quality tools and methods include the seven QC tools, the 5 Whys analysis method, the cause-effect diagram method, and the Six Sigma DMAIC method; S32: Invoke the digital constraint rules to perform structured analysis on the abnormal quality indicators, automatically execute logical reasoning and output analysis results containing the logical analysis path.

[0009] Furthermore, step S32 includes: S321: Use the control chart method from the seven QC tools to generate a control chart based on the measured data of the abnormal quality indicators, and identify the fluctuation patterns and process runaway states of the abnormal quality indicators; S322: Based on the identified fluctuation pattern and process runaway state, the 5 Whys analysis method is invoked to automatically generate a multi-round questioning logic chain according to the deviation direction and deviation magnitude of the abnormal quality indicators, and to locate the direct cause of the quality defect layer by layer; S323: Use the cause-effect diagram method to construct causal branches related to the direct cause from six dimensions: people, machine, material, method, environment, and measurement, in order to identify the degree of contribution of each dimension factor to the abnormal quality index, and mark the dimension factor with the highest degree of contribution as the main influencing branch; S324: Invoke the Six Sigma DMAIC method, and in accordance with the phase sequence of definition, measurement, analysis, improvement, and control, incorporate the analysis conclusions of the multi-round questioning logic chain and the main influencing branches into the DMAIC framework to generate the analysis results that include problem definition, data measurement, root cause analysis, improvement schemes, and control thresholds.

[0010] Furthermore, step S4 includes: S41: Using the order number corresponding to the associated index identifier as an index, retrieve the process log, equipment operation data and alarm records for the corresponding production period from the production process control system; The production process parameter data includes the process log, equipment operation data, and alarm records.

[0011] Furthermore, step S5 includes: S51: Using the production period corresponding to the associated index identifier as a time window, align the production process parameter data with the abnormal quality indicators on the time axis. S52: Perform hierarchical analysis of the production process parameter data according to work section, equipment and shift, and extract the fluctuation characteristics of the production process parameter data at each level; S53: Perform a correlation analysis on the changing trends of the fluctuation characteristics and the abnormal quality indicators, and calculate the correlation coefficient between each production process parameter and the abnormal quality indicators; S54: Identify the process parameters and fluctuation periods that are highly correlated with the abnormal quality indicators based on the correlation coefficient, and locate the root cause of the quality abnormality.

[0012] Furthermore, step S6 includes: S61: Using the abnormal process parameter type and fluctuation pattern in the process root cause as matching features, query the historical optimization cases in the process optimization strategy library that correspond to the matching features; the historical optimization cases stored in the process optimization strategy library include the following fields: abnormal process parameter type, fluctuation pattern, root cause description, standardized adjustment scheme, verified quality improvement effect and applicable operating conditions. S62: If a historical optimization case with a matching degree that meets the preset threshold is found, the standardized adjustment scheme recorded in the historical optimization case is extracted, and the adjustment parameters of the standardized adjustment scheme are adaptively corrected according to the current production conditions before being output. S63: If no historical optimization cases with a matching degree that meets the preset threshold are found, the process root cause is matched with the preset process parameter adjustment rule library to generate an initial process adjustment plan and output it. S64: The output standardized adjustment scheme or initial process adjustment scheme shall be used as the final standardized adjustment scheme.

[0013] Furthermore, step S7 includes: S71: Collect the quality inspection data of the new batch of products produced after implementing the standardized adjustment plan, and calculate the pass rate and improvement range of each quality indicator; S72: Determine whether the pass rate reaches the preset pass rate threshold and whether the improvement range meets the preset improvement range standard; S73: If the pass rate and the improvement range both meet the verification pass criteria, then the process root cause, the standardized adjustment scheme and the quality improvement effect are associated and stored in the process optimization strategy library as historical optimization cases; S74: If the pass rate or the improvement range does not meet the verification pass standard, a new round of analysis is triggered based on the quality inspection data of the new batch of products. Abnormal quality index identification, process parameter tracing and strategy library matching operations are re-executed until the quality improvement effect meets the verification pass standard. The final effective process adjustment plan is then updated to the process optimization strategy library.

[0014] Secondly, a process industry quality intelligent traceability and process optimization system, the system being used to implement the steps of any of the methods described above, the system comprising: Acquisition module: Acquires external feedback data, which includes associated index identifiers; Identification module: retrieves the corresponding quality inspection data based on the associated index identifier, and identifies abnormal quality indicators from the quality inspection data; Calling module: Calls a preset quality analysis rule base to perform logical analysis on the abnormal quality indicators and determine the logical analysis path of the quality problem; Retrieval Module: Retrieves production process parameter data based on the production time period corresponding to the associated index identifier; The positioning module determines the target process segment corresponding to the logical analysis path based on the logical analysis path, and retrieves the production process parameter data of the target process segment from the production process control system using the production period corresponding to the associated index as the time window. Matching module: Performs time-series correlation analysis on the production process parameter data and the abnormal quality indicators to locate the root cause of the quality abnormality; Update module: Collects quality inspection data after implementing the standardized adjustment scheme, and iteratively updates the process optimization strategy library based on the quality improvement effect.

[0015] Beneficial Effects: This application proposes a method and system for intelligent quality traceability and process optimization in the process industry. By constructing a fully automated traceability model from external feedback data to production process parameters, it achieves rapid diagnosis of quality anomalies and precise location of process root causes. Utilizing a digital lean rule engine, traditional quality control methods are transformed into structured analysis paths, completely eliminating reliance on manual experience. Simultaneously, by introducing a self-learning process strategy library and a closed-loop feedback mechanism, the system can automatically iterate and update optimization strategies based on the implementation effects of solutions, shifting quality control from post-processing to predictive prevention, significantly improving the efficiency of production quality management and process robustness in the process industry. Attached Figure Description

[0016] Figure 1 This is a flowchart of a process industry quality intelligent traceability and process optimization method proposed in this application.

[0017] Figure 2 This is a schematic diagram of a process industry quality intelligent traceability and process optimization system proposed in this application.

[0018] Labeling Explanation: 201. Acquisition Module; 202. Identification Module; 203. Calling Module; 204. Retrieval Module; 205. Location Module; 206. Matching Module; 207. Update Module. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. The components of the embodiments of this application described and marked in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] In process industries (such as chemical fibers, chemicals, food, and pharmaceuticals), production processes are generally characterized by continuous production, dynamic fluctuations in process parameters, and a strong correlation between finished product quality and the process. When product quality issues arise, they often first manifest as external feedback. However, in traditional production management models, external feedback data, internal quality inspection data, and production process parameter data are typically stored in different business systems, isolated from each other and lacking end-to-end traceability. This makes it difficult to quickly and accurately trace the specific process anomaly from massive amounts of production data when facing quality anomalies. Furthermore, adjustments to identified process problems often rely on manual experience, lacking standardized anomaly response strategies, leading to recurring issues and hindering the formation of a continuous improvement loop. To address the traceability difficulties caused by data silos and the lack of a standardized process optimization loop, a mechanism capable of achieving end-to-end automated correlation and intelligent optimization is needed.

[0022] For this, please refer to Figure 1 This application proposes a method for intelligent traceability and process optimization in the process industry, the method including the following steps: S1: Obtain external feedback data, which includes the associated index identifier; S2: Retrieve the corresponding quality inspection data based on the associated index identifier, and identify abnormal quality indicators from the quality inspection data; S3: Call the preset quality analysis rule base to perform logical analysis on abnormal quality indicators and determine the logical analysis path of quality problems; S4: Based on the logical analysis path, determine the target process segment corresponding to the logical analysis path, and use the production period corresponding to the associated index as the time window to retrieve the production process parameter data of the target process segment from the production process control system. S5: Perform time-series correlation analysis on production process parameter data and abnormal quality indicators to pinpoint the root cause of quality abnormalities; S6: Match the root causes of the process with the process optimization strategy library and output the corresponding standardized adjustment plan; S7: Collect quality inspection data after implementing the standardized adjustment plan, and iteratively update the process optimization strategy library based on the quality improvement effect.

[0023] Process industries refer to industrial sectors that involve continuous production through physical or chemical changes. In these industries, process parameters such as temperature, pressure, and flow rate have a decisive impact on the quality of the final product. For example, in the stretching and setting process of synthetic fibers, temperature fluctuations directly affect the molecular orientation; in food sterilization, the holding time and medium temperature directly affect the stability of the finished product; and in pharmaceutical compounding, stirring speed and feeding cycle time affect the uniformity of content. Therefore, quality problems in process industries are usually not caused by a single static factor, but are related to multiple continuously changing process parameters within a specific production period.

[0024] External feedback data refers to evaluations or anomaly reports obtained by manufacturers from external channels regarding product quality, user experience, and other aspects. This data can originate from customer complaint records, after-sales service work orders, email feedback, online forms, or anomaly descriptions uploaded by distributors. External feedback data is typically in text format and is characterized by its unstructured nature, inconsistent descriptive terminology, and strong business semantics. For example, descriptions such as a product being brittle and easily broken, shrinking excessively during use, or having a sticky surface, while not standard testing terms, often directly reflect quality issues exposed during final use.

[0025] A correlation index identifier is a data tag that can uniquely identify a specific production task or product batch across different business systems (such as customer complaint systems, quality systems, and production systems). This identifier can be one of the following: order number, production batch number, roll number, work order number, or batch code, or a combined identifier formed by a combination of these fields. The purpose of using this identifier is to unify and anchor customer complaint information, inspection information, and production information that were originally scattered across different systems to a single object, thereby ensuring accurate and unconfused data correspondence during subsequent traceability processes.

[0026] A quality analysis rule base refers to a pre-built digital collection containing various logical reasoning rules and analysis paths for quality issues, used for structured analysis of abnormal phenomena. This rule base does not merely store simple problem-cause correspondences, but can further store hierarchical analysis content including abnormal quality indicators, potentially affecting processes, priority investigation parameters, and auxiliary verification conditions. For example, when an abnormal quality indicator manifests as insufficient strength, the rule base can prioritize pointing to related processes such as stretching, heat setting, and cooling setting; when an abnormal quality indicator manifests as abnormal shrinkage, the rule base can prioritize pointing to processes such as thermal history control, cooling stability, and tension matching status. This structured rule organization method provides clear direction for subsequent analysis, rather than blindly searching through all process parameters.

[0027] A process optimization strategy library is a database that stores the root causes of historical process anomalies and their corresponding standardized adjustment solutions. It has the function of recording and guiding subsequent production adjustments. The content stored in this strategy library may include: the name of the process root cause, the applicable product type, the applicable process segment, the suggested adjustment action, the preconditions for execution, the quality improvement results after execution, and whether its effectiveness has been verified. Through continuous accumulation and updating, the process optimization strategy library can gradually accumulate standardized handling experience applicable to different product specifications and different anomaly types, enabling process adjustments to gradually shift from relying on individual experience to relying on reusable digital knowledge.

[0028] In a specific embodiment, the core technical means for obtaining external feedback data in step S1 can be achieved by manually entering the text record of external feedback into the system, or by automatically capturing the title and body of the feedback email as external feedback data through the email system.

[0029] To ensure a stable transition to subsequent traceability processes, the acquired external feedback data should ideally be stored in a unified data structure, including at least the feedback source, receipt time, original text content, and associated index identifier field. If the external feedback initially does not provide an associated index identifier in a structured field format, the original text can be retained first, and the order number, batch number, or work order number can be identified from the text in subsequent information extraction steps.

[0030] Taking email capture as an example, the system can monitor new emails in a specified after-sales mailbox and write the email subject, body, attachment name, and receipt time into an external feedback data table. If keywords such as "order," "batch," or "work order" appear in the body, the system will further extract the continuous encoded string that follows as a candidate association index identifier and compare and verify it with the encoding format in the enterprise's order master data.

[0031] If the verification passes, the code is written to the associated index identifier field; if the verification fails, the feedback is marked as pending confirmation to prevent the error identifier from entering the subsequent traceability chain. In this way, step S1 not only completes the receipt of external feedback, but also completes the preparation of the basic anchor points required for subsequent cross-system association.

[0032] Regarding the core technical means of retrieving quality inspection data and identifying anomalies in step S2, the corresponding quality inspection result table can be directly retrieved from the database based on the associated index identifier. Unqualified inspection items can be filtered out as abnormal quality indicators by manually setting a simple threshold.

[0033] To make this step more feasible, preferably, the quality inspection result table includes at least the fields of inspection object identification, inspection time, inspection item name, measured value, judgment result and corresponding specification requirements.

[0034] After receiving the associated index identifier, the system first searches for the inspection record that matches the identifier in the quality inspection system. If the same order corresponds to multiple inspection batches, it matches them according to the production completion time, inspection time, or finished product warehousing time, and prioritizes the inspection report that matches the shipment batch corresponding to the external feedback.

[0035] Subsequently, the system performs item-by-item judgment on each test item: for indicators with clearly defined upper and lower limits, it determines whether the measured value falls within the allowable range; for indicators with target value requirements, it determines whether the measured value meets the company's preset judgment rules; for cases where the same indicator has multiple sampling results, it summarizes all test records for that indicator in that batch, counts the number of those judged as qualified and unqualified, and forms the overall compliance status of that indicator. If an indicator significantly exceeds the limit, or if the proportion of unqualified records for that indicator in that batch reaches the preset judgment conditions, it is identified as an abnormal quality indicator.

[0036] In this way, step S2 can not only identify explicit anomalies in a single detection, but also identify stability anomalies at the batch level, thereby avoiding making one-sided judgments based on a single record.

[0037] Regarding the core technical means of calling the quality analysis rule base for logical analysis in step S3, the quality analysis rule base can contain a pre-set decision tree model. By inputting abnormal quality indicators, the decision tree model makes judgments layer by layer according to the preset node conditions, and finally outputs a logical analysis path pointing to a specific process link.

[0038] To describe this step more clearly and completely, the process of building a quality analysis rule base may include the following: First, collect historical quality anomaly cases, with each case recording at least the abnormal quality indicators, the finally confirmed process root cause, the process sections involved, key process parameters, and the handling results; then, process management personnel and quality management personnel classify and organize these cases, extracting judgment rules for which process sections are usually associated with a certain type of abnormal indicator and which type of parameter should be investigated first under what accompanying phenomena; finally, these rules are entered into the rule base in a tree-like hierarchy.

[0039] Taking the production scenario of packing tape as an example, if the input abnormal quality indicators are abnormal strength and abnormal boiling water shrinkage rate, the rule base can first determine whether these two indicators occur at the same time. If they occur at the same time, the branch related to the shaping process will be entered first. Under this branch, it will be determined whether the abnormal performance is more inclined to insufficient dimensional stability or insufficient mechanical properties. If both exist at the same time, the cooling and shaping process will be further pointed to as the priority for investigation.

[0040] The logical analysis path does not directly give the final root cause, but rather provides an analysis route that gradually converges from the abnormal phenomenon to the key process section, such as "abnormal strength / abnormal shrinkage rate → shaping-related links → cooling stability investigation priority".

[0041] By pre-building such a rule base, scattered experience-based judgments can be transformed into repeatable digital analysis paths, making subsequent production process parameter retrieval more targeted.

[0042] Regarding the core technical means of retrieving production process parameter data in step S4, the system first determines the target process segment pointed to by the logical analysis path output in step S3. For example, when the logical analysis path points to the cooling and shaping stage, the cooling and shaping stage is determined as the target process segment; when the logical analysis path points to the stretching stage, the stretching stage is determined as the target process segment.

[0043] After identifying the target process segment, to ensure that the retrieved results accurately correspond to the actual production process, the system queries the production execution system or order scheduling system for the corresponding production event records such as start-up time, roll change time, downtime, and completion time based on the associated index identifier. This determines the actual production time period corresponding to the order. If the same order spans multiple consecutive production segments, the segments are merged into a complete production time period for the order, or the data from each segment is extracted separately and then analyzed uniformly.

[0044] Subsequently, the system uses a defined production period as a time window and prioritizes retrieving production process parameter data related to the target process segment from the production process control system, rather than indiscriminately exporting all parameters from the entire plant. For example, when the logical analysis path points to the cooling and shaping stage, it prioritizes retrieving data directly related to that stage, such as cooling water tank temperature, circulating medium temperature, relevant pump operating status, linear speed, tension setpoint, and temperature records before and after the shaping stage; when the logical analysis path points to the stretching stage, it prioritizes retrieving data directly related to that stage, such as stretching roller temperature, stretching ratio setting, and roller speed synchronization status.

[0045] The retrieved data should ideally retain the timestamp, parameter name, parameter value, acquisition device identifier, and status identifier for subsequent time alignment and abnormal fluctuation identification. If both real-time sampled values ​​and system-calculated statistical values ​​exist in the historical database, the original sampled values ​​should be retained first, supplemented by statistical results summarized by time window for subsequent analysis of parameter fluctuation characteristics.

[0046] In this way, step S4 can accurately match the logical analysis path, target process segment, production period, and specific parameters, providing real, continuous, and targeted underlying data for root cause localization in step S5.

[0047] Regarding the core technical means of performing time-series correlation analysis on production process parameter data and abnormal quality indicators to locate the root cause of the process in step S5, in actual processing, the system first organizes the production process parameter data retrieved in step S4 before entering the time-series correlation analysis process. Specifically, firstly, parameter data from different sources are aligned according to a unified time benchmark to eliminate the misalignment caused by different sampling periods; then, abnormal records such as equipment downtime, sensor malfunction, and obviously invalid values ​​are removed; then, focusing on the production batch corresponding to the abnormal quality indicators, parameter features that can characterize the process status are extracted, such as the average level of a certain parameter during the production period, fluctuation range, period characteristics of continuous high or low values, number of abrupt changes, and synchronous changes with other key parameters.

[0048] After data processing, the system performs correlation analysis on the fluctuation characteristics and the changing trends of the abnormal quality indicators, calculating the correlation coefficient between each production process parameter and the abnormal quality indicators. Based on this, the system compares the analysis results with historical normal batches and historical abnormal batches, selecting parameters that are consistent with the changing trends of the abnormal quality indicators, have significant correlation coefficients, and have clear process significance within the process segment pointed to by the logical analysis path.

[0049] Taking the packaging strap scenario as an example, if both strength and boiling water shrinkage rate are abnormal, and the rule base has narrowed the investigation focus to the cooling and shaping stage, the system will focus on analyzing whether the cooling-related temperature parameters continuously deviate from the stable range, fluctuate frequently, or are not synchronized with changes in linear velocity during the production period of this order. The system calculates the correlation coefficient between the cooling water tank temperature and the strength index. If the correlation coefficient is significant and the fluctuation characteristics repeatedly appear in similar historical abnormal cases, but are absent or significantly reduced in normal batches, the system will identify the process parameters and fluctuation periods that are highly correlated with the abnormal quality indicators based on the correlation coefficient, and locate the drastic fluctuation of the cooling and shaping water tank temperature as the root cause of the process.

[0050] Therefore, step S5 does not draw conclusions based on a single statistical indicator, but rather combines time axis alignment, hierarchical analysis, fluctuation feature extraction, correlation analysis, correlation coefficient calculation, historical comparison, and process semantic interpretation to jointly complete the root cause localization.

[0051] Regarding the core technical means of matching the output scheme of the process optimization strategy library in step S6, the root cause of the process can be input into the process optimization strategy library, and the corresponding strategy text containing the root cause can be found by keyword matching, and then output as a standardized adjustment scheme.

[0052] To improve the accuracy of matching results, preferably, each strategy record in the process optimization strategy library includes at least the following fields: process root cause name, applicable product specification range, applicable process segment, suggested adjustment action, execution conditions, historical verification results, and remarks. During matching, the system can not only directly match using the process root cause name, but also perform secondary filtering by combining product specifications, abnormal quality indicator types, and the current production line identifier, avoiding the incorrect output of historical strategies that are not applicable to the current scenario.

[0053] For example, when the root cause of a process problem is identified as drastic temperature fluctuations in the cooling and shaping bath, the system can first retrieve all strategy records related to cooling and shaping, temperature fluctuations, and insufficient shaping stability, and then filter out strategies applicable to the current product specifications and production line. The output standardized adjustment plan should preferably include explicit descriptions of the actions to be taken, rather than simply providing abstract suggestions.

[0054] For example, it can output information such as checking the stability of the cooling medium circulation, confirming that the temperature regulation unit is in closed-loop control, rechecking the upper and lower limits of the temperature control of the cooling shaping section according to the requirements of the current product process card, switching to automatic stable control mode if there is frequent manual intervention, and inspecting the equipment at the source of temperature fluctuations and recording the processing results.

[0055] If multiple candidate solutions exist in the strategy library, they can be sorted according to historical verification effectiveness, number of times they have been applied, or the time of the most recent verification, prioritizing the output of solutions with more stable verification results. In this way, step S6 can directly transform the root cause localization results into executable process adjustment actions, reducing arbitrariness in the handling process.

[0056] Regarding the core technical means of collecting data and iteratively updating in step S7, after the adjustment plan is implemented on the production line, the quality inspection data of the new batch of products can be collected periodically, the average pass rate can be calculated, and if the pass rate improves, the adjustment plan can be marked as effective and saved in the strategy library.

[0057] To ensure that this step forms a truly closed-loop update mechanism, preferably, after implementing the standardized adjustment plan, the system first records the plan's execution time, execution object, execution content, and the corresponding abnormal quality indicators and process root causes before execution; then, new quality inspection data is collected in one or more subsequent continuous production batches and compared with the quality performance of similar batches before execution.

[0058] The effectiveness of quality improvement can be judged comprehensively from multiple dimensions, such as whether abnormal quality indicators have returned to the specified range, whether the proportion of non-conforming records of the indicator has decreased, whether similar problems have decreased in external feedback, and whether the corresponding process parameters have returned to stability.

[0059] If the aforementioned improvement phenomena continue to occur, the correspondence between "process root cause - adjustment plan - improvement result" will be written into the process optimization strategy library, and the strategy will be marked as verified as effective. If the improvement is not obvious, or if the same anomaly reappears after only a short period of improvement, the strategy will be marked as pending review, and the execution record will be retained for subsequent revision.

[0060] Furthermore, the iterative updates to the strategy library go beyond simply saving a new record; they also synchronously update the applicable conditions for the strategy, such as applicable product specifications, applicable abnormal combination types, and applicable production line status. In this way, step S7 allows the strategy library to continuously accumulate effective experience validated in actual production, thereby making the handling of subsequent similar problems faster, more standardized, and more reliable.

[0061] For a specific example, in a packaging tape production scenario, the system obtains external feedback data, which includes the associated index identifier order number SDX260311091. Based on this order number, the system retrieves the corresponding quality inspection data across systems and identifies strength and boiling water shrinkage rate as abnormal quality indicators. Next, the system calls a preset quality analysis rule base to perform logical analysis on these two abnormal indicators, determining that the quality problem mainly points to the logical analysis path of the cooling and shaping stage.

[0062] Subsequently, the system retrieves production process parameter data such as cooling water tank temperature and pump body temperature within the production period corresponding to the order number. By constructing a correlation analysis model between abnormal quality indicators and these temperature parameters, the system identifies the root cause of the quality anomaly as drastic fluctuations in the cooling water tank temperature. This root cause is then matched with a process optimization strategy library, and the system outputs corresponding standardized adjustment schemes, such as adjusting the temperature control threshold to 25-30℃ and adding a PID control module.

[0063] Finally, the system collects the quality inspection data of the new batch of packing tape after implementing the solution. After confirming that the pass rate of the strength index has improved, the system updates the optimization record to the process optimization strategy library.

[0064] Furthermore, to avoid directly embedding unverified specific parameters into the solution, the temperature control threshold is preferably set according to the target temperature range T1 to T2 in the process card corresponding to the current product, where T2 is greater than T1. If the existing control method involves frequent manual adjustment, the addition of a PID control module can be understood as introducing a closed-loop automatic adjustment unit to ensure that the temperature of the cooling water bath operates stably around the target range.

[0065] This approach retains the core technological concept of improving quality through stable cooling temperature while avoiding applicability deviations caused by directly applying fixed values ​​to different product specifications and equipment conditions.

[0066] To further illustrate the implementation process, the system first extracts order number SDX260311091 from customer complaint records and retrieves the corresponding finished product inspection report from the quality inspection system. The report reveals that multiple measured strength values ​​are below target requirements, and the boiling water shrinkage rate also shows abnormal deviations. Subsequently, based on the combined characteristics of abnormal mechanical properties and dimensional stability, the rule base narrows the investigation focus to the shaping-related stages. The system then extracts cooling-related parameter curves from the production history database for this order, finding frequent temperature fluctuations in the cooling shaping water tank, while normal orders of the same specifications show more stable temperature changes during the same period. Based on this, the system identifies temperature fluctuations as a priority process root cause and retrieves historically effective solutions corresponding to this root cause from the strategy library, outputting a standardized adjustment plan for the current order. After subsequent batches are executed, the system continues to collect quality inspection data. If both strength and boiling water shrinkage rate return to normal, the execution result of this plan is written into the strategy library as a new verification record. Through this complete process, the entire implementation path from external feedback triggering to cross-system tracing, root cause localization, and strategy write-back can be clearly seen.

[0067] Furthermore, step S1 includes: S11: Receive unstructured customer complaint information and identify it as external feedback data; S12: Extract the order number, complaint type, and product specifications from customer complaint information, and determine the order number as the associated index identifier.

[0068] In actual industrial production and operation, external feedback data often comes from direct customer complaints. These complaints are usually in natural language, with varying formats and complex content. Simply acquiring this raw text makes it difficult for the system to directly understand and utilize the key information for subsequent accurate traceability.

[0069] Unstructured customer complaint information refers to textual information that is not entered according to fixed fields but is expressed in a free-form descriptive manner, such as transcripts of telephone records, email bodies, online customer service conversations, and after-sales work order notes. Although the expression of this type of information is inconsistent, it usually contains key information such as order number, product specifications, anomalies, and usage scenarios. However, this information is scattered in natural language and cannot be directly used as database query conditions. Therefore, it needs to be converted into structured fields before it can serve as the starting point for subsequent cross-system tracing.

[0070] This specific implementation plan receives unstructured customer complaint information and uses information extraction technology to accurately extract core elements such as order number, complaint type, and product specifications. More specifically, the system receives an unstructured customer complaint text: "Extra-large roll customer reports severe brittleness and cracking; order number SDX260311091, specification 120080." The system uses natural language processing technology to parse this text, extracting the order number SDX260311091, the complaint type "brittleness and cracking," and the product specification 120080.

[0071] Subsequently, the system identified the extracted order number SDX260311091 as the associated index identifier for use in subsequent data concatenation.

[0072] Furthermore, the extraction process can be implemented as follows: First, the original text is preprocessed to remove meaningless spaces, unify full-width and half-width characters, and retain continuous strings of numbers and letters; then, candidate order numbers are identified based on the company's internal order coding rules, such as identifying strings composed of a fixed prefix plus continuous numbers or alphanumeric combinations; next, complaint types and product specifications are extracted by combining keyword proximity relationships, for example, extracting abnormal phenomenon phrases after descriptive words such as feedback, occurrence, and existence, and extracting specification codes after keywords such as specifications and models; finally, the extraction results are checked for consistency with the order master data and product master data to confirm whether the order number exists and whether the specification matches the order record.

[0073] If a match is found, the order number is written to the associated index identifier field; if multiple candidate order numbers exist, the order number that matches the product specifications in the complaint text is selected first, or the record is marked as awaiting manual confirmation. This step-by-step processing method allows key business information in unstructured text to be extracted reliably, rather than relying on coarse matching of a single keyword.

[0074] Furthermore, step S2 includes: S21: Using the associated index identifier as the query key, retrieve the finished product quality inspection report of the corresponding production order from the quality inspection system. The finished product quality inspection report contains the measured values ​​of multiple key quality indicators and the corresponding preset specification limits. S22: Compare the measured values ​​of each key quality indicator with the corresponding preset specification limits item by item, and calculate the pass rate of each quality indicator; S23: Mark quality indicators with a pass rate lower than a preset threshold and / or measured values ​​exceeding preset specification limits as abnormal quality indicators.

[0075] After determining the associated index identifier, it is necessary to retrieve the quality inspection data. Simply retrieving the inspection results table and performing rough threshold filtering often fails to comprehensively and accurately reflect the product's true compliance with various key quality indicators, and may easily overlook potential quality hazards.

[0076] The finished product quality inspection report can be a structured record generated by the quality inspection system according to the order, batch, roll number, or inspection report. It usually includes the name of the inspection item, inspection time, sample number, measured value, specification requirements, and judgment result. The preset specification limits are the quality judgment boundaries set by the enterprise in advance according to product standards, customer requirements, or process cards. They can be expressed as upper limits, lower limits, range requirements, or target requirements. By matching the measured values ​​with the preset specification limits item by item, the "poor quality" described by the customer can be further refined into specific, quantifiable, and verifiable abnormal indicators.

[0077] This specific implementation plan clarifies the scientific logic for judging abnormal quality indicators. For example, using order number SDX260311091 as the query key, the system retrieves the finished product quality inspection report for that order from the quality inspection system. The report includes strength as a key quality indicator, with a preset specification limit of the target value of 32.0 MPa.

[0078] The system extracts multiple sets of measured strength values ​​(e.g., 25.14 MPa to 26.72 MPa) for this batch of products and compares them item by item with the preset specification limit of 32.0 MPa. Calculations show that the pass rate for the strength index of this batch of products is 0.0%. Since the pass rate of 0.0% is far below the preset threshold, and none of the measured values ​​meet the preset specification limit, the system marks the strength as an abnormal quality indicator.

[0079] Furthermore, this step can be implemented as follows: The system first retrieves all finished product testing records corresponding to the order based on the order number; if there are multiple roll numbers or multiple batches submitted for testing under the same order, these records are merged according to the testing items; for the strength indicator, the system reads the measured value and judgment requirements of each sample record, and judges whether the specifications are met one by one; after all samples are judged, the overall compliance status of this indicator under the order is statistically analyzed.

[0080] For other indicators, such as boiling water shrinkage rate, if the specification requirements are within an allowable range, the system will also check whether each measured value falls within that allowable range and calculate the overall compliance status. If a certain indicator has individual samples exceeding the limit, but the overall compliance still meets the company's preset release rules, it can be temporarily not marked as an abnormal quality indicator; if the overall compliance of a certain indicator deviates significantly, or multiple consecutive samples exceed the limit, it will be marked as an abnormal quality indicator.

[0081] The preset threshold is preferably configured in the system by the enterprise according to the actual quality management rules. For example, it can be set separately according to different product types and different testing items, rather than being directly limited to a single fixed value, so as to ensure adaptability in different application scenarios.

[0082] Furthermore, the quality analysis rule base includes a general lean quality analysis rule engine, which is a configurable and pluggable digital component; Step S3 includes: S31: Call the preset quality analysis rule base to perform logical analysis on abnormal quality indicators and determine the logical analysis path of quality problems; the quality analysis rule base contains digital constraint rules transformed from lean quality tools and methods; lean quality tools and methods include the seven QC tools, 5 Whys analysis, cause-effect diagram method and Six Sigma DMAIC method; S32: Invoke digital constraint rules to perform structured analysis on abnormal quality indicators, automatically execute logical reasoning, and output analysis results containing the logical analysis path.

[0083] This specific implementation plan transforms traditional lean quality tools and methods into system-executable digital constraint rules. The general lean quality analysis rule engine is a low-level computing component independent of specific business logic. Its configurable and pluggable nature means that enterprises can flexibly add or remove specific analysis tool modules according to their own quality management maturity without refactoring the entire system.

[0084] To more clearly and completely describe the structure of the engine, in one specific implementation, the general lean quality analysis rule engine may include a rule configuration module, a rule parsing module, a rule execution module, a result output module, and a version management module.

[0085] The rule configuration module is used to input rule items corresponding to different lean quality tools, such as input fields, trigger conditions, analysis levels, output templates, and call priorities; the rule parsing module is used to parse configuration files or rule scripts into system-recognizable execution instructions; the rule execution module is used to retrieve abnormal quality indicators, historical quality data, production context information, and preset analysis templates in a predetermined order; the result output module is used to generate visualized or structured analysis results; and the version management module is used to record the history of rule additions, deletions, and modifications, ensuring that the specific rule version used in a particular analysis can be traced during subsequent reviews.

[0086] When transforming the seven QC tools, the 5 Whys analysis method, the cause-effect diagram method, and the Six Sigma DMAIC method into digital constraint rules, they are not simply written as a number of conditional judgment statements. Instead, the prerequisites for using each tool, the input content, the reasoning process, and the output format are standardized and defined.

[0087] Specifically, the different tools in the seven QC tools can be used in different scenarios. For example, when abnormal quality indicators have continuous time series data, the control chart method or Pareto chart analysis module should be called first; when it is necessary to sort the frequency of multiple defect categories, the Pareto chart method should be called first; when it is necessary to describe the discrete distribution of a certain process parameter, the histogram analysis module can be called.

[0088] The digital constraint rules corresponding to the 5 Whys analysis method can be pre-configured as multi-level questioning templates. Each level of questioning includes at least the current phenomenon description, the conclusion of the previous level, the optional cause category, the evidence verification field, and whether to continue drilling down.

[0089] The digitization rules of the cause-and-effect diagram method can preset six dimensions—human, machine, material, method, environment, and measurement—as main branches, and configure several expandable secondary cause labels and evidence source labels for each main branch. For example, the machine dimension can include subcategories such as temperature control units, transmission components, actuators, and detection elements; the method dimension can include subcategories such as process formulas, setpoint switching, operating procedures, and changeover strategies; and the environment dimension includes the potential impact of external conditions and auxiliary systems on product quality, such as temperature, humidity, cleanliness, vibration, noise, lighting, and the stability of the supply of utilities (such as water, electricity, gas, and steam).

[0090] The Six Sigma DMAIC method can be defined as a master control process template, which receives the intermediate results output by the aforementioned analysis tools and merges them according to the stages of definition, measurement, analysis, improvement, and control.

[0091] In one specific embodiment, the digital constraint rules can be first compiled by process experts, quality engineers, and IT implementation personnel with production quality experience to sort out the existing quality analysis process and form a rule collection list.

[0092] The rule collection list can include the rule name, applicable anomaly type, required input items, optional input items, trigger conditions, exclusion conditions, output fields, and manual confirmation method. These rules are then entered into the rule management platform in the form of a structured configuration file. For example, "When multiple consecutive detection points of abnormal quality indicators deviate from the target range and the deviation direction is consistent, prioritize calling the control chart method to identify whether the process is out of control" can be configured as a rule item; and "When the control chart method identifies the process as out of control, automatically enter the first round of follow-up questions using the 5 Whys analysis method" can be configured as the next connecting rule.

[0093] After a small-scale trial run, the conclusions of manual review during the trial period are compared with the conclusions of system reasoning. If the consistency between the logical analysis path output by the system and the manual conclusions meets the preset requirements, the corresponding rule version is solidified as a releasable version and included in the quality analysis rule base. The rule base obtained in this way is not statically established all at once, but can be continuously revised in subsequent production processes based on newly added defect types, process adjustment records, and analysis and review results.

[0094] Regarding information acquisition, before the system calls digital constraint rules to analyze abnormal quality indicators, it needs to obtain the abnormal quality indicators themselves and the basic data supporting the analysis. Abnormal quality indicators can originate from laboratory testing systems, online quality testing equipment, or finished product sampling and data entry terminals.

[0095] For example, in the context of chemical fiber production, the breaking strength can be obtained from the quality inspection results recorded after the laboratory performs tensile tests on the retained fiber samples; in the context of chemical production, the low viscosity can be obtained from the real-time detection data of the online viscometer or the laboratory retest data.

[0096] To ensure the traceability of the analysis, abnormal quality indicators are usually accompanied by context fields such as batch number, order number, inspection time, inspection location, inspection method, target range, deviation direction and deviation magnitude before entering the general lean quality analysis rule engine.

[0097] The rule execution module uses these fields to determine the call path. For example, when the detection time corresponds to a stable production batch and the deviation persists, the structured root cause analysis process is entered first. When the indicator is only occasionally abnormal at a single point and lacks sufficient context, the abnormality confirmation process can be entered first to avoid misjudgment due to insufficient input information.

[0098] In one specific embodiment, when the system identifies the breaking strength of a batch of chemical fiber products as an abnormal quality indicator, the general lean quality analysis rule engine is triggered. Here, identification can mean that the quality inspection system compares the measured breaking strength of the batch of samples with the corresponding product specifications, determines that it is below the acceptable threshold, and then sends an abnormal event record containing the order number, batch number, inspection time, product model, inspection equipment number, and deviation direction to the quality analysis rule base.

[0099] After receiving the abnormal event record, the system first performs a data integrity check to confirm that the abnormal event has at least the quality indicator name, detection result value, target range, detection time, and associated index identifier; if any key field is missing, the system will first proceed to the data completion or manual confirmation process.

[0100] After the data integrity requirements are met, the system first calls the QC seven tools module, which is converted into digital constraint rules, to perform Pareto chart analysis on the historical fracture strength data of the batch of products to confirm whether the anomaly is a frequent problem or an occasional problem.

[0101] In practice, records of fracture strength anomalies for the product model within a preset period can be retrieved from the quality database and statistically summarized according to defect category, shift, equipment location, or work section affiliation. The Pareto chart analysis module generates a defect frequency ranking list based on the statistical results. If the cumulative frequency of fracture strength anomalies within the preset period is among the highest, it is marked as a priority issue for in-depth investigation; if it only occurs in a few batches, an occasional anomaly label can be added so that subsequent analysis can focus more on instantaneous operating conditions or temporary intervention factors. The preset period can be configured as a week, several days, or several batches according to the company's production rhythm, and is not limited to a fixed value.

[0102] Subsequently, the engine automatically invokes the digital rules of the cause-and-effect diagram method to conduct a structured analysis from six dimensions: people, machine, material, method, environment, and measurement, and identifies machine (such as the temperature of the stretching roller) and method (such as the setting of the stretching ratio) as potentially high-risk dimensions.

[0103] Specifically, the system first retrieves basic information corresponding to the batch from the equipment management system, process formulation system, production execution system, and quality database. This information includes production shift, operator, equipment number, raw material batch number, process settings, on-site alarms, and testing instrument status. Then, the system categorizes different information into six dimensions—human, machine, material, method, environment, and measurement—according to a preset dimensional mapping relationship. For example, the temperature of the stretching roller, the synchronization status of the roller speed, and the working status of the heating unit can be categorized as machine; the stretching ratio setting, the cooling zone switching strategy, and the parameter curve of the shaping section can be categorized as method; the raw yarn batch, oil content, and pretreatment status can be categorized as material; the ambient temperature and humidity can be categorized as environment; and the calibration status of the testing fixture and the calibration record of the testing machine can be categorized as measurement.

[0104] For each dimension, the system will further verify whether there is any abnormal evidence adjacent to the time of abnormal fracture strength, such as sudden changes in set values, increased parameter fluctuations, maintenance overdue, alarm triggering, raw material switching, etc. If a dimension simultaneously meets the conditions of having abnormal evidence and having a high frequency of association with the quality problem in history, then that dimension is upgraded to a high-risk dimension.

[0105] Next, the engine further invokes the rules of the 5 Whys analysis method to automatically generate a follow-up logic chain for abnormal temperature of the stretching roll, and finally outputs a logic analysis path pointing to the temperature control unit of the stretching and shaping section.

[0106] Specifically, the 5WHY analysis module does not mechanically follow up with five fixed questions, but rather drills down layer by layer based on the sufficiency of the evidence. For example, the first layer might ask why the fracture strength decreased. Based on the identified high-risk dimensions, the system suggests the possibility of insufficient molecular orientation during the stretching process. The second layer might ask why the molecular orientation was insufficient. The system further combines the stretching roll temperature curve and stretching ratio setting records to conclude that the stretching roll temperature did not stably reach the target range. The third layer might ask why the stretching roll temperature did not stably reach the target range. The system reads equipment inspection records and control signal trends to determine if it is related to sluggish response of the temperature control unit, abnormal operation of the heating actuator, or deviation of the feedback measurement point. The fourth layer might further combine alarm records and maintenance records to narrow down the issue to control deviation of the temperature control unit in the stretching and shaping section. If sufficient evidence is available, the drilling stops, and this conclusion becomes the endpoint of the logical analysis path.

[0107] The entire process requires no human intervention; the system automatically completes complex logical reasoning. "No human intervention" means that, given complete input data and published rules, the system can automatically complete data reading, rule invocation, path switching, conclusion generation, and result archiving. However, in cases of missing data, conflicting conclusions, or insufficient confidence, the system can still initiate a manual review task according to a preset mechanism to ensure the robustness of the analysis results.

[0108] Furthermore, step S32 includes: S321: Use the control chart method from the seven QC tools to generate control charts based on the measured data of abnormal quality indicators, and identify the fluctuation patterns of abnormal quality indicators and the state of process runaway. S322: Based on the identified fluctuation patterns and process runaway states, the 5 Whys analysis method is invoked to automatically generate a multi-round questioning logic chain according to the deviation direction and magnitude of abnormal quality indicators, and to locate the direct cause of quality defects layer by layer. S323: Use the cause-effect diagram method to construct causal branches related to the direct cause from six dimensions: people, machine, material, method, environment, and measurement, in order to identify the degree of contribution of each dimension factor to the abnormal quality index, and mark the dimension factor with the highest degree of contribution as the main influencing branch. S324: Invoke the Six Sigma DMAIC method, and in accordance with the phase sequence of definition, measurement, analysis, improvement and control, incorporate the analysis conclusions of the multi-round questioning logic chain and the main influencing branches into the DMAIC framework to generate analysis results that include problem definition, data measurement, root cause analysis, improvement plan and control threshold.

[0109] This specific implementation plan details the collaborative execution logic of various lean quality tools in a digital environment. First, by invoking the control chart method, the system can transform discrete measured data into intuitive statistical control charts, thereby accurately identifying whether abnormal quality indicators are in a state of random fluctuation or systematic out-of-control, and clarifying the specific pattern of fluctuation (such as continuous rise, periodic oscillation, etc.).

[0110] To achieve this step, the system first needs to acquire measured data of abnormal quality indicators. This measured data can come from online quality inspection equipment, laboratory testing systems, or manual data entry terminals. Taking online inspection as an example, the system periodically receives the test values ​​of specified product indicators through an industrial communication interface, and simultaneously records the sampling time, inspection station, batch number, and equipment status. Taking laboratory testing as an example, after the laboratory personnel complete the test, they enter the results into the laboratory information management system, and the system then synchronously acquires the results through an interface. Regardless of the source, the control chart method module performs basic data verification before processing, including removing obviously missing records, identifying duplicate uploaded records, standardizing the time format, and marking inspection station switching points to avoid loss of control and distorted judgments due to dirty data.

[0111] During the implementation of the control chart method, the system can automatically select the appropriate control chart template according to the type of quality indicator. For example, for indicators such as viscosity, thickness, and breaking strength that are sampled continuously in chronological order, a control chart template oriented towards continuous variables can be used; for the number of defective items or defects counted by batch, a control chart template oriented towards count data can be used.

[0112] After generating the control chart, the system does not limit itself to simply displaying the curve, but further combines preset identification rules to determine whether there is a process runaway state. The process runaway state can include the detection point deviating from the central area for a long time, drifting continuously in a single direction, abnormally amplified fluctuation amplitude in a short period of time, or obvious periodic oscillations.

[0113] The preset identification rules can be obtained by first selecting typical anomaly cases that have been confirmed by the quality team in the past, summarizing the discrimination conditions for different out-of-control behaviors, and then solidifying these conditions into rule items. For example, defining multiple consecutive detection points located on the same side of the center line as a unilateral offset type of out-of-control, and defining a sudden and significant increase in the fluctuation amplitude of adjacent detection points as a discrete aggravation type of out-of-control. After the rules are solidified, they are then verified by replaying historical normal batches. If the false trigger rate of normal batches is within an acceptable range, the control chart identification rule is then incorporated into the official version.

[0114] Next, the 5 Whys analysis method uses the aforementioned fluctuation pattern as input. Through a pre-set logical judgment tree, it automatically triggers multiple rounds of follow-up questions based on the direction and magnitude of the deviation, stripping away surface phenomena to reach the direct cause of the quality defect. The logical judgment tree can be jointly constructed using a historical defect case library and human experience. Specifically, similar quality anomalies can first be categorized according to the direction of deviation into types such as too high, too low, excessive fluctuation, and unstable. Then, under each category, common direct and indirect causes and investigation evidence confirmed by humans can be compiled.

[0115] For example, for low viscosity, common causes may include incomplete reaction, insufficient temperature, insufficient residence time, and deviation in raw material ratio; for uneven thickness, common causes may include pressure fluctuations, partial blockage of the die, unstable traction speed, and uneven cooling.

[0116] After organizing these empirical paths into a multi-layered decision tree, the system automatically selects the corresponding branch during runtime based on the type, deviation direction, fluctuation pattern, and on-site evidence of the current abnormal quality indicator. It also verifies the sufficiency of the evidence after each round of follow-up questioning. If the evidence for the current branch is insufficient, it allows a rollback and redirection to other branches to avoid misleading results from mechanical, single-path reasoning.

[0117] Subsequently, the cause-and-effect diagram method (i.e., fishbone diagram) takes the direct cause as the core and expands outwards to six dimensions: people, machine, material, method, environment, and measurement. By quantitatively evaluating the contribution of factors in each dimension, it identifies the main influencing branches, avoiding blind spots in the analysis process. The contribution level is not limited to the output of a single algorithm, but can be ranked by comprehensively considering information such as historical co-occurrence frequency, consistency of anomaly occurrence time, strength of parameter fluctuations, accompanying alarm conditions, and upstream and downstream influence relationships in the process.

[0118] For example, if equipment temperature control deviations in the "situational" dimension repeatedly occur during the abnormal period and are accompanied by related alarms, and have historically occurred multiple times alongside similar quality anomalies, then the system can assign a higher priority to the "situational" dimension. Conversely, factors that exhibit minor anomalies but are inconsistent with the time window and have weak historical correlations are ranked lower. Through this ranking, the system can compress a large number of candidate factors across the six dimensions into a few key influencing branches, facilitating focused processing when subsequently incorporated into the DMAIC framework.

[0119] Finally, the Six Sigma DMAIC method, as a macro-level framework, integrates the analytical conclusions of the preceding tools and generates a complete and logically rigorous analysis result following a standardized process of definition, measurement, analysis, improvement, and control. Specifically, the system can automatically fill in the content of each stage using preset templates during the generation process.

[0120] The process is as follows: the definition phase includes the name of the abnormal quality indicator, the abnormal batch, the deviation performance, and the business impact; the measurement phase includes the fluctuation pattern identified by the control chart method, the time period of the measured data, and the key test records; the analysis phase includes the multi-round questioning logic chain generated by the 5 Whys analysis method and the main impact branches screened by the causal relationship diagram method; the improvement phase includes placeholder items for suggested measures for the main impact branches, such as equipment inspection, parameter rollback, operation specification review, and raw material re-inspection; and the control phase includes the control thresholds, retesting requirements, and early warning conditions that need to be continuously monitored.

[0121] This template can also be configured according to the characteristics of the enterprise's industry. For example, different fields can be used for different scenarios such as chemical fiber, chemical industry, thin film, and polymerization, but the overall stage order remains unchanged to ensure that the analysis results have a unified reading method and archiving standard.

[0122] Furthermore, step S4 includes: S41: Using the order number corresponding to the associated index as the index, retrieve the process log, equipment operation data and alarm records for the corresponding production period from the production process control system; Production process parameter data includes process logs, equipment operation data, and alarm records.

[0123] This specific implementation plan clarifies the multidimensional composition of production process parameter data. Using the order number as an index, the system can accurately pinpoint the time window for a specific production task. Within this time window, the system synchronously retrieves three types of key data from the production process control system: process logs, equipment operation data, and alarm records.

[0124] Among them, the process log mainly records information such as human operation trajectory, formula issuance record, and parameter setting value modification during the production process, which can reflect the factors of law and human; the equipment operation data includes continuous physical quantities (such as temperature, pressure, flow rate, speed, etc.) collected by various sensors and the start-up and shutdown status of the equipment, which is the core carrier reflecting the factors of machine and environment; the alarm record captures abnormal events that trigger system warnings during the production process, such as over-limit alarms and interlock shutdowns, which can directly indicate the sudden instability points in the production process.

[0125] These three types of data are integrated into unified production process parameter data.

[0126] Furthermore, step S5 includes: S51: Using the production period corresponding to the associated index as the time window, align the production process parameter data and abnormal quality indicators with the time axis. S52: Perform hierarchical analysis of production process parameter data by work section, equipment and shift, and extract the fluctuation characteristics of production process parameter data at each level; S53: Conduct correlation analysis on the changing trends of fluctuation characteristics and abnormal quality indicators, and calculate the correlation coefficient between each production process parameter and the abnormal quality indicators; S54: Identify process parameters and fluctuation periods that are highly correlated with abnormal quality indicators based on correlation coefficients, and locate the root cause of the quality abnormality.

[0127] This specific implementation plan details the complete computational process for constructing a correlation analysis model and locating the root cause of the process. First, step S51 addresses the problem of data misalignment in time and space. Since quality inspection is usually delayed (e.g., sampling inspection is only carried out after the finished product has come off the production line), while production process parameters are collected in real time, it is necessary to use the production period as a benchmark, consider the dwell time of materials on the production line, and map the quality inspection results back to the corresponding production time to achieve precise alignment of the time axis.

[0128] To achieve this, the system can pre-maintain a process flow table. This table records typical flow paths, quality inspection locations, and approximate dwell times from key stages to inspection locations for different products and processes. This table can be obtained by manually calibrating it under stable production conditions, combining production cycle time records, conveyor length, buffer dwell times, and winding or packaging times, and then verifying it with historical data from several batches to form a usable configuration. When the system receives an abnormal quality indicator, it first queries the process flow table based on the product model and inspection location, then maps the inspection time forward to the candidate time window of the corresponding production stage, thereby aligning it with the production process parameter data.

[0129] Next, step S52 performs structured dimensionality reduction on the complex data. Based on the physical and management logic of work sections, equipment, and shifts, the data is stratified, and statistical fluctuation characteristics such as mean drift, variance increase, and kurtosis anomalies are extracted, transforming continuous raw values ​​into feature vectors with process characterization significance. This stratified analysis can be understood as first splitting the data into a first layer based on the physical structure of the production line, then assigning it to a second layer based on equipment objects, and finally labeling it by shift or operating time period. For example, in a film production scenario, the data can first be divided into extrusion, casting, traction, and winding sections, then further subdivided within the extrusion section into main extruders, auxiliary machines, metering pumps, etc., and finally labeled with day shifts, night shifts, or the time periods before and after shift changes. After stratification, the system extracts fluctuation characteristics for continuous parameters within each level. Taking melt pressure as an example, the system can extract the overall stability of this parameter within the alignment time window, short-term sudden changes, continuous high or low states, and fluctuation range changes; taking screw speed as an example, it can extract features such as whether there are frequent small adjustments, sudden large changes, or long-term deviations from the set value.

[0130] Through this process, the massive amount of raw sampling points that were originally difficult to compare directly are compressed into feature descriptions with clear technological significance.

[0131] Subsequently, step S53 uses statistical algorithms to quantitatively evaluate the synchronicity or lag correlation between these fluctuation characteristics and the changing trends of abnormal quality indicators. To avoid introducing unnecessary complex formulas, the correlation analysis here can be understood as follows: the system sequentially examines whether the fluctuation characteristics of each process parameter and the changing trends of abnormal quality indicators correspond in time, are consistent in direction, and echo each other in terms of intensity. If a parameter continuously exhibits abnormal fluctuations within the corresponding window before and after the occurrence of a quality anomaly, and the more obvious the fluctuation, the more obvious the deviation of the quality indicator, then the parameter is assigned a higher degree of correlation, i.e., a larger correlation coefficient.

[0132] To improve the reliability of the results, the system can also first exclude parameters that are obviously not related to the process, such as auxiliary equipment measurement points that are unrelated to the product, label fields that remain unchanged for a long time, or data points that are obviously in a shutdown state.

[0133] Furthermore, correlation analysis can be conducted at the level of a single device or at the level of a combination of work sections. For example, the combined effects of temperature deviation and pressure fluctuation on a certain type of quality anomaly can be examined simultaneously to avoid ignoring multi-factor coupling scenarios.

[0134] Finally, step S54, based on the calculated correlation coefficients, sets a threshold to filter out weakly correlated parameters, identifying highly correlated process parameters and the specific time periods during which they exhibit abnormal fluctuations, thereby definitively pinpointing the root cause of the process issue. This threshold can be configured according to different product types, quality indicators, or historical data scales, and is not limited to a fixed value.

[0135] When multiple parameters simultaneously exhibit a high degree of correlation, the system can also perform cross-validation using the logical analysis path obtained in step S3. For example, if the logical analysis path points to a temperature control unit, and the correlation analysis shows that the temperature measurement point is highly consistent with the quality anomaly within the alignment time window, the system can increase the priority of this parameter as a process root cause. Conversely, if a parameter has a high degree of correlation but no overlap with the preceding logical analysis, it can be marked as a factor to be reviewed rather than directly identified as the final root cause. This method of cross-validation between logical analysis and data analysis can further reduce the risk of misjudgment caused by spurious correlations.

[0136] Furthermore, step S6 includes: S61: Using the abnormal process parameter type and fluctuation pattern in the process root cause as matching features, query the historical optimization cases in the process optimization strategy library that correspond to the matching features; the historical optimization cases stored in the process optimization strategy library include the following fields: abnormal process parameter type, fluctuation pattern, root cause description, standardized adjustment plan, verified quality improvement effect and applicable operating conditions. S62: If a historical optimization case with a matching degree that meets the preset threshold is found, the standardized adjustment scheme recorded in the historical optimization case is extracted, and the adjustment parameters of the standardized adjustment scheme are adaptively corrected according to the current production conditions before being output. S63: If no historical optimization cases with matching degree matching the preset threshold are found, the initial process adjustment plan is generated and output based on the process root cause and the preset process parameter adjustment rule library. S64: Use the output standardized adjustment scheme or initial process adjustment scheme as the final standardized adjustment scheme.

[0137] This specific implementation plan details how to intelligently output practical process adjustment solutions based on the root causes of the process. First, in step S61, the system does not simply use a single parameter name for retrieval; instead, it combines the type of abnormal process parameter with its fluctuation pattern as a joint matching feature. For example, even with abnormal melt pressure in an extruder, the fluctuation pattern might be consistently high or exhibit high-frequency oscillations, and the corresponding solution strategies for these two patterns are drastically different.

[0138] By using joint feature queries, the system can accurately locate historical optimization cases with high similarity within the process optimization strategy library. The rich fields pre-structured and stored in the strategy library, especially the applicable operating conditions range (such as the applicable product thickness range, the applicable production line speed range, etc.), provide multi-dimensional filtering conditions for subsequent accurate matching.

[0139] Next, step S62 introduces a crucial adaptive correction mechanism. When the system finds a historical optimization case with a matching degree that meets a preset threshold (e.g., feature similarity of 80% or more), it will not directly output the historical solution unchanged.

[0140] The system extracts the core adjustment logic from historical solutions and dynamically fine-tunes it in conjunction with the current production conditions. In practice, the system can preset a series of correction coefficients or formulas. Taking film production as an example, assuming the historical optimization case was for a product with a thickness of 50 micrometers and a linear speed of 100 meters per minute, the standardized adjustment solution was to lower the temperature of the second zone of the main extruder by 5°C; while the current abnormal production condition is a thickness of 50 micrometers and a linear speed of 120 meters per minute.

[0141] The system recognizes that the current linear speed is 20% higher than historical cases. Based on the preset process mechanism correction formula that shows a positive correlation between temperature adjustment and linear speed, the system may adaptively adjust the temperature reduction to 6°C. In this way, historical experience is given the ability to dynamically adapt to current operating conditions, avoiding blind adjustments based on outdated methods.

[0142] Meanwhile, to address cold starts or rare anomalies where relevant experience has not yet been accumulated in the strategy library, step S63 provides a fallback mechanism. When no historical cases meeting the matching criteria are found, the system will switch to the preset process parameter adjustment rule library.

[0143] This rule base differs from the historical case base that records specific values. Instead, it stores general adjustment rules extracted based on process mechanism models or the prior knowledge of senior experts (such as qualitative rules that prioritize reducing the speed or increasing the front-end temperature when the pressure is too high).

[0144] Based on these general rules and the root causes of the current process, the system deduces an initial process adjustment plan that is correct in direction and conservative in magnitude. Finally, in step S64, both the historical plan that has undergone adaptive correction and the initial plan derived from the rules are uniformly formatted into standard output for execution by field operators or automated control systems.

[0145] Furthermore, step S7 includes: S71: Collect quality inspection data of new batches of products produced after the implementation of the standardized adjustment plan, and calculate the pass rate and improvement range of each quality indicator; S72: Determine whether the pass rate has reached the preset pass rate threshold and whether the improvement range meets the preset improvement range standard; S73: If the pass rate and improvement range both meet the verification pass criteria, the process root cause, standardized adjustment plan and quality improvement effect will be associated and stored in the process optimization strategy library as a historical optimization case. S74: If the pass rate or improvement does not meet the verification pass standard, a new round of analysis will be triggered based on the quality inspection data of the new batch of products. Abnormal quality index identification, process parameter tracing and strategy library matching operations will be re-executed until the quality improvement effect meets the verification pass standard. The final effective process adjustment plan will be updated to the process optimization strategy library.

[0146] This specific implementation plan clarifies the closed-loop verification and self-learning logic of the process optimization strategy. In step S71, after confirming that the adjustment plan has been implemented on-site, the system automatically marks the newly produced batches of products and collects quality inspection data for these new batches. The system not only recalculates the pass rate of previously abnormal quality indicators but also compares it with historical data before the adjustment, quantifying the specific improvement (e.g., the pass rate increases from 85% to 98%, an improvement of 13 percentage points). Simultaneously, the system monitors other relevant quality indicators to confirm whether the process adjustment has caused deterioration in other indicators.

[0147] In steps S72 and S73, the system compares the calculated pass rate and improvement range with the corresponding verification pass criteria. The verification pass criteria are that the pass rate reaches the preset pass rate threshold, the improvement range meets the preset improvement range standard, the pass rate increases to above 95%, and the improvement range remains stable for three consecutive batches.

[0148] Only when the actual improvement results strictly meet these conditions will the system determine that the adjustment plan is truly effective. Subsequently, the system packages and links the complete event chain, including the initial root cause of the process, the actual standardized adjustment plan implemented, and the quantified quality improvement results, and formally writes it into the process optimization strategy library as a high-value historical optimization case. This process realizes the explicitification and digital accumulation of tacit knowledge.

[0149] More importantly, there is the error correction and iteration mechanism described in step S74. If the quality improvement of the new batch of products does not meet expectations, it indicates that the previously identified root cause of the process may only be a symptom, or that the output adjustment plan is insufficient or misguided.

[0150] At this point, the system will not terminate the process. Instead, it will use the quality inspection data of the new batch of products with anomalies as a new input source, automatically triggering a new round of traceability and analysis. In this new cycle, the system may exclude parameters that were adjusted but were ineffective in the previous round, and instead delve deeper into the secondary influence branches ranked slightly lower in the causal relationship graph method, or re-evaluate other candidate parameters in the correlation analysis. This cycle of execution, verification, feedback, and re-analysis will continue until a process adjustment solution that can truly solve the problem is found and updated to the strategy library.

[0151] The above technical solutions ensure that every piece of experience accumulated in the process optimization strategy library is a practical strategy verified by real production data, eliminating the misleading influence of ineffective or erroneous strategies on subsequent production. Simultaneously, the mechanism of automatically triggering a new round of analysis when verification fails endows the system with strong fault tolerance and a self-learning ability to continuously approach the essence of the problem. As the system operates for longer and handles more cases, the process optimization strategy library will become increasingly rich and accurate, and the enterprise's quality management level will gradually leap from reactive remediation to rapid and accurate troubleshooting based on a reliable knowledge base.

[0152] Please refer to Figure 2 This application also proposes a process industry quality intelligent traceability and process optimization system, which is used to implement the steps of any of the above methods. The system includes: Module 201: Acquires external feedback data, which includes the associated index identifier; Identification module 202: retrieves the corresponding quality inspection data based on the associated index identifier, and identifies abnormal quality indicators from the quality inspection data; Module 203: Calls the preset quality analysis rule base to perform logical analysis on abnormal quality indicators and determine the logical analysis path of quality problems; Retrieval Module 204: Based on the logical analysis path, determine the target process segment corresponding to the logical analysis path, and retrieve the production process parameter data of the target process segment from the production process control system using the production period corresponding to the associated index as the time window; Location module 205: Performs time-series correlation analysis on production process parameter data and abnormal quality indicators to locate the root cause of quality abnormalities; Matching module 206: Matches the root causes of the process with the process optimization strategy library and outputs the corresponding standardized adjustment plan; Update module 207: Collects quality inspection data after implementing the standardized adjustment plan, and iteratively updates the process optimization strategy library based on the quality improvement effect.

[0153] Through the above technical solution, the system, through the coordinated operation of its various functional modules, fully maps the aforementioned intelligent traceability and process optimization methods in both physical architecture and software logic. The acquisition module 201 and the identification module 202 break down the data barriers between external customer complaints and the internal quality system, transforming unstructured customer feedback information into structured, associated index identifiers and abnormal quality indicators.

[0154] Module 203 performs structured logical analysis on abnormal quality indicators and outputs a logical analysis path pointing to a specific process step, thus achieving initial convergence from quality phenomena to analysis guidance.

[0155] The retrieval module 204 determines the target process segment based on the logical analysis path, and uses the corresponding production period identified by the associated index as the time window to retrieve the production process parameter data of the target process segment in a targeted manner. This avoids information overload and processing redundancy caused by full data retrieval, and realizes precise drilling down from the logical analysis path to the micro-production parameters.

[0156] The positioning module 205 performs time-series correlation analysis on the retrieved production process parameter data and abnormal quality indicators. Through time axis alignment, hierarchical analysis, fluctuation feature extraction and correlation coefficient calculation, it objectively identifies the root cause of the quality abnormality, replacing traditional manual experience judgment with quantitative analysis.

[0157] The matching module 206 matches the identified process root causes with the process optimization strategy library and outputs the corresponding standardized adjustment scheme, realizing the structured reuse of historical effective experience.

[0158] The update module 207 collects quality inspection data after the implementation of the adjustment plan, and iteratively updates the process optimization strategy library based on the quality improvement effect, thus constructing a complete closed loop from plan output, effect verification to knowledge accumulation.

[0159] Through the aforementioned end-to-end automation mechanism, the system can effectively solve long-standing problems in the process industry, such as data silos, traceability difficulties, and over-reliance on human experience, providing enterprises with a highly automated, intelligent, and continuously evolving quality management and process optimization infrastructure.

[0160] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent quality traceability and process optimization in process industries, characterized in that, The method includes the following steps: S1: Obtain external feedback data, wherein the external feedback data includes an associated index identifier; S2: Retrieve the corresponding quality inspection data based on the associated index identifier, and identify abnormal quality indicators from the quality inspection data; S3: Call the preset quality analysis rule base to perform logical analysis on the abnormal quality indicators and determine the logical analysis path of the quality problem; S4: Based on the logical analysis path, determine the target process segment corresponding to the logical analysis path, and use the production period corresponding to the associated index as the time window to retrieve the production process parameter data of the target process segment from the production process control system; S5: Perform time-series correlation analysis on the production process parameter data and the abnormal quality indicators to locate the root cause of the quality abnormality; S6: Match the root causes of the process with the process optimization strategy library and output the corresponding standardized adjustment scheme; S7: Collect quality inspection data after implementing the standardized adjustment scheme, and iteratively update the process optimization strategy library based on the quality improvement effect.

2. The method for intelligent traceability and process optimization of process industry quality according to claim 1, characterized in that, Step S1 includes: S11: Receive unstructured customer complaint information and identify the customer complaint information as the external feedback data; S12: Extract the order number, complaint type, and product specifications from the customer complaint information, and determine the order number as the associated index identifier.

3. The intelligent traceability and process optimization method for process industry quality according to claim 1, characterized in that, Step S2 includes: S21: Using the associated index identifier as the query key, retrieve the finished product quality inspection report of the corresponding production order from the quality inspection system. The finished product quality inspection report contains the measured values ​​of multiple key quality indicators and the corresponding preset specification limits. S22: Compare the measured values ​​of each key quality indicator with the corresponding preset specification limits item by item, and calculate the pass rate of each quality indicator. S23: Mark quality indicators with a pass rate lower than a preset threshold and / or measured values ​​exceeding the preset specification limits as abnormal quality indicators.

4. The intelligent traceability and process optimization method for process industry quality according to claim 1, characterized in that, The quality analysis rule base includes a general lean quality analysis rule engine, which is a configurable and pluggable digital component; Step S3 includes: S31: Call the preset quality analysis rule base to perform logical analysis on the abnormal quality indicators and determine the logical analysis path of the quality problem; wherein, the quality analysis rule base contains digital constraint rules transformed from lean quality tools and methods; the lean quality tools and methods include the seven QC tools, the 5 Whys analysis method, the cause-effect diagram method, and the Six Sigma DMAIC method; S32: Invoke the digital constraint rules to perform structured analysis on the abnormal quality indicators, automatically execute logical reasoning and output analysis results containing the logical analysis path.

5. The intelligent traceability and process optimization method for process industry quality according to claim 4, characterized in that, Step S32 includes: S321: Use the control chart method from the seven QC tools to generate a control chart based on the measured data of the abnormal quality indicators, and identify the fluctuation patterns and process runaway states of the abnormal quality indicators. S322: Based on the identified fluctuation pattern and process runaway state, the 5 Whys analysis method is invoked to automatically generate a multi-round questioning logic chain according to the deviation direction and deviation magnitude of the abnormal quality indicators, and to locate the direct cause of the quality defect layer by layer; S323: Use the cause-effect diagram method to construct causal branches related to the direct cause from six dimensions: people, machine, material, method, environment, and measurement, in order to identify the degree of contribution of each dimension factor to the abnormal quality index, and mark the dimension factor with the highest degree of contribution as the main influencing branch; S324: Invoke the Six Sigma DMAIC method, and in accordance with the phase sequence of definition, measurement, analysis, improvement, and control, incorporate the analysis conclusions of the multi-round questioning logic chain and the main influencing branches into the DMAIC framework to generate the analysis results that include problem definition, data measurement, root cause analysis, improvement schemes, and control thresholds.

6. The intelligent traceability and process optimization method for process industry quality according to claim 1, characterized in that, Step S4 includes: S41: Using the order number corresponding to the associated index identifier as an index, retrieve the process log, equipment operation data and alarm records for the corresponding production period from the production process control system; The production process parameter data includes the process log, equipment operation data, and alarm records.

7. The intelligent traceability and process optimization method for process industry quality according to claim 1, characterized in that, Step S5 includes: S51: Using the production period corresponding to the associated index identifier as a time window, align the production process parameter data with the abnormal quality indicators on the time axis. S52: Perform hierarchical analysis of the production process parameter data according to work section, equipment and shift, and extract the fluctuation characteristics of the production process parameter data at each level; S53: Perform a correlation analysis on the changing trends of the fluctuation characteristics and the abnormal quality indicators, and calculate the correlation coefficient between each production process parameter and the abnormal quality indicators; S54: Identify the process parameters and fluctuation periods that are highly correlated with the abnormal quality indicators based on the correlation coefficient, and locate the root cause of the quality abnormality.

8. The intelligent traceability and process optimization method for process industry quality according to claim 1, characterized in that, Step S6 includes: S61: Using the abnormal process parameter type and fluctuation pattern in the process root cause as matching features, query the historical optimization cases in the process optimization strategy library that correspond to the matching features; the historical optimization cases stored in the process optimization strategy library include the following fields: abnormal process parameter type, fluctuation pattern, root cause description, standardized adjustment scheme, verified quality improvement effect and applicable operating conditions. S62: If a historical optimization case with a matching degree that meets the preset threshold is found, the standardized adjustment scheme recorded in the historical optimization case is extracted, and the adjustment parameters of the standardized adjustment scheme are adaptively corrected according to the current production conditions before being output. S63: If no historical optimization cases with a matching degree that meets the preset threshold are found, the process root cause is matched with the preset process parameter adjustment rule library to generate an initial process adjustment plan and output it. S64: The output standardized adjustment scheme or initial process adjustment scheme shall be used as the final standardized adjustment scheme.

9. The intelligent traceability and process optimization method for process industry quality according to claim 1, characterized in that, Step S7 includes: S71: Collect the quality inspection data of the new batch of products produced after implementing the standardized adjustment plan, and calculate the pass rate and improvement range of each quality indicator; S72: Determine whether the pass rate reaches the preset pass rate threshold and whether the improvement range meets the preset improvement range standard; S73: If the pass rate and the improvement range both meet the verification pass criteria, then the process root cause, the standardized adjustment scheme and the quality improvement effect are associated and stored in the process optimization strategy library as historical optimization cases; S74: If the pass rate or the improvement range does not meet the verification pass standard, a new round of analysis is triggered based on the quality inspection data of the new batch of products. Abnormal quality index identification, process parameter tracing and strategy library matching operations are re-executed until the quality improvement effect meets the verification pass standard. The final effective process adjustment plan is then updated to the process optimization strategy library.

10. A process industry quality intelligent traceability and process optimization system, characterized in that, The system is used to implement the steps of the method according to any one of claims 1-9 above, and the system includes: Acquisition module: Acquires external feedback data, which includes associated index identifiers; Identification module: retrieves the corresponding quality inspection data based on the associated index identifier, and identifies abnormal quality indicators from the quality inspection data; Calling module: Calls a preset quality analysis rule base to perform logical analysis on the abnormal quality indicators and determine the logical analysis path of the quality problem; Retrieval Module: Retrieves production process parameter data based on the production time period corresponding to the associated index identifier; The positioning module determines the target process segment corresponding to the logical analysis path based on the logical analysis path, and retrieves the production process parameter data of the target process segment from the production process control system using the production period corresponding to the associated index as the time window. Matching module: Performs time-series correlation analysis on the production process parameter data and the abnormal quality indicators to locate the root cause of the quality abnormality; Update module: Collects quality inspection data after implementing the standardized adjustment scheme, and iteratively updates the process optimization strategy library based on the quality improvement effect.

Images