Top-mounted ball valve production quality digital twin tracing system

Abstract

The invention discloses a top-entry ball valve production quality digital twin tracing system, and relates to the technical field of industrial quality management. The system comprises a quality core data definition module used for defining core performance indexes, corresponding data dimensions and quality judgment standards of the top-entry ball valve; the digital twinborn model building module is used for generating a comprehensive weight matrix and a dynamically optimized digital twinborn model; the multi-source data fusion processing module outputs fusion quality data to the digital twinborn model construction module; the full-process quality tracing module is used for realizing forward tracing and reverse tracing of the quality of the top-mounted ball valve based on the fused quality data and outputting a tracing result; according to the method, quality control is promoted to be converted into active early warning and data driving, and the intelligent and efficient level is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of industrial quality management technology, specifically to a digital twin traceability system for the production quality of top-mounted ball valves. Background Technology

[0002] Top-entry ball valves are key equipment in oil and gas pipelines, used to distribute and change the direction of media flow. Their sealing performance, corrosion resistance, adaptability to ultra-low temperatures of -109 degrees Celsius, and the precision of hard-seal assembly directly determine the safety of pipeline operation. Product manufacturing requires multiple precision processes and rigorous testing. The multi-source quality data generated throughout the process, including raw material information, processing parameters, assembly data, and test results, are the core basis for accurate quality traceability.

[0003] Current quality traceability technologies have prominent problems: multi-source quality data are scattered across various production systems, equipment records, and test reports, lacking unified standardization, integration, and end-to-end correlation; digital twin models are mostly designed with fixed parameters, which cannot adapt to real-time quality characteristics such as fluctuations in hard seal assembly precision and changes in ultra-low temperature test data, resulting in a disconnect between the model and the actual quality status, making it difficult to achieve accurate and efficient end-to-end quality traceability. Summary of the Invention

[0004] The purpose of this invention is to provide a digital twin traceability system for the production quality of top-mounted ball valves to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A digital twin traceability system for the production quality of top-loading ball valves includes: The core quality data definition module is used to define the core performance indicators, corresponding data dimensions, and quality judgment standards of top-loading ball valves, and outputs standardized quality data. The digital twin model construction module builds a unique twin based on standardized quality data, calculates the comprehensive weight of quality features through a dual-weight dynamic coupling algorithm, and generates a comprehensive weight matrix and a dynamically optimized digital twin model. The multi-source data fusion processing module receives standardized quality data and data from the entire production process of top-mounted ball valves. After cleaning and correlation processing, it outputs the fused quality data to the digital twin model construction module. The end-to-end quality traceability module calls the dynamically optimized digital twin model to realize forward and reverse traceability of the quality of the top-mounted ball valve based on the fused quality data, and outputs the traceability results. The quality anomaly early warning and analysis module, based on a comprehensive weight matrix, a dynamically optimized digital twin model, and fused quality data, generates quality anomaly early warning information through an early warning model, conducts correlation analysis between quality characteristics and performance, and outputs early warning results and quality analysis reports.

[0006] Preferably, the quality core data definition module includes: The performance index definition unit is used to define the core performance indexes of the top-loading ball valve. The core performance indexes include sealing performance, corrosion resistance, adaptability to ultra-low temperature environment and hard seal structure, and also support extended performance indexes. The data dimension definition unit is used to define the corresponding data dimensions for core performance indicators. The data dimensions include performance-specific data dimensions and general data dimensions. The performance-specific data dimensions cover sealing, corrosion resistance, ultra-low temperature and hard sealing adaptation scenarios. The general data dimensions include raw material information, process parameters, test parameters and production batch information. The judgment criterion definition unit is used to set judgment thresholds for each data dimension, forming a judgment criterion library. The judgment thresholds can be adjusted by the user, and the judgment criterion library is associated with standardized quality data.

[0007] Preferably, the digital twin model construction module includes a dedicated twin construction unit, which is used to construct product twins, process twins, and test twins; The product twin uses the product's unique identifier as its core and links it to standardized quality data throughout the entire lifecycle of the top-loading ball valve, forming a unique mapping relationship between the product and the twin. The process twin covers the entire production process of top-loading ball valves, linking the processing parameters, equipment operation data, and quality inspection data of each process; The test twin constructs a virtual test environment model for the specific test scenario of the top-mounted ball valve, and associates the test equipment parameters, test process data and test result data.

[0008] Preferably, the digital twin model building module further includes a dynamic coupling engine, which includes: The dual-weight calculation unit is used to calculate the feature importance weight and the quality impact weight. The formula for calculating the feature importance weight is: W 1i =(σ i / Σσ i )×0.5, where σ i Let Σσ be the real-time data variance of the i-th quality feature. i The sum of the variances of all quality characteristics; the formula for calculating the weight of quality influence is: W 2i =(F i / ΣF i )×0.5, where F i Let ΣF be the historical number of times the i-th quality characteristic causes the top-loading ball valve to fail. i The total number of nonconformities caused by all quality characteristics; The model parameter iteration unit is used to calculate the comprehensive weight based on the feature importance weight and the quality impact weight. The formula for calculating the comprehensive weight is: W i =W 1i +W 2i The model parameters are updated based on the comprehensive weights. These parameters include feature mapping coefficients, virtual simulation rules, and traceability association weights. The formula for calculating the feature mapping coefficients is: K ij =W i ×K0, where K0 is the basic mapping coefficient; The deviation verification unit is used to calculate the deviation between the prediction results of the dynamically optimized digital twin model and the actual quality results. If the deviation value is greater than 5%, the model parameter iteration unit is triggered to update the model parameters again.

[0009] Preferably, the multi-source data fusion processing module includes: The data access unit is used to access various types of production data through a preset standardized interface. These various types of production data include sensor-collected data, test equipment data, system synchronization data, and manually entered data. The data cleaning unit is used to process missing and outlier values ​​in various types of production data. The data association unit is used to establish a full-link association between multiple types of production data and standardized quality data, with the product's unique identifier as the primary association key and batch number, process ID, and test equipment number as secondary association keys, to form integrated quality data.

[0010] Preferably, the full-process quality traceability module includes a forward traceability unit, which is used to receive the product's unique identifier or batch number input by the user, call the dynamically optimized digital twin model, display the full-link quality data of the top-mounted ball valve from raw material warehousing to factory inspection in chronological order, clarify the correlation between the data of each link and the core performance indicators, and output the traceability results.

[0011] Preferably, the full-process quality traceability module also includes a reverse traceability unit. The reverse traceability unit is used to receive the description of the quality problem or the unique identifier of the non-conforming product input by the user, and based on the comprehensive weight matrix, prioritize the production link corresponding to the high-weight quality feature, clarify the root link, related data and scope of impact of the quality problem, and output the reverse traceability result.

[0012] Preferably, the quality anomaly early warning and analysis module includes an early warning rule configuration unit, which is used to configure three types of early warning rules: threshold early warning, trend early warning, and correlation early warning. Threshold alerts are used to trigger warnings when real-time quality data exceeds a judgment threshold; Trend alerts are used to trigger warnings when the trend of quality data deviates from the historical average by 10%. The correlation alert is used to trigger an alert for related quality features when high-weight quality feature data is abnormal.

[0013] Preferably, the quality anomaly early warning and analysis module also includes an early warning model construction unit. The early warning model construction unit adopts a fusion algorithm of decision tree and long short-term memory network, and completes model training using historical quality data of top-mounted ball valves as the training set. The accuracy of the early warning model is ≥92%. The early warning model building unit is also used to generate early warning information, which includes the early warning type, the unique identifier of the product involved, abnormal data, related performance indicators and early warning level, and is pushed to the designated terminal through a preset push method.

[0014] Preferably, the quality anomaly early warning and analysis module further includes a quality analysis unit, which is used to conduct correlation strength analysis between quality characteristics and performance based on a comprehensive weight matrix; The quality analysis unit is also used to statistically analyze the trends of quality data from multiple batches of products, identify patterns of quality fluctuations, analyze the root causes of quality anomalies, including raw material defects, equipment failures, process parameter deviations, and operator errors, and output quality analysis results.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By standardizing and defining core quality data and integrating multi-source data across the entire supply chain, it provides comprehensive traceability data support. Relying on a dual-weight dynamic coupling algorithm, it achieves self-optimization of the digital twin model, improving the model's adaptability to actual working conditions and the accuracy of traceability analysis. Combining forward full-chain traceability with a reverse high-weight priority positioning mechanism improves the efficiency of quality problem identification. Through multi-dimensional early warning rules and intelligent early warning models, it enables early prediction of quality risks, reducing the outflow of non-conforming products and production losses. Simultaneously, it possesses flexible expansion capabilities for performance indicators and judgment standards, adapting to the management needs of different working conditions and product specifications, promoting the transformation of quality control towards proactive early warning and data-driven approaches, and significantly improving the level of intelligence and efficiency. Attached Figure Description

[0016] Figure 1 This is a structural block diagram of a digital twin traceability system for the production quality of top-mounted ball valves provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 This invention provides a digital twin traceability system for the production quality of top-mounted ball valves, comprising: The core quality data definition module is used to define the core performance indicators, corresponding data dimensions, and quality judgment standards of top-loading ball valves, and outputs standardized quality data. The digital twin model construction module builds a unique twin based on standardized quality data, calculates the comprehensive weight of quality features through a dual-weight dynamic coupling algorithm, and generates a comprehensive weight matrix and a dynamically optimized digital twin model. The multi-source data fusion processing module receives standardized quality data and data from the entire production process of top-mounted ball valves. After cleaning and correlation processing, it outputs the fused quality data to the digital twin model construction module. The end-to-end quality traceability module calls the dynamically optimized digital twin model to realize forward and reverse traceability of the quality of the top-mounted ball valve based on the fused quality data, and outputs the traceability results. The quality anomaly early warning and analysis module, based on a comprehensive weight matrix, a dynamically optimized digital twin model, and fused quality data, generates quality anomaly early warning information through an early warning model, conducts correlation analysis between quality characteristics and performance, and outputs early warning results and quality analysis reports.

[0019] In one optional embodiment, the quality core data definition module includes: The performance index definition unit is used to define the core performance indexes of the top-loading ball valve. The core performance indexes include sealing performance, corrosion resistance, adaptability to ultra-low temperature environment and hard seal structure, and also support extended performance indexes. The data dimension definition unit is used to define the corresponding data dimensions for core performance indicators. The data dimensions include performance-specific data dimensions and general data dimensions. The performance-specific data dimensions cover sealing, corrosion resistance, ultra-low temperature and hard sealing adaptation scenarios. The general data dimensions include raw material information, process parameters, test parameters and production batch information. The judgment criterion definition unit is used to set judgment thresholds for each data dimension, forming a judgment criterion library. The judgment thresholds can be adjusted by the user, and the judgment criterion library is associated with standardized quality data.

[0020] It should be noted that in this embodiment, the "core performance indicators" are key performance parameters adapted to special application scenarios such as subsea oil extraction of top-mounted ball valves. Sealing performance ensures no leakage of the medium, corrosion resistance adapts to seawater and oil and gas media erosion, ultra-low temperature environment adaptability is for extreme working conditions of -109℃, and hard seal structure adaptability ensures sealing reliability under metal-to-metal contact friction. The purpose of its definition is to focus on the core quality control points of the product and avoid redundant irrelevant indicators. The "performance-specific data dimensions" are the quantification and visualization of the core performance indicators. For example, the sealing performance dimension includes directly measurable parameters such as leakage and sealing surface contact pressure, while the general data dimensions cover common influencing factors throughout the production process. The combination of the two achieves comprehensive coverage of quality data. The "judgment standard library" is formulated based on industry standards (such as GB / T12224-2019) and enterprise internal control requirements. The judgment threshold can be adjusted to adapt to the characteristics of different batches of raw materials, production equipment status, or customized needs of end customers. The judgment standard library is associated with standardized quality data to ensure the consistency of data collection and judgment.

[0021] In one possible implementation, the performance index expansion may include "impact resistance performance", and the corresponding data dimensions may include parameters such as impact pressure and number of impacts, to adapt to the impact conditions required by the top-mounted ball valve during transportation or installation.

[0022] For example, in one feasible implementation, for top-entry ball valves, the performance index definition unit clearly defines the core performance indicators as sealing performance, corrosion resistance, ultra-low temperature environment adaptability, and hard seal structure adaptability; the data dimension definition unit correspondingly sets the sealing performance dimension including leakage (unit: mL / min) and sealing surface contact pressure (unit: MPa), the corrosion resistance dimension including salt spray test duration (unit: h) and corrosion spot area (unit: mm²), the ultra-low temperature adaptability dimension including test temperature (unit: ℃) and low temperature leakage (unit: mL / min), the hard seal adaptability dimension including assembly gap (unit: mm) and coaxiality (unit: mm), and the general data dimension including raw material alloy grade, CNC machine tool speed, test equipment number, and production batch number; the judgment standard definition unit sets thresholds such as leakage ≤0.1mL / min, salt spray test ≥72h, and assembly gap ≤0.02mm, forming a judgment standard library, and allows the ultra-low temperature leakage threshold to be adjusted to ≤0.12mL / min according to customer requirements.

[0023] In one optional embodiment, the digital twin model construction module includes a dedicated twin construction unit, which is used to construct product twins, process twins, and test twins; The product twin uses the product's unique identifier as its core and links it to standardized quality data throughout the entire lifecycle of the top-loading ball valve, forming a unique mapping relationship between the product and the twin. The process twin covers the entire production process of top-loading ball valves, linking the processing parameters, equipment operation data, and quality inspection data of each process; The test twin constructs a virtual test environment model for the specific test scenario of the top-mounted ball valve, and associates the test equipment parameters, test process data and test result data.

[0024] It should be noted that in this embodiment, the "unique product identifier" uses RFID coding in the format "BQ + year + batch number + serial number". Its function is to assign a unique identifier to each top-mounted ball valve, ensuring accurate association of quality data throughout its entire lifecycle and avoiding data confusion. The "process twin" covers all production processes, including raw material processing, precision manufacturing of the ball / seat, hard seal assembly, cryogenic testing, salt spray testing, and factory inspection. The associated process data can intuitively present the quality transfer relationship between processes, facilitating the location of process-level quality problems. The "test twin" constructs a virtual test environment model that recreates the physical conditions of actual testing (such as the liquid nitrogen injection rate and temperature fluctuation range of the test chamber in cryogenic testing), achieving virtual reproduction and data traceability of the testing process. The associated test data can verify whether the product performance meets the judgment criteria. Furthermore, it should be noted that the core of constructing the dedicated twin is "data mapping rather than physical structure restoration," requiring no equipment structure diagrams; the virtual presentation of the quality status can be achieved solely through data association. In one possible implementation, the process twin can add associated data for "process quality inspection points," with each inspection point corresponding to specific inspection equipment parameters and inspection results, further refining the granularity of process quality traceability.

[0025] For example, in one feasible implementation, for the top-mounted ball valve with batch number 202504, the dedicated twin building unit assigns a unique RFID code (such as BQ2025040001) to each product. The product twin is associated with the raw material alloy grade (Inconel 625) corresponding to this code, the processing data of each process, all test results and factory reports; the process twin covers core processes such as ball machining, hard seal assembly, and ultra-low temperature testing, and is associated with data such as CNC machine tool speed (3000 r / min), assembly torque (5 N·m), and test equipment operating voltage (220 V); the test twin builds a virtual test chamber model for the ultra-low temperature test scenario, and is associated with data such as liquid nitrogen filling volume (50 L), heat preservation time (24 h), and low temperature leakage (0.12 mL / min), forming a complete dedicated twin system.

[0026] In an optional embodiment, the digital twin model building module further includes a dynamic coupling engine, which includes: The dual-weight calculation unit is used to calculate the feature importance weight and the quality impact weight. The formula for calculating the feature importance weight is: W1i =(σ i / Σσ i )×0.5, where σ i Let Σσ be the real-time data variance of the i-th quality feature. i The sum of the variances of all quality characteristics; the formula for calculating the weight of quality influence is: W 2i =(F i / ΣF i )×0.5, where F i Let ΣF be the historical number of times the i-th quality characteristic causes the top-loading ball valve to fail. i The total number of nonconformities caused by all quality characteristics; The model parameter iteration unit is used to calculate the comprehensive weight based on the feature importance weight and the quality impact weight. The formula for calculating the comprehensive weight is: W i =W 1i +W 2i The model parameters are updated based on the comprehensive weights. These parameters include feature mapping coefficients, virtual simulation rules, and traceability association weights. The formula for calculating the feature mapping coefficients is: K ij =W i ×K0, where K0 is the basic mapping coefficient; The deviation verification unit is used to calculate the deviation between the prediction results of the dynamically optimized digital twin model and the actual quality results. If the deviation value is greater than 5%, the model parameter iteration unit is triggered to update the model parameters again.

[0027] It should be noted that in this embodiment, the "feature importance weight" is calculated based on the real-time data variance. A larger variance indicates more significant fluctuations in the current batch of data and a more pronounced impact on quality results. For example, when the variance of cryogenic leakage data is large, increasing its weight can make the model focus on the mapping accuracy of that feature. The "quality impact weight" is based on statistics from nearly 1000 batches of historical non-conforming data. The more times a feature causes non-conformity, the higher its weight, reflecting long-term quality risk points. The comprehensive weight W... i By integrating real-time fluctuations and historical risks, the optimization of model parameters is ensured to be targeted. In terms of algorithm implementation, σ... i Σσ is obtained by calculating the standard deviation of data from the same batch for a certain quality characteristic. i F is the sum of the standard deviations of all characteristics. i By querying the database of quality issues, the statistical count of non-conforming records corresponding to this feature is obtained. i This represents the sum of all non-compliant records for all features; the default value of the basic mapping coefficient K0 is 1.2, which is an initial value obtained through calibration using data from at least three batches of historical compliant products. The feature mapping coefficient K... ij With comprehensive weight W iPositive correlation ensures the mapping priority of high-weight features in the model.

[0028] In one possible implementation, the deviation threshold can be adjusted according to the importance of product performance. The deviation threshold for core performance can be set to 3%, and for non-core performance it can be set to 7%.

[0029] For example, in one feasible implementation, for a certain batch of top-loading ball valves, the dual-weighted calculation unit statistically obtains the cryogenic leakage amount (i=1) as σ1=0.006, Σσ i =0.012, therefore W 11 =(0.006 / 0.012)×0.5=0.25; In historical data, this feature resulted in 60 non-compliant F1 values, ΣF i =80, therefore W 21 =(60 / 80)×0.5=0.375, the comprehensive weight W1=0.25+0.375=0.625; the model parameter iteration unit calculates the feature mapping coefficient W 1j =0.625×1.2=0.75, update the mapping rule of this feature in the model; the deviation verification unit calculates the deviation between the sealing performance result predicted by the model and the actual test result. If the deviation is less than 5%, the model parameters remain stable. If the deviation of a certain batch is 6%, a new iteration is triggered.

[0030] In one optional embodiment, the multi-source data fusion processing module includes: The data access unit is used to access various types of production data through a preset standardized interface. These various types of production data include sensor-collected data, test equipment data, system synchronization data, and manually entered data. The data cleaning unit is used to process missing and outlier values ​​in various types of production data. The data association unit is used to establish a full-link association between multiple types of production data and standardized quality data, with the product's unique identifier as the primary association key and batch number, process ID, and test equipment number as secondary association keys, to form integrated quality data.

[0031] It should be noted that in this embodiment, the "preset standardized interface" includes a RESTful API interface, an OPCUA protocol interface, and a file upload interface. Its function is to be compatible with the access requirements of data from different sources and to ensure that the data formats of different carriers such as sensors, testing equipment, and MES / ERP systems are unified. The "multi-type production data" covers the entire production process. Sensor-collected data includes real-time parameters such as assembly gaps and temperature. Testing equipment data includes the results of ultra-low temperature tests and salt spray tests. System-synchronized data includes production plans and equipment maintenance records. Manually entered data includes operator information and anomaly handling records, comprehensively covering factors affecting quality. For data cleaning, missing values ​​are handled using the K-nearest neighbor imputation algorithm (K=5), which fills in missing values ​​by finding the mean of the corresponding data of 5 similar products in the same batch, ensuring data integrity. Outlier handling adopts the 3σ criterion, marking data values ​​that exceed the range of [μ-3σ,μ+3σ] (μ is the mean of the data in the same batch, and σ is the standard deviation) as outliers. Invalid data is removed by manual review to avoid outliers affecting model accuracy. The data association unit establishes links through primary and secondary association keys to ensure the data traceability and continuity of "raw materials-process-testing-performance".

[0032] In one possible implementation, the data access unit supports two modes: batch import and real-time synchronization. Batch import is suitable for historical data supplementation, while real-time synchronization is suitable for production process data collection.

[0033] For example, in one feasible implementation, the data access unit accesses the hard seal assembly gap data (0.018mm) collected by the laser rangefinder through the OPCUA protocol interface, synchronizes the CNC machine tool speed data (3000r / min) in the MES system through the RESTful API interface, and imports the salt spray test report data through the file upload interface; the data cleaning unit finds that the ultra-low temperature leakage data of a certain product is missing, and fills it by finding the average leakage of 5 products in the same batch (0.11mL / min) through the K-nearest neighbor algorithm, and identifies one assembly gap data (0.05mm) that exceeds the 3σ range and marks it as abnormal; the data association unit uses the product unique identifier BQ2025040001 as the primary key, and associates all data corresponding to batch number 202504, process IDZM-03, and test equipment number CS-08 to form fused quality data.

[0034] In one optional embodiment, the full-process quality traceability module includes a forward traceability unit. The forward traceability unit is used to receive the product's unique identifier or batch number input by the user, call the dynamically optimized digital twin model, display the full-link quality data of the top-mounted ball valve from raw material warehousing to factory inspection in chronological order, clarify the correlation between the data of each link and the core performance indicators, and output the traceability results.

[0035] It should be noted that in this embodiment, "user input methods" include manual input, RFID barcode scanning, and batch number dropdown selection, adapting to different usage scenarios (such as on-site barcode scanning in the workshop and remote querying in the office); "full-chain quality data" is arranged in chronological order, clearly presenting the time flow of quality data from raw material warehousing (T0 time), sphere processing (T1 time), hard seal assembly (T2 time) to factory inspection (Tn time); "correlation between data at each stage and core performance indicators" refers to clarifying the impact of data at a certain stage on specific performance indicators. For example, hard seal assembly gap data is directly related to sealing performance indicators, and ultra-low temperature test data is directly related to ultra-low temperature compatibility indicators, helping users understand the logic of quality formation. The core purpose of forward traceability is to achieve "traceable source and traceable process." When end customers report quality problems, the entire process quality status of the product can be quickly located, and any data anomalies in the production process can be investigated.

[0036] In one possible implementation, the traceability results support the visualization of charts, including data time-series trend charts and process-performance correlation heatmaps, which intuitively present the patterns of quality changes. For example, in one feasible implementation, the user scans the product's unique identifier BQ2025040001 and the forward traceability unit calls the dynamically optimized digital twin model, displaying the following in chronological order: T0 (2025-04-01) Raw material warehousing (alloy grade Inconel625, material inspection qualified) → T1 (2025-04-02) Spherical machining (CNC machine tool speed 3000r / min, cutting depth 0.2mm) → T2 (2025-04-03) Hard seal assembly (assembly gap 0.018mm, coaxiality 0.008mm) → T3 (2025-04-04) Ultra-low temperature test (test temperature -109℃, leakage 0.12mL / min) → T4 (2025-04-05) Factory inspection (sealing performance qualified, corrosion resistance qualified), clarifying the correlation between assembly gap data and sealing performance indicators, and outputting traceability results including structured tables and visualized links.

[0037] In an optional embodiment, the end-to-end quality traceability module further includes a reverse traceability unit. The reverse traceability unit is used to receive a description of the quality problem or a unique identifier of the non-conforming product input by the user, and based on the comprehensive weight matrix, prioritize the production links corresponding to high-weight quality characteristics, clarify the root cause of the quality problem, related data and scope of impact, and output the reverse traceability results.

[0038] It should be noted that in this embodiment, the "Quality Problem Description" supports keyword input (such as "ultra-low temperature seal failure") or preset problem type selection (such as sealing, corrosion, assembly), facilitating users to quickly initiate traceability. High-weighted quality features in the "Comprehensive Weight Matrix" are key influencing factors for the current batch's quality; prioritizing their location can shorten root cause search time. For example, the assembly process corresponding to the assembly gap feature with the highest comprehensive weight is the priority for investigation. The "Root Cause Link" refers to the specific production or testing link that leads to the quality problem. "Related Data" includes the processing parameters, equipment status, and operator information for that link. "Scope of Impact" includes the number of non-conforming products in the same batch and the product serial numbers of products already shipped, facilitating rapid control of quality risks. The core idea of ​​reverse traceability is "result-based backtracking and root cause localization," solving the efficiency pain point of "finding a needle in a haystack" in traditional traceability by reverse-linking quality problems to data throughout the entire process.

[0039] In one possible implementation, reverse tracing supports multi-dimensional filtering, which can narrow the scope of investigation by time range, equipment number, operator, etc.

[0040] For example, in one feasible implementation, the user inputs a quality problem description "ultra-low temperature seal failure". The reverse tracing unit, based on a comprehensive weight matrix (ultra-low temperature leakage weight 0.625, assembly gap weight 0.58), prioritizes locating the ultra-low temperature testing stage and the hard seal assembly stage with high weights. Through correlation data investigation, it is found that the leakage of a certain product in the ultra-low temperature test is 0.16mL / min (exceeding the standard). Further tracing leads to the assembly gap of 0.025mm (exceeding the standard) in the hard seal assembly stage. Correlation data shows that the operator in this process is OP-08, the CNC machine tool used is CNC-12, and the equipment accuracy drift is 0.005mm. The affected range is 12 out of 200 products in the same batch that have similar assembly gap exceeding the standard, of which 8 have already left the factory. The output includes the root cause (hard seal assembly), correlation data, and the scope of impact.

[0041] In one optional embodiment, the quality anomaly warning and analysis module includes a warning rule configuration unit, which is used to configure three types of warning rules: threshold warning, trend warning, and correlation warning. Threshold alerts are used to trigger warnings when real-time quality data exceeds a judgment threshold; Trend alerts are used to trigger warnings when the trend of quality data deviates from the historical average by 10%. The correlation alert is used to trigger an alert for related quality features when high-weight quality feature data is abnormal.

[0042] It should be noted that in this embodiment, the "threshold warning" is based on real-time monitoring of thresholds in the judgment standard library. When real-time data exceeds the threshold (e.g., assembly gap > 0.02mm), it is triggered immediately, which is an "instant warning" and can quickly intercept unqualified products. The "trend warning" analyzes the quality data trend of 5-10 consecutive products. If it deviates from the historical average by 10% (e.g., the ultra-low temperature leakage rate continuously increases from 0.08mL / min to 0.12mL / min), it is a "potential risk warning" and can predict batch quality fluctuations. The "correlation warning" is based on the correlation between quality characteristics (e.g., assembly gap and sealing leakage are positively correlated). When the assembly gap data with high weight is abnormal, a sealing performance warning is triggered in association, which is a "chain risk warning" and can comprehensively cover potential quality hazards. The three types of warning rules complement each other to form a comprehensive warning system. The purpose of its configuration is to discover quality risks in advance and avoid batch non-conformities.

[0043] In one possible implementation, the deviation ratio of the trend warning can be adjusted according to the feature type, and the deviation ratio of key features can be set to eight percent.

[0044] For example, in one feasible implementation, the early warning rule configuration unit is configured with: threshold early warning (triggered when assembly gap > 0.02mm), trend early warning (triggered when the ultra-low temperature leakage rate of 5 consecutive products deviates from the historical average by 10%), and correlation early warning (correlated triggering of sealing leakage rate early warning when the assembly gap is abnormal); during the production process, the measured assembly gap of a certain product is 0.023mm, triggering the threshold early warning; the ultra-low temperature leakage rates of 5 consecutive products are 0.10mL / min, 0.11mL / min, 0.12mL / min, 0.13mL / min, and 0.14mL / min, respectively, deviating from the historical average (0.10mL / min) by 40%, triggering the trend early warning; at the same time, the abnormal assembly gap triggers the correlation early warning, and the system simultaneously pushes three early warning messages to the management personnel terminal.

[0045] In an optional embodiment, the quality anomaly warning and analysis module further includes a warning model construction unit. The warning model construction unit adopts a fusion algorithm of decision tree and long short-term memory network, and completes model training using historical quality data of top-mounted ball valves as the training set. The accuracy of the warning model is ≥92%. The early warning model building unit is also used to generate early warning information, which includes the early warning type, the unique identifier of the product involved, abnormal data, related performance indicators and early warning level, and is pushed to the designated terminal through a preset push method.

[0046] It should be noted that in this embodiment, the "fusion algorithm of decision tree and long short-term memory network" combines the classification advantages of decision tree (rapidly identifying clear abnormal patterns) with the time-series data processing advantages of LSTM (capturing data trend changes). The model training uses nearly 1000 batches of historical quality data, with 70% used for training and 30% for validation to ensure the model's generalization ability. The key elements included in the "early warning information" help managers quickly grasp the core content of the warning. The warning levels are divided into Level 1 (severe, such as core performance exceeding the standard), Level 2 (important, such as potential trend risks), and Level 3 (general, such as non-core data anomalies), facilitating graded handling. The "preset push methods" include system pop-ups, SMS, and emails, with a push delay of no more than one minute to ensure timely response to warnings. In one possible implementation, the early warning model supports online updates and can incorporate the latest batch of data to continuously optimize accuracy.

[0047] For example, in one feasible implementation, the early warning model construction unit adopts a fusion algorithm of decision tree (C4.5 algorithm) and LSTM, using the quality data of 1,000 batches of top-mounted ball valves from January 2024 to March 2025 as the training set. After training, the model accuracy rate is 94%. During the production process, the model identifies that the cryogenic leakage of product BQ2025040056 is 0.15mL / min (close to the threshold of 0.15mL / min), and three consecutive products show an upward trend. The model generates early warning information: warning type (trend warning), unique identifier of the product involved (BQ2025040056), abnormal data (cryo-cryo leakage of 0.15mL / min), associated performance index (cryo-cryo environment adaptability), and warning level (level 2). The warning is pushed to the workshop management personnel terminal through a system pop-up window and an SMS is sent to the quality supervisor's mobile phone at the same time.

[0048] In an optional embodiment, the quality anomaly warning and analysis module further includes a quality analysis unit, which is used to conduct correlation strength analysis between quality characteristics and performance based on a comprehensive weight matrix. The quality analysis unit is also used to statistically analyze the trends of quality data from multiple batches of products, identify patterns of quality fluctuations, analyze the root causes of quality anomalies, including raw material defects, equipment failures, process parameter deviations, and operator errors, and output quality analysis results.

[0049] It should be noted that in this embodiment, "correlation strength analysis" is achieved by calculating the correlation coefficient between quality characteristics and performance indicators. The larger the absolute value of the correlation coefficient, the stronger the correlation. For example, the correlation coefficient between assembly gap and sealing performance is 0.85, indicating that the two are highly correlated. The analysis results can guide the key control of this characteristic during the production process. The statistical period for "multi-batch product quality data trend" can be customized (such as monthly or quarterly). Through trend analysis, common causes of quality fluctuations can be identified. For example, if the average ultra-low temperature leakage increases in a certain quarter, correlation with changes in raw material batches reveals that it is a problem with the alloy material batch of supplier A. "Root cause type" is classified based on data correlation analysis. Raw material defects correspond to abnormal raw material test data, equipment failures correspond to equipment operating parameter drift, process parameter deviations correspond to processing parameters exceeding the set range, and operator errors correspond to non-standard operation process records. The purpose of this analysis is to provide data support for production optimization. In addition, it should be noted that the output results of the quality analysis unit can be directly connected to the production management system to achieve quality control.

[0050] In one possible implementation, the correlation strength analysis supports filtering by performance index, allowing for individual viewing of key quality characteristics corresponding to sealing performance, facilitating targeted optimization.

[0051] For example, in one feasible implementation, the quality analysis unit calculates the correlation strength between assembly clearance and sealing performance as 0.85 and the correlation strength between cryogenic leakage and cryogenic compatibility as 0.78 based on a comprehensive weight matrix. Statistical analysis of data trends from 50 batches of products in the first quarter of 2025 reveals that the average cryogenic leakage increased from 0.10 mL / min to 0.13 mL / min. Correlation with raw material data shows that the average thickness of the corrosion-resistant coating of alloy materials from supplier A decreased from 0.12 mm to 0.09 mm during this quarter. Analysis of the root causes of 10 recent quality anomalies reveals that 4 were due to equipment failure (CNC machine tool accuracy drift), 3 to process parameter deviations (insufficient assembly torque), 2 to raw material defects (excessive alloy impurities), and 1 to operator error (incorrect assembly sequence). The output includes quality analysis results encompassing correlation strength analysis, trend analysis, and root cause analysis, providing a basis for changing raw material suppliers, overhauling CNC machine tools, and optimizing assembly processes.

[0052] In this embodiment, comprehensive traceability data support is provided through the standardized definition of core quality data and the full-link association and fusion of multi-source data; the digital twin model is self-optimized by relying on a dual-weight dynamic coupling algorithm, improving the model's adaptability to actual working conditions and the accuracy of traceability analysis; the efficiency of quality problem investigation is improved by combining forward full-link traceability and reverse high-weight priority positioning mechanism; quality risk is predicted in advance through multi-dimensional early warning rules and intelligent early warning models, reducing the outflow of non-conforming products and production losses; at the same time, it has the flexible expansion capability of performance indicators and judgment standards to adapt to the management and control needs of different working conditions and product specifications, promoting the transformation of quality control towards proactive early warning and data-driven approaches, and significantly improving the level of intelligence and efficiency.

[0053] Furthermore, it should be noted that the combination of the various technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0054] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0055] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An upper-mounted ball valve production quality digital twin traceability system, characterized in that, Comprise: A quality core data definition module for defining the core performance indicators of the top-mounted ball valve, corresponding data dimensions and quality judgment standards, and outputting standardized quality data; A digital twin model construction module that constructs a dedicated twin based on the standardized quality data, calculates the comprehensive weight of the quality characteristics through a double-weight dynamic coupling algorithm, generates a comprehensive weight matrix and a dynamically optimized digital twin model; A multi-source data fusion processing module that accesses the standardized quality data and the entire production process data of the top-mounted ball valve, processes the data after cleaning and association, and outputs the fusion quality data to the digital twin model construction module; A full-process quality traceability module that calls the dynamically optimized digital twin model and realizes forward and reverse traceability of the quality of the top-mounted ball valve based on the fusion quality data, and outputs the traceability results; A quality anomaly early warning and analysis module that generates quality anomaly early warning information through a warning model based on the comprehensive weight matrix, the dynamically optimized digital twin model and the fusion quality data, and conducts correlation analysis of quality characteristics and performance, and outputs early warning results and quality analysis reports.

2. The upper-mounted ball valve production quality digital twin traceability system according to claim 1, wherein, The quality core data definition module comprises: A performance indicator definition unit for defining the core performance indicators of the top-mounted ball valve, the core performance indicators including sealing performance, corrosion resistance, ultra-low temperature environment adaptability and hard seal structure adaptability, and supporting extended performance indicators; A data dimension definition unit for defining corresponding data dimensions for the core performance indicators, the data dimensions including performance-specific data dimensions and general data dimensions, the performance-specific data dimensions covering sealing, corrosion resistance, ultra-low temperature and hard seal adaptation scenarios, and the general data dimensions including raw material information, process parameters, test parameters and production batch information; A judgment standard definition unit for setting judgment thresholds for each data dimension to form a judgment standard library, the judgment thresholds being adjustable by the user, and the judgment standard library being associated with the standardized quality data.

3. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The digital twin model construction module comprises a dedicated twin construction unit, which is used to construct product twins, process twins and test twins; The product twin takes the product unique identifier as the core, associates the standardized quality data of the top-mounted ball valve throughout its life cycle, and forms a unique mapping relationship between the product and the twin; The process twin covers the entire production process of the top-mounted ball valve, and is associated with the processing parameters, equipment operation data and quality detection data of each process; The test twin constructs a virtual test environment model for the dedicated test scenarios of the top-mounted ball valve, and is associated with test equipment parameters, test process data and test result data.

4. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The digital twin model construction module further comprises a dynamic coupling engine, which comprises: A double weight calculation unit is configured to calculate a feature importance weight and a quality influence degree weight, wherein the feature importance weight is calculated according to the formula: W 1i =(σ i / Σσ i )×0.5, wherein σ i is the real-time data variance of the i th quality feature, and Σσ i is the total variance of all quality features; and the quality influence degree weight is calculated according to the formula: W 2i =(F i / ΣF i )×0.5, wherein F i is the historical number of times that the i th quality feature causes the upper-mounted ball valve to be unqualified, and ΣF i is the total number of times that all quality features cause the valve to be unqualified. a model parameter iteration unit, configured to calculate a comprehensive weight based on the feature importance weight and the quality influence degree weight, a calculation formula of the comprehensive weight being: W i = 1i + W 2i , and update a model parameter according to the comprehensive weight, the model parameter including a feature mapping coefficient, a virtual simulation rule and a trace correlation weight, a calculation formula of the feature mapping coefficient being: K ij = W i ×K0, wherein K0 is a basic mapping coefficient. A bias checking unit for calculating the bias value between the prediction results of the dynamically optimized digital twin model and the actual quality results, and triggering the model parameter iteration unit to update the model parameters if the bias value is greater than 5%.

5. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The multi-source data fusion processing module comprises: A data access unit is configured to access multiple types of production data through a preset standardized interface, the multiple types of production data including sensor collected data, test equipment data, system synchronization data and manually entered data; A data cleaning unit is configured to process missing values and abnormal values of the multiple types of production data; A data correlation unit is configured to establish full-link correlation between the multiple types of production data and the standardized quality data by taking product unique identification as a main correlation key and taking batch number, process ID and test equipment number as auxiliary correlation keys, and form the fused quality data.

6. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The full-process quality traceability module includes a forward traceability unit configured to receive product unique identification or batch number input by a user, call the dynamically optimized digital twin model, display full-link quality data of the top-mounted ball valve from raw material warehousing to factory detection in time sequence, make clear the correlation between each link data and the core performance index, and output traceability results.

7. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The full-process quality traceability module also includes a reverse traceability unit configured to receive quality problem description or unqualified product unique identification input by a user, prioritize the production link corresponding to a high-weight quality feature based on the comprehensive weight matrix, make clear the root link, related data and influence range of the quality problem, and output reverse traceability results.

8. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The quality anomaly early warning and analysis module includes a warning rule configuration unit configured to configure three types of warning rules, i.e., threshold warning, trend warning and correlation warning; The threshold warning is configured to trigger a warning when real-time quality data exceeds the judgment threshold; The trend warning is configured to trigger a warning when the quality data trend deviates from the historical mean value by 10%; The correlation warning is configured to trigger a warning of related quality features when high-weight quality feature data is abnormal.

9. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The quality anomaly early warning and analysis module also includes a warning model construction unit configured to use a fusion algorithm of decision tree and long short-term memory network, complete model training by taking historical quality data of the top-mounted ball valve as a training set, and the accuracy of the warning model is ≥92%; The warning model construction unit is also configured to generate warning information, the warning information including warning type, involved product unique identification, abnormal data, related performance index and warning level, and push the warning information to a specified terminal through a preset pushing mode.

10. The upper-mounted ball valve production quality digital twin traceability system of claim 1, wherein, The quality anomaly early warning and analysis module also includes a quality analysis unit configured to analyze the correlation strength between quality features and performance based on the comprehensive weight matrix; The quality analysis unit is also configured to statistically analyze quality data trends of multiple batches of products, identify quality fluctuation rules, analyze root cause types of quality anomalies, the root cause types including raw material defects, equipment failures, process parameter deviations and operator errors, and output quality analysis results.

Images