CN118011990A

CN118011990A - Industrial data quality monitoring and improving system based on artificial intelligence

Info

Publication number: CN118011990A
Application number: CN202410424860.1A
Authority: CN
Inventors: 岳高峰; 李文武; 高亮; 王淑敏
Original assignee: China National Institute of Standardization
Current assignee: China National Institute of Standardization
Priority date: 2024-04-10
Filing date: 2024-04-10
Publication date: 2024-05-10

Abstract

The invention relates to the technical field of industrial production monitoring, in particular to an industrial data quality monitoring and improving system based on artificial intelligence. The system comprises the following modules: the industrial data acquisition module S1 is used for acquiring workshop equipment space layout data; acquiring high-density sensing industrial data of a target production workshop according to workshop equipment space layout data to obtain a heterogeneous multi-source industrial data set, wherein the heterogeneous multi-source industrial data set comprises product production data, equipment performance parameters and product quality indexes; the data quality influence factor mining module S2 is used for acquiring historical product production data and historical equipment performance parameters; and detecting low-quality data of the historical product production data to obtain the historical low-quality production data. The invention can realize the capabilities of automation, intellectualization, high-efficiency real-time monitoring, comprehensive industrial data quality control and improvement.

Description

Industrial data quality monitoring and improving system based on artificial intelligence

Technical Field

The invention relates to the technical field of industrial production monitoring, in particular to an industrial data quality monitoring and improving system based on artificial intelligence.

Background

With the advent of the 4.0 era of industry, automation and informatization have become a hallmark of modern industrial production. During this process, sensors, control systems and intelligent devices on the production line produce vast amounts of data. Such data includes, but is not limited to, machine performance parameters, process monitoring data, product quality metrics, etc., which are critical to achieving efficient production management and quality control.

However, conventional data monitoring and analysis methods suffer from a number of deficiencies. First, they typically rely on manual operations, which are not only time consuming and labor intensive, but also prone to errors due to human factors. Secondly, the traditional method has low efficiency when processing large-scale data, and real-time monitoring and quick response are difficult to realize. Furthermore, it is possible to provide a device for the treatment of a disease. The traditional data quality control method mainly relies on expert experience rules for data verification, but the method has the following obvious defects: the workload required by rule setting is large, and all possible abnormal conditions are difficult to cover; the novel abnormality cannot be identified, and only the known abnormality mode can be detected; each rule is independently set, and the correlation between data cannot be considered; data quality analysis and promotion lacks systematic and intelligent support.

Disclosure of Invention

Accordingly, there is a need for an artificial intelligence based industrial data quality monitoring and promotion system that solves at least one of the above-mentioned problems.

To achieve the above object, an industrial data quality monitoring and improving system based on artificial intelligence, the system comprises the following modules:

The industrial data acquisition module S1 is used for acquiring workshop equipment space layout data; acquiring high-density sensing industrial data of a target production workshop according to workshop equipment space layout data to obtain a heterogeneous multi-source industrial data set, wherein the heterogeneous multi-source industrial data set comprises product production data, equipment performance parameters and product quality indexes;

the data quality influence factor mining module S2 is used for acquiring historical product production data and historical equipment performance parameters; detecting low-quality data of the historical product production data to obtain historical low-quality production data; mining data quality influence factors based on historical product production data and historical equipment performance parameters to obtain a product production data quality influence factor set; calculating a data quality critical value range according to the product production data quality influence factor set and the historical low-quality production data to obtain a production data acceptable quality interval;

The data quality evaluation module S3 is used for carrying out data quality quantitative evaluation on the product production data according to the product production data quality influence factor set to obtain a product production data quality index; comparing the product production data quality index with the acceptable quality interval of the production data;

The data quality improvement decision module S4 is used for performing intelligent production stop line operation on a target production workshop when the quality index of the production data of the product is smaller than the lower limit of the acceptable quality interval of the production data, and performing data anomaly root cause tracing on the production data of the product to obtain anomaly incentive data of the production data; carrying out data quality improvement optimization decision on a target production workshop based on historical product production data, historical low-quality production data and production data abnormality cause data to obtain a production data quality enhancement strategy; carrying out strategy deep fusion implementation on a target production workshop according to the production data quality enhancement strategy;

The production process knowledge graph construction module S5 is used for carrying out causal relation dynamic knowledge graph construction on the quality index of the production data of the product, the abnormal incentive data of the production data and the quality enhancement strategy of the production data so as to obtain a dynamic data quality causal knowledge graph;

The production data situation visualization module S6 is used for carrying out space-time labeling on the performance parameters of the production data equipment and the quality indexes of the products when the quality indexes of the production data of the products are equal to or larger than the acceptable quality interval data of the production data, so as to obtain real-time production environment state portrait data; carrying out multidimensional data fusion on the dynamic data quality causal knowledge graph and the real-time production environment state portrait data to obtain a production quality visual situation graph; uploading the production quality visualization situation map to a preset industrial data monitoring platform.

According to the invention, the industrial data acquisition module is used for acquiring the industrial data through high-density perception, and the system can timely acquire real-time data of workshop equipment, including product production data, equipment performance parameters and product quality indexes. The method can provide accurate and real-time production data, eliminates the defect of the traditional dependence on manual operation, and improves the efficiency and accuracy of data acquisition. Such real-time monitoring and acquisition can help solve the problem of inefficiency of conventional methods in processing large-scale data, and achieve real-time monitoring and rapid response to the production line. The module includes data from different devices and systems by acquiring heterogeneous multi-source industrial data sets. Such data integration can provide a comprehensive data perspective, so that subsequent data analysis and decision-making can comprehensively consider information of different data sources, and more comprehensively understand relevant data on the whole production line. The data quality influence factor mining module is capable of mining influence factor sets of data quality by analyzing and modeling historical data. These influencing factors can be used for subsequent quantitative evaluation and promotion decisions of data quality, and provide basis and reference for evaluating data quality. Based on the historical data, the system may calculate an acceptable quality interval for the production data. The interval can be used as a standard for measuring the quality of the current data, and helps to judge whether the data meets the preset quality requirement. The module can comprehensively and systematically consider the relativity between data, and can identify novel anomalies, thereby not only reducing the workload of manual rule setting, but also improving the accuracy and the intelligent level of data quality analysis. The data quality evaluation module performs data quality quantitative evaluation on the product production data and compares the data quality quantitative evaluation with an acceptable quality interval of the production data, so that automatic evaluation on the product production data is realized. By automated evaluation, the system is able to quickly and accurately quantitatively evaluate the quality of product production data. Compared with manual evaluation, the automatic evaluation can improve the accuracy and efficiency of evaluation and save human resources. By comparing with the acceptable quality interval, the system can determine whether the product production data meets the preset quality requirement. The method provides basis for subsequent data quality improvement decision, and helps users take measures in time to improve the data quality. The data quality improvement decision module judges whether the data quality improvement optimization decision is needed or not through comparing the quality index of the product production data with the acceptable quality interval of the production data. When the quality index is lower than the lower limit of the acceptable quality interval, the system can conduct intelligent production line stop operation, trace the abnormal source of the data and obtain abnormal cause data. Therefore, the method can help users to timely find and process abnormal data quality on the production line, and reduce production and circulation of unqualified products. Based on the historical data and the anomaly incentive data, the system is able to generate targeted production data quality enhancement policies. These strategies may include improvements in equipment maintenance, process optimization, etc., to help improve data quality and line stability. The module can generate strategies based on historical data and abnormal incentive data more systematically, and improves the improvement effect of data quality and the stability of a production line. The production process knowledge graph construction module constructs a causal relation dynamic knowledge graph, and correlates the quality index of the production data, the abnormal incentive data and the quality enhancement strategy of the product to form the dynamic data quality causal knowledge graph. The data quality related information is structured and visually represented, so that a user is helped to better understand the influencing factors and improvement strategies of the data quality. The form of the knowledge graph can present the connection and the association between the data, so that the user can intuitively grasp the overall condition of the data quality. Constructing a dynamic data quality causal knowledge graph may facilitate sharing and propagation of knowledge. The form of the knowledge graph can intuitively present the connection and the association between the data, help users better understand the influencing factors and the improvement strategies of the data quality, and promote the sharing and the propagation of the knowledge. Different stakeholders can know relevant information of the data quality through the knowledge graph, and discuss and improve schemes of the data quality together. The production data situation visualization module performs space-time labeling on the production data of the product, performs multidimensional data fusion with a dynamic data quality causal knowledge graph, and generates a production quality visualization situation graph. The multi-source data are comprehensively analyzed and visually displayed, so that a user is helped to intuitively know the state and trend of the production data. Through the visual situation map, a user can clearly grasp the data quality condition on the production line at a glance, and quickly find out abnormal conditions and improvement directions. In conclusion, the invention can realize high-efficiency automatic data monitoring and analysis, reduce the complexity and error of manual operation and improve the data processing efficiency. The intelligent data quality control method can reduce the workload of expert rule setting, identify novel anomalies, consider the relativity between data and improve the accuracy and the comprehensiveness of data quality analysis. The systematic and intelligent data quality improvement method provides comprehensive support and guidance, helps to improve data quality and enhances the stability of a production line. In addition, the structured and visualized data quality information representation form is beneficial to users to better understand data quality related information, so that knowledge sharing and spreading are promoted, and the efficiency and accuracy of data quality management are improved.

Drawings

Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of a non-limiting implementation, made with reference to the accompanying drawings in which:

FIG. 1 is a block flow diagram of an embodiment of an artificial intelligence based industrial data quality monitoring and promotion system.

FIG. 2 shows a detailed step flow diagram of S44 of an embodiment.

Detailed Description

The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.

Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

To achieve the above objective, referring to fig. 1 to 2, the present invention provides an industrial data quality monitoring and enhancing system based on artificial intelligence, the system comprises the following modules:

s1: the industrial data acquisition module S1 is used for acquiring workshop equipment space layout data; acquiring high-density sensing industrial data of a target production workshop according to workshop equipment space layout data to obtain a heterogeneous multi-source industrial data set, wherein the heterogeneous multi-source industrial data set comprises product production data, equipment performance parameters and product quality indexes;

Specifically, for example, a sensor network may be installed or a wireless communication device may be used in a target production plant, and various sensor devices such as a temperature sensor, a humidity sensor, a pressure sensor, a vibration sensor, and the like may be deployed to acquire real-time status information of each device in the plant. These sensor devices are strategically placed to cover the equipment space layout of the entire plant. Next, the sensor device is configured to be able to collect various industrial data such as product production data, device performance parameters, and product quality indicators. By setting the proper sampling frequency and sampling precision, high-density data perception and accuracy are ensured. Finally, the collected data is integrated and processed from different sensors and devices to form a heterogeneous multi-source industrial dataset. This data set includes product production data such as quantity produced, rate of production, time of production, etc.; device performance parameters such as temperature, pressure, vibration frequency, etc.; and product quality indicators such as size, weight, mass, etc.

S2: the data quality influence factor mining module S2 is used for acquiring historical product production data and historical equipment performance parameters; detecting low-quality data of the historical product production data to obtain historical low-quality production data; mining data quality influence factors based on historical product production data and historical equipment performance parameters to obtain a product production data quality influence factor set; calculating a data quality critical value range according to the product production data quality influence factor set and the historical low-quality production data to obtain a production data acceptable quality interval;

specifically, historical product production data and historical device performance parameters may be obtained, for example, through a system integration or data interface. For historical product production data, relevant data, such as production quantity, production rate, etc., may be extracted from a production database, production record, or production management system. For historical device performance parameters, such as device temperature, pressure, etc., may be obtained through device monitoring systems, sensor data, or device logging. These data may be obtained by suitable data extraction methods and communication protocols. Then, low-quality data detection is performed on the historical product production data to identify historical low-quality production data samples. This may be achieved by applying techniques such as statistical analysis methods, anomaly detection algorithms or rule engines. The detection method can be selected and customized according to specific data characteristics and quality criteria. And then, mining the data quality influence factors based on the historical product production data and the historical equipment performance parameters. This may employ data mining techniques such as machine learning algorithms, statistical analysis methods, or correlation analysis, etc., to find factors that have a significant impact on the quality of the product production data. These factors may be equipment status, process parameters, environmental conditions, etc. And finally, calculating a data quality critical value range according to the product production data quality influence factor set and the historical low-quality production data. By analyzing the correlation between the historical low quality production data and the quality impact factor set, an acceptable quality interval for the production data can be determined.

S3: the data quality evaluation module S3 is used for carrying out data quality quantitative evaluation on the product production data according to the product production data quality influence factor set to obtain a product production data quality index; comparing the product production data quality index with the acceptable quality interval of the production data;

Specifically, for example, product production data may be quantitatively evaluated using a set of data quality impact factors. The influence degree of each factor on the data quality can be comprehensively considered by carrying out weighted calculation on the influence factors of each data sample or establishing a proper model, so that the quality index of the product production data is obtained. The index may reflect quality aspects of accuracy, integrity, consistency, etc. of the data. Then, the quality index of the product production data is compared with the acceptable quality interval of the production data.

S4: the data quality improvement decision module S4 is used for performing intelligent production stop line operation on a target production workshop when the quality index of the production data of the product is smaller than the lower limit of the acceptable quality interval of the production data, and performing data anomaly root cause tracing on the production data of the product to obtain anomaly incentive data of the production data; carrying out data quality improvement optimization decision on a target production workshop based on historical product production data, historical low-quality production data and production data abnormality cause data to obtain a production data quality enhancement strategy; carrying out strategy deep fusion implementation on a target production workshop according to the production data quality enhancement strategy;

Specifically, for example, when the quality index of the product production data is lower than the lower limit of the acceptable quality interval, the module automatically triggers the intelligent production stop operation to avoid the adverse effect of the low quality data on the subsequent production and quality. Then, the module performs anomaly source tracing on the product production data, and can determine incentive data causing data anomalies by analyzing the characteristics, related factors and historical data of the low-quality data. These incentive data may include equipment anomalies, process variations, material problems, and the like. Based on historical product production data, historical low-quality production data and production data anomaly incentive data, the module can make data quality improvement optimization decisions for the target production plant. The decision can be based on technologies such as data mining, statistical analysis and machine learning, and the like, and combines expert experience and domain knowledge to identify factors with great influence on the quality of the data, and make corresponding strategies and measures to improve the quality of the production data. And finally, according to the production data quality enhancement strategy, the module carries out strategy deep fusion implementation on the target production workshop. The method comprises the steps of implementing a data quality monitoring and feedback mechanism, monitoring and analyzing data in the production process in real time, and timely finding and processing abnormal data quality. Meanwhile, the quality level of the production data is comprehensively improved by optimizing the production flow, improving the equipment maintenance, adjusting the technological parameters and the like.

S5: the production process knowledge graph construction module S5 is used for carrying out causal relation dynamic knowledge graph construction on the quality index of the production data of the product, the abnormal incentive data of the production data and the quality enhancement strategy of the production data so as to obtain a dynamic data quality causal knowledge graph;

Specifically, for example, the product production data quality index, production data anomaly incentive data, and production data quality enhancement strategies may be further analyzed using causal relationship analysis methods, such as causal inference, causal graph analysis, and the like. This can help determine causal relationships between different factors and reveal their extent of influence on data quality. Based on the results of the causal relationship analysis, a knowledge graph construction tool or method, such as a graph database, a knowledge graph modeling tool and the like, is used for constructing the causal knowledge graph of the dynamic data quality. The knowledge graph can graphically represent the association relations among different factors, indexes and strategies, and supports dynamic updating and query.

S6: the production data situation visualization module S6 is used for carrying out space-time labeling on the performance parameters of the production data equipment and the quality indexes of the products when the quality indexes of the production data of the products are equal to or larger than the acceptable quality interval data of the production data, so as to obtain real-time production environment state portrait data; carrying out multidimensional data fusion on the dynamic data quality causal knowledge graph and the real-time production environment state portrait data to obtain a production quality visual situation graph; uploading the production quality visualization situation map to a preset industrial data monitoring platform.

Specifically, for example, when the quality index of the production data of the product is equal to or greater than the acceptable quality interval data of the production data, the performance parameters of the production data equipment and the quality index of the product may be space-time labeled for the data satisfying the quality requirement. This means that the data is marked to indicate the device performance parameters and product quality indicators to which the data corresponds. Labeling may be performed by automatic recognition or model inference, etc. And obtaining real-time production environment state portrait data based on the production data labeling result. These data reflect the status of the device performance parameters and product quality indicators in the current production environment and may include time, space, and other relevant attribute information. And carrying out multidimensional data fusion on the dynamic data quality causal knowledge graph and the real-time production environment state portrait data. This step involves correlating and integrating the causal relationships in the knowledge-graph with the real-time environmental data to build a comprehensive data model. The multidimensional data fusion may be performed using a data integration tool, a data mining algorithm, a machine learning method, or the like. And generating a production quality visual situation map by carrying out visual processing on the multi-dimensional data fusion result. The graphical display interactively displays the dynamic data quality causal knowledge graph and the real-time production environment state portrait data so as to present the situation change and trend of the production quality. The visualization map may be implemented using data visualization tools, a graphic library, or dashboard software, etc. And finally, uploading the generated production quality visualization situation map to a preset industrial data monitoring platform. The platform can be a data monitoring system inside an enterprise or a data analysis platform in the cloud. After uploading, the production quality visualization situation map can be further analyzed, shared and applied to real-time monitoring and decision support.

Preferably, the industrial data acquisition module S1 is specifically:

s11: acquiring workshop equipment space layout data;

Specifically, for example, the position, size, and relative position of the equipment within the plant may be measured, marked, and recorded using a measuring tool (e.g., a measuring instrument, a laser rangefinder, etc.). From the measurement data, a plan view of the plant is drawn using Computer Aided Design (CAD) software or other drawing tools. The plan view should include detailed information of the locations of the devices, the distances between the devices, and the relative locations.

S12: according to the workshop equipment space layout data, carrying out automatic acquisition and manual input equipment division on a target production workshop to obtain automatic acquisition equipment identification data and manual input equipment identification data;

Specifically, for example, the type, function, and location information of the device may be known based on the plant space layout data. According to the characteristics and functions of the devices, determining which devices can acquire data through an automatic acquisition system. For example, an automated device provided with a sensor, a monitoring device, or the like may be classified as an automatic acquisition device. Each automated collection device is assigned a unique identifier, such as a device number, bar code, or RFID tag. These identifiers can be used to identify the device and to interface with the automated acquisition system. For equipment which cannot acquire data through an automatic acquisition system, manual entry is required. According to the nature of the devices and the feasibility of data acquisition, it is determined which devices need to manually input data by a human. Each manual entry device is assigned a unique identifier, such as a device number, bar code or RFID tag. These identifiers may be used to identify devices and to associate with data entry processes. The identification data of each device is recorded, including the identifiers of the automatic acquisition devices and the manual entry devices.

S13: deploying an industrial Internet of things sensor array for corresponding equipment in a target production workshop according to the automatic acquisition equipment identification data, and carrying out asynchronous multi-source data acquisition to obtain product production data and equipment performance parameters;

Specifically, for example, the type and number of sensors to be deployed may be determined to cover the equipment in the target production plant based on automatically collecting the equipment identification data. According to the device identification data and the sensor deployment plan, the industrial Internet of things sensor array is deployed on a corresponding device in the target production plant. The sensor comprises a temperature sensor, a pressure sensor, a vibration sensor and the like, and is used for collecting the running state and the performance parameters of the equipment. Meanwhile, in the production link of the product, according to the production data of the product to be monitored, an appropriate sensor type is selected and deployed on a production line, for example, a weighing sensor is used for monitoring the weight of the product. The configuration data acquisition module is connected with the sensor to realize asynchronous multi-source data acquisition. The sensor begins to collect data from the device including parameters such as temperature, pressure, vibration, etc., as well as product production data such as product weight. These data are received by a data collection node or gateway device and stored in a local database or cloud platform.

S14: according to the identification data of the manual input equipment, carrying out man-machine interaction data acquisition on corresponding equipment in the target production workshop to obtain manual input production operation data;

Specifically, for example, the equipment required to perform human-computer interaction data acquisition can be determined according to manually input equipment identification data. And designing a corresponding human-computer interaction interface aiming at each device needing to collect data. The interface may be a computer application, a mobile device application, or a specialized data acquisition terminal. An operator or related personnel inputs and records production job data related to the equipment through a human-machine interaction interface. Such data may include equipment operation records, maintenance records, fault information, and the like.

S15: performing self-adaptive manual error correction on the manual input production operation data to obtain optimized manual input production operation data;

Specifically, for example, data mining and statistical analysis methods may be used to identify potential data errors and anomalies in manually entered production job data. And establishing a self-adaptive artificial error correction model based on the result of the data quality analysis. The model can automatically correct errors in manually entered data according to the characteristics and error modes of the data. And correcting the manually input production operation data according to the established error correction model. An automated script, data processing tool, or specialized data correction system may be employed to implement the error correction process.

S16: carrying out multi-mode product quality index evaluation on the product production data and the optimized manually-input production operation data to obtain product quality indexes;

In particular, for example, product production data and optimized manually entered production job data may be integrated to create a complete data set. And ensuring the correspondence and consistency of the data. According to specific application requirements and product characteristics, multi-mode product quality indexes are defined. These indicators may include indicators of production efficiency, product quality, failure rate, etc. And selecting a proper evaluation method according to the definition of the product quality index. For example, statistical analysis, data mining, machine learning, etc. may be used to evaluate the quality index of the multi-modal product. Based on the integrated dataset, data analysis and model construction are performed. A predictive model or classification model may be built using appropriate algorithms and techniques to evaluate the product quality index. And evaluating the product production data and optimizing the manually input production operation data by using the established model to obtain the result of the product quality index.

S17: and carrying out intelligent weighted heterogeneous data aggregation on the product production data, the equipment performance parameters and the product quality indexes to obtain a heterogeneous multi-source industrial data set, wherein the heterogeneous multi-source industrial data set comprises the product production data, the equipment performance parameters and the product quality indexes.

Specifically, for example, product production data, equipment performance parameters, and product quality metrics may be intelligently weighted. Weighting may be based on factors such as data quality, data reliability, importance of the data, and the like. Weighting algorithms, such as weighted averaging or weighted regression methods, are used to calculate the weights of the different data sources. And carrying out heterogeneous data aggregation on the weighted product production data, the equipment performance parameters and the product quality indexes. This may be achieved by a data connection, a merge or an association operation. And according to the association relation and the characteristics of the data, the data are aggregated to form a heterogeneous multi-source industrial data set.

The invention can obtain the position and layout information of each device in the workshop by obtaining the space layout data of the workshop device. This provides a basis for subsequent automatic acquisition and manual entry of device identification data, helping to determine the relative location and relevance of the devices, and thus better understanding of the source and relationship of the data. According to the space layout data of the workshop equipment, the equipment is divided into automatic acquisition equipment and manual input equipment, so that the data sources are distinguished. The data sources of the equipment represented by the automatic acquisition equipment identification data can be acquired asynchronously and multisource from the industrial Internet of things sensor array, and the data sources of the equipment represented by the manual entry equipment identification data can be acquired through man-machine interaction. Such partitioning provides a source of data collected in different ways, providing diversity and flexibility for subsequent data processing and quality assessment. Based on the automatic acquisition of the equipment identification data, aiming at equipment in a target production workshop, an industrial Internet of things sensor array is deployed, and data acquisition of the equipment is realized. Product production data and equipment performance parameters can be acquired through asynchronous multi-source data acquisition. Such data is of great value for monitoring and analysis of the production process. And recording important data in the production operation process by performing man-machine interaction data acquisition. The production operation data manually recorded are supplementary, can provide different visual angles and information from the automatically collected data, and enrich the diversity of data sources. And carrying out self-adaptive manual error correction on the production operation data recorded by the manual operation, and correcting possible human errors through a correction process. Thus, the accuracy and consistency of the manually entered data can be improved, and the reliability and comparability of the data are ensured. The multi-mode evaluation is carried out on the product production data and the optimized manually-recorded production operation data, and the information of a plurality of data sources is comprehensively considered, so that the product quality index is obtained. The evaluation mode can more comprehensively understand the quality condition of the product, comprehensively analyze the information of different data sources and provide more accurate and comprehensive quality indexes of the product. The information of different data sources is integrated into a comprehensive industrial data set through heterogeneous data aggregation by intelligently weighting the product production data, the equipment performance parameters and the product quality indexes. Such a dataset includes product production data, equipment performance parameters, and product quality metrics, providing a comprehensive view of the entire production process.

Preferably, the data quality impact factor mining module S2 specifically is:

S21: acquiring historical product production data and historical equipment performance parameters;

Specifically, for example, a data source of historical product production data and historical equipment performance parameters may be determined. This may be a production database, an equipment monitoring system, log files, sensor data, etc. Ensuring that a suitable data source containing the required data is available. Depending on the source of the data, it may be desirable to retrieve and extract the data using corresponding data extraction tools, API calls, database queries, or log analysis techniques.

S22: detecting low-quality data of the historical product production data to obtain historical low-quality production data;

Specifically, for example, an evaluation index of low quality data may be defined according to specific data quality requirements and application requirements. These metrics may include requirements in terms of data integrity, accuracy, consistency, etc. And selecting a proper low-quality data detection method according to the defined evaluation index. For example, methods such as statistical analysis, data mining, rule detection, etc. may be used to detect low quality data. And detecting historical product production data according to the selected low-quality data detection method. The detection script may be written or an existing data quality tool may be applied using corresponding tools and techniques, such as the pandas library and the numpy library of Python. Historical low quality production data is extracted from the test results. And identifying and extracting the low-quality data according to the mark or the threshold value of the detection result to form a data set containing historical low-quality production data.

S23: performing time sequence synchronous alignment on the historical product production data and the historical equipment performance parameters to obtain a synchronous historical production-equipment data set;

In particular, for example, the timestamp fields of the product production data and the device performance parameter data may be determined and the two data sets time-series aligned to ensure that they correspond in time. This can be done using the pandas library of Python for time series operations and alignment. And combining the aligned product production data and the equipment performance parameter data to form a synchronous historical production-equipment data set. The two datasets may be merged according to the timestamp field using a merge operation (e.g., merge) of the pandas library of Python.

S24: performing equipment operation efficiency evaluation on the historical equipment performance parameters in the corresponding time period of the historical low-quality production data based on the synchronous historical production-equipment data set to obtain equipment operation efficiency data;

Specifically, for example, a period of time to be evaluated may be determined from historical low quality production data. Device performance parameter data corresponding to the evaluation period is extracted from the synchronized historical production-device dataset. The pandas library of Python can be used for time-scale screening and extraction operations. The extracted device performance parameter data is evaluated for operational efficiency using an appropriate method or model. The specific evaluation method can be selected according to the business requirements and domain knowledge, and can be evaluated by using a statistical analysis method, a machine learning model or an algorithm in the professional domain. And obtaining the equipment operation efficiency data in the corresponding time period according to the equipment operation efficiency evaluation result.

S25: mining the data quality influence factors according to the equipment operation efficiency data based on the historical product production data and the historical equipment performance parameters to obtain a product production data quality influence factor set;

Specifically, for example, relevant features may be extracted from historical product production data and historical equipment performance parameters, e.g., product production rates, equipment failure times, equipment run times, etc., may be calculated. This can be done using the pandas library of Python for data processing and feature extraction. Appropriate data mining methods (e.g., correlation analysis, regression analysis, decision trees, etc.) are used to mine the data quality impact factors based on the plant operational efficiency data and related characteristics. These methods can help discover which factors have a significant impact on the quality of the product production data. And obtaining a quality influence factor set of the product production data according to the data mining result. These factors may be features or variables that have a significant impact on the quality of the product production data. These factors may be organized into a collection or data table that records information such as their names and degrees of influence.

S26: and calculating a data quality critical value range according to the product production data quality influence factor set and the historical low-quality production data to obtain a production data acceptable quality interval.

Specifically, for example, the quality influence factor set of the product production data may be used as a reference to analyze the historical low quality production data, and understand the value range of each quality influence factor in the low quality data. Statistical analysis methods (e.g., box graphs, histograms) can be used to observe data distribution and anomalies. A threshold range for each quality impact factor is calculated based on the historical low quality production data and the product production data quality impact factor set. The calculation method of the critical value can be determined according to statistical indexes (such as mean value and standard deviation) or domain knowledge. And determining an acceptable quality interval of the production data according to the calculated critical value range of the quality influence factor. This interval may be defined as a range between low quality data and high quality data, with data smaller than the range being considered low quality data.

The present invention can create a data set containing past production and equipment information by acquiring historical product production data and historical equipment performance parameters. These data are records of past production processes and can provide detailed information about product production and equipment performance. By performing low quality data detection on historical product production data, low quality data present in the data can be identified and marked. Low quality data may include outliers, missing values, duplicate values, etc., which negatively impact the accuracy and reliability of the data. By identifying low-quality data, guidance can be provided for subsequent data cleaning and processing, and the improvement of data quality is ensured. By synchronizing alignment of the time series of historical product production data and historical equipment performance parameters, it is ensured that they have the same time axis and sampling interval. In this way, a correspondence between production data and equipment data can be achieved, and a synchronized historical production-equipment dataset is constructed. Based on the synchronized historical production-plant data sets, plant performance parameters over a corresponding period of time for the historical low quality production data are evaluated. Thus, the operation efficiency of the equipment when the low-quality data occur can be analyzed, and the operation efficiency data of the equipment can be obtained. The equipment operation efficiency data can help to know the operation conditions of the equipment under different conditions, and provide important basis for subsequent data quality influence factor mining. And mining the data quality influence factors based on the historical product production data, the historical equipment performance parameters and the equipment operation efficiency data. By analyzing the relationship between these data, factors that have an important impact on the quality of the product production data can be identified. These factors may include device performance, data acquisition methods, environmental conditions, etc., that have a positive or negative impact on data quality. By mining these influencing factors, the formation mechanism of data quality can be deeply understood, and guidance and optimization suggestions are provided for improving the data acquisition and production process. And calculating a data quality critical value range from the product production data quality influence factor set and the historical low-quality production data. By analyzing the relation between the historical low quality production data and the quality impact factor, a critical value range of the data quality, i.e. a data quality within an acceptable range, can be determined. Thus, a reference can be provided for subsequent data quality control and anomaly detection, and the production data is ensured to be within an acceptable quality interval.

Preferably, S25 is specifically:

s251: comparing the equipment operation efficiency data with a preset equipment operation efficiency critical value;

Specifically, for example, a logic determination operation (e.g., greater than, less than, equal to) may be used to determine whether the device operating efficiency data satisfies a preset device operating efficiency threshold condition.

S252: when the equipment operation efficiency data is smaller than or equal to a preset equipment operation efficiency critical value, performing Taylor matrix decomposition on the historical equipment performance parameters to obtain an equipment performance influence factor tensor;

Specifically, for example, when the device operating efficiency data is less than or equal to a preset device operating efficiency threshold, the historical device performance parameter data may be taylor matrix decomposed (Tucker decomposition). The tyla matrix decomposition is a high-order tensor decomposition method, and can decompose high-dimensional data into products of low-dimensional core tensors and modal matrices. Taylor matrix decomposition may be implemented using the TensorLy library of Python or other tensor decomposition tool library. And obtaining the device performance influence factor tensor according to the Taylor matrix decomposition result. The tyla matrix decomposition decomposes historical device performance parameter data into a core tensor and a modal matrix, wherein the modal matrix represents different influencing factors of device performance. The influence factors can be organized into a tensor or data structure, and information such as names and influence degrees on the device performance is recorded.

S253: performing hypergraph convolution feature extraction on the device performance influence factor tensor to obtain a device performance influence factor set;

Specifically, for example, a hypergraph may be constructed based on the device performance impact factor tensor. Hypergraph is a graph structure in which nodes represent influence factors and edges represent relationships between influence factors. The definition of nodes and edges in the hypergraph can be determined according to the characteristics and domain knowledge of the device performance impact factor tensor. The hypergraph is feature extracted using a hypergraph convolutional neural network (Graph Convolutional Network, GCN) or other hypergraph convolutional method. Hypergraph convolution is a graph convolution method suitable for hypergraph structures, and can capture high-order relations between nodes. Hypergraph convolution feature extraction may be implemented using the DGL library, pyTorch Geometric library, or other graph neural network library of Python. And obtaining a device performance influence factor set according to the result of the hypergraph convolution feature extraction. These impact factors may be node features extracted by the hypergraph convolution, representing the importance and relevance of the device performance impact factors. These factors may be organized into a collection or data table that records information such as their names and the degree of impact on device performance.

S254: when the equipment operation efficiency data is larger than a preset equipment operation efficiency critical value, performing supply chain tracing on a target production workshop to obtain supply chain link data;

specifically, for example, when the device operation efficiency data is greater than a preset device operation efficiency threshold, supply chain tracing is performed. And acquiring supply chain link data related to the target production workshop by tracking information such as raw materials, parts or assembly processes. This may involve data exchanges and queries with suppliers and other partners. And obtaining the supply chain link data of the target production workshop according to the supply chain tracing result. Such data may include raw material sources, supplier information, production lots, quality control records, and the like. The data may be organized into a data set or table that records the relevant supply chain information.

S255: performing supply chain stability impact simulation on the supply chain link data to obtain a supply chain disturbance factor set;

specifically, for example, the supply chain link data may be simulated using a supply chain stability impact simulation method. This may simulate uncertainty and disturbance conditions in the supply chain by introducing random variables, model parameter adjustments, or other simulation methods. Supply chain stability impact simulation can be implemented using a simulation library of Python (e.g., simPy, numPy). And obtaining a supply chain disturbance factor set according to the result of the supply chain stability impact simulation. These factors may be disturbance variables introduced in the simulation, representing instability and risk factors in the supply chain. These factors may be organized into a collection or data table that records information such as their name and the extent of impact on the stability of the supply chain.

S256: collecting historical man-machine interaction data of a target production workshop to obtain historical man-machine interaction data; performing behavior pattern mining and deviation detection on the historical human-computer interaction data to obtain a factor set affected by manual input operation errors;

Specifically, for example, a sensor, recording device, or other data collection method may be used to obtain data related to human-machine interaction, including operational records, input data, timestamps, and the like. And analyzing and mining the historical human-computer interaction data by using a behavior pattern mining method. Machine learning algorithms (e.g., clustering, association rule mining) or sequence pattern mining methods (e.g., sequence pattern, sequence frequent item mining) may be used to discover patterns of behavior of human-machine interactions. Behavior pattern mining may be implemented using a machine learning library of Python (e.g., scikit-learn, tensorFlow) or a sequence analysis library (e.g., prefixSpan, GSPY). And detecting deviation of the behavior mode, and identifying potential manual input operation errors. Deviations may be detected using anomaly detection algorithms (e.g., outlier detection, statistical methods) or specially designed rules and models. And according to the result of the anomaly detection, determining possible manual input operation errors. And obtaining a factor set affected by the manual input operation errors according to the deviation detection result. These factors may be characteristics, patterns or other indicators related to human input errors, indicative of the extent and frequency of their impact on the operation of the production plant. These factors can be organized into a collection or data table that records information about their names and the extent of impact on the operation of the production plant.

S257: and taking the equipment performance influence factor set, the supply chain disturbance factor set and the manual input operation error influence factor set as the product production data quality influence factor set.

Specifically, for example, the set of device performance impact factors, the set of supply chain disturbance factors, and the set of manual entry misoperation impact factors may be integrated. Depending on the structure and requirements of the data, the sets of factors may be consolidated into one data set or data table using a data processing tool (e.g., the pandas library of Python). Ensuring that each factor has a corresponding identifier or association information for subsequent analysis and processing. And taking the integrated factor set as a product production data quality influence factor set. This set of influencing factors includes factors that affect the quality of the product production data in several aspects, such as equipment performance, supply chain stability, and manual entry operations. These factors can be organized into a collection or data table that records information such as their names and the degree of impact on the quality of the product production data.

The invention can evaluate the operation condition of the equipment by comparing the equipment operation efficiency data with the preset equipment operation efficiency critical value. If the equipment operation efficiency data is lower than or equal to a preset critical value, the operation state of the equipment can be judged to have problems, and the influence factors of the equipment performance need to be further analyzed. When the equipment operation efficiency data is smaller than or equal to a preset equipment operation efficiency critical value, analyzing the historical equipment performance parameters by adopting a Taylor matrix decomposition method. The tyla matrix decomposition may extract the impact factor of the device performance, i.e. the device performance impact factor tensor, from the historical data. These factors can be used for subsequent data analysis and mining to help understand the changes in device performance and influencing factors. By performing hypergraph convolution feature extraction on the device performance influence factor tensor, key features of the device performance can be extracted from the hypergraph convolution feature extraction. Hypergraph convolution is a method for processing graph data that captures complex relationships between nodes in a graph. By means of hypergraph convolution feature extraction, a device performance influence factor set can be obtained, the factors have higher characterization capability, and the influence factors of the device performance can be better described. And when the equipment operation efficiency data is larger than a preset equipment operation efficiency critical value, indicating that the operation state of the equipment is normal. In this case, the supply chain link data of the target production plant, including information of stock of raw materials, suppliers, production process, etc., can be tracked by a supply chain tracing method. And (3) performing supply chain stability impact simulation on the supply chain link data, simulating the supply chain stability under different conditions, and extracting disturbance factors in the supply chain from the supply chain stability. These perturbation factors can help understand the vulnerability and stability of the supply chain, providing basis for improving supply chain management and coping with potential risks. By performing behavior pattern mining and deviation detection on the historical human-computer interaction data, deviation and errors in manual input operation can be identified. The extracted manual input misoperation influence factor set can help to analyze and improve the human-computer interaction process, and reduce the influence of human factors on the quality of product production data. And integrating the previously extracted equipment performance influence factor set, the supply chain disturbance factor set and the manual input operation error influence factor set together to form a product production data quality influence factor set. The set contains key factors influencing the quality of the product production data, and can be used for evaluating and optimizing the product production process and improving the data quality and the production efficiency.

Preferably, S26 is specifically:

s261: performing defect cost estimation on the historical low-quality production data to obtain defect cost data;

Specifically, for example, historical low quality production data may be analyzed and calculated using a defect cost estimation method. Defect costs include direct costs (e.g., waste disposal, rework costs), indirect costs (e.g., loss of production, customer complaint handling costs), and the like. The specific cost calculation method can be selected and customized according to the requirements and actual conditions of enterprises. Excel or a specialized cost estimation tool can be used for calculation and analysis. And obtaining defect cost data according to the defect cost estimation result. The data may be cost values for each defect type or production lot for measuring the economic impact of low quality production data on the enterprise. The data can be organized into a collection or data table, and information such as the name, the corresponding defect type or production lot, the influence degree on the enterprise economy and the like are recorded.

S262: setting an acceptable loss threshold value for the defect cost data to obtain the acceptable loss threshold value;

Specifically, for example, an acceptable loss threshold may be set using an appropriate method or model. Common methods include risk assessment, cost-effectiveness analysis, quality objective setting, and the like. The appropriate method may be selected according to the needs of the enterprise. This threshold should be able to balance the economic investment and quality risks, ensuring that the enterprise operates within acceptable limits of loss. And obtaining an acceptable loss threshold according to the result of the threshold setting. This threshold may be a specific value that represents the maximum defect cost or loss amount that the enterprise can accept. A range or interval is also possible to take into account the different situations and flexibilities. Ensuring that specific values of threshold values and associated definitions or specifications are recorded.

S263: performing acceptable historical low-quality production data inverse calculation on the historical low-quality production data according to an acceptable loss threshold value to obtain acceptable historical low-quality production data;

Specifically, for example, historical low quality production data may be screened for data meeting threshold requirements based on an acceptable cost range of defects. Screening and filtering can be performed according to the defect cost data and the threshold value, and production data with the defect cost within an acceptable range can be found. Acceptable historical low quality production data is obtained that meets the acceptable defect cost range.

S264: acquiring corresponding equipment performance parameters of acceptable historical low-quality production data;

Specifically, for example, corresponding equipment performance parameters may be obtained based on acceptable historical low quality production data. Parameters and metrics related to device performance are determined. These parameters may be the operating status of the device, the operating efficiency, the fault record, etc. According to the requirements and the actual conditions, proper performance parameters are selected to evaluate the running quality of the equipment. Acceptable historical low quality production data is integrated and correlated with equipment performance parameters. And according to the production batch or the product defect information, corresponding equipment performance parameter data are found. This may be matched and associated by lot number, time stamp, or other unique identifier. And acquiring corresponding equipment performance parameters according to the associated equipment performance parameter definition and the data integration result. This may involve extracting desired performance parameters from a device monitoring system, sensor data, maintenance records, or other related data sources.

S265: weighting the influence factors of the acceptable historical low-quality production data and the corresponding equipment performance parameters of the acceptable historical low-quality production data based on the product production data quality influence factor set to obtain weighted influence factor data;

Specifically, the influence factors may be weighted using a suitable method or model, for example. Methods that may be used include expert assessment, analytic Hierarchy Process (AHP), weighted average, and the like. And selecting the most suitable weighting method according to the requirements of enterprises and available data. Ensuring that the weighting process is repeatable and has reasonable logic. And weighting the influence factors of the acceptable historical low-quality production data and the corresponding equipment performance parameters according to the selected weighting method. This may involve calculating the weight of each influence factor using a weighting table or formula from a defined set of influence factors. Calculation and analysis were performed using Excel or similar tools. And weighting the influence factors of the acceptable historical low-quality production data and the corresponding equipment performance parameters according to the selected weighting method.

S266: calculating a data quality critical value range of the acceptable historical low-quality production data and corresponding equipment performance parameters of the acceptable historical low-quality production data according to the product production data quality influence factor set and the weighted influence factor data, and obtaining a production data acceptable quality interval;

Specifically, for example, a suitable method or model may be used to calculate the critical value range of data quality. Methods that may be used include statistical analysis based, quality control charts, expert judgment, and the like. And selecting a proper method according to the requirements of enterprises. The reasonable and reliable calculation method is ensured and the calculation method accords with the target. And carrying out weighted summation on the acceptable historical low-quality production data and the equipment performance parameters according to the weighted result of the influence factors to obtain a comprehensive quality evaluation value. This may be achieved by multiplying the weighted impact factor for each data point with the corresponding data value using a weighted itemized addition method and summing the results to obtain a composite estimate. And converting the comprehensive quality evaluation value into a data quality critical value range according to the data quality critical value calculation method. This may be a range determination by accepting data quality thresholds corresponding to historical low quality production data, or by setting upper and lower thresholds according to enterprise needs and criteria to determine an acceptable range of data quality. The data quality threshold range is calculated for the integrated quality assessment value using an appropriate tool or method. This may involve calculation and processing using statistical analysis software (e.g., excel, the numpy library of Python). And obtaining an acceptable quality interval of the production data according to the calculation result. This may be a range of values, e.g., upper and lower, representing an acceptable range of data quality.

The invention can calculate the cost related to the defects, such as the reworking cost, the rejection cost and the like by estimating the defect cost of the historical low-quality production data. The defect cost data is helpful to evaluate the influence of low-quality data on the production process and the enterprise economy, and provides quantitative basis for subsequent decisions. By analyzing and considering the defect cost data, an acceptable loss threshold may be set. This threshold represents the maximum degree of loss an enterprise is willing to bear during production, and low quality data exceeding this threshold will be considered an unacceptable quality problem. The setting of the acceptable loss threshold helps to clarify the goals and limitations of the quality of the production data. And reversely solving the historical low-quality production data according to the set acceptable loss threshold value, and screening out the data meeting the acceptable standard. These acceptable historical low quality production data are data within acceptable losses of the enterprise. From the acceptable historical low quality production data, the device performance parameters associated with these data are obtained. These parameters reflect key performance indicators of the device during production, and may have an association with the generation of low quality data. By obtaining these device performance parameters, the impact of the device on low quality data can be further analyzed and understood. And weighting the acceptable historical low-quality production data and corresponding equipment performance parameters according to the quality influence factor set of the production data. These impact factors may be weighted according to their importance to the quality of the product production data to reflect their relative importance to quality. By weighting the impact factor data, the degree of impact of equipment performance on acceptable historical low quality production data can be more accurately assessed. Based on the product production data quality impact factor set and the weighted impact factor data, a data quality threshold range of acceptable historical low quality production data and corresponding equipment performance parameters is calculated. This quality threshold range defines an acceptable data quality boundary beyond which data will be considered unacceptably low quality data. By determining an acceptable quality interval for production data, guidance can be provided for quality control and improvement of the data

Preferably, the data quality evaluation module S3 specifically includes:

S31: according to the quality influence factor set of the product production data, non-influence factor attribute data are removed from the product production data, and simplified product production data are obtained;

specifically, for example, for each data attribute, it may be determined whether it is an influence factor attribute according to the product production data quality influence factor set. The influence factor attributes are those that have an important impact on the quality of the product production data. And screening and eliminating the production data of the product based on the definition and the rule of the influence factor attribute. And eliminating data which do not belong to the influence factor attribute set. Ensuring that the data culling process is traceable and repeatable. Recording the data attribute and the corresponding rule. And after the non-influence factor attribute data is removed, a simplified product production data set is obtained.

S32: constructing a product production data quality evaluation model based on the product production data quality influence factor set;

Specifically, for example, an appropriate evaluation model may be selected to construct a product production data quality evaluation model. Evaluation models that may be used include logistic regression, support vector machines, decision trees, etc. And selecting the most suitable model according to the characteristics and the evaluation targets of the data. Based on the product production data quality impact factor set, a data set for model construction is prepared. This includes extracting and sorting data attributes associated with the impact factors from the product production data, and marking quality assessment results (e.g., good/bad, pass/fail) for each data sample. The prepared data set is divided into a training set and a test set. Training of the model is performed using the training set data. According to the selected evaluation model algorithm, the quality of the product production data can be accurately evaluated by learning and adjusting model parameters. The trained model is validated using the test set data. And (3) evaluating the performance of the model on unseen data, and checking the evaluation accuracy and reliability of the model on the quality of the product production data, thereby obtaining the quality evaluation model of the product production data.

S33: inputting the simplified product production data into a product production data quality evaluation model to perform data quality quantitative evaluation to obtain a product production data quality index; the product production data quality index is compared with the production data acceptable quality interval.

Specifically, for example, the reduced product production data may be input into a product production data quality evaluation model for evaluation. The input product production data is predicted or classified using an evaluation model. The model evaluates the quality of the data according to the learned rules and generates a corresponding quality index. And obtaining a corresponding product production data quality index for each data sample. The index represents the data quality level of the sample and may be a continuous value or a discrete evaluation result. And obtaining the product production data quality index of each data sample according to the output result of the evaluation model. And comparing the quality index of the production data of the product with the acceptable quality interval of the production data to determine whether the quality of the data meets the requirement.

According to the invention, the simplified product production data is obtained by eliminating the non-influence factor attribute in the product production data, and the quality and the relativity of the data are improved. And identifying and removing attribute data which have no direct influence or no influence on the quality evaluation of the product production data according to the quality influence factor set of the product production data. The value of this is to reduce the complexity of the data set, remove redundant information, and make subsequent data analysis and evaluation more accurate and efficient. By utilizing the quality influence factor sets of the product production data, an evaluation model is established, and the model can be calculated according to the weight of each influence factor and related data, so that the quality score of the product production data is obtained. The value of constructing the product production data quality evaluation model is to provide an objective and quantifiable evaluation means, so that enterprises can know the quality level of the product production data more accurately. The quality index of the product production data is obtained through calculation and analysis by inputting the product production data subjected to the simplification treatment into a product production data quality evaluation model. The quality index may reflect the quality level of the data to further determine whether the data meets expected quality requirements. By comparing the quality index of the production data of the product with the acceptable quality interval of the production data, the qualification of the data can be determined, measures can be timely taken to improve and optimize the production process, and the quality of the data is ensured to reach an acceptable level.

Preferably, the data quality improvement decision module S4 is specifically:

s41: when the quality index of the production data of the product is smaller than the lower limit of the acceptable quality interval of the production data, performing intelligent production stop line operation on a target production workshop;

specifically, for example, when the product production data quality index is less than the lower limit of the acceptable quality interval of the production data, the production line of the target production plant is stopped by the automated control system according to a preset production stop line strategy. The intelligent shutdown may be implemented using rule-based logic control or intelligent control methods based on machine learning.

S42: carrying out causal tracing rule construction on the product production data according to the product production data quality influence factor set to obtain causal tracing rule data;

Specifically, for example, association rule mining algorithms (e.g., apriori algorithms) may be used to discover the association and causal relationships between product production data quality impact factor sets to product production data. The rule building process is implemented using a data mining and association rule mining library of Python (e.g., mlxtend). And converting the rule into a form with high readability according to the constructed causal tracing rule, and generating causal tracing rule data. The rule data may include information on the condition, result, and support of the rule.

S43: acquiring current supply chain data; carrying out data anomaly root tracing on the production data of the product based on the causal tracing rule data, the equipment performance parameters and the current supply chain data to obtain production data anomaly incentive data;

Specifically, for example, various links associated with the supply chain, including raw material suppliers, logistics information, production plans, etc., may be provided with data interfaces or collection means to obtain current supply chain data. The data may be obtained using an Enterprise Resource Planning (ERP) system, a Supply Chain Management (SCM) system, or other related system. And (3) correlating and analyzing the current supply chain data and the equipment performance parameters based on the causal tracing rule data so as to trace the abnormal source of the product production data. Data mining and association rule mining methods can be used to match and infer based on conditions and results in the rule data to find out causes of anomalies in the production data. Data processing and analysis is performed using a Python data analysis and association rule mining library (e.g., mlxtend). And extracting incentive data causing production data abnormality according to the result of tracing the source of the data abnormality. These incentive data may include information regarding supply chain links, equipment status, process parameters, and the like.

S44: carrying out data quality improvement optimization decision on a target production workshop based on historical product production data, historical low-quality production data and production data abnormality cause data to obtain a production data quality enhancement strategy;

Specifically, for example, an optimization decision for data quality improvement may be made in combination with historical product production data, historical low quality production data, and production data anomaly causing data. Based on the results of the data analysis, correlations between factors leading to low quality production and anomaly causing data are identified and analyzed. And according to the analysis result, making a series of optimization decisions for improving the quality of the data, including adjusting production process parameters, improving supply chain management, optimizing equipment maintenance plans and the like.

S45: and carrying out strategy deep fusion implementation on the target production workshop according to the production data quality enhancement strategy.

Specifically, for example, the target production shop may be applied in time according to the countermeasure in the production data quality enhancement policy.

According to the invention, the quality index of the production data of the product is monitored, and if the index is found to be lower than the lower limit of the acceptable quality interval of the production data, the intelligent production stop line operation is triggered. Low quality industrial data can negatively impact the supply chain. Bad data can be prevented from being transmitted to a downstream link through timely line stopping operation, and the data quality in a supply chain is ensured. By constructing the causal tracing rule, the factors with larger influence on the quality of the production data of the product can be identified, and a reference basis of causal relation is provided, so that the understanding and solving of the data quality problem are facilitated. These rules can help the enterprise understand how much each factor affects the quality of the product production data, thereby better managing and improving the data quality. The generation of causal trace rule data provides enterprises with the ability to learn data anomalies in depth, helping to determine the root cause of data quality problems. And carrying out abnormal root tracing on the production data of the product by acquiring the current supply chain data and combining the previously constructed causal tracing rule data and the equipment performance parameters. By analyzing the cause and source of the data anomalies, specific causes of production data anomalies, including problems in the supply chain, equipment performance, etc., can be obtained. Such traceability helps to pinpoint the problem, provides a basis for solutions, and provides guidance for improving the quality of production data. And carrying out data quality improvement optimization decision on the target production workshop by utilizing the historical product production data, the historical low-quality production data and the production data anomaly incentive data obtained before. By analyzing the historical data and the abnormal cause data, the main reasons of the quality problems of the production data can be identified, and corresponding improvement strategies and measures are formulated. These decisions may relate to aspects of supply chain management, equipment maintenance, process adjustments, etc., with the aim of eliminating or reducing factors that cause anomalies in the data, improving the quality and stability of the production data. And deeply fusing the strategies into the actual production process of the target production workshop according to the formulated production data quality enhancement strategies. This may include updating and adjusting supply chain management procedures, equipment maintenance planning, data collection and monitoring systems, etc. to ensure efficient implementation and continued improvement of policies. By implementing the deep fusion of the strategy, the stability and the sustainability of the data quality can be improved.

Preferably, S44 is specifically:

s441: acquiring a historical production data anomaly incentive data set;

Specifically, for example, query statements may be written to obtain historical production data anomaly incentive datasets based on the type and structure of the enterprise database.

S442: carrying out heterogeneous incentive recognition and classification on the historical production data abnormal incentive data set to obtain a heterogeneous incentive data set, wherein the heterogeneous incentive data set comprises equipment fault heterogeneous incentive data, manual operation error heterogeneous incentive data, supply chain interruption heterogeneous incentive data and environment influence heterogeneous incentive data;

Specifically, for example, relevant features may be extracted from a historical production data anomaly incentive dataset for identification and classification of heterogeneous incentives. Features may include, but are not limited to, device identification, time stamps, operator information, supply chain information, environmental parameters, and the like. And constructing a heterogeneous incentive recognition and classification model according to the feature extraction result. The model may be constructed using various machine learning algorithms (e.g., decision trees, support vector machines, random forests, etc.) or deep learning methods (e.g., neural networks). A labeled training data set is prepared, wherein the training data set comprises heterogeneous incentive category labels which are manually labeled or heterogeneous incentive category labels which are automatically learned and labeled by a machine. And predicting and classifying the historical production data abnormal incentive data set by using the trained heterogeneous incentive recognition and classification model. And correlating the prediction result with the original data set to generate a heterogeneous incentive data set comprising equipment fault heterogeneous incentive data, manual operation error heterogeneous incentive data, supply chain interruption heterogeneous incentive data and environment influence heterogeneous incentive data.

S443: carrying out abnormal occurrence probability statistical analysis on each type of heterogeneous incentive data in the heterogeneous incentive data set based on the historical product production data, the historical low-quality production data and the historical production data abnormal incentive data set to obtain various abnormal incentive occurrence probability data;

Specifically, for example, a statistical analysis of the occurrence probability of abnormality may be performed for each type of heterogeneous incentive data in the heterogeneous incentive data set. For each type of heterogeneous incentive data, corresponding low-quality data in the historical product production data is associated with the historical production data anomaly incentive data set. The statistical analysis may calculate the probability of occurrence of an anomaly for each type of heterogeneous incentive data based on the associated data, such as using frequency statistics or probability distribution fitting.

S444: carrying out economic loss risk distribution analysis on each type of heterogeneous incentive data in the heterogeneous incentive data set to obtain economic loss risk distribution data;

In particular, for example, economic loss data associated with each type of heterogeneous incentive may be collected, ensuring that the data is accurate, complete and reliable. And correlating the economic loss data with each type of data in the heterogeneous incentive data set to ensure that the corresponding relation is correct. And carrying out economic loss risk distribution analysis on each type of heterogeneous incentive data in the heterogeneous incentive data set. The economic loss risk profile for each type of heterogeneous incentive data is calculated using appropriate statistical methods and risk assessment models, such as probability distribution analysis, risk value (VaR) calculation, and the like. These methods may be selected according to specific requirements, such as normal distribution, monte Carlo simulation, and the like.

S445: constructing a data abnormal incentive knowledge graph for the heterogeneous incentive data set, the occurrence probability data of various abnormal incentives and the economic loss risk distribution data to obtain the data abnormal incentive knowledge graph;

Specifically, for example, each type of heterogeneous incentive data in the heterogeneous incentive data set may be used as a node, and the probability data of occurrence of abnormality obtained by the probability statistical analysis may be used as a node attribute. And correlating the economic loss risk distribution data with various heterogeneous incentive data to form edges, wherein the edges represent the relationship between abnormal incentive and economic loss. The data anomaly predisposition knowledge graph may be constructed using graph databases or knowledge graph construction tools such as Neo4j, apache Jena, etc.

S446: carrying out abnormal source matching on the production data abnormal incentive data and the data abnormal incentive knowledge graph to obtain production data abnormal incentive matching result data;

Specifically, for example, production data abnormality cause data may be matched with a data abnormality cause knowledge map. And according to the attribute of the production data abnormal incentive data, matching with the node attribute in the data abnormal incentive knowledge graph, and searching similar or related nodes. Query and matching operations may be performed using a graph database or a knowledge graph query language (e.g., the Cypher query language).

S447: and carrying out data quality improvement optimization decision on the target production workshop based on the data anomaly incentive knowledge graph and the production data anomaly incentive matching result data to obtain a production data quality enhancement strategy.

Specifically, for example, the data quality problem and abnormal situation of the target production shop may be analyzed based on the production data abnormality cause matching result data. And evaluating and analyzing the data quality problems by considering information such as the occurrence probability of the abnormality, economic loss risk distribution, node attributes related to the cause of the abnormality and the like. And formulating a production data quality enhancement strategy based on the data anomaly causing knowledge graph and the data quality analysis result. And determining the target and the direction of data quality improvement according to the relationship between the attribute and the edge of the abnormal incentive node in the knowledge graph. Data quality improvement measures aiming at different abnormal causes can be formulated, such as data acquisition optimization, sensor calibration, process parameter adjustment and the like.

The present invention provides a data set of historical production data anomaly causes, including various factors that cause data anomalies, by acquiring a data set of historical production data anomaly causes. This provides the basis for subsequent anomaly cause analysis and optimization decisions. By identifying and classifying the historical production data abnormal causes, different types of abnormal causes are classified. This helps enterprises to clearly understand the different types of sources of anomalies and to conduct anomaly root cause analysis and processing specifically. Heterogeneous incentive data sets are generated, including different types of anomaly incentive data for equipment failure, human operational errors, supply chain interruptions, and environmental impact. This provides the basis for subsequent statistical analysis and risk assessment. Through statistical analysis, the abnormal occurrence probability of each type of heterogeneous incentive data is calculated. This provides a basis for the enterprise to assess the risk level of different causes of anomalies. The method helps enterprises to know the occurrence frequency of different types of abnormal incentive data, thereby being beneficial to formulating corresponding risk prevention and data quality improvement strategies. And (3) carrying out economic loss risk analysis on each type of abnormal incentive data in the heterogeneous incentive data set to help enterprises evaluate the economic loss risk degrees of different abnormal incentives. Economic loss risk distribution data is provided that helps businesses understand and prioritize those anomaly incentives that may lead to greater economic losses. And organically combining the heterogeneous incentive data, the abnormality occurrence probability data and the economic loss risk distribution data by constructing a data abnormality incentive knowledge graph. The data anomaly incentive knowledge graph provides comprehensive anomaly incentive information, helps enterprises to better understand the association and influence among different anomaly incentives, and provides basis for subsequent root cause matching and data quality improvement decision. And matching the production data anomaly causing data with the data anomaly causing knowledge graph to determine the root cause of the production data anomaly. By matching the result data, enterprises can accurately know specific abnormal causes causing data abnormality, and provide guidance and basis for solving the data quality problem. And based on the data anomaly incentive knowledge graph and the production data anomaly incentive matching result data, making a data quality improvement optimization decision aiming at the target production workshop. By formulating a production data quality enhancement strategy, enterprises can effectively improve the data quality, reduce the possibility of abnormality occurrence and improve the product quality and the production efficiency. The method can help enterprises to deeply understand the cause of abnormality, evaluate risks, solve the problem of data quality and finally improve the quality and benefit of production data.

Preferably, S447 is specifically:

S4471: when the production data abnormal cause matching result data is equipment fault heterogeneous cause data, scanning the health state of equipment in a target production workshop to obtain fault equipment identification data; performing equipment maintenance decision support on equipment corresponding to the fault equipment identification data to obtain an optimized maintenance strategy of workshop equipment;

specifically, for example, when the production data anomaly cause matching result data is equipment failure heterogeneous cause data, the equipment in the target production plant may perform health status scanning. The operation state, fault information and the like of the equipment can be obtained in real time or periodically by using technical means such as a sensor, monitoring equipment or a remote monitoring system. And identifying the equipment with faults according to the equipment health state scanning result, and generating fault equipment identification data. The failed device identification data may include information of device number, device type, failure level, etc. for subsequent device repair decision support. And carrying out equipment maintenance decision support based on the fault equipment identification data. The identification data of the fault equipment can be utilized to carry out fault cause analysis and maintenance priority ranking by combining equipment maintenance records, maintenance history data and the like. And according to factors such as maintenance priority and resource availability, a workshop equipment optimizing maintenance strategy such as equipment replacement, maintenance plan adjustment and the like is formulated.

S4472: when the production data abnormal cause matching result data is heterogeneous cause data of manual operation errors, performing man-machine interaction analysis on key processes in a target production workshop to obtain man-machine optimization demand data; virtual reality interactive training is implemented based on the man-machine optimization demand data, and a workshop personnel skill improvement strategy is obtained;

Specifically, for example, when the production data abnormal cause matching result data is the artificial misoperation heterogeneous cause data, the key process in the target production shop may be focused on according to the artificial misoperation heterogeneous cause data in the production data abnormal cause matching result data. And (3) performing man-machine interaction analysis, and evaluating the problem of manual operation errors possibly caused by man-machine interface design, process operation flow, tool equipment and the like. And obtaining the human-computer optimization demand data based on the human-computer interaction analysis result. The human-machine optimization requirement data may include improvement requirements for interface design, adjustment suggestions for process flows, requirements for training and skill improvement, and the like. Based on the human-machine optimization demand data, virtual reality interactive training is implemented. And a simulated production working environment is constructed by utilizing a virtual reality technology, so that workshop personnel can carry out practical operation and skill improvement through interactive training. The training content can comprise a correct operation flow, a processing method for coping with abnormal conditions and the like, so that the operation skills of workshop staff are improved and manual operation errors are reduced. And formulating a workshop personnel skill improvement strategy according to the result and feedback of the virtual reality interactive training. Further training plans and measures, such as periodic training, skill certification, etc., may be determined based on the performance assessment of the training to ensure continued improvement in skill levels of the plant personnel.

S4473: when the production data abnormal cause matching result data is supply chain interruption heterogeneous cause data, a supply chain digital twin model is constructed, and workshop raw material logistics data are obtained; carrying out intelligent logistics scheduling optimization implementation on a target production workshop by combining a supply chain digital twin model with workshop raw material flow data to obtain a workshop supply chain stabilization strategy;

Specifically, for example, when the production data abnormal cause matching result data is supply chain interruption heterogeneous cause data, a supply chain digital twin model is established based on the structure and operation flow of the supply chain. The digital twin model is a virtual supply chain image that can model the links and interrelationships in the supply chain. Feed stream data for a target production plant is collected. Such data may include supplier information of the feedstock, time of transportation, inventory levels, etc. The plant feed stream data is combined with a supply chain digital twin model. The supply chain is scheduled and optimized using intelligent algorithms and optimization techniques to achieve more stable logistics operation and reduce the risk of disruption. The goals of optimization may include reducing the number of supply chain interruptions, reducing logistics costs, improving logistics efficiency, etc. And formulating a workshop supply chain stability strategy according to the intelligent logistics scheduling optimization result. These policies may include collaborative agreement adjustments with the suppliers, inventory management policy optimization, logistics transportation style adjustments, etc., with the aim of ensuring stability and reliability of the supply chain.

S4474: when the production data abnormal incentive matching result data is environment influence heterogeneous incentive data, performing workshop environment parameter sensing network deployment on a target production workshop to obtain workshop real-time environment parameters; performing intelligent environment control optimization implementation on a target production workshop according to workshop real-time environment parameters to obtain a workshop environment optimization strategy;

Specifically, for example, when the production data abnormal incentive match result data is environmental impact heterogeneous incentive data, an environmental parameter perception network is deployed in the target production plant. This may include sensors, monitoring devices, data acquisition systems, etc. for sensing environmental parameters of the plant, such as temperature, humidity, air quality, etc., in real time. And acquiring real-time environmental parameter data of the target production workshop through the deployed environmental parameter sensing network. And the workshop environment is optimized and regulated by combining with the workshop real-time environment parameter data and using an intelligent algorithm and a control technology. For example, automatic adjustment of air conditioning and ventilation systems is performed based on temperature and humidity data to maintain a suitable operating environment. And formulating a workshop environment optimization strategy according to the intelligent environment control optimization result. These policies may include setting target ranges for environmental parameters, optimizing device scheduling and control policies, etc., in order to provide a comfortable, safe, and efficient working environment.

S4475: taking a workshop equipment optimization maintenance strategy or a workshop personnel skill improvement strategy or a workshop supply chain stabilization strategy or a workshop environment optimization strategy as a production data quality enhancement strategy;

Specifically, for example, when the production data abnormal cause matching result data is one of equipment failure heterogeneous cause data, manual operation error heterogeneous cause data, supply chain interruption heterogeneous cause data, and environmental impact heterogeneous cause data, a quality enhancement policy corresponding to the heterogeneous cause data is taken as the production data quality enhancement policy.

S4476: when the production data abnormal cause matching result data is any one of equipment fault heterogeneous cause data, manual operation error heterogeneous cause data, supply chain interruption heterogeneous cause data and environment influence heterogeneous cause data, the corresponding combination of workshop equipment optimization maintenance strategy, workshop personnel skill improvement strategy, workshop supply chain stability strategy and workshop environment optimization strategy is used as a production data quality enhancement strategy.

Specifically, for example, when the production data abnormal cause matching result data is any one of the equipment failure heterogeneous cause data, the manual operation failure heterogeneous cause data, the supply chain interruption heterogeneous cause data, and the environmental impact heterogeneous cause data, a corresponding combination of the shop floor equipment optimization maintenance policy, the shop floor personnel skill improvement policy, the shop floor supply chain stabilization policy, and the shop floor environment optimization policy is taken as the production data quality enhancement policy.

According to the method, when the production data abnormal cause matching result data is equipment fault heterogeneous cause data, fault equipment existing in a target production workshop is timely found through equipment health state scanning, and the fault equipment is accurately identified. This helps the enterprise to take timely maintenance measures, prevents equipment failure from causing bigger influence to production. And providing support for equipment maintenance decision-making based on the fault equipment identification data. The enterprise can formulate corresponding equipment optimization maintenance strategies according to factors such as the severity degree, the influence range and the like of equipment faults. Effective equipment maintenance can reduce production interruption and industrial production data consistency quality problems, and improve production efficiency and stability. When the abnormal incentive matching result data of the production data is heterogeneous incentive data of the manual operation errors, the problem of the manual operation errors possibly existing in the key working procedures of the target production workshop is deeply known through man-machine interaction analysis. This helps identify potential problem points, improve process design and operational flow, and reduce the occurrence of human operator error. Based on the human-machine optimization demand data, virtual reality interactive training is implemented. The training mode can simulate the real working environment, provide the experience of being in the scene, and help workshop staff to improve the skill and the operation level, thereby ensuring the quality of manually input data in the production process. When the production data abnormal cause matching result data is supply chain interruption heterogeneous cause data, a supply chain digital twin model is built, workshop raw material flow data are obtained, and comprehensive analysis and modeling of the supply chain interruption heterogeneous cause can be achieved. This allows businesses to better understand problems and bottlenecks in the supply chain and predict possible outage situations ahead of time, providing more accurate data support for supply chain management. And carrying out intelligent logistics scheduling optimization implementation on the target production workshop by combining the workshop raw material logistics data through a supply chain digital twin model. By optimizing the logistics scheduling, the on-time supply of raw materials, the reduction of stock backlog, the optimization of transportation routes and the like can be realized. This helps to reduce the risk of supply chain interruption and improves the supply chain stability and operating efficiency of the production plant. When the production data abnormal cause matching result data is environment influence heterogeneous cause data, environmental parameters of a target production workshop, such as temperature, humidity, gas concentration and the like, can be perceived in real time by deploying a workshop environment parameter perception network. This facilitates accurate monitoring and data collection of environmental impact factors. Based on the real-time environmental parameters, the environment of the target production workshop is adjusted and optimized through intelligent environment control optimization implementation. The environmental parameters such as temperature, humidity and the like are adjusted to meet the requirements in the production process, so that the product quality is improved, the production flow is stabilized, and the product quality index is improved. The workshop equipment optimization maintenance strategy is used as a production data quality enhancement strategy, so that the normal operation of equipment can be ensured, the reliability of the equipment can be improved, and the influence of equipment faults on the quality of production data can be reduced. The workshop personnel skill improvement strategy is used as a production data quality enhancement strategy, so that the operation level and skill of staff can be improved, and production data abnormality and quality problems caused by human factors are reduced. The workshop supply chain stability strategy is used as a production data quality enhancement strategy, so that the stability of the supply chain can be ensured, and the influence of supply chain interruption on production data is reduced. The workshop environment optimization strategy is used as a production data quality enhancement strategy, so that the production environment can be optimized, and the influence of environmental factors on production data and product quality is reduced.

Preferably, the production process knowledge graph construction module S5 specifically includes:

s51: carrying out fractal dimension analysis on the quality index of the production data of the product and the abnormal cause data of the production data to obtain fractal correlation data of abnormal data quality;

specifically, for example, fractal dimension analysis algorithms such as Hurst index, fractal dimension calculation, and the like may be used. Inputting the quality index of the production data and the abnormal inducement data of the production data into a selected fractal dimension analysis tool. And running an analysis tool, and calculating the fractal dimension of the data. This will reveal the self-similar features and fractal structure of the data. And integrating the quality index of the production data of the product and the fractal dimension analysis result of the abnormal incentive data of the production data. And extracting fractal correlation data with abnormal data quality, namely correlation information of the abnormal data in fractal dimension.

S52: carrying out process semantic analysis on the production data abnormal incentive data and the production data quality enhancement strategy to obtain abnormal incentive strategy semantic data;

Specifically, for example, natural Language Processing (NLP) technology, semantic analysis algorithms, and the like may be used. The production data anomaly causing data and the production data quality enhancement strategy data are input into a selected process semantic analysis tool. And operating an analysis tool, and extracting semantic information and association relations in the data. This will reveal the semantic association between the anomaly incentive data and the policy data. And combining the abnormal cause and the semantic analysis result of the strategy data. And extracting semantic data of the abnormal incentive strategy, namely semantic relation and semantic representation between the abnormal incentive data and the strategy data. And explaining and analyzing the abnormal incentive strategy semantic data. From the analysis results, the semantic relationship and logic between the anomaly causes and the policies can be understood.

S53: carrying out heterogeneous data graph embedding fusion on the product production data quality index, the production data abnormal incentive data and the production data quality enhancement strategy based on the data quality abnormal fractal association data and the abnormal incentive strategy semantic data to obtain an initial data quality causal knowledge graph;

Specifically, for example, a tool or method suitable for heterogeneous data map embedding may be selected. The heterogeneous data graph embedding tool can embed different types of data and association relationships into a unified graph structure. And expressing the data quality anomaly fractal association data, anomaly incentive strategy semantic data, product production data quality index, production data anomaly incentive data and production data quality enhancement strategy in a heterogeneous data graph mode. In the figure, nodes represent data and policy elements, and edges represent an association relationship between them. The heterogeneous data map is input into a selected heterogeneous data map embedding tool. And running an embedding tool, such as GRAPHSAGE, DEEPWALK, embedding different types of data into a unified graph structure, and reserving the association relation between the data. This will generate an embedded causal knowledge-graph of the initial data quality.

S54: carrying out real-time monitoring on the Internet of things of the target production workshop to obtain a real-time production state monitoring data set; and carrying out time sequence knowledge graph evolution on the initial data quality causal knowledge graph based on the real-time production state monitoring data set to obtain the dynamic data quality causal knowledge graph.

Specifically, for example, the internet of things monitoring system may be deployed in a target production plant. The system may collect real-time data of various sensors and devices, such as temperature, humidity, pressure, device status, etc. And collecting real-time production state monitoring data in the target production workshop by using the internet of things monitoring system. The data may be real-time sensor readings, device operating status, workflow information, etc. And selecting an algorithm or a method suitable for carrying out time sequence knowledge graph evolution. For example Temporal Graph Convolutional Networks (TGCN): TGCN is a graph-rolling network-based algorithm for processing timing graph data. It can capture the evolution of the relationship between nodes and edges at different time steps. The time sequence knowledge graph evolution algorithm can embed time sequence data into the knowledge graph to reflect time sequence change and evolution process of the data. And inputting the preprocessed real-time production state monitoring data set and the initial data quality causal knowledge graph into a selected time sequence knowledge graph evolution algorithm. And (3) operating an evolution algorithm, and embedding the time sequence change and the evolution process of the real-time data into an initial data quality causal knowledge graph. This will generate a dynamic data quality causal knowledge graph, wherein nodes and edges represent the change and evolution of data and policy elements over time.

According to the method, through fractal dimension analysis, the association relationship between the quality index of the production data of the product and the cause data of the abnormality of the production data can be revealed. Fractal dimension analysis is a method for studying the self-similarity and complexity of data, and can extract hidden modes and features of the data. The fractal correlation data of the data quality anomalies provides quantitative measurement and visual representation of the data quality anomalies, is beneficial to data quality monitoring, anomaly detection and problem investigation of enterprises, and improves the data quality management effect. Through process semantic analysis, key semantic information between production data anomaly causing data and production data quality enhancement strategies can be extracted, and enterprises are helped to understand and explain reasons and improvement strategies of data anomalies. The abnormal incentive strategy semantic data can be used as an important element in the knowledge graph, so that semantic richness and accuracy are provided for knowledge representation and reasoning, and the application effect and decision support capability of the knowledge graph are improved. The different types of data and information are integrated through the initial data quality causal knowledge graph, so that a comprehensive view is provided for understanding and analyzing the quality causal relation of the product production data. It can help enterprises discover the root cause of data quality anomalies, identify potential improvement strategies, and provide guidance for data quality management and decision making. Through heterogeneous data graph embedding and fusion, data and information in the knowledge graph are related to each other, so that a richer and comprehensive knowledge representation is formed. This can improve the awareness and understanding of the enterprise for data quality problems, enhancing data driven decision support and problem solving capabilities. The real-time production state monitoring data set provides real-time knowledge of a target production workshop, helps enterprises to timely master changes and abnormal conditions in the production process, and supports real-time monitoring and decision adjustment. The dynamic data quality causal knowledge graph reflects the time sequence evolution of the data quality causal relation, can help enterprises to find the law and trend of the data quality change, forecast potential data quality problems, and timely take corresponding measures to adjust and improve.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An artificial intelligence-based industrial data quality monitoring and promotion system, which is characterized by comprising the following modules:

2. The industrial data quality monitoring and promotion system according to claim 1, wherein the industrial data acquisition module S1 is specifically:

s11: acquiring workshop equipment space layout data;

3. The industrial data quality monitoring and promotion system according to claim 1, wherein the data quality impact factor mining module S2 specifically comprises:

4. The industrial data quality monitoring and promotion system of claim 3, wherein S25 is specifically:

5. The industrial data quality monitoring and promotion system of claim 3, wherein S26 is specifically:

s266: and calculating a data quality critical value range of the acceptable historical low-quality production data and corresponding equipment performance parameters of the acceptable historical low-quality production data according to the product production data quality influence factor set and the weighted influence factor data, and obtaining a production data acceptable quality interval.

6. The industrial data quality monitoring and promotion system according to claim 1, wherein the data quality evaluation module S3 is specifically:

7. The industrial data quality monitoring and promotion system according to claim 1, wherein the data quality promotion decision module S4 is specifically:

8. The industrial data quality monitoring and promotion system of claim 7, wherein S44 is specifically:

s441: acquiring a historical production data anomaly incentive data set;

9. The industrial data quality monitoring and promotion system of claim 8, wherein S447 is specifically:

10. The industrial data quality monitoring and promotion system based on artificial intelligence according to claim 1, wherein the production process knowledge graph construction module S5 specifically comprises: