Boric anhydride production quality tracing system

Abstract

The invention relates to the technical field of boric anhydride production, in particular to a boric anhydride production quality tracing system. The system comprises a data acquisition and preprocessing module, a data storage and management module, a data analysis and insight module, a quality early warning and feedback module, a tracing query and display module, a system authority management module and the like, wherein the data acquisition and preprocessing module is responsible for collecting and standardizing multi-source production data; the data storage and management module guarantees safe storage and efficient retrieval of data; the data analysis and insight module deeply mines data association through a plurality of professional units, evaluates a quality state and predicts quality risks and traceability abnormities; the quality early warning and feedback module timely notifies and provides improvement suggestions based on the risk level; the traceability query and display module is convenient for a user to intuitively query and export quality traceability information; and the system authority management module strictly controls the data access authority. Through cooperative operation of all the modules, the quality stability of boric anhydride production is effectively improved, the quality risk is reduced, and the requirement of boric anhydride production for high-quality refined production quality tracing is met.

Description

Technical Field

[0001] The present invention relates to the technical field of boric anhydride production, and specifically to a quality traceability system for boric anhydride production. Background Art

[0002] Boric anhydride is an important inorganic chemical raw material, widely used in industries such as glass, ceramics, and metallurgy. The quality of boric anhydride directly affects the performance and quality of downstream products. In glass production, high-quality boric anhydride helps to improve the transparency, heat resistance, and chemical stability of glass; in the metallurgy industry, it can optimize the properties of metal materials. Therefore, ensuring the stable and reliable quality of boric anhydride during the production process is crucial for the normal operation of the entire industrial chain and the improvement of product quality.

[0003] However, the quality control in the traditional boric anhydride production process mainly relies on manual records and sampling inspections, with problems such as imperfect data collection and processing, inaccurate quality analysis, difficult problem traceability, and untimely warnings, making it difficult to meet the increasingly strict quality management requirements. Once a quality problem occurs, it is often difficult to quickly and accurately locate the problem link and the responsible entity, increasing production costs and quality risks. Summary of the Invention

[0004] The present invention provides a quality traceability system for boric anhydride production, which solves the problems existing in the quality control during the boric anhydride production process, such as imperfect data collection and processing, inaccurate quality analysis, difficult problem traceability, and untimely warnings, realizes the full - range traceability and control of the quality of boric anhydride production from raw materials to finished products, improves the quality stability of products, and reduces quality risks. A quality traceability system for boric anhydride production includes: Data collection and pre - processing module: responsible for collecting various data in the whole process of boric anhydride production, and performing pre - processing operations on the collected raw data to convert it into a structured data form that is standardized, complete, and convenient for subsequent analysis. Data storage and management module: communicatively connected to the data collection and pre - processing module, receiving the pre - processed structured data, using a distributed storage architecture, and storing the data into the corresponding database tables according to the preset data classification rules, facilitating the efficient retrieval, query, and long - term management of data, and providing stable data support for the system. Data analysis and insight module: includes a quality correlation analysis unit, a quality status evaluation unit, a quality risk prediction unit, and an abnormal traceability analysis unit. This module obtains the required data from the data storage and management module, analyzes and mines the data information, and realizes the comprehensive analysis and insight of the quality of boric anhydride production. Quality warning and feedback module: communicatively connected to the data analysis and insight module, receiving its analysis results, and performing hierarchical warnings according to its analysis results. Traceability Query and Display Module: Receive the traceability query instructions input from the outside, retrieve relevant data from the Data Storage and Management Module, and combine with the analysis results of the Data Analysis and Insights Module to display the detailed quality traceability information of the whole process of boric anhydride products from raw material input to finished product output through a visual interface.

[0005] Preferably, the Data Analysis and Insights Module includes: Quality Correlation Analysis Unit: Explore the internal correlation relationships among various types of production data, construct a correlation chain covering raw material impurities, equipment operation parameters, production process execution conditions, key intermediate state quality, and final product quality indicators, and provide a basic data correlation basis for the subsequent units; Quality Status Evaluation Unit: Based on the correlation relationships determined by the Quality Correlation Analysis Unit and the preset quality standard range, use evaluation algorithms to accurately evaluate the quality indicators at each production stage by comparing and analyzing real-time production data and historical data, and timely mark the quality abnormal points and corresponding links that appear; Quality Risk Prediction Unit: With the help of a prediction model, according to the laws presented in the current production data and historical data, considering the multi-factor influence revealed by the Quality Correlation Analysis Unit, predict the quality of the upcoming boric anhydride products, and give early warnings of potential quality risks; Abnormal Root Cause Analysis Unit: When the Quality Status Evaluation Unit discovers a quality abnormality or the Quality Risk Prediction Unit indicates a high risk, start to trace back the whole process of production data, analyze the root cause of the abnormality, accurately locate the cause of the quality problem, and generate a detailed traceability report.

[0006] Preferably, the exploration of the internal correlation relationships among various types of production data and the construction of a correlation chain covering raw material impurities, equipment operation parameters, production process execution conditions, key intermediate state quality, and final product quality indicators specifically include: Data Acquisition: Obtain various types of data in the boric anhydride production process from the Data Storage and Management Module, including raw material information, production equipment operation parameters, production process execution parameters, and the quality information of key intermediates and final products; Feature Selection: Calculate the Pearson correlation coefficients of raw material impurities, equipment operation parameters, production process execution conditions, key intermediate state quality, and final product quality indicators, set the correlation threshold, and screen out the data pairs greater than the correlation threshold as the input of the decision tree; Decision Tree Construction: Select the Splitting Criterion: Adopt the information gain criterion, calculate the information gain of each feature, and select the feature with the largest information gain as the splitting attribute of the current node; Recursive Splitting: Start from the root node, divide the data set into different subsets according to the selected splitting attribute, and recursively repeat the above steps for each subset until the stopping condition is met; Association relationship mining: According to the structure of the constructed decision tree and the node division rules, mine the internal association relationships among various types of production data; Result verification and optimization: Use an independent test data set to verify the mined association relationships, evaluate their accuracy and reliability. If the results are not satisfactory, adjust the construction parameters of the decision tree and re - mine until satisfactory results are obtained.

[0007] Preferably, based on the association relationships determined by the quality association analysis unit and the preset quality standard range, use an evaluation algorithm to accurately evaluate the quality indicators at each production stage by comparing and analyzing real - time production data and historical data. Specifically, it includes: Data acquisition: Obtain real - time production data during the boric anhydride production process from the data storage and management module, and at the same time obtain the corresponding historical production data; Feature extraction and weight determination: According to the association relationships determined by the quality association analysis unit, extract features closely related to quality evaluation from real - time production data and historical production data, and determine corresponding weights for each feature according to expert experience; Construct a quality evaluation model: Use an evaluation algorithm optimized by a large amount of historical production data to construct a quality evaluation model; Real - time quality evaluation: Substitute the features extracted from real - time production data into the quality evaluation model to calculate the real - time quality evaluation score. At the same time, calculate the range of quality evaluation scores under normal production conditions according to historical production data; Quality status judgment: Compare the real - time quality evaluation score with the normal score range. If the real - time quality evaluation score is within the normal score range, it is determined that the current production process is in a normal quality state. If the real - time quality evaluation score is not within the normal score range, it is determined that the current production process has a quality abnormality, and mark the abnormal points and corresponding links. Among them, the formula for calculating the quality evaluation score is: ; In the formula, is the number of features, is the weight of the th feature, is the th feature score.

[0008] Preferably, the prediction model includes: Input layer: The input data is the historical data of the boric anhydride production process obtained from the data storage and management module and the current production real - time data; Time - series encoding layer: Use the long short - term memory network (LSTM) to capture the time - dependent relationships in historical data; Dynamic regression layer: A regression equation is constructed based on historical data and time factors through a multiple linear regression model. , where is time the predicted value of the product quality index, is time the input feature value, is time the historical value of the product quality index, and are regression coefficients, is the error term; Fusion layer: The outputs of the time series encoding layer and the dynamic regression layer are fused through a fully connected network to generate the final predicted value. , where is the fully connected network, is time the output of the LSTM hidden layer, is the parameter set of the fully connected layer; Output layer: The fused result is converted into the actual predicted value through a linear transformation.

[0009] Preferably, in the quality early warning and feedback module, hierarchical early warning is performed according to the analysis results of the data analysis and insight module, specifically including: Data acquisition and preparation: Key quality-related data is obtained from the data analysis and insight module, including the actual value of the current quality index output by the quality status assessment unit, the predicted quality index value generated by the quality risk prediction unit, and the standard quality index range corresponding to each production link. At the same time, production process-related data is collected to determine the influence range information of each production link on the subsequent processes and the performance of the final product; Calculate the deviation degree of the quality index: For each key quality index, calculate its deviation degree , where is the actual value of the th quality index or the predicted value generated by the quality risk prediction unit, is the standard value corresponding to the th quality index. By quantifying the deviation degree between the actual value of the quality index and the standard value, the deviation percentage of each quality index relative to its standard value is obtained; Determine the influence range coefficient: According to the influence range of each production link on the subsequent processes and the performance of the final product, an influence range coefficient is set for each link; Calculate the comprehensive risk degree: Calculate the comprehensive risk degree corresponding to each quality index; Risk level classification: According to the calculated comprehensive risk degree, the quality risk is classified into levels, and the specific classification criteria are as follows: Level 1 Risk: If the comprehensive risk degree is greater than or equal to , it indicates that the quality index deviates severely and the affected production processes have a wide range of impacts, which may lead to the product failing to meet the basic quality requirements, seriously affecting subsequent processing and product performance. Immediate emergency measures such as halting production for rectification are required. Meanwhile, the highest-level warning notice is triggered. Level 2 Risk: If the comprehensive risk degree is between and , it indicates that the quality index has a relatively obvious deviation, and the involved production processes have a certain impact on subsequent processes and product performance. Personnel should be arranged as soon as possible to investigate and adjust the corresponding production processes, and the person in charge of the corresponding process should be notified. Level 3 Risk: If the comprehensive risk degree is less than or equal to , it indicates that the deviation degree of the quality index is small, and the impact range of the corresponding production process is limited. At this time, correction is carried out through routine operations, and the operator is reminded to pay attention to the corresponding situation, continuously track subsequent quality data to ensure quality stability.

[0010] The formula for calculating the comprehensive risk degree corresponding to each quality index is: ; In the formula, is the deviation degree of this quality index, is the number of production processes, is the influence range coefficient of the th process, is the correlation indication factor between this quality index and the th process.

[0011] Preferably, according to the influence range of each production process on subsequent processes and the performance of the final product, an influence range coefficient is set for each process, specifically including: Determine influence factors and quantification: Influence of the number of subsequent processes: Count the number of subsequent processes affected by quality abnormalities in each production process; Influence on final product performance indicators: Based on experimental data and historical production records, set an influence degree value for the influence of each performance indicator, and sum up the influence degree values of all affected performance indicators to obtain the influence value of the final product performance indicators; Calculate the weights of each factor: Use the analytic hierarchy process to determine the weights of the influence of the number of subsequent processes and the influence of the final product performance indicators; Calculate the influence range coefficient: According to the determined values of each influence factor and their weights, calculate the influence range coefficient of each production process; The formula for calculating the influence range coefficient of each production process is: ; Where, is the influence range coefficient of the production link, is the impact of this link on subsequent processes, It is the maximum value of the number of impacts on subsequent processes in all production links. Influence on the performance indicators of the final product, and are the weights of the impact of the number of subsequent processes and the impact of the final product performance indicators.

[0012] Preferably, a system permissions management module is also included: Based on the role settings of personnel at different levels (such as operators, managers, quality control personnel, etc.), they are assigned corresponding data access, operation, and traceability query permissions. A combination of role-based and resource-based access control is used, and multiple verification mechanisms (such as password input combined with fingerprint recognition or dynamic verification codes) are set up for sensitive data and key operations to prevent abuse of permissions. When personnel positions change, job responsibilities are adjusted, or system functions are upgraded, permissions can be dynamically adjusted. Detailed logs are recorded for each permission change, facilitating audit and traceability of permission management operations, ensuring the security and compliance of system data.

[0013] Compared with the prior art, the advantages of the present invention are: Comprehensive data collection: Ensure that multi-source data covering all aspects of boric anhydride production is accurately collected and converted into a standardized format suitable for analysis, avoiding data issues that affect subsequent quality analysis and traceability, and improving data availability and reliability.

[0014] Accurate analysis and insights: With the help of various data analysis units, we can deeply explore the correlations behind the data, accurately assess the current quality status, predict quality risks in advance, and quickly trace the root causes of anomalies, providing a comprehensive and scientific basis for production decisions. This helps to adjust production strategies in a timely manner, optimize the production process, and ensure that product quality is stable and meets standards.

[0015] Rapid quality warning and feedback: Based on reasonable risk level classification and multiple forms of warning notification methods, relevant personnel can receive quality problem prompts in a timely manner and take targeted measures quickly based on detailed traceability reports and improvement suggestions to avoid further expansion of quality problems and reduce losses caused by quality risks.

[0016] Convenient quality traceability display: Users can query detailed quality traceability information on the visual interface through simple operations and export reports on demand, which facilitates internal management, reporting and external quality communication, and improves the efficiency and transparency of quality traceability. DETAILED DESCRIPTION

[0017] Example In one embodiment, an implementation manner of a data acquisition and preprocessing module is provided: Data acquisition: Raw material data acquisition: In the raw material procurement link, the supplier information, batch number, purity, impurity content, etc. of the raw materials are recorded by scanning codes or manual entry. During the storage of raw materials, environmental parameters such as the temperature and humidity of the raw materials are collected in real time by sensors.

[0018] Equipment operation data acquisition: The operation parameters such as the temperature, pressure, rotation speed, and current of the production equipment are collected in real time by sensors on the production equipment. For some key equipment, data such as the vibration frequency and noise level of the equipment can also be collected to monitor the operation status of the equipment.

[0019] Process execution data acquisition: The execution parameters of the production process, such as reaction time, feeding speed, reaction temperature curve, etc., are recorded through an automated control system. At the same time, the operation records of the operators are collected, including operation time, operation content, etc.

[0020] Environmental monitoring data acquisition: Environmental monitoring equipment is installed in the production workshop to collect environmental parameters such as the temperature, humidity, and air quality of the workshop in real time.

[0021] Quality inspection data acquisition: At the key nodes of the production process and before the finished products leave the factory, quality inspection equipment is used to detect the quality indicators of the products, such as purity, particle size, crystal form, etc.

[0022] Data preprocessing: Data cleaning: The collected data is cleaned to remove noise data, duplicate data, and error data. For example, for abnormal high or low temperature data collected by temperature sensors, judgment and correction are performed.

[0023] Data format conversion: Convert data in different formats into a unified format for subsequent analysis and processing. For example, convert the binary data collected by sensors into decimal numerical values.

[0024] Data normalization: Perform normalization processing on numerical data so that different data have the same scale.

[0025] Data encoding: Encode non-numerical data and convert it into numerical data. For example, encode the supplier names of raw materials for use in data analysis.

[0026] In another embodiment, an implementation manner of a data storage and management module is provided: Data storage: Distributed storage: The preprocessed data is scattered and stored on multiple storage nodes according to preset rules. For example, the raw material information data is stored on one storage node, and the equipment operation parameter data is stored on another storage node, improving the read and write performance and reliability of the data through distributed storage.

[0027] Data encryption: During the data storage process, a symmetric encryption algorithm is used to encrypt sensitive data.

[0028] Data management: Data classification and indexing: The data is classified and stored according to the type and time dimension of the data, and corresponding indexes are established. For example, an index structure with the raw material batch number as the index is established for the raw material information table, and an index structure with the timestamp as the index is established for the equipment operation parameter table, improving the retrieval efficiency of the data.

[0029] Data backup and recovery: The database is fully backed up and incrementally backed up regularly, and the backup data is stored in a data center in a different location to prevent data loss due to local storage device failures or natural disasters.

[0030] In another embodiment, an implementation manner of a data analysis and insight module is provided: Quality correlation analysis unit: Data preparation: Extract the required production data from the data storage and management module, including raw material data, equipment operation data, process execution data, and product quality data, etc., and perform further feature engineering processing, such as standardizing and normalizing the data to improve the accuracy of data analysis.

[0031] Association relationship mining: Use the decision tree algorithm to analyze the data and mine the internal association relationships between various types of production data. For example, analyze the association relationships between raw material impurity content, equipment operation temperature, and final product purity through the decision tree algorithm to construct an association model.

[0032] Model evaluation and optimization: Use methods such as cross-validation to evaluate the association model, and optimize and adjust the model according to the evaluation results, such as adjusting the parameters of the decision tree, increasing or decreasing the support and confidence thresholds of association rules, etc., to improve the accuracy and stability of the model.

[0033] Quality status evaluation unit: Feature extraction and weight determination: According to the association relationships determined by the quality correlation analysis unit, extract features closely related to quality evaluation from the production data, such as raw material purity, equipment operation stability, etc. Determine the corresponding weights for each feature through expert experience.

[0034] Quality assessment model construction: Use a weighted comprehensive scoring model or a machine learning-based classification model (such as a support vector machine model) to evaluate the quality indicators at each production stage. Input the real-time production data into the quality assessment model to calculate the quality assessment score.

[0035] Quality status judgment: Compare the quality assessment score with the preset quality standard range to judge the quality status of the current production process. If the score is within the normal range, it is determined that the production is in a normal quality state; if the score exceeds the normal range, it is determined that a quality anomaly has occurred, and the anomaly points and corresponding links are marked.

[0036] Quality risk prediction unit: Model selection and construction: Select a long short-term memory network (LSTM) model combined with a dynamic regression model for quality risk prediction. The LSTM model is used to process time series data and capture the long-term dependencies in the production data; the dynamic regression model considers the dynamic change characteristics of the data and combines historical data and time factors for prediction. Take the outputs of the two models as the inputs of the fully connected layer, and fuse the outputs of the two through the fully connected layer to generate the final predicted value.

[0037] Model training and optimization: Use historical production data to train the model, and adopt a backpropagation algorithm (such as the Adam optimizer) to update the model parameters to minimize the loss function between the predicted value and the true value. During the training process, monitor the performance of the model through the validation set to prevent overfitting. Regularly update and optimize the model according to new production data to improve the prediction accuracy and timeliness of the model.

[0038] Risk prediction and early warning: Input the current production data into the trained model to obtain the predicted value of the future product quality. Compare the predicted value with the preset risk threshold to judge whether there is a quality risk. If there is a risk, an early warning message will be sent in a timely manner. The early warning message includes the risk level, the production links that may be affected, and the recommended measures, etc.

[0039] Abnormal origin tracing analysis unit: Abnormal trigger and data backtracking: When the quality status assessment unit discovers a quality anomaly or the quality risk prediction unit indicates a high risk, the abnormal origin tracing analysis unit is activated. Trace back the entire process of production data, and query and analyze the data from various links such as raw material procurement, production equipment operation, and process execution.

[0040] Application of origin tracing analysis method: Adopt the methods of forward reasoning and backward tracing, and combine the association relationships determined by the quality association analysis unit and historical similar abnormal cases to gradually investigate the possible causes of quality anomalies. For example, starting from the quality anomaly of the final product, trace back to the production process link, and further analyze whether there are problems with the equipment operation parameters and raw material quality.

[0041] Traceability report generation: According to the results of traceability analysis, a detailed traceability report is generated. The report content includes the time and location of the anomaly, possible causes, scope of impact, and recommended corrective measures, etc. The traceability report is presented in a visual way to facilitate understanding and decision-making by relevant personnel.

[0042] In another embodiment, an implementation manner of a quality early warning and feedback module is provided: Risk level classification: According to the impact degree of quality risks on subsequent links of boric anhydride production and the scope of impact on the performance of the final product, the quality risks are divided into different levels. For example, a first-level risk indicates that the quality index seriously deviates from the standard range, which may cause the product to be unusable and requires immediate production suspension for rectification; a second-level risk means that the quality index deviates to a certain extent and adjustment measures need to be taken as soon as possible; a third-level risk is that the quality index slightly deviates from the standard range and can be corrected through appropriate monitoring or fine-tuning. The classification criteria of risk levels can be dynamically adjusted according to the actual production situation.

[0043] Early warning notification: For different levels of early warnings, various methods are used to promptly convey the quality early warning information to the corresponding personnel. For first-level risks, a strong alarm signal is issued in the production workshop through an audible and visual alarm, and at the same time, a text message is sent to the mobile phones of relevant management personnel, and an emergency early warning window pops up in the system; for second-level risks, the person in charge of the corresponding link is notified by text message, and a key reminder is made in the system message; for third-level risks, the operator is reminded to pay attention to the corresponding situation through the system message.

[0044] Feedback and improvement: The traceability report generated by the anomaly traceability analysis unit and the corresponding improvement suggestions are fed back to the relevant production departments. The production departments formulate specific rectification measures according to the feedback information, and record the rectification process and results in the system. The system tracks and evaluates the rectification situation to ensure that quality problems are effectively solved.

[0045] In another embodiment, an implementation manner of a traceability query and display module is provided: Query interface design: Develop a user-friendly query interface that supports users to input traceability query instructions in various ways, such as product batch number, production time period, quality index range, etc. The query interface adopts a RESTful-style API design to ensure the simplicity and scalability of the interface.

[0046] Data retrieval and integration: According to the query conditions input by the user, relevant data is retrieved from the data storage and management module and integrated in combination with the analysis results of the data analysis and insight module. During the retrieval process, the index structure of the database is used to improve the query efficiency and ensure that the required data can be obtained within a short time.

[0047] Visual display: The integrated data is presented to the user in the form of a visual interface. The visual interface combines intuitive charts (such as line charts, bar charts, pie charts, etc.), graphs (such as flowcharts, Gantt charts, etc.) and text to present quality traceability information, supports multi-dimensional filtering functions, and facilitates users to quickly locate the required information. At the same time, a data export function is provided, and users can export the query results as a traceability report in PDF format or Excel format.

[0048] In another embodiment, an implementation manner of a system permission management module is provided: Role definition and permission assignment: Different roles are defined according to the job responsibilities and position requirements of different personnel within the enterprise, such as operators, managers, quality supervisors, etc. Corresponding data access, operation, and traceability query permission scopes are assigned to each role. For example, operators can only access and operate production data related to their own work; managers can query and analyze quality data of the entire production process; quality supervisors have higher permissions and can audit and modify data.

[0049] Permission verification and control: When a user logs in to the system, the user's identity is verified. After verification, corresponding permissions are assigned according to the user's role. When the user accesses, operates, and queries data, the system checks the user's permissions in real time, and only users with corresponding permissions can perform corresponding operations. For sensitive data and critical operations, a multi-factor verification mechanism is set up, such as password input combined with fingerprint recognition or dynamic verification codes.

[0050] Permission change management: When there are changes in the positions of personnel within the enterprise, adjustments in job responsibilities, or system function upgrades, etc., the system supports dynamically adjusting the permissions of users. Permission change operations require a strict approval process. After approval, the system records the detailed logs of each permission change, including the change time, the person making the change, the change content, etc., which is convenient for auditing and tracing permission management operations.

[0051] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0052] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A quality traceability system for boric anhydride production, characterized in that, Including: Data acquisition and preprocessing module: Responsible for collecting various types of data throughout the boric anhydride production process, and performing preprocessing operations on the collected raw data to convert it into a structured data form that is standardized, complete, and convenient for subsequent analysis; Data storage and management module: Communicatively connected to the data acquisition and preprocessing module, receives the preprocessed structured data, uses a distributed storage architecture, and stores the data into the corresponding database tables according to the preset data classification rules, facilitating efficient retrieval, query, and long-term management of the data, and providing stable data support for the system; Data analysis and insight module: Includes a quality correlation analysis unit, a quality status assessment unit, a quality risk prediction unit, and an anomaly traceability analysis unit. This module obtains the required data from the data storage and management module, analyzes and mines the data information, and realizes a comprehensive analysis and insight into the quality of boric anhydride production; Quality early warning and feedback module: Communicatively connected to the data analysis and insight module, receives its analysis results, and issues graded early warnings according to its analysis results; Traceability query and display module: Receives the externally input traceability query instruction, retrieves relevant data from the data storage and management module, and combines the analysis results of the data analysis and insight module to display the detailed quality traceability information of the entire process from raw material input to finished product output of the boric anhydride product through a visual interface.

2. The boron anhydride production quality traceability system according to claim 1, wherein The data analysis and insight module includes: Quality correlation analysis unit: Mines the internal correlation relationships between various types of production data, constructs a correlation chain covering raw material impurities, equipment operation parameters, production process execution conditions, key intermediate state quality, and final product quality indicators, and provides a basic data correlation basis for the subsequent units; Quality status assessment unit: Based on the correlation relationships determined by the quality correlation analysis unit and the preset quality standard range, uses an assessment algorithm to accurately assess the quality indicators at each production stage by comparing and analyzing real-time production data and historical data, and timely marks the quality anomaly points and corresponding links that appear; Quality risk prediction unit: With the help of a prediction model, based on the laws presented in the current production data and historical data, considering the multi-factor influence revealed by the quality correlation analysis unit, predicts the quality of the upcoming boric anhydride product, and early warns of potential quality risks; Anomaly traceability analysis unit: When the quality status assessment unit discovers a quality anomaly or the quality risk prediction unit indicates a high risk, it starts to trace back the entire process of production data, analyzes the root cause of the anomaly, accurately locates the cause of the quality problem, and generates a detailed traceability report.

3. The quality traceability system for boric anhydride production according to claim 2, characterized in that, The mining of the internal correlation relationships between various types of production data and the construction of a correlation chain covering raw material impurities, equipment operation parameters, production process execution conditions, key intermediate state quality, and final product quality indicators specifically include: Data acquisition: Obtains various types of data during the boric anhydride production process from the data storage and management module, including raw material information, production equipment operation parameters, production process execution parameters, and the quality information of key intermediate states and final products; Feature selection: Calculate the Pearson correlation coefficients of raw material impurities, equipment operation parameters, production process execution, key intermediate state quality, and final product quality indicators, set a correlation threshold, and filter out data pairs with a correlation greater than the threshold as the input to the decision tree; Decision tree construction: Select a splitting criterion: Adopt the information gain criterion, calculate the information gain of each feature, and select the feature with the largest information gain as the splitting attribute of the current node; Recursive splitting: Starting from the root node, divide the data set into different subsets according to the selected splitting attribute, and recursively repeat the above steps for each subset until the stopping condition is met; Association relationship mining: According to the structure of the constructed decision tree and the node splitting rules, mine the internal association relationships between various types of production data; Result verification and optimization: Use an independent test data set to verify the mined association relationships, evaluate their accuracy and reliability. If the results are not satisfactory, adjust the construction parameters of the decision tree and re-mine until satisfactory results are obtained.

4. The quality traceability system for boric anhydride production according to claim 2, wherein Based on the association relationships determined by the quality association analysis unit and the preset quality standard range, use an evaluation algorithm to accurately evaluate the quality indicators at each production stage by comparing and analyzing real-time production data and historical data. Specifically, it includes: Data acquisition: Obtain real-time production data during the production process of boric anhydride from the data storage and management module, and at the same time obtain the corresponding historical production data; Feature extraction and weight determination: According to the association relationships determined by the quality association analysis unit, extract features closely related to quality evaluation from real-time production data and historical production data, and determine corresponding weights for each feature according to expert experience; Construct a quality evaluation model: Use an evaluation algorithm optimized by a large amount of historical production data to construct a quality evaluation model; Real-time quality evaluation: Substitute the features extracted from the real-time production data into the quality evaluation model to calculate the real-time quality evaluation score. At the same time, calculate the quality evaluation score range under normal production conditions based on the historical production data; Quality status judgment: Compare the real-time quality evaluation score with the normal score range. If the real-time quality evaluation score is within the normal score range, it is determined that the current production process is in a normal quality state. If the real-time quality evaluation score is not within the normal score range, it is determined that the current production process has a quality abnormality, and mark the abnormal points and corresponding links. Among them, the formula for calculating the quality evaluation score is: ; In the formula, is the number of features, is the weight of the -th feature, is the score of the -th feature.

5. The boron anhydride production quality traceability system according to claim 2, wherein The prediction model includes: Input layer: The input data is the historical data of the boric anhydride production process obtained from the data storage and management module and the current production real-time data; Time series encoding layer: Use the long short-term memory network (LSTM) to capture the time-dependent relationships in the historical data; Dynamic regression layer: A regression equation is constructed through a multiple linear regression model based on historical data and time factors , where is the time predicted value of the product quality index, is the input feature value at time , is the historical value of the product quality index at time , and are regression coefficients, is the error term; Fusion layer: The outputs of the time series encoding layer and the dynamic regression layer are fused through a fully connected network to generate the final predicted value , where is the fully connected network, is the output of the LSTM hidden layer of time , and is the parameter set of the fully connected layer; Output layer: Convert the fused result into an actual predicted value through a linear transformation.

6. The boron anhydride production quality traceability system according to claim 1, characterized in that, Based on its analysis results, conduct hierarchical early warnings, specifically including: Data acquisition and preparation: Obtain key quality-related data from the data analysis and insight module, including the actual current quality indicator values output by the quality status assessment unit, the predicted quality indicator values generated by the quality risk prediction unit, and the standard quality indicator ranges corresponding to each production link. At the same time, collect production process-related data to determine the impact range information of each production link on subsequent processes and the performance of the final product; Calculate the deviation degree of quality indicators: For each key quality indicator, calculate its deviation degree , where is the actual value of the th quality indicator or the predicted value generated by the quality risk prediction unit, is the standard value corresponding to the th quality indicator. By quantifying the deviation degree between the actual value and the standard value of the quality indicator, the deviation percentage of each quality indicator relative to its standard value is obtained; Determine the impact range coefficient: Set an impact range coefficient for each link according to the impact range of each production link on subsequent processes and the performance of the final product; Calculate the comprehensive risk degree: Calculate the comprehensive risk degree corresponding to each quality indicator; Risk level classification: Classify the quality risks according to the calculated comprehensive risk degree. The specific classification criteria are as follows: Level 1 risk: If the comprehensive risk degree is greater than or equal to 80%, it indicates that the quality indicator deviates severely and the impact range of the production links involved is wide, which may cause the product to fail to meet the basic quality requirements, seriously affecting subsequent processing and the performance of the product in use. Immediate emergency measures such as stopping production need to be taken for rectification, and at the same time, trigger the highest-level warning notice; Level 2 risk: If the comprehensive risk degree is between 30% and 80%, it indicates that the quality indicator has a relatively obvious deviation, and the production links involved have a certain impact on subsequent processes and product performance. Personnel should be arranged as soon as possible to check and adjust the corresponding production links, and notify the person in charge of the corresponding link; Level 3 risk: If the comprehensive risk degree is less than or equal to 30%, it indicates that the deviation degree of the quality indicator is small, and the impact range of the corresponding production link is limited. At this time, correction is carried out through routine operations, and the operator is reminded to pay attention to the corresponding situation, continuously track the subsequent quality data to ensure quality stability.

7. The formula for calculating the comprehensive risk degree corresponding to each quality indicator is: ； In the formula, is the deviation degree of the quality index, is the number of production links, is the influence range coefficient of the th link, and is the correlation indication factor between the quality index and the th link.

8. A boron anhydride production quality traceability system according to claim 6, characterized in that, The impact range coefficient is set for each link according to the impact range of each production link on subsequent processes and the performance of the final product, specifically including: Determine the influencing factors and quantification: Impact on the number of subsequent processes: Count the number of impacts of quality abnormalities in each production link on subsequent processes; Impact on the performance indicators of the final product: Through experimental data and historical production records, set an impact degree value for the impact of each performance indicator, and sum the impact degree values of all affected performance indicators to obtain the impact value of the performance indicators of the final product; Calculate the weights of each factor: Use the analytic hierarchy process to determine the weights of the above-mentioned impact on the number of subsequent processes and the impact on the performance indicators of the final product; Calculate the impact range coefficient: Calculate the impact range coefficient of each production link according to the determined values of each influencing factor and their weights; The formula for calculating the impact range coefficient of each production link is: ; In the formula, is the influence range coefficient of the production process, is the influence quantity of this process on the subsequent processes, is the maximum value of the influence quantity on the subsequent processes in all production processes, is the influence on the performance index of the final product, and are the weights of the influence of the quantity of subsequent processes and the influence of the performance index of the final product respectively.

Images