Food additive production line real-time monitoring method based on industrial internet

Abstract

The application discloses a food additive production line real-time monitoring method based on an industrial internet, belongs to the field of intelligent manufacturing of food industry, and comprises a monitoring system, wherein the monitoring system comprises a data acquisition and uploading module, an edge processing node module, a cloud analysis and processing module, an abnormality monitoring and early warning module and a data archiving and management module; the data acquisition and uploading module is used for acquiring original process data of each process node of the food additive production line, encapsulating and formatting the original process data to form regular process data, and acquiring a production process of the food additive.The application solves problems such as disordered data of the food additive production line, misjudgment, abnormal response lag and difficulty in traceability, can identify production abnormalities, optimize control parameters, reduce production downtime probability and quality loss, realizes production full-process traceability, improves production line control efficiency, product quality stability and equipment operation safety, and meets food safety supervision compliance requirements.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing in the food industry. Specifically, it is a method for real-time monitoring of food additive production lines based on the Industrial Internet. Background Technology

[0002] Food additives are essential raw materials for the food industry, and their production process requires strict control to comply with industry regulatory standards. Currently, most food additive production lines employ traditional monitoring methods, which suffer from severe data silos, insufficient monitoring accuracy, and delayed anomaly response. Data from individual devices is stored independently and cannot be interconnected. Manual inspections and offline testing struggle to achieve real-time control of key parameters, easily leading to product quality fluctuations, increased non-compliance rates, and failing to meet the regulatory requirements for full traceability. Existing automated control systems are mostly limited to single devices or single stages, lacking global collaboration and intelligent analysis capabilities, making them ill-suited to the industry's digital transformation needs. Therefore, there is an urgent need for a real-time monitoring system integrating industrial internet technology to address the pain points of traditional monitoring and achieve transparent and intelligent control of the entire production process. This has become a pressing requirement for the transformation and upgrading of the food additive industry.

[0003] A Chinese patent (application number CN202511906948.8) discloses a real-time monitoring and anomaly handling system and method for a food additive production line. This system collects data from each stage of the production line, processes the data, and uses an anomaly detection module. It employs multiple algorithms to extract anomaly features, establishes and optimizes historical data models, enabling real-time and accurate analysis and detection of various parameter anomalies at different stages. The system performs logical operations on the detection result matrix, matches preset signals, and converts and sends them. Corresponding signals are generated for different anomalies, and targeted measures are taken based on the number and cycle of signals. A convolutional neural network is constructed to calculate and optimize the historical data model, providing comprehensive data support for production monitoring and anomaly handling, which is beneficial for process optimization and quality improvement.

[0004] However, the above solutions suffer from fixed anomaly detection and signal matching processes, resulting in low response efficiency and an inability to quickly address on-site anomalies, which can easily lead to production downtime and quality losses. Furthermore, the historical data model optimization lacks dynamic adaptation, and the application of convolutional neural networks is limited to modeling, failing to achieve parameter adjustment and optimization. This leads to low management efficiency, insufficient product quality stability, and the absence of data traceability mechanisms, making it impossible to achieve full-process traceability of production. Consequently, it is difficult to meet food safety regulatory compliance requirements, and the safety of equipment operation cannot be effectively guaranteed. Summary of the Invention

[0006] The present invention aims to provide a method for real-time monitoring of food additive production lines based on the Industrial Internet, in order to solve the problems mentioned in the background art.

[0007] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows: The real-time monitoring method for food additive production lines based on the Industrial Internet includes a data acquisition and uploading module, an edge processing node module, a cloud analysis and processing module, an anomaly monitoring and early warning module, and a data archiving and management module. The data acquisition and upload module is used to acquire the original process data of each process node in the food additive production line, encapsulate and format it into regular process data, acquire the production process logic and historical time sequence data of food additives, and transmit the above data to the edge processing node module. The cloud-based analysis and processing module communicates bidirectionally with the edge processing node module. Based on the preliminary judgment results, the regularized process data of all process nodes, the production process logic, and historical time series data, it performs analysis to obtain global analysis results and pushes them to the anomaly monitoring and early warning module. Based on the global analysis results, it issues optimization instructions to the edge processing node module to update the preliminary judgment rules. The anomaly monitoring and early warning module generates tiered early warning signals based on global analysis results and pushes them to the control terminal; The data archiving and management module is used to encrypt and store data from the data acquisition and upload module, the edge processing node module, and the cloud analysis and processing module, and provides data retrieval and traceability query interfaces to the control terminal.

[0008] The above solution, through the collaboration of various modules, reduces data misjudgment and delays in handling anomalies, achieves production data traceability, and comprehensively improves production line management efficiency and product quality stability. Specifically, given the continuous and real-time requirements of food additive production, the edge processing node module enables local preliminary assessment, eliminating the need to upload all data to the cloud. This reduces network transmission pressure and processing latency, avoiding missed anomalies and delayed handling due to cloud lag. For example, if the temperature at a reaction node suddenly rises, the edge node can quickly and initially determine the anomaly, buying time for on-site handling and preventing material deterioration and equipment damage. The bidirectional communication design of the cloud analysis and processing module enables both global operational condition analysis and dynamic optimization of edge node assessment rules, adapting to actual operational changes such as production line batch switching and raw material fluctuations, solving the problem of fixed and unadaptable traditional monitoring rules.

[0009] Optionally, the data acquisition and uploading module includes: The raw data acquisition and encapsulation module is used to acquire raw process data and encapsulate and format it into regular process data. The raw process data includes material process parameters, equipment operating parameters, and process execution parameters. The process baseline acquisition module is used to acquire the preset production process logic, which includes the process parameter thresholds, process execution sequence and process coupling constraints for each process node. The historical time series data acquisition module is used to acquire historical normal time series data and historical abnormal time series data. The historical normal time series data is the process parameter time series data under qualified operating conditions of the production line, and the historical abnormal time series data is the process parameter time series data when the production line experiences abnormal parameters, abnormal processes, abnormal equipment, or abnormal quality conditions. The data upload module is used to upload standardized process data, production process logic, historical normal time series data, and historical abnormal time series data to the edge processing node module.

[0010] Optionally, the edge processing node module includes: The time-series feature extraction module extracts the time-series change trends, fluctuation amplitudes, and steady-state durations of parameters for corresponding food additive production process nodes based on the regularized process data, forming time-series features for a single node. The local operating condition analysis module matches the time-series characteristics of a single node with the process parameter thresholds and process coupling constraints in the production process logic. At the same time, it compares historical normal time-series data with historical abnormal time-series data. Based on the matching and comparison results, it makes a preliminary assessment of the operating condition of a single food additive production process node, generates a preliminary assessment result, and synchronously uploads the preliminary assessment result and the regularized process data of a single node to the cloud analysis and processing module. The judgment rule update module receives the judgment optimization parameters sent by the cloud analysis and processing module, and updates the judgment rules of the local working condition judgment module according to the judgment optimization parameters.

[0011] Optionally, the local operating condition assessment module includes: The threshold comparison submodule compares the material process parameters, equipment operating parameters, and process execution parameters in the time-series characteristics of a single node with the process parameter thresholds of the corresponding process node one by one. If any parameter exceeds the process parameter threshold range, the process node is marked as a suspected anomaly and the corresponding abnormal parameters are recorded and transmitted to the trend matching submodule. If all parameters are within the process parameter threshold range, the process node is determined to be in normal operating condition. The trend matching submodule constructs a parameter trend comparison benchmark based on historical normal time series data. It compares and analyzes the time series characteristics of individual nodes corresponding to suspected abnormal process nodes with the parameter trend comparison benchmark, eliminates false suspected anomalies caused by instantaneous material fluctuations and data acquisition errors, confirms the remaining suspected anomalies as real anomalies, marks the anomaly type and records the parameters, and then transmits them to the process association verification submodule. The process correlation verification submodule verifies actual anomalies based on the process execution sequence, process coupling constraints, and historical anomaly time series data. It verifies the current anomaly characteristics by combining historical anomaly time series data to eliminate duplicate misjudgments. At the same time, based on the process execution sequence and process coupling constraints, it determines whether the anomaly is a normal process fluctuation in the process connection of the food additive production line. If it is a normal process fluctuation, the process node is determined to be in normal operating condition. If it is still an anomaly after excluding normal process fluctuations, a preliminary judgment result containing operating condition judgment, abnormal process node, and abnormal parameter type is generated. The preliminary judgment result is uploaded to the cloud analysis and processing module, and the judgment process data is pushed to the data archiving management module.

[0012] In the above solution, the three-level judgment logic ensures the accuracy of anomaly judgment, reduces the interference of false anomalies on production, and accurately identifies real anomalies, providing data support for cloud-based global traceability and optimization. Simultaneously, it reduces the workload of on-site personnel and minimizes quality defects and production losses caused by misjudgments or omissions. Specifically, the threshold comparison submodule performs initial screening, marking nodes exceeding the threshold as suspected anomalies. Then, the trend matching submodule constructs a comparison benchmark based on historical normal time-series data, eliminating false anomalies caused by instantaneous material fluctuations and collection errors, such as instantaneous material ratio deviations in the batching process, which are not real anomalies and can be quickly eliminated through trend comparison, avoiding unnecessary downtime and improving production efficiency. The process correlation verification submodule, combining the process execution sequence and process coupling constraints, can effectively distinguish between normal process fluctuations and real anomalies. For example, slight temperature fluctuations in preceding reaction nodes may cause parameters in subsequent purification nodes to temporarily deviate from thresholds. This is a normal fluctuation during process transitions and can be identified as a normal operating condition through correlation verification, avoiding misjudgment. If the anomaly persists after excluding transition fluctuations, the anomaly type and parameters are accurately labeled, providing clear guidance for on-site personnel and reducing investigation time. Simultaneously, combining historical anomaly time-series data can eliminate duplicate misjudgments. For instance, if a node previously experienced a similar anomaly due to sensor failure, comparing historical data can quickly pinpoint the cause, avoiding redundant investigations. Furthermore, the data from the analysis process is pushed to the data archiving management module, enabling full-process traceability of anomaly analysis. When similar anomalies occur subsequently, the handling plan can be optimized by referring to historical analysis processes, gradually improving the accuracy of local analysis.

[0013] Optionally, the cloud-based analysis and processing module includes: The global anomaly tracing module collects and sorts the preliminary judgment results and regularized process data of all process nodes in a unified manner, extracts the abnormal process nodes and forms a list of nodes to be traced, and identifies the process node where the anomaly first occurred by combining the production process logic and the sequence of the process. It tracks the transmission and diffusion process and the scope of influence of the abnormal parameters along the process flow, and eliminates instantaneous random fluctuations by globally matching with historical time series data, thus forming a global anomaly tracing result. The multi-node collaborative verification module summarizes the preliminary judgment results of all process nodes to establish a global operating condition judgment dataset. Based on the production process logic, it performs consistency verification on the judgment results of adjacent processes, compares the global operating condition distribution with historical normal time series data, identifies and eliminates single point suspicious misjudgments, confirms systemic real anomalies, and forms global operating condition data. The quality stability analysis module binds and collects global operating condition data and standardized process data according to production batches, establishes a global correlation mapping between finished product quality indicators and parameters of each process, statistically analyzes parameter fluctuations and consistency deviations between batches, locates key process nodes that affect quality stability, completes quality stability classification assessment, and outputs quality stability analysis results. The multi-node parameter linkage module performs multi-process node parameter linkage analysis based on global operating condition data, historical normal time series data, and quality stability analysis results, and generates parameter linkage analysis results.

[0014] Optionally, the multi-node parameter linkage module includes: The full-process linkage model construction module, based on the production process logic, combined with global operating condition data, historical normal time sequence data and quality stability analysis results, determines the linkage relationship between parameters of each process in the food additive production line, extracts the core process parameters of each process, and constructs a full-process parameter linkage model through industrial Internet data association algorithms. The parameter quality weighting module, based on the results of quality stability analysis, statistically analyzes the correlation between parameters of each process and quality indicators, calculates the quality deviation corresponding to parameter fluctuations, and determines the priority of parameter optimization through weight quantification and assignment. The global parameter adjustment module determines the basic adjustment range of individual parameters and the coordinated adjustment range of parameters for each process based on historical normal time-series data and production process logic. The analysis and optimization module is used to generate analysis and optimization parameters and corresponding working condition optimization instructions that are adapted to the global working conditions based on the full-process parameter linkage model, parameter optimization priority, basic adjustment range of individual parameters and coordinated adjustment range of parameters in each process. The parameter command sending module is used to receive the analysis and optimization parameters and operating condition optimization commands output by the analysis and optimization module, send the analysis and optimization parameters to the edge processing node module, and send the operating condition optimization commands to the field equipment of the food additive production line.

[0015] Optionally, the analysis and optimization module includes: The parameter priority matching submodule is used to combine parameter optimization priority with the full-process parameter linkage model, and according to the production process logic of the food additive production line, prioritize various parameters, classify them into core control parameters and auxiliary control parameters, determine the order of parameter control, and output the parameter control ranking results. The control parameter calculation submodule is used to receive the parameter control sorting results, combine the full process parameter linkage model, the basic adjustment range of a single parameter and the coordinated adjustment range of parameters in each process, refer to historical normal time series data and production process requirements, and calculate the control parameter combination and corresponding equipment control parameter value that are suitable for the current production conditions through the industrial Internet data association algorithm. The working condition optimization instruction generation submodule is used to receive equipment control parameter values, combine them with the linkage relationship between parameters of each process determined by the full process parameter linkage model, and convert them into working condition optimization instructions that can be recognized by the equipment on the production line. The parameter instruction verification submodule is used to verify the compliance of control parameter combinations and operating condition optimization instructions. It checks whether the parameter adjustment range conforms to the basic adjustment range of a single parameter and the coordinated adjustment range of parameters in each process. It also verifies whether the instructions match the equipment operation logic. At the same time, it verifies the consistency between parameters in conjunction with the full process parameter linkage model. When the verification is abnormal, it generates a parameter correction prompt containing abnormal parameter identifier, out-of-range value, and allowable adjustment range, and feeds it back to the control parameter calculation submodule for recalculation and adjustment. After the verification is qualified, the control parameter combination and operating condition optimization instructions are transmitted to the global operating condition adaptation submodule. The global operating condition adaptation submodule is used to combine the global operating condition data output by the multi-node collaborative verification module, refer to the linkage relationship between parameters of each process determined by the full process parameter linkage model and the adjustment range defined by the global parameter adjustment module, and quantitatively adapt and adjust the qualified control parameter combination and operating condition optimization instructions respectively, and output the judgment optimization parameters and operating condition optimization instructions adapted to the global operating condition.

[0016] This solution prioritizes parameters, calculates parameter combinations suitable for the current operating conditions, converts parameters into equipment-recognizable commands and verifies compliance, dynamically adapts to changes in operating conditions, and pushes optimized parameters to edge nodes for optimization and judgment rules. This reduces the anomaly rate and improves production line quality stability, equipment safety, and control efficiency. Specifically, the parameter priority matching submodule, combined with quality stability analysis results, distinguishes between core and auxiliary control parameters, determines the control sequence, avoids blind control, and prioritizes ensuring that core parameters affecting product quality meet standards, thereby improving control efficiency. The control parameter calculation submodule, combining the full-process linkage model and historical data, calculates parameter combinations suitable for the current operating conditions, preventing single parameter adjustments from causing anomalies in other processes, improving product quality consistency, and resolving batch-to-batch quality deviations. The operating condition optimization command generation submodule converts parameters into equipment-recognizable commands. The parameter command verification submodule avoids parameters exceeding range and command conflicts, preventing equipment damage and material deterioration. The global operating condition adaptation submodule, combined with cloud data, dynamically adapts to changes in operating conditions such as batch switching and raw material fluctuations. The analysis and optimization parameters are pushed to the edge nodes, and the local analysis rules can be continuously optimized to reduce the anomaly rate. Overall, the parameters are accurately and compliantly controlled, and the quality stability and equipment safety of the production line are improved.

[0017] In summary, this invention effectively reduces data misjudgment and delays in handling anomalies, improving the real-time performance and accuracy of anomaly responses. Through dynamic optimization and holistic design, it can adapt to changes in production conditions, achieving precise parameter control and thus improving production line management efficiency and product quality stability. Simultaneously, production data is traceable, ensuring industry compliance requirements, reducing regulatory penalties, enhancing equipment operational safety, minimizing production downtime and quality losses, and ultimately achieving efficient, stable, and compliant production line operation. Attached Figure Description

[0018] The present invention can be further illustrated by the non-limiting embodiments given in the accompanying drawings; Figure 1 A structural framework diagram of a real-time monitoring method for food additive production lines based on the Industrial Internet; Figure 2 This is a structural framework diagram of the data acquisition and uploading module in this invention; Figure 3 This is a structural framework diagram of the edge processing node module in this invention; Figure 4 This is a structural framework diagram of the cloud-based analysis and processing module in this invention; Figure 5 This is a flowchart of the local working condition assessment module of the present invention. Figure 6 This is a flowchart of the analysis and optimization module of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0021] like Figures 1 to 6 As shown, the real-time monitoring method for food additive production lines based on the Industrial Internet includes a data acquisition and uploading module, an edge processing node module, a cloud analysis and processing module, an anomaly monitoring and early warning module, and a data archiving and management module. The data acquisition and upload module is used to acquire the original process data of each process node in the food additive production line, encapsulate and format it into regular process data, acquire the production process logic and historical time sequence data of food additives, and transmit the above data to the edge processing node module. The cloud-based analysis and processing module communicates bidirectionally with the edge processing node module. Based on the preliminary judgment results, the regularized process data of all process nodes, the production process logic, and historical time series data, it performs analysis to obtain global analysis results and pushes them to the anomaly monitoring and early warning module. Based on the global analysis results, it issues optimization instructions to the edge processing node module to update the preliminary judgment rules. The anomaly monitoring and early warning module generates tiered early warning signals based on global analysis results and pushes them to the control terminal; The data archiving and management module is used to encrypt and store data from the data acquisition and upload module, the edge processing node module, and the cloud analysis and processing module, and provides data retrieval and traceability query interfaces to the control terminal.

[0022] The data acquisition and upload module includes: The raw data acquisition and encapsulation module is used to acquire raw process data and encapsulate and format it into regular process data. The raw process data includes material process parameters, equipment operating parameters, and process execution parameters. The process baseline acquisition module is used to acquire the preset production process logic, which includes the process parameter thresholds, process execution sequence and process coupling constraints for each process node. The historical time series data acquisition module is used to acquire historical normal time series data and historical abnormal time series data. The historical normal time series data is the process parameter time series data under qualified operating conditions of the production line, and the historical abnormal time series data is the process parameter time series data when the production line experiences abnormal parameters, abnormal processes, abnormal equipment, or abnormal quality conditions. The data upload module is used to upload standardized process data, production process logic, historical normal time series data, and historical abnormal time series data to the edge processing node module.

[0023] The edge processing node module includes: The time-series feature extraction module extracts the time-series change trends, fluctuation amplitudes, and steady-state durations of parameters for corresponding food additive production process nodes based on the regularized process data, forming time-series features for a single node. The local operating condition analysis module matches the time-series characteristics of a single node with the process parameter thresholds and process coupling constraints in the production process logic. At the same time, it compares historical normal time-series data with historical abnormal time-series data. Based on the matching and comparison results, it makes a preliminary assessment of the operating condition of a single food additive production process node, generates a preliminary assessment result, and synchronously uploads the preliminary assessment result and the regularized process data of a single node to the cloud analysis and processing module. The judgment rule update module receives the judgment optimization parameters sent by the cloud analysis and processing module, and updates the judgment rules of the local working condition judgment module according to the judgment optimization parameters.

[0024] The local operating condition analysis module includes: The threshold comparison submodule compares the material process parameters, equipment operating parameters, and process execution parameters in the time-series characteristics of a single node with the process parameter thresholds of the corresponding process node one by one. If any parameter exceeds the process parameter threshold range, the process node is marked as a suspected anomaly and the corresponding abnormal parameters are recorded and transmitted to the trend matching submodule. If all parameters are within the process parameter threshold range, the process node is determined to be in normal operating condition. The trend matching submodule constructs a parameter trend comparison benchmark based on historical normal time series data. It compares and analyzes the time series characteristics of individual nodes corresponding to suspected abnormal process nodes with the parameter trend comparison benchmark, eliminates false suspected anomalies caused by instantaneous material fluctuations and data acquisition errors, confirms the remaining suspected anomalies as real anomalies, marks the anomaly type and records the parameters, and then transmits them to the process association verification submodule. The process correlation verification submodule verifies actual anomalies based on the process execution sequence, process coupling constraints, and historical anomaly time series data. It verifies the current anomaly characteristics by combining historical anomaly time series data to eliminate duplicate misjudgments. At the same time, based on the process execution sequence and process coupling constraints, it determines whether the anomaly is a normal process fluctuation in the process connection of the food additive production line. If it is a normal process fluctuation, the process node is determined to be in normal operating condition. If it is still an anomaly after excluding normal process fluctuations, a preliminary judgment result containing operating condition judgment, abnormal process node, and abnormal parameter type is generated. The preliminary judgment result is uploaded to the cloud analysis and processing module, and the judgment process data is pushed to the data archiving management module.

[0025] Implementation details for selecting citric acid in food additive production: like Figure 2As shown, the specific implementation of the data acquisition and upload module is as follows: The raw data acquisition and encapsulation module collects the raw process data of the core crystallization process of the citric acid production line, specifically including: material process parameters (crystallization temperature, pH value of the liquid), equipment operating parameters (stirring motor speed, material conveying pump power), and process execution parameters (crystallization time, feed rate); the acquisition frequency is 1 time / second, forming regular process data, and the encapsulation format is: timestamp + parameter name + parameter value + acquisition equipment number.

[0026] The process baseline acquisition module pre-sets the citric acid production process logic, which includes: the process parameter thresholds for each process node, the process execution sequence (feeding, heating, crystallization, and discharging), and process coupling constraints (crystallization temperature is linked to pH value, and pH value needs to be finely adjusted synchronously when the temperature increases to avoid abnormal crystal morphology).

[0027] The historical time-series data acquisition module stores two types of historical time-series data, both taken from the past 12 months of citric acid production records, including both qualified and abnormal production records. Historical normal time-series data: time series under qualified operating conditions, with crystallization temperature of 43~47℃, pH value of 1.9~2.1, and stirring speed of 300~350r / min, accumulating over 1000 batches of data. Historical abnormal time-series data: time series under abnormal operating conditions such as crystallization temperature exceeding 50℃ (leading to a decrease in crystallization rate), pH value <1.8 (leading to excessive acidity), and abnormal stirring speed shutdown (leading to uneven crystallization), accumulating 86 batches of abnormal data.

[0028] The data upload module uploads the encapsulated and standardized process data to the edge processing node module in real time, and simultaneously uploads the preset production process logic, historical normal time series data and historical abnormal time series data to ensure the real-time nature of local analysis.

[0029] like Figure 3 As shown, the specific implementation of the edge processing node module is as follows: The temporal feature extraction module, based on the uploaded regularized process data, extracts three types of temporal features for the citric acid crystallization process (a single process node): Parameter temporal change trend: such as the rise and fall of crystallization temperature every 10 seconds, and the continuous fluctuation trend of pH value; Fluctuation amplitude: such as the maximum fluctuation value of temperature per minute (normally ≤0.5℃), and the fluctuation range of pH value (normally ≤0.1); Steady-state duration: such as the continuous duration of temperature maintained at 43~47℃, and the continuous duration of stirring speed stable at 300~350 r / min. After extraction, a single node temporal feature dataset is formed.

[0030] like Figure 5As shown, the specific implementation of the local operating condition judgment module includes a threshold comparison submodule. This submodule compares the material process parameters, equipment operating parameters, and process execution parameters in the time-series characteristics with the parameter thresholds in the process baseline one by one. It sets corresponding judgment criteria based on the production requirements of different grades of citric acid. Two examples are provided here: Standard criteria for judging food-grade citric acid: Temperature exceeding 43-47℃, pH value exceeding 1.9-2.1, or stirring speed exceeding 300-350 r / min. If any parameter exceeds the standard, it is marked as a suspected anomaly. The name of the abnormal parameter, the exceeding value, and the duration are recorded and transmitted to the trend matching submodule. If all parameters meet the standards, it is judged as normal operating condition. Standard criteria for judging high-purity citric acid: Temperature exceeding 44-46℃, pH value exceeding 1.95-2.05, or stirring speed exceeding 310-340 r / min. If any parameter exceeds the standard, it is marked as a suspected anomaly. If all parameters meet the standards, it is judged as normal operating condition.

[0031] The trend matching submodule constructs a parameter trend comparison benchmark based on historical normal time series data (such as the normal rise and fall trend of temperature over 10 consecutive seconds, and the stable fluctuation benchmark of pH value). It uses the DTW dynamic time warping algorithm to compare and analyze the time series characteristics of suspected abnormal process nodes with the benchmark, thereby completing the elimination of false anomalies. The process correlation verification submodule verifies the actual anomalies confirmed by the trend matching submodule based on the execution sequence of the citric acid crystallization process (feeding, heating, crystallization, and discharging), process coupling constraints (temperature and pH linkage), and historical anomaly time series data. It combines historical anomaly data to verify current anomaly characteristics (e.g., temperature exceeding 47℃ for ≥10 seconds) and eliminates duplicate false alarms (e.g., repeated alarms for the same parameter in the same batch). Based on process coupling constraints, it determines whether the anomaly is a normal fluctuation in the process connection (e.g., after feeding, the temperature briefly rises to 48℃ for <5 seconds, and the pH value rises to 2.15 simultaneously, which is considered a normal fluctuation). If the anomaly persists after excluding normal fluctuations, a preliminary judgment result is generated (including operating condition judgment: anomaly; anomaly process node: crystallization process; anomaly parameter type: temperature exceeding the limit). This preliminary judgment result is uploaded to the cloud analysis and processing module, and the judgment process data is pushed to the data archiving management module. The judgment rule update module receives judgment optimization parameters issued by the cloud analysis and processing module based on the global analysis results, and updates the judgment rules of the local working condition judgment module according to the judgment optimization parameters (such as adjusting the temperature threshold of conventional citric acid from 43~47℃ to 42~48℃).

[0032] like Figure 4As shown, the cloud-based analysis and processing module achieves bidirectional communication with the edge processing node module. On the one hand, it receives the preliminary judgment results uploaded by the edge nodes and the regularized process data of individual nodes in real time. On the other hand, based on the received data, production process logic, and historical time series data, it completes global analysis (such as global comparison of crystallization process parameters, summary of abnormal working conditions, and batch production trend analysis). After obtaining the global analysis results, it pushes them to the anomaly monitoring and early warning module. At the same time, based on the global analysis results, it issues optimization instructions (such as threshold adjustment instructions and judgment standard optimization instructions) to the edge processing node module to update the preliminary judgment rules, thereby realizing bidirectional linkage between the cloud and the edge.

[0033] The anomaly monitoring and early warning module generates tiered early warning signals based on global analysis results pushed from the cloud. The specific tiering criteria are determined according to the production characteristics of different food additives. The following example shows the early warning threshold for citric acid: Yellow alert: Parameters slightly deviate from the threshold (e.g., temperature 47.5℃ for ≥10s for conventional citric acid, pH 2.15 for ≥10s). This threshold is set based on the principle that "exceeding the conventional parameter threshold (43~47℃, 1.9~2.1) but not reaching the abnormal impact threshold, only minor adjustments are needed." A notification is sent to the on-site control terminal (audio-visual alarm to alert operators). Red alert: Parameters severely exceed the standard (e.g., temperature ≥50℃ for ≥5s for conventional citric acid, pH ≤1.8 for ≥5s). This threshold is set based on the principle that "exceeding the abnormal impact threshold will lead to a decrease in crystallization rate and excessive acidity, requiring emergency handling." A notification is simultaneously sent to the on-site control terminal and the mobile phone of management personnel (audio-visual alarm + SMS reminder, mandating a shutdown for inspection). All alert signals include abnormal parameters, abnormal duration, and abnormal process nodes for rapid response.

[0034] The data archiving and management module adopts the AES-256 encryption standard to encrypt and store the entire process data from the data acquisition and upload module, edge processing node module, and cloud analysis and processing module. It provides specific data retrieval and traceability query interfaces to the control terminals. The interface format includes a web-based query interface and a terminal software interface, meeting the hardware specifications and communication protocols of the on-site control terminals to ensure smooth data retrieval. The query method involves entering the batch number and process name to retrieve the corresponding data. Queryable content includes: production parameters for each batch, anomaly assessment records, early warning records, and process adjustment records, achieving full traceability of the production process and meeting food safety regulatory requirements.

[0035] The cloud-based analysis and processing module includes: The global anomaly tracing module collects and sorts the preliminary judgment results and regularized process data of all process nodes in a unified manner, extracts the abnormal process nodes and forms a list of nodes to be traced, and identifies the process node where the anomaly first occurred by combining the production process logic and the sequence of the process. It tracks the transmission and diffusion process and the scope of influence of the abnormal parameters along the process flow, and eliminates instantaneous random fluctuations by globally matching with historical time series data, thus forming a global anomaly tracing result. The multi-node collaborative verification module summarizes the preliminary judgment results of all process nodes to establish a global operating condition judgment dataset. Based on the production process logic, it performs consistency verification on the judgment results of adjacent processes, compares the global operating condition distribution with historical normal time series data, identifies and eliminates single point suspicious misjudgments, confirms systemic real anomalies, and forms global operating condition data. The quality stability analysis module binds and collects global operating condition data and standardized process data according to production batches, establishes a global correlation mapping between finished product quality indicators and parameters of each process, statistically analyzes parameter fluctuations and consistency deviations between batches, locates key process nodes that affect quality stability, completes quality stability classification assessment, and outputs quality stability analysis results. The multi-node parameter linkage module performs multi-process node parameter linkage analysis based on global operating condition data, historical normal time series data, and quality stability analysis results, and generates parameter linkage analysis results.

[0036] The multi-node parameter linkage module includes: The full-process linkage model construction module, based on the production process logic, combined with global operating condition data, historical normal time sequence data and quality stability analysis results, determines the linkage relationship between parameters of each process in the food additive production line, extracts the core process parameters of each process, and constructs a full-process parameter linkage model through industrial Internet data association algorithms. The parameter quality weighting module, based on the results of quality stability analysis, statistically analyzes the correlation between parameters of each process and quality indicators, calculates the quality deviation corresponding to parameter fluctuations, and determines the priority of parameter optimization through weight quantification and assignment. The global parameter adjustment module determines the basic adjustment range of individual parameters and the coordinated adjustment range of parameters for each process based on historical normal time-series data and production process logic. The analysis and optimization module is used to generate analysis and optimization parameters and corresponding working condition optimization instructions that are adapted to the global working conditions based on the full-process parameter linkage model, parameter optimization priority, basic adjustment range of individual parameters and coordinated adjustment range of parameters in each process. The parameter command sending module is used to receive the analysis and optimization parameters and operating condition optimization commands output by the analysis and optimization module, send the analysis and optimization parameters to the edge processing node module, and send the operating condition optimization commands to the field equipment of the food additive production line.

[0037] The analysis and optimization module includes: The parameter priority matching submodule is used to combine parameter optimization priority with the full-process parameter linkage model, and according to the production process logic of the food additive production line, prioritize various parameters, classify them into core control parameters and auxiliary control parameters, determine the order of parameter control, and output the parameter control ranking results. The control parameter calculation submodule is used to receive the parameter control sorting results, combine the full process parameter linkage model, the basic adjustment range of a single parameter and the coordinated adjustment range of parameters in each process, refer to historical normal time series data and production process requirements, and calculate the control parameter combination and corresponding equipment control parameter value that are suitable for the current production conditions through the industrial Internet data association algorithm. The working condition optimization instruction generation submodule is used to receive equipment control parameter values, combine them with the linkage relationship between parameters of each process determined by the full process parameter linkage model, and convert them into working condition optimization instructions that can be recognized by the equipment on the production line. The parameter instruction verification submodule is used to verify the compliance of control parameter combinations and operating condition optimization instructions. It checks whether the parameter adjustment range conforms to the basic adjustment range of a single parameter and the coordinated adjustment range of parameters in each process. It also verifies whether the instructions match the equipment operation logic. At the same time, it verifies the consistency between parameters in conjunction with the full process parameter linkage model. When the verification is abnormal, it generates a parameter correction prompt containing abnormal parameter identifier, out-of-range value, and allowable adjustment range, and feeds it back to the control parameter calculation submodule for recalculation and adjustment. After the verification is qualified, the control parameter combination and operating condition optimization instructions are transmitted to the global operating condition adaptation submodule. The global operating condition adaptation submodule is used to combine the global operating condition data output by the multi-node collaborative verification module, refer to the linkage relationship between parameters of each process determined by the full process parameter linkage model and the adjustment range defined by the global parameter adjustment module, and quantitatively adapt and adjust the qualified control parameter combination and operating condition optimization instructions respectively, and output the judgment optimization parameters and operating condition optimization instructions adapted to the global operating condition.

[0038] The following section uses sodium carboxymethyl cellulose in the production of food additives to illustrate implementation details: The data acquisition and upload module collects raw process data based on the four core processes of the CMC production line: ingredient preparation, emulsification, homogenization, and drying. Specific material process parameters include: ingredient concentration (2.5%~3.5%), emulsification temperature (80~90℃), homogenization pressure (32~38MPa), drying temperature (115~125℃), and finished product viscosity (500~1000mPa·s). Depending on the scenario, it is divided into two levels: Level 1 CMC viscosity of 500~800mPa·s and Level 2 CMC viscosity of 800~1000mPa·s, meeting the production needs of different levels of CMC. The data acquisition and upload module also adds multi-process data association storage (such as the association between ingredient concentration and finished product viscosity). After being uploaded to the edge processing node module, it is synchronously pushed to the cloud analysis and processing module, adapting to the highly coupled production needs of CMC multiple processes. The edge processing node module now includes multi-process adaptation: a time-series feature extraction module extracts the time-series features of individual nodes for four processes; a local operating condition analysis module performs three-level analysis on each of the four processes, generating preliminary analysis results for each process and simultaneously uploading them to the cloud analysis and processing module; and an analysis rule update module receives optimization instructions from the cloud and updates the analysis and judgment rules for each process. After extraction, the local operating condition analysis module generates a dataset of time-series features for each node (such as time-series fluctuations in ingredient concentration and trends in homogenization pressure) for subsequent local operating condition analysis. After generating preliminary analysis results, the local operating condition analysis module simultaneously uploads the preliminary analysis results and the regularized process data for each node to the cloud analysis and processing module, while also pushing the analysis process data to the data archiving and management module.

[0039] Specifically, using a CMC production scenario example: When an abnormal ingredient concentration occurs in the batching process (measured at 2.2%, lower than the process threshold of 2.5%~3.5%), the threshold comparison submodule marks the process as a suspected anomaly, records the abnormal parameter (ingredient concentration 2.2%) and the comparison result (lower than the lower limit of 0.3%); the trend matching submodule compares historical anomaly time-series data, eliminates false anomalies caused by instantaneous feed fluctuations, confirms the anomaly as a real anomaly, labels the anomaly type (low ingredient concentration) and records the parameter time-series data (concentration ≤ 2.2% for 15 consecutive seconds); process correlation verification. The submodule combines the process execution sequence (ingredient preparation → emulsification → homogenization) and process coupling constraints (low ingredient concentration leads to insufficient emulsification), verifies the current abnormal characteristics and historical abnormal time series data (low ingredient concentration in the past 3 batches all caused emulsification abnormalities), excludes normal process fluctuations, and generates preliminary judgment results (including operating condition judgment: abnormal; abnormal process node: ingredient preparation process; abnormal parameter type: low ingredient concentration). At the same time, all the data of the above judgment process (threshold comparison records, trend matching results, abnormal time series data, and verification conclusions) are pushed to the data archiving management module.

[0040] like Figure 5As shown, the cloud-based analysis and processing module specifically includes: a global anomaly tracing module, which uniformly collects and sorts the preliminary judgment results and standardized process data of the four processes, extracts abnormal process nodes, and forms a list of nodes to be traced; combining the CMC production process logic (ingredient preparation, emulsification, homogenization, drying) and the sequence of processes, it identifies the process node where the anomaly first occurred, and tracks the transmission and diffusion process and impact range of the abnormal parameters along the process flow; through global matching with historical time-series data, it eliminates instantaneous random fluctuations (such as short-term fluctuations in homogenization pressure <3s), forming a global anomaly tracing result. Example: The viscosity of a batch of finished products does not meet the standard (first-level CMC <500mPa·s). After tracing analysis, the source of the anomaly is the ingredient preparation process (ingredient concentration is only 2.2%). This anomaly is transmitted to the emulsification and homogenization processes, resulting in insufficient emulsification and low homogenization pressure, ultimately affecting the viscosity of the finished product. The tracing result is simultaneously pushed to the control terminal and data archiving module. The multi-node collaborative verification module summarizes the preliminary judgment results of four process nodes: batching, emulsification, homogenization, and drying, and establishes a global operating condition judgment dataset. Based on the CMC production process logic (e.g., homogenization pressure must be adjusted after the batching concentration reaches the standard), it performs consistency verification on the judgment results of adjacent processes (e.g., batching and emulsification). Specific implementation scenario: If the batching process is judged as "normal" (concentration 2.8%), but the emulsification process is judged as "abnormal" (insufficient emulsification), the system will trigger a consistency verification to check whether it is a fault in the emulsifier equipment. This eliminates potential false alarms at downstream points caused by misjudgments of upstream parameters, confirms systemic anomalies, and forms a global operating condition dataset containing process consistency deviations. The quality stability analysis module binds and aggregates global operating condition data and standardized process data by production batch, establishing a global correlation mapping between finished product quality indicators (viscosity, moisture content) and parameters of each process; it statistically analyzes parameter fluctuations and consistency deviations between batches (such as the fluctuation range of ingredient concentrations between different batches and the deviation value of homogenization pressure); it identifies key process nodes affecting quality stability (such as the homogenization process, where pressure fluctuations have the greatest impact on finished product viscosity), completes a quality stability classification assessment (Grade A: batch deviation ≤ 5%; Grade B: batch deviation 5%~10%; Grade C: batch deviation > 10%), and outputs the quality stability analysis results.

[0041] The multi-node parameter linkage module includes the following modules: A full-process linkage model construction module, based on CMC production process logic, combined with global operating condition data, historical normal time-series data, and quality stability analysis results, determines the linkage relationship between four process parameters (e.g., a 0.5% increase in ingredient concentration requires a 2°C increase in emulsification temperature and a 1MPa increase in homogenization pressure); extracts the core process parameters of each process (ingredient concentration, homogenization pressure, drying temperature); and uses the Pearson correlation coefficient algorithm to construct a full-process parameter linkage model for parameter linkage analysis and optimization. A parameter quality weighting module, based on quality stability analysis results, statistically analyzes the correlation between each process parameter and finished product quality indicators (viscosity, moisture content), calculates the quality deviation corresponding to parameter fluctuations (e.g., a 1MPa fluctuation in homogenization pressure results in a 50mPa·s fluctuation in finished product viscosity); and uses the AHP (Analytic Hierarchy Process) to determine parameter optimization priorities through weight quantification. A global parameter adjustment module, based on historical normal time-series data and CMC production process logic, determines the basic adjustment range of individual parameters and the coordinated adjustment range of parameters in each process. Two schemes are used as examples below: Adjustment Scheme 1 (Level 1 CMC, used for beverage thickening, balancing efficiency and cost, allowing for minor parameter fluctuations): Basic adjustment range for individual parameters: homogenization pressure ±2MPa, ingredient concentration ±0.3%, drying temperature ±3℃; Coordination adjustment range: when homogenization pressure increases by 2MPa, ingredient concentration increases by 0.2% simultaneously, and drying temperature decreases by 2℃ simultaneously to avoid excessive viscosity of the finished product; Triggering condition: when the adjustment range of the core parameter (homogenization pressure) reaches 50% of the basic adjustment range (i.e., ±1MPa), coordination adjustment is triggered. The core parameter is adjusted first, and then the auxiliary parameters are adjusted simultaneously. The adjustment process strictly follows the parameter linkage relationship determined by the full-process linkage model to ensure coordinated matching of parameters in each process. Adjustment Plan 2 (Second-level CMC, used for jelly and yogurt, requiring high parameter precision to avoid fluctuations that could lead to substandard taste and stability of the finished product): Basic adjustment range for individual parameters: homogenization pressure ±1MPa, ingredient concentration ±0.2%, drying temperature ±2℃; Coordinated adjustment range: when the homogenization pressure increases by 1MPa, the ingredient concentration increases by 0.1% simultaneously, and the drying temperature decreases by 1℃ simultaneously, ensuring that the viscosity of the finished product remains stable at 800~1000mPa·s; Triggering condition: when the adjustment range of the core parameter (homogenization pressure) reaches 50% of the basic adjustment range (i.e., ±0.5MPa), coordinated adjustment is triggered, strictly following the logic of prioritizing core parameters and synchronizing auxiliary parameters. The adjustment process strictly follows the parameter linkage relationship determined by the full-process linkage model to ensure coordinated matching of parameters in each process.

[0042] The optimization module's specific process is as follows: The parameter priority matching submodule, combining the optimization priorities (homogenization pressure > ingredient concentration > drying temperature > emulsification temperature) determined by the parameter quality weighting module with the full-process parameter linkage model, prioritizes various parameters according to the CMC production process logic, classifying them into core control parameters (homogenization pressure) and auxiliary control parameters (ingredient concentration, drying temperature, emulsification temperature). It then determines the order of parameter control (adjusting homogenization pressure first, then ingredient concentration, and finally drying and emulsification temperatures), outputting the parameter control ranking result, which is transmitted to the control parameter calculation submodule. The control parameter calculation submodule receives the parameter control ranking result, and, combining the full-process parameter linkage model, the basic adjustment range of individual parameters, and the coordinated adjustment range of parameters in each process, referencing historical normal time-series data and CMC production process requirements, uses a multiple linear regression algorithm to calculate the appropriate combination of control parameters and corresponding equipment control parameter values ​​for the current production conditions. The operating condition optimization instruction generation submodule receives equipment control parameter values ​​and, based on the linkage relationship between parameters in each process determined by the full-process parameter linkage model, converts them into operating condition optimization instructions recognizable by the production line equipment (e.g., homogenizer: pressure adjusted to 35MPa, speed adjusted to 1100r / min; batching tank: concentration adjusted to 2.8%, stirring speed adjusted to 220r / min). The instruction format conforms to the communication protocol of the field equipment (e.g., Modbus protocol). The parameter instruction verification submodule performs compliance verification on the combination of control parameters and operating condition optimization instructions: it checks whether the parameter adjustment range conforms to the basic adjustment range and coordinated adjustment interval of a single parameter (e.g., homogenizing pressure 35MPa, within the range of 32~38MPa, meets the requirements); it verifies whether the instructions match the equipment operating logic (e.g., before adjusting the homogenizer pressure, confirm that the equipment is fault-free); and it verifies the consistency between parameters based on the full-process parameter linkage model (e.g., when homogenizing pressure is adjusted to 35MPa, the batching concentration is simultaneously adjusted to 2.8%, meeting the coordination requirements). Combining the global operating condition data output by the multi-node collaborative verification module (current global operating condition is level B, batch deviation 7%), and referring to the linkage relationship between parameters of each process determined by the full-process parameter linkage model and the adjustment range defined by the global parameter adjustment module, the verified control parameter combinations and operating condition optimization instructions are quantitatively adapted and adjusted (e.g., the homogenization pressure is fine-tuned from 35MPa to 35.5MPa to meet the current global operating condition), and the judgment optimization parameters and operating condition optimization instructions adapted to the global operating condition are output. The parameter instruction issuing module is used to receive the judgment optimization parameters and operating condition optimization instructions output by the judgment optimization module, where the judgment optimization parameters are generated by the judgment optimization module to ensure logical closed loop; the judgment optimization parameters are sent to the edge processing node module to update the local judgment rules, and the operating condition optimization instructions are sent to the CMC production line field equipment, such as homogenizers, batching tanks, dryers, etc.; at the same time, the multi-node parameter linkage module works collaboratively through the above sub-modules to generate parameter linkage analysis results.The anomaly monitoring and early warning module generates tiered early warning signals based on the global analysis results output by the cloud-based analysis and processing module. The following is an example of the tiering criteria: Yellow alert: Slight deviation of a single process parameter without related anomalies (e.g., ingredient concentration 2.4%, slightly lower than 2.5%, no other process anomalies), pushed to the on-site control terminal, prompting operators to fine-tune the parameters; Orange alert: Severe deviation of a single process parameter, or related anomalies (e.g., homogenization pressure 30MPa, lower than 32MPa, and finished product viscosity not meeting standards), pushed to the on-site control terminal and the team leader's mobile phone, requiring shutdown and inspection of the process; Red alert: Systemic anomalies in multiple processes, or batch deviation in quality stability assessment >10%, simultaneously pushed to the on-site personnel, management personnel's mobile phones, and the enterprise monitoring platform, forcing shutdown for rectification and investigation of all process anomalies.

[0043] The above description of specific embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A method for real-time monitoring of food additive production lines based on the Industrial Internet, characterized in that, The system includes a monitoring system, which comprises a data acquisition and uploading module, an edge processing node module, a cloud analysis and processing module, an anomaly monitoring and early warning module, and a data archiving and management module. The data acquisition and upload module is used to acquire the original process data of each process node in the food additive production line, encapsulate and format it into regular process data, acquire the production process logic and historical time sequence data of food additives, and transmit the above data to the edge processing node module. The cloud-based analysis and processing module communicates bidirectionally with the edge processing node module. Based on the preliminary judgment results, the regularized process data of all process nodes, the production process logic, and historical time series data, it performs analysis to obtain global analysis results and pushes them to the anomaly monitoring and early warning module. Based on the global analysis results, it issues optimization instructions to the edge processing node module to update the preliminary judgment rules. The anomaly monitoring and early warning module generates tiered early warning signals based on global analysis results and pushes them to the control terminal; The data archiving and management module is used to encrypt and store data from the data acquisition and upload module, the edge processing node module, and the cloud analysis and processing module, and provides data retrieval and traceability query interfaces to the control terminal.

2. The method for real-time monitoring of a food additive production line based on the Industrial Internet according to claim 1, characterized in that, The data acquisition and upload module includes: The raw data acquisition and encapsulation module is used to acquire raw process data and encapsulate and format it into regular process data. The raw process data includes material process parameters, equipment operating parameters, and process execution parameters. The process baseline acquisition module is used to acquire the preset production process logic, which includes the process parameter thresholds, process execution sequence and process coupling constraints for each process node. The historical time series data acquisition module is used to acquire historical normal time series data and historical abnormal time series data. The historical normal time series data is the process parameter time series data under qualified operating conditions of the production line, and the historical abnormal time series data is the process parameter time series data when the production line experiences abnormal parameters, abnormal processes, abnormal equipment, or abnormal quality conditions. The data upload module is used to upload standardized process data, production process logic, historical normal time series data, and historical abnormal time series data to the edge processing node module.

3. The method for real-time monitoring of a food additive production line based on the Industrial Internet according to claim 2, characterized in that, The edge processing node module includes: The time-series feature extraction module extracts the time-series change trends, fluctuation amplitudes, and steady-state durations of parameters for corresponding food additive production process nodes based on the regularized process data, forming time-series features for a single node. The local operating condition analysis module matches the time-series characteristics of a single node with the process parameter thresholds and process coupling constraints in the production process logic. At the same time, it compares historical normal time-series data with historical abnormal time-series data. Based on the matching and comparison results, it makes a preliminary assessment of the operating condition of a single food additive production process node, generates a preliminary assessment result, and synchronously uploads the preliminary assessment result and the regularized process data of a single node to the cloud analysis and processing module. The judgment rule update module receives the judgment optimization parameters sent by the cloud analysis and processing module, and updates the judgment rules of the local working condition judgment module according to the judgment optimization parameters.

4. The method for real-time monitoring of a food additive production line based on the Industrial Internet according to claim 3, characterized in that, The local operating condition analysis module includes: The threshold comparison submodule compares the material process parameters, equipment operating parameters, and process execution parameters in the time-series characteristics of a single node with the process parameter thresholds of the corresponding process node one by one. If any parameter exceeds the process parameter threshold range, the process node is marked as a suspected anomaly and the corresponding abnormal parameters are recorded and transmitted to the trend matching submodule. If all parameters are within the process parameter threshold range, the process node is determined to be in normal operating condition. The trend matching submodule constructs a parameter trend comparison benchmark based on historical normal time series data. It compares and analyzes the time series characteristics of individual nodes corresponding to suspected abnormal process nodes with the parameter trend comparison benchmark, eliminates false suspected anomalies caused by instantaneous material fluctuations and data acquisition errors, confirms the remaining suspected anomalies as real anomalies, marks the anomaly type and records the parameters, and then transmits them to the process association verification submodule. The process correlation verification submodule verifies actual anomalies based on the process execution sequence, process coupling constraints, and historical anomaly time series data. It verifies the current anomaly characteristics by combining historical anomaly time series data to eliminate duplicate misjudgments. At the same time, based on the process execution sequence and process coupling constraints, it determines whether the anomaly is a normal process fluctuation in the process connection of the food additive production line. If it is a normal process fluctuation, the process node is determined to be in normal operating condition. If it is still an anomaly after excluding normal process fluctuations, a preliminary judgment result containing operating condition judgment, abnormal process node, and abnormal parameter type is generated. The preliminary judgment result is uploaded to the cloud analysis and processing module, and the judgment process data is pushed to the data archiving management module.

5. The method for real-time monitoring of a food additive production line based on the Industrial Internet according to claim 1, characterized in that, The cloud-based analysis and processing module includes: The global anomaly tracing module collects and sorts the preliminary judgment results and regularized process data of all process nodes in a unified manner, extracts the abnormal process nodes and forms a list of nodes to be traced, and identifies the process node where the anomaly first occurred by combining the production process logic and the sequence of the process. It tracks the transmission and diffusion process and the scope of influence of the abnormal parameters along the process flow, and eliminates instantaneous random fluctuations by globally matching with historical time series data, thus forming a global anomaly tracing result. The multi-node collaborative verification module summarizes the preliminary judgment results of all process nodes to establish a global operating condition judgment dataset. Based on the production process logic, it performs consistency verification on the judgment results of adjacent processes, compares the global operating condition distribution with historical normal time series data, identifies and eliminates single point suspicious misjudgments, confirms systemic real anomalies, and forms global operating condition data. The quality stability analysis module binds and collects global operating condition data and standardized process data according to production batches, establishes a global correlation mapping between finished product quality indicators and parameters of each process, statistically analyzes parameter fluctuations and consistency deviations between batches, locates key process nodes that affect quality stability, completes quality stability classification assessment, and outputs quality stability analysis results. The multi-node parameter linkage module performs multi-process node parameter linkage analysis based on global operating condition data, historical normal time series data, and quality stability analysis results, and generates parameter linkage analysis results.

6. The method for real-time monitoring of a food additive production line based on the Industrial Internet according to claim 5, characterized in that, The multi-node parameter linkage module includes: The full-process linkage model construction module, based on the production process logic, combined with global operating condition data, historical normal time sequence data and quality stability analysis results, determines the linkage relationship between parameters of each process in the food additive production line, extracts the core process parameters of each process, and constructs a full-process parameter linkage model through industrial Internet data association algorithms. The parameter quality weighting module, based on the results of quality stability analysis, statistically analyzes the correlation between parameters of each process and quality indicators, calculates the quality deviation corresponding to parameter fluctuations, and determines the priority of parameter optimization through weight quantification and assignment. The global parameter adjustment module determines the basic adjustment range of individual parameters and the coordinated adjustment range of parameters for each process based on historical normal time-series data and production process logic. The analysis and optimization module is used to generate analysis and optimization parameters and corresponding working condition optimization instructions that are adapted to the global working conditions based on the full-process parameter linkage model, parameter optimization priority, basic adjustment range of individual parameters and coordinated adjustment range of parameters in each process. The parameter command sending module is used to receive the analysis and optimization parameters and operating condition optimization commands output by the analysis and optimization module, send the analysis and optimization parameters to the edge processing node module, and send the operating condition optimization commands to the field equipment of the food additive production line.

7. The method for real-time monitoring of a food additive production line based on the Industrial Internet according to claim 6, characterized in that, The analysis and optimization module includes: The parameter priority matching submodule is used to combine parameter optimization priority with the full-process parameter linkage model, and according to the production process logic of the food additive production line, prioritize various parameters, classify them into core control parameters and auxiliary control parameters, determine the order of parameter control, and output the parameter control ranking results. The control parameter calculation submodule is used to receive the parameter control sorting results, combine the full process parameter linkage model, the basic adjustment range of a single parameter and the coordinated adjustment range of parameters in each process, refer to historical normal time series data and production process requirements, and calculate the control parameter combination and corresponding equipment control parameter value that are suitable for the current production conditions through the industrial Internet data association algorithm. The working condition optimization instruction generation submodule is used to receive equipment control parameter values, combine them with the linkage relationship between parameters of each process determined by the full process parameter linkage model, and convert them into working condition optimization instructions that can be recognized by the equipment on the production line. The parameter instruction verification submodule is used to verify the compliance of control parameter combinations and operating condition optimization instructions. It checks whether the parameter adjustment range conforms to the basic adjustment range of a single parameter and the coordinated adjustment range of parameters in each process. It also verifies whether the instructions match the equipment operation logic. At the same time, it verifies the consistency between parameters in conjunction with the full process parameter linkage model. When the verification is abnormal, it generates a parameter correction prompt containing abnormal parameter identifier, out-of-range value, and allowable adjustment range, and feeds it back to the control parameter calculation submodule for recalculation and adjustment. After the verification is qualified, the control parameter combination and operating condition optimization instructions are transmitted to the global operating condition adaptation submodule. The global operating condition adaptation submodule is used to combine the global operating condition data output by the multi-node collaborative verification module, refer to the linkage relationship between parameters of each process determined by the full process parameter linkage model and the adjustment range defined by the global parameter adjustment module, and quantitatively adapt and adjust the qualified control parameter combination and operating condition optimization instructions respectively, and output the judgment optimization parameters and operating condition optimization instructions adapted to the global operating condition.

Images