Feed production line monitoring control method and system based on joint machine learning model

Abstract

The invention provides a feed production line monitoring control method and system based on a joint machine learning model, and relates to the technical field of machine learning, and the method comprises the steps: firstly obtaining a process data set of a feed production line, including equipment operation state data, material flow track data and process instruction execution data; then performing feature collaborative coding processing on the process data set to generate a production line state feature set containing equipment operation adaptation features, material flow matching features and process execution coordination features, and then calling a pre-constructed joint machine learning model to perform abnormal development deduction processing on the production line state feature set; the method comprises the steps of obtaining a production line abnormal development deduction result, identifying key abnormal links of a production line according to the deduction result, including an equipment performance degradation link, a material quality deterioration link and a process connection dislocation link, finally generating a production line regulation and control instruction combination based on the key abnormal links, sending the combination to a production line execution unit, and triggering targeted control operation. And effective monitoring and control of the feed production line are realized.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and more specifically, to a method and system for monitoring and controlling feed production lines based on a joint machine learning model. Background Technology

[0002] In the feed production industry, the stable operation of feed production lines is crucial for ensuring feed yield, quality, and production efficiency. Traditional feed production line monitoring and control methods mainly rely on manual inspections and simple sensor data monitoring. Manual inspections are not only labor-intensive and time-consuming, but also difficult to grasp the real-time and comprehensive operating status of the production line, easily overlooking some potential anomalies. While simple sensor data monitoring can obtain some equipment operating parameters, this data is often isolated and lacks comprehensive analysis of data on equipment operating status, material flow trajectory, and process command execution.

[0003] In existing technologies, even those monitoring methods based on machine learning mostly model and analyze single data sources or single processes, failing to fully consider the complex relationships between equipment, materials, and processes in the feed production line. This makes it difficult to accurately predict the development trend of abnormalities in the production line during actual production, and to monitor key abnormal processes in a timely manner. Consequently, it is impossible to take effective control measures in a timely manner, which can easily lead to production interruptions, product quality decline, and other problems, resulting in significant economic losses for feed production enterprises. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a feed production line monitoring and control method based on a joint machine learning model, the method comprising: Acquire a process data set of the feed production line, the process data set including equipment operating status data, material flow trajectory data and process instruction execution data; The process data set is subjected to feature co-coding to generate a production line status feature set, which includes equipment operation adaptation features, material flow matching features, and process execution coordination features. The pre-built joint machine learning model is invoked to perform abnormal development inference processing on the set of production line state features, and the abnormal development inference results of the production line are generated. Based on the results of the production line anomaly development simulation, key abnormal links in the production line are identified. These key abnormal links include equipment performance degradation, material quality deterioration, and process connection misalignment. Based on the aforementioned critical abnormal process, a combination of production line control instructions is generated, and the combination of control instructions is sent to the production line execution unit to trigger targeted control operations.

[0005] In another aspect, the present invention also provides a feed production line monitoring and control system based on a joint machine learning model, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the machine-readable storage medium to implement the above-mentioned method.

[0006] Based on the above, this invention acquires a process data set from a feed production line, including equipment operating status data, material flow trajectory data, and process instruction execution data. It then performs feature co-coding to generate a production line status feature set containing equipment operating adaptation features, material flow matching features, and process execution coordination features. This comprehensively reflects the actual operating status of the feed production line, fully considering the complex relationships between equipment, materials, and processes. It overcomes the limitations of traditional methods that rely on isolated data analysis. By calling a pre-built joint machine learning model to perform anomaly prediction on the production line status feature set, it can accurately generate anomaly prediction results, enabling precise prediction of anomaly trends and early detection of potential problems. Based on the prediction results, it identifies key abnormal links in the production line and generates a combination of production line control instructions, sending them to the production line execution unit to trigger targeted control operations. This allows for timely and effective control of production line anomalies, ensuring stable operation of the feed production line, improving feed yield and quality, and reducing economic losses during production. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the execution flow of the feed production line monitoring and control method based on a joint machine learning model provided in an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of exemplary hardware and software components of a feed production line monitoring and control system based on a joint machine learning model provided in an embodiment of the present invention. Detailed Implementation

[0009] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a feed production line monitoring and control method based on a joint machine learning model, according to an embodiment of the present invention. The following is a detailed description of this feed production line monitoring and control method based on a joint machine learning model.

[0010] Step S110: Obtain the process data set of the feed production line, wherein the process data set includes equipment operating status data, material flow trajectory data, and process instruction execution data.

[0011] In actual feed production scenarios, the collection of equipment operating status data relies on various sensors installed on different types of equipment on the feed production line. For example, in the pelleting process, the pellet mill is the core equipment. To accurately monitor its operating status, temperature sensors, pressure sensors, and speed sensors can be installed. Temperature sensors can detect real-time temperature changes inside the pellet mill, as a suitable and stable temperature is crucial for the quality of feed pelleting. If the temperature is too high, it may destroy the nutrients in the feed; if the temperature is too low, the pelleting effect may be poor. Pressure sensors monitor the pressure inside the pellet mill; the stability of the pressure directly affects the pellet density and forming effect. Speed ​​sensors record the operating speed of the pellet mill; a stable speed ensures uniformity of pelleting and production efficiency. Similarly, in the cooling process, temperature sensors can be installed on the cooler to monitor the temperature drop during the cooling process, ensuring that the feed is effectively cooled to a suitable storage temperature. Airflow sensors are also installed to monitor the airflow of the cooler; an appropriate airflow ensures uniform cooling. Weight sensors can be installed on the baling machine during the baling process to ensure that the weight of each bale of feed meets the specified standards; speed sensors can be installed to monitor the baling speed and ensure baling efficiency. These sensors continuously collect the operating parameters of the equipment, forming equipment operating status data.

[0012] The acquisition of material flow trajectory data relies on Radio Frequency Identification (RFID) and visual recognition technologies. For example, RFID tags can be attached to material containers or packaging bags. These tags contain basic material information, such as material type and batch number. RFID readers are installed at key locations on the feed production line, such as the raw material silo exit, the junctions between processes, and the finished product warehouse entrance. When materials pass through these locations, the readers automatically read the tag information, recording the material's position and the time it passed through that location. For instance, when materials enter the pelleting process from the raw material silo, the reader at the raw material silo exit reads the tag information and records the time the material leaves the silo. At the junctions between processes, readers can monitor whether materials enter the next process on time and in the correct order. Visual recognition technology uses cameras installed at key locations on the production line to identify the appearance, shape, and other characteristics of the materials. The cameras capture images of the materials, and image analysis algorithms determine whether the material's appearance is normal and whether there is any damage or deformation. These data, when combined, provide a detailed record of the entire material flow process from the raw material warehouse, through various processes, to the finished product warehouse, forming material flow trajectory data.

[0013] Process instruction execution data is obtained from the feed production line's control system, which is responsible for issuing various process instructions and recording their execution status. When a process instruction for a pelleting process is issued, the control system records the specific content of the instruction, including pellet specifications, output requirements, temperature and pressure setpoints, etc. It also records the start time of the process instruction execution and the actual output and quality of the pellets. For example, if the process instruction requires pellets to have specific dimensions and shapes, the control system records whether the actual produced feed meets these specifications. If the actual output deviates from the required output, the control system also records this. Similarly, the process instructions for cooling and packaging processes are recorded in detail, along with their execution results. This data reflects the execution status of process instructions during the actual operation of the production line, forming the process instruction execution data.

[0014] Step S120: Perform feature co-coding processing on the process data set to generate a production line status feature set, which includes equipment operation adaptation features, material flow matching features, and process execution coordination features.

[0015] After obtaining the process data set of the feed production line, it is necessary to perform feature co-coding on these data to generate a production line status feature set that can accurately reflect the production line status.

[0016] Step S121: Perform operation mode recognition processing on the equipment operation status data, and extract the operation parameter matching features of the equipment in typical processes as equipment operation adaptation features. The equipment operation adaptation features include intra-process parameter stability description and inter-process parameter connection description. The typical processes include granulation process, cooling process and packaging process.

[0017] For equipment operating status data, the first step is to identify the operating mode. In typical feed production line processes, the pelleting, cooling, and packaging processes each have their unique operating requirements and characteristics. In the pelleting process, operating parameters of the pellet mill, such as temperature, pressure, and speed, need to remain relatively stable to ensure pellet quality. The description of parameter stability within the process involves a detailed analysis of the fluctuations of these parameters during pelleting. Through the study and analysis of a large amount of historical data, the normal fluctuation range for each parameter is determined. For example, by analyzing the pellet mill's temperature data over a period of time, it is observed that under normal circumstances, the temperature fluctuates within a specific range. The actual operating temperature parameters are compared with this normal fluctuation range. If the temperature fluctuates within the normal range, it indicates that the pellet mill's heating system is working normally; if the temperature frequently exceeds the normal range, it may mean that the heating system is malfunctioning or the process settings are unreasonable. Similarly, a similar analysis is performed on pressure and speed parameters to determine their stability within the process.

[0018] The description of inter-process parameter connectivity focuses on the changes in equipment parameters between different processes. Taking the pelleting process to the cooling process as an example, there should be a reasonable connection between the pellet mill's discharge temperature and the cooler's feed temperature. Under normal circumstances, the pellet mill's discharge temperature will be at a relatively high level, and the cooler's feed temperature should be able to match it, reducing the temperature to a suitable storage temperature through a reasonable cooling process. If the pellet mill's discharge temperature is too high and the cooler's feed temperature is set too low, it may lead to quality problems in the feed during the cooling process, such as surface condensation and uneven internal moisture distribution. By analyzing the changing trends and interrelationships of equipment parameters in adjacent processes, it is determined whether the parameters are reasonably connected. For example, by comparing historical data of the pellet mill's discharge temperature and the cooler's feed temperature, a reasonable relationship model between the two can be established. In actual production, this model is used to determine whether the current parameter connection meets the requirements. Combining the description of intra-process parameter stability and the description of inter-process parameter connectivity yields the equipment operation adaptation characteristics.

[0019] Step S122: Perform path consistency analysis on the material flow trajectory data, and extract the matching characteristics of the material flow path from the raw material warehouse to the finished product warehouse with the standard path as the material flow matching characteristics. The material flow matching characteristics include a description of path direction matching degree and a description of path speed balance degree.

[0020] Path consistency analysis of material flow trajectory data is performed to assess whether the material flow meets standard requirements. The flow of materials from the raw material warehouse to the finished product warehouse follows a pre-defined standard path, determined by the design and process requirements of the feed production line. Path direction matching describes a detailed comparison of the actual material flow path with the standard path in terms of direction. In a feed production line, the material flow direction should be clear and fixed. For example, materials should flow in the order of raw material warehouse to pelleting, then to cooling, and finally to the finished product warehouse. By analyzing the directional information at each key point on the path, the degree of directional matching between the actual path and the standard path is calculated. The path can be divided into multiple segments, and the direction of each segment can be compared. If the direction of most segments matches the standard path, the path direction matching is high; if many segments deviate significantly from the standard path, it may indicate an anomaly in the material flow process, such as accidentally entering other processes or an incorrect flow route.

[0021] Path speed uniformity description focuses on the speed changes of materials during the flow process. The material flow speed should remain relatively stable across different process stages. During the flow from the raw material silo to the pelleting process, if the material flow speed is too fast, it may lead to uneven feeding into the pellet mill, affecting pellet quality; if the flow speed is too slow, it may result in low production efficiency. The material flow speed is calculated by analyzing the flow time and distance at each location. Comparing the flow speeds at different locations determines whether the speeds are uniform. For example, a standard speed fluctuation range can be established. When the material flow speed at each location fluctuates within this range, it indicates good path speed uniformity; if the flow speed at some locations exceeds this range, further investigation is needed to check for equipment malfunctions, material blockages, etc. Combining the path direction matching description and the path speed uniformity description yields the material flow matching characteristics.

[0022] Step S123: Perform instruction response analysis on the process instruction execution data, and extract the matching features between the actual executed actions and the preset instructions as process execution coordination features. The process execution coordination features include a description of the timeliness of instruction response and a description of the completeness of action execution.

[0023] Analyzing the execution data of process instructions is crucial for evaluating their effectiveness in actual production. Preset instructions are formulated based on the process requirements of feed production, encompassing specific operational details and timeframes. The timeliness of instruction response primarily analyzes the speed at which equipment or operators respond to process instructions. For example, when a process instruction for the pelleting process is issued, the pellet mill should begin executing the instruction within a specified time. The control system records the time the instruction is issued and the actual start time of the pellet mill. By comparing these two times, the timeliness of the instruction response is determined. If the pellet mill starts executing quickly after the instruction is issued, it indicates a fast response speed and high production efficiency. If the pellet mill executes the instruction late, it may be due to equipment malfunction, operator unfamiliarity, or other reasons.

[0024] The description of action execution integrity focuses on whether the actual actions performed fully comply with the requirements of the preset instructions. In the pelleting process, preset instructions may require pellet specifications, output, quality, etc., to meet set standards. During actual execution, a detailed inspection of the actual pelleting situation is necessary. For example, by inspecting the appearance, size, and density of the pellets, it is determined whether the specifications meet the requirements; by statistically analyzing the actual pellet output and comparing it with the output required by the instructions, it is determined whether the output meets the standard; by testing the nutritional components and moisture content of the feed, it is determined whether the quality meets the standards. If there are deviations between the actual actions performed and the preset instructions, it may lead to product quality problems, requiring timely adjustments to the production process or maintenance of the equipment. Combining the description of instruction response timeliness and the description of action execution integrity yields the process execution coordination characteristics.

[0025] Step S124: Verify the logical correlation between the equipment operation adaptation feature, material flow matching feature and process execution coordination feature, and eliminate redundant features whose logical correlation does not meet the set correlation conditions.

[0026] After obtaining the equipment operation adaptation characteristics, material flow matching characteristics, and process execution coordination characteristics, it is necessary to verify the logical correlation between them. These characteristics should have a certain inherent connection, because the operating status of the equipment affects the material flow, and the material flow, in turn, affects the execution effect of process instructions. For example, if the pellet mill's operating status is unstable, it may cause abnormalities in the material flow during the pelleting process, thereby affecting the execution of pellet specifications and output in the process instructions. Through the analysis and mining of a large amount of historical data, a logical relationship model between the characteristics is established. In this logical relationship model, the degree of correlation between each characteristic and other characteristics is analyzed to determine whether their logical relationships are reasonable.

[0027] The correlation conditions are determined based on the actual operation and process requirements of the feed production line. If the logical correlation between a certain feature and other features does not meet the set conditions, it indicates that the feature may be redundant and has no practical significance for judging the production line status. For example, if the fluctuation of a certain equipment operating parameter has no obvious correlation with material flow and process execution, then the feature corresponding to that parameter can be eliminated. By removing redundant features in this way, the effectiveness and accuracy of the production line status feature set are improved.

[0028] Step S125: Standardize and splice the verified equipment operation adaptation features, material flow matching features, and process execution coordination features to generate a production line status feature set.

[0029] After verification and removal of redundant features, the validated equipment operation adaptation features, material flow matching features, and process execution coordination features were obtained. The next step is to standardize and stitch these features together. Standardization ensures that different features have the same scale and range, avoiding the impact of different dimensions on subsequent analysis and processing. For example, temperature and pressure parameters in the equipment operation adaptation features may have different dimensions; standardization converts them to the same scale.

[0030] The splicing process combines standardized equipment operation adaptation characteristics, material flow matching characteristics, and process execution coordination characteristics in a predetermined order to form a complete set of production line status characteristics. This set of production line status characteristics includes information on multiple aspects such as equipment operation, material flow, and process execution, and can comprehensively reflect the status of the feed production line.

[0031] Step S130: Call the pre-built joint machine learning model to perform abnormal development inference processing on the production line state feature set, and generate abnormal development inference results for the production line.

[0032] After generating the production line status feature set, a pre-built joint machine learning model needs to be invoked to perform anomaly prediction processing, in order to predict possible anomalies and their development trends on the production line. The joint machine learning model is a complex system containing multiple modules, each with its specific function. The processing procedure of each module is described in detail below.

[0033] Step S131: Input the production line status feature set into the historical status memory module of the joint machine learning model to extract the historical status feature sequence of the production line's historical normal operation stage.

[0034] First, the set of production line status features is input into the historical status memory module of the joint machine learning model. This module stores a large amount of status feature data from the feed production line during its normal historical operation. This data, collected over a period of time during normal production, includes feature information on various aspects such as equipment operation, material flow, and process execution. Through analysis and processing of this historical data, a historical status feature sequence is extracted. This sequence, arranged chronologically, demonstrates the changes in the production line's status features at different points in time.

[0035] In practical applications, the historical state memory module categorizes and organizes historical data, associating production line state characteristics at different points in time. For example, it combines equipment operation adaptation characteristics, material flow matching characteristics, and process execution coordination characteristics within the same time period to form a complete production line state feature vector. Then, it arranges the production line state feature vector in chronological order to construct a historical state feature sequence. By comparing the current production line state feature set with the historical state feature sequence, the difference between the current production line state and the historical normal state can be monitored. If there is a significant difference between the current state and the historical normal state, it may indicate an anomaly in the production line.

[0036] Step S132: Input the historical state feature sequence and the current production line state feature set into the time series extrapolation module of the joint machine learning model to calculate the evolutionary difference features between the current state and the historical normal state.

[0037] Next, the historical state feature sequence and the current production line state feature set are input into the time series extrapolation module of the joint machine learning model. The time series extrapolation module mainly focuses on the changes in the production line state over time.

[0038] Step S1321: Arrange the historical state feature sequence in chronological order to construct a historical state timeline.

[0039] This historical state timeline is a continuous time series, with each point in time corresponding to the state characteristics of the production line at that moment. Through the historical state timeline, the state changes of the production line during its historical normal operation can be monitored intuitively. For example, fluctuations in equipment operating parameters over different time periods, trends in material flow rate, and the stability of process command execution can be observed.

[0040] Step S1322: Map the current production line status feature set to the historical status time axis to determine the corresponding position of the current status on the time axis.

[0041] This step is achieved by comparing the similarity between the current state features and the historical state features. The time series extrapolation module calculates the similarity between the current state features and the state features at each time point on the historical state timeline, finds the time point with the highest similarity, maps the current state to that time point, and thus determines the relative position of the current state on the historical state timeline.

[0042] Step S1323: Extract the state features of the same process stage in the timeline of the current state and the historical state through the time series extrapolation module of the joint machine learning model, and calculate the feature difference value between the state features.

[0043] In a feed production line, the state characteristics of different process stages have different features and importance. For example, the equipment operating parameters in the pelleting process may differ significantly from those in the cooling process. Therefore, when calculating the characteristic difference value, only the state characteristics of the same process stage are compared.

[0044] The time-series simulation module compares the equipment operation adaptation characteristics, material flow matching characteristics, and process execution coordination characteristics of the same process stage in the current and historical states. For each characteristic, it calculates the degree of difference between the current and historical states. For example, for the temperature parameter in the equipment operation adaptation characteristics of the granulation process, it compares the current temperature value with the historical normal operating temperature value and calculates the difference between the two. The degree of difference of each characteristic is then combined to obtain the characteristic difference value between the state characteristics. The characteristic difference value reflects the degree of difference between the current state and the historical normal state at the same process stage.

[0045] Step S1324: Analyze the changing trend of the feature difference value at continuous time points to identify abnormal fluctuation patterns of the state features.

[0046] If the characteristic difference value remains relatively stable over a continuous period of time, it indicates that the production line status changes are relatively smooth. However, if the characteristic difference value suddenly increases or decreases, it suggests that the status characteristics may be experiencing abnormal fluctuations. For example, if the characteristic difference value of the equipment operation adaptation characteristics in the granulation process suddenly increases over a certain period, it may mean that the granulator has malfunctioned or that the process settings have changed. By analyzing the trend of the characteristic difference value, abnormal changes in the production line status can be monitored in a timely manner.

[0047] Step S1325: Construct evolutionary difference features between the current state and the historical normal state based on the feature difference value and the abnormal fluctuation pattern. The evolutionary difference features include feature deviation distribution information and abnormal fluctuation pattern information.

[0048] Evolutionary difference features include information on feature deviation distribution and abnormal fluctuation patterns. Feature deviation distribution describes the distribution of the degree of deviation of the current state features from historical normal state features across various feature dimensions. For example, it allows us to statistically analyze the distribution range of feature difference values ​​for different features, identifying which features have larger deviations and which are relatively stable. Abnormal fluctuation pattern information records the abnormal fluctuations in feature difference values ​​over continuous time points, including the amplitude and frequency of these fluctuations.

[0049] Step S133: Input the evolutionary difference features into the spatial correlation module of the joint machine learning model, analyze the mutual influence relationship between equipment operation adaptation features, material flow matching features and process execution coordination features, and generate a feature correlation influence matrix.

[0050] After obtaining the evolutionary difference features, they are input into the spatial correlation module of the joint machine learning model. The main task of this module is to analyze the mutual influence relationship between equipment operation adaptation features, material flow matching features, and process execution coordination features.

[0051] Step S1331: Assign feature identifiers to the equipment operation adaptation feature, material flow matching feature, and process execution coordination feature respectively.

[0052] In this embodiment, a feature identifier is a symbol or code used to uniquely identify each feature. By assigning feature identifiers, it becomes easier to distinguish and operate different features in subsequent processing. For example, a specific identifier can be assigned to the equipment operation adaptation feature, another different identifier can be assigned to the material flow matching feature, and yet another unique identifier can be assigned to the process execution coordination feature. Thus, when analyzing the mutual influence between features, these identifiers can be used to accurately determine which features are involved.

[0053] Step S1332: Construct a feature association analysis table, which includes records of the impact of equipment operation adaptation features on material flow matching features, records of the impact of equipment operation adaptation features on process execution coordination features, and records of the impact of material flow matching features on process execution coordination features.

[0054] The feature correlation analysis table records the interrelationships between different features. Specifically, it includes records of the impact of equipment operation adaptation features on material flow matching features, the impact of equipment operation adaptation features on process execution coordination features, and the impact of material flow matching features on process execution coordination features. In a feed production line, equipment operating status directly affects material flow. If the operating parameters of the pellet mill are unstable, the material flow rate in the pelleting process may change, thus affecting the material flow matching features. Similarly, equipment operating status also affects process execution coordination features; for example, a pellet mill malfunction may prevent process instructions from being executed on time or accurately. Material flow also affects process execution coordination features; if the material flow path is abnormal, it may lead to deviations in the execution of process instructions. The feature correlation analysis table records these relationships in detail.

[0055] Step S1333: Using the spatial association module of the joint machine learning model, extract the influence direction and degree information between the equipment operation adaptation features, material flow matching features and process execution coordination features from the evolutionary difference features.

[0056] The spatial correlation module extracts information on the direction and degree of influence among equipment operation adaptation features, material flow matching features, and process execution coordination features from evolutionary difference features. The direction of influence describes whether one feature affects another positively or negatively. For example, an increase in equipment temperature (a feature related to equipment operation adaptation) might lead to a faster material flow rate (a feature related to material flow matching), which is a positive influence; conversely, equipment failure (a feature related to equipment operation adaptation) might lead to a decrease in the timeliness of command response (a feature related to process execution coordination), which is a negative influence. The degree of influence describes the magnitude of one feature's impact on another. By analyzing the changes in each feature within the evolutionary difference features, and combining historical data and experience, the spatial correlation module can determine the direction and degree of influence between different features. For example, by comparing changes in equipment operating parameters with changes in material flow rate, the correlation between the two can be analyzed to determine the degree of influence of equipment operation adaptation features on material flow matching features.

[0057] Step S1334: Fill the influence direction and influence degree information into the feature association analysis table to generate a feature association influence matrix containing the influence relationship between the equipment operation adaptation features, material flow matching features and process execution coordination features.

[0058] The feature correlation influence matrix is ​​a two-dimensional matrix where rows and columns correspond to different features, and each element represents the influence relationship between one feature and another. The elements contain information about the direction and degree of influence. Using the feature correlation influence matrix, the mutual influence relationships between different features can be visually monitored. For example, an element in a specific row and column of the feature correlation influence matrix represents the direction and degree of influence of the equipment operation adaptation feature on the material flow matching feature.

[0059] Step S1335: Standardize the feature association influence matrix to generate a standardized feature association influence matrix.

[0060] The purpose of standardization is to ensure that the elements in the feature correlation matrix have the same scale and range, avoiding the impact of large differences in the numerical values ​​of the influence of different features on subsequent analysis and processing. Standardization can be achieved by transforming the elements in the feature correlation matrix, for example, mapping the values ​​of the elements to a set interval. After standardization, the standardized feature correlation matrix can more accurately reflect the mutual influence relationships between different features.

[0061] Step S1344: Based on the equipment dimension deviation distribution and equipment dimension fluctuation pattern, identify possible paths for the equipment operation adaptation feature to transition from the current state to subsequent states, and generate abnormal state transition paths for the equipment dimension by combining the normal state transition rules of the equipment dimension in the historical state feature sequence; based on the material dimension deviation distribution and material dimension fluctuation pattern, identify possible paths for the material flow matching feature to transition from the current state to subsequent states, and generate abnormal state transition paths for the material dimension by combining the normal state transition rules of the material dimension in the historical state feature sequence; based on the process dimension deviation distribution and process dimension fluctuation pattern, identify possible paths for the process execution coordination feature to transition from the current state to subsequent states, and generate abnormal state transition paths for the process dimension by combining the normal state transition rules of the process dimension in the historical state feature sequence.

[0062] Step S13441: Based on the device dimension deviation distribution and device dimension fluctuation pattern, identify possible paths for the device operation adaptation features to transition from the current state to subsequent states, and generate abnormal state transition paths for the device dimension by combining the normal state transition rules of the device dimension in the historical state feature sequence.

[0063] For example, step S13441-1: Extract the feature deviation value between the current state and the historical normal state from the device dimension deviation distribution, and mark the state node whose deviation value deviates from the normal range.

[0064] The anomaly prediction module extracts the deviation values ​​between the current equipment's operational adaptation features and the corresponding features under historical normal conditions from the equipment-level deviation distribution. For example, for the temperature parameters of a pellet mill, it calculates the deviation between the current temperature and the historical normal temperature. State nodes whose deviation values ​​exceed a preset normal range are marked. If the normal temperature range is a specific interval, and the current temperature deviates from that interval, the corresponding state node will be marked.

[0065] Step S13441-2: Extract the trend of feature deviation value changes over continuous time points from the device dimension fluctuation pattern, and mark the state nodes where abnormal fluctuation patterns occur.

[0066] Next, the trend of characteristic deviation values ​​over continuous time points is extracted from the equipment-level fluctuation patterns. The changes in the pellet mill temperature deviation value at multiple time points are analyzed to determine if any abnormal fluctuation patterns exist, such as sudden large increases or decreases. State nodes exhibiting abnormal fluctuation patterns are marked, as these state nodes may indicate abnormal equipment operation.

[0067] Step S13441-3: State nodes that deviate from the normal range and state nodes that exhibit abnormal fluctuation patterns are designated as abnormal state nodes at the device level.

[0068] The previously marked state nodes that deviate from the normal range and those exhibiting abnormal fluctuation patterns are combined to identify device-level abnormal state nodes. These device-level abnormal state nodes represent abnormal states in the device's operational adaptation characteristics.

[0069] Step S13441-4: Extract normal state transition paths of equipment dimensions in the same process stage from the historical state feature sequence, and establish a set of normal state transition paths.

[0070] Extract normal state transition paths for equipment at the same process stage (such as granulation) from historical state feature sequences. These normal state transition paths represent the transition relationships between state nodes during normal equipment operation. Organize these normal state transition paths into a set, which serves as a reference for determining abnormal state transition paths.

[0071] Step S13441-5: Compare the subsequent state transition paths of the abnormal state nodes in the device dimension with the set of normal state transition paths, and filter out the transition paths that do not appear in the set of normal state transition paths as the abnormal state transition paths in the device dimension.

[0072] The subsequent state transition paths of abnormal state nodes at the device level are compared with the set of normal state transition paths. If a subsequent state transition path of an abnormal state node does not appear in the set of normal state transition paths, it indicates that the transition path is abnormal, and it is filtered out as an abnormal state transition path at the device level.

[0073] Step S13442: Based on the material dimension deviation distribution and material dimension fluctuation pattern, identify possible paths for the material flow matching characteristics to transition from the current state to subsequent states, and generate abnormal state transition paths for material dimensions by combining the normal state transition rules of material dimensions in the historical state feature sequence.

[0074] Similar to the equipment dimension, for the material dimension, firstly, the deviation values ​​of material flow matching features are extracted from the material dimension deviation distribution, and state nodes that deviate from the normal range are marked. Then, the changing trends of feature deviation values ​​are extracted from the material dimension fluctuation patterns, and state nodes exhibiting abnormal fluctuation patterns are marked, thus identifying abnormal state nodes in the material dimension. Next, normal state transition paths for the material dimension at the same flow stage are extracted from historical state feature sequences, establishing a set of normal state transition paths. Finally, the subsequent state transition paths of abnormal state nodes in the material dimension are compared with the set of normal state transition paths, and abnormal transition paths are selected as abnormal state transition paths for the material dimension.

[0075] Step S13443: Based on the process dimension deviation distribution and process dimension fluctuation pattern, identify possible paths for the process execution coordination feature to transition from the current state to the subsequent state, and generate abnormal state transition paths for the process dimension by combining the normal state transition rules of the process dimension in the historical state feature sequence.

[0076] For the process dimension, deviation values ​​of process execution coordination features are extracted from the process dimension deviation distribution to mark state nodes that deviate from the normal range. The changing trends of feature deviation values ​​are extracted from the process dimension fluctuation patterns to mark state nodes exhibiting abnormal fluctuation patterns, thus identifying abnormal state nodes in the process dimension. Normal state transition paths for the same process stage are extracted from historical state feature sequences to establish a set of normal state transition paths. The subsequent state transition paths of abnormal state nodes in the process dimension are compared with the set of normal state transition paths to filter out abnormal transition paths, which are then used as abnormal state transition paths for the process dimension.

[0077] Step S1345: Through the anomaly inference module of the joint machine learning model, the abnormal state transfer paths of the equipment dimension, the abnormal state transfer paths of the material dimension, and the abnormal state transfer paths of the process dimension are correlated and verified with the set of influence relationships between features to determine the expansion direction of the abnormal state in each dimension under the influence constraints between features, and generate vector data of the development trend of abnormal states in the equipment dimension, material dimension, and process dimension.

[0078] For example, step S13451: Extract the direction and degree of influence of the equipment operation adaptation feature on the material flow matching feature in the set of influence relationships between features, match the abnormal state transfer path of the equipment dimension with the abnormal state transfer path of the material dimension, if the starting node of the abnormal state transfer path of the equipment dimension and the starting node of the abnormal state transfer path of the material dimension have a temporal correspondence and the degree of influence meets the set influence degree condition, then take the part of the abnormal state transfer path of the material dimension that corresponds to the abnormal state transfer path of the equipment dimension as the direction of the expansion of the abnormal state of the equipment dimension to the material dimension.

[0079] The anomaly inference module extracts the direction and degree of influence of equipment operation adaptation features on material flow matching features from the set of influence relationships between features. Then, it matches the equipment-dimensional anomaly state transition paths with the material-dimensional anomaly state transition paths. It checks whether there is a temporal correspondence between the starting nodes of the equipment-dimensional and material-dimensional anomaly state transition paths; for example, if an equipment anomaly quickly followed by a material flow anomaly, it determines whether the degree of influence meets the set influence degree conditions. If both are met, the portion of the material-dimensional anomaly state transition path corresponding to the equipment-dimensional anomaly state transition path is identified as the direction of the equipment-dimensional anomaly state's expansion into the material dimension.

[0080] Step S13452: Extract the influence direction and degree of the equipment operation adaptation feature on the process execution coordination feature in the feature influence relationship set, match the abnormal state transfer path of the equipment dimension with the abnormal state transfer path of the process dimension. If the starting node of the abnormal state transfer path of the equipment dimension and the starting node of the abnormal state transfer path of the process dimension have a temporal correspondence and the degree of influence meets the set influence degree condition, then the part of the abnormal state transfer path of the process dimension that corresponds to the abnormal state transfer path of the equipment dimension is taken as the direction of the abnormal state of the equipment dimension to the process dimension.

[0081] Similarly, the influence direction and degree of equipment operation adaptation features on process execution coordination features are extracted from the set of influence relationships between features. The abnormal state transition paths of the equipment dimension are matched with those of the abnormal state transition paths of the process dimension. The temporal correspondence and degree of influence between the starting nodes of the equipment dimension and process dimension abnormal state transition paths are determined to meet certain conditions. If they do, the portion of the process dimension abnormal state transition path corresponding to the equipment dimension abnormal state transition path is identified as the direction of the equipment dimension abnormal state's expansion into the process dimension.

[0082] Step S13453: Extract the influence direction and degree of the material flow matching feature on the process execution coordination feature in the set of influence relationships between features, match the material dimension abnormal state transfer path with the process dimension abnormal state transfer path. If the starting node of the material dimension abnormal state transfer path and the starting node of the process dimension abnormal state transfer path have a temporal correspondence and the degree of influence meets the set influence degree condition, then the part of the process dimension abnormal state transfer path that corresponds to the material dimension abnormal state transfer path is taken as the direction of the material dimension abnormal state to expand to the process dimension.

[0083] Similarly, the influence direction and degree of material flow matching features on process execution coordination features are extracted from the set of influence relationships between features. The abnormal state transfer paths of the material dimension are matched with those of the abnormal state transfer paths of the process dimension. The temporal correspondence and degree of influence between the starting nodes of the abnormal state transfer paths of the material dimension and the process dimension are checked to see if they meet the conditions. If they do, the portion of the abnormal state transfer path of the process dimension corresponding to the abnormal state transfer path of the material dimension is determined as the direction of the expansion of the abnormal state from the material dimension to the process dimension.

[0084] Step S13454: Based on the direction of the expansion of the abnormal states in each dimension to other dimensions, correct the path direction of the abnormal development trend vector data of the equipment dimension, the abnormal development trend vector data of the material dimension, and the abnormal development trend vector data of the process dimension.

[0085] Based on the previously identified directions of anomalous states expanding into other dimensions, the path directions of the anomaly development trend vector data for the equipment, material, and process dimensions are revised. This involves considering the directions of equipment-level anomalies expanding into the material and process dimensions, as well as the direction of material-level anomalies expanding into the process dimension, to ensure that the anomaly development trend vector data for each dimension more accurately reflects the expansion of the anomalous state.

[0086] Step S13455: Generate vector data on the development trend of abnormal states in the equipment, material, and process dimensions.

[0087] Following the preceding correlation verification and path correction, the final result is vector data on the development trends of abnormal states across the equipment, material, and process dimensions. This vector data contains information on the expansion direction and trend of abnormal states in each dimension.

[0088] Step S135: Generate a production line anomaly development simulation result based on the development trend vector data, including the anomaly initiation stage, expansion direction, and impact range.

[0089] For example, step S1351: Extract the starting node information of the abnormal state transfer path of the equipment dimension, the abnormal state transfer path of the material dimension, and the abnormal state transfer path of the process dimension from the development trend vector data. The starting node information includes the process stage identifier where the abnormal state transfer path of each dimension begins to deviate from the normal state transition path.

[0090] First, the starting node information for abnormal state transition paths at the equipment, material, and process dimensions is extracted from the development trend vector data. The starting node information includes the process stage identifier where the abnormal state transition path in each dimension begins to deviate from the normal state transition path. For example, an abnormal state transition path at the equipment dimension might begin to deviate from the normal state at the granulation process; therefore, the starting node information would record the identifier of the granulation process. By analyzing the starting node information, it is possible to determine at which process stage the abnormal state initially occurred.

[0091] Step S1352: Compare the order of occurrence of the starting nodes of the abnormal state transfer paths in the equipment dimension, material dimension, and process dimension on the time axis, filter out the starting node with the earliest time sequence, and determine the process stage corresponding to the starting node as the abnormal starting link.

[0092] Compare the order of the starting nodes of the abnormal state transfer paths on the timeline across the equipment, material, and process dimensions. Identify the earliest starting node in the time sequence; the process stage corresponding to this starting node is the abnormal initiation point. For example, if the starting node of the abnormal state transfer path in the equipment dimension is the granulation process, the starting node of the abnormal state transfer path in the material dimension is the cooling process, and the starting node of the abnormal state transfer path in the process dimension is the packaging process, and the granulation process is the earliest in the time sequence, then the granulation process is identified as the abnormal initiation point.

[0093] Step S1353: Analyze the extension direction of the abnormal state transfer path in the equipment dimension to obtain the extension direction of the equipment abnormal state from the starting node to the subsequent process stage or related features; analyze the extension direction of the abnormal state transfer path in the material dimension to obtain the extension direction of the material abnormal state from the starting node to the subsequent process stage or related features; analyze the extension direction of the abnormal state transfer path in the process dimension to obtain the extension direction of the process abnormal state from the starting node to the subsequent process stage or related features; combine the extension directions of the abnormal state transfer paths in the equipment, material, and process dimensions to determine the mutual expansion direction of the abnormal state between equipment and materials, equipment and processes, and materials and processes.

[0094] Analyzing the extension direction of abnormal state transfer paths at the equipment level helps understand how equipment abnormalities extend from their starting point to subsequent process stages or related features. For example, an equipment abnormality might begin with a malfunction in the granulation process, affecting the operating parameters of the equipment in the cooling process, or impacting material flow matching characteristics and process execution coordination characteristics. Similarly, analyzing the extension direction of abnormal state transfer paths at the material level determines how material abnormalities extend from their starting point to subsequent process stages or related features. Analyzing the extension direction of abnormal state transfer paths at the process level determines how process abnormalities extend from their starting point to subsequent process stages or related features.

[0095] By considering the extension directions of abnormal status transfer paths across equipment, materials, and processes, the mutual expansion directions of abnormal statuses between equipment and materials, equipment and processes, and materials and processes can be determined. For example, abnormal equipment status may lead to changes in material flow paths, thereby affecting the execution of process instructions; abnormal material status may affect the operating status of equipment and the execution of processes; and abnormal process status may also affect equipment and materials.

[0096] Step S1354: Count the number of process stages covered by the abnormal state transfer path in the equipment dimension and the number of associated material flow matching features and process execution coordination features to form equipment dimension impact range information; count the number of process stages covered by the abnormal state transfer path in the material dimension and the number of associated equipment operation adaptation features and process execution coordination features to form material dimension impact range information; count the number of process stages covered by the abnormal state transfer path in the process dimension and the number of associated equipment operation adaptation features and material flow matching features to form process dimension impact range information; combine the impact range information of equipment, material, and process dimensions to determine the set of process stages and feature association set involved in the abnormal state in the entire production line process, as the impact range.

[0097] The system analyzes the number of process stages covered by abnormal state transfer paths at the equipment level, as well as the number of associated material flow matching features and process execution coordination features. For example, an abnormal state transfer path at the equipment level might cover the granulation and cooling processes, and be associated with features such as material flow speed and the timeliness of process command response, thus forming information on the scope of influence at the equipment level.

[0098] Similarly, by statistically analyzing the number of process stages covered by abnormal state transfer paths at the material level, as well as the number of associated equipment operation adaptation characteristics and process execution coordination characteristics, information on the scope of influence at the material level is generated. Similarly, by statistically analyzing the number of process stages covered by abnormal state transfer paths at the process level, as well as the number of associated equipment operation adaptation characteristics and material flow matching characteristics, information on the scope of influence at the process level is generated.

[0099] By integrating information on the impact of equipment, materials, and processes, the set of process stages and characteristic associations involved in the abnormal state throughout the entire production line are determined, which serves as the scope of impact. This scope of impact information allows staff to clearly understand the degree and extent of the abnormal state's influence on the production line.

[0100] Step S1355: Integrate the process stage identifier of the abnormal initiation link, the extension direction information, the process stage set of the affected scope, and the feature association set to generate a production line abnormal development simulation result containing the abnormal initiation link, the extension direction, and the affected scope.

[0101] Finally, the process stage identifier of the anomaly initiation point, the extension direction information, the set of process stages affecting the scope of influence, and the set of feature associations are integrated to generate the production line anomaly development simulation results. The production line anomaly development simulation results comprehensively describe the initiation, development direction, and scope of influence of the production line anomaly.

[0102] Step S140: Identify key abnormal links in the production line based on the deduction results of the abnormal development of the production line. The key abnormal links include equipment performance degradation, material quality deterioration, and process connection misalignment.

[0103] Step S141: Analyze the abnormal initiation information in the production line abnormal development simulation results to determine the process stage in which the abnormal state first appears.

[0104] The results of the production line anomaly development simulation are analyzed to extract information on the anomaly initiation stage. This information includes the process stage identifier where the anomaly first appeared. This process stage identifier allows us to determine which process stage in the production line the anomaly first occurred in. For example, if the anomaly initiation stage information indicates granulation, then we can know that the anomaly initially started in the granulation process.

[0105] Step S142: Extract the equipment operation adaptation features, material flow matching features, and process execution coordination features corresponding to the abnormal initiation link.

[0106] After identifying the initiating point of the anomaly, it is necessary to extract the corresponding equipment operation adaptation characteristics, material flow matching characteristics, and process execution coordination characteristics. If the initiating point of the anomaly is the granulation process, the equipment operation adaptation characteristics should include detailed information on all operating parameters of the granulator in that process. In addition to the previously mentioned temperature, pressure, and speed, this also includes the granulator's motor power and the wear condition of the die. Temperature parameters reflect whether the granulator's heating system is working properly; pressure parameters relate to the density and molding effect of the granules; speed affects the uniformity of granulation; changes in motor power may indicate potential equipment malfunctions; and the wear condition of the die directly affects the shape and quality of the granules.

[0107] Regarding material flow matching characteristics, in addition to the material entry speed, attention should also be paid to the material residence time in the granulation process and the uniformity of material distribution. An excessively fast or slow material entry speed can lead to poor granulation results, an inappropriate residence time will affect the maturity of the granules, and uneven material distribution will result in inconsistent granule sizes.

[0108] In terms of process execution coordination, in addition to focusing on whether the granulation specifications and output meet the preset instructions, it is also necessary to consider whether the operation steps in the granulation process are standardized, such as the order of feeding materials and the timing of additive addition.

[0109] Step S143: Analyze the intra-process parameter stability description and inter-process parameter connectivity description of the equipment operation adaptation characteristics to determine whether there are any signs of equipment performance degradation.

[0110] For example, step S1431: obtain the stability description of the process parameters of the equipment operation adaptation characteristics, and check whether there are frequent fluctuations or exceed the normal range of parameters.

[0111] First, obtain a description of the stability of the in-process parameters for equipment operation adaptation characteristics. In the granulation process, conduct a detailed inspection of all parameters of the granulator. For temperature parameters, observe their fluctuations during granulation. Under normal circumstances, the temperature should fluctuate within a relatively stable range. If the temperature fluctuates frequently and significantly, or exceeds the preset normal range, this may be a sign of a problem with the equipment's heating system. The same applies to pressure parameters; unstable pressure may lead to inconsistent granule density, affecting product quality. Fluctuations in rotational speed may cause uneven granule size. By monitoring these parameters in real time and comparing them with historical data, determine if there are any instances of frequent parameter fluctuations or deviations from the normal range.

[0112] Step S1432: Obtain the inter-process parameter connectivity description of the equipment operation adaptation characteristics, and check whether there are sudden changes or mismatches in the equipment parameters of adjacent processes.

[0113] Next, obtain a description of the inter-process parameter connectivity for equipment operation adaptation characteristics. Taking the pelleting process to the cooling process as an example, check the connectivity between the pellet mill's discharge temperature and the cooler's feed temperature. Under normal circumstances, there should be a reasonable transition relationship between these two temperatures. If the pellet mill's discharge temperature suddenly rises while the cooler's feed temperature does not adjust accordingly, or if the temperature difference between the two is too large, it indicates a sudden change or mismatch in the equipment parameters of adjacent processes. Similarly, the parameter connectivity between the pellet mill and the packaging machine, such as whether the pelleting output and packaging speed match, also needs to be checked.

[0114] Step S1433: Extract historical equipment maintenance records and find historical cases similar to the current parameter fluctuations and connection anomalies.

[0115] Extract historical equipment maintenance records and search through a large amount of historical data for cases similar to the current equipment parameter fluctuations and connection anomalies. The pellet mill previously experienced frequent temperature fluctuations, which were found to be caused by a damaged heating element during maintenance. Compare the current parameter fluctuations with these historical cases in detail to analyze whether similar symptoms and causes exist.

[0116] Step S1434: Analyze the processing results of the historical cases to determine whether parameter fluctuations and connection anomalies are caused by equipment performance degradation.

[0117] Analyze the results of these historical cases to determine whether the current parameter fluctuations and connection anomalies are caused by equipment performance degradation. If similar parameter fluctuations and connection anomalies in historical cases were caused by equipment aging, component wear, or other reasons leading to decreased equipment performance, and the current equipment also has a similar service life and operating environment, then there is a high probability that the current anomalies are also caused by equipment performance degradation.

[0118] Step S1435: If there are frequent fluctuations in parameters, exceeding the normal range, or sudden mismatches in parameters of adjacent processes, and historical cases show that the situation is caused by equipment performance degradation, then it is determined that there are signs of equipment performance degradation.

[0119] If, after the above checks, frequent parameter fluctuations, deviations from normal ranges, or sudden mismatches in parameters between adjacent processes are found, and historical cases indicate that these situations are caused by equipment performance degradation, then it can be determined that the equipment shows signs of performance degradation. For example, if the temperature of the pellet mill frequently exceeds the normal range, and similar situations have historically been caused by uneven heating due to aging heating elements, then it can be determined that the pellet mill may have a performance degradation problem.

[0120] Step S144: Analyze the path direction matching degree description and path speed balance degree description of the material flow matching characteristics to determine whether there are signs of material quality deterioration.

[0121] Step S1441: Check the path direction matching degree description of the material flow matching feature to determine whether the material flow path deviates from the standard path.

[0122] Check the path direction matching description of material flow characteristics. Materials should flow from the raw material silo to the granulation process along a preset standard path. Analyze the material flow trajectory data to determine if the actual flow path matches the standard path. If the material deviates from the standard path during flow, it may be subjected to unnecessary compression or collisions, affecting material quality. For example, if material should be transported through the target pipeline from the raw material silo to the granulator but actually enters another bypass pipeline, it may become contaminated or mix with other impurities, leading to deterioration in material quality.

[0123] Step S1442: Check the path speed balance description of the material flow matching feature to determine whether there is a sudden change or imbalance in the material flow speed.

[0124] Examine the path speed uniformity description of the material flow matching characteristics. During material flow, the speed should remain relatively stable. Sudden increases or decreases in material flow speed, or significant speed differences between different process stages, can affect material quality. In the granulation process, a sudden increase in the speed at which material enters the granulator may lead to uneven granulation, affecting product quality. Speed ​​imbalances can also cause material to accumulate in certain areas, leading to localized temperature increases, which can destroy nutrients in the material and cause quality deterioration.

[0125] Step S1443: Based on the physical and chemical properties of the material, determine whether abnormal path direction and speed will cause changes in material quality.

[0126] By combining the physical and chemical properties of the material, we can further determine whether abnormalities in path direction and velocity will lead to changes in material quality. Some materials are sensitive to factors such as temperature and pressure. If the material flow path deviates from the standard path, it may expose the material to different environmental conditions, thereby affecting its physical and chemical properties. Sudden changes in material flow velocity may alter the shear forces acting on the material, leading to the destruction of its particle structure. Through analysis of material characteristics and comprehensive consideration of abnormal path direction and velocity, we can determine whether there is a possibility of material quality deterioration.

[0127] Step S1444: If the path direction deviates from the standard path or there is a sudden change or imbalance in speed, and it is determined that the material characteristics will lead to a change in quality, then it is determined that there are signs of deterioration in material quality.

[0128] If, after inspection and monitoring, the material flow path deviates from the standard path, or there are sudden changes or imbalances in speed, and based on the material's characteristics, it is determined that these abnormalities will lead to changes in material quality, then it can be determined that there are signs of material quality deterioration. For example, if a certain material is temperature-sensitive, and its flow path deviates from the standard path, passing through a high-temperature area, and the flow speed suddenly slows down, causing the material to remain in the high-temperature environment for too long, then based on a comprehensive assessment of these circumstances, it can be determined that there are signs of material quality deterioration.

[0129] Step S145: Analyze the instruction response timeliness description and action execution integrity description of the process execution coordination characteristics to determine whether there are any signs of process connection misalignment.

[0130] Step S1451: Obtain the instruction response timeliness description of the process execution coordination characteristics, and check whether the equipment or operators respond to the process instructions in a timely manner.

[0131] Obtain a description of the timeliness of instruction response to process execution coordination characteristics, and check the response of equipment or operators to process instructions. In the granulation process, after a process instruction is issued, the granulator should begin executing the instruction within the specified time. Check the instruction issuance time and the actual start time of the equipment recorded by the control system. If the time difference between the two exceeds a reasonable range, it indicates that the equipment or operators have not responded to the process instruction in a timely manner. This may be due to equipment malfunction, operator unfamiliarity, or poor information transmission.

[0132] Step S1452: Obtain the action execution integrity description of the process execution coordination feature, and check whether the actual executed actions fully meet the requirements of the preset instructions.

[0133] Obtain a complete description of the action execution characteristics of the process execution coordination features, and check whether the actual executed actions fully comply with the requirements of the preset instructions. In the granulation process, preset instructions may require that the granulation specifications, output, and operating procedures meet set standards. Check the actual granulation situation, including whether the granulation size, density, and output are consistent with the instruction requirements, and whether the operating procedures are performed in the prescribed order and method. If there are deviations in the actual executed actions, such as the granulation specifications not meeting the requirements, or certain key steps being omitted in the operating procedures, this indicates that the actual executed actions have not fully complied with the requirements of the preset instructions.

[0134] Step S1453: Analyze the execution of process instructions in adjacent processes to determine whether there are any instruction conflicts or poor connections.

[0135] Analyze the execution of process instructions in adjacent processes to determine if there are any conflicts or misalignments. The process instructions for the granulation and cooling processes may be interrelated. If the output instruction for the granulation process does not match the cooling capacity of the cooling process, or if the operation time of the granulation process is not coordinated with the start time of the cooling process, this can lead to conflicts or misalignments between the process instructions of adjacent processes. A detailed analysis of the execution of process instructions in adjacent processes can help identify potential problems.

[0136] Step S1454: If the instruction response is not timely, the action is not fully executed, or there are instruction conflicts or poor connections between adjacent processes, it is determined that there are signs of process connection misalignment.

[0137] If an inspection reveals that equipment or operator commands are not responding promptly, actions are not being performed completely, or there are conflicting or poorly coordinated commands between adjacent processes, then it can be determined that there are signs of misalignment in process coordination. For example, if the granulator does not respond to process commands to start granulation in a timely manner, the granules do not meet the required specifications, and the commands between the granulation and cooling processes are not well coordinated, resulting in untimely cooling, these combined situations indicate a misalignment in process coordination.

[0138] Step S146: Based on the judgment results, select the links that have the greatest impact on the abnormal development as key abnormal links.

[0139] Based on the judgment results of the signs of equipment performance degradation, material quality deterioration, and process connection misalignment, screen out the link that has the greatest impact on the abnormal development as the key abnormal link. If the parameter abnormality caused by equipment performance degradation has seriously affected the quality of pelletizing and further affected the normal operation of subsequent processes, while the impacts of material quality deterioration and process connection misalignment are relatively small, then the equipment performance degradation link can be determined as the key abnormal link. By evaluating the impact degree of each link, accurately identify the key abnormal link.

[0140] Step S150: Generate a production line control instruction combination based on the key abnormal link, and send the control instruction combination to the production line execution unit to trigger targeted control operations.

[0141] Step S151: Query the preset production line exception handling rule library, and extract the control strategy template that matches the type of the key abnormal link.

[0142] If the key abnormal link is an equipment performance degradation link, extract the control strategy template related to equipment performance degradation from the rule library. These templates contain a series of framework of control measures that may be taken for equipment performance degradation, such as basic strategies for equipment maintenance and adjustment of operating parameters.

[0143] Step S152: Select the applicable control action type from the control strategy template according to the specific performance characteristics of the key abnormal link.

[0144] Select the applicable control action type from the control strategy template according to the specific performance characteristics of the key abnormal link. For the equipment performance degradation link, if the specific performance is that the temperature of the pelletizer is unstable, then select the action type related to temperature control from the template, such as checking the heating element and adjusting the heating power. If it is a material quality deterioration link, if the performance is abnormal material flow path, select action types such as adjusting the material flow path and checking the conveying equipment. For the process connection misalignment link, if the performance is conflict between adjacent process instructions, select action types such as coordinating process instructions and optimizing the operation process.

[0145] Step S153: Configure specific execution parameters for the control action type, and the execution parameters include the action object, execution timing, and action method of the control action.

[0146] Step S1531: Determine the action object of the control action type. If the key abnormal link is an equipment performance degradation link, the action object is the corresponding equipment component; if it is a material quality deterioration link, the action object is the corresponding material flow path; if it is a process connection misalignment link, the action object is the corresponding process instruction node.

[0147] Determine the target of the control action. When the critical abnormality is due to equipment performance degradation, if the problem lies with the pellet mill's heating element, then the target is the pellet mill's heating element. If it's due to material quality deterioration, and the material flow path deviates from the standard path due to a blockage in a conveying pipe, then the target is that conveying pipe. For process connection misalignment, if conflicts between adjacent process instructions are due to unreasonable instruction timing settings for the pelletizing and cooling processes, then the target is the process instruction nodes of those two processes.

[0148] Step S1532: Determine the timing of the control action. Based on the abnormal expansion direction information in the production line abnormal development simulation results, select the time point before the abnormal expands to the next stage as the timing of the action.

[0149] Determine the timing of control actions based on the anomaly propagation direction information in the production line anomaly development simulation. If the anomaly propagation direction indicates that equipment performance degradation is about to affect material quality, then choose a time point before this impact occurs to execute the control action. For example, an abnormal rise in pellet mill temperature may cause quality problems in subsequent processes; therefore, adjust the pellet mill before the temperature rises to a critical value that could affect material quality.

[0150] Step S1533: Determine the mode of action of the control action. If it is a process of equipment performance degradation, the mode of action is to adjust the equipment operating parameters or start the maintenance program; if it is a process of material quality deterioration, the mode of action is to correct the material flow path or adjust the flow speed; if it is a process connection misalignment, the mode of action is to calibrate the process command or optimize the command response process.

[0151] Determine the mode of action for the control measures. For equipment performance degradation, if the pellet mill temperature is unstable, the temperature can be stabilized by adjusting the heating power, or the equipment maintenance program can be initiated to inspect the heating elements. For material quality deterioration, if the material flow path is abnormal, the material flow path can be corrected by controlling the valves of the conveyor equipment, or the flow speed can be adjusted by changing the speed of the conveyor equipment. For process misalignment, if adjacent process instructions conflict, the timing and content of the process instructions can be recalibrated, or the instruction response process can be optimized to improve the efficiency of information transmission.

[0152] Step S1534: Combine the target object, execution timing and execution mode to generate specific execution parameters.

[0153] By combining the target object, timing of execution, and mode of action, specific execution parameters are generated. For situations where the pellet mill temperature is unstable, the execution parameters could include checking and adjusting the heating power of the pellet mill's heating elements before the temperature reaches a certain critical value. These specific execution parameters clarify the detailed implementation of the control actions.

[0154] Step S154: Combine the control action type and execution parameters to generate a control sub-instruction for the critical abnormal link.

[0155] By combining the control action type and specific execution parameters, control sub-instructions are generated for critical abnormal processes. For equipment performance degradation, a control sub-instruction might be to check the heating element of the pellet mill before a specific time point and adjust the heating power based on the check results. These sub-instructions detail the specific control operations for critical abnormal processes.

[0156] Step S155: Prioritize the control sub-instructions corresponding to different key abnormal links to generate a combination of production line control instructions with execution order.

[0157] If multiple critical anomalies exist, the control sub-instructions corresponding to different critical anomalies are prioritized. The priorities are based on the impact and urgency of each critical anomaly on the production line, derived from the anomaly development simulation. If equipment performance degradation has the most severe and urgent impact on the production line, its corresponding control sub-instruction is prioritized. This prioritization generates a combination of production line control instructions with an execution order, ensuring that control operations are performed sequentially according to importance and urgency.

[0158] Step S156: The production line control command combination is sent to the corresponding production line execution unit through the production line communication network. The production line execution unit performs the control operation according to the priority and execution parameters of the command.

[0159] Production line control commands are combined and sent to the corresponding production line execution units via the production line communication network. These execution units include equipment control systems, material conveying equipment controllers, and operator terminals. Upon receiving the command combinations, the production line execution units perform control operations according to the command priority and execution parameters. Operators maintain and adjust equipment parameters as required by the commands, the equipment control system automatically executes the corresponding control actions, and the material conveying equipment controller adjusts the material flow path and speed. Through these operations, targeted control of critical abnormalities in the production line is achieved, ensuring the normal operation of the production line.

[0160] Figure 2The illustration shows exemplary hardware and software components of a feed production line monitoring and control system 100 based on a joint machine learning model, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the feed production line monitoring and control system 100 based on a joint machine learning model and to perform the functions in this application.

[0161] The feed production line monitoring and control system 100 based on a joint machine learning model can be a general-purpose server or a special-purpose server; both can be used to implement the feed production line monitoring and control method based on a joint machine learning model of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0162] For example, a feed production line monitoring and control system 100 based on a joint machine learning model may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the feed production line monitoring and control system 100 based on a joint machine learning model may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The feed production line monitoring and control system 100 based on a joint machine learning model also includes an I / O interface 150 between the computer and other input / output devices.

[0163] For ease of explanation, only one processor is described in the feed production line monitoring and control system 100 based on a joint machine learning model. However, it should be noted that the feed production line monitoring and control system 100 based on a joint machine learning model in this application may also include multiple processors. Therefore, the steps executed by one processor as described in this application may also be executed jointly by multiple processors or individually. For example, if the processor of the feed production line monitoring and control system 100 based on a joint machine learning model executes steps A and B, it should be understood that steps A and B may also be executed jointly by two different processors or individually by one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0164] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned feed production line monitoring and control method based on a joint machine learning model is implemented.

[0165] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for monitoring and controlling feed production lines based on a joint machine learning model, characterized in that, The method includes: Acquire a process data set of the feed production line, the process data set including equipment operating status data, material flow trajectory data and process instruction execution data; The process data set is subjected to feature co-coding to generate a production line status feature set, which includes equipment operation adaptation features, material flow matching features, and process execution coordination features. The pre-built joint machine learning model is invoked to perform abnormal development inference processing on the set of production line state features, and the abnormal development inference results of the production line are generated. Based on the results of the production line anomaly development simulation, key abnormal links in the production line are identified. These key abnormal links include equipment performance degradation, material quality deterioration, and process connection misalignment. Based on the aforementioned critical abnormal process, a combination of production line control instructions is generated, and the combination of control instructions is sent to the production line execution unit to trigger targeted control operations.

2. The feed production line monitoring and control method based on a joint machine learning model according to claim 1, characterized in that, The step of performing feature co-coding on the process data set to generate a production line status feature set includes: The equipment operating status data is processed for operating mode recognition, and the operating parameter matching features of the equipment in typical processes are extracted as equipment operating adaptation features. The equipment operating adaptation features include intra-process parameter stability description and inter-process parameter connectivity description. The typical processes include granulation process, cooling process, and packaging process. The material flow trajectory data is subjected to path consistency analysis to extract the matching characteristics of the material flow path from the raw material warehouse to the finished product warehouse with the standard path as the material flow matching characteristics. The material flow matching characteristics include a description of path direction matching degree and a description of path speed balance. The process instruction execution data is subjected to instruction response analysis and processing. The matching features between the actual executed actions and the preset instructions are extracted as process execution coordination features. The process execution coordination features include a description of the timeliness of instruction response and a description of the completeness of action execution. Verify the logical correlation between the equipment operation adaptation features, material flow matching features, and process execution coordination features, and eliminate redundant features whose logical correlation does not meet the set correlation conditions; The verified equipment operation adaptation features, material flow matching features, and process execution coordination features are standardized and spliced ​​together to generate a set of production line status features.

3. The feed production line monitoring and control method based on a joint machine learning model according to claim 2, characterized in that, The step of calling a pre-built joint machine learning model to perform anomaly prediction processing on the production line state feature set, generating anomaly prediction results for the production line, includes: The production line status feature set is input into the historical status memory module of the joint machine learning model to extract the historical status feature sequence of the production line's historical normal operation phase. The historical state feature sequence and the current production line state feature set are input into the time series extrapolation module of the joint machine learning model to calculate the evolutionary difference features between the current state and the historical normal state. The evolutionary difference features are input into the spatial correlation module of the joint machine learning model to analyze the mutual influence relationship between equipment operation adaptation features, material flow matching features and process execution coordination features, and generate a feature correlation influence matrix. The evolutionary difference features and feature association influence matrix are input into the anomaly inference module of the joint machine learning model, and the state transition analysis algorithm is used to predict the development trend vector data of the abnormal state in the equipment dimension, material dimension, and process dimension. Based on the aforementioned development trend vector data, a production line anomaly development projection result is generated, including the anomaly initiation stage, expansion direction, and scope of impact.

4. The feed production line monitoring and control method based on a joint machine learning model according to claim 3, characterized in that, The step of inputting the historical state feature sequence and the current production line state feature set into the time series extrapolation module of the joint machine learning model to calculate the evolutionary difference features between the current state and the historical normal state includes: Arrange the historical state feature sequence in chronological order to construct a historical state timeline; Map the current production line status feature set to the historical status time axis to determine the corresponding position of the current status on the time axis; The time-series extrapolation module of the joint machine learning model extracts the state features of the same process stage in the timeline of the current state and the historical state, and calculates the feature difference value between the state features. Analyze the changing trend of the feature difference value at continuous time points to identify abnormal fluctuation patterns of state features; Based on the feature difference value and the abnormal fluctuation pattern, an evolutionary difference feature between the current state and the historical normal state is constructed. The evolutionary difference feature includes feature deviation distribution information and abnormal fluctuation pattern information.

5. The feed production line monitoring and control method based on a joint machine learning model according to claim 3, characterized in that, The step involves inputting the evolutionary difference features into the spatial correlation module of the joint machine learning model to analyze the mutual influence relationships among equipment operation adaptation features, material flow matching features, and process execution coordination features, generating a feature correlation influence matrix, including: Assign feature identifiers to the equipment operation adaptation features, material flow matching features, and process execution coordination features respectively; Construct a feature association analysis table, which includes records of the impact of equipment operation adaptation features on material flow matching features, records of the impact of equipment operation adaptation features on process execution coordination features, and records of the impact of material flow matching features on process execution coordination features. The spatial association module of the joint machine learning model extracts information on the direction and degree of influence among the equipment operation adaptation features, material flow matching features, and process execution coordination features from the evolutionary difference features. The influence direction and degree information are filled into the feature association analysis table to generate a feature association influence matrix containing the influence relationship between the equipment operation adaptation feature, material flow matching feature and process execution coordination feature. The feature association influence matrix is ​​standardized to generate a standardized feature association influence matrix.

6. The feed production line monitoring and control method based on a joint machine learning model according to claim 3, characterized in that, The step of inputting the evolutionary difference features and feature correlation influence matrix into the anomaly inference module of the joint machine learning model, and predicting the development trend vector data of the abnormal state in the equipment dimension, material dimension, and process dimension through the state transition analysis algorithm, includes: Extract the feature deviation distribution information from the evolutionary difference features, and construct the equipment dimension deviation distribution, material dimension deviation distribution, and process dimension deviation distribution according to the classification criteria of equipment operation adaptation features, material flow matching features, and process execution coordination features; Extract the abnormal fluctuation pattern information from the evolutionary difference features, and construct equipment-dimensional fluctuation patterns, material-dimensional fluctuation patterns, and process-dimensional fluctuation patterns according to the classification criteria of equipment operation adaptation features, material flow matching features, and process execution coordination features. Extract the influence direction and degree of equipment operation adaptation features on material flow matching features, the influence direction and degree of equipment operation adaptation features on process execution coordination features, and the influence direction and degree of material flow matching features on process execution coordination features from the feature association influence matrix to form a set of influence relationships between features; Based on the equipment dimension deviation distribution and equipment dimension fluctuation pattern, possible paths for equipment operation adaptation features to transition from the current state to subsequent states are identified. Combined with the normal state transition rules for equipment dimensions in the historical state feature sequence, abnormal state transition paths for equipment dimensions are generated. Similarly, based on the material dimension deviation distribution and material dimension fluctuation pattern, possible paths for material flow matching features to transition from the current state to subsequent states are identified. Combined with the normal state transition rules for material dimensions in the historical state feature sequence, abnormal state transition paths for material dimensions are generated. Finally, based on the process dimension deviation distribution and process dimension fluctuation pattern, possible paths for process execution coordination features to transition from the current state to subsequent states are identified. Combined with the normal state transition rules for process dimensions in the historical state feature sequence, abnormal state transition paths for process dimensions are generated. The anomaly inference module of the joint machine learning model is used to verify the correlation between the abnormal state transfer paths of the equipment dimension, the abnormal state transfer paths of the material dimension, and the abnormal state transfer paths of the process dimension and the set of influence relationships between features. This determines the expansion direction of the abnormal state in each dimension under the influence constraints between features, and generates vector data of the development trend of abnormal states in the equipment dimension, material dimension, and process dimension.

7. The feed production line monitoring and control method based on a joint machine learning model according to claim 1, characterized in that, The process of identifying key abnormal links in the production line based on the deduction results of the abnormal development of the production line includes: Analyze the information on the abnormal initiation stage in the production line abnormal development simulation results to determine the process stage in which the abnormal state first appears; Extract the equipment operation adaptation features, material flow matching features, and process execution coordination features corresponding to the abnormal initiation link; Analyze the stability description of intra-process parameters and the inter-process parameter connectivity description of the equipment's operational adaptation characteristics to determine whether there are signs of equipment performance degradation. Analyze the path direction matching degree description and path speed balance degree description of the material flow matching characteristics to determine whether there are signs of material quality deterioration; Analyze the timeliness of instruction response and the completeness of action execution of the process execution coordination characteristics to determine whether there are any signs of process connection misalignment; Based on the judgment results, the links that have the greatest impact on the abnormal development are selected as key abnormal links.

8. The feed production line monitoring and control method based on a joint machine learning model according to claim 1, characterized in that, The process of generating a production line control instruction combination based on the critical abnormal process and sending the control instruction combination to the production line execution unit to trigger targeted control operations includes: Query the preset production line anomaly handling rule library and extract the control strategy template that matches the type of the key anomaly link; Based on the specific characteristics of the key abnormal links, select the appropriate type of control action from the control strategy template; Configure specific execution parameters for the control action type, wherein the execution parameters include the target of the control action, the timing of execution, and the mode of action; The control action type and execution parameters are combined to generate control sub-instructions for key abnormal links; Prioritize the control sub-instructions corresponding to different key abnormal links to generate a combination of production line control instructions with execution order; The production line control commands are combined and sent to the corresponding production line execution units through the production line communication network. The production line execution units then perform control operations according to the priority and execution parameters of the commands.

9. The feed production line monitoring and control method based on a joint machine learning model according to claim 8, characterized in that, The specific execution parameters for configuring the control action type are specified, and the execution parameters include the target object, execution timing, and action mode of the control action, including: The target of the control action type is determined as follows: if the critical abnormal link is the equipment performance degradation link, the target is the corresponding equipment component; if it is the material quality deterioration link, the target is the corresponding material flow path; if it is the process connection misalignment link, the target is the corresponding process instruction node. Determine the timing of the control action. Based on the abnormal expansion direction information in the production line abnormal development simulation results, select the time point before the abnormal expands to the next stage as the execution timing. The mode of action of the control action is determined. If it is a stage of equipment performance degradation, the mode of action is to adjust the equipment operating parameters or start the maintenance program; if it is a stage of material quality deterioration, the mode of action is to correct the material flow path or adjust the flow speed; if it is a stage of process connection misalignment, the mode of action is to calibrate the process command or optimize the command response process. The target object, execution timing, and execution method are combined to generate specific execution parameters.

10. A feed production line monitoring and control system based on a joint machine learning model, characterized in that, The method includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the feed production line monitoring and control method based on a joint machine learning model as described in any one of claims 1-9.

Images