A grain winnowing parameter optimization method and system based on digital twinning

Abstract

This invention relates to the field of data processing, specifically to a method and system for optimizing grain air separation parameters based on digital twins. First, a dynamic production data graph model is constructed, including entity nodes such as grain batches, air separation equipment, raw grain sources, and screen components, as well as edges representing air separation processing activities. When substandard air separation performance is detected, the system calculates the suspicion weight of parameter adjustment for each processing stage based on multiple dimensions. Finally, using a best-first search algorithm, starting from the substandard grain batch node on the graph model, the system traces backward along the path with the highest suspicion, thereby quickly and accurately locating the root cause combination leading to poor air separation performance, providing a basis for subsequent parameter optimization adjustments. This invention achieves intelligent air separation parameter optimization and fault diagnosis by constructing a graph model of the air separation process and calculating and integrating the suspicion weights of historical data, real-time parameters, and process criticality.

Description

Technical Field

[0001] This invention relates to the field of data processing, and specifically to a method and system for optimizing grain air separation parameters based on digital twins. Background Technology

[0002] Air separation, a key technology in grain processing that separates materials and impurities based on differences in suspension velocity and aerodynamic characteristics, is widely used in several core stages of grain processing, including initial cleaning upon receiving raw grains, destoning by gravity, and fine grading. Modern grain air separation processing lines are often complex systems integrating lifting, screening, air separation, and dust removal processes, involving a wide variety of equipment and generating massive amounts of highly coupled process data such as airflow, air pressure, screen vibration frequency, and raw grain flow rate. To ensure the purity of the finished grain, reduce breakage rates, and strictly control loss rates, when online monitoring detects quality deviations such as excessive impurity content, poor destoning, or unclear grading at the discharge end, in-depth data traceability of the entire air separation process is essential to quickly and accurately pinpoint the root cause of air separation failure, thereby guiding real-time optimization and adjustment of process parameters. Therefore, establishing an efficient and intelligent grain air separation parameter optimization and quality traceability system is crucial for improving the manufacturing level and intelligent management capabilities of the grain and oil machinery industry.

[0003] In existing technologies, with the development of the Industrial Internet and digital twin technologies, graph models have become an effective way to organize and associate data. They can clearly show the complex topological relationships between wind-separating equipment, raw grain batches, operators, and processing parameters. However, traditional breadth-first search or depth-first search algorithms have significant shortcomings when faced with the massive and time-varying digital twin graphs of grain processing. These algorithms are non-heuristic blind searches; they treat all possible processing paths and parameter nodes equally during traversal, failing to distinguish the weight differences in the impact of different wind-separating processes on the final grain quality, and failing to effectively utilize the valuable experience contained in historical equipment operation and maintenance data and past parameter misalignment cases. Therefore, they often explore a large number of paths unrelated to the current wind-separating anomalies, resulting in low traceability efficiency, inaccurate problem localization, and difficulty in quickly focusing on the root causes of high risks in massive real-time sensor data, posing a huge challenge to parameter optimization and quality analysis. Summary of the Invention

[0004] To address the problem that conventional graph search algorithms struggle to quickly pinpoint high-risk root causes in massive datasets, this invention proposes a method for optimizing grain wind-sorting parameters based on digital twins. The method includes: constructing a wind-sorting processing data graph model, where nodes represent wind-sorting processing entities and directed edges represent wind-sorting processing activities; acquiring wind-sorting anomaly characteristics of the target grain batch; determining historical correlation indicators for each node based on the similarity between the anomaly characteristics and historical anomaly events corresponding to each node; calculating parameter deviation correction coefficients based on the deviation of actual values ​​of key process parameters in the corresponding wind-sorting processing activities from their set specification center values; using these parameter deviation correction coefficients to weight and amplify the historical correlation indicators to obtain path instability indicators for each edge; multiplying the path instability indicators by the preset process inherent coefficients of the corresponding wind-sorting processing activities to obtain the suspicion weight of each edge; and tracing back along the reverse path of the graph model, starting from the node corresponding to the target grain batch, prioritizing the expansion of tracing paths with higher cumulative suspicion weights.

[0005] Compared to existing technologies that rely on trial and error based on human experience, post-event statistical analysis, or simple database queries to adjust wind separation parameters, this invention constructs a wind separation processing data graph model to structure and correlate discrete grain processing information, and creatively proposes a suspicion weighting. This method not only considers the recurrence of historical problems with specific equipment or grain sources, but also dynamically reflects the stability of current process parameters such as airflow, air pressure, and vibration. This makes parameter optimization and fault tracing no longer blind troubleshooting, but rather a data-driven, clearly targeted intelligent search. This greatly improves the efficiency and accuracy of locating the root cause of poor wind separation performance, reduces reliance on the experience of senior grain and oil processing engineers, and achieves automation and intelligence in grain processing parameter optimization.

[0006] Furthermore, the air separation processing entity includes at least one of the following: grain batch, raw grain source, air separation equipment, operator, and screen assembly.

[0007] Furthermore, determining the historical correlation index for each node also includes applying a time decay coefficient to the historical anomaly event, the magnitude of which is inversely proportional to the time interval between the historical anomaly event and the current time.

[0008] Compared to analytical methods that treat all historical equipment maintenance data and anomaly records equally, this invention introduces a time decay coefficient, giving higher weight to recent wind-separation anomalies when calculating historical correlation. This aligns with the objective laws governing equipment wear and tear and changes in grain properties over time in grain processing practice, ensuring the real-time nature and effectiveness of the model analysis. In this way, the model can better capture the main current problems and potential risks of the wind-separation system, avoiding interference from outdated, resolved historical issues, thereby improving the accuracy of the suspicion weight calculation.

[0009] Furthermore, the time decay coefficient ranges from 0.1 to 0.5.

[0010] Furthermore, the specific method for calculating the path instability index is as follows: calculate the absolute value of the difference between the actual value and the set specification center value of all key process parameters, divide it by the difference between the set specification upper limit and the center value to obtain the normalized deviation; calculate the average value of the normalized deviation of all key process parameters; add 1 to the average value and multiply it by the historical correlation index to obtain the path instability index.

[0011] By dividing the difference between the actual value and the center value by the difference between the upper limit of the specification and the center value, the incomparability caused by the different dimensions and specification ranges of different process parameters is eliminated. This allows the model to fairly and objectively measure the impact of fluctuations in all key process parameters on production stability, thereby more accurately calculating the path instability index and providing a more reliable basis for obtaining the subsequent suspicion weight.

[0012] Furthermore, the abnormal characteristics of the wind separation include at least one of the following: abnormal type, abnormal morphology, and the section in which the abnormality occurred.

[0013] Furthermore, the data in the air separation processing data graph model comes from the production execution system, fan and motor sensors, and online quality inspection system.

[0014] Furthermore, the step of prioritizing the expansion of tracing paths with higher cumulative suspicion weights specifically includes: maintaining a priority queue for storing tracing paths to be expanded, wherein the priority of each path is determined by its cumulative suspicion weight; each time, the path with the highest priority is taken from the priority queue as the current path, all adjacent edges and adjacent nodes of the current path's end node are searched, the cumulative suspicion weight of the new path extending to each adjacent node is calculated, and the new path is added to the priority queue.

[0015] Compared to traditional blind search algorithms such as breadth-first or depth-first search, this invention expands paths by maintaining a priority queue based on accumulated suspicion weights. This method ensures that the tracing process always explores the most suspicious path, representing a heuristic intelligent search strategy. It significantly reduces the search of invalid paths, substantially improves the computational efficiency of finding the root cause, and enables the system to respond and locate problems as quickly as possible.

[0016] Furthermore, the tracing of the reverse path terminates when a preset stopping condition is met; the preset stopping condition includes reaching a set tracing depth or returning to a set number of paths.

[0017] Compared to scenarios that could lead to infinite tracing or excessively large tracing ranges, this invention sets a preset tracing depth or number of paths as a stopping condition. This provides a clear stopping mechanism for the intelligent tracing algorithm, ensuring that the analysis process is controllable and limited, and avoiding waste of computational resources. It allows the system to output a set of the most likely root cause paths within a reasonable timeframe and scope, enhancing the method's engineering practicality.

[0018] In a second aspect, the present invention provides a grain wind separation parameter optimization system based on digital twins, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a grain wind separation parameter optimization method based on digital twins of the present invention is implemented.

[0019] The technical effects of this invention are as follows: The core innovation of this invention lies in proposing a digital twin-based intelligent parameter optimization and traceability method for grain winnowing. It constructs a winnowing data graph model incorporating multi-dimensional information such as grain batches, equipment, and personnel, and uniquely integrates two dimensions: historical anomaly similarity and real-time process parameter deviation, to quantify the suspicion weight of each processing stage. Based on this, the system can start from grain batches with substandard winnowing performance and perform efficient heuristic reverse tracing, prioritizing the identification of the most likely root cause. Compared to traditional manual investigation relying on experience, this invention achieves quantification and precision in parameter optimization, significantly improving the quality control level and production efficiency of finished grains in complex grain and oil processing scenarios. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart illustrating a grain wind sorting parameter optimization method based on digital twins according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the working principle of the air separation device and the airflow material direction in an embodiment of the present invention; Figure 3This schematically illustrates the digital twin model and node statistics diagram of wind separation processing data constructed in an embodiment of the present invention, wherein... Figure 3 (a) is the topological structure diagram of the graph model for multidimensional entity association. Figure 3 (b) A bar chart showing the number of nodes for various production entities; Figure 4 This is a schematic illustration of a heatmap showing the historical status assessment of device nodes calculated based on historical fault data in an embodiment of the present invention. Figure 5 This is a schematic heatmap showing the distribution of normalized deviation of key process parameters in each processing step in an embodiment of the present invention. Figure 6 This is a schematic diagram illustrating the structural block of a grain wind separation parameter optimization system based on digital twins, according to an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] An example of a grain air separation parameter optimization method based on digital twins: like Figure 1 As shown, the grain wind sorting parameter optimization method based on digital twins of the present invention includes: S1. Construct a data graph model of air separation processing and obtain defect information.

[0024] In this embodiment, to achieve data association and parameter traceability throughout the entire lifecycle of grain air separation processing, it is first necessary to construct an air separation processing data graph model. For example, this graph model can be implemented based on graph database technology (such as Neo4j).

[0025] Specifically, the model uses various entities in the air separation process as nodes. These nodes encompass all elements required for grain processing and parameter optimization, such as: grain batch nodes specific to a single processing batch, raw grain source nodes identifying the origin or silo, air separation equipment nodes performing specific air separation processes on the production line (such as circulating air separators, gravity destoners, and rotary screens), operator nodes operating the equipment or performing inspections, and key consumable screen component nodes used for physical screening and separation. Each node contains its own attribute information, such as the batch ID and variety information of the grain batch node, the rated power and maintenance records of the equipment node, and the aperture specifications and material of the screen component node.

[0026] To gain a clearer understanding of the physical meaning of nodes and edges in a graph model, please refer to [reference needed]. Figure 2 This demonstrates the working principle of the air separation equipment and the flow of air and materials. For example... Figure 2 As shown, after the raw grain enters through the feed inlet, it moves above the screen; the airflow generated by the fan below passes through the screen, using the difference in the suspension velocity of the materials to separate the lighter impurities from the heavier clean grain. The graphical model of this invention is abstractly constructed based on this kind of physical connection relationship.

[0027] Combination Figure 3 As shown, it displays the completed digital twin model of the air separation processing data. Among them, Figure 3 (a) The topology of the production panorama is presented intuitively. The nodes of different colors in the figure represent grain batches, air separation equipment, raw grain sources, screen components and operators, respectively. The directed lines between the nodes show the material flow and process control relationship. Figure 3 b) further analyzed the scale of various nodes in the current graph. For example, in the subgraph associated with the current batch, there are 25 grain batch nodes and 12 wind separation equipment nodes. These nodes constitute the basic search space for subsequent parameter tracing.

[0028] The air-separation processing activity serves as the connecting edges, all of which are directed, reflecting the direction of grain processing and material flow. The edge types clearly define the relationships between nodes. For example, the initial air-separation edge connects the raw grain source node, initial cleaning equipment node, screen assembly node, operator node, and the work-in-process node formed after initial cleaning; the gravity destoning edge connects the grain node after initial cleaning, the destoner node, the blower assembly node, and the clean grain node after destoning; the online detection edge connects the processed grain node and the visual inspection equipment node, indicating whether the result indicates a qualified or abnormal product. By collecting data in real-time from the Supervisory Control and Data Acquisition (SCADA) system, blower frequency converter feedback data, air pressure sensors, and online quality inspection instruments, this air-separation processing data model is dynamically updated, forming a digital twin map that accurately reflects the entire air-separation process.

[0029] When a batch of grain is determined to have substandard air separation performance, such as excessive impurity content, incomplete stone removal, or excessive grain loss, the parameter optimization and traceability process is initiated. First, a unique identifier for the batch of grain, such as the batch ID, is obtained, and its corresponding grain batch node is precisely located in the constructed graph model. This node will serve as the starting point for this parameter optimization.

[0030] At the same time, the characteristics of this wind-separation anomaly are described in a structured manner to generate a standardized anomaly feature vector. Understandably, this vector serves as the initial input for all subsequent calculations, and its accuracy directly impacts the effectiveness of parameter optimization. In one exemplary implementation, the anomaly feature vector... It may include, but is not limited to, the following dimensions: Abnormal types, such as "incomplete removal of light impurities (glumes remaining)," "side stones not removed," and "loss of normal grains," can be addressed using One-Hot Encoding. Abnormal forms / values: such as the "percentage deviation" of impurity content, the "average aerodynamic diameter" of residual impurities, and the "thousand-grain weight" of lost grains, are numerical values; Location / Section of Occurrence: If the anomaly is detected at the "exit of the fine selection section" or "second-stage screening material", it can be coded.

[0031] This step completes the initialization of the parameter optimization task, that is, it clarifies the starting point of the tracing and the target feature to be optimized.

[0032] S2. Determine the historical correlation index of a node based on the current defect features and the defect feature vectors corresponding to historical defect events.

[0033] To effectively integrate past maintenance records and parameter tuning experience of air separation equipment into this optimization process, this embodiment quantifies the historical correlation strength between any production entity node in the air separation processing diagram model and the current air separation anomaly type to be addressed. This allows the model to refer to historical data and prioritize nodes that have previously caused similar impurity or classification problems (e.g., a certain suction separator that has historically frequently resulted in incomplete removal of light impurities).

[0034] Specifically, a historical anomaly database needs to be built and maintained in advance. This database records in detail every confirmed event of substandard wind separation performance, including the anomaly feature vector at the time (such as "high impurity content", "severe grain loss", etc.), the root cause node confirmed after review (such as the frequency converter of a certain wind turbine, a batch of high moisture raw grain, or a specific screen component, etc.), and the time when the anomaly occurred.

[0035] When a new parameter optimization task is started, for any node in the graph model Its characteristics are consistent with current wind-separation anomalies. Historical correlation indicators The calculation is performed using the following formula: ; in Represents a node Regarding the current abnormal characteristics of wind selection Historical correlation indicators; This indicates that in the historical anomaly database, all confirmed cases originating from nodes... As a collection of all historical anomalies resulting from the root cause; Represents a set The anomalous feature vector corresponding to a specific historical anomaly event; This represents an anomaly feature similarity function used to calculate the similarity between a current anomaly and a historical anomaly. As a preferred approach, this function can be implemented using a cosine similarity algorithm, and its output value ranges from [value missing]. The closer the value is to 1, the more similar the manifestations of the two wind separation anomalies are (e.g., both are "excessive husk residue rate"). This indicates the current timestamp of the parameter optimization task being executed. Indicates historical anomalies The timestamp of when it occurred; This represents a positive number, which serves as a time decay coefficient to adjust the weight of historical events on the current judgment.

[0036] From the above formula, it can be seen that if a node (For example, a certain circulating air separator) has historically caused anomalies similar to the current one on multiple occasions. Highly similar problems, i.e. The higher the value, and the more recent these historical events are from the present time, the better. A smaller value indicates a lower historical correlation index for that node. The higher the value, the greater the likelihood, based on historical experience, that this node is a potential source of poor current wind selection performance.

[0037] As a preferred option, the time decay coefficient The value can be set according to the seasonal changes of the raw grain variety and the maintenance cycle of the equipment, and its preferred range is: For example, when the properties of the raw grain are relatively stable during the storage period, a smaller [size / weight] can be selected. A value (e.g., 0.1) is chosen to retain longer-term historical equipment condition experience; a larger value can be selected when the new grain is on the market, the moisture and impurity content of the raw grain changes rapidly, or the equipment has just undergone a major overhaul and technological upgrade. A value (such as 0.5) is used to focus more on recent anomalous patterns and avoid interference from outdated data.

[0038] like Figure 4 As shown, this is a heatmap illustrating the historical status assessment of device nodes based on historical fault data. The horizontal and vertical axes represent different device numbers or location indices, and the color intensity of the blocks represents the historical operational stability score of that device. For example, the device at coordinates (1.5, 2.0) in the graph displays a value of 0.98, indicating extremely stable historical operation with very few associated historical anomalies. Therefore, in the calculation... The value is relatively low at this point; while the device at coordinates (1.5, 1.0) has a value of 0.72, indicating that it has been involved in parameter misalignment or failure events multiple times in the past. Therefore, the system will assign it a higher historical correlation index and give it special attention during tracing.

[0039] S3. Historical correlation indicators and extreme value calculation path instability indicators of key process parameters obtained in step S2.

[0040] After obtaining the historical correlations at the node level, further analysis is needed in conjunction with the real-time status of air classification processing activities. The instability of an air classification processing path depends not only on the historical maintenance records of its associated equipment nodes, but also on the process parameter control during the execution of that specific processing activity. This step aims to integrate the historical risks of nodes with the real-time parameter deviations of the air classification process.

[0041] For the source node in the wind separation processing diagram model With the target node one side This represents the path instability index of a specific air classification process (such as "gravity destoning" or "circular air classification"). By using the result node generated by this activity, i.e., the target node Historical relevance This serves as the base risk value, which is then amplified based on the actual deviations of the critical process parameters (KPPs) in the activity. The calculation formula is as follows: ; in Representing an edge Regarding the current abnormal characteristics of wind selection Path instability index; This represents the edge calculated in step S2. The target node Historical correlation indicators; This indicates the air separation processing activity This is a collection of key process parameters. For air classification processes, these parameters typically include: static pressure in the air chamber (Pa), frequency of the blower (Hz), vibration amplitude of the screen vessel (mm), damper opening (%), and feed flow rate (t / h), etc. These parameters and their specification ranges can be obtained from the company's air classification process standard documents or operating procedures (SOPs). Indicates the number of key process parameters; Indicates parameters The actual recorded values ​​during this processing activity can be obtained in real time from the production SCADA system, fan controller, or smart sensors installed on site. and Representing parameters respectively The upper and lower limits of the specifications are specified in the process standards, and there are .

[0042] The above formula calculates the average normalized deviation of all key air separation process parameters from their specification center values. Represents parameters The ideal center setting value (e.g., the wind pressure value corresponding to the optimal air separation efficiency), while This represents the maximum permissible deviation on one side. Therefore, This item maps the actual parameter value to a... The larger the value, the further away from the ideal process center.

[0043] As can be seen from the above formula, when all key process parameters When all values ​​are precisely at their specified center, for example, when the wind pressure is stable at the optimal operating point, the summation term is 0. Path instability is entirely determined by the historical performance of the nodes. When process parameters gradually deviate from the center and approach their specification boundaries, for example, when air pressure drifts to a critical value due to filter clogging, the deviation term approaches 1. In the extreme case where all parameters reach their specification boundaries, the value within parentheses approaches 2, making... The value was amplified to This reflects the fact that if a historically problematic equipment node experiences unstable air separation parameter control and drastic airflow fluctuations during a processing activity, its likelihood of causing impurities or loss increases dramatically.

[0044] like Figure 5As shown in the figure, this is a heatmap illustrating the distribution of normalized deviations of key process parameters in each processing step. The vertical axis represents different processing steps such as "initial cleaning," "destoning," "air separation," and "fine cleaning," while the horizontal axis represents key process parameters (KPPs) such as "air pressure," "frequency," "amplitude," "damper," and "flow rate." The darker the color, the greater the deviation of that parameter from the set center value in the current real-time production.

[0045] S4. Calculate the edge suspicion weight using the inherent coefficient of the air separation process and the path instability index obtained in step S3.

[0046] Path instability indicators alone are insufficient to fully reflect their potential role as the root cause of air separation anomalies, as different air separation processing steps have vastly different impacts on the final product quality. For example, even minor deviations in air pressure during the gravity destoning process can lead to stones being mixed into the finished product or a large amount of raw grain being discharged as impurities, causing serious quality incidents or economic losses. While parameter deviations in the initial cleaning or dust removal processes may also affect the environment or preliminary cleanliness, subsequent processes usually offer opportunities for remediation, and their fatal impact on the final product is relatively small. Therefore, it is necessary to incorporate the inherent criticality of the processes themselves into the calculations.

[0047] Suspicion weight In path instability index Based on this, the final weight value obtained by considering the criticality of the air separation process itself is used to directly guide the graph traversal algorithm. The calculation formula is as follows: ; in Representing an edge Regarding the current wind selection anomaly The final suspicion weight is the core indicator for guided graph traversal; This represents the path instability index calculated in step S3; Representing an edge The process-inherent coefficient representing the air separation process. This coefficient is a preset static value, pre-assigned to each type of air separation process by the process engineering team based on Failure Mode and Effects Analysis (FMEA) or grain processing expert scoring methods.

[0048] By dynamically calculating path instability Compared to static, expert-knowledge-based process criticality Multiplication enables an effective combination of dynamic data and static knowledge. Even a path with significant wind pressure fluctuations can be considered valid if the process it involves is not critical (i.e., ...). If the value is low, its final suspicion weight will not be too high. Conversely, if it is located in a core or critical process (i.e., A moderate level of parameter fluctuation (high value) may be more suspicious than a highly unstable path located on a non-critical process, thus guiding the system to prioritize the investigation of core links.

[0049] As a preferred option, the inherent coefficient of the process The value range can be set to An example: For core processes that directly determine the cleanliness and grade of the finished grain, such as gravity destoning, circulating air separation, and fine grading, their... It can be set to 4.0 or 5.0; for important but relatively stable processes such as planar rotary screening and magnetic separation, it can be set to 3.0; while for auxiliary processes such as dust removal and ventilation, conveying and lifting, it can be set to 1.0.

[0050] S5. Guided graph tracing based on suspicion weight.

[0051] The suspicion weights of relevant edges in the wind-sorting processing diagram model are calculated. Then, a guided best-first search algorithm is used to replace the traditional breadth-first search (BFS) or depth-first search (DFS) algorithm, and reverse tracing is performed starting from the grain batch node where the wind separation effect is not up to standard.

[0052] The core of this algorithm lies in maintaining a priority queue, where each element represents a tracing path to be expanded. The priority of a path is determined by its cumulative suspicion score, which is defined as the suspicion weight of all edges traversed along the path. sum.

[0053] The specific traceability process is as follows: Initialization: First, the starting node for tracing, i.e. the target grain batch node that is determined to be abnormal (e.g., "Batch_20251025_wheat_03"), is placed into the priority queue as the initial path.

[0054] Path selection: When the priority queue is not empty, the path with the highest cumulative suspicion is selected as the current expansion path. This means that the system always prioritizes exploration along the path with "many historical failures, large current parameter fluctuations, and critical processes".

[0055] Reverse expansion: For the end node of the path, such as a certain suction separator node, find all its adjacent edges and adjacent nodes that point to the node, that is, find the upstream feeding equipment, the source of raw grain or the previous screening process. This is the reverse tracing step.

[0056] Weight accumulation and enqueueing: Subsequently, for each newly discovered, unvisited adjacent path, calculate its new cumulative suspicion score (original path suspicion score + new edge suspicion score). ), and add it to the priority queue.

[0057] Iterative loop: Repeat the above iterative process of path expansion and enqueueing until the preset stopping condition is met.

[0058] The stopping conditions can be set flexibly to balance computational efficiency and tracing depth. For example, a tracing depth threshold can be set, such as tracing back a maximum of 10 process levels; or a limit can be set on the number of return paths, such as returning the 5 root source paths with the highest cumulative suspicion.

[0059] The algorithm ultimately outputs one or more traceability paths, sorted in descending order of cumulative suspicion. The nodes and edges pointed to by these paths represent the most suspicious combination of production processes that led to the current anomaly in the air sorting quality.

[0060] Through the above steps, the system can provide a clear diagnostic conclusion, such as: "Path A: points to the No. 2 gravity destoner, whose damper opening showed abnormal fluctuations when processing this batch of high-moisture raw grain ( (High), and this process is a critical process ( Based on this output, on-site process personnel or automatic control systems can directly focus their attention on these high-suspicion links, specifically checking whether the air inlet of a particular fan is blocked, adjusting the frequency converter parameters, or replacing worn screen plates. This enables rapid and accurate optimization of air separation parameters and location of the root cause of the fault, effectively avoiding the dilemma of blindly checking through massive amounts of irrelevant equipment data, and significantly improving the response speed of the production line and the stability of the finished grain quality.

[0061] An example of a grain air separation parameter optimization system based on digital twins: On the other hand, the present invention also provides a grain air separation parameter optimization system based on digital twins. For example... Figure 6 The memory stores computer program instructions, which, when executed by the processor, implement a grain wind sorting parameter optimization method based on digital twins according to the first aspect of the present invention.

[0062] A grain air separation parameter optimization system based on digital twins also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0063] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

Claims

1. A method for optimizing grain air separation parameters based on digital twins, characterized in that, The method includes: Construct a data graph model for air separation processing, where nodes represent air separation processing entities and directed edges represent air separation processing activities; Obtain the abnormal characteristics of the target grain batch in the air separation process; determine the historical correlation index of each node based on the similarity between the abnormal characteristics and the historical abnormal events corresponding to each node; calculate the parameter deviation correction coefficient according to the degree of deviation between the actual value of the key process parameter in the corresponding air separation processing activity and its set specification center value; use the parameter deviation correction coefficient to weight and amplify the historical correlation index to obtain the path instability index of each edge. The path instability index is multiplied by the preset process inherent coefficient of the corresponding wind separation processing activity to obtain the suspicion weight of each edge. Starting from the node corresponding to the target grain batch, trace along the reverse path of the graph model, prioritizing the expansion of traceability paths with higher cumulative suspicion weights.

2. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, The air separation processing entity includes at least one of the following: grain batch, raw grain source, air separation equipment, operator, and screen assembly.

3. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, Determining the historical correlation indicators for each node also includes: A time decay coefficient is applied to the historical anomaly, the magnitude of which is inversely proportional to the time interval between the historical anomaly and the current time.

4. The method for optimizing grain air separation parameters based on digital twins according to claim 3, characterized in that, The time decay coefficient ranges from 0.1 to 0.

5.

5. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, The specific method for calculating the path instability index is as follows: Calculate the absolute value of the difference between the actual value and the set specification center value of all key process parameters, divide it by the difference between the set specification upper limit and the center value, and obtain the normalized deviation. Calculate the average normalized deviation of all key process parameters; The path instability index is obtained by multiplying the average value by 1 and then by the historical correlation index.

6. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, The abnormal characteristics of wind separation include: At least one of the following: abnormal type, abnormal form, and the section in which the abnormality occurred.

7. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, The data in the wind separation processing data graph model comes from the production execution system, wind turbine and motor sensors, and online quality inspection system.

8. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, The priority extension of the tracing path with a higher cumulative suspicion weight specifically includes: Maintain a priority queue for storing traceability paths to be expanded, where the priority of each path is determined by its cumulative suspicion weight; Each time, the highest priority path is taken from the priority queue as the current path, all adjacent edges and adjacent nodes of the end node of the current path are found, the cumulative suspicion weight of the new path extending to each adjacent node is calculated, and the new path is added to the priority queue.

9. The method for optimizing grain air separation parameters based on digital twins according to claim 1, characterized in that, The tracing of the reverse path terminates when a preset stopping condition is met; The preset stopping conditions include reaching a set tracing depth or returning to a set number of paths.

10. A grain air separation parameter optimization system based on digital twins, characterized in that, It includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, it implements the grain wind sorting parameter optimization method based on any one of claims 1 to 9.

Images