A threshing and redrying batch dividing system and method

Abstract

The application relates to the tobacco production technical field and discloses a threshing and redrying batch division system and method, which comprises the following steps: S1, collecting actual running time lengths of multiple threshing and redrying batches, and calculating theoretical running time lengths of various processes; S2, calculating batch running time lengths of corresponding processes according to the theoretical running time lengths of the various processes; S3, establishing corresponding production time sequence queues for various processes of a current threshing and redrying batch; S4, detecting actual running time lengths of various processes of a current batch according to the batch running time lengths, and detecting running time length abnormalities; S5, defining an advanced running time length; defining a stagnation running time length; S6, when the advanced running time length, delaying a time stamp of a process to the batch running time length, and then switching to a next batch number; when the stagnation running time length, adjusting the time stamp of the process forward to the batch running time length, and then switching to the next batch number. The application realizes a leap from "passive record errors" to "active real-time correction".

Description

Technical Field

[0001] This application relates to the field of tobacco production technology, and specifically to a batch classification system and method for leaf re-drying. Background Technology

[0002] Leaf threshing and re-drying is a crucial link between agriculture and industry in tobacco processing. Its core task is to process the initially cured raw tobacco through a series of steps including re-moistening, threshing, air-sifting, and re-drying, into sheet tobacco products with uniform moisture and temperature and meeting required leaf structure requirements. These sheet tobacco products are typically stored and aged in batches before being supplied to downstream cigarette factories. Therefore, establishing a precise and reliable batch management system is the cornerstone for ensuring product quality traceability and stability.

[0003] Currently, the most common technology used in threshing and re-drying production lines is to use a Manufacturing Execution System (MES) or Programmable Logic Controller (PLC) to preset a fixed threshold (such as cumulative weight, 9600kg of tobacco leaves) at the beginning of the threshing and re-drying process. When the production data reaches this threshold, the MES system automatically generates a new batch number and issues it to the subsequent processes of that batch.

[0004] The batch of tobacco leaves is measured by accumulating their weight at each stage (i.e., when the detector measures 9600 kg, it is considered to include all the tobacco leaves in that batch). However, each stage performs corresponding processing on the tobacco leaves in that batch (uneven moisture content of the incoming tobacco leaves, fluctuations in the distribution of material in the storage bins and the output of material, changes in the leaf-breaking rate of the leaf-breaking unit, etc.), which causes the weight of the tobacco leaves in that batch to change. Therefore, when the fixed threshold is measured at each stage, it does not completely include the tobacco leaves in that batch, resulting in confusion in the batch division of tobacco leaves at each stage.

[0005] The aforementioned issues pose a significant challenge to the end-to-end quality traceability of the leaf re-drying processing technology. This could not only lead to distorted quality evaluation of the entire batch of products, but also cause a break in the quality traceability chain, resulting in major quality management loopholes. Summary of the Invention

[0006] This application provides a batch division system and method for leaf re-drying, which aims to solve the above-mentioned problems.

[0007] One approach provides a method for batch division of leaf re-drying, which includes the following steps: S1. Collect the actual running time of each process in multiple batches of leaf re-drying as historical data, and calculate the theoretical running time of each process based on the historical data; S2. Calculate the batch runtime tolerance threshold for the corresponding process based on the theoretical runtime of each process; the value range of the batch runtime is within... ;in, This refers to the theoretical runtime. This is the tolerance threshold; S3. Establish a corresponding production time sequence queue for each process of the current leaf re-drying batch, associate the timestamp of the production time sequence queue with the current leaf re-drying batch number, and execute steps S1 and S2 for each process to obtain the running time corresponding to each process. S4. Detect whether the actual running time of each process in the current batch is within the range of the batch running time based on the batch running time. If not, determine that the running time is abnormal. S5. Regarding the runtime anomaly, if the batch runtime of each process in the current batch is less than... Time is defined as the advance runtime; the batch runtime of each process in the current batch is greater than 100%. Time is defined as the duration of the stay. S6. When the process exceeds the allotted time, stop dividing the current batch number and postpone the timestamp of the process until the batch corresponding to that process reaches the allotted time. Then switch to the next batch number; When the process experiences a runtime delay, the division of the current batch number is stopped, and the timestamp of the process is adjusted forward until the runtime of the batch corresponding to that process reaches [a certain value]. Then switch to the next batch number.

[0008] In one approach, the theoretical runtime is obtained using the following formula:

[0009] in, ; The set of data representing the actual runtime of the same process in n batches of the historical data; for The median of the dataset.

[0010] In one approach, in step S2, the tolerance threshold is obtained using the following formula:

[0011] Where k is a set multiple, and is a constant; This represents the absolute median difference.

[0012] In one scheme, step S4 further includes real-time detection of whether two adjacent batch number sequences of each process are consecutive. If not, and the missing batch number does not appear in the batch of the previous process, it is determined to be a missing batch anomaly. In real time, it is detected whether the sequences of two adjacent batch numbers in each process are consecutive. If not, and the missing batch number has appeared in the batch of the previous process, it is determined to be a missing batch anomaly.

[0013] In one approach, for the missing batch anomaly, the later batch number in two adjacent batch numbers is corrected to restore the continuity of the batch sequence numbers. For the aforementioned batch shortage anomaly, the missing process position is filled with the corresponding batch runtime, and the timestamps of all subsequent processes are adjusted accordingly.

[0014] In one scheme, in step S4, an alarm signal is output according to the runtime anomaly, the missing batch anomaly, and the few batch anomaly, and the process position corresponding to the occurrence of the corresponding runtime anomaly, the missing batch anomaly, and the few batch anomaly is located.

[0015] In one approach, another aspect provides a batch segmentation system for leaf re-drying, including: The data acquisition and processing module is used to collect the actual running time of each process in multiple batches of leaf re-drying, and to calculate the theoretical running time of each process. The statistical calculation module is used to calculate the batch runtime of each process, wherein the value of the batch runtime is within a certain range. ;in, This refers to the theoretical runtime. This is the tolerance threshold; The time-series management module is used to establish corresponding production time-series queues for each process of the current leaf-picking and re-drying batch, and associate the timestamps of the production time-series queues with the current leaf-picking and re-drying batch number one by one. The real-time verification module is used to determine whether the actual runtime of each process is abnormal. For abnormal runtimes, if the batch runtime of each process in the current batch is less than [a certain value], [further verification is required]. Time is defined as the advance runtime; the batch runtime of each process in the current batch is greater than 100%. Time is defined as the duration of the stay. The correction module is used to adjust the process timestamp to the corresponding batch runtime for the process based on the advanced runtime or the delayed runtime. or Then switch to the next batch number.

[0016] One approach also includes an optimization module, which processes the historical data using a random forest algorithm to adjust the calculated theoretical runtime of each process and the tolerance threshold.

[0017] One solution also includes an early warning and diagnosis module, which is used to output alarm signals for the runtime anomaly, the missing batch anomaly, and the few batch anomaly, and to locate the process position corresponding to the occurrence of the corresponding runtime anomaly, the missing batch anomaly, and the few batch anomaly.

[0018] The beneficial effects of this application are: Existing technologies rely on measuring fixed weight thresholds at each process stage to divide batches. However, due to weight variations caused by material moisture evaporation and losses during processing, the actual batch boundaries of each process cannot be aligned. This invention abandons the "fixed weight" threshold and uses a dynamic time window based on the theoretical running time and tolerance threshold of each process as the criterion for batch division and synchronization. The time parameter is not affected by changes in the absolute weight of the material, thus ensuring the consistency of the time logic of the same batch across processes and eliminating the root cause of batch boundary misalignment.

[0019] Existing technologies involve one-way, rigid instruction issuance. This invention, by monitoring actual runtime in real time and comparing it with a dynamic tolerance range, can automatically identify "ahead of time" or "delayed" anomalies. It can also proactively intervene in batch switching logic, automatically aligning the batch end time to near a reasonable theoretical duration by delaying or advancing the batch number switch. This represents a leap from "passively recording errors" to "proactive real-time correction," giving batch management intelligent and adaptive capabilities.

[0020] By establishing a production timeline queue for each process and strongly linking timestamps with batch numbers, this method precisely timestamps the "trajectory" of each batch of materials throughout the entire production process. When quality traceability is required, a clear, continuous, and logically consistent batch timeline can be used to accurately pinpoint the specific processing time of a particular batch within any given process, and retrieve all process parameters for that time period. This ensures the integrity and accuracy of the quality traceability chain, avoiding the problem of unclear quality responsibility due to batch confusion.

[0021] This method elevates management from a broad "weight accumulation" approach to a refined "process timing" control. By monitoring and optimizing the runtime of each process, managers can more accurately assess the cycle time matching, equipment stability, and process stability of each process. This provides data-driven decision support for production scheduling, equipment maintenance, and process optimization, propelling the leaf re-drying production towards a higher level of refined and digital management. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic flowchart of a partitioning method according to an embodiment of this application; Figure 2 This is a schematic diagram of the system structure flow according to an embodiment of this application; The system comprises: 1. Data acquisition and processing module; 2. Statistical calculation module; 3. Time-series chain management module; 4. Real-time verification module; 5. Correction module; 6. Optimization module; 7. Early warning and diagnosis module; 8. Mobile terminal; and 9. MES system. Detailed Implementation

[0024] The specific embodiments of this application will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application. Similarly, the following examples are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] In this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0026] This application proposes improvements and innovations, and presents the following embodiments.

[0027] In some implementations, please refer to Figures 1 to 2 A method for batch division of re-drying leaves is provided, including the following steps: S1. Collect the actual running time of each process in multiple batches of leaf re-drying as historical data, and calculate the theoretical running time of each process based on the historical data. S2. Calculate the tolerance threshold for the batch runtime of the corresponding process based on the theoretical runtime of each process; the value range of the batch runtime is within... ;in, This refers to the theoretical runtime. This is the tolerance threshold; S3. Establish a corresponding production time sequence queue for each process of the current leaf re-drying batch, associate the timestamp of the production time sequence queue with the current leaf re-drying batch number, execute steps S1 and S2 for each process, and obtain the runtime corresponding to each process. S4. Check whether the actual running time of each process in the current batch is within the range of the batch running time. If not, it is determined to be an abnormal running time. S5. For runtime anomalies, if the batch runtime of each process in the current batch is less than... Time is defined as the advance runtime; the batch runtime of each process in the current batch is greater than 100%. Time is defined as the duration of the stay. S6. When a process runs ahead of schedule, stop dividing the current batch number and postpone the timestamp of the process until the corresponding batch runs out. Then switch to the next batch number; When a process experiences a runtime delay, the current batch number division is stopped, and the timestamp of the process is adjusted forward until the corresponding batch runtime reaches [a certain value]. Then switch to the next batch number.

[0028] Existing technologies rely on measuring fixed weight thresholds at each process stage to divide batches. However, due to weight variations caused by material moisture evaporation and losses during processing, the actual batch boundaries of each process cannot be aligned. This invention abandons the "fixed weight" threshold and uses a dynamic time window based on the theoretical running time and tolerance threshold of each process as the criterion for batch division and synchronization. The time parameter is not affected by changes in the absolute weight of the material, thus ensuring the consistency of the time logic of the same batch across processes and eliminating the root cause of batch boundary misalignment.

[0029] Existing technologies involve one-way, rigid instruction issuance. This invention, by monitoring actual runtime in real time and comparing it with a dynamic tolerance range, can automatically identify "ahead of time" or "delayed" anomalies. It can also proactively intervene in batch switching logic, automatically aligning the batch end time to near a reasonable theoretical duration by delaying or advancing the batch number switch. This represents a leap from "passively recording errors" to "proactive real-time correction," giving batch management intelligent and adaptive capabilities.

[0030] By establishing a production timeline queue for each process and strongly linking timestamps with batch numbers, this method precisely timestamps the "trajectory" of each batch of materials throughout the entire production process. When quality traceability is required, a clear, continuous, and logically consistent batch timeline can be used to accurately pinpoint the specific processing time of a particular batch within any given process, and retrieve all process parameters for that time period. This ensures the integrity and accuracy of the quality traceability chain, avoiding the problem of unclear quality responsibility due to batch confusion.

[0031] This method elevates management from a broad "weight accumulation" approach to a refined "process timing" control. By monitoring and optimizing the runtime of each process, managers can more accurately assess the cycle time matching, equipment stability, and process stability of each process. This provides data-driven decision support for production scheduling, equipment maintenance, and process optimization, propelling the leaf re-drying production towards a higher level of refined and digital management.

[0032] Specifically, for a particular raw material grade (such as top-grade upper tobacco, top-grade middle tobacco, etc.) at a specific process, historical batch data of its process flow rate are collected. Using the process flow rate signal of that process as a benchmark, the instantaneous point when the flow rate signal jumps from zero to non-zero is defined as the batch start time. The instantaneous point at which the flow rate signal drops from non-zero to zero is the batch end time. For each historical batch The actual runtime is calculated based on the flow signal:

[0033] In step S3, within the Manufacturing Execution System (MES), fixed thresholds (such as cumulative weight, running time, or number of boxes) are preset to divide batches. A globally unique batch number is generated for each newly generated production batch and distributed in real time to key process nodes on the production line (such as re-moistening, leaf trimming, re-drying, and packaging) through the production management system, driving collaborative production across the entire process. Simultaneously, at each process node, the MES system establishes an independent local batch production timeline queue. At the moment materials arrive or a process begins, the timestamp of the production timeline queue is strongly bound to the batch number, written as a composite primary key to the local queue and database, and the actual start and end times of the batch at the process node are recorded, along with the preset theoretical running time. The tolerance threshold Δ is compared in real time to form a continuous, complete, and verifiable batch production timeline.

[0034] In one embodiment, the theoretical runtime is obtained using the following formula:

[0035] in, ; A dataset representing the actual runtime of the same process in n batches of historical data; for The median of the dataset.

[0036] Specifically, will The duration data is sorted in ascending order: If n is odd It is the (n+1) / 2th value after sorting.

[0037] If n is even It is the arithmetic mean of the n / 2th and (n / 2)+1th values ​​after sorting.

[0038] The median is insensitive to extreme outliers (such as excessively long or short batches due to temporary equipment failures), which allows the calculated theoretical duration to more accurately reflect the central trend under stable production conditions, avoiding the problem that the mean is easily "skewed" by a few outliers.

[0039] This ensures the representativeness and stability of the benchmark parameters, making subsequent tolerance calculations and anomaly detection more reliable and reducing false alarms caused by inaccurate benchmarks.

[0040] The method is simple and effective with low computational complexity, making it suitable for real-time and efficient deployment and updates in industrial control systems.

[0041] In one embodiment, in step S2, the tolerance threshold is obtained by the following formula:

[0042] Where k is a set multiple, and is a constant; This represents the absolute median difference.

[0043] Specifically, tolerance threshold This is used to reflect the normal fluctuation range of controlled processes and equipment under stable production conditions. Its calculation is based on data distribution characteristics, rather than fixed values. The calculation steps are as follows: (1) Calculate the absolute median: Calculate the absolute deviation of each sample from the median. And find the median of its absolute value sequence, i.e., the absolute median difference. ;

[0044] (2) Scaling MAD to obtain standard deviation estimate: Compared with the normal distribution, the relationship between MAD and standard deviation σ is approximately .

[0045] (3) Setting the multiplier k: Determined based on process requirements and equipment performance, usually taken as (Covering approximately 95.4% of the data) or (Covering approximately 99.7% of the data).

[0046] Compared with traditional experience-based fixed thresholds, this method quantifies the normal fluctuation range based on the actual distribution characteristics of historical data (measured by MAD), so that the tolerance threshold naturally fits the inherent rhythmic characteristics of a specific production line and a specific process.

[0047] Together with the median, it forms a complete robust statistical modeling scheme that can effectively resist the interference of outliers in historical production data. The calculated tolerance range more reasonably covers "controlled normal fluctuations" rather than including accidental abnormal events, thereby improving the signal-to-noise ratio and accuracy of anomaly detection.

[0048] By using the scaling factor k, the system can flexibly adjust the severity of the judgment according to the process control accuracy requirements of different processes (e.g., k=2 corresponds to general control, k=3 corresponds to strict requirements), thereby enhancing the applicability and configurability of the method.

[0049] In one embodiment, step S4 further includes real-time detection of whether two adjacent batch number sequences of each process are consecutive. If not, and the missing batch number does not appear in the batch of the previous process, it is determined to be a missing batch anomaly. The system monitors in real time whether two adjacent batch number sequences in each process are consecutive. If not, and the missing batch number has appeared in a batch in the previous process, it is determined to be a missing batch anomaly.

[0050] The various processes in the leaf-beating and re-drying process are arranged sequentially according to the production line, namely, re-moistening, leaf beating, air separation, and re-drying.

[0051] Specifically, the system monitors in real time whether the current leaf-picking and re-drying batch has completed all corresponding processes in sequence. If not, it is determined to be a missing batch anomaly. A missing batch anomaly occurs when there are missing batches from the second process to the final process of the current leaf-picking and re-drying batch.

[0052] Specifically, when a non-continuous interval is detected between the batch number of a later-started batch and the batch number of a previously completed batch, i.e. a missing batch number exists, and the missing batch does not appear in the active batch record of the previous process, the system will immediately determine it as a "missing batch anomaly" and generate an anomaly alarm.

[0053] Specifically, the process involves checking whether a batch has flowed through all necessary steps according to the preset process path. If a batch is lost in the intermediate process queue, the final process queue, or the starting process queue, it is determined to be a "missing batch".

[0054] This embodiment enables online monitoring of the logical integrity and continuity of batches. It can instantly detect "breaks" (missing batches) and "absences" (few batches) in the batch chain, filling the monitoring blind spot of traditional methods that only focus on the duration of a single process while ignoring the logical connections between processes.

[0055] The clear rules for determining "missing batches" and "incomplete batches" (such as checking the continuity of batch numbers and the integrity of process paths) enable the system to make a quick and accurate diagnosis of the serious problem of interrupted batch information flow, providing an early warning to prevent the quality traceability chain from breaking.

[0056] This mechanism upgrades batch management from "single-point management" to "process link management," ensuring the synchronization of production information flow and physical material flow.

[0057] In one embodiment, for a missing batch anomaly, the later batch number in two adjacent batch numbers is corrected to restore the continuity of the batch sequence number. For missing batch anomalies, fill in the corresponding batch runtime at the missing process position, and adjust the timestamps of all subsequent processes accordingly.

[0058] Specifically, the average value method is used to infer the runtime of the missing batch, that is, the average of the runtime of the previous batch and the runtime of the next batch in the process is used as the batch runtime of the process.

[0059] For "missing batches," the continuity of the sequence numbers is directly restored by correcting the batch number; for "incomplete batches," the missing batch time period information is supplemented by interpolation (such as the averaging method). This is not just an alarm, but a proactive action to maintain the integrity of the logical world of batches.

[0060] The MES system 9 automatically performs corrective actions, generating logically coherent batch records. This ensures that even if there is a brief loss of signal in physical production, a traceable and continuous link can still be maintained in the information system, thus preventing quality traceability failures that might result from missing information.

[0061] Corrective logic (such as average inference) is simple, reasonable, and traceable, providing a clear basis for manual review.

[0062] In one embodiment, in step S4, alarm signals are output based on runtime anomalies, missing batch anomalies, and insufficient batch anomalies, and the process positions corresponding to the occurrence of the corresponding runtime anomalies, missing batch anomalies, and insufficient batch anomalies are located.

[0063] Specifically, alarms are issued in the form of sound and light on the central control room dashboard and related operating terminals, clearly indicating the type of abnormality (such as "batch N in the re-drying process is stuck"), the location and time of occurrence.

[0064] The MES system 9 automatically associates and displays the operating parameters of relevant equipment during the abnormal time period (such as motor current, valve opening, and belt scale reading), assisting technicians in quickly locating whether the batch problem is caused by equipment failure, process abnormality, or operational error.

[0065] This embodiment provides a multi-dimensional, visualized anomaly management interface. By using audible and visual alarms on the central control panel and clearly indicating the anomaly type, location, and time, it greatly improves the timeliness of problem response and the accuracy of location, reducing the troubleshooting time for operators.

[0066] The function of automatically linking and displaying equipment parameters during abnormal periods enables a rapid association between "abnormal phenomena" and "potential process causes," upgrading simple batch alarms into diagnostic tools that assist in root cause analysis. This helps technicians quickly move from "knowing there is a problem" to "knowing where the problem might be," thus improving troubleshooting efficiency.

[0067] It enhances human-machine collaboration capabilities, making the system not only a monitor but also an intelligent assistant for production and maintenance personnel.

[0068] In one embodiment, another aspect provides a leaf re-drying batch segmentation system, comprising: Data acquisition and processing module 1 is used to collect the actual running time of each process in multiple batches of leaf re-drying and calculate the theoretical running time of each process. Statistical calculation module 2 is used to calculate the batch runtime of each process. The value range of the batch runtime is within... ;in, This refers to the theoretical runtime. This is the tolerance threshold; The timing chain management module 3 is used to establish corresponding production timing queues for each process of the current leaf re-drying batch, and associate the timestamps of the production timing queues with the current leaf re-drying batch number one by one. Real-time verification module 4 is used to determine whether the actual runtime of each process is abnormal. If the runtime is abnormal, and the batch runtime of each process in the current batch is less than 100%, then the verification is performed. Time is defined as the advance runtime; the batch runtime of each process in the current batch is greater than 100%. Time is defined as the duration of the stay. Correction module 5 is used to adjust the process timestamp for advanced or delayed runtimes until the corresponding batch runtime for that process is reached. or Then switch to the next batch number.

[0069] From data acquisition, model calculation, real-time verification to dynamic correction and optimization, each module has a clear division of labor and works collaboratively to form an intelligent system covering the entire process of "perception-analysis-decision-execution".

[0070] Modular design makes the system easy to deploy, maintain, and expand. For example, new process monitoring points can be flexibly added, or more advanced optimization algorithm modules can be replaced.

[0071] By translating the methodology into concrete software and hardware systems, the industrial-scale implementation of this invention is clarified, and the scope of protection is extended from the method to the device, thereby enhancing the practical value and protective strength of the patent.

[0072] In one embodiment, the system further includes an optimization module 6, which processes historical data using a random forest algorithm to adjust the calculated theoretical runtime and tolerance threshold for each process.

[0073] The introduction of machine learning capabilities has enabled the system to evolve from a "rule-based static model" to a "data-driven dynamic learning model." Algorithms such as random forests can handle complex relationships involving multiple variables and nonlinearity, and can more accurately uncover the deep correlations between process parameters, equipment status, and ideal batch duration.

[0074] It achieves continuous adaptive optimization and prediction of theoretical runtime. The model can periodically learn from new normal production data and automatically adjust T_standard and Δ, enabling the system to keep up with slow changes in equipment performance, process improvements, or seasonal fluctuations in raw material characteristics, and always maintain the advanced nature of the judgment benchmark, achieving an intelligent effect of "becoming more accurate with use".

[0075] This represents a key step in batch management moving from "automation" to "intelligence," enabling continuous autonomous optimization of production processes.

[0076] In one embodiment, the system further includes an early warning and diagnosis module 7, which is used to output alarm signals for runtime anomalies, missing batch anomalies, and insufficient batch anomalies, and to locate the process position corresponding to the occurrence of the corresponding runtime anomaly, missing batch anomaly, or insufficient batch anomaly.

[0077] This module can immediately alert production control personnel when an anomaly occurs by outputting alarm signals in real time, shifting the problem-solving window from "post-event remediation" to "in-process correction", which greatly reduces the losses and correction costs caused by the continuous accumulation of batch errors or their spread downstream.

[0078] The module not only issues alerts but also pinpoints the specific process where the anomaly occurred (e.g., explicitly pointing to the "re-drying process" rather than the general "production line anomaly"). This allows maintenance personnel to skip tedious troubleshooting steps and go directly to the problem location to handle it, minimizing the disruption to the production process and significantly improving the availability and operational efficiency of the production system.

[0079] In one embodiment, the implementation process of each module is as follows: I. Data Acquisition and Processing Module 1; Algorithm support: Accurate batch start and end time extraction algorithm based on process flow signal.

[0080] Input: Real-time process flow signal F(t) (unit: kg / s or t / h) from each process of the production line (such as re-moistening, leaf removal outlet, etc.).

[0081] Processing steps: a. Signal filtering: Apply a first-order low-pass digital filter (such as a Butterworth filter) to the original flow signal F(t) to smooth high-frequency noise and obtain the filtered signal F. filt (t). b. Gradient calculation: Calculate F filt The discrete difference dF / dt of (t) is used to detect signal jumps. c. Change point detection: batch start time t start When F filt (t) jumps from below the threshold ϵ (close to zero, such as 0.05 times full scale) for 3 consecutive sampling points to above ϵ, and dF / dt exceeds the positive threshold θ. rise When the first point exceeds ϵ is denoted as t. start Batch end time t end When F filt (t) decreases from above ϵ for 3 consecutive sampling points to below ϵ, and dF / dt is below the negative threshold θ. fall When the first point below ϵ is denoted as t, end .

[0082] Output: The start and end times of each batch in this process (t) starti ,t endi ), and calculate the actual runtime of the batch. .

[0083] Storage: Store Ti and its associated raw material grade, process code, and timestamp into the historical batch duration database.

[0084] II. Statistical Calculation Module 2; Algorithm support: Benchmark parameter calculation algorithm based on robust statistics.

[0085] Input: A set of n normal batch runtimes for a specific raw material grade in a specific process, retrieved from a historical database. .

[0086] Processing steps: a. Theoretical runtime Calculate: 1. Sort by value in ascending order. 2. If n is odd, It is the (n+1) / 2th value after sorting.

[0087] 3. If n is even, It is the arithmetic mean of the n / 2th and (n / 2)+1th values ​​after sorting.

[0088] b. Tolerance threshold Δ calculation: 1. Calculate the absolute deviation sequence: D={∣ |, | ∣,...,∣ ∣}。 2. Calculate the absolute median difference (MAD): Sort D in ascending order and take the median (calculation method is the same as above). 3. Calculate Δ = k × 1.4826 × MAD, where k is a preset multiple (usually 2 or 3).

[0089] Output: Baseline parameters for this grade and this process. And Δ.

[0090] Configuration distribution: The output baseline parameters are distributed to the local control unit of the corresponding process for real-time verification.

[0091] III. Timing Chain Management Module 3; Algorithm support: Distributed queue synchronization algorithm based on strong association between timestamp and batch number.

[0092] Input: Global batch partitioning event from MES (including the new batch number BatchID).

[0093] Batch start and end signals (including BatchID and precise timestamp) from the data acquisition modules of each process.

[0094] Processing Steps: a. Queue Structure: Maintain a production time-series queue for each process in memory and persistent storage. Each element in the queue is a four-tuple: (BatchID, StartTime, EndTime, Status), where Status can be Running or Completed. b. Enqueue Operation: When a process detects the start of a batch (BatchID), it generates a StartTime and creates or updates the status of that batch element in the queue to Running. c. Update and Dequeue Operations: When a batch (BatchID) is detected to have ended, it generates an EndTime, updates the EndTime of the corresponding element in the queue, and sets the status to Completed. Completed batch elements can be archived according to a strategy. d. Association and Synchronization: Using BatchID as the primary key, ensure that records with the same BatchID in different process queues are logically linked by timestamps, forming a cross-process time-series chain.

[0095] Output: Maintain a globally queryable, time-consistent batch production chain view.

[0096] Consistency guarantee: Distributed transactions or eventual consistency protocols are used to ensure that the records of batch events in each process queue are reliable and ordered.

[0097] IV. Real-time verification module 4; Algorithm support: Multi-dimensional parallel sliding window verification algorithm.

[0098] enter: Real-time status of each process sequence queue.

[0099] The actual running time T of the current batch in each process actual .

[0100] Baseline parameters (T) from statistical calculation module 2 standard ,Δ).

[0101] Processing steps (the following three verification threads are executed in parallel): a. Continuity verification thread: Upon receiving a batch end event for each process, immediately query its downstream process queue to check if there is a start record matching BatchID. If not, and the downstream process has already started with BatchID+2, then trigger a "missing batch" exception. b. Timing consistency verification thread: For each running batch, calculate T in real time. actual Set a sliding timer, when T... actual <T standard Triggering an "advanced anomaly" warning when T = −Δ; actual>T standard +Δ triggers a "Delay Anomaly" warning. c. Logical Integrity Verification Thread: Periodically (e.g., every second) scans all active BatchIDs to check if they are recorded in all necessary process queues of the preset process path (e.g., re-moistening → leaf trimming → re-drying, etc.) (status is Running or Completed). If a batch is missing in an intermediate or final process, a "Missing Batch Anomaly" is triggered.

[0102] Output: Real-time generation of exception events, including exception type, BatchID, process position, and timestamp.

[0103] V. Correction Module 5; Algorithm support: Intelligent error correction decision and execution algorithm based on state machine.

[0104] Input: An exception event from Real-time Validation Module 4.

[0105] Processing steps (state machine driven): a. Exception classification and state initialization: Initialize the corrective state machine according to the exception type. b. Corrective decision: For "advanced exception" / "delayed exception": the state machine enters a waiting adjustment state. The system overrides the original fixed threshold switching logic of the MES and sets a new target end time: Advanced exception: Target end time = StartTime + T standard −Δ; Abnormal Delay: Target End Time = StartTime + T standard +Δ; Continuous monitoring continues until CurrentTime ≥ the target end time before issuing a batch switching instruction to that process. For "missing batch anomaly": The state machine enters logic repair mode. It identifies the missing batch number (MissingID), modifies the timing queue of downstream processes, and corrects the logical associations of all subsequent related batches. For "missing batch anomaly": The state machine enters path completion mode. In the queue of processes missing this batch, it interpolates and generates a record for that batch, the duration of which can be determined based on the process's T. standard Alternatively, the duration of adjacent batches can be inferred, and the timestamp chain of all subsequent related batches can be recalculated and adjusted. c. Instruction issuance and confirmation: The corrective decision is transformed into specific control instructions (such as: modifying the batch number, sending a batch switching command, updating the queue record), issued to the corresponding MES service or process controller, and awaits execution confirmation.

[0106] Output: Corrective execution instructions; updated, logically consistent batch timing chain.

[0107] VI. Optimize Module 6; Algorithm support: Adaptive optimization algorithm for baseline parameters based on random forest.

[0108] Input: An expanded historical dataset, each sample containing: the actual run time T of the batch, and a set of feature variables, such as raw material grade, ambient temperature and humidity, average values ​​of key equipment parameters (such as current and speed), and the duration of the preceding process.

[0109] Processing steps: a. Feature engineering and data preparation: Clean and normalize historical data to construct the feature matrix X and the target variable y (i.e., T). b. Model training: 1. Use a random forest regressor. Set hyperparameters such as the number of decision trees (e.g., 100 trees) and maximum depth. 2. Divide the data into training and testing sets. Train the model on the training set to learn the complex mapping relationship from multidimensional features X to batch duration y. c. Prediction and optimization: 1. For the current production conditions (given by real-time features X...) current (Representation), using the trained model to predict an "estimated value of the expected optimal runtime" T. pred 2. Put this T pred Compared with the currently used T standard Perform weighted fusion to generate a new theoretical runtime: T standard−new =α×T pred +(1−α)×T standard (α is the learning rate, such as 0.1). 3. Based on T pred The prediction error distribution is then used to recalculate MAD and Δnew.

[0110] Output: Optimized baseline parameters (T) standard−new ,Δ new ).

[0111] Triggering mechanism: The optimization process is automatically triggered periodically (e.g., every 100 batches produced) or when a significant drift in production conditions is detected.

[0112] 7. Early Warning and Diagnosis Module 7; Algorithm support: Algorithm for alarm generation and intelligent distribution based on multi-source information fusion.

[0113] enter: The original exception event from Real-time Verification Module 4.

[0114] Real-time device parameter snapshot from MES.

[0115] Preset alarm rules and priority matrix.

[0116] Processing Steps: a. Alarm Information Enrichment: 1. Receive abnormal events (type, BatchID, process, time). 2. Automatically query key equipment parameters (such as motor current, valve opening, temperature deviation) for the process within a period before and after the abnormal time window, and attach them as "contextual data" to the alarm information. b. Priority Determination: Assign priority to alarms according to preset rules. For example, "missing batches" may cause a break in the traceability chain, so the priority is "urgent"; "delay" may affect scheduling, so the priority is "high". c. Multi-channel Distribution: 1. Central Dashboard: Push highlighted alarm information, including text and sound prompts. 2. Workstation Terminals: Based on the process location, push alarms to the relevant operator's mobile terminal or workstation computer. 3. Logs and Reports: Write complete alarm records (including abnormal events and contextual data) to the database, which can be used as part of the "Batch Division Effect Report".

[0117] Output: Structured and actionable alarm notifications pushed to different human-machine interfaces; storage of complete diagnostic logs.

[0118] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for batch division of leaves after re-drying, characterized in that, Includes the following steps: S1. Collect the actual running time of each process in multiple batches of leaf re-drying as historical data, and calculate the theoretical running time of each process based on the historical data; S2. Calculate the tolerance threshold for the batch runtime of the corresponding process based on the theoretical runtime of each process. The range of the batch runtime is within ;in, This refers to the theoretical runtime. This is the tolerance threshold; S3. Establish a corresponding production time sequence queue for each process of the current leaf re-drying batch, associate the timestamp of the production time sequence queue with the current leaf re-drying batch number, execute steps S1 and S2 for each process, and obtain the running time corresponding to each process. S4. Detect whether the actual running time of each process in the current batch is within the range of the batch running time based on the batch running time. If not, determine that the running time is abnormal. S5. Regarding the runtime anomaly, if the batch runtime of each process in the current batch is less than... Time is defined as the advance runtime; the batch runtime of each process in the current batch is greater than 100%. Time is defined as the duration of the stay. S6. When the process exceeds the allotted time, stop dividing the current batch number and postpone the timestamp of the process until the batch corresponding to that process reaches the allotted time. Then switch to the next batch number; When the process experiences a runtime delay, the division of the current batch number is stopped, and the timestamp of the process is adjusted forward until the runtime of the batch corresponding to that process reaches [a certain value]. Then switch to the next batch number.

2. The batch division method according to claim 1, characterized in that, The theoretical runtime is obtained using the following formula: in, ; The set of data representing the actual runtime of the same process in n batches of the historical data; for The median of the dataset.

3. The batch division method according to claim 2, characterized in that, In step S2, the tolerance threshold is obtained by the following formula: Where k is a set multiple, and is a constant; This represents the absolute median difference.

4. The batch division method according to claim 1, characterized in that, In step S4, it is also included to detect in real time whether the two adjacent batch number sequences of each process are consecutive. If not, and the missing batch number does not appear in the batch of the previous process, it is determined to be a missing batch abnormality. In real time, it is detected whether the sequences of two adjacent batch numbers in each process are consecutive. If not, and the missing batch number has appeared in the batch of the previous process, it is determined to be a missing batch anomaly.

5. The batch division method according to claim 4, characterized in that, For the aforementioned batch missing anomaly, the later batch number in two adjacent batch numbers is corrected to restore the continuity of the batch sequence number; For the aforementioned batch shortage anomaly, the corresponding batch runtime is filled into the missing process position, and the timestamps of all subsequent processes are adjusted accordingly.

6. The batch division method according to claim 5, characterized in that, In step S4, alarm signals are output according to the runtime anomaly, the missing batch anomaly, and the few batch anomaly, and the process position corresponding to the occurrence of the corresponding runtime anomaly, the missing batch anomaly, and the few batch anomaly is located.

7. A batch division system for leaf re-drying, used to implement the division method as described in claim 6; characterized in that, include: The data acquisition and processing module is used to collect the actual running time of each process in multiple batches of leaf re-drying, and to calculate the theoretical running time of each process. The statistical calculation module is used to calculate the batch runtime of each process, wherein the value of the batch runtime is within a certain range. ;in, This refers to the theoretical runtime. This is the tolerance threshold; The time-series management module is used to establish corresponding production time-series queues for each process of the current leaf-picking and re-drying batch, and associate the timestamps of the production time-series queues with the current leaf-picking and re-drying batch number one by one. The real-time verification module is used to determine whether the actual runtime of each process is abnormal. For abnormal runtimes, if the batch runtime of each process in the current batch is less than [a certain value], [further verification is required]. Time is defined as the advance runtime; the batch runtime of each process in the current batch is greater than 100%. Time is defined as the duration of the stay. The correction module is used to adjust the timestamp of the process based on the advanced runtime or the delayed runtime until the corresponding batch runtime of that process reaches the specified value. or Then switch to the next batch number.

8. The batch division system according to claim 7, characterized in that, It also includes an optimization module, which processes the historical data using a random forest algorithm to adjust the calculated theoretical runtime of each process and the tolerance threshold.

9. The batch division system according to claim 1, characterized in that, It also includes an early warning and diagnosis module, which is used to output alarm signals for the runtime anomaly, the missing batch anomaly, and the few batch anomaly, and to locate the process position corresponding to the corresponding runtime anomaly, the missing batch anomaly, and the few batch anomaly.

Images