Tobacco industry production regulation method and system for reducing cost and increasing efficiency in silk rolling and packaging

Abstract

The application discloses a production regulation and control method and system for cost reduction and efficiency increase of tobacco industry silk making and baling, relates to the technical field of data processing, and acquires and pre-processes silk making multi-line equipment operation data and baling single machine raw and auxiliary material metering data into a time sequence characteristic matrix; adjusts silk making rhythm and material distribution through a dynamic load index and a load balancing model, guarantees continuous and stable upstream incoming material, compares time-sharing instantaneous unit consumption with a self-adaptive standard consumption interval, identifies baling material consumption abnormalities in real time, fits a working condition-material consumption mapping relationship through a gradient boosting regression tree, generates a cooperative regulation and control instruction set, realizes two-way parameter adjustment under global constraints of finished product qualification rate, distinguishes chain disturbance traceability results and baling local equipment abnormalities through Granger causality test, accurately locates cross-process implicit loss sources, and predicts working condition trends through a long short-term memory network and feeds back and updates a causality correlation strength matrix, so that closed-loop iteration of production regulation and control is realized.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a production control method and system for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging. Background Technology

[0002] The tobacco processing and packaging industry forms a tightly coupled chain production system. Multiple production lines operate in parallel within the processing workshop, each completing processes such as tobacco rehydration, feeding, drying, and blending according to its own pace. This outputs key quality indicators such as tobacco moisture content, filling value, and uniformity, which are then supplied to the packaging workshop in batches via a material conveying system. The packaging workshop, equipped with multiple rolling and packing machines, has extremely low tolerance for upstream material instability. Fluctuations in the cycle time of the processing, material interruptions, and deviations in tobacco quality parameters can propagate downstream along the production chain, leading to abnormal consumption of auxiliary materials such as splicing paper, filter rods, and small boxes, as well as fluctuations in the finished product qualification rate at the packaging end.

[0003] To ensure stable production throughout the entire process, it is necessary to deploy multi-line equipment operation data collection in the yarn making workshop and single-machine raw material metering data collection in the packaging workshop. Cross-process production control can be achieved through collaborative analysis of the two types of data.

[0004] Existing cross-process production control technologies mainly rely on single quality index correction or single parameter stability control. They achieve work order error prevention and batch matching through data comparison, but have not established a causal coupling relationship between yarn-making process parameters and winding and packaging material consumption indicators.

[0005] When there is excessive consumption of auxiliary materials or fluctuations in the finished product qualification rate at the packaging end, only the abnormal data can be identified, but the specific cause of the fluctuation in the upstream yarn production conditions cannot be located. Hidden losses across processes occur repeatedly, which restricts cost reduction and efficiency improvement throughout the entire process. Summary of the Invention

[0006] The main purpose of this application is to provide a production control method and system for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging, in order to solve the problem that in the existing cross-process production control technology, when there is excessive consumption of auxiliary materials or fluctuation in the finished product qualification rate at the packaging end, it can only identify data anomalies but cannot locate the specific cause of fluctuations in the upstream manufacturing conditions, resulting in repeated hidden losses across processes and restricting cost reduction and efficiency improvement throughout the entire process.

[0007] To achieve the above objectives, this application provides the following technical solution: A production control method for synergistic cost reduction and efficiency improvement in tobacco processing and packaging, wherein the production control method is applied to multi-line equipment in the tobacco processing workshop and single machines in the packaging workshop, and the production control method includes: Step S1: Collect the operating data of multi-line equipment in the yarn making workshop and the metering data of raw and auxiliary materials of single machine in the packaging workshop. After preprocessing, obtain the time series feature matrix of yarn making conditions and the time series feature matrix of packaging material consumption. Step S2: Calculate the dynamic load index of each silk production line based on the silk production condition time sequence feature matrix, and adjust the cycle time and material allocation of each line through the load balancing model to obtain the silk production scheduling parameters. Step S3: Calculate the instantaneous unit consumption of raw and auxiliary materials for each single machine based on the time-series feature matrix of the package material consumption, and compare it with the adaptively updated standard consumption interval to obtain the abnormal labeling sequence of package material consumption. Step S4: Input the yarn processing condition time series feature matrix and the packaging material consumption time series feature matrix into the causal coupling model. With the finished product qualification rate as the global constraint, the mapping relationship between the yarn processing condition parameters and the packaging material consumption index is fitted by the gradient boosting regression tree to obtain the condition-material consumption causal correlation strength matrix and the collaborative control instruction set. Step S5: Perform time-series alignment and backtracking of the abnormal material consumption marker sequence of the package and the working condition-material consumption causal correlation strength matrix, and distinguish the chain disturbance tracing results from the local equipment abnormality of the package through Granger causality test to obtain the abnormal tracing classification result; Step S6: Based on the collaborative control instruction set and the anomaly tracing classification results, predict the future trend of the silk-making process time-series feature matrix through a long short-term memory network, and input the prediction results into the causal coupling model to generate production control instructions.

[0008] Beneficial effects of steps S1 to S6: This method addresses the technical problem of existing cross-process production control technologies failing to establish a causal coupling relationship between yarn-making operating parameters and winding and packaging material consumption indicators, and the recurrence of hidden losses across processes. By collecting data from both yarn-making and winding and packaging ends and constructing a causal coupling model, the method uses the finished product qualification rate as a global constraint to fit the mapping relationship between operating parameters and material consumption indicators, thereby achieving closed-loop iteration of upstream and downstream collaborative control and anomaly tracing, and curbing the transmission and recurrence of chain production losses from the source.

[0009] Step S1 involves collecting operating data from multi-line spinning equipment and raw material metering data from single-machine bundling, and preprocessing them into a time-series feature matrix to provide a unified data foundation for cross-process collaborative analysis. Step S2 uses a dynamic load index and load balancing model to adjust the spinning cycle and material allocation, ensuring continuous and stable upstream material supply. Step S3 compares time-sharing instantaneous unit consumption with adaptive standard consumption intervals to identify bundling material consumption anomalies in real time. Step S4 uses a gradient boosting regression tree to fit the working condition-material consumption mapping relationship and generate a collaborative control instruction set, achieving bidirectional parameter adjustment under the global constraint of finished product qualification rate. Step S5 uses Granger causality tests to distinguish between chain disturbance tracing results and local bundling equipment anomalies, accurately locating the root cause of hidden losses across processes. Step S6 uses a long short-term memory network to predict working condition trends and update the causal correlation strength matrix, achieving closed-loop iteration of production control.

[0010] As a further improvement to this application, step S1 involves collecting operating data of multi-line equipment in the yarn-making workshop and metering data of raw and auxiliary materials for single machines in the packaging workshop. After preprocessing, the time-series characteristic matrix of yarn-making conditions and the time-series characteristic matrix of packaging material consumption are obtained, including: Step S1.1: Collect the equipment operating efficiency, process progress, material conveying capacity, and start-up / stop frequency of each production line in the silk-making workshop to obtain the original silk-making working condition data sequence; Step S1.2: Collect metering data of paper loading, filter rods, small boxes, inner lining paper, and strip boxes from each individual machine in the roll packaging workshop to obtain the original material consumption data sequence of roll packaging; Step S1.3: Perform missing value imputation and 3σ anomaly filtering on the original yarn making working condition data sequence and the original package material consumption data sequence respectively to obtain yarn making cleaning working condition data and package cleaning material consumption data. Step S1.4: Perform Min-Max standardization on the yarn making and cleaning condition data and the package cleaning material consumption data respectively to obtain standardized yarn making condition data and standardized package material consumption data. Step S1.5: Extract statistical features from the standardized yarn processing data and the standardized packaged material consumption data according to a preset time window to obtain the time-series feature matrix of yarn processing and the time-series feature matrix of packaged material consumption.

[0011] Beneficial effects of steps S1.1 to S1.5: This series of steps addresses the technical issues of inconsistent data acquisition formats at both the yarn processing and winding / bundling ends, and the impact of missing and outlier values ​​on subsequent collaborative analysis. By separately collecting yarn processing condition data and winding / bundling material consumption data and sequentially performing missing value imputation, outlier filtering, standardization, and statistical feature extraction, a time-series feature matrix with a unified structure is provided for cross-process collaborative control.

[0012] Specifically, step S1.1 involves collecting data on the operating efficiency, process progress, material conveying capacity, and start-stop frequency of each production line in the yarn-making process to construct the original operating condition data sequence; step S1.2 involves collecting raw and auxiliary material metering data for each individual machine in the yarn-making and packaging process to construct the original material consumption data sequence for the yarn-making and packaging process; step S1.3 involves eliminating the interference of missing data and outliers on subsequent processing through missing value imputation and 3σ anomaly filtering; step S1.4 involves eliminating the differences in the dimensions of each indicator through Min-Max standardization to make the features of different dimensions comparable; and step S1.5 involves extracting statistical features according to time windows to transform the original sequence into a structured time-series feature matrix.

[0013] As a further improvement to this application, step S2 involves calculating the dynamic load index of each silk-making production line based on the silk-making operating condition time-series characteristic matrix, and adjusting the cycle time and material allocation of each line through a load balancing model to obtain silk-making production scheduling parameters, including: Step S2.1: Extract the load congestion degree, process progress deviation, and material conveying margin ratio of each filament production line based on the filament production condition time sequence feature matrix, and calculate the dynamic load index. Step S2.2: Input the dynamic load index into the load balancing model to calculate the cycle adjustment coefficient and material allocation ratio of each yarn production line; Step S2.3: Fine-tune the operating speed of each yarn production line according to the cycle adjustment coefficient, and adjust the material conveying rate according to the material distribution ratio to obtain yarn production scheduling parameters; Step S2.4: Calculate the probability of material shortage risk based on the material conveying margin and process cycle difference in the filament making time sequence feature matrix. Step S2.5: Generate a diversion instruction for the yarn production line where the probability of material shortage exceeds a preset threshold, and incorporate the diversion instruction into the yarn production scheduling parameters.

[0014] Beneficial effects of steps S2.1 to S2.5: This series of steps addresses the technical problem of load imbalance and material shortage shutdowns caused by fixed-cycle scheduling in the yarn-making workshop. It assesses the operating status of each line through dynamic load index and adjusts the cycle time and material allocation through a load balancing model. At the same time, it predicts material risks and triggers diversion based on the material balance and cycle time difference to ensure continuous and stable upstream material supply.

[0015] Specifically, step S2.1 calculates the dynamic load index by extracting load congestion, process progress deviation, and material conveying margin ratio to quantify the operating status of each line; step S2.2 calculates the cycle adjustment coefficient and material allocation ratio using the load balancing model to provide a basis for dynamic balancing control; step S2.3 fine-tunes the operating speed and adjusts the conveying rate using the cycle adjustment coefficient and material allocation ratio to obtain the yarn production scheduling parameters; step S2.4 calculates the material shortage risk probability by the difference between the material conveying margin and the process cycle, and identifies the material supply gap in advance; step S2.5 generates diversion instructions for production lines whose material shortage risk probability exceeds the threshold to avoid passive shutdown and material shortage.

[0016] As a further improvement to this application, step S3 calculates the instantaneous unit consumption of raw and auxiliary materials for each individual machine based on the time-series feature matrix of the package material consumption, and compares it with the adaptively updated standard consumption interval to obtain the package material consumption anomaly marker sequence, including: Step S3.1: Extract the raw material consumption and finished product output of each individual machine based on the time-series feature matrix of the package material consumption, and calculate the time-sharing instantaneous unit consumption; Step S3.2: Calculate the rolling mean and rolling standard deviation of each raw material category based on historical time-sharing instantaneous unit consumption data, and construct an adaptively updated standard consumption interval; Step S3.3: Compare the time-division instantaneous unit consumption with the standard consumption range, and mark the unit consumption data points that exceed the upper and lower boundaries of the range with an abnormal label; Step S3.4: Summarize the abnormal tags of each individual machine in chronological order to obtain the pack material consumption abnormal tag sequence.

[0017] Beneficial effects of steps S3.1 to S3.4: This series of steps addresses the technical problem that the post-production statistics of raw and auxiliary materials in the packaging workshop cannot capture abnormal consumption in real time. By calculating the instantaneous unit consumption of each machine at different times and comparing it with the adaptively updated standard consumption range, it realizes the real-time marking and serialized output of material consumption anomalies.

[0018] Specifically, step S3.1 calculates the instantaneous unit consumption by extracting the consumption of raw and auxiliary materials and the output of finished products, thus quantifying the material consumption level of a single machine; step S3.2 constructs a standard consumption interval by using the rolling average and rolling standard deviation of historical unit consumption data, thereby achieving adaptive updating of the benchmark; step S3.3 accurately identifies excessive consumption data points by comparing the instantaneous unit consumption with the standard consumption interval and marking abnormal tags; and step S3.4 obtains a sequence of abnormal material consumption tags for packaging by summarizing the abnormal tags in chronological order, providing structured input for cross-process traceability.

[0019] As a further improvement of this application, in step S4, the time-series feature matrix of the spinning process and the time-series feature matrix of the packaging material consumption are input into the causal coupling model. Using the finished product qualification rate as a global constraint, the mapping relationship between the spinning process parameters and the packaging material consumption index is fitted using a gradient boosting regression tree to obtain the causal correlation strength matrix of the process and material consumption and the collaborative control instruction set, including: Step S4.1: Align and splice the yarn making process time sequence feature matrix and the package material consumption time sequence feature matrix according to the timestamp to obtain a cross-process joint feature matrix; Step S4.2: With the finished product qualification rate as the global constraint objective, construct a gradient boosting regression tree model between the yarn making process parameters and the winding and packaging material consumption index. Step S4.3: Input the cross-process joint feature matrix into the gradient boosting regression tree model to train and obtain the feature importance scores of each yarn processing condition parameter to each package material consumption index. Step S4.4: Assemble the feature importance scores into the working condition-material consumption causal correlation strength matrix; Step S4.5: Based on the working condition parameters whose scores in the working condition-material consumption causal correlation strength matrix exceed the preset threshold, generate corresponding yarn-making scheduling adjustment amounts and winding and packaging equipment parameter adjustment amounts, and assemble them into a collaborative control instruction set.

[0020] Beneficial effects of steps S4.1 to S4.5: This series of steps addresses the technical problem that existing cross-process control technologies fail to establish a causal coupling relationship between yarn-making operating parameters and packaging material consumption indicators, and that local parameter adjustments lead to global degradation. By using gradient boosting regression trees to fit the mapping relationship between operating parameters and material consumption indicators, and using the finished product qualification rate as a global constraint, a two-way collaborative control instruction set is generated to achieve the linkage adjustment of upstream and downstream parameters.

[0021] Specifically, step S4.1 involves aligning and splicing the yarn-making and bundling feature matrices according to timestamps to construct a cross-process joint feature matrix; step S4.2 involves constructing a gradient boosting regression tree model with the finished product qualification rate as the global constraint objective to establish a fitting framework for working conditions and material consumption; step S4.3 involves obtaining the feature importance scores of each yarn-making working condition parameter to each bundling material consumption index through training to quantify the cross-process coupling strength; step S4.4 involves assembling the feature importance scores into a causal correlation strength matrix to provide a structured basis for anomaly tracing; and step S4.5 involves generating yarn-making scheduling adjustment amounts and bundling equipment parameter adjustment amounts based on the working condition parameters whose scores in the causal correlation strength matrix exceed the threshold, and assembling them into a collaborative control instruction set.

[0022] As a further improvement to this application, step S5 involves performing time-series alignment and backtracking between the packing material consumption anomaly marker sequence and the working condition-material consumption causal correlation strength matrix, and using the Granger causality test to distinguish between chain disturbance tracing results and packing local equipment anomalies, thereby obtaining anomaly tracing classification results, including: Step S5.1: Align the abnormal material consumption marker sequence of the package with the working condition-material consumption causal correlation strength matrix according to the timestamp, and extract the yarn making working condition fluctuation events within a preset time window before and after the abnormal event; Step S5.2: Calculate the Granger causality test statistic for each pair of abnormal events and yarn-making condition fluctuation events; Step S5.3: Mark the silk-making process fluctuation events whose statistics exceed the significance threshold as chain disturbance tracing results; Step S5.4: Mark abnormal events whose statistics do not exceed the significance threshold as local device abnormalities in the package. Step S5.5: Summarize the chain disturbance tracing results and the local device anomaly of the package to obtain the anomaly tracing classification results.

[0023] Beneficial effects of steps S5.1 to S5.5: This series of steps addresses the technical problems of existing linkage technologies, which can only match data but cannot achieve cross-process causal traceability and the recurrence of hidden losses. By aligning abnormal bundling material consumption with fluctuations in yarn production conditions in a time sequence and using Granger causality testing, the steps accurately distinguish between chain disturbances and local bundling equipment abnormalities.

[0024] Specifically, step S5.1 involves aligning the material consumption anomaly marker sequence with the causal correlation strength matrix by timestamp to extract the yarn processing condition fluctuation events before and after the anomaly events, thus constructing a source tracing window; step S5.2 involves calculating the Granger causality test statistic for each pair of anomaly events and condition fluctuation events to quantify cross-process causal relationships; step S5.3 involves marking condition fluctuation events with statistic exceeding the significance threshold as chain disturbance source tracing results to locate the upstream yarn processing disturbance causes; step S5.4 involves marking anomaly events with statistic not exceeding the threshold as local packaging equipment anomalies to eliminate upstream interference misjudgments; and step S5.5 involves summarizing the chain disturbance source tracing results and local packaging equipment anomalies to obtain anomaly source tracing classification results, providing a classification basis for closed-loop control.

[0025] As a further improvement to this application, step S6, based on the collaborative control instruction set and the anomaly tracing classification results, predicts the future trend of the silk-making operating condition time-series feature matrix through a long short-term memory network, and inputs the prediction results into the causal coupling model to generate production control instructions, including: Step S6.1: Construct a joint input sequence based on the coordinated control instruction set and the anomaly tracing classification results; Step S6.2: Concatenate the joint input sequence with the silk-making condition time series feature matrix, input the sequence into the long short-term memory network for time series prediction, and obtain the silk-making condition prediction matrix. Step S6.3: Input the yarn processing condition prediction matrix into the causal coupling model to generate production control instructions; Step S6.4: Use the production control instructions and the collaborative control instruction set as feedback inputs to update the working condition-material consumption causal correlation strength matrix.

[0026] Beneficial effects of steps S6.1 to S6.4: This series of steps addresses the technical problems of lacking forward-looking prediction and closed-loop feedback updates in production control, and the inability to proactively address fluctuations in operating conditions. By using a long short-term memory network to predict future trends in yarn-making conditions and feeding the prediction results back to a causal coupling model, the dynamic generation of production control instructions and iterative updates of the causal correlation strength matrix are achieved.

[0027] Specifically, step S6.1 integrates control and tracing information by constructing a joint input sequence from the coordinated control instruction set and the anomaly tracing classification results; step S6.2 concatenates the joint input sequence with the silk-making condition time-series feature matrix and inputs it into a long short-term memory network to obtain the silk-making condition prediction matrix; step S6.3 inputs the silk-making condition prediction matrix into a causal coupling model to generate production control instructions; and step S6.4 updates the condition-material consumption causal correlation strength matrix as feedback input by using the production control instructions and the coordinated control instruction set as feedback inputs to achieve closed-loop iteration.

[0028] To achieve the above objectives, this application also provides the following technical solutions: A production control system for synergistic cost reduction and efficiency improvement in tobacco processing and packaging, wherein the production control system is applied to the production control method described above, and the production control system includes: The time-series feature matrix acquisition module is used to collect the operating data of multi-line equipment in the yarn making workshop and the metering data of raw and auxiliary materials of single machine in the packaging workshop. After preprocessing, the time-series feature matrix of yarn making conditions and the time-series feature matrix of packaging material consumption are obtained. The silk production scheduling parameter calculation module is used to calculate the dynamic load index of each silk production line based on the silk working condition time sequence feature matrix, and adjust the cycle time and material allocation of each line through the load balancing model to obtain the silk production scheduling parameters. The roll and package material consumption anomaly marker sequence acquisition module is used to calculate the instantaneous unit consumption of raw and auxiliary materials for each single machine according to the roll and package material consumption time series feature matrix, and compare it with the adaptively updated standard consumption interval to obtain the roll and package material consumption anomaly marker sequence. The mapping relationship fitting module is used to input the time series feature matrix of the yarn processing condition and the time series feature matrix of the packaging material consumption into the causal coupling model. With the finished product qualification rate as the global constraint, the mapping relationship between the yarn processing condition parameters and the packaging material consumption index is fitted by the gradient boosting regression tree to obtain the causal correlation strength matrix of the condition-material consumption and the collaborative control instruction set. The anomaly tracing and classification module is used to perform time-series alignment and backtracking of the pack material consumption anomaly marker sequence with the working condition-material consumption causal correlation strength matrix, and to distinguish between the chain disturbance tracing results and the pack local equipment anomalies through Granger causality test to obtain the anomaly tracing and classification results. The production control instruction generation module is used to predict the future trend of the silk-making working condition time series feature matrix based on the collaborative control instruction set and the anomaly tracing classification results, and input the prediction results into the causal coupling model to generate production control instructions.

[0029] To achieve the above objectives, this application also provides the following technical solutions: An electronic device includes a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the production control method described above for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging.

[0030] To achieve the above objectives, this application also provides the following technical solutions: A computer-readable storage medium storing program instructions, which, when executed by a processor, enable the production control method described above for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging. Attached Figure Description

[0031] Figure 1 This is a schematic flowchart illustrating one embodiment of a production control method for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging according to this application. Figure 2 This is a schematic diagram of the functional modules of a production control system for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging, according to an embodiment of this application. Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of this application; Figure 4 This is a schematic diagram of the structure of one embodiment of the storage medium of this application. Detailed Implementation

[0032] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0033] The terms "first," "second," and "third" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (e.g., as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive having. For example, a process, method, system, product, or device having a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0034] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be present in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0035] It should be noted that, due to the limited types and number of symbols or letters that can represent specific meanings, for embodiments with many formulas or codes, there may be situations where symbols or letters cannot meet the usage requirements. Therefore, the interpretation of formula symbols in the steps or sub-steps of the embodiments is only valid for the current step or sub-step.

[0036] If the same symbol has different interpretations in different steps or sub-steps, the interpretation in the current step or sub-step shall prevail; if the same symbol appears in different steps or sub-steps, but no interpretation is given in subsequent steps or sub-steps after its first appearance, the interpretation in the first step or sub-step shall be used.

[0037] For example, "i=1,2,...,n" is a writing convention and a well-known meaning. If "i=1,2,...,n" has already appeared once in an embodiment and needs to be used again in formulas with different scenarios and meanings in the future, then the meaning of "i=1,2,...,n" should be interpreted according to the corresponding formulas. If "i=1,2,...,n" is not used for the first time and is changed to a symbol or letter with a non-well-known meaning, such as "p=1,2,...,q", to avoid repetition, it is more likely to cause confusion, ambiguity, and unclear problems.

[0038] like Figure 1 As shown, this embodiment provides an example of a production control method for synergistic cost reduction and efficiency improvement in tobacco processing and packaging. In this embodiment, the production control method is applied to multi-line equipment in the tobacco processing workshop and single machines in the packaging workshop.

[0039] Specifically, this production control method includes the following steps: Step S1: Collect the operating data of multi-line equipment in the yarn making workshop and the metering data of raw and auxiliary materials of single machine in the packaging workshop. After preprocessing, obtain the time series feature matrix of yarn making conditions and the time series feature matrix of packaging material consumption.

[0040] Furthermore, step S1 specifically includes the following steps: Step S1.1: Collect the equipment operating efficiency, process progress, material conveying capacity, and start-up / stop frequency of each production line in the silk-making workshop to obtain the original silk-making operating condition data sequence.

[0041] Preferably, the core of this step is to collect the operating data of multiple production lines in the silk-making workshop. By collecting data on the equipment operating efficiency, process progress, material conveying capacity, and start-up / stop frequency of each production line, the original operating conditions are provided for subsequent preprocessing and time-series feature extraction.

[0042] In this context, the tobacco processing workshop refers to the production area in the tobacco industry where processes such as tobacco rehydration, feeding, drying, and blending are completed. The total number of production lines operating in parallel within the tobacco processing workshop is denoted as [missing information]. , No. The production line number is recorded as , .

[0043] Among them, equipment operating efficiency refers to the ratio of the actual output of the silk-making production line to its theoretical capacity within the statistical period. The equipment operating efficiency of each production line is denoted as value range .

[0044] Among them, process progress refers to the deviation ratio of the current process completion status of each production line from the planned progress. The process progress of each production line is denoted as value range Positive values ​​indicate leading and negative values ​​indicate lagging.

[0045] Among them, the material conveying surplus refers to the proportion of the current amount of tobacco residue in the buffer bin of the tobacco processing production line to the buffer bin capacity. The material conveying margin of each production line is denoted as value range .

[0046] Among them, the start-stop frequency refers to the number of unplanned shutdowns and restarts of the silk production line within the statistical period. The start-up and shutdown frequency of each production line is recorded as follows: The unit is times per hour (times / h).

[0047] For the raw data sequence of yarn production, the above four types of indicators are arranged by timestamp, and the first... The production line at the The original operating condition data at each sampling time point are denoted as follows: The original working condition data sequence of the yarn making process is denoted as ,in This represents the total number of sampling periods for silk production data.

[0048] Preferably, the total number of silk production lines The sampling interval is seconds, total number of sampling periods (correspond (hours), equipment operating efficiency Process progress Material conveying margin Start-stop frequency times / hour.

[0049] For example, a silk processing workshop , No. The production line at the Equipment operating efficiency at each sampling time Process progress Material conveying margin Start-stop frequency The frequency is per hour, and the corresponding device identifier is Line-P-YYYY-MM-DD-001.

[0050] Step S1.2: Collect metering data of paper loading, filter rods, small boxes, inner lining paper, and strip boxes from each individual machine in the roll packaging workshop to obtain the original material consumption data sequence of roll packaging.

[0051] Preferably, the core of this step is to collect the metering data of raw and auxiliary materials for each machine in the packaging workshop. By collecting the metering data of the paper, filter rod, small box, inner liner paper and strip box of each machine, the original material consumption input is provided for subsequent preprocessing and time series feature extraction.

[0052] The cigarette rolling and packaging workshop refers to the production area in the tobacco industry where cigarettes are rolled, filled, and packaged. The total number of single machines operating within the cigarette rolling and packaging workshop is denoted as... , No. The unit number is recorded as , .

[0053] Tipping paper, in particular, refers to the special paper used in cigarette production to connect the filter tip to the cigarette. The paper loading metering data for a single machine is recorded as follows: The unit is meters (m).

[0054] Among them, filter rods refer to the rod-shaped materials used to filter cigarette smoke during cigarette production. The filter rod metering data for a single unit is recorded as follows: The unit is a branch.

[0055] Among them, "small box" refers to the inner cardboard box used to package cigarettes. The metering data of the small box of a single machine is recorded as follows: The unit is "one".

[0056] Among them, the inner lining paper refers to the inner packaging paper used for moisture retention and preservation inside the small box. The metering data of the inner lining paper for a single machine is recorded as follows: The unit is sheets.

[0057] Among them, "carton" refers to a strip-shaped cardboard box used for outer packaging. The carton metering data for a single unit is recorded as follows: The unit is "one".

[0058] For the raw material consumption data sequence of roll packaging, the metering data of the above five types of auxiliary materials are arranged by timestamp, the first... The single machine in the first The raw material consumption data at each sampling time point are denoted as follows: The original material consumption data sequence of the roll package is denoted as ,in This represents the total number of sampling periods for the volume package data.

[0059] Preferably, the total number of roll packaging machines The sampling interval is seconds, total number of sampling periods (correspond (hours), paper metering m, filter rod measurement Branches, small boxes for measurement One, inner lining paper measurement Zhang, carton measurement indivual.

[0060] For example, a roll packaging workshop , No. The single machine in the first The sampling paper at each sampling time m, filter rod Branch, small box One, inner lining paper Zhang, carton One, corresponding to the device identifier Pack-M-YYYY-MM-DD-001.

[0061] Step S1.3: Perform missing value imputation and 3σ anomaly filtering on the original yarn processing condition data sequence and the original packaged material consumption data sequence to obtain yarn processing cleaning condition data and packaged cleaning material consumption data.

[0062] Preferably, the core of this step is to complete missing value imputation and outlier filtering for two types of raw data sequences. This is achieved by performing missing value imputation and outlier filtering on the raw yarn processing data sequence and the raw bundling material consumption data sequence, respectively. Anomaly filtering eliminates missing data and outlier interference, resulting in clean operating condition and material consumption data.

[0063] Among them, for missing value imputation, when the original working condition data sequence of the yarn making process is... When a field is missing in a record, it is filled using linear interpolation of adjacent time periods from the same production line. The first production line The sampling time of the first sampling moment The fill value for field number is denoted as When the original material consumption data sequence of the roll package When a certain field of a record is missing, linear interpolation of adjacent time periods on the same machine is used to fill the missing field, and the filling method is the same as above.

[0064] Among them, for Anomaly filtering involves calculating the mean for each field in the raw silk-making process data sequence. and standard deviation When the value of a certain record satisfy Values ​​identified as outliers are removed, and the mean value is calculated for each field in the original material consumption data sequence of the package. and standard deviation When the value of a certain record satisfy Values ​​that are identified as outliers are removed.

[0065] Among them, the data for yarn making and cleaning processes refers to the data after missing value imputation and... The fiber-making process data after abnormal filtration is denoted as ,in This refers to the total number of data periods for yarn production after cleaning; for data on material consumption during package cleaning, this refers to the data after missing value imputation and... The anomaly-filtered package material consumption data is denoted as ,in This represents the total number of data periods in the cleaned roll package.

[0066] Preferably, the missing rate of silk production data Volume package data missing rate Detection rate of abnormal values ​​in silk production Packaging outlier detection rate Linear interpolation fills in the error .

[0067] For example, in a certain silk-making workshop The equipment operating efficiency field of the production line is in the first... Missing sampling time, values ​​from adjacent time periods , Linear interpolation fill value ; a certain roll packaging workshop Average value of filter rod metering field per unit support, standard deviation branch, a certain record value Support satisfies Values ​​identified as outliers are removed.

[0068] Step S1.4: Perform Min-Max standardization on the yarn making and cleaning working condition data and the package cleaning material consumption data to obtain standardized yarn making working condition data and standardized package material consumption data.

[0069] Preferably, the core of this step is to complete the Min-Max standardization of the data after cleaning. By performing linear normalization on the data of yarn cleaning and the data of material consumption for package cleaning, the differences in the dimensions of each indicator are eliminated, so that the characteristics of different dimensions are comparable.

[0070] For Min-Max standardization, each field is mapped to... In the interval, the first in the silk-making and cleaning condition data The first production line The sampling time of the first sampling moment The standardized value of field number is denoted as ,in and The first The first production line The minimum and maximum values ​​of field number in the cleaned data; the number of data points in the roll package cleaning material consumption data. single-player machine The sampling time of the first sampling moment The standardized value of field number is denoted as ,in and The first single-player machine The minimum and maximum values ​​of the field after data cleaning.

[0071] Among them, the standardized operating condition data for silk production is denoted as For standardized material consumption data of roll packaging, it is denoted as All fields have a value range of 0.05. .

[0072] Preferably, the difference between the minimum and maximum values ​​of each field in the yarn washing process data is... To avoid division by zero, the difference between the minimum and maximum values ​​of each field in the roll package cleaning material consumption data. To avoid division by zero.

[0073] For example, in a certain silk-making workshop Production line equipment operating efficiency field , Cleaning value at a certain sampling time Standardized value ; a certain roll packaging workshop Unit filter rod metering field branch, Support, cleaning value at a certain sampling time Support, standardized value .

[0074] Step S1.5: Extract statistical features from the standardized working condition data of yarn making and the standardized material consumption data of winding and packaging according to the preset time window to obtain the time series feature matrix of yarn making and the time series feature matrix of winding and packaging material consumption.

[0075] Preferably, the core of this step is to extract the statistical features of the standardized data. By extracting the mean, variance, maximum value, minimum value, and skewness of the standardized working condition data of yarn making and the standardized material consumption data of packaging according to the preset time window, the original time series is transformed into a structured time series feature matrix.

[0076] Here, the preset time window refers to the time interval used for statistical feature calculation, and the window length is denoted as... The sliding step size is denoted as , No. The start time of each time window is denoted as The end time is recorded as ,in This is the initial time.

[0077] Among them, for statistical characteristics, the first The production line at the Within the first time window The mean of the number field is denoted as Variance is denoted as The maximum value is denoted as The minimum value is denoted as skewness is denoted as ,in For the first The set of sampling times within a time window The number of elements in the set.

[0078] For the time-series characteristic matrix of the silk-making process, the five statistical characteristics of each production line within each time window are arranged into a matrix, denoted as follows: ,in Number of fields for silk production conditions , The total number of yarn-making time windows; for the time-series characteristic matrix of packaged material consumption, denoted as... ,in For the number of material consumption fields in the package , This represents the total number of time windows for the package.

[0079] Preferably, the time window length seconds, sliding step seconds, total number of silk-making time windows (correspond (hours), total number of roll package time windows Extracting from each production line at each time window Each silk-making feature is extracted for each machine within each time window. Individual package characteristics.

[0080] For example, a silk processing workshop , , No. The production line at the The average value of the device operating efficiency field within a time window ,variance maximum value minimum value skewness Timing characteristic matrix of silk-making operation The dimension is A certain roll packaging workshop , Time-series feature matrix of material consumption of roll package The dimension is .

[0081] Beneficial effects of steps S1.1 to S1.5: This series of steps addresses the technical issues of inconsistent data acquisition formats at both the yarn processing and winding / bundling ends, and the impact of missing and outlier values ​​on subsequent collaborative analysis. By separately collecting yarn processing condition data and winding / bundling material consumption data and sequentially performing missing value imputation, outlier filtering, standardization, and statistical feature extraction, a time-series feature matrix with a unified structure is provided for cross-process collaborative control.

[0082] Specifically, step S1.1 involves collecting data on the operating efficiency, process progress, material conveying capacity, and start-stop frequency of each production line in the yarn-making process to construct the original operating condition data sequence; step S1.2 involves collecting raw and auxiliary material metering data for each individual machine in the yarn-making and packaging process to construct the original material consumption data sequence for the yarn-making and packaging process; step S1.3 involves eliminating the interference of missing data and outliers on subsequent processing through missing value imputation and 3σ anomaly filtering; step S1.4 involves eliminating the differences in the dimensions of each indicator through Min-Max standardization to make the features of different dimensions comparable; and step S1.5 involves extracting statistical features according to time windows to transform the original sequence into a structured time-series feature matrix.

[0083] Step S2: Calculate the dynamic load index of each silk production line based on the time sequence characteristic matrix of silk production conditions, and adjust the cycle time and material allocation of each line through the load balancing model to obtain the silk production scheduling parameters.

[0084] Furthermore, step S2 specifically includes the following steps: Step S2.1: Extract the load congestion degree, process progress deviation, and material conveying margin ratio of each filament production line based on the filament production condition time sequence feature matrix, and calculate the dynamic load index.

[0085] Preferably, the core of this step is to calculate the dynamic load index of the yarn production line. This is achieved by extracting three indicators from the yarn production condition time sequence feature matrix: load congestion, process progress deviation, and material conveying margin ratio, and then weighting and integrating them to quantify the operating status of each line.

[0086] Among them, load congestion degree refers to the ratio of the current material backlog in the buffer bin of the silk production line to the rated bin capacity. The production line at the The load congestion level of each time window is denoted as . ,in For the first The production line at the The average material conveying capacity within a time window, and the range of load congestion values. The larger the value, the more severe the congestion.

[0087] Among them, process schedule deviation refers to the absolute deviation between the actual process progress and the planned progress of the yarn production line. The production line at the The process progress deviation within a time window is denoted as... ,in For the first The production line at the The average progress of the process within a time window, and the range of process progress deviation. The larger the value, the more severe the deviation.

[0088] Among them, the material conveying allowance ratio refers to the standardized value of the average material conveying allowance of the silk production line. The production line at the The material conveying capacity ratio for each time window is denoted as: value range .

[0089] For the dynamic load index, the above three indicators are weighted and combined. The production line at the The dynamic load index for each time window is denoted as ,in , , For the preset weighting coefficients, satisfy The range of values ​​for the dynamic load index A larger value indicates a heavier load.

[0090] Preferably, the weighting coefficient , , The calculation period for the dynamic load index is consistent with the length of the time window.

[0091] For example, a silk processing workshop , No. The production line at the Load congestion within a time window Process progress deviation Material conveying margin ratio Dynamic load index .

[0092] Step S2.2: Input the dynamic load index into the load balancing model to calculate the cycle adjustment coefficient and material allocation ratio of each yarn production line.

[0093] Preferably, the core of this step is to complete the calculation of the load balancing model. By inputting the dynamic load index of each line into the load balancing model, the cycle adjustment coefficient and material allocation ratio are calculated with the goal of balancing the load of each line.

[0094] For the load balancing model, the objective is to minimize the variance of the dynamic load exponent for each line. The cycle time adjustment coefficient for each production line is denoted as: ,in This is the average of the dynamic load indexes of all production lines. This is the beat adjustment coefficient. Indicates acceleration, This indicates a slowdown.

[0095] Among them, regarding the material allocation ratio, the first The material allocation ratio of each production line is denoted as: That is, the material allocation is normalized according to the reciprocal of the dynamic load index of each line, and the production line with the lighter load receives a larger material allocation ratio.

[0096] Preferably, the beat adjustment coefficient Beat adjustment coefficient Material allocation ratio The sum of the material allocation ratios for each production line .

[0097] For example, a silk processing workshop , No. The dynamic load indices of the three production lines within each time window are as follows: , , mean , No. Production line cycle time adjustment coefficient Material allocation ratio .

[0098] Step S2.3: Fine-tune the operating speed of each yarn production line according to the cycle adjustment coefficient, and adjust the material conveying rate according to the material distribution ratio to obtain the yarn production scheduling parameters.

[0099] Preferably, the core of this step is to generate the silk production scheduling parameters. The operating speed of each line is fine-tuned by the cycle adjustment coefficient, and the material conveying rate is adjusted by the material distribution ratio. The results of the two adjustments are then encapsulated as the silk production scheduling parameters.

[0100] Among them, for fine-tuning the running speed, the first The adjusted operating speed of the production line is denoted as: ,in For the first The baseline operating speed of the production line, expressed in kilograms per hour (kg / h).

[0101] Among them, for the adjustment of material conveying rate, the first The adjusted material conveying rate of the production line is denoted as: ,in This refers to the total material conveying rate in the silk-making workshop, expressed in kilograms per hour (kg / h).

[0102] Specifically, for the silk production scheduling parameters, the adjusted operating speed and material conveying rate of each line are encapsulated into a scheduling parameter vector, denoted as . .

[0103] Preferably, the reference operating speed kg / h, total material conveying rate kg / h, the fine-tuning range of the operating speed does not exceed the benchmark value. .

[0104] For example, a silk processing workshop , No. Standard operating speed of production line kg / h, cycle time adjustment factor Adjusted running speed kg / h, material distribution ratio Adjusted material conveying rate kg / h, silk production scheduling parameters .

[0105] Step S2.4: Calculate the probability of material shortage risk based on the material conveying margin and process cycle difference in the filament processing time sequence feature matrix.

[0106] Preferably, the core of this step is to calculate the probability of material shortage risk. By extracting the material conveying margin and process cycle difference from the yarn making working condition time sequence feature matrix, the probability of material shortage occurring in each line within the current scheduling cycle is quantified.

[0107] Among them, the process cycle time difference refers to the difference in the average process progress between adjacent production lines, the first... The production line and the first The difference in cycle time between production lines is denoted as .

[0108] Among them, regarding the probability of material shortage risk, the first The production line at the The probability of material shortage risk within a time window is denoted as . ,in For the Sigmoid function, Material balance weight For the beat difference weight, As a bias term, the range of values ​​for the probability of material shortage risk is... A higher value indicates a higher risk of material shortage.

[0109] Preferably, material surplus weight Beat difference weight Bias term The calculation period for the probability of material shortage is consistent with the length of the time window.

[0110] For example, a silk processing workshop , No. The production line at the Average material conveying capacity within each time window , No. Article and No. The difference in cycle time between production lines Probability of material shortage .

[0111] Step S2.5: Generate a diversion command for the yarn production line where the probability of material shortage exceeds the preset threshold, and incorporate the diversion command into the yarn production scheduling parameters.

[0112] Preferably, the core of this step is to generate and incorporate the material shortage risk diversion instruction. By generating diversion instructions for production lines where the probability of material shortage risk exceeds a preset threshold and incorporating them into the yarn production scheduling parameters, passive shutdown due to material shortage can be avoided.

[0113] The preset threshold refers to the critical value of the probability of material shortage risk that triggers the diversion command, denoted as . ,when Time to determine the first There is a risk of material shortage on this production line.

[0114] Among these, the diversion instruction allocates some materials from production lines at risk of material shortage to production lines with lighter loads. The amount of material diverted from each production line is denoted as ,in The current splitting ratio coefficient is denoted as . ,in A collection of production lines at risk of material shortages.

[0115] Specifically, for the diversion command incorporated into the yarn production scheduling parameters, the diversion command is superimposed on the original scheduling parameters, and the updated yarn production scheduling parameters are denoted as follows: .

[0116] Preferably, a preset threshold Diversion ratio coefficient After diversion, the material conveying speed of each line shall not be lower than the reference operating speed. .

[0117] For example, a silk processing workshop , No. The first time window Probability of material shortage on production line Determined to be at risk of material shortage, material volume was diverted. kg / h, diversion command Updated silk production scheduling parameters .

[0118] Beneficial effects of steps S2.1 to S2.5: This series of steps addresses the technical problem of load imbalance and material shortage shutdowns caused by fixed-cycle scheduling in the yarn-making workshop. It assesses the operating status of each line through dynamic load index and adjusts the cycle time and material allocation through a load balancing model. At the same time, it predicts material risks and triggers diversion based on the material balance and cycle time difference to ensure continuous and stable upstream material supply.

[0119] Specifically, step S2.1 calculates the dynamic load index by extracting load congestion, process progress deviation, and material conveying margin ratio to quantify the operating status of each line; step S2.2 calculates the cycle adjustment coefficient and material allocation ratio using the load balancing model to provide a basis for dynamic balancing control; step S2.3 fine-tunes the operating speed and adjusts the conveying rate using the cycle adjustment coefficient and material allocation ratio to obtain the yarn production scheduling parameters; step S2.4 calculates the material shortage risk probability by the difference between the material conveying margin and the process cycle, and identifies the material supply gap in advance; step S2.5 generates diversion instructions for production lines whose material shortage risk probability exceeds the threshold to avoid passive shutdown and material shortage.

[0120] Step S3: Calculate the instantaneous unit consumption of raw and auxiliary materials for each single machine based on the time-series feature matrix of package material consumption, and compare it with the adaptively updated standard consumption interval to obtain the abnormal label sequence of package material consumption.

[0121] Furthermore, step S3 specifically includes the following steps: Step S3.1: Extract the raw material consumption and finished product output of each individual machine based on the time-series feature matrix of package material consumption, and calculate the time-sharing instantaneous unit consumption.

[0122] Preferably, the core of this step is to calculate the instantaneous unit consumption in time-sharing mode. By extracting the raw material consumption and finished product output of each machine from the time-series feature matrix of package material consumption, and calculating the ratio, the material consumption level of each machine is quantified.

[0123] Among them, the consumption of raw and auxiliary materials refers to the total amount of various raw and auxiliary materials consumed by each individual machine within a time window. The single machine in the first Within the first time window The consumption of auxiliary materials is recorded as follows: The unit depends on the type of auxiliary material (meter, piece, unit, sheet).

[0124] Among them, the output of finished products refers to the number of qualified finished products produced by each individual machine within a time window. The single machine in the first The output of finished products in each time window is denoted as: The unit is a branch.

[0125] Among them, for time-sharing instantaneous unit consumption, the first The single machine in the first Within the first time window The instantaneous unit consumption of auxiliary material is denoted as: This indicates the amount of time consumed in producing each qualified product. The quantity of auxiliary material, and the time-sharing instantaneous unit consumption of each auxiliary material in each individual machine are denoted as: .

[0126] Preferably, the calculation cycle and time window length of the time-sharing instantaneous unit consumption are... Consistent in seconds, finished product output To avoid division by zero.

[0127] For example, a roll packaging workshop , No. The single machine in the first Paper consumption within a time window m, finished product output Support, instantaneous single consumption of paper during loading and unloading m / piece.

[0128] Step S3.2: Calculate the rolling mean and rolling standard deviation of each raw material category based on historical time-sharing instantaneous unit consumption data, and construct an adaptively updated standard consumption interval.

[0129] Preferably, the core of this step is to construct the adaptive standard consumption interval, calculate the rolling mean and rolling standard deviation of each category using historical time-sharing instantaneous unit consumption data, and dynamically update the upper and lower boundaries of the standard consumption interval.

[0130] Among them, for the rolling mean, the first single-player machine No. 1 auxiliary material in The rolling mean of the time windows is denoted as . ,in This is the length of the scroll window.

[0131] Among them, for the rolling standard deviation, the first single-player machine No. 1 auxiliary material in The rolling standard deviation of each time window is denoted as . .

[0132] Among them, for the standard consumption interval of adaptive update, the first single-player machine No. 1 auxiliary material in The standard consumption interval of each time window is denoted as ,in This is the interval control coefficient, and the interval changes adaptively with the updates of the rolling mean and rolling standard deviation.

[0133] Preferably, the scroll window length Interval control coefficient The rolling mean and rolling standard deviation are updated at the end of each time window.

[0134] For example, a roll packaging workshop , No. The single-machine paper receiving unit is in the first Rolling mean of time windows m / piece, rolling standard deviation m / piece, standard consumption range m / piece.

[0135] Step S3.3: Compare the instantaneous unit consumption with the standard consumption range, and mark the unit consumption data points that exceed the upper and lower boundaries of the range as abnormal.

[0136] Preferably, the core of this step is to complete the abnormal material consumption marking. By comparing the time-sharing instantaneous unit consumption of each auxiliary material of each unit with the corresponding standard consumption range, abnormal labels are marked for unit consumption data points that exceed the upper and lower boundaries.

[0137] Among them, for comparison judgment, when or At that time, the judgment of the first single-player machine No. 1 auxiliary material in The unit consumption data points within each time window are abnormal.

[0138] For the abnormal label, it is denoted as When the unit consumption data points are normal When the unit consumption is lower than the lower boundary of the interval When the unit consumption is higher than the upper boundary of the interval .

[0139] Preferably, the calculation of anomaly labels is performed synchronously with the update of the standard consumption interval, and the set of anomaly labels for the current window is output at the end of each time window.

[0140] For example, a roll packaging workshop , No. The single-machine paper receiving unit is in the first Instantaneous unit consumption per time window m / piece, standard consumption range m / piece Exceeding the upper boundary, exception label .

[0141] Step S3.4: Summarize the abnormal tags of each individual machine in chronological order to obtain the pack material consumption abnormal tag sequence.

[0142] Preferably, the core of this step is to generate the abnormal label sequence of package material consumption. By summarizing the abnormal tags of each machine and each auxiliary material in chronological order, a structured abnormal label sequence is formed for subsequent cross-process tracing.

[0143] Specifically, for the abnormal material consumption marker sequence of roll packaging, the abnormal tags of each auxiliary material for each individual machine in each time window are arranged by timestamp and denoted as follows: The records corresponding to non-zero tags constitute a subset of abnormal events. .

[0144] Preferably, the sequence of abnormal packing material consumption markers is arranged in ascending order according to the time window, and the subset of abnormal events... The number of records does not exceed the total number of records. .

[0145] For example, a roll packaging workshop , , , No. The first time window was detected Paper loading abnormality on single machine , No. Single-unit filter rod malfunction Exception event subset In the Each time window contains 1 abnormal record, a sequence of abnormal material consumption markers for roll packaging. Total number of records is strip.

[0146] Beneficial effects of steps S3.1 to S3.4: This series of steps addresses the technical problem that the post-production statistics of raw and auxiliary materials in the packaging workshop cannot capture abnormal consumption in real time. By calculating the instantaneous unit consumption of each machine at different times and comparing it with the adaptively updated standard consumption range, it realizes the real-time marking and serialized output of material consumption anomalies.

[0147] Specifically, step S3.1 calculates the instantaneous unit consumption by extracting the consumption of raw and auxiliary materials and the output of finished products, thus quantifying the material consumption level of a single machine; step S3.2 constructs a standard consumption interval by using the rolling average and rolling standard deviation of historical unit consumption data, thereby achieving adaptive updating of the benchmark; step S3.3 accurately identifies excessive consumption data points by comparing the instantaneous unit consumption with the standard consumption interval and marking abnormal tags; and step S3.4 obtains a sequence of abnormal material consumption tags for packaging by summarizing the abnormal tags in chronological order, providing structured input for cross-process traceability.

[0148] Step S4: Input the time-series feature matrix of yarn processing conditions and the time-series feature matrix of packaging material consumption into the causal coupling model. With the finished product qualification rate as the global constraint, the mapping relationship between yarn processing condition parameters and packaging material consumption index is fitted by gradient boosting regression tree to obtain the causal correlation strength matrix of working conditions and material consumption and the coordinated control instruction set.

[0149] Furthermore, step S4 specifically includes the following steps: Step S4.1: Align and splice the yarn making process time sequence feature matrix and the package material consumption time sequence feature matrix according to the timestamp to obtain the cross-process joint feature matrix.

[0150] Preferably, the core of this step is to construct a cross-process joint feature matrix. This is achieved by aligning and splicing the yarn processing condition time-series feature matrix with the packaging material consumption time-series feature matrix according to timestamps, forming a cross-process joint feature matrix for subsequent causal modeling.

[0151] For timestamp alignment, since yarn production and bundling use the same time window length... and sliding step size The time windows of the two matrices correspond one-to-one, and the total number of aligned joint time windows is denoted as . .

[0152] Among them, for splicing, the time sequence feature matrix of the silk-making process is used. Feature vectors and time series feature matrices of material consumption for each production line and time window in China The feature vectors of each individual machine and each time window are concatenated according to the time window alignment. The cross-process joint feature vector of each time window is denoted as . The joint feature matrix across processes is denoted as , dimension .

[0153] Preferably, the total number of joint time windows The dimension of the cross-process joint feature vector is dimension.

[0154] For example, a silk processing workshop A certain roll packaging workshop , , No. The dimension of the silk-making feature vector for each time window is The dimension of the feature vector of the volume packet is . Dimension, joint feature vector across processes after splicing Dimensions dimension.

[0155] Step S4.2: With the finished product qualification rate as the global constraint objective, construct a gradient boosting regression tree model between the yarn making process parameters and the winding and packaging material consumption index.

[0156] Preferably, the core of this step is to construct a gradient boosting regression tree model, with the finished product qualification rate as the global constraint objective, and to establish a regression mapping framework for the yarn processing parameters on the material consumption index of the winding and packaging.

[0157] Among them, the finished product qualification rate refers to the rate at which the roll packaging workshop reaches the required qualification rate on the [number]th [day]. The proportion of qualified finished products to the total output of finished products within a time window is denoted as . value range .

[0158] Among them, for the gradient boosting regression tree model, the characteristics of the yarn processing conditions are taken as input, the material consumption index of the roll packaging is taken as output, and the finished product qualification rate is taken as the global constraint. The model is denoted as follows: ,in For the number of regression trees, For the first There are 10 regression trees, each fitted based on the residuals of the previous tree. The global constraint objective is to... The loss function is denoted as minimizing the material consumption prediction error under the given conditions. ,in The regularization coefficient is . For the first The structural complexity penalty term for each tree.

[0159] Preferably, the number of regression trees Learning rate Maximum tree depth Minimum number of leaf samples Regularization coefficient Lower limit of finished product qualification rate .

[0160] For example, the number of regression trees in a joint model for a certain silk roll packaging Learning rate , No. Tree of return Fit the initial residual, the first Tree of return Fitting The residuals are iterated to the th iteration. Trees, finished product qualification rate Satisfy global constraints.

[0161] Step S4.3: Input the cross-process joint feature matrix into the gradient boosting regression tree model to train and obtain the feature importance scores of each yarn processing condition parameter to each package material consumption index.

[0162] Preferably, the core of this step is to complete the training and extraction of feature importance scores. This is done by inputting the cross-process joint feature matrix into the gradient boosting regression tree model for training, and calculating the feature importance scores of each yarn processing condition parameter to each roll material consumption index.

[0163] Among them, for feature importance scores, the first The characteristics of the No. 1 silk-making working condition The characteristic importance score of the material consumption index of roll package is denoted as: ,in For the first The first of the 100 regression trees Feature number for the first The reduction in impurity of the output target, and the range of feature importance score values. The larger the value, the stronger the causal influence of the operating condition parameter on the material consumption index.

[0164] Preferably, the feature importance score matrix has the following dimensions: Training rounds Stop early and be patient. .

[0165] For example, a joint feature matrix Dimensions After training Characteristics of the No. 1 silk-making working condition (No. 1) (Average operating efficiency of equipment on production line) for the first No. 1 Packaging Material Consumption Index (No. 1) The characteristic importance score of the average paper loading per unit machine , No. Characteristics of the No. 1 silk-making working condition (No. 1) (Average material conveying allowance of production line) for the first The characteristic importance score of the material consumption index of roll pack .

[0166] Step S4.4: Assemble the feature importance scores into a working condition-material consumption causal correlation strength matrix.

[0167] Preferably, the core of this step is to assemble the causal correlation strength matrix of working conditions and material consumption. By assembling the characteristic importance scores of each yarn-making working condition feature to each package material consumption index into a matrix, a structured causal basis is provided for subsequent anomaly tracing and coordinated control.

[0168] For the causal relationship strength matrix of working conditions and material consumption, the feature importance scores are arranged into a matrix, denoted as follows: ,in As a dimension of silk-making process characteristics, For the dimension of material consumption characteristics of the package, the matrix No. Line number Column elements Indicates the first The characteristics of the No. 1 silk-making working condition The strength of the causal relationship between the material consumption characteristics of roll packs.

[0169] Preferably, the causal relationship strength matrix between operating conditions and material consumption The dimension is Range of values ​​for matrix elements The sum of elements in each row .

[0170] For example, the causal relationship strength matrix of a certain working condition and material consumption Dimensions , No. Line number Column elements Indicates the first The characteristics of the No. 1 silk-making working condition The causal correlation strength of the material consumption characteristics of roll packs is: , No. The sum of all elements in the row .

[0171] Step S4.5: Based on the working condition parameters whose scores in the working condition-material consumption causal correlation strength matrix exceed the preset threshold, generate the corresponding yarn making scheduling adjustment amount and the winding and packaging equipment parameter adjustment amount, and assemble them into a collaborative control instruction set.

[0172] Preferably, the core of this step is to generate a coordinated control instruction set. This is achieved by filtering out operating parameters whose scores exceed a preset threshold from the causal correlation strength matrix and generating adjustment amounts for yarn making scheduling and winding equipment parameters, which are then assembled into a coordinated control instruction set.

[0173] Among them, the preset threshold refers to the critical value of the feature importance score that triggers the control command, denoted as ,when Time to determine the first Characteristics of No. 1 silk-making working conditions and No. 1 There is a significant causal relationship between the material consumption characteristics of each roll and package.

[0174] Among them, for the silk-making scheduling adjustment amount, the first The scheduling adjustment amount for the characteristic of the No. 1 silk-making working condition is denoted as: ,in For adjusting the gain coefficient during fiber fabrication, In order to be with the first The set of package material consumption characteristics that exhibit a significant causal relationship with the characteristics of yarn production conditions. For the first The target value for the material consumption index.

[0175] Among them, for the parameter adjustment amount of the roll packaging equipment, the first The equipment parameter adjustment amount for the No. 1 roll / pack material consumption index is denoted as: ,in This is the gain coefficient for the roll package adjustment. .

[0176] Specifically, for the coordinated control instruction set, all significant causal relationships corresponding to the yarn-making scheduling adjustment and the winding and packaging equipment parameter adjustment are assembled into an instruction set, denoted as . .

[0177] Preferably, a preset threshold gain coefficient for fiber fabrication Packaging control gain coefficient Target value Take the midpoint of the standard consumption interval .

[0178] For example, a causal correlation strength matrix middle Determine the first Characteristics of No. 1 silk-making working conditions and No. 1 There is a significant causal relationship between the material consumption characteristics of roll package number 1 and roll package number 2. Current value of material consumption index m / branch, target value m / count, yarn production scheduling adjustment amount Adjustment amount of parameters for roll packaging equipment Coordinated control instruction set This includes the control instruction.

[0179] Beneficial effects of steps S4.1 to S4.5: This series of steps addresses the technical problem that existing cross-process control technologies fail to establish a causal coupling relationship between yarn-making operating parameters and packaging material consumption indicators, and that local parameter adjustments lead to global degradation. By using gradient boosting regression trees to fit the mapping relationship between operating parameters and material consumption indicators, and using the finished product qualification rate as a global constraint, a two-way collaborative control instruction set is generated to achieve the linkage adjustment of upstream and downstream parameters.

[0180] Specifically, step S4.1 involves aligning and splicing the yarn-making and bundling feature matrices according to timestamps to construct a cross-process joint feature matrix; step S4.2 involves constructing a gradient boosting regression tree model with the finished product qualification rate as the global constraint objective to establish a fitting framework for working conditions and material consumption; step S4.3 involves obtaining the feature importance scores of each yarn-making working condition parameter to each bundling material consumption index through training to quantify the cross-process coupling strength; step S4.4 involves assembling the feature importance scores into a causal correlation strength matrix to provide a structured basis for anomaly tracing; and step S4.5 involves generating yarn-making scheduling adjustment amounts and bundling equipment parameter adjustment amounts based on the working condition parameters whose scores in the causal correlation strength matrix exceed the threshold, and assembling them into a collaborative control instruction set.

[0181] Step S5: Time-series alignment and backtracking are performed between the abnormal material consumption marker sequence of the package and the causal correlation strength matrix of the working condition and material consumption. The Granger causality test is used to distinguish between the chain disturbance tracing results and the local equipment abnormality of the package, and the abnormal tracing classification results are obtained.

[0182] Furthermore, step S5 specifically includes the following steps: Step S5.1: Align the abnormal material consumption marker sequence of the package with the causal correlation strength matrix of working condition and material consumption according to the timestamp, and extract the yarn making working condition fluctuation events within the preset time window before and after the abnormal event.

[0183] Preferably, the core of this step is to complete the temporal alignment and backtracking extraction of abnormal events and yarn processing condition fluctuation events. By aligning the abnormal material consumption mark sequence of the roll and package with the causal correlation strength matrix according to the timestamp, the yarn processing condition fluctuation events within the preset time window before and after the abnormal event are extracted, and a source tracing and backtracking window is constructed.

[0184] Specifically, for timing alignment, a subset of abnormal events is selected. Each abnormal record Causal correlation strength matrix between working conditions and material consumption Associated by timestamp, where This is the time window number where the abnormal event occurred.

[0185] The preset time window refers to the time range for backtracking before and after the abnormal event, and the length of the backtracking window is denoted as [missing information]. The backtracking interval before the abnormal event is denoted as The backtracking interval after an abnormal event is denoted as .

[0186] Among them, the silk-making condition fluctuation event refers to the record in which the change amplitude of the silk-making condition characteristic value exceeds the fluctuation threshold within the traceback period, the first... The characteristics of the silk-making process in the first... The fluctuation within each time window is denoted as ,when Time to determine the first Feature number in the first Operating condition fluctuation events occurred within a time window, among which The set of extracted yarn-making condition fluctuation events is denoted as the fluctuation threshold. .

[0187] Preferably, the backtracking window length Time windows (corresponding to) (seconds), fluctuation threshold .

[0188] For example, a subset of abnormal events The Middle The abnormal record is That is, the first single-player machine No. 1 auxiliary material in An anomaly occurred within a certain time window, and the backtracking interval is... , No. The characteristics of the silk-making process in the first... Fluctuation within a time window This was determined to be an operating condition fluctuation event and extracted to... .

[0189] Step S5.2: Calculate the Granger causality test statistic for each pair of abnormal events and yarn-making condition fluctuation events.

[0190] Preferably, the core of this step is to calculate the Granger causality test statistic. By constructing an autoregressive model and a restricted model for each pair of abnormal events and silk-making condition fluctuation events, the Granger causality test statistic is calculated to quantify the causal relationship across processes.

[0191] For the autoregressive model, an abnormal event sequence is used as the dependent variable and a silk-making operating condition fluctuation event sequence is used as the independent variable, constructing a model that includes lag order. Full model The sum of squared residuals is denoted as .

[0192] For the restricted model, the restricted model is constructed using only the lag term of the anomalous event sequence itself as the independent variable. The sum of squared residuals is denoted as .

[0193] Among them, the Granger causality test statistic is denoted as . ,in For sample size, the statistic is... obey distributed.

[0194] Preferably, the lag order Sample size significance level .

[0195] For example, an abnormal event Fluctuations in silk production conditions This constitutes a pair of test objects, the sum of squared residuals of the entire model. Sum of squared residuals in a constrained model Granger causality test statistic .

[0196] Step S5.3: Mark the silk-making process fluctuation events whose statistics exceed the significance threshold as chain disturbance tracing results.

[0197] Preferably, the core of this step is to mark the chain disturbance source tracing results. By marking the silk-making condition fluctuation events whose Granger causality test statistic exceeds the significance threshold as chain disturbance source tracing results, the upstream silk-making disturbance causes can be located.

[0198] Among them, the significance threshold refers to Distribution at significance level The critical value below is denoted as ,when Time to determine the first The characteristic fluctuation of the No. 1 silk-making operating condition is the first Granger cause of abnormal material consumption in roll pack #1.

[0199] Specifically, for chain disturbance tracing results, operating condition fluctuation events determined to be Granger causes are marked as chain disturbance tracing results, denoted as... .

[0200] Preferably, the significance threshold .

[0201] For example, a certain test statistic Determine the first The characteristics of the silk-making process in the first... The fluctuation of the first time window is the first single-player machine No. 1 auxiliary material in The Granger causation of anomalies within a specific time window is marked as a result of chain perturbation tracing. .

[0202] Step S5.4: Mark abnormal events whose statistics do not exceed the significance threshold as local device abnormalities of the volume package.

[0203] Preferably, the core of this step is to mark the local equipment anomalies of the package. By marking abnormal events whose Granger causality test statistic does not exceed the significance threshold as local equipment anomalies of the package, upstream interference and misjudgment are eliminated.

[0204] Specifically, for local device anomalies in the package, anomalies not determined to be due to Granger malfunctions are marked as local device anomalies in the package, denoted as... ,in An abnormal event The set of silk-making feature numbers corresponding to all silk-making condition fluctuation events within the backtracking interval.

[0205] Preferably, the number of abnormal records on the local device of the package does not exceed a subset of abnormal events. Total number of records .

[0206] For example, an abnormal event The Granger test statistic for all yarn-making condition fluctuation events within the backtracking interval did not exceed [the specified value]. The abnormal event was determined to be a local device error and marked as such. .

[0207] Step S5.5: Summarize the chain disturbance tracing results and local device anomalies in the package to obtain the anomaly tracing classification results.

[0208] Preferably, the core of this step is to summarize and generate the anomaly tracing and classification results. By summarizing the chain disturbance tracing results and the local equipment anomalies of the package according to the anomaly event number, a classification result is formed for subsequent closed-loop control.

[0209] Among them, for the anomaly tracing and classification results, the chain perturbation tracing results will be... Local equipment malfunction related to roll package Summarize by abnormal event number and record as follows: Each record includes the abnormal event number, category, associated silk-making condition feature number, and time window number.

[0210] Preferably, the anomaly tracing and classification results are arranged in ascending order according to the time window, and the chain disturbance tracing results do not overlap with the local equipment anomalies of the package.

[0211] For example, a subset of abnormal events In the Each time window contains The results of chain perturbation tracing of the abnormal records after Granger causality test. Includes Record: Local device malfunction in the package. Includes Records, anomaly tracing and classification results .

[0212] Beneficial effects of steps S5.1 to S5.5: This series of steps addresses the technical problems of existing linkage technologies, which can only match data but cannot achieve cross-process causal traceability and the recurrence of hidden losses. By aligning abnormal bundling material consumption with fluctuations in yarn production conditions in a time sequence and using Granger causality testing, the steps accurately distinguish between chain disturbances and local bundling equipment abnormalities.

[0213] Specifically, step S5.1 involves aligning the material consumption anomaly marker sequence with the causal correlation strength matrix by timestamp to extract the yarn processing condition fluctuation events before and after the anomaly events, thus constructing a source tracing window; step S5.2 involves calculating the Granger causality test statistic for each pair of anomaly events and condition fluctuation events to quantify cross-process causal relationships; step S5.3 involves marking condition fluctuation events with statistic exceeding the significance threshold as chain disturbance source tracing results to locate the upstream yarn processing disturbance causes; step S5.4 involves marking anomaly events with statistic not exceeding the threshold as local packaging equipment anomalies to eliminate upstream interference misjudgments; and step S5.5 involves summarizing the chain disturbance source tracing results and local packaging equipment anomalies to obtain anomaly source tracing classification results, providing a classification basis for closed-loop control.

[0214] Step S6: Based on the collaborative control instruction set and the anomaly tracing classification results, predict the future trend of the silk-making process time-series feature matrix through a long short-term memory network, and input the prediction results into the causal coupling model to generate production control instructions.

[0215] Furthermore, step S6 specifically includes the following steps: Step S6.1: Construct a joint input sequence based on the coordinated control instruction set and the anomaly tracing classification results.

[0216] Preferably, the core of this step is to construct the joint input sequence. By aligning and integrating the coordinated regulation instruction set and the anomaly tracing classification results by timestamp, a joint input sequence containing regulation instruction information and anomaly tracing classification information is formed for subsequent LSTM model prediction.

[0217] Among them, the coordinated control instruction set refers to the one generated in step S4. This includes adjustments to the yarn-making schedule and the parameters of the packaging equipment within each time window. The coordinated control command vector for each time window is denoted as... ,in .

[0218] Among them, the anomaly tracing and classification results refer to those generated in step S5. It includes the chain disturbance tracing results for each time window and local device anomalies in the volume package. The anomaly tracing classification feature vector for each time window is denoted as . ,in and The first The number of chain disturbance tracing results and the number of local device anomalies within a time window. For the first Sum of the causal association strengths corresponding to the chain perturbation tracing results within a time window For the first Sum of the absolute values ​​of local device anomaly labels within a given time window.

[0219] Specifically, for constructing a joint input sequence, the instruction vector will be coordinated. and anomaly source tracing classification feature vector Attach the pieces according to timestamps, the first... The joint input vector of the time windows is denoted as The joint input sequence is denoted as .

[0220] Preferably, the dimension of the coordinated control command vector depends on The number of central control instructions and the dimension of the anomaly tracing classification feature vector are: The dimension of the joint input vector is . ,in To coordinate and control the command vector dimension.

[0221] For example, in a scenario of joint control of a certain silk roll packaging, the first Coordinated control instruction set for each time window Include One control instruction, coordinated control instruction vector Dimensions Dimensions (each instruction includes yarn preparation adjustment and package adjustment), anomaly tracing classification feature vector Dimensions dimension, joint input vector Dimensions Dimension, joint input sequence Length is .

[0222] Step S6.2: Concatenate the joint input sequence with the time series feature matrix of the silk-making process, input it into the long short-term memory network for time series prediction, and obtain the silk-making process prediction matrix.

[0223] Preferably, the core of this step is to generate the silk-making condition prediction matrix. This is achieved by concatenating the joint input sequence with the silk-making condition time-series feature matrix and inputting it into the LSTM model. The gating mechanism is then used to capture the time-series dependencies and predict the future trend of the silk-making condition time-series feature matrix.

[0224] For concatenation, the input sequence is combined. Joint input vector for each time window Timing characteristic matrix of yarn making operation Feature vectors corresponding to the time window splicing, the first The concatenated vector of the time windows is denoted as... The LSTM input sequence is denoted as .

[0225] For Long Short-Term Memory (LSTM) networks, there are three sets of gated structures: input gate, forget gate, and output gate. The forgetting gate at each time step is recorded as follows: The input gate is denoted as The output gate is denoted as Cell status updated to The hidden state has been updated to ,in For the Sigmoid function, For element-wise multiplication, , , These are the model parameters.

[0226] Specifically, for the yarn processing condition prediction matrix, the hidden state of the LSTM model at the final time step is mapped through a fully connected layer to the future trend prediction value of the yarn processing condition time series feature matrix for the next time period. The length of the prediction period is denoted as... The yarn-making condition prediction matrix is ​​denoted as , dimension .

[0227] Preferably, the LSTM hidden layer dimension Predicted time period length Time windows (corresponding to) (seconds), the fully connected layer activation function is a linear function, training epochs Batch size Learning rate Dropout ratio .

[0228] For example, a silk processing workshop LSTM input sequence Dimensions Hidden layer dimensions ,through The final hidden state after forward propagation at each time step Dimensions The dimension of the fully connected layer is mapped to a fiber fabrication condition prediction matrix. Dimensions Dimension, the Predicted operating efficiency of equipment on each production line .

[0229] Step S6.3: Input the yarn processing condition prediction matrix into the causal coupling model to generate production control instructions.

[0230] Preferably, the core of this step is to generate production control instructions. By inputting the LSTM-predicted yarn processing condition prediction matrix into the causal coupling model trained in step S4, forward-looking production control instructions are generated based on the condition-material consumption causal correlation strength matrix.

[0231] For the causal coupling model, the gradient boosting regression tree model trained in step S4 is used. The causal relationship strength matrix between working conditions and material consumption The yarn making condition prediction matrix Input Model Obtain predicted material consumption indicators .

[0232] For production control instructions, based on predicted material consumption indicators The deviation from the standard consumption range generates a control command, the first The prediction deviation of the material consumption index is denoted as... ,when Timely generation of control commands, among which To adjust the trigger threshold, the adjustment command is denoted as: ,in This is a sign function; a positive value indicates that material consumption needs to be reduced, and a negative value indicates that material supply needs to be increased.

[0233] Preferably, the trigger threshold is adjusted. The calculation cycle of the control instructions and the length of the forecast period Consistent.

[0234] For example, a prediction matrix for a certain silk-making process. After inputting the causal coupling model, the first... Predicted values ​​of material consumption index m / branch, current rolling average m / piece, prediction bias , Generate control instructions This indicates that the first [level] needs to be lowered. Material consumption level of auxiliary materials, control instructions Include this instruction.

[0235] Step S6.4: Use the production control instructions and collaborative control instruction set as feedback inputs to update the causal relationship strength matrix of operating conditions and material consumption.

[0236] Preferably, the core of this step is to complete the feedback update of the causal relationship strength matrix of working conditions and material consumption. By using production control instructions and collaborative control instruction sets as feedback inputs, the causal relationship strength is corrected by the feedback of the effect after the control is executed, thereby realizing closed-loop iterative optimization.

[0237] Among them, the feedback input will include production control instructions. The characteristic number of each control target and the coordinated control instruction set for the silk-making process. The characteristic numbers of the yarn-making conditions corresponding to each control command are merged, and the actual material consumption changes are collected after the control is executed in the next time period. The actual change in the material consumption index is denoted as: .

[0238] Specifically, for updating the causal relationship strength matrix of operating conditions and material consumption, the corresponding causal relationship strength is enhanced when the direction of the control command is consistent with the actual change direction, and weakened when the directions are inconsistent. The update formula is denoted as follows: ,in To provide feedback on the learning rate, the updated working condition-material consumption causal relationship strength matrix is ​​denoted as follows: .

[0239] Preferably, the feedback learning rate The feedback update cycle is consistent with the calculation cycle of the control command, and each element is updated afterward. Cut off to Interval.

[0240] For example, a production control instruction and coordinated control instruction set The corresponding number in the middle The control commands based on the characteristics of the yarn-making process serve as feedback input, and the actual material consumption change in the next period is... (Reduced material consumption), control direction With actual direction Inconsistent, before update After the update After being cut off (Minimum change), updated working condition-material consumption causal relationship strength matrix For coordinated regulation in the next period.

[0241] Beneficial effects of steps S6.1 to S6.4: This series of steps addresses the technical problems of lacking forward-looking prediction and closed-loop feedback updates in production control, and the inability to proactively address fluctuations in operating conditions. By using a long short-term memory network to predict future trends in yarn-making conditions and feeding the prediction results back to a causal coupling model, the dynamic generation of production control instructions and iterative updates of the causal correlation strength matrix are achieved.

[0242] Specifically, step S6.1 integrates control and tracing information by constructing a joint input sequence from the coordinated control instruction set and the anomaly tracing classification results; step S6.2 concatenates the joint input sequence with the silk-making condition time-series feature matrix and inputs it into a long short-term memory network to obtain the silk-making condition prediction matrix; step S6.3 inputs the silk-making condition prediction matrix into a causal coupling model to generate production control instructions; and step S6.4 updates the condition-material consumption causal correlation strength matrix as feedback input by using the production control instructions and the coordinated control instruction set as feedback inputs to achieve closed-loop iteration.

[0243] In summary, the overall beneficial effects of steps S1 to S6 in this embodiment are as follows: This method addresses the technical problem of existing cross-process production control technologies failing to establish a causal coupling relationship between yarn-making operating parameters and winding and packaging material consumption indicators, and the recurrence of hidden losses across processes. By collecting data from both yarn-making and winding and packaging ends and constructing a causal coupling model, the method uses the finished product qualification rate as a global constraint to fit the mapping relationship between operating parameters and material consumption indicators, thereby achieving closed-loop iteration of upstream and downstream collaborative control and anomaly tracing, and curbing the transmission and recurrence of chain production losses from the source.

[0244] Step S1 involves collecting operating data from multi-line spinning equipment and raw material metering data from single-machine bundling, and preprocessing them into a time-series feature matrix to provide a unified data foundation for cross-process collaborative analysis. Step S2 uses a dynamic load index and load balancing model to adjust the spinning cycle and material allocation, ensuring continuous and stable upstream material supply. Step S3 compares time-sharing instantaneous unit consumption with adaptive standard consumption intervals to identify bundling material consumption anomalies in real time. Step S4 uses a gradient boosting regression tree to fit the working condition-material consumption mapping relationship and generate a collaborative control instruction set, achieving bidirectional parameter adjustment under the global constraint of finished product qualification rate. Step S5 uses Granger causality tests to distinguish between chain disturbance tracing results and local bundling equipment anomalies, accurately locating the root cause of hidden losses across processes. Step S6 uses a long short-term memory network to predict working condition trends and update the causal correlation strength matrix, achieving closed-loop iteration of production control.

[0245] like Figure 2 As shown, this embodiment provides an example of a production control system for synergistic cost reduction and efficiency improvement in tobacco manufacturing and packaging. In this embodiment, the production control system is applied to the production control method described above.

[0246] Specifically, the production control system includes a time-series feature matrix acquisition module 1, a yarn production scheduling parameter calculation module 2, a roll and package material consumption anomaly marker sequence acquisition module 3, a mapping relationship fitting module 4, an anomaly tracing and classification module 5, and a production control instruction generation module 6, which are connected electrically or by communication in sequence.

[0247] The module 1, which acquires the time-series feature matrix, collects the operating data of multiple equipment lines in the spinning workshop and the raw material metering data of a single machine in the packaging workshop. After preprocessing, it obtains the time-series feature matrix of spinning conditions and the time-series feature matrix of packaging material consumption. The module 2, which calculates the dynamic load index of each spinning production line based on the time-series feature matrix of spinning conditions, adjusts the cycle time and material allocation of each line through a load balancing model to obtain the spinning production scheduling parameters. The module 3, which acquires the abnormal label sequence of packaging material consumption, calculates the instantaneous unit consumption of raw materials and auxiliary materials of each single machine based on the time-series feature matrix of packaging material consumption and compares it with the adaptively updated standard consumption interval to obtain the abnormal label sequence of packaging material consumption. The module 4, which fits the mapping relationship, combines the time-series feature matrix of spinning conditions with the time-series feature matrix of packaging material consumption. The time-series feature matrix is ​​input into the causal coupling model, with the finished product qualification rate as a global constraint. The mapping relationship between the spinning process parameters and the packaging material consumption index is fitted by a gradient boosting regression tree to obtain the causal correlation strength matrix of the process and material consumption and the collaborative control instruction set. The anomaly tracing and classification module 5 is used to perform time-series alignment and backtracking of the packaging material consumption anomaly marker sequence with the process-material consumption causal correlation strength matrix, and to distinguish between the chain disturbance tracing results and the local packaging equipment anomalies by Granger causality test to obtain the anomaly tracing and classification results. The production control instruction generation module 6 is used to predict the future trend of the spinning process time-series feature matrix based on the collaborative control instruction set and the anomaly tracing and classification results by using a long short-term memory network, and input the prediction results into the causal coupling model to generate production control instructions.

[0248] like Figure 3 As shown, the electronic device 7 includes a processor 71 and a memory 72 coupled to the processor 71.

[0249] The memory 72 stores program instructions for implementing the production control method for synergistic cost reduction and efficiency improvement in tobacco industry tobacco processing and packaging according to any of the above embodiments.

[0250] The processor 71 is used to execute program instructions stored in the memory 72 to perform production control for coordinated cost reduction and efficiency improvement in tobacco manufacturing and packaging.

[0251] The processor 71 can also be referred to as a CPU (Central Processing Unit). The processor 71 may be an integrated circuit chip with signal processing capabilities. The processor 71 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.

[0252] Furthermore, Figure 4 This is a schematic diagram of the structure of a storage medium according to an embodiment of this application. See also: Figure 4 The storage medium 8 in this embodiment stores program instructions 81 capable of implementing all the above methods. These program instructions 81 can be stored in the storage medium as a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods in each embodiment of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0253] In the embodiments provided in this application, it should be understood that the disclosed systems, methods, and approaches can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the production control of tobacco processing and packaging in a coordinated manner to reduce costs and increase efficiency is only a logical function. In actual implementation, there may be other production control methods for tobacco processing and packaging in a coordinated manner to reduce costs and increase efficiency. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or the indirect coupling or communication connection between systems or units may be electrical, mechanical, signal, or other forms.

[0254] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A production control method for synergistic cost reduction and efficiency improvement in tobacco processing and packaging, wherein the production control method is applied to multi-line equipment in the tobacco processing workshop and single machines in the packaging workshop, characterized in that, The production control method includes: Step S1: Collect the operating data of multi-line equipment in the yarn making workshop and the metering data of raw and auxiliary materials of single machine in the packaging workshop. After preprocessing, obtain the time series feature matrix of yarn making conditions and the time series feature matrix of packaging material consumption. Step S2: Calculate the dynamic load index of each silk production line based on the silk production condition time sequence feature matrix, and adjust the cycle time and material allocation of each line through the load balancing model to obtain the silk production scheduling parameters. Step S3: Calculate the instantaneous unit consumption of raw and auxiliary materials for each single machine based on the time-series feature matrix of the package material consumption, and compare it with the adaptively updated standard consumption interval to obtain the abnormal labeling sequence of package material consumption. Step S4: Input the yarn processing condition time series feature matrix and the packaging material consumption time series feature matrix into the causal coupling model. With the finished product qualification rate as the global constraint, the mapping relationship between the yarn processing condition parameters and the packaging material consumption index is fitted by the gradient boosting regression tree to obtain the condition-material consumption causal correlation strength matrix and the collaborative control instruction set. Step S5: Perform time-series alignment and backtracking of the abnormal material consumption marker sequence of the package and the working condition-material consumption causal correlation strength matrix, and distinguish the chain disturbance tracing results from the local equipment abnormality of the package through Granger causality test to obtain the abnormal tracing classification result; Step S6: Based on the collaborative control instruction set and the anomaly tracing classification results, predict the future trend of the silk-making process time-series feature matrix through a long short-term memory network, and input the prediction results into the causal coupling model to generate production control instructions.

2. The production control method according to claim 1, characterized in that, Step S1: Collect operating data of multi-line equipment in the yarn processing workshop and raw material metering data of single machine in the packaging workshop. After preprocessing, obtain the time-series feature matrix of yarn processing conditions and the time-series feature matrix of packaging material consumption, including: Step S1.1: Collect the equipment operating efficiency, process progress, material conveying capacity, and start-up / stop frequency of each production line in the silk-making workshop to obtain the original silk-making working condition data sequence; Step S1.2: Collect metering data of paper loading, filter rods, small boxes, inner lining paper, and strip boxes from each individual machine in the roll packaging workshop to obtain the original material consumption data sequence of roll packaging; Step S1.3: Perform missing value imputation and 3σ anomaly filtering on the original yarn making working condition data sequence and the original package material consumption data sequence respectively to obtain yarn making cleaning working condition data and package cleaning material consumption data. Step S1.4: Perform Min-Max standardization on the yarn making and cleaning working condition data and the package cleaning material consumption data respectively to obtain standardized yarn making working condition data and standardized package material consumption data. Step S1.5: Extract statistical features from the standardized yarn processing data and the standardized packaged material consumption data according to a preset time window to obtain the time-series feature matrix of yarn processing and the time-series feature matrix of packaged material consumption.

3. The production control method according to claim 1, characterized in that, Step S2: Calculate the dynamic load index of each silk-making production line based on the silk-making operating condition time-series characteristic matrix, and adjust the cycle time and material allocation of each line through a load balancing model to obtain silk-making production scheduling parameters, including: Step S2.1: Extract the load congestion degree, process progress deviation, and material conveying margin ratio of each filament production line based on the filament production condition time sequence feature matrix, and calculate the dynamic load index. Step S2.2: Input the dynamic load index into the load balancing model to calculate the cycle adjustment coefficient and material allocation ratio of each yarn production line; Step S2.3: Fine-tune the operating speed of each yarn production line according to the cycle adjustment coefficient, and adjust the material conveying rate according to the material distribution ratio to obtain yarn production scheduling parameters; Step S2.4: Calculate the probability of material shortage risk based on the material conveying margin and process cycle difference in the filament making time sequence feature matrix. Step S2.5: Generate a diversion instruction for the yarn production line where the probability of material shortage exceeds a preset threshold, and incorporate the diversion instruction into the yarn production scheduling parameters.

4. The production control method according to claim 1, characterized in that, Step S3: Calculate the instantaneous unit consumption of raw materials for each individual machine based on the time-series feature matrix of package material consumption, and compare it with the adaptively updated standard consumption interval to obtain the package material consumption anomaly marker sequence, including: Step S3.1: Extract the raw material consumption and finished product output of each individual machine based on the time-series feature matrix of the package material consumption, and calculate the time-sharing instantaneous unit consumption; Step S3.2: Calculate the rolling mean and rolling standard deviation of each raw material category based on historical time-sharing instantaneous unit consumption data, and construct an adaptively updated standard consumption interval; Step S3.3: Compare the time-division instantaneous unit consumption with the standard consumption range, and mark the unit consumption data points that exceed the upper and lower boundaries of the range with an abnormal label; Step S3.4: Summarize the abnormal tags of each individual machine in chronological order to obtain the pack material consumption abnormal tag sequence.

5. The production control method according to claim 1, characterized in that, Step S4: Input the time-series feature matrix of the yarn processing condition and the time-series feature matrix of the bundling material consumption into the causal coupling model. Using the finished product qualification rate as a global constraint, fit the mapping relationship between the yarn processing condition parameters and the bundling material consumption index through gradient boosting regression tree to obtain the causal correlation strength matrix of the condition-material consumption and the collaborative control instruction set, including: Step S4.1: Align and splice the yarn making process time sequence feature matrix and the package material consumption time sequence feature matrix according to the timestamp to obtain a cross-process joint feature matrix; Step S4.2: With the finished product qualification rate as the global constraint objective, construct a gradient boosting regression tree model between the yarn making process parameters and the winding and packaging material consumption index. Step S4.3: Input the cross-process joint feature matrix into the gradient boosting regression tree model to train and obtain the feature importance scores of each yarn processing condition parameter to each package material consumption index. Step S4.4: Assemble the feature importance scores into the working condition-material consumption causal correlation strength matrix; Step S4.5: Based on the working condition parameters whose scores in the working condition-material consumption causal correlation strength matrix exceed the preset threshold, generate corresponding yarn-making scheduling adjustment amounts and winding and packaging equipment parameter adjustment amounts, and assemble them into a collaborative control instruction set.

6. The production control method according to claim 1, characterized in that, Step S5: Perform time-series alignment and backtracking between the package material consumption anomaly marker sequence and the operating condition-material consumption causal correlation strength matrix, and use Granger causality test to distinguish between chain disturbance tracing results and package local equipment anomalies to obtain anomaly tracing classification results, including: Step S5.1: Align the abnormal material consumption marker sequence of the package with the working condition-material consumption causal correlation strength matrix according to the timestamp, and extract the yarn making working condition fluctuation events within a preset time window before and after the abnormal event; Step S5.2: Calculate the Granger causality test statistic for each pair of abnormal events and yarn-making condition fluctuation events; Step S5.3: Mark the silk-making process fluctuation events whose statistics exceed the significance threshold as chain disturbance tracing results; Step S5.4: Mark abnormal events whose statistics do not exceed the significance threshold as local device abnormalities in the package. Step S5.5: Summarize the chain disturbance tracing results and the local device anomaly of the package to obtain the anomaly tracing classification results.

7. The production control method according to claim 1, characterized in that, Step S6: Based on the coordinated control instruction set and the anomaly tracing classification results, predict the future trend of the silk-making process time-series feature matrix using a long short-term memory network, and input the prediction results into the causal coupling model to generate production control instructions, including: Step S6.1: Construct a joint input sequence based on the coordinated control instruction set and the anomaly tracing classification results; Step S6.2: Concatenate the joint input sequence with the silk-making condition time series feature matrix, input the sequence into the long short-term memory network for time series prediction, and obtain the silk-making condition prediction matrix. Step S6.3: Input the yarn processing condition prediction matrix into the causal coupling model to generate production control instructions; Step S6.4: Use the production control instructions and the collaborative control instruction set as feedback inputs to update the working condition-material consumption causal correlation strength matrix.

8. A production control system for synergistic cost reduction and efficiency improvement in tobacco processing and packaging, wherein the production control system is applied to the production control method as described in any one of claims 1 to 7, characterized in that, The production control system includes: The time-series feature matrix acquisition module is used to collect the operating data of multi-line equipment in the yarn making workshop and the metering data of raw and auxiliary materials of single machine in the packaging workshop. After preprocessing, the time-series feature matrix of yarn making conditions and the time-series feature matrix of packaging material consumption are obtained. The silk production scheduling parameter calculation module is used to calculate the dynamic load index of each silk production line based on the silk working condition time sequence feature matrix, and adjust the cycle time and material allocation of each line through the load balancing model to obtain the silk production scheduling parameters. The roll and package material consumption anomaly marker sequence acquisition module is used to calculate the instantaneous unit consumption of raw and auxiliary materials for each single machine according to the roll and package material consumption time series feature matrix, and compare it with the adaptively updated standard consumption interval to obtain the roll and package material consumption anomaly marker sequence. The mapping relationship fitting module is used to input the time series feature matrix of the yarn processing condition and the time series feature matrix of the packaging material consumption into the causal coupling model. With the finished product qualification rate as the global constraint, the mapping relationship between the yarn processing condition parameters and the packaging material consumption index is fitted by the gradient boosting regression tree to obtain the causal correlation strength matrix of the condition-material consumption and the collaborative control instruction set. The anomaly tracing and classification module is used to perform time-series alignment and backtracking of the pack material consumption anomaly marker sequence with the working condition-material consumption causal correlation strength matrix, and to distinguish between the chain disturbance tracing results and the pack local equipment anomalies through Granger causality test to obtain the anomaly tracing and classification results. The production control instruction generation module is used to predict the future trend of the silk-making working condition time series feature matrix based on the collaborative control instruction set and the anomaly tracing classification results, and input the prediction results into the causal coupling model to generate production control instructions.

9. An electronic device, characterized in that, The method includes a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the production control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that, when executed by a processor, enable the production control method as described in any one of claims 1 to 7.

Images