Production rhythm dynamic evaluation method and system based on multi-attribute process constraints

Abstract

The application discloses a production rhythm dynamic evaluation method and system based on multi-attribute process constraints, relates to the tobacco technical field, and is suitable for the process manufacturing scene of mixed production arrangement of multiple brands. The method solves the problems of distortion of traditional static evaluation standards, broken cross-day production data, and difficulty in quantifying process coordination efficiency, provides dynamic and accurate benchmarks for production rhythm evaluation, and supports lean production landing in a tobacco silk workshop. In the method, step S1 realizes semantic alignment of production data, guarantees production shift analysis integrity; step S2 constructs a multi-attribute process constraint system, adapts to different process production requirements; step S3 generates differentiated theoretical working hour benchmarks, eliminates systematic deviation of the evaluation system; step S4 separates time deviation interference components, improves delay judgment accuracy; step S5 realizes automatic identification of delay events, reduces manual statistical lag; and step S6 locates key links of capacity constraints, and provides quantitative basis for production scheduling optimization.

Description

Technical Field

[0001] This application relates to the field of tobacco technology, and in particular to a method and system for dynamic evaluation of production cycle time based on multi-attribute process constraints. Background Technology

[0002] In modern manufacturing, the core of lean production lies in eliminating waste and continuous improvement. As a key upstream link in cigarette production, the tobacco processing workshop is transforming from traditional extensive processing to just-in-time production. This process is long and complex (covering slicing, rehydration, cutting, drying, blending, etc.), and involves the mixed production scheduling of multiple brands.

[0003] With the advancement of digital transformation, "standardized production" has become a key indicator for measuring workshop efficiency. This model requires each process to complete tasks strictly according to a predetermined time rhythm to ensure process continuity and optimal equipment utilization. However, in actual production management, a "black box effect" often exists: managers often only focus on raw material input and finished product output, making it difficult to see the micro-dynamics and efficiency losses in the production process.

[0004] Furthermore, delays in processes are frequent due to equipment malfunctions, material supply delays, or differences in personnel operation. Traditional management relies on manual recording or post-event statistics, which lacks real-time and targeted capabilities, making it difficult to accurately pinpoint bottlenecks. Summary of the Invention

[0005] The main objective of this application is to provide a method and system for dynamic evaluation of production cycle time based on multi-attribute process constraints, in order to solve the problem of frequent process delays caused by equipment failure, material supply delays, or differences in personnel operation in the existing technology. Traditional management relies on manual recording or post-event statistics, which lacks real-time and targeted capabilities and makes it difficult to accurately locate bottlenecks.

[0006] To achieve the above objectives, this application provides the following technical solution: A dynamic evaluation method for production cycle time based on multi-attribute process constraints is proposed. This method is applied to discrete operation data extracted from an external industrial database. The dynamic evaluation method for production cycle time includes: Step S1: Generate a dynamic time window covering production shifts using an event-driven time slicing algorithm, and preprocess the discrete operation data acquired from the outside using the dynamic time window to construct a continuous batch operation event sequence. Step S2: Extract process attribute features from the continuous batch operation event sequence, and construct a multi-attribute process constraint rule library through gradient boosting regression tree; Step S3: Match the correction factor of the corresponding process attribute from the multi-attribute process constraint rule library, and calculate the dynamic theoretical working hours based on the rated production rate and batch operation volume. Step S4: Calculate the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and separate the residual components of the time deviation using the STL time series decomposition algorithm; Step S5: Calculate the dynamic tolerance threshold of the residual component based on the 3σ principle, and determine the generation of a set of delay events based on the dynamic tolerance threshold; Step S6: Extract the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and count the maximum value of the synchronization gap between processes according to the set of delay events to identify the maximum damping point of the production line.

[0007] Beneficial effects of steps S1 to S6: This method is suitable for process manufacturing scenarios with mixed production scheduling of multiple grades. It solves common industry problems such as distortion of traditional static evaluation standards, fragmentation of cross-day production data, and difficulty in quantifying process collaboration efficiency. It provides a dynamic and accurate benchmark for production cycle evaluation, promotes the transformation of the production process from result traceability to process control, supports the implementation of lean production concepts in the silk making workshop, and improves the refinement and scientific nature of production management.

[0008] Specifically, step S1 achieves semantic alignment of production data to ensure the integrity of production shift analysis; step S2 constructs a multi-attribute process constraint system to adapt to the production needs of different process attributes; step S3 generates differentiated theoretical working hour benchmarks to eliminate systematic biases in the evaluation system; step S4 separates interference components in time deviations to improve the accuracy of delay judgment; step S5 achieves automatic identification of delay events to reduce the lag in manual statistics; and step S6 locates key links that restrict production capacity to provide quantitative basis for production scheduling optimization.

[0009] As a further improvement to this application, step S1 involves generating a dynamic time window covering production shifts using an event-driven time slicing algorithm, and preprocessing externally acquired discrete job data using the dynamic time window to construct a continuous batch job event sequence, including: Step S1.1: Obtain the original timestamp set of the discrete operation data from an external industrial database; Step S1.2: Obtain the production shift start and end time pairs of the original timestamp set using an event-driven time slicing algorithm; Step S1.3: Generate a dynamic time window covering the production shifts based on the start and end times of the production shifts; Step S1.4: Filter the discrete operation data through the dynamic time window covering the production shift to obtain a subset of effective data within the shift; Step S1.5: Filter out invalid null values ​​and duplicate records from the valid data subset within the shift, and aggregate them according to the batch identifier to obtain a continuous batch operation event sequence.

[0010] Beneficial effects of steps S1.1 to S1.5: This series of steps completes the time-series preprocessing and structured transformation of the original discrete operation data, solving the problems of cross-day production cycle breaks and invalid data interference in traditional data processing. It provides complete and semantically consistent basic data for subsequent production cycle evaluation, ensuring the sample integrity and logical coherence of subsequent analysis.

[0011] Specifically, step S1.1 provides the original time sequence basis for identifying the time boundary of production shifts; step S1.2 adapts to the production scheduling mode of non-natural days to achieve accurate definition of the production cycle; step S1.3 constructs a time range that matches the actual production logic; step S1.4 removes irrelevant data outside the time range to narrow the data range for subsequent processing; and step S1.5 purifies the data quality and forms a continuous production event structure to support subsequent process attribute extraction and analysis.

[0012] As a further improvement to this application, step S2 involves extracting process attribute features from the continuous batch operation event sequence and constructing a multi-attribute process constraint rule library using a gradient boosting regression tree, including: Step S2.1: Extract all process attribute fields from the continuous batch operation event sequence to generate the original process attribute dataset; Step S2.2: Perform numerical encoding conversion on the original process attribute dataset to obtain a numerical process attribute matrix; Step S2.3: Extract historical actual working hour data from the continuous batch operation event sequence and generate a historical working hour tag set; Step S2.4: Align the numerical process attribute matrix with the historical working time tag set according to the batch identifier to generate a gradient boosting regression tree training sample set; Step S2.5: Initialize the gradient boosting regression tree model and input the gradient boosting regression tree training sample set into the gradient boosting regression tree model for iterative training; Step S2.6: Output the trained gradient boosting regression tree model, which is the mapping relationship between process attributes and time correction factors. Step S2.7: The mapping relationship is stored in a structured manner to obtain a multi-attribute process constraint rule library.

[0013] Beneficial effects of steps S2.1 to S2.7: This series of steps completes the construction of a multi-attribute process constraint rule library, solving the problem that traditional fixed working hour standards cannot adapt to different process requirements, establishing a quantitative correlation between process attributes and working hour correction factors, providing reusable constraint basis for dynamic evaluation of production cycle time, and improving the adaptability of the evaluation system to multi-variety production scenarios.

[0014] Specifically, step S2.1 provides basic feature sources for process constraint modeling; step S2.2 standardizes non-numerical process information to meet algorithm input requirements; step S2.3 provides target labeled data for model training; step S2.4 ensures the correspondence between features and labels to improve the effectiveness of training samples; step S2.5 mines the inherent laws of process attributes and time correction; step S2.6 forms a directly applicable quantitative mapping relationship; and step S2.7 implements structured storage of constraint rules to support subsequent rapid matching and retrieval.

[0015] As a further improvement to this application, step S3, matching the correction factor of the corresponding process attribute from the multi-attribute process constraint rule library, and calculating the dynamic theoretical working hours based on the rated production rate and batch operation volume, includes: Step S3.1: Extract the process attribute features of the batch to be evaluated from the continuous batch operation event sequence; Step S3.2: Input the process attribute features of the batch to be evaluated into the multi-attribute process constraint rule library, and match to obtain the correction factor of the corresponding process attribute; Step S3.3: Extract the rated production rate and batch quantity of the batch to be evaluated from the continuous batch operation event sequence; Step S3.4: Calculate the dynamic theoretical working hours using the correction factor of the corresponding process attribute, the rated production rate of the batch to be evaluated, and the workload of the batch to be evaluated.

[0016] Beneficial effects of steps S3.1 to S3.4: This series of steps enables accurate calculation of dynamic theoretical working hours, solving the problem that traditional fixed working hour standards cannot be adapted to different process batches. Combined with multi-attribute process constraints, it enables adaptive adjustment of the evaluation benchmark, providing an objective and differentiated reference for subsequent production cycle deviation analysis, and improving the rationality and credibility of the evaluation results.

[0017] Specifically, step S3.1 obtains the core process characteristics of the batch to be evaluated, providing input for correction factor matching; step S3.2 achieves accurate correlation between process attributes and corresponding correction factors, reflecting the differences in production characteristics of different processes; step S3.3 obtains the basic production parameters required for theoretical time calculation; and step S3.4 generates a theoretical time benchmark adapted to the current batch, supporting subsequent cycle time deviation determination.

[0018] As a further improvement to this application, step S4, calculating the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and separating the residual components of the time deviation using the STL time series decomposition algorithm, includes: Step S4.1: Extract the actual production hours of the batches corresponding to the dynamic theoretical working hours from the continuous batch operation event sequence; Step S4.2: Match the actual production hours of the batch corresponding to the dynamic theoretical working hours with the dynamic theoretical working hours one by one according to the batch identifier to generate working hour matching data pairs; Step S4.3: Sort the time matching data pairs according to the production time sequence to generate an equally spaced time sequence dataset; Step S4.4: Calculate the difference between each set of matched data in the equally spaced time-series dataset to obtain the original time deviation sequence; Step S4.5: Input the original time deviation sequence into the STL time series decomposition algorithm to perform time series component decomposition operation to obtain the residual components of the original time deviation sequence.

[0019] Beneficial effects of steps S4.1 to S4.5: This series of steps completes the extraction and time series decomposition of time deviations, solving the problem of traditional deviation analysis being affected by systemic fluctuations. It separates the core components reflecting production anomalies, providing a clean analytical basis for subsequent delay judgment and improving the accuracy of cycle time anomaly identification.

[0020] Specifically, step S4.1 provides a true basis for production duration for deviation calculation; step S4.2 establishes the correspondence between theoretical and actual working hours to ensure the correspondence of deviation calculation; step S4.3 forms a deviation sequence that conforms to the production time sequence to meet the input requirements of time sequence decomposition; step S4.4 obtains the difference between theoretical and actual production duration; and step S4.5 eliminates the influence of trends and seasonal fluctuations and extracts the core components that reflect production anomalies.

[0021] As a further improvement to this application, step S5, calculating the dynamic tolerance threshold of the residual component based on the 3σ principle, and determining the generation of a delay event set based on the dynamic tolerance threshold, includes: Step S5.1: Calculate the arithmetic mean and sample standard deviation of the residual components; Step S5.2: Calculate the upper control limit of the residual based on the arithmetic mean and the sample standard deviation, using the 3σ principle, and define it as the dynamic tolerance threshold. Step S5.3: Traverse the original time deviation sequence and filter out time deviation items that are greater than the dynamic tolerance threshold; Step S5.4: Associate the time deviation items that are greater than the dynamic tolerance threshold with the corresponding batch identifier to obtain the set of delay events.

[0022] Beneficial effects of steps S5.1 to S5.4: This series of steps completes the calculation of dynamic tolerance thresholds and automatic identification of delay events, solving the problem that traditional fixed thresholds cannot adapt to production fluctuations. It generates reasonable judgment criteria based on statistical laws, reduces the impact of subjective factors on delay judgment, and provides accurate abnormal event basis for subsequent production bottleneck location.

[0023] Specifically, step S5.1 provides basic statistics for tolerance threshold calculation; step S5.2 generates dynamic judgment criteria adapted to the current production status; step S5.3 filters out time deviations that exceed the normal fluctuation range; and step S5.4 associates abnormal deviations with corresponding production batches to form a traceable set of delay events.

[0024] As a further improvement to this application, step S6, extracting the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and statistically analyzing the maximum value of the synchronization gap between processes based on the set of delay events to identify the maximum damping point of the production line, includes: Step S6.1: Extract the start and end timestamps of all processes in the continuous batch operation event sequence to generate a process start and end timestamp set; Step S6.2: Pair adjacent processes in the process start and end timestamp set according to the preset process topology order to generate a set of adjacent process timestamp pairs; Step S6.3: Calculate the difference between the start time of the subsequent process and the end time of the preceding process for each group of data in the adjacent process timestamp pair set to obtain the synchronization gap set between processes; Step S6.4: Extract all associated batch identifiers from the delay event set to generate a delay batch identifier set; Step S6.5: Select the synchronization gap data corresponding to the delayed batch identifier set from the set of synchronization gaps between processes to obtain the delay-associated synchronization gap subset; Step S6.6: Traverse the subset of delay-related synchronization gaps and extract the maximum value to obtain the maximum synchronization gap value; Step S6.7: Locate the adjacent process pair corresponding to the maximum synchronization gap value and define it as the maximum damping point of the production line.

[0025] Beneficial effects of steps S6.1 to S6.7: This series of steps calculates the synchronization gap between processes and identifies the maximum damping point, solving the problem that traditional bottleneck analysis only focuses on the operating status of a single machine and cannot quantify the efficiency of process collaboration. It establishes the correlation between delay events and process connection efficiency, accurately locates the key links that restrict the overall capacity of the production line, and provides targeted directions for production process optimization.

[0026] Specifically, step S6.1 provides basic time data for process collaboration analysis; step S6.2 establishes the temporal correlation between processes; step S6.3 quantifies the connection efficiency of adjacent processes; step S6.4 extracts batch identifiers related to delays; step S6.5 filters out synchronization gap data related to abnormal production; step S6.6 determines the process connection delay with the greatest impact; and step S6.7 locates the core bottleneck link of the production line.

[0027] To achieve the above objectives, this application also provides the following technical solutions: A dynamic evaluation system for production cycle time based on multi-attribute process constraints, wherein the dynamic evaluation system for production cycle time is applied to the dynamic evaluation method for production cycle time described above, and the dynamic evaluation system for production cycle time includes: The continuous batch operation event sequence construction module is used to generate a dynamic time window covering production shifts through an event-driven time slicing algorithm, and to preprocess externally acquired discrete operation data through the dynamic time window to construct a continuous batch operation event sequence. The multi-attribute process constraint rule library construction module is used to extract process attribute features from the continuous batch operation event sequence and construct a multi-attribute process constraint rule library through gradient boosting regression tree; The dynamic theoretical working time calculation module is used to match the correction factor of the corresponding process attribute from the multi-attribute process constraint rule library, and calculate the dynamic theoretical working time according to the rated production rate and batch operation volume. The time deviation and residual component calculation module is used to calculate the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and to separate the residual components of the time deviation through the STL time series decomposition algorithm. The delay event set acquisition module is used to calculate the dynamic tolerance threshold of the residual component based on the 3σ principle, and generate a delay event set based on the dynamic tolerance threshold. The production line maximum damping point identification module is used to extract the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and to count the maximum value of the synchronization gap between processes based on the set of delay events to identify the maximum damping point of the production line.

[0028] 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 dynamic evaluation method for production cycle time based on multi-attribute process constraints as described above.

[0029] 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 implementation of the aforementioned dynamic evaluation method for production cycle time based on multi-attribute process constraints. Attached Figure Description

[0030] Figure 1 This is a schematic flowchart illustrating the steps of an embodiment of a dynamic evaluation method for production cycle time based on multi-attribute process constraints according to this application. Figure 2 This is a schematic diagram of the functional modules of an embodiment of a dynamic evaluation system for production cycle time based on multi-attribute process constraints according to 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

[0031] 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.

[0032] 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 technical features indicated. 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 inclusion. For example, a process, method, apparatus, product, or device that includes 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.

[0033] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included 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.

[0034] 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.

[0035] 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.

[0036] like Figure 1 As shown, this embodiment provides an example of a dynamic evaluation method for production cycle time based on multi-attribute process constraints. In this embodiment, the dynamic evaluation method for production cycle time is applied to discrete operation data extracted from an external industrial database.

[0037] Specifically, the dynamic evaluation method for production cycle time includes the following steps: Step S1: Generate a dynamic time window covering production shifts using an event-driven time slicing algorithm, and preprocess the discrete operation data obtained from external sources using the dynamic time window to construct a continuous batch operation event sequence.

[0038] Preferably, this step addresses the problem of data truncation and invalid data interference in production analysis caused by cross-day scheduling in the silk-making workshop. By replacing traditional fixed time slices with event-driven time slices, a structured data foundation fully aligned with actual production logic is constructed, providing semantically consistent input for subsequent process constraint modeling and cycle time evaluation.

[0039] Furthermore, step S1 specifically includes the following steps: Step S1.1: Obtain the original timestamp set of discrete operation data from an external industrial database.

[0040] Preferably, the data source is the process event table, equipment operation log table, and batch tracking table of the Manufacturing Execution System (MES), from which all production-related timestamp fields are extracted, including equipment start / stop time, process start / end time, material feeding time, and finished product discharge time.

[0041] Preferably, the time precision can be uniformly converted to UTC millisecond-level timestamps to avoid timing chaos caused by time synchronization errors between different systems.

[0042] Preferably, the data range can extract the full data of the most recent 30 production shifts, covering production scenarios of different brands and different outputs, to ensure the representativeness of the sample.

[0043] Step S1.2: Obtain the start and end time pairs of the production shifts from the original timestamp set using an event-driven time slicing algorithm.

[0044] Preferably, the algorithm is based on time slices triggered by changes in the state of core equipment in the production system, rather than relying on natural day boundaries, thus adapting to the shift scheduling pattern of three shifts and cross-day production in the silk-making workshop. The threshold settings are as follows: (1) Threshold for starting a shift: The slicer runs continuously for ≥10 minutes and at least 3 subsequent processes are started in sequence.

[0045] (2) Threshold for determining the end of shift: All production equipment has been shut down for ≥30 minutes without any new material feeding events.

[0046] For example, this step can be implemented using the following pseudocode: def get_shift_time_pairs(timestamps, device_status): shift_pairs = [] current_start = None for i in range(len(timestamps)): # Determine the shift start if device_status[i] == "running" and not current_start: if check_consecutive_running(timestamps, device_status, i, 10*60*1000): current_start = timestamps[i] # Determine the end of the shift elif device_status[i] == "stopped" and current_start: if check_consecutive_stopped(timestamps, device_status, i, 30*60*1000): shift_pairs.append((current_start, timestamps[i])) current_start = None return shift_pairs Step S1.3: Generate a dynamic time window covering the production shifts based on the start and end times of the production shifts.

[0047] Preferably, the time window format corresponds to a left-closed and right-closed time interval [shift_start, shift_end] for each shift, and the length of the interval is determined by the actual production time and has no fixed value.

[0048] Preferably, for short-term shutdowns within a shift (≤5 minutes), the time window is not divided to maintain the integrity of the shift; for shift interruptions caused by equipment failure (downtime >30 minutes), the time window is automatically split into two independent time windows.

[0049] Step S1.4: Filter discrete operation data by using a dynamic time window covering production shifts to obtain a subset of effective data within each shift.

[0050] Preferably, the filtering logic is to retain all records whose timestamps fall within any dynamic time window interval, and to remove non-production data such as equipment debugging, maintenance, and shift handover outside the time window.

[0051] For example, the SQL filter statement is as follows: SELECT * FROM mes_production_data WHERE event_time BETWEEN :shift_start AND :shift_end AND production_status ='normal'; Step S1.5: Filter out invalid null values ​​and duplicate records from the valid data subset within the shift, and aggregate them according to the batch identifier to obtain a continuous batch operation event sequence.

[0052] Preferably, for filtering invalid null values, records with any of the batch number, process number, or equipment number fields being empty can be deleted, and batches with more than 10% null values ​​can be completely removed.

[0053] Preferably, for duplicate record filtering, the first record is retained and subsequent duplicates are deleted, using "batch number + process number + event timestamp" as the unique key.

[0054] Preferably, for batch aggregation, the batch number is used as the primary key to aggregate the start and end times, process parameters, equipment numbers, and other information of all processes in the batch, forming a structured batch event sequence. Each sequence element contains complete information on the entire production process of a single batch.

[0055] Beneficial effects of steps S1.1 to S1.5: This series of steps completes the time-series preprocessing and structured transformation of the original discrete operation data, solving the problems of cross-day production cycle breaks and invalid data interference in traditional data processing. It provides complete and semantically consistent basic data for subsequent production cycle evaluation, ensuring the sample integrity and logical coherence of subsequent analysis.

[0056] Specifically, step S1.1 provides the original time sequence basis for identifying the time boundary of production shifts; step S1.2 adapts to the production scheduling mode of non-natural days to achieve accurate definition of the production cycle; step S1.3 constructs a time range that matches the actual production logic; step S1.4 removes irrelevant data outside the time range to narrow the data range for subsequent processing; and step S1.5 purifies the data quality and forms a continuous production event structure to support subsequent process attribute extraction and analysis.

[0057] Step S2: Extract process attribute features from the continuous batch operation event sequence, and construct a multi-attribute process constraint rule library through gradient boosting regression tree.

[0058] Preferably, this step addresses the problem of distortion in fixed working hours standards caused by mixed production scheduling of multiple grades and fluctuations in process parameters in the silk-making workshop. It uses gradient boosting regression trees to mine the nonlinear quantitative correlation between process attributes and production working hours, replacing the traditional manually formulated experience-based working hour standards, and constructing a multi-attribute process constraint rule library that can dynamically adapt to different production scenarios.

[0059] Furthermore, step S2 specifically includes the following steps: Step S2.1: Extract all process attribute fields from the continuous batch job event sequence to generate the original process attribute dataset.

[0060] Preferably, for field classification extraction, three types of core process attributes are extracted from the batch event sequence, including discrete attributes (tobacco leaf brand, equipment number, shift type, tobacco leaf grade), continuous attributes (rehydration temperature, cutting width, drying moisture content, blending ratio), and time-series attributes (feeding interval, process switching time).

[0061] Preferably, the variance threshold method can be used to remove attribute fields with no discriminative power. The variance threshold is set to 0.01, and fields with variance less than the threshold are deleted to reduce redundant calculations in model training.

[0062] Preferably, continuous attributes can be filled with the mean of the same brand and shift, and discrete attributes can be filled with the mode. Fields with a missing rate of more than 20% can be directly removed.

[0063] Step S2.2: Perform numerical encoding transformation on the original process attribute dataset to obtain the numerical process attribute matrix.

[0064] Preferably, for the encoding strategy, one-hot encoding is used for unordered discrete attributes, label encoding is used for ordered discrete attributes, and Z-score normalization is used for continuous attributes.

[0065] Preferably, if the feature dimension generated by a single attribute after one-hot encoding exceeds 10, the feature hashing method is used to reduce the dimension to 8 to avoid the curse of dimensionality.

[0066] Step S2.3: Extract historical actual working hour data from the continuous batch job event sequence and generate a historical working hour tag set.

[0067] Preferably, the working hours are calculated as follows: the total actual working hours for a single batch is the difference between the end time of the last process and the start time of the first process in that batch, with the unit being seconds.

[0068] Preferably, for abnormal working hours filtering, the quartile method can be used to remove abnormal working hour data. The first quartile Q1 and the third quartile Q3 of the sample are calculated, and records less than Q1-1.5×IQR or greater than Q3+1.5×IQR are deleted, where IQR=Q3-Q1.

[0069] Step S2.4: Align the numerical process attribute matrix with the historical working time label set according to the batch identifier to generate a gradient boosting regression tree training sample set.

[0070] Preferably, the alignment logic is to use the batch number as the unique primary key, and to map the process attribute feature vector of each batch to the corresponding total actual working hours label to form a "feature-label" sample pair.

[0071] Preferably, for the partitioning of the dataset, a stratified random partitioning method can be used to divide the sample set into a training set and a test set in a ratio of 8:2, with a random seed of 42, to ensure that the distribution of samples of different grades is consistent in the training set and the test set.

[0072] Step S2.5: Initialize the gradient boosting regression tree model and input the gradient boosting regression tree training sample set into the gradient boosting regression tree model for iterative training.

[0073] Preferably, the model parameters can be set using the Gradient BoostingRegressor implemented in scikit-learn, one of the core parameters of which is shown in the following pseudocode: from sklearn.ensemble import GradientBoostingRegressor gbr_model = GradientBoostingRegressor( n_estimators=100, learning_rate=0.1, max_depth=5, min_samples_split=2, min_samples_leaf=1, loss='squared_error', random_state=42 ) gbr_model.fit(X_train, y_train) Preferably, training is terminated early when the mean square error of the test set decreases by less than 0.001 over five consecutive iterations to prevent overfitting.

[0074] Step S2.6: Output the trained gradient boosting regression tree model, which is the mapping relationship between process attributes and time correction factors.

[0075] Preferably, for feature importance calculation, the model outputs the feature importance score for each process attribute, and the calculation formula is as follows:

[0076] Gain t,i Let represent the information gain brought by the i-th feature in the t-th tree, where T is the total number of trees and N is the total number of features.

[0077] Preferably, for the derivation of the correction factor, the importance score of each feature can be multiplied by the value range of that feature to obtain the correction coefficient of the process attribute to the total working hours. The set of all correction coefficients is the mapping relationship between the process attribute and the working hour correction factor.

[0078] Step S2.7: The mapping relationship is stored in a structured manner to obtain a multi-attribute process constraint rule library.

[0079] Preferably, the storage format adopts JSON format to structure and store the rule base. Each rule item contains four fields: process attribute combination, corresponding correction factor, applicable batch range, and confidence level.

[0080] Preferably, the rule base can be set to have an automatic update mechanism, which automatically extracts new production data from the past 7 days every Sunday morning, performs incremental training on the model, updates the rule base content, and ensures the timeliness of the rules.

[0081] Beneficial effects of steps S2.1 to S2.7: This series of steps completes the construction of a multi-attribute process constraint rule library, solving the problem that traditional fixed working hour standards cannot adapt to different process requirements, establishing a quantitative correlation between process attributes and working hour correction factors, providing reusable constraint basis for dynamic evaluation of production cycle time, and improving the adaptability of the evaluation system to multi-variety production scenarios.

[0082] Specifically, step S2.1 provides basic feature sources for process constraint modeling; step S2.2 standardizes non-numerical process information to meet algorithm input requirements; step S2.3 provides target labeled data for model training; step S2.4 ensures the correspondence between features and labels to improve the effectiveness of training samples; step S2.5 mines the inherent laws of process attributes and time correction; step S2.6 forms a directly applicable quantitative mapping relationship; and step S2.7 implements structured storage of constraint rules to support subsequent rapid matching and retrieval.

[0083] Step S3: Match the correction factor of the corresponding process attribute from the multi-attribute process constraint rule library, and calculate the dynamic theoretical working hours based on the rated production rate and batch operation volume.

[0084] Preferably, this step addresses the problem that traditional fixed rated working hours cannot adapt to the production characteristics of different process batches. By dynamically matching a multi-attribute process constraint rule library, a theoretical working hour benchmark that accurately corresponds to the process conditions of the batch to be evaluated is generated, replacing the unified static standard and eliminating evaluation bias caused by process differences.

[0085] Furthermore, step S3 specifically includes the following steps: Step S3.1: Extract the process attribute features of the batch to be evaluated from the continuous batch operation event sequence.

[0086] Preferably, the triggering time is when the batch to be evaluated completes the first process of feeding and generates a unique batch identifier, the feature extraction process is automatically triggered without manual intervention.

[0087] Preferably, the feature extraction range matching step S2.1 defines three types of core process attribute fields to ensure that the dimensions and types of the input features are completely consistent with those of the model training features.

[0088] Preferably, for real-time preprocessing, the extracted original features can be subjected to the same numerical encoding transformation as in step S2.2, and the preprocessing parameters reuse the training set mean and standard deviation stored in the rule base to ensure the consistency of the input distribution.

[0089] Preferably, the threshold for the proportion of missing features can be set to 5%. When the threshold is exceeded, the completion process is automatically triggered, and the average value of the corresponding features of the three most recent normal batches of the same brand is used to fill the gap.

[0090] Step S3.2: Input the process attribute characteristics of the batch to be evaluated into the multi-attribute process constraint rule library, and match to obtain the correction factor of the corresponding process attribute.

[0091] Preferably, the matching priority adopts the exact matching logic. When there is a rule in the rule base that is completely consistent with the combination of process attributes of the batch to be evaluated, the corresponding correction factor is directly extracted.

[0092] Preferably, for the fuzzy matching mechanism, when there is no exact match, the cosine similarity is used to calculate the similarity between the feature to be evaluated and all feature vectors in the rule base. The similarity threshold is set to 0.85. The correction factors of the three rule items with the highest similarity and greater than the threshold are taken, and the final correction factor is obtained by weighted averaging based on similarity.

[0093] Preferably, when the similarity of all rule items is less than 0.85, a global average correction factor of 1.0 is used for the corresponding grade, and the process combination is recorded and included in the training samples when the rule base is updated weekly.

[0094] For example, this step can be implemented using the following pseudocode: import numpy as np from sklearn.metrics.pairwise import cosine_similarity def match_correction_factor(batch_features, rule_base): # Exact Match exact_match = next((r for r in rule_base if r["feature_vector"] ==batch_features.tolist()), None) if exact_match: return exact_match["correction_factor"] # Fuzzy Matching similarities = [cosine_similarity(batch_features.reshape(1,-1),np.array(r["feature_vector"]).reshape(1,-1))[0][0] for r in rule_base] top_indices = np.argsort(similarities)[-3:][::-1] top_sims = [similarities[i] for i in top_indices] if top_sims[0]>= 0.85: weights = np.array(top_sims) / np.sum(top_sims) top_factors = [rule_base[i]["correction_factor"] for i in top_indices] return float(np.dot(weights, top_factors)) # Default correction factor return 1.0 Step S3.3: Extract the rated production rate and batch quantity of the batch to be evaluated from the continuous batch operation event sequence.

[0095] Preferably, the rated production rate is extracted from the process specification database of the MES system, corresponding to the standard production rate of the batch grade to be evaluated, with the unit uniformly set as kg / s, and the rated rates of different grades and different production lines are stored independently.

[0096] Preferably, for calculating the workload of the batch to be evaluated, the actual weight of the first process of the batch can be taken. The data is uploaded to the MES system in real time by the on-site weighing sensor and automatically extracted without manual input. The unit is uniformly kilogram.

[0097] Preferably, the rated range verification threshold can be set to ±10%. When the extracted workload exceeds this range of the rated workload for the corresponding grade, a secondary verification is automatically triggered. The data on the material feeding sheet is compared with the sensor data, and the consistent value is taken as the final workload.

[0098] Step S3.4: Calculate the dynamic theoretical working hours using the correction factor of the corresponding process attribute, the rated production rate of the batch to be evaluated, and the workload of the batch to be evaluated.

[0099] Preferably, the calculation formula is as follows:

[0100] Where T dynamic , where W is the dynamic theoretical working time (in seconds), W is the batch quantity to be evaluated (in kilograms), Rrated is the rated production rate of the batch to be evaluated (in kilograms / second), and k is the correction factor for the corresponding process attribute.

[0101] Preferably, the calculation results are automatically retained to two decimal places and converted to seconds to maintain consistency with the units of subsequent actual production hours, avoiding unit conversion errors. The calculated dynamic theoretical working hours are associated with the batch identifier and extraction time and stored in a dedicated database for production cycle evaluation for subsequent deviation analysis.

[0102] Beneficial effects of steps S3.1 to S3.4: This series of steps enables accurate calculation of dynamic theoretical working hours, solving the problem that traditional fixed working hour standards cannot be adapted to different process batches. Combined with multi-attribute process constraints, it enables adaptive adjustment of the evaluation benchmark, providing an objective and differentiated reference for subsequent production cycle deviation analysis, and improving the rationality and credibility of the evaluation results.

[0103] Specifically, step S3.1 obtains the core process characteristics of the batch to be evaluated, providing input for correction factor matching; step S3.2 achieves accurate correlation between process attributes and corresponding correction factors, reflecting the differences in production characteristics of different processes; step S3.3 obtains the basic production parameters required for theoretical time calculation; and step S3.4 generates a theoretical time benchmark adapted to the current batch, supporting subsequent cycle time deviation determination.

[0104] Step S4: Calculate the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and separate the residual components of the time deviation using the STL time series decomposition algorithm.

[0105] Preferably, this step addresses the problem that traditional production cycle deviation analysis cannot distinguish between systematic fluctuations and abnormal fluctuations. By using time-series decomposition technology, it removes periodic and trend-related disturbances such as equipment aging, shift changes, and environmental temperature and humidity variations, extracting only residual components that reflect production anomalies, thus providing a clean analytical basis for subsequent accurate delay determination.

[0106] Furthermore, step S4 specifically includes the following steps: Step S4.1: Extract the actual production hours of the batch corresponding to the dynamic theoretical working hours from the continuous batch operation event sequence.

[0107] Preferably, when the batch to be evaluated completes the last process and generates a completion event, the start time of the first process and the end time of the last process of the batch are automatically extracted from the MES batch tracking table, and the total actual production time of a single batch is calculated, with the unit uniformly in seconds, consistent with the unit of dynamic theoretical time.

[0108] Preferably, for batches spanning two or more production shifts, the production time across all shifts is accumulated, and non-production time such as shift handover and equipment maintenance is excluded to ensure that the actual working hours only reflect the effective production time.

[0109] Preferably, the reasonable range of actual working hours can be set to 0.5 to 3 times the corresponding dynamic theoretical working hours. Records exceeding this range are marked as abnormal and will not be included in subsequent analysis, while triggering a data review alarm.

[0110] Step S4.2: Match the actual production hours of the batch corresponding to the dynamic theoretical working hours with the dynamic theoretical working hours according to the batch identifier to generate working hour matching data pairs.

[0111] Preferably, the matching unique key can be a globally unique batch number as the matching primary key, ensuring that each actual working hour record corresponds to a unique dynamic theoretical working hour record.

[0112] Preferably, records with dynamic theoretical working hours but no corresponding actual working hours, or records with actual working hours but no corresponding dynamic theoretical working hours, are automatically marked as mismatches, stored in the abnormal data log, and reviewed in batches every morning.

[0113] Preferably, the batch matching success rate threshold can be set to 98%. When the matching success rate of three consecutive shifts is lower than this threshold, the system alarm will be automatically triggered to investigate data transmission or calculation logic problems.

[0114] Step S4.3: Sort the time-matching data pairs according to the production time sequence to generate an equally spaced time sequence dataset.

[0115] Preferably, for time-series sorting, the start time of the first process in the batch is used as the basis, and all successfully matched time data pairs are arranged in ascending order according to their time sequence.

[0116] Preferably, for equal-interval interpolation, since the intervals of actual production batches are not uniform, linear interpolation is used to convert the batch data with unequal intervals into equal-interval time-series data, and the basic time interval is set to 15 minutes (to match the average production time of a single batch in the yarn-making workshop).

[0117] Preferably, the interpolation constraint is to perform linear interpolation when there are no more than 3 consecutive missing time intervals (45 minutes); when there are more than 3 consecutive missing time intervals, the time interval is marked as a missing data segment and is not included in the subsequent decomposition analysis.

[0118] Preferably, the interpolation formula is as follows:

[0119] Where t is the interpolation time point, t1 and t2 are two adjacent known time points, and y1 and y2 are the working hours at the corresponding time points.

[0120] Step S4.4: Calculate the difference between each set of matched data in the equally spaced time-series dataset to obtain the original time deviation sequence.

[0121] Preferably, the original time deviation is the actual production time minus the dynamic theoretical time, that is: ΔT(t)=T actual (t)-T dynamic (t). Where ΔT(t) is the time deviation at time t, T actual (t) represents the actual production hours at time t, where T is the production time at time t. dynamic (t) represents the dynamic theoretical working time at time t.

[0122] Preferably, a positive deviation indicates a production delay for that batch, a negative deviation indicates that the production is ahead of schedule, and a zero deviation indicates that the production is in line with the theoretical cycle time.

[0123] Preferably, for interpolation deviation processing, the deviation value for the time point obtained by interpolation is the difference between the actual working hours calculated by interpolation and the corresponding theoretical working hours.

[0124] Step S4.5: Input the original time skew sequence into the STL time series decomposition algorithm to perform time series component decomposition operation to obtain the residual components of the original time skew sequence.

[0125] Preferably, STL (Seasonal and Trend decomposition using Loess) is a time series decomposition algorithm based on local weighted regression, which decomposes the time series into a trend component T(t), a seasonal component S(t), and a residual component R(t), satisfying ΔT(t)=T(t)+S(t)+R(t).

[0126] For example, this step can be implemented using the following pseudocode: from statsmodels.tsa.seasonal import STL import pandas as pd # Original time deviation sequence (indexed by equally spaced timestamps) ts = pd.Series(deviation_list, index=time_index) # STL Decomposition Parameters stl = STL( ts, period=96, # The period is 24 hours, corresponding to 96 data points at 15-minute intervals. seasonal=7, # Seasonal component smoothing window trend=15, # Trend component smoothing window low_pass=13, # Low-pass filter window robust=True # Enable robustness and reduce the impact of outliers on the decomposition results ) # Execution Decomposition result = stl.fit() # Extracting residual components residual_component = result.resid Among them, the trend component reflects the slowly changing systematic factors such as equipment aging, improved personnel proficiency, and long-term adjustments to process parameters.

[0127] Among them, the seasonal component reflects periodic factors such as shift rotation, diurnal temperature and humidity changes, and cyclical differences in raw material batches.

[0128] Among them, the residual component is the random fluctuation remaining after removing the above-mentioned systematic and periodic factors, which only reflects abnormal events such as sudden equipment failure, material supply interruption, and operational errors.

[0129] Beneficial effects of steps S4.1 to S4.5: This series of steps completes the extraction and time series decomposition of time deviations, solving the problem of traditional deviation analysis being affected by systemic fluctuations. It separates the core components reflecting production anomalies, providing a clean analytical basis for subsequent delay judgment and improving the accuracy of cycle time anomaly identification.

[0130] Specifically, step S4.1 provides a true basis for production duration for deviation calculation; step S4.2 establishes the correspondence between theoretical and actual working hours to ensure the correspondence of deviation calculation; step S4.3 forms a deviation sequence that conforms to the production time sequence to meet the input requirements of time sequence decomposition; step S4.4 obtains the difference between theoretical and actual production duration; and step S4.5 eliminates the influence of trends and seasonal fluctuations and extracts the core components that reflect production anomalies.

[0131] Step S5: Calculate the dynamic tolerance threshold of the residual components based on the 3σ principle, and generate a set of delay events based on the dynamic tolerance threshold.

[0132] Preferably, this step addresses the problem that traditional fixed tolerance thresholds cannot adapt to the dynamic fluctuations of the production process. It generates adaptive dynamic judgment criteria based on the statistical characteristics of the current production status, avoiding misjudgments and omissions caused by manually setting thresholds, and realizing the automated and objective identification of delay events, providing accurate abnormal event basis for subsequent bottleneck location.

[0133] Furthermore, step S5 specifically includes the following steps: Step S5.1: Calculate the arithmetic mean and sample standard deviation of the residual components.

[0134] Preferably, a sliding window mechanism is used: a sliding time window is used to calculate the statistics, the window size is set to the most recent 24 hours (corresponding to 96 data points at 15-minute intervals in step S4.3), the window sliding step is 1 data point, and the statistics are automatically updated once after each batch is completed to ensure that the threshold reflects the current production status in real time.

[0135] The formula for calculating the arithmetic mean is as follows: Where n is the number of residual components within the sliding window, and R i Let be the value of the i-th residual component.

[0136] The formula for calculating the sample standard deviation (using an unbiased estimate) is as follows: .

[0137] Preferably, before calculating the statistics, extreme outliers with an absolute residual value greater than 10 times the standard deviation within the sliding window are removed to avoid excessive influence of individual extreme data on the statistical results.

[0138] Step S5.2: Calculate the upper control limit of the residuals based on the arithmetic mean and sample standard deviation, using the 3σ principle, and define it as the dynamic tolerance threshold.

[0139] Preferably, the core calculation formula is: T threshold =μ+3s. Where T threshold This is the dynamic tolerance threshold.

[0140] Among them, for boundary constraints, when the calculated T threshold When <0, force T to be set. threshold =0, to avoid the unreasonable situation where a negative threshold would cause all positive deviations to be judged as delays.

[0141] Preferably, based on the 3σ principle, 99.73% of normal production fluctuations will fall within the range of [μ-3s, μ+3s], and deviations exceeding the upper control limit can be regarded as non-random abnormal delays.

[0142] For example, this step can be implemented using the following pseudocode: import numpy as np def calculate_dynamic_threshold(residual_window): # Remove extreme outliers q1, q3 = np.percentile(residual_window, [25, 75]) iqr = q3 - q1 lower_bound = q1 - 1.5 * iqr upper_bound = q3 + 1.5 * iqr filtered_residuals = residual_window[(residual_window>= lower_bound)&(residual_window<= upper_bound)] # Calculate statistics mu = np.mean(filtered_residuals) s = np.std(filtered_residuals, ddof=1) # Calculate and constrain thresholds threshold = mu + 3 * s return max(threshold, 0) Step S5.3: Traverse the original time deviation sequence and filter out time deviation items that are greater than the dynamic tolerance threshold.

[0143] Preferably, the traversal range only traverses the original time deviation sequence within the sliding window, without involving repeated calculations of historical data.

[0144] Preferably, for the original time deviation ΔT(t) at each time point, if ΔT(t) > T threshold If so, the batch corresponding to that time point is marked as a suspected delayed batch.

[0145] Preferably, if the deviation at three or more consecutive time points exceeds the threshold, it is marked as a persistent delay event and marked separately to prompt dispatchers to pay close attention.

[0146] Step S5.4: Associate the time deviation items that are greater than the dynamic tolerance threshold with the corresponding batch identifier to obtain the set of delay events.

[0147] Preferably, the association logic can associate abnormal deviation items with the time matching data pairs generated in step S4.2 through timestamps to obtain corresponding metadata such as batch number, process information, and equipment information.

[0148] Preferably, the batch number is used as the unique key for deduplication to avoid the same batch being recorded multiple times due to deviations exceeding the standard at multiple time points; each delay event includes fields such as batch number, deviation value, occurrence time, dynamic tolerance threshold, production line involved, and associated equipment number, and is stored in a structured manner in the production anomaly database; real-time push: after a delay event is generated, it is automatically pushed to the production scheduling module of the MES system through the API interface, highlighted in the scheduling interface, and at the same time triggers SMS / DingTalk alarm notifications to the relevant responsible persons.

[0149] Beneficial effects of steps S5.1 to S5.4: This series of steps completes the calculation of dynamic tolerance thresholds and automatic identification of delay events, solving the problem that traditional fixed thresholds cannot adapt to production fluctuations. It generates reasonable judgment criteria based on statistical laws, reduces the impact of subjective factors on delay judgment, and provides accurate abnormal event basis for subsequent production bottleneck location.

[0150] Specifically, step S5.1 provides basic statistics for tolerance threshold calculation; step S5.2 generates dynamic judgment criteria adapted to the current production status; step S5.3 filters out time deviations that exceed the normal fluctuation range; and step S5.4 associates abnormal deviations with corresponding production batches to form a traceable set of delay events.

[0151] Step S6: Extract the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and count the maximum value of the synchronization gap between processes based on the set of delay events to identify the maximum damping point of the production line.

[0152] Preferably, this step addresses the problem that traditional bottleneck analysis only focuses on the efficiency of a single machine and cannot quantify the collaborative losses between processes. By establishing the correlation between delay events and the gaps between processes, it accurately locates the biggest damping point that restricts the overall capacity of the production line, rather than an isolated single-machine bottleneck, providing targeted quantitative basis for production scheduling optimization and process reengineering.

[0153] Furthermore, step S6 specifically includes the following steps: Step S6.1: Extract the start and end timestamps of all processes in the continuous batch operation event sequence to generate a set of process start and end timestamps.

[0154] Preferably, the data source is extracted from the process execution record table of the MES system. Each record contains six core fields: globally unique batch number, process ID, process name, process start time, process end time, and execution device number.

[0155] Preferably, the timestamps can be uniformly converted to UTC millisecond-level timestamps to maintain consistency with the time precision of step S1. Records with process statuses of "debugging," "maintenance," or "rework" are discarded, retaining only the start and end times of processes under normal production conditions. For processes spanning multiple shifts, the effective production time across all shifts is accumulated without splitting the process time.

[0156] Preferably, each batch must contain all the processes specified in the process specification. A batch missing any process is marked as an incomplete batch and will not be included in subsequent analysis.

[0157] Step S6.2: Pair adjacent processes in the process start and end timestamp set according to the preset process topology order to generate a set of adjacent process timestamp pairs.

[0158] Preferably, the preset process topology is loaded from the process specification database of the MES system, corresponding to the standard process flow of the batch grade to be evaluated. The typical topology of the silk processing workshop is: slicing → rehydration → leaf moistening → slicing → drying → blending → fragrance addition → silk storage.

[0159] Preferably, for each batch, the end time of the i-th process is paired with the start time of the (i+1)-th process in the order of the process topology to form a timestamp pair of (end time of process i, start time of process i+1).

[0160] Preferably, for processes with multiple parallel production lines (such as multiple yarn drying lines), the end time of all parallel processes is paired with the start time of the next common process (such as blending). If a batch skips a process due to process adjustments, the preceding and following processes of the skipped process are automatically paired directly.

[0161] Step S6.3: Calculate the difference between the start time of the subsequent process and the end time of the preceding process for each group of data in the adjacent process timestamp pair set to obtain the synchronization gap set between processes.

[0162] Preferably, the calculation formula is: G i,i+1 =T start(i+1) -T end(i) Among them, G i,i+1 T is the synchronization gap between process i and process i+1. start(i+1) T represents the start time of the subsequent process i+1. end(i) This represents the end time of the preceding process i.

[0163] Preferably, a positive gap value represents the time a subsequent process waits for a preceding process to complete, i.e., the waiting time between processes; a negative gap value indicates that the subsequent process has started prematurely before the preceding process has completed, which is marked as an abnormal process connection. Preferably, the maximum reasonable threshold for the synchronization gap can be set to 2 hours. Gaps exceeding this threshold are marked as data anomalies and stored in the anomaly log for review.

[0164] For example, this step can be implemented using the following pseudocode: def calculate_sync_gaps(batch_processes, process_topology): gaps = [] for i in range(len(process_topology) - 1): prev_process = next(p for p in batch_processes if p["process_id"]== process_topology[i]) next_process = next(p for p in batch_processes if p["process_id"]== process_topology[i+1]) gap = next_process["start_time"] - prev_process["end_time"] gaps.append({ "batch_id": batch_processes[0]["batch_id"], "prev_process": process_topology[i], "next_process": process_topology[i+1], "gap_value": gap }) return gaps Step S6.4: Extract all associated batch identifiers from the delay event set and generate a delay batch identifier set.

[0165] Preferably, the extraction scope is drawn from the production anomaly database generated in step S5, extracting the batch numbers corresponding to all delay events within the current sliding window (last 24 hours).

[0166] Preferably, the batch number can be used as a unique key for deduplication to ensure that each delayed batch appears only once. Only delayed batches that have completed all processes are retained, while batches still in production are discarded to avoid errors in calculating synchronization gaps due to incomplete processes.

[0167] Step S6.5: Select the synchronization gap data corresponding to the delayed batch identifier set from the inter-process synchronization gap set to obtain the delayed associated synchronization gap subset.

[0168] Preferably, the batch number is used as the matching key to traverse the set of synchronization gaps between processes and retain all synchronization gap records where the batch number exists in the set of delayed batch identifiers.

[0169] Preferably, missing synchronization gap records in delayed batches are marked as missing data and excluded from subsequent statistical analysis; if a single delayed batch is missing more than two synchronization gap records, the entire batch is removed. The filtered subset retains four core fields: batch number, preceding process ID, succeeding process ID, and synchronization gap value.

[0170] Step S6.6: Traverse the subset of delay-related synchronization gaps and extract the maximum value to obtain the maximum synchronization gap value.

[0171] Preferably, the traversal range is all delay-related synchronization gap records within the current sliding window.

[0172] For determining the maximum value, a numerical comparison method can be used to extract the largest synchronization gap value. If multiple identical maximum values ​​exist, all of them are retained. Simultaneously, the average synchronization gap, median synchronization gap, and frequency of occurrence for each adjacent process pair are calculated to provide supplementary data for subsequent analysis.

[0173] For example, this step can be implemented using the following pseudocode: def find_max_gap(delay_gaps): if not delay_gaps: return None, [] max_value = max(gap["gap_value"] for gap in delay_gaps) max_gaps = [gap for gap in delay_gaps if gap["gap_value"] == max_value] return max_value, max_gaps Step S6.7: Locate the adjacent process pair corresponding to the maximum synchronization gap value and define it as the maximum damping point of the production line.

[0174] Preferably, the positioning logic can extract the preceding process ID and the following process ID based on the record corresponding to the maximum synchronization gap value, map them to specific process names, and form adjacent process pairs.

[0175] The maximum damping point refers to the connection link with the longest waiting time between processes among all delayed batches, which is the most critical bottleneck restricting the overall capacity improvement of the production line. A maximum damping point report can be generated, which includes information such as the names of adjacent process pairs, the maximum synchronization gap value, the number of delayed batches involved, the time period in which it occurred, and the associated equipment number.

[0176] Preferably, the maximum damping point information can be pushed to the production scheduling module of the MES system via the API interface, highlighted on the electronic Kanban board, and automatically generated scheduling optimization suggestions to prompt the scheduler to prioritize adjusting the material supply and equipment arrangement for this process pair.

[0177] Beneficial effects of steps S6.1 to S6.7: This series of steps calculates the synchronization gap between processes and identifies the maximum damping point, solving the problem that traditional bottleneck analysis only focuses on the operating status of a single machine and cannot quantify the efficiency of process collaboration. It establishes the correlation between delay events and process connection efficiency, accurately locates the key links that restrict the overall capacity of the production line, and provides targeted directions for production process optimization.

[0178] Specifically, step S6.1 provides basic time data for process collaboration analysis; step S6.2 establishes the temporal correlation between processes; step S6.3 quantifies the connection efficiency of adjacent processes; step S6.4 extracts batch identifiers related to delays; step S6.5 filters out synchronization gap data related to abnormal production; step S6.6 determines the process connection delay with the greatest impact; and step S6.7 locates the core bottleneck link of the production line.

[0179] In summary, the overall beneficial effects of steps S1 to S6 of this application are as follows: This method is suitable for process manufacturing scenarios with mixed production scheduling of multiple grades. It solves common industry problems such as distortion of traditional static evaluation standards, fragmentation of cross-day production data, and difficulty in quantifying process collaboration efficiency. It provides a dynamic and accurate benchmark for production cycle evaluation, promotes the transformation of the production process from result traceability to process control, supports the implementation of lean production concepts in the silk making workshop, and improves the refinement and scientific nature of production management.

[0180] Specifically, step S1 achieves semantic alignment of production data to ensure the integrity of production shift analysis; step S2 constructs a multi-attribute process constraint system to adapt to the production needs of different process attributes; step S3 generates differentiated theoretical working hour benchmarks to eliminate systematic biases in the evaluation system; step S4 separates interference components in time deviations to improve the accuracy of delay judgment; step S5 achieves automatic identification of delay events to reduce the lag in manual statistics; and step S6 locates key links that restrict production capacity to provide quantitative basis for production scheduling optimization.

[0181] like Figure 2 As shown, this embodiment provides an example of a dynamic evaluation system for production cycle time based on multi-attribute process constraints. In this embodiment, the dynamic evaluation system for production cycle time is applied to the dynamic evaluation method for production cycle time as described in the above embodiment.

[0182] Specifically, the production cycle dynamic evaluation system includes a continuous batch operation event sequence construction module 1, which is electrically or communicatively connected in sequence; a multi-attribute process constraint rule library construction module 2; a dynamic theoretical working time calculation module 3; a time deviation and residual component calculation module 4; a delay event set acquisition module 5; and a production line maximum damping point identification module 6.

[0183] The continuous batch operation event sequence construction module 1 generates dynamic time windows covering production shifts using an event-driven time slicing algorithm, and preprocesses externally acquired discrete operation data through these dynamic time windows to construct continuous batch operation event sequences. The multi-attribute process constraint rule library construction module 2 extracts process attribute features from the continuous batch operation event sequences and constructs a multi-attribute process constraint rule library using a gradient boosting regression tree. The dynamic theoretical working time calculation module 3 matches correction factors for corresponding process attributes from the multi-attribute process constraint rule library and calculates the dynamic theoretical working time based on the rated production rate and batch operation volume. The dynamic theoretical working time; the time deviation and residual component calculation module 4 is used to calculate the time deviation between the dynamic theoretical working time and the actual production working time of the corresponding batch, and separates the residual component of the time deviation through the STL time sequence decomposition algorithm; the delay event set acquisition module 5 is used to calculate the dynamic tolerance threshold of the residual component based on the 3σ principle, and generate the delay event set according to the dynamic tolerance threshold; the production line maximum damping point identification module 6 is used to extract the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and to count the maximum value of the synchronization gap between processes according to the delay event set to identify the maximum damping point of the production line.

[0184] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Figure 3 As shown, the electronic device 7 includes a processor 71 and a memory 72 coupled to the processor 71.

[0185] The memory 72 stores program instructions for implementing the production cycle dynamic evaluation method based on multi-attribute process constraints in any of the above embodiments.

[0186] The processor 71 is used to execute program instructions stored in the memory 72 to perform dynamic evaluation of production cycle time based on multi-attribute process constraints.

[0187] 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.

[0188] 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.

[0189] In the several embodiments provided in this application, it should be understood that the disclosed apparatus, devices, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, 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; the indirect coupling or communication connection between devices or units may be electrical, mechanical, signal, or other forms.

[0190] 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 description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A dynamic evaluation method for production cycle time based on multi-attribute process constraints, wherein the dynamic evaluation method for production cycle time is applied to discrete operation data extracted from an external industrial database, characterized in that, The dynamic evaluation method for production cycle time includes: Step S1: Generate a dynamic time window covering production shifts using an event-driven time slicing algorithm, and preprocess the discrete operation data acquired from the outside using the dynamic time window to construct a continuous batch operation event sequence. Step S2: Extract process attribute features from the continuous batch operation event sequence, and construct a multi-attribute process constraint rule library through gradient boosting regression tree; Step S3: Match the correction factor of the corresponding process attribute from the multi-attribute process constraint rule library, and calculate the dynamic theoretical working hours based on the rated production rate and batch operation volume. Step S4: Calculate the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and separate the residual components of the time deviation using the STL time series decomposition algorithm; Step S5: Calculate the dynamic tolerance threshold of the residual component based on the 3σ principle, and determine the generation of a set of delay events based on the dynamic tolerance threshold; Step S6: Extract the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and count the maximum value of the synchronization gap between processes according to the set of delay events to identify the maximum damping point of the production line.

2. The method for dynamic evaluation of production cycle time according to claim 1, characterized in that, Step S1 involves generating a dynamic time window covering production shifts using an event-driven time slicing algorithm, and preprocessing externally acquired discrete job data using the dynamic time window to construct a continuous batch job event sequence, including: Step S1.1: Obtain the original timestamp set of the discrete operation data from an external industrial database; Step S1.2: Obtain the production shift start and end time pairs of the original timestamp set using an event-driven time slicing algorithm; Step S1.3: Generate a dynamic time window covering the production shifts based on the start and end times of the production shifts; Step S1.4: Filter the discrete operation data through the dynamic time window covering the production shift to obtain a subset of effective data within the shift; Step S1.5: Filter out invalid null values ​​and duplicate records from the valid data subset within the shift, and aggregate them according to the batch identifier to obtain a continuous batch operation event sequence.

3. The method for dynamic evaluation of production cycle time according to claim 1, characterized in that, Step S2: Extract process attribute features from the continuous batch operation event sequence, and construct a multi-attribute process constraint rule base using gradient boosting regression trees, including: Step S2.1: Extract all process attribute fields from the continuous batch operation event sequence to generate the original process attribute dataset; Step S2.2: Perform numerical encoding conversion on the original process attribute dataset to obtain a numerical process attribute matrix; Step S2.3: Extract historical actual working hour data from the continuous batch operation event sequence and generate a historical working hour tag set; Step S2.4: Align the numerical process attribute matrix with the historical working time tag set according to the batch identifier to generate a gradient boosting regression tree training sample set; Step S2.5: Initialize the gradient boosting regression tree model and input the gradient boosting regression tree training sample set into the gradient boosting regression tree model for iterative training; Step S2.6: Output the trained gradient boosting regression tree model, which is the mapping relationship between process attributes and time correction factors. Step S2.7: The mapping relationship is stored in a structured manner to obtain a multi-attribute process constraint rule library.

4. The method for dynamic evaluation of production cycle time according to claim 1, characterized in that, Step S3 involves matching the correction factor for the corresponding process attribute from the multi-attribute process constraint rule library, and calculating the dynamic theoretical working hours based on the rated production rate and batch workload, including: Step S3.1: Extract the process attribute features of the batch to be evaluated from the continuous batch operation event sequence; Step S3.2: Input the process attribute features of the batch to be evaluated into the multi-attribute process constraint rule library, and match to obtain the correction factor of the corresponding process attribute; Step S3.3: Extract the rated production rate and batch quantity of the batch to be evaluated from the continuous batch operation event sequence; Step S3.4: Calculate the dynamic theoretical working hours using the correction factor of the corresponding process attribute, the rated production rate of the batch to be evaluated, and the workload of the batch to be evaluated.

5. The method for dynamic evaluation of production cycle time according to claim 1, characterized in that, Step S4, calculate the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and separate the residual components of the time deviation using the STL time series decomposition algorithm, including: Step S4.1: Extract the actual production hours of the batches corresponding to the dynamic theoretical working hours from the continuous batch operation event sequence; Step S4.2: Match the actual production hours of the batch corresponding to the dynamic theoretical working hours with the dynamic theoretical working hours one by one according to the batch identifier to generate working hour matching data pairs; Step S4.3: Sort the time matching data pairs according to the production time sequence to generate an equally spaced time sequence dataset; Step S4.4: Calculate the difference between each set of matched data in the equally spaced time-series dataset to obtain the original time deviation sequence; Step S4.5: Input the original time deviation sequence into the STL time series decomposition algorithm to perform time series component decomposition operation to obtain the residual components of the original time deviation sequence.

6. The method for dynamic evaluation of production cycle time according to claim 1, characterized in that, Step S5: Calculate the dynamic tolerance threshold of the residual component based on the 3σ principle, and determine the generation of a delay event set based on the dynamic tolerance threshold, including: Step S5.1: Calculate the arithmetic mean and sample standard deviation of the residual components; Step S5.2: Calculate the upper control limit of the residual based on the arithmetic mean and the sample standard deviation, using the 3σ principle, and define it as the dynamic tolerance threshold. Step S5.3: Traverse the original time deviation sequence and filter out time deviation items that are greater than the dynamic tolerance threshold; Step S5.4: Associate the time deviation items that are greater than the dynamic tolerance threshold with the corresponding batch identifier to obtain the set of delay events.

7. The method for dynamic evaluation of production cycle time according to claim 1, characterized in that, Step S6 involves extracting the timestamp information of adjacent processes from the continuous batch operation event sequence to calculate the synchronization gap between processes, and statistically analyzing the maximum value of the synchronization gap between processes based on the set of delay events to identify the maximum damping point of the production line, including: Step S6.1: Extract the start and end timestamps of all processes in the continuous batch operation event sequence to generate a process start and end timestamp set; Step S6.2: Pair adjacent processes in the process start and end timestamp set according to the preset process topology order to generate a set of adjacent process timestamp pairs; Step S6.3: Calculate the difference between the start time of the subsequent process and the end time of the preceding process for each group of data in the adjacent process timestamp pair set to obtain the synchronization gap set between processes; Step S6.4: Extract all associated batch identifiers from the delay event set to generate a delay batch identifier set; Step S6.5: Select the synchronization gap data corresponding to the delayed batch identifier set from the set of synchronization gaps between processes to obtain the delay-associated synchronization gap subset; Step S6.6: Traverse the subset of delay-related synchronization gaps and extract the maximum value to obtain the maximum synchronization gap value; Step S6.7: Locate the adjacent process pair corresponding to the maximum synchronization gap value and define it as the maximum damping point of the production line.

8. A dynamic evaluation system for production cycle time based on multi-attribute process constraints, wherein the dynamic evaluation system for production cycle time is applied to the dynamic evaluation method for production cycle time as described in any one of claims 1 to 7, characterized in that, The production cycle dynamic evaluation system includes: The continuous batch operation event sequence construction module is used to generate a dynamic time window covering production shifts through an event-driven time slicing algorithm, and to preprocess externally acquired discrete operation data through the dynamic time window to construct a continuous batch operation event sequence. The multi-attribute process constraint rule library construction module is used to extract process attribute features from the continuous batch operation event sequence and construct a multi-attribute process constraint rule library through gradient boosting regression tree; The dynamic theoretical working time calculation module is used to match the correction factor of the corresponding process attribute from the multi-attribute process constraint rule library, and calculate the dynamic theoretical working time according to the rated production rate and batch operation volume. The time deviation and residual component calculation module is used to calculate the time deviation between the dynamic theoretical working hours and the actual production working hours of the corresponding batch, and to separate the residual components of the time deviation through the STL time series decomposition algorithm. The delay event set acquisition module is used to calculate the dynamic tolerance threshold of the residual component based on the 3σ principle, and generate a delay event set based on the dynamic tolerance threshold. The production line maximum damping point identification module is used to extract the timestamp information of adjacent processes in the continuous batch operation event sequence to calculate the synchronization gap between processes, and to count the maximum value of the synchronization gap between processes based on the set of delay events to identify the maximum damping point of the production line.

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 cycle dynamic evaluation 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 cycle dynamic evaluation method as described in any one of claims 1 to 7.

Images