Milk tea raw material production line intelligent regulation and control method based on internet of things and digital twinning

Abstract

The application discloses a milk tea raw material production line intelligent regulation and control method based on the Internet of Things and digital twinning, and aims to solve the problems of high production task switching cost, serious cleaning waste and slow response to temporary order insertion in the existing milk tea raw material production line. The application collects real-time production data through the Internet of Things sensor, constructs a raw material production line optimization model reflecting the current state of the production line, calculates the similarity of the material characteristics of the production task and the production switching cost dimension, and generates a comprehensive similarity index by weighting. Based on the comprehensive similarity, multiple candidate regulation and control schemes are generated, the production cost of each scheme is simulated and predicted, and the optimal target scheme is selected. According to the actual production effect, the dimension weight is adjusted adaptively, the hierarchical processing and local cleaning optimization of temporary order insertion are supported, the production switching cost of milk tea raw materials is reduced, the waste of materials and cleaning consumables is reduced, and the flexible production capacity and order response speed of the production line are improved.

Description

Technical Field

[0001] This application relates to the field of control methods for milk tea raw material production lines, and in particular to intelligent control methods for milk tea raw material production lines based on the Internet of Things and digital twins. Background Technology

[0002] In recent years, my country's freshly made tea beverage industry has experienced explosive growth, with the market size continuously expanding. Consumers' demands for diversified and personalized product flavors are constantly increasing, driving the upstream milk tea raw material production to shift from the traditional large-scale, single-variety model to a flexible production model with multiple varieties, small batches, and high-frequency switching. Milk tea raw material production is characterized by a wide variety of materials, precise component ratios, high risk of cross-contamination, and a large proportion of cleaning costs during production switching. Traditional manual scheduling and fixed-process production methods can no longer meet the current market's requirements for production efficiency, product quality, and order response speed.

[0003] In the production process of milk tea raw materials, the production line usually needs to continuously handle the production tasks of multiple different formulas and material components. The switching between different production tasks often involves complex operations such as equipment cleaning, material replacement and parameter adjustment.

[0004] Currently, production scheduling and switchover rely primarily on manual experience or simple static rules, such as prioritizing tasks based solely on delivery dates or material categories. However, this extensive management approach fails to adequately consider the real-time physical state of the production line, such as the relationship between residual materials in equipment, current processing progress, and the deep connections between different tasks in terms of material availability and switchover costs. This leads to frequent problems in actual production, including over-cleaning, material waste, and prolonged downtime due to improper task coordination, severely restricting the overall efficiency and flexible response capabilities of the production line. Summary of the Invention

[0005] The purpose of this application is to address the problem that in existing milk tea ingredient production lines, the switching and control of production tasks fails to combine the real-time status of the production line with the multi-dimensional characteristics of the tasks, resulting in high switching costs and low production efficiency. This application provides an intelligent control method for milk tea ingredient production lines based on the Internet of Things and digital twins.

[0006] To achieve the above objectives, the intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins provided in this application adopts the following technical solution:

[0007] This application discloses a method for intelligent control of a milk tea ingredient production line based on the Internet of Things and digital twins, including:

[0008] By collecting production data in real time through IoT sensors deployed on the milk tea raw material production line, an optimization model for the raw material production line reflecting the current production status of the milk tea raw material production line is constructed based on the real-time production data.

[0009] Acquire feature data for each production task in multiple production tasks. The feature data includes at least the material feature dimension and the production changeover cost dimension. For any two production tasks, determine the dimension similarity index corresponding to each dimension based on the feature data of the two production tasks in each dimension.

[0010] Based on the similarity indicators of each dimension and the corresponding dimension weight coefficients, a comprehensive similarity indicator between two production tasks is generated by weighted fusion. The dimension weight coefficients are used to characterize the degree of influence of the corresponding dimension on the comprehensive similarity of the production tasks.

[0011] By using the raw material production line optimization model, multiple candidate control schemes are generated based on the comprehensive similarity index between each production task. The predicted production costs corresponding to each candidate control scheme are simulated, and the target control scheme is selected from the multiple candidate control schemes based on the predicted production costs.

[0012] The target control plan is implemented, and the actual production effect data after implementation is collected. Based on the deviation between the actual production effect data and the predicted production cost, the dimensional weight coefficients corresponding to the similarity indicators of each dimension are adjusted.

[0013] Preferably, a raw material production line optimization model is constructed based on real-time production data to reflect the current production status of the milk tea raw material production line, including:

[0014] Obtain an initial digital twin model of the milk tea raw material production line. The initial digital twin model includes models of each production equipment in the milk tea raw material production line and the material flow paths between each production equipment.

[0015] Acquire real-time production data collected by IoT sensors. Real-time production data includes equipment operating status data and material handling data of each production device.

[0016] For each production equipment, the equipment operation status data and material handling data are mapped to the corresponding equipment model to update the operation status parameters, material handling status parameters and material residue parameters. The material handling status parameters include the type of material currently being processed by the corresponding production equipment and the processing progress. The material residue parameters include the type of material remaining in the corresponding production equipment and the residue value.

[0017] Based on the updated equipment models and material flow paths, the real-time distribution of materials in the milk tea raw material production line is determined, and a raw material production line optimization model reflecting the current overall production status of the milk tea raw material production line is generated.

[0018] Preferably, based on the feature data of the two production tasks in each dimension, the dimensional similarity index corresponding to each dimension is determined, including:

[0019] Based on the raw material production line optimization model, the operating status parameters, material handling status parameters, material residue parameters, and real-time material distribution data in the production line of each production equipment model are extracted as real-time benchmark data for calculating the dimensional similarity index.

[0020] Based on the material characteristics dimension, extract the material type and component ratio characteristic data corresponding to the two production tasks. Combine the real-time material distribution data, the material types in each production equipment and the material residue parameters to calculate the common material ratio, component ratio deviation value and matching degree with the residual material in the current production line for the two production tasks respectively.

[0021] The weights of common material proportion, component ratio deviation and residual material matching degree are set respectively, and the dimensional similarity index corresponding to the material feature dimension is obtained by weighted summation.

[0022] For the production switchover cost dimension, we extract the characteristic data of equipment cleaning requirements and material adjustment requirements required for the production switchover of two production tasks. Combined with real-time material distribution data, material types in each production equipment and material residue parameters, we calculate the estimated values ​​of equipment cleaning cost, material adjustment cost and downtime cost corresponding to the switchover of two production tasks.

[0023] For equipment cleaning costs, material adjustment costs, and downtime costs, preset benchmark cost thresholds are set. The estimated values ​​of each item are normalized with the benchmark cost thresholds to obtain the normalization coefficients corresponding to each cost. Based on the normalization coefficients, the dimensional similarity index of the two production tasks in the production switching cost dimension is calculated.

[0024] Preferably, the dimensional similarity index of two production tasks in the production switchover cost dimension is calculated based on the normalization coefficient, including:

[0025] For any two production tasks to be produced, denoted as production task i and production task j respectively, the expression for the dimensional similarity index of the two production tasks in terms of material characteristics is as follows:

[0026]

[0027] In the formula,

[0028] This is the dimensional similarity index between production task i and production task j in terms of material characteristics.

[0029] , , These are the sub-weights corresponding to the proportion of common materials, the component ratio deviation value, and the residual material matching degree, respectively. + + =1;

[0030] The percentage of materials shared by production task i and production task j is the ratio of the number of shared material types between the two production tasks to the total number of material types in the two production tasks.

[0031] is the component ratio deviation value between production task i and production task j, and is the normalized mean of the relative deviation of the component ratio of the common materials in the two production tasks.

[0032] The matching degree between production task i and production task j and the residual materials in the current production line;

[0033] The similarity index between production task i and production task j in terms of production switchover cost is expressed as follows:

[0034]

[0035] In the formula,

[0036] This is the dimensional similarity index between production task i and production task j in the production switching cost dimension.

[0037] , , These are the weighting coefficients corresponding to equipment cleaning costs, material adjustment costs, and downtime costs, respectively. + + =1;

[0038] The normalization coefficient for the equipment cleaning cost when switching between production task i and production task j;

[0039] The normalization coefficient for the material adjustment cost corresponding to the switching between production task i and production task j;

[0040] It is the normalization coefficient of the downtime cost corresponding to the switching between production task i and production task j.

[0041] Preferably, multiple candidate control schemes are generated based on the comprehensive similarity index between various production tasks using a raw material production line optimization model, including:

[0042] The current equipment operating status parameters, equipment capacity limits, material supply sequence, and delivery time requirements of each production task are extracted from the raw material production line optimization model and used as the basic constraints for generating candidate control schemes.

[0043] With maximizing the comprehensive similarity index between adjacent production tasks as the core ranking objective and using basic constraints as boundaries, all production tasks to be produced are combined and ranked in multiple rounds to generate multiple initial production task ranking combinations.

[0044] For each initial production task sorting combination, based on the similarity index of material characteristics and the similarity index of production switching cost of adjacent production tasks, the component compensation strategy and graded cleaning strategy are matched when switching between adjacent production tasks to form an initial control plan.

[0045] The initial control schemes of each group were pre-simulated and verified using the raw material production line optimization model. Infeasible schemes that did not meet the requirements of equipment capacity, material supply and delivery time were eliminated, and the initial control schemes that passed the verification were determined as candidate control schemes.

[0046] Preferably, the predicted production costs corresponding to each candidate control scheme include:

[0047] The unit time operating cost, unit material loss cost, unit cleaning consumable cost, and unit time downtime loss cost of each production equipment are extracted from the raw material production line optimization model and used as the basic cost parameters for predicting production costs.

[0048] Each candidate control scheme is broken down into multiple continuous production task execution stages and adjacent production task switching stages. Each stage is then virtually simulated using a raw material production line optimization model.

[0049] During the virtual simulation of the production task execution phase, the entire process of material transportation, batching and metering, mixing and processing and finished product output is simulated simultaneously, and the equipment operating cost, material loss cost and production cycle of the corresponding stage are calculated.

[0050] During the virtual simulation of the transition between adjacent production tasks, based on the matching component compensation strategy and graded cleaning strategy in the candidate control scheme, the material component adjustment process and equipment cleaning operation process are simulated simultaneously, and the material adjustment cost, cleaning consumable cost and downtime loss cost of the corresponding stage are calculated.

[0051] The costs of all production task execution phases and adjacent production task switching phases are summarized to obtain the predicted production costs of the corresponding candidate control schemes. The predicted production costs include at least the total equipment operating costs, total material loss costs, total cleaning consumable costs, and total downtime loss costs.

[0052] Preferably, it also includes emergency order scheduling tasks, which include:

[0053] Receive urgent production task instructions and obtain full-dimensional feature data of the production task that needs to be expedited. The full-dimensional feature data includes at least material feature dimension data, production changeover cost dimension data, and delivery time requirements.

[0054] Extract the feature data of the currently executing production task, real-time production data, material residue parameters in each production equipment, and real-time material distribution data from the raw material production line optimization model;

[0055] Calculate the comprehensive similarity index between the required expedited production tasks and the currently executed production tasks, and compare the comprehensive similarity index with the preset order insertion similarity threshold;

[0056] When the comprehensive similarity index is greater than or equal to the first preset insertion similarity threshold, it is determined to be a highly adaptable insertion task. After the currently executing production task is completed, the emergency production task is started.

[0057] When the comprehensive similarity index is less than the first preset insertion similarity threshold and greater than or equal to the second preset insertion similarity threshold, it is determined to be a medium-adaptation insertion task. The target component compensation amount is calculated based on the raw material production line optimization model, and the component compensation strategy is executed before the emergency production task is carried out.

[0058] When the overall similarity index is less than the second preset insertion similarity threshold, it is judged as a low-adaptability insertion task. After the production task currently being executed is completed, each production equipment is cleaned.

[0059] Preferably, after the emergency order scheduling task is completed, it also includes:

[0060] Extract real-time status data of the production line after the completion of the emergency production task from the raw material production line optimization model. The real-time status data of the production line includes at least the operating status parameters of each production equipment, material residual parameters, real-time material distribution data, and the actual production cycle and actual production cost of the emergency production task.

[0061] Determine the set of subsequent production tasks affected by the emergency production task in the original target control plan. The set of subsequent production tasks is all the unfinished production tasks originally planned to be executed after the time point of the emergency production task insertion.

[0062] Based on the full-dimensional feature data of the emergency production task and the extracted real-time status data of the production line, the comprehensive similarity index between each production task in the subsequent production task set and the emergency production task, as well as the comprehensive similarity index between any two production tasks in the subsequent production task set, are recalculated.

[0063] With the core objective of maximizing the comprehensive similarity index between adjacent production tasks, and with the real-time status of the production line after the execution of emergency production tasks and the delivery date requirements of each subsequent production task as constraints, multiple updated candidate control schemes are generated.

[0064] The predicted production cost and total order delay time corresponding to each updated candidate control scheme are simulated by the raw material production line optimization model. Based on the weighted comprehensive value of the predicted production cost and total order delay time, the optimal updated target control scheme is selected from multiple updated candidate control schemes.

[0065] The updated target control plan is implemented to produce the tasks in the subsequent set of tasks awaiting production.

[0066] Preferably, the method generates multiple candidate control schemes based on the comprehensive similarity index between various production tasks using a raw material production line optimization model, and also includes:

[0067] For any pair of adjacent production tasks in each initial production task sorting combination, extract the production process sequence and material type data of each production process corresponding to the two production tasks from the raw material production line optimization model.

[0068] By comparing the consistency of material types in the corresponding production stages of two production tasks, the set of production stages with inconsistent material types is identified.

[0069] When the set of differential production links is not empty and contains only some production links, a local cleaning strategy is generated for each production equipment in the set of differential production links. The local cleaning strategy only performs cleaning operations on the production equipment corresponding to the production links with inconsistent material types.

[0070] Preferably, after identifying the set of production stages with inconsistent material types, the estimated value of downtime cost is calculated, including:

[0071] Based on the cleaning time parameters of each production equipment in the raw material production line optimization model and the material residue parameters of each production equipment in the differential production process set, the estimated value of the local downtime cost corresponding to the execution of the local cleaning strategy is calculated.

[0072] Calculate the estimated downtime cost of implementing the full-line cleaning strategy, and compare the estimated downtime cost of the local downtime with the estimated downtime cost of the full line.

[0073] When the estimated cost of local downtime is less than or equal to the preset local cleaning time threshold, the local cleaning strategy will be determined as the optimal switching cleaning strategy for the group of adjacent production tasks.

[0074] When the estimated cost of local downtime exceeds the preset local cleaning time threshold, the order of adjacent production tasks in the initial production task sorting combination is adjusted, and the initial control plan is regenerated.

[0075] Compared with existing technologies, this application provides an intelligent control method for milk tea raw material production lines based on the Internet of Things and digital twins, which has the following beneficial effects:

[0076] 1. By constructing a raw material production line optimization model, the similarity of production tasks in the dimensions of material characteristics and production switching costs is calculated and weighted to generate a comprehensive similarity index. With the core objective of maximizing comprehensive similarity, candidate control schemes are generated. The predicted production costs of each scheme are predicted through virtual simulation to select the optimal target control scheme. By anchoring the goal of reducing production switching costs from the source of production scheduling, the problem of the disconnect between static production scheduling and actual switching costs in existing technologies is effectively solved. It can significantly reduce unnecessary cleaning operations and material waste, and improve the overall production efficiency of the production line.

[0077] 2. A three-level hierarchical decision-making mechanism for emergency order scheduling is adopted. By comparing the comprehensive similarity index of the emergency task and the currently executed task with the first preset order similarity threshold and the second preset order similarity threshold, the direct order insertion, component compensation order insertion, and cleaning order insertion strategies are matched respectively. After the emergency task is completed, an updated candidate control plan is regenerated based on the real-time status. This achieves rapid response to emergency orders, minimizes the disturbance of the order insertion operation to the original production plan, and balances the delivery of emergency tasks with the regular production order.

[0078] 3. By comparing the consistency of material types in adjacent production tasks at each stage, the system identifies sets of differing production stages and generates corresponding local cleaning strategies. At the same time, it calculates the estimated cost of local downtime and compares it with the cost of cleaning the entire production line to select the optimal switching cleaning strategy. This avoids the full-line cleaning mode used in existing technologies, achieves precise control of cleaning scope and cleaning time, further reduces cleaning consumable costs and downtime losses, and improves the efficiency of connection between production tasks. Attached Figure Description

[0079] Figure 1 This is a flowchart illustrating the steps of an intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins, as described in this application.

[0080] Figure 2 This is a flowchart illustrating the steps involved in constructing a raw material production line optimization model based on real-time production data to reflect the current production status of the milk tea raw material production line in this method.

[0081] Figure 3 This is a flowchart illustrating the steps in this method to determine the dimensional similarity index for each dimension based on the feature data of two production tasks in each dimension.

[0082] Figure 4 This is a flowchart of the steps involved in executing the emergency order scheduling task in this method. Detailed Implementation

[0083] First, it should be noted that the data and values ​​involved in this embodiment are the values ​​taken for the example and are used to assist in the explanation of the formula or technical solution. The source of the data will not be explained in detail in the following description.

[0084] The following is in conjunction with the appendix Figure 1-4 This application will be described in further detail.

[0085] This application discloses an intelligent control method for a milk tea ingredient production line based on the Internet of Things and digital twins. (Refer to...) Figure 1 The intelligent control method for milk tea raw material production lines based on the Internet of Things and digital twins includes:

[0086] S1. Real-time production data is collected by IoT sensors deployed on the milk tea raw material production line, and a raw material production line optimization model reflecting the current production status of the milk tea raw material production line is constructed based on the real-time production data.

[0087] The production equipment for the milk tea raw material production line includes raw material storage tanks, ingredient tanks, mixers, homogenizers, sterilizers, and filling machines.

[0088] Among them, IoT sensors include temperature sensors, pressure sensors, flow sensors, liquid level sensors, speed sensors, and material identification sensors.

[0089] The production data includes the operating status data of each production equipment and material handling data;

[0090] The operating status data includes equipment start / stop status, operating speed, operating temperature, and operating pressure;

[0091] The material handling data includes material type, material flow rate, material level, and processing progress.

[0092] S2. Obtain the feature data of each production task in multiple production tasks. The feature data includes at least the material feature dimension and the production switching cost dimension. For any two production tasks, determine the dimension similarity index corresponding to each dimension based on the feature data of the two production tasks in each dimension.

[0093] The material characteristic dimension data includes the material type, component ratio, and material usage corresponding to the production task; the production switchover cost dimension data includes the equipment cleaning requirements, material adjustment requirements, and equipment parameter adjustment requirements required for the production task switchover.

[0094] S3. Based on the similarity indicators of each dimension and the corresponding dimension weight coefficients of each dimension similarity indicator, a comprehensive similarity indicator between two production tasks is generated by weighted fusion. The dimension weight coefficient is used to characterize the degree of influence of the corresponding dimension on the comprehensive similarity of the production tasks.

[0095] S4. Using the raw material production line optimization model, multiple candidate control schemes are generated based on the comprehensive similarity index between each production task. The predicted production costs corresponding to each candidate control scheme are simulated respectively, and the target control scheme is selected from the multiple candidate control schemes based on the predicted production costs.

[0096] S5. Execute the target control plan, organize production according to the production task sorting and switching strategy in the target control plan, continuously collect actual production effect data after execution through IoT sensors, and adjust the dimension weight coefficients corresponding to the similarity indicators of each dimension based on the deviation between the actual production effect data and the predicted production cost. The actual production effect data includes actual equipment operating costs, actual material loss costs, actual cleaning consumable costs, actual downtime loss costs, and the actual completion time of each production task.

[0097] Furthermore, refer to Figure 2 Based on real-time production data, a raw material production line optimization model reflecting the current production status of the milk tea raw material production line is constructed, including:

[0098] S11. Obtain the initial digital twin model of the milk tea raw material production line. The initial digital twin model includes the models of each production equipment in the milk tea raw material production line and the material flow path between each production equipment.

[0099] The initial digital twin model is a one-to-one mapping of the equipment composition, geometric structure, and basic production logic of the milk tea raw material physical production line. It is constructed using a combination of 3D solid modeling and logical modeling, wherein:

[0100] Each production equipment model includes static attributes such as the geometric dimensions, rated parameters, operating characteristics, and fault thresholds of the corresponding physical equipment, as well as initial values ​​for dynamic attributes such as operating status parameters, material handling status parameters, and material residue parameters.

[0101] Here is a brief description of its construction method:

[0102] In this embodiment, the initial digital twin model is constructed using the Unity3D engine combined with an industrial digital twin platform. The specific implementation steps are as follows:

[0103] A 3D laser scan was performed on the milk tea raw material production line to obtain the precise geometric dimensions of each production equipment and the overall layout data of the production line, and a 3D geometric model of the production line was constructed.

[0104] Import the technical manual parameters of each production equipment and add static attributes to each equipment model, including: the rated volume of the raw material storage tank, the stirring power of the batching tank, the maximum mixing capacity of the mixer, the rated sterilization temperature of the sterilizer, and the filling speed of the filling machine.

[0105] Based on the actual process flow of the production line, define the material flow path between each production equipment, including: the conveying pipeline from raw material storage tank to batching tank, the conveying pipeline from batching tank to mixer, the conveying pipeline from mixer to homogenizer, the conveying pipeline from homogenizer to sterilizer, and the conveying pipeline from sterilizer to filling machine. Add valve control logic and conveying speed parameters to each conveying pipeline.

[0106] Initialize dynamic parameters for each equipment model, including: the initial value for the operating status parameter is "stopped", the initial value for the material handling status parameter is "no material", and the initial value for the material residue parameter is 0L (International System of Units, liters).

[0107] The material flow paths between each production equipment are defined in advance according to the standard process flow. The conveying sequence, pipeline connection relationship, conveying speed range and switching valve control logic of the material from the raw material storage tank through the batching tank, mixer, homogenizer, sterilizer to the filling machine are clearly defined. This is the initial digital twin model commonly used in existing milk tea raw material production lines. Its specific working principle will not be elaborated here.

[0108] S12. Acquire real-time production data collected by IoT sensors. Real-time production data includes equipment operating status data and material handling data of each production device.

[0109] Equipment operating status data includes the start / stop status, operating speed, working temperature, working pressure, motor current, and valve opening parameters of each production equipment, such as the stirring speed of the batching tank, the mixing time of the mixer, and the sterilization temperature and holding time of the sterilizer.

[0110] Material handling data includes the types of materials, flow rates, liquid levels, and temperatures of materials in each production device, as well as processing stage indicators, such as the remaining amount of non-dairy creamer in the raw material storage tank, the liquid level of the tea base, and the components and proportions of materials currently being mixed in the mixing tank.

[0111] S13. For each production equipment, map the equipment operation status data and material handling data to the corresponding equipment model to update the operation status parameters, material handling status parameters and material residue parameters. The material handling status parameters include the type of material currently being processed by the corresponding production equipment and the processing progress. The material residue parameters include the type of material remaining in the corresponding production equipment and the residue value.

[0112] The material residue parameter is calculated using a material balance algorithm to determine the material residue in each production device. The expression is as follows:

[0113]

[0114] In the formula, Let L be the residual amount of the nth material in the kth production equipment at time t;

[0115] The residual amount of the nth material in the kth production equipment at the end of the previous calculation cycle, in L, where k is the unique number of the production equipment, ranging from 1, 2, ..., K, and K is the total number of production equipment in the production line.

[0116] n is the unique number of the material type, with a value range of 1, 2, ..., N, where N is the total number of material types involved in the production line.

[0117] For the time of the previous calculation cycle, in this embodiment, Δt = t - =100ms is a fixed calculation period;

[0118] for The feed flow rate of the nth material in the kth production equipment at time k, in L / min;

[0119] )for The discharge flow rate of the nth material in the kth production equipment at time k, in L / min;

[0120] To illustrate with an example, if the material residue of the mixer at the previous moment was 20L, and in the next 10 minutes, 480L of material was fed and 475L of material was discharged, then the current material residue is calculated to be 25L. The material residue parameter of the mixer model is then updated to the following: material type: original milk tea base, residue value: 25L.

[0121] Furthermore, the total material residue in a single production unit is the sum of all material residues within that unit, expressed as:

[0122]

[0123] In the formula, Let L be the total amount of material remaining in the k-th production equipment at time t, and N be the total number of material types involved in the production line.

[0124] Furthermore, for any conveying pipeline and any material, the material inventory is calculated using the pipeline volume-velocity model under three operating conditions, expressed as follows:

[0125]

[0126] In the formula, Let L be the quantity of the nth type of material in the mth conveying pipeline at time t;

[0127] Let m be the cross-sectional area of ​​the m-th conveying pipeline, in units of ;

[0128] The total length of the m-th transport pipeline, in meters;

[0129] The density of the nth material is expressed in kg / L. In this embodiment, the density of the liquid milk tea raw material is approximately 1.02-1.05 kg / L, which has been pre-calibrated through experiments.

[0130] Let be the flow velocity of the material in the m-th conveying pipeline at time t, in m / s;

[0131] The calculation period is 100ms.

[0132] Furthermore, based on the calculated inventory of the nth material in the m-th conveying pipeline at time t, the total material inventory in a single conveying pipeline is the sum of the inventories of all materials in that pipeline, expressed as:

[0133]

[0134] In the formula, Let L be the total material inventory in the m-th conveying pipeline at time t.

[0135] Based on the above steps, the real-time material distribution expression is derived as follows:

[0136]

[0137] In the formula, Let t be the real-time material distribution matrix at time t; K is the total number of production equipment in the production line; M is the total number of conveying pipes in the production line; the first K rows of the matrix correspond to each production equipment, and the last M rows correspond to each conveying pipe.

[0138] S14. Based on the updated equipment models and material flow paths, determine the real-time distribution of materials in the milk tea raw material production line and generate a raw material production line optimization model that reflects the current overall production status of the milk tea raw material production line.

[0139] This involves integrating all updated production equipment models, material flow paths, and real-time material distribution matrices to generate a raw material production line optimization model. This model reflects the equipment operating status of the production line, as well as the location, quantity, and flow status of each material within the production line.

[0140] Furthermore, refer to Figure 3 Based on the feature data of the two production tasks in each dimension, the dimensional similarity index corresponding to each dimension is determined, including:

[0141] S21. Based on the raw material production line optimization model, extract the operating status parameters, material handling status parameters, material residue parameters, and real-time material distribution data in the production line of each production equipment model as real-time benchmark data for calculating the dimensional similarity index.

[0142] To provide a better explanation, the parameters are described below:

[0143] Operating status parameters include equipment start / stop status, operating speed, operating temperature, and equipment health, which are used to determine whether the equipment is ready to switch production tasks.

[0144] Material handling status parameters, including the type of material currently being processed, processing progress, and remaining processing time, are used to assess the timing of production task switching.

[0145] The material residue parameter includes the residue value of each material in each piece of equipment, which is used to calculate the material compatibility and cleaning cost;

[0146] Real-time material distribution data, i.e., the complete real-time material distribution matrix at time t, is used to globally assess the matching degree between existing materials on the production line and the tasks to be produced.

[0147] It should be noted that the extraction cycle of the above data is synchronized with the update frequency of the raw material production line optimization model.

[0148] S22. Based on the material characteristic dimension, extract the material type and component ratio characteristic data corresponding to the two production tasks. Combine the real-time material distribution data, the material types in each production equipment and the material residue parameters to calculate the common material ratio, component ratio deviation value and matching degree with the residual material in the current production line for the two production tasks respectively.

[0149] Specifically, for any two production tasks to be produced, denoted as production task i and production task j respectively, the number of material types common to both tasks and the total number of material types are counted. The expression for the proportion of common materials is as follows:

[0150]

[0151] In the formula, The quantity of material types common to production task i and production task j;

[0152] The union quantity of all material types for two production tasks.

[0153] To illustrate with an example: Production task i is for original flavor milk tea base, and the materials include non-dairy creamer, white sugar, and black tea extract; production task j is for vanilla milk tea base, and the materials include non-dairy creamer, white sugar, black tea extract, and vanilla flavoring. Therefore, the total number of material types is 3, and the total number of material types is 4. The percentage of common materials is... .

[0154] S23. Set corresponding sub-weights for the proportion of common materials, the deviation value of component ratio, and the matching degree of residual materials, and obtain the dimension similarity index corresponding to the material feature dimension after weighted summation.

[0155] Furthermore, for each material common to both production tasks, the relative deviation of its component ratio is calculated, and the normalized mean is taken to obtain the component ratio deviation value, expressed as:

[0156] In the formula, Let n be the mass percentage of the nth common material in production task i. Let be the mass percentage of the nth common material in production task j.

[0157] To better illustrate, based on the previous example, the components of production task i are 20% non-dairy creamer, 10% white sugar, and 70% black tea extract; the components of production task j are 20% non-dairy creamer, 10% white sugar, 65% black tea extract, and 5% vanilla flavoring. The relative deviations of the three common materials are 0, 0, and (70%-65%) / 70%≈0.0714, respectively. Therefore, the component ratio deviation value is... .

[0158] Correspondingly, based on the real-time material distribution matrix D(t), the overlap between production task i and production task j and the residual materials in the production line is calculated, and the average value is taken to obtain the residual material matching degree, which is expressed as:

[0159]

[0160] In the formula, The amount of overlap between the material type of production task i and the material type remaining in the production line;

[0161] The total number of material types for production task i;

[0162] The amount of overlap between the material type of production task j and the material type remaining in the production line;

[0163] This represents the total number of material types for production task j.

[0164] Based on the above example, the remaining material in the current production line is the original flavor milk tea base, containing non-dairy creamer, white sugar, and black tea extract. Therefore, the overlap between production task i and the remaining material is 3, with a matching degree of 1; the overlap between production task j and the remaining material is 3, with a matching degree of 0.75; the matching degree of the remaining material is... =0.875.

[0165] Furthermore, the expression for the dimensional similarity index between the two production tasks in terms of material characteristics is as follows:

[0166]

[0167] In the formula,

[0168] This is the dimensional similarity index between production task i and production task j in terms of material features. The value range is [0,1], where a larger value indicates a higher degree of matching of material features between the two production tasks. When the calculation result exceeds this range, it is automatically truncated to the boundary value, that is: less than 0 is taken as 0, and greater than 1 is taken as 1.

[0169] , , These are the sub-weights corresponding to the proportion of common materials, the component ratio deviation value, and the residual material matching degree, respectively. + + =1;

[0170] It should be noted here that the percentage of common materials is determined based on comprehensive production experience. Component proportion deviation value weight =0.3, weight of residual material matching degree =0.2.

[0171] The percentage of materials shared by production task i and production task j is the ratio of the number of shared material types between the two production tasks to the total number of material types in the two production tasks.

[0172] is the component ratio deviation value between production task i and production task j, and is the normalized mean of the relative deviation of the component ratio of the common materials in the two production tasks.

[0173] The matching degree between production task i and production task j and the residual materials in the current production line;

[0174] Continuing with the example above, substituting the values ​​yields:

[0175]

[0176] S24. For the production switchover cost dimension, extract the characteristic data of equipment cleaning requirements and material adjustment requirements required for the production switchover of the two production tasks. Combine the real-time material distribution data, the types of materials in each production equipment and the material residue parameters to calculate the estimated values ​​of equipment cleaning cost, material adjustment cost and downtime cost corresponding to the switchover of the two production tasks.

[0177] S25. Preset baseline cost thresholds for equipment cleaning costs, material adjustment costs, and downtime costs respectively. Normalize the estimated values ​​of each item with the baseline cost thresholds to obtain the normalization coefficients corresponding to each cost. Calculate the dimensional similarity index of the two production tasks in the production switching cost dimension based on the normalization coefficients.

[0178] The numerical expression for the estimated cleaning cost of computing equipment is as follows:

[0179]

[0180] In the formula, This represents the total amount of residual material currently inside the k-th device;

[0181] The cleaning difficulty coefficient is the residual material in the k-th equipment, with a value ranging from 1.0 to 3.0. The coefficient is 1.0 for easy-to-clean materials and 3.0 for easy-to-scale materials. The cost of cleaning per unit of residual amount is expressed in yuan / L.

[0182] Furthermore, the estimated material adjustment cost is calculated using the following expression:

[0183]

[0184] In the formula, The cost of materials required to supplement production task j; The waste cost of unused materials remaining for production task i.

[0185] Furthermore, the estimated cost of downtime is calculated, which includes the production loss corresponding to equipment cleaning time, material adjustment time, and equipment parameter adjustment time. The expression is as follows:

[0186]

[0187] In the formula, Total equipment cleaning time; Adjust the total time for materials; Total time for adjusting equipment parameters; The cost of downtime is expressed as yuan / min.

[0188] Based on the above steps, normalization is performed to obtain the normalization coefficients for each cost.

[0189] It should be noted that baseline thresholds for various costs are set in advance based on historical production line data, including:

[0190] Cleaning cost benchmark threshold =500 yuan; Material adjustment cost benchmark threshold =1000 yuan; Downtime baseline threshold Normalization is performed to obtain the normalized coefficients:

[0191] =

[0192] =

[0193] =

[0194] Furthermore, corresponding weighting coefficients are set for the three normalization coefficients mentioned above. , , These are the weighting coefficients corresponding to equipment cleaning costs, material adjustment costs, and downtime costs, respectively. + + =1, in this embodiment, the initial weight is set to: =0.2, =0.3, =0.5.

[0195] The expression for the similarity index in the production switchover cost dimension is as follows:

[0196]

[0197] In the formula,

[0198] This is the dimensional similarity index between production task i and production task j in the production switching cost dimension.

[0199] In summary, the formula for calculating the comprehensive similarity index is:

[0200]

[0201] In the formula, This is a comprehensive similarity index between production task i and production task j, with a value range of [0,1]. The larger the value, the higher the overall similarity between the two production tasks, making them suitable for scheduling as adjacent production tasks. This is the dimension weight coefficient corresponding to the material feature dimension, with an initial default value of 0.6; This refers to the dimension weight coefficient corresponding to the production switching cost dimension, with an initial default value of 0.4. .

[0202] Furthermore, through the raw material production line optimization model, multiple candidate control schemes are generated based on the comprehensive similarity index between various production tasks, including:

[0203] S41. Extract the current equipment operating status parameters, equipment capacity limit, material supply sequence, and delivery time requirements of each production task from the raw material production line optimization model as the basic constraints for generating candidate control schemes.

[0204] Before generating the initial production task sorting combination, the basic constraints at the current moment are first extracted from the raw material production line optimization model. All basic constraints are hard constraints, and any sorting combination that does not meet the basic constraints will be directly eliminated.

[0205] A. Equipment operating status parameters: Extract the start / stop status, health status, and remaining available time of each production equipment, and exclude equipment that is in a fault or maintenance state;

[0206] B. Equipment capacity limit: Extract the rated capacity of each production equipment;

[0207] C. Material supply sequence: Extract the remaining inventory and estimated arrival time of each raw material;

[0208] D. Delivery time requirements for each production task: Extract the latest completion time for all tasks to be produced.

[0209] S42. Taking the maximization of the comprehensive similarity index between adjacent production tasks as the core sorting objective and the basic constraints as the boundary, perform multiple rounds of combined sorting on all production tasks to be produced to generate multiple sets of initial production task sorting combinations.

[0210] Based on the comprehensive similarity index, an N×N dimensional comprehensive similarity matrix is ​​constructed. Where N is the total number of tasks to be produced:

[0211]

[0212] In the formula, Let be the comprehensive similarity index between task i and task j, and .

[0213] Furthermore, based on the aforementioned comprehensive similarity matrix The fitness function of a genetic algorithm is used, where the fitness function is defined as the sum of the comprehensive similarity indices of all adjacent production tasks in the sorting combination, with a constraint penalty term added. The expression is:

[0214]

[0215] In the formula, F is the fitness value, and the larger the value, the better the ranking combination;

[0216] Sort and combine production tasks;

[0217] This is the penalty coefficient, with a value of 10;

[0218] The penalty value for the c-th constraint condition is when the constraint is satisfied. =0, when not satisfied =1, where C is the total number of constraints.

[0219] Furthermore, based on the above process, the genetic algorithm parameters are set as follows: population size 50, crossover probability 0.8, mutation probability 0.1, and number of iterations 100. During the execution of the genetic algorithm, the top 10 sorting combinations with the highest fitness values ​​in each generation of the population are retained, and finally 10 initial production task sorting combinations with the highest fitness values ​​are generated.

[0220] S43. For each group of initial production task sorting combinations, based on the similarity index of material characteristics dimension and the similarity index of production switching cost dimension of adjacent production tasks, match the component compensation strategy and graded cleaning strategy when switching adjacent production tasks to form an initial control plan.

[0221] The matching rules are shown in Table 1:

[0222] ≥0.8 ≥0.7 Direct switching, no cleaning or compensation required. 0.7～0.8 0.6～0.7 Component compensation strategy 0.6～0.7 0.5～0.6 Tiered cleaning strategy, i.e., spot cleaning <0.6 <0.5 Full-line cleaning strategy

[0223] Table 1

[0224] In the example, for the adjacent production task A → production task B in combination 1, the similarity in the material feature dimension is 0.8429, and the similarity in the production switching cost dimension is 0.643. The matching component compensation strategy is to add 5kg of vanilla flavoring to the ingredients of task B without the need for equipment cleaning.

[0225] S44. The initial control schemes of each group are pre-simulated and verified using the raw material production line optimization model. Infeasible schemes that do not meet the requirements of equipment capacity, material supply and delivery time are eliminated, and the initial control schemes that pass the verification are determined as candidate control schemes.

[0226] The initial control schemes for each group were pre-simulated and verified using a raw material production line optimization model. The simulation time step was 1 minute, and the simulation duration covered the latest completion time of all production tasks. The following indicators were monitored in real time during the simulation:

[0227] Does the actual operating load of each production piece of equipment exceed the equipment's maximum capacity?

[0228] Whether the material supply at each stage of production is timely and whether there are any material shortages;

[0229] Whether the actual completion time of each production task meets the delivery deadline requirements;

[0230] Based on the above process, all infeasible schemes that do not meet the above indicators are eliminated, and the initial control schemes that pass the verification are determined as candidate control schemes.

[0231] Furthermore, through the raw material production line optimization model, multiple candidate control schemes are generated based on the comprehensive similarity index between various production tasks, including:

[0232] S41a. For any pair of adjacent production tasks in each initial production task sorting combination, extract the production process sequence and material type data of each production process corresponding to the two production tasks from the raw material production line optimization model.

[0233] include:

[0234] The production sequence of each of the two production tasks;

[0235] The set of input material types and the set of output material types for each production stage;

[0236] Each production stage corresponds to a unique equipment number.

[0237] S41b. Compare the consistency of material types between two production tasks at corresponding production stages to identify the set of production stages with inconsistent material types.

[0238] S41c When the set of differential production links is not empty and contains only some production links, generate a local cleaning strategy for each production equipment in the set of differential production links. The local cleaning strategy only performs cleaning operations on the production equipment corresponding to the production links with inconsistent material types.

[0239] include:

[0240] Extract the material residue parameters and cleaning difficulty coefficients of the corresponding equipment in the different production stages;

[0241] Match the equipment with the corresponding standard operating procedures for cleaning, and determine the cleaning level, cleaning time, and amount of cleaning consumables required;

[0242] Generate a local cleaning instruction, specifying that cleaning operations should only be performed on the production equipment corresponding to the different production processes, while the remaining production equipment should remain in standby mode and do not require cleaning operations.

[0243] The generated local cleaning strategy replaces the corresponding hierarchical cleaning strategy in the initial control scheme to form an updated initial control scheme, which then enters the pre-simulation verification stage.

[0244] To better illustrate this point, we will use an example below.

[0245] Two adjacent production tasks are defined as Production Task B and Production Task C, respectively. Production Task B is for vanilla milk tea base, and Production Task C is for chocolate milk tea base. The production process and material data for Production Task B and Production Task C are shown in Table 2.

[0246] Raw material pretreatment Non-dairy creamer solution, white sugar solution, black tea extract, vanilla flavoring Non-dairy creamer solution, white sugar solution, black tea extract, cocoa powder E1 Raw Material Storage Tank Group Ingredients Non-dairy creamer solution, white sugar solution, black tea extract, vanilla flavoring Non-dairy creamer solution, white sugar solution, black tea extract, cocoa powder E2 Ingredient Tank mix Vanilla Milk Tea Premix Chocolate Milk Tea Premix E3 Mixer homogenization Vanilla Milk Tea (coarse liquid) Chocolate Milk Tea (coarse liquid) E4 Homogenizer sterilization Vanilla Milk Tea Homogenizer Chocolate Milk Tea Homogenizer E5 sterilizer Filling Vanilla milk tea finished product Chocolate milk tea finished product E6 Filling Machine

[0247] Table 2

[0248] Furthermore, the types of materials in Task B and Task C are compared step by step:

[0249] Raw material pretreatment stage: The only difference between the input materials is vanilla flavoring and cocoa powder, but this stage is only responsible for material transportation and does not produce material mixing residue, so it is judged that the material types are the same;

[0250] Ingredient mixing process: The only difference between the input materials is the vanilla flavoring and the cocoa powder, but after the mixing tank is completed, all materials will be output to the mixer without any residue, which means that the materials are the same.

[0251] Mixing stage: The input materials are vanilla milk tea premix and chocolate milk tea premix, and the output materials are vanilla milk tea crude liquid and chocolate milk tea crude liquid, respectively. There is an essential difference between them, so it is marked as a differential production stage.

[0252] Homogenization, sterilization, and filling processes: The input materials are all intermediate products of the corresponding tasks, differing only in flavor components, but all are continuously processed through the same equipment. The inner wall of the equipment will retain materials from previous tasks, which is considered to indicate that the material types are consistent.

[0253] It should be noted that the above judgment logic is based on the process characteristics of milk tea raw material production. The residual flavor components will not affect the quality of subsequent products, and will not be elaborated here.

[0254] Based on the above process, the set of differentiated production links is identified as a mixed set, containing only one production link, which satisfies the condition of "containing only a portion of the production links".

[0255] Furthermore, for the mixing stage within the differentiated production process, the real-time parameters of mixer E3 are extracted: current residual vanilla milk tea base 25L, cleaning difficulty coefficient 1.2. Matching the primary cleaning SOP, the following localized cleaning strategy is generated:

[0256] Cleaning equipment: Mixer E3 only;

[0257] Cleaning process: Rinse with clean water → circulate detergent → rinse with clean water → blow dry;

[0258] Estimated cleaning time: 10 minutes;

[0259] Estimated cleaning supplies: 50L of clean water, 0.5kg of food-grade cleaning agent;

[0260] The remaining devices (E1, E2, E4, E5, E6) remain in standby mode and do not require cleaning.

[0261] This localized cleaning strategy was incorporated into the task B→task C switching strategy in the initial control plan, replacing the original full-line cleaning strategy and shortening the cleaning time of the localized cleaning strategy.

[0262] Simultaneously, after identifying the set of production processes with inconsistent material types, the estimated cost of downtime is calculated, including:

[0263] S41b1. Based on the cleaning time parameters of each production equipment in the raw material production line optimization model and the material residue parameters of each production equipment in the differential production process set, calculate the estimated value of the local downtime cost corresponding to the execution of the local cleaning strategy.

[0264] Specifically, based on the pre-stored cleaning time parameters of each production equipment in the raw material production line optimization model and the real-time material residue parameters of each equipment in the set of differential production links, the estimated value of local downtime cost is calculated through the equipment-level cleaning time accumulation model. The expression is as follows:

[0265]

[0266] In the formula, The estimated cost of local downtime for implementing a local cleaning strategy, in minutes;

[0267] This is the set of production equipment numbers corresponding to the set of differentiated production stages.

[0268] Preparation time for cleaning the k-th device, including device shutdown and safety isolation, in minutes;

[0269] The total amount of residual material in the k-th device, in L;

[0270] The cleaning time for a unit of residual amount of the kth equipment is expressed in min / L and is determined by both the equipment characteristics and the difficulty coefficient of cleaning the material.

[0271] The time for quality inspection confirmation after cleaning the kth piece of equipment, in minutes.

[0272] Based on the above-defined production tasks B and C, the equipment corresponding to the set of differential production links is mixer E3. Parameters are extracted from the raw material production line optimization model: min, =25L, =0.3 min / L, =0.5min, substitute into the above The expression is calculated as follows: =10min

[0273] S41b2. Calculate the estimated value of the total downtime cost corresponding to the implementation of the full-line cleaning strategy, and compare the estimated value of the partial downtime cost with the estimated value of the total downtime cost.

[0274] A full-line cleaning strategy refers to performing standard cleaning procedures on all production equipment on the production line. The estimated downtime cost is calculated based on the standard full-line cleaning time parameters, expressed as:

[0275]

[0276] In the formula, The estimated cost of total downtime for implementing the full-line cleaning strategy, in minutes;

[0277] The standard cleaning baseline time for the entire production line is determined by the production line design parameters, and is set to 30 minutes in this embodiment.

[0278] In this embodiment, the correction time for material residue is set to 5 minutes.

[0279] is the average normalization coefficient for the residual material amount of all equipment in the production line, with a value range of [0,1].

[0280] The normalization factor for the average material residue of all equipment in the current production line is set to [value missing]. =0.8, calculated as follows =34min.

[0281] S41b3 When the estimated cost of local downtime is less than or equal to the preset local cleaning time threshold, the local cleaning strategy is determined as the optimal switching cleaning strategy for the adjacent production tasks in that group.

[0282] In the specific implementation process, the calculated estimated value of local downtime cost will be used. With the preset local cleaning time threshold Perform a comparison and execute different decision-making processes based on the comparison results:

[0283] when ≤ When the time cost is optimal, the local cleaning strategy is determined to be the optimal switching cleaning strategy for the group of adjacent production tasks.

[0284] when > If the time cost of the local cleaning strategy is deemed too high, the local cleaning strategy is abandoned, the order of the pair of adjacent production tasks in the initial production task sorting combination is adjusted, and the initial control plan is regenerated.

[0285] Among them, the preset local cleaning time threshold Set as the baseline time for full-line cleaning 80% of × In this embodiment, The value is 24 minutes, which is the preset local cleaning time threshold. Adjustments can be made based on the actual operation of the production line; no restrictions are imposed here.

[0286] Regarding the above examples, =10min≤ =24min, the local cleaning strategy is determined to be the optimal strategy, and the local cleaning strategy of the mixer is determined to be the optimal switching cleaning strategy from task B to task C.

[0287] S41b4 When the estimated cost of local downtime exceeds the preset local cleaning time threshold, adjust the order of adjacent production tasks in the initial production task sorting combination and regenerate the initial control plan.

[0288] The determined optimal switching cleaning strategy is incorporated into the adjacent production task switching strategy of the corresponding initial production task sorting combination to form a complete initial control scheme, which then enters the pre-simulation verification stage. For the initial control scheme that adjusts the production task order due to excessive local cleaning time, the process of material consistency comparison and cleaning strategy generation for each stage needs to be re-executed.

[0289] Furthermore, the predicted production costs corresponding to each candidate control scheme are simulated, including:

[0290] S401. Extract the unit time operating cost, unit material loss cost, unit cleaning consumable cost, and unit time downtime loss cost of each production equipment from the raw material production line optimization model, and use them as the basic cost parameters for predicting production costs.

[0291] The unit time operating cost of each production equipment includes equipment energy consumption, equipment depreciation, and labor cost allocation, with the unit being yuan / min;

[0292] Unit material loss cost: includes the procurement cost of various raw materials and the cost of waste disposal, in yuan / L;

[0293] Unit cleaning consumables cost: includes the cost of clean water, food-grade cleaning agents, and disinfectant, expressed in yuan / L;

[0294] Downtime loss cost per unit time: includes capacity loss and order delay loss amortization caused by production line downtime, in yuan / minute;

[0295] S402. Each candidate control scheme is broken down into multiple continuous production task execution stages and adjacent production task switching stages. Each stage is then simulated segment by segment using the raw material production line optimization model.

[0296] The number of stages in the production task execution is equal to the total number of tasks to be produced, and each stage corresponds to the complete production process of one production task.

[0297] The number of adjacent production task switching stages is equal to the total number of tasks to be produced minus 1. Each stage corresponds to the switching process between two adjacent production tasks.

[0298] Here, it needs to be defined that, assuming there are four production tasks, namely production task A, production task B, production task C, and production task D, two sets of candidate control schemes are obtained for the above four production tasks, where:

[0299] Candidate Solution 1:

[0300] Production task A → Production task B → Production task C → Production task D can be broken down into four production execution stages and three switching stages.

[0301] The four production execution stages are: Production Task A execution, Production Task B execution, Production Task C execution, and Production Task D execution.

[0302] The three switching phases are: switching between production task A and production task B, switching between production task B and production task C, and switching between production task C and production task D.

[0303] Candidate Solution 2:

[0304] Production task A → Production task C → Production task B → Production task D can be broken down into four production execution stages and three switching stages.

[0305] The four production execution stages are: production task A execution, production task C execution, production task B execution, and production task D execution.

[0306] The three switching phases are: switching between production task A and production task C, switching between production task C and production task B, and switching between production task B and production task D.

[0307] S403. During the virtual simulation of the production task execution phase, the entire process of material conveying, batching and metering, mixing and processing and finished product output is simulated simultaneously, and the equipment operating cost, material loss cost and production cycle of the corresponding stage are calculated.

[0308] The raw material production line optimization model is used to perform segmented virtual simulation of each production task execution stage. The simulation time step is 1 minute, synchronously simulating the entire process of material transportation, batching and metering, mixing and processing, homogenization, sterilization, and filling. The three costs for the corresponding stage are calculated, including:

[0309] Equipment operating cost calculation: Equipment operating cost is the sum of the unit-time operating costs of all operating equipment during this period multiplied by the operating time. The expression is:

[0310]

[0311] In the formula, The unit for the equipment operating cost during the execution phase of this production task is: yuan; The running time of the k-th device in this stage, in minutes; This represents the unit time operating cost of the kth device, expressed in yuan / min.

[0312] Define production task A as having a production time of 60 minutes, with all equipment operating at full capacity, and substitute the values ​​into the calculation. =60×(0.5+1.2+1.8+2.0+2.5+3.0)=60×11.0=660 yuan.

[0313] Furthermore, the material loss cost is the sum of the loss amount of each type of material during this stage multiplied by the unit material loss cost. The material loss rate is set at 0.5% based on historical production line data, and the expression is as follows:

[0314]

[0315] In the formula, This represents the material loss cost during the execution phase of this production task, expressed in yuan. This represents the total consumption of the nth material, expressed in L. This represents the material loss rate, with a value of 0.005. This represents the unit material loss cost of the nth material, expressed in yuan / L.

[0316] Production task A is defined as consuming 200L of non-dairy creamer solution, 100L of white sugar solution, and 500L of black tea extract. Substituting these values ​​into the calculation, we get:

[0317] (200×8+100×5+500×12)×0.005=(1600+500+6000)×0.005=8100×0.005=40.5 yuan.

[0318] S404. In the virtual simulation process of switching between adjacent production tasks, based on the matching component compensation strategy and graded cleaning strategy in the candidate control scheme, the material component adjustment process and equipment cleaning operation process are simulated simultaneously, and the material adjustment cost, cleaning consumable cost and downtime loss cost of the corresponding stage are calculated.

[0319] Material adjustment cost calculation: Material adjustment cost is the sum of the cost of replenishing materials and the waste cost of remaining materials in the component compensation strategy, expressed as:

[0320]

[0321] In the formula, The material adjustment cost for this switching phase is in yuan. The cost of supplementary materials is expressed in yuan. The waste cost of remaining materials is expressed in yuan.

[0322] In the above example, during actual production, 5L of vanilla flavoring needs to be added, costing 5 × 200 = 1000 yuan; there is no waste of leftover materials, so the calculation is as follows: 000 yuan.

[0323] Cleaning consumables cost calculation: The cost of cleaning consumables is the sum of the amount of various cleaning consumables consumed during the cleaning process multiplied by the unit cost of the cleaning consumables. The expression is:

[0324]

[0325] In the formula, The cost of cleaning consumables during this switching phase is in yuan.

[0326] The consumption of the m-th type of cleaning consumable is expressed in liters (L).

[0327] The unit cost of the m-th type of cleaning consumable is expressed in yuan / L.

[0328] During the switchover phase between production task B and production task C, a localized cleaning strategy for the mixer is adopted:

[0329] The consumption is 50L of clean water and 0.5L of food-grade detergent. Substituting these into the above expression, we get... 50×0.01+0.5×20=0.5+10=10.5 yuan.

[0330] The calculation of downtime loss cost is as follows: downtime loss cost is the total downtime during the switchover phase multiplied by the downtime loss cost per unit time, expressed as:

[0331]

[0332] In the formula, The downtime cost during this switching phase is expressed in yuan. Total downtime during the switchover phase, in minutes; The cost of downtime is expressed as yuan per minute.

[0333] For example, if the total downtime is 10 minutes, substituting it into the above expression yields the following result:

[0334]

[0335] S405. Summarize the costs of all production task execution stages and adjacent production task switching stages to obtain the predicted production costs of the corresponding candidate control schemes. The predicted production costs include at least the total equipment operating costs, total material loss costs, total cleaning consumable costs, and total downtime loss costs.

[0336] The costs of all production task execution phases and adjacent production task switching phases are summarized to obtain the predicted production costs of the corresponding candidate control schemes. The predicted production costs include at least four core indicators: total equipment operating costs, total material loss costs, total cleaning consumable costs, and total downtime loss costs. A comparison table of the predicted production costs of candidate scheme 1 and candidate scheme 2 is shown in Table 3.

[0337] Total equipment operating cost 2640 yuan 2640 yuan Total material loss cost 162 yuan 162 yuan Total material adjustment cost 2800 yuan 3300 yuan Total cleaning consumable costs 31.5 yuan 63 yuan Total downtime loss cost 2000 yuan 3000 yuan Total projected production costs 7633.5 yuan 9165

[0338] Table 3

[0339] As can be seen from the comparison in Table 3, the total predicted production cost of candidate scheme 1 is significantly lower than that of candidate scheme 2. Therefore, candidate scheme 1 is selected as the target control scheme.

[0340] Furthermore, refer to Figure 4It also includes emergency order scheduling tasks, which include:

[0341] S51. Receive emergency production task instructions and obtain full-dimensional feature data of the production task that needs to be expedited. The full-dimensional feature data includes at least material feature dimension data, production changeover cost dimension data, and delivery time requirements.

[0342] It should be noted that the emergency production task instruction is received by the production line control system from the ERP system or issued manually. The system automatically parses the instruction and extracts full-dimensional feature data of the required expedited production task. All data conforms to the production line's unified data format standard.

[0343] S52. Extract the feature data of the currently executing production task, real-time production data, material residue parameters in each production equipment, and real-time material distribution data from the raw material production line optimization model.

[0344] The characteristic data of the currently executing production task includes task number, material type, component ratio, production time, remaining production time, and completed output. Among them, the real-time material distribution data is a complete real-time material distribution matrix D(t), which includes the material inventory data in all production equipment and pipelines.

[0345] S53. Calculate the comprehensive similarity index between the required expedited production task and the currently executed production task, and compare the comprehensive similarity index with the preset order insertion similarity threshold.

[0346] Specifically, the similarity indexes of material characteristics and production switching costs between expedited production tasks and currently executed production tasks are calculated separately. These are then weighted and fused to obtain a comprehensive similarity index, which will not be detailed here.

[0347] S54. When the comprehensive similarity index is greater than or equal to the first preset insertion similarity threshold, it is determined to be a highly adaptable insertion task. After the currently executing production task is completed, the emergency production task is started.

[0348] In this embodiment, for better illustration, two levels of insertion similarity thresholds are preset, wherein the first preset insertion similarity threshold... The second preset insertion similarity threshold The calculated comprehensive similarity index is compared with the two-level insertion similarity threshold, and the corresponding differentiated insertion strategy is executed based on the comparison result.

[0349] Among them, the first preset single-single similarity threshold Second preset single similarity threshold The determination is based on a comprehensive approach that considers both product quality and safety baselines and the cost-effectiveness of production switching. It is solidified through historical data statistical analysis and on-site verification, and the specific method for setting it will not be detailed here.

[0350] S55. When the comprehensive similarity index is less than the first preset insertion similarity threshold and greater than or equal to the second preset insertion similarity threshold, it is determined to be a medium-adaptation insertion task. The target component compensation amount is calculated based on the raw material production line optimization model, and the component compensation strategy is executed before the emergency production task is carried out.

[0351] Specifically, when considering the comprehensive similarity index When the task is identified as a highly adaptable insertion task, the material characteristics of the two production tasks are highly similar, the production switching cost is extremely low, and there is no need to clean the equipment or adjust a large number of materials.

[0352] It is important to note that emergency production tasks can only be initiated after the currently executing production task has been completed, in order to reduce the cost of switching production tasks.

[0353] S56. When the comprehensive similarity index is less than the second preset insertion similarity threshold, it is determined to be a low-adaptability insertion task. After the production task currently being executed is completed, each production equipment is cleaned.

[0354] Among them, when the comprehensive similarity index When the situation is determined to be a medium-adaptation interruption task, there is a certain material difference between the production task in progress and the interruption task. However, the difference can be eliminated through the component compensation strategy, and there is no need to perform full-line cleaning.

[0355] Component compensation amount calculation: Based on the real-time material distribution data in the raw material production line optimization model, the target material quantity that needs to be replenished is calculated, and the expression is:

[0356]

[0357] In the formula, The target compensation amount for the nth material; This represents the total demand for the nth type of material in an emergency mission. Let n be the residual amount of the nth material in the kth piece of equipment; Let be the availability of residual materials in the k-th equipment, with a value ranging from 0.8 to 1.0.

[0358] Therefore, based on the calculated target component compensation amount, the component compensation strategy is executed, and the emergency production task is directly started after adjusting the ingredient ratio.

[0359] Correspondingly, when the comprehensive similarity index When a task is identified as a low-fitness insertion task, the material characteristics of the two tasks are significantly different, and the risk of cross-contamination cannot be eliminated through component compensation strategies. Therefore, a full-line cleaning must be carried out. After the currently executing production task is completed, a standard full-line cleaning process is performed on all production equipment. After cleaning is completed, an emergency production task is initiated.

[0360] The following example will be used to further illustrate steps S51-S56:

[0361] For example, the production line receives an urgent production task instruction at 10:00. The urgent task is production task E (original milk tea base), and the full-dimensional feature data is as follows:

[0362] Material characteristics: 20% non-dairy creamer, 10% white sugar, 70% black tea extract, single batch yield 500L;

[0363] Switching costs: Cleanliness level requirement is Level 1, material adjustment difficulty coefficient is 1.0, and no equipment parameter adjustment is required;

[0364] Delivery time requirement: Production must be completed by 11:00 no later than 60 minutes.

[0365] Furthermore, the current production task being executed is Production Task B (Vanilla Milk Tea Base), and the real-time status data is as follows:

[0366] Task characteristics: Production has been underway for 40 minutes, with 20 minutes remaining to complete the task. Total output is 300L.

[0367] Equipment status: All equipment is operating normally, with a load rate of 85%;

[0368] Material residue: 25L of vanilla milk tea base remained in the mixer, 120L in the sterilizer, and 18.6L in the filling machine;

[0369] Real-time material distribution: obtained through the real-time material distribution matrix D(t).

[0370] Furthermore, the similarity of material characteristic dimensions is calculated: the proportion of common materials between Task E and Task B is R=0.75, and the component ratio deviation value is... The residual material matching degree M=1.0. Substituting this into the expression for the dimensional similarity index of the two production tasks in terms of material feature dimension, we obtain... =0.8429;

[0371] Calculate the similarity across the production changeover cost dimension, where the normalization coefficient for equipment cleaning costs is included. Material adjustment cost normalization coefficient =0.2, substitute into the dimension similarity index corresponding to the production switching cost dimension. get =0.868;

[0372] Furthermore, calculate the overall similarity: dimension weight coefficients. 0.4, obtained =0.6×0.8429+0.4×0.868=0.853.

[0373] Therefore, in the above scenarios =0.853 0.7 indicates a highly adaptable task. Task B is scheduled to complete at 10:20, and urgent task E will start at 10:20 and is expected to complete at 11:20, meeting the delivery schedule requirements.

[0374] If the urgent task is production task F, its overall similarity to the current task B is... =0.62, which is determined to be a medium-adaptive insertion task. It is calculated that 20kg of white sugar needs to be added. After executing the component compensation strategy, production is started without cleaning the equipment.

[0375] If the urgent production task is production task G, its overall similarity to the current production task B is... =0.42, judged as a low-adaptability insertion task. The current task B will complete at 10:20, and will perform a 30-minute full-line cleaning. Emergency task G will be launched at 10:50 and is expected to be completed at 11:50.

[0376] Furthermore, after the emergency order scheduling task is completed, it also includes:

[0377] S5a. Extract real-time status data of the production line after the completion of the emergency production task from the raw material production line optimization model. The real-time status data of the production line shall include at least the operating status parameters of each production equipment, material residual parameters, real-time material distribution data, and the actual production cycle and actual production cost of the emergency production task.

[0378] Equipment operating status parameters: start / stop status, operating parameters, equipment health, and remaining available time for each production device;

[0379] Material residue parameters: the precise residue amount and type of each material in each production unit;

[0380] Real-time material distribution data: The updated complete real-time material distribution matrix D(t) includes the material inventory in all equipment and pipelines;

[0381] Emergency mission actual operation data: the actual production cycle and actual production cost of emergency production missions, including equipment operating costs, material loss costs, switching costs, as well as actual output and pass rate.

[0382] Among them, the mechanism of capturing status snapshots after order insertion, accurately identifying affected tasks, dynamically reconstructing the similarity matrix, and optimizing and rearranging multiple objectives solves the technical problem in the existing technology that after order insertion is completed, production is carried out directly according to the original planned order, resulting in large-scale delays in subsequent tasks and a significant increase in overall production costs.

[0383] Once the emergency production task is completed and the last batch of finished products is output, immediately extract a snapshot of the complete production line status at the current moment from the raw material production line optimization model. All data is synchronously collected at the instant the task is completed.

[0384] Equipment operating status parameters: start / stop status, operating parameters, equipment health, and remaining available time for each production device;

[0385] Material residue parameters: the precise residue amount and type of each material in each production unit;

[0386] Real-time material distribution data: The updated complete real-time material distribution matrix D(t) includes the material inventory in all equipment and pipelines;

[0387] Emergency mission actual operation data: the actual production cycle and actual production cost of emergency production missions, including equipment operating costs, material loss costs, switching costs, as well as actual output and pass rate.

[0388] S5b. Determine the set of subsequent production tasks affected by the emergency production task in the original target control plan. The set of subsequent production tasks is all the unfinished production tasks that were originally planned to be executed after the time point when the emergency production task was inserted.

[0389] Extract the production schedule from the original target control plan, and define all unfinished production tasks originally scheduled to be executed after the insertion point of emergency production tasks as the set of subsequent pending production tasks. It also records the estimated start time and latest completion time of each subsequent task in the original plan, and calculates the initial delay time caused by the insertion of orders.

[0390] S5c: Based on the full-dimensional feature data of the emergency production task and the extracted real-time status data of the production line, recalculate the comprehensive similarity index between each production task in the subsequent production task set and the emergency production task, as well as the comprehensive similarity index between any two production tasks in the subsequent production task set.

[0391] Specifically, based on the full-dimensional feature data of the emergency production task and the real-time status data of the production line extracted in step S5a, the data are analyzed according to the dimensional similarity index expression. Recalculate the two sets of similarity indices.

[0392] Among them, the comprehensive similarity index between each task in the subsequent set of tasks to be produced and the emergency task is based on the residual state of the production line after the current production task is completed.

[0393] Furthermore, the comprehensive similarity index between any two tasks within the subsequent set of tasks to be produced will be calculated and integrated into a new comprehensive similarity matrix, which will serve as the benchmark for subsequent target control schemes.

[0394] S5d takes maximizing the comprehensive similarity index between adjacent production tasks as the core objective, and uses the real-time status of the production line after the execution of emergency production tasks and the delivery time requirements of each subsequent production task as constraints to generate multiple updated candidate control schemes.

[0395] Specifically, an improved genetic algorithm is used to generate multiple sets of updated initial production task sorting combinations:

[0396] Hard constraint 1: Real-time status of the production line after the emergency task is executed;

[0397] Hard constraint 2: The latest completion time of each subsequent task to be produced;

[0398] Hard constraint 3: Inventory balance and expected delivery time of each raw material.

[0399] For each initial sorted combination, a corresponding switching strategy is matched, and infeasible schemes are eliminated through pre-simulation verification, ultimately generating multiple updated candidate control schemes.

[0400] S5e: Simulate the predicted production cost and total order delay time corresponding to each updated candidate control scheme through the raw material production line optimization model. Based on the weighted comprehensive value of the predicted production cost and total order delay time, select the optimal updated target control scheme from multiple updated candidate control schemes.

[0401] Among them, by simulating each updated candidate control scheme using a raw material production line optimization model, two core evaluation indicators were calculated:

[0402] The production cost is predicted and obtained from steps S401-S405, including the equipment operating cost, material loss cost, and changeover cost of all subsequent tasks.

[0403] Total order delay time: The sum of the differences between the actual completion time of all subsequent tasks and the originally planned latest completion time.

[0404] Set a multi-objective weighted comprehensive evaluation function:

[0405]

[0406] In the formula, To comprehensively evaluate the score, the smaller the value, the better the solution; This is the production cost weight, with an initial value of 0.6; Order delay weight, initial value 0.4, and + =1; Predicted production costs for candidate solutions; The average predicted production cost for all candidate solutions; The total order delay time for the candidate solutions; The average total order delay time for all candidate solutions.

[0407] S5f: Execute the updated target control plan to produce the production tasks in the subsequent set of tasks to be produced.

[0408] Specifically, the updated target control plan is sent to the production line control system, and production is organized according to the new production task order. During the production process, real-time data is continuously collected through IoT sensors, the raw material production line optimization model is updated synchronously, and the dimension weight coefficients are adaptively adjusted.

[0409] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for intelligent control of a milk tea ingredient production line based on the Internet of Things and digital twins, characterized in that, include: By collecting production data in real time through IoT sensors deployed on the milk tea raw material production line, an optimization model for the raw material production line reflecting the current production status of the milk tea raw material production line is constructed based on the real-time production data. Acquire feature data for each production task in multiple production tasks. The feature data includes at least the material feature dimension and the production changeover cost dimension. For any two production tasks, determine the dimension similarity index corresponding to each dimension based on the feature data of the two production tasks in each dimension. Based on the similarity indicators of each dimension and the corresponding dimension weight coefficients, a comprehensive similarity indicator between two production tasks is generated by weighted fusion. The dimension weight coefficients are used to characterize the degree of influence of the corresponding dimension on the comprehensive similarity of the production tasks. By using the raw material production line optimization model, multiple candidate control schemes are generated based on the comprehensive similarity index between each production task. The predicted production costs corresponding to each candidate control scheme are simulated, and the target control scheme is selected from the multiple candidate control schemes based on the predicted production costs. The target control plan is implemented, and the actual production effect data after implementation is collected. Based on the deviation between the actual production effect data and the predicted production cost, the dimensional weight coefficients corresponding to the similarity indicators of each dimension are adjusted.

2. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 1, characterized in that, A raw material production line optimization model is constructed based on real-time production data to reflect the current production status of the milk tea raw material production line, including: Obtain an initial digital twin model of the milk tea raw material production line. The initial digital twin model includes models of each production equipment in the milk tea raw material production line and the material flow paths between each production equipment. Acquire real-time production data collected by IoT sensors. Real-time production data includes equipment operating status data and material handling data of each production device. For each production equipment, the equipment operation status data and material handling data are mapped to the corresponding equipment model to update the operation status parameters, material handling status parameters and material residue parameters. The material handling status parameters include the type of material currently being processed by the corresponding production equipment and the processing progress. The material residue parameters include the type of material remaining in the corresponding production equipment and the residue value. Based on the updated equipment models and material flow paths, the real-time distribution of materials in the milk tea raw material production line is determined, and a raw material production line optimization model reflecting the current overall production status of the milk tea raw material production line is generated.

3. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 1, characterized in that, Based on the feature data of the two production tasks in each dimension, the dimensional similarity indexes corresponding to each dimension are determined, including: Based on the raw material production line optimization model, the operating status parameters, material handling status parameters, material residue parameters, and real-time material distribution data in the production line of each production equipment model are extracted as real-time benchmark data for calculating the dimensional similarity index. Based on the material characteristics dimension, extract the material type and component ratio characteristic data corresponding to the two production tasks. Combine the real-time material distribution data, the material types in each production equipment and the material residue parameters to calculate the common material ratio, component ratio deviation value and matching degree with the residual material in the current production line for the two production tasks respectively. The weights of common material proportion, component ratio deviation and residual material matching degree are set respectively, and the dimensional similarity index corresponding to the material feature dimension is obtained by weighted summation. For the production switchover cost dimension, we extract the characteristic data of equipment cleaning requirements and material adjustment requirements required for the production switchover of two production tasks. Combined with real-time material distribution data, material types in each production equipment and material residue parameters, we calculate the estimated values ​​of equipment cleaning cost, material adjustment cost and downtime cost corresponding to the switchover of two production tasks. For equipment cleaning costs, material adjustment costs, and downtime costs, preset benchmark cost thresholds are set. The estimated values ​​of each item are normalized with the benchmark cost thresholds to obtain the normalization coefficients corresponding to each cost. Based on the normalization coefficients, the dimensional similarity index of the two production tasks in the production switching cost dimension is calculated.

4. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 3, characterized in that, Based on the normalized coefficient, the dimensional similarity index of the two production tasks in terms of production switchover cost is calculated, including: For any two production tasks to be produced, denoted as production task i and production task j respectively, the expression for the dimensional similarity index of the two production tasks in terms of material characteristics is as follows: In the formula, The dimensional similarity index between production task i and production task j in terms of material characteristics. , , These are the sub-weights corresponding to the proportion of common materials, the component ratio deviation value, and the residual material matching degree, respectively. + + =1; The percentage of materials shared by production task i and production task j is the ratio of the number of shared material types between the two production tasks to the total number of material types in the two production tasks. Let be the component ratio deviation between production task i and production task j, and let be the normalized mean of the relative deviation of the component ratio of the common materials in the two production tasks. The matching degree between production task i and production task j and the residual materials in the current production line; The similarity index between production task i and production task j in terms of production switchover cost is expressed as follows: In the formula, This is the dimensional similarity index between production task i and production task j in the production switching cost dimension. , , These are the weighting coefficients corresponding to equipment cleaning costs, material adjustment costs, and downtime costs, respectively. + + =1; The normalization coefficient for the equipment cleaning cost when switching between production task i and production task j; The normalization coefficient for the material adjustment cost when switching between production task i and production task j; This is the normalization coefficient for the downtime cost when switching between production task i and production task j.

5. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 1, characterized in that, Based on the comprehensive similarity index between various production tasks, a raw material production line optimization model is used to generate multiple candidate control schemes, including: The current equipment operating status parameters, equipment capacity limits, material supply sequence, and delivery time requirements of each production task are extracted from the raw material production line optimization model and used as the basic constraints for generating candidate control schemes. With maximizing the comprehensive similarity index between adjacent production tasks as the core ranking objective and using basic constraints as boundaries, all production tasks to be produced are combined and ranked in multiple rounds to generate multiple initial production task ranking combinations. For each initial production task sorting combination, based on the similarity index of material characteristics and the similarity index of production switching cost of adjacent production tasks, the component compensation strategy and graded cleaning strategy are matched when switching between adjacent production tasks to form an initial control plan. The initial control schemes of each group were pre-simulated and verified using the raw material production line optimization model. Infeasible schemes that did not meet the requirements of equipment capacity, material supply and delivery time were eliminated, and the initial control schemes that passed the verification were determined as candidate control schemes.

6. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 1, characterized in that, Simulate the predicted production costs corresponding to each candidate control scheme, including: The unit time operating cost, unit material loss cost, unit cleaning consumable cost, and unit time downtime loss cost of each production equipment are extracted from the raw material production line optimization model and used as the basic cost parameters for predicting production costs. Each candidate control scheme is broken down into multiple continuous production task execution stages and adjacent production task switching stages. Each stage is then virtually simulated using a raw material production line optimization model. During the virtual simulation of the production task execution phase, the entire process of material transportation, batching and metering, mixing and processing and finished product output is simulated simultaneously, and the equipment operating cost, material loss cost and production cycle of the corresponding stage are calculated. During the virtual simulation of the transition between adjacent production tasks, based on the matching component compensation strategy and graded cleaning strategy in the candidate control scheme, the material component adjustment process and equipment cleaning operation process are simulated simultaneously, and the material adjustment cost, cleaning consumable cost and downtime loss cost of the corresponding stage are calculated. The costs of all production task execution phases and adjacent production task switching phases are summarized to obtain the predicted production costs of the corresponding candidate control schemes. The predicted production costs include at least the total equipment operating costs, total material loss costs, total cleaning consumable costs, and total downtime loss costs.

7. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 1, characterized in that, This also includes emergency order scheduling tasks, which include: Receive urgent production task instructions and obtain full-dimensional feature data of the production task that needs to be expedited. The full-dimensional feature data includes at least material feature dimension data, production changeover cost dimension data, and delivery time requirements. Extract the feature data of the currently executing production task, real-time production data, material residue parameters in each production equipment, and real-time material distribution data from the raw material production line optimization model; Calculate the comprehensive similarity index between the required expedited production tasks and the currently executed production tasks, and compare the comprehensive similarity index with the preset order insertion similarity threshold; When the comprehensive similarity index is greater than or equal to the first preset insertion similarity threshold, it is determined to be a highly adaptable insertion task. After the currently executing production task is completed, the emergency production task is started. When the comprehensive similarity index is less than the first preset insertion similarity threshold and greater than or equal to the second preset insertion similarity threshold, it is determined to be a medium-adaptation insertion task. The target component compensation amount is calculated based on the raw material production line optimization model, and the component compensation strategy is executed before the emergency production task is carried out. When the overall similarity index is less than the second preset insertion similarity threshold, it is judged as a low-adaptability insertion task. After the production task currently being executed is completed, each production equipment is cleaned.

8. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 7, characterized in that, After the emergency order scheduling task is completed, it also includes: Extract real-time status data of the production line after the completion of the emergency production task from the raw material production line optimization model. The real-time status data of the production line includes at least the operating status parameters of each production equipment, material residual parameters, real-time material distribution data, and the actual production cycle and actual production cost of the emergency production task. Determine the set of subsequent production tasks affected by the emergency production task in the original target control plan. The set of subsequent production tasks is all the unfinished production tasks originally planned to be executed after the time point of the emergency production task insertion. Based on the full-dimensional feature data of the emergency production task and the extracted real-time status data of the production line, the comprehensive similarity index between each production task in the subsequent production task set and the emergency production task, as well as the comprehensive similarity index between any two production tasks in the subsequent production task set, are recalculated. With the core objective of maximizing the comprehensive similarity index between adjacent production tasks, and with the real-time status of the production line after the execution of emergency production tasks and the delivery date requirements of each subsequent production task as constraints, multiple updated candidate control schemes are generated. The predicted production cost and total order delay time corresponding to each updated candidate control scheme are simulated by the raw material production line optimization model. Based on the weighted comprehensive value of the predicted production cost and total order delay time, the optimal updated target control scheme is selected from multiple updated candidate control schemes. The updated target control plan is implemented to produce the tasks in the subsequent set of tasks awaiting production.

9. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 5, characterized in that, Based on the comprehensive similarity index between various production tasks, a raw material production line optimization model is used to generate multiple candidate control schemes, including: For any pair of adjacent production tasks in each initial production task sorting combination, extract the production process sequence and material type data of each production process corresponding to the two production tasks from the raw material production line optimization model. By comparing the consistency of material types in the corresponding production stages of two production tasks, the set of production stages with inconsistent material types is identified. When the set of differential production links is not empty and contains only some production links, a local cleaning strategy is generated for each production equipment in the set of differential production links. The local cleaning strategy only performs cleaning operations on the production equipment corresponding to the production links with inconsistent material types.

10. The intelligent control method for a milk tea raw material production line based on the Internet of Things and digital twins according to claim 9, characterized in that, After identifying the set of production processes with inconsistent material types, the estimated cost of downtime is calculated, including: Based on the cleaning time parameters of each production equipment in the raw material production line optimization model and the material residue parameters of each production equipment in the differential production process set, the estimated value of the local downtime cost corresponding to the execution of the local cleaning strategy is calculated. Calculate the estimated downtime cost of implementing the full-line cleaning strategy, and compare the estimated downtime cost of the local downtime with the estimated downtime cost of the full line. When the estimated cost of local downtime is less than or equal to the preset local cleaning time threshold, the local cleaning strategy will be determined as the optimal switching cleaning strategy for the group of adjacent production tasks. When the estimated cost of local downtime exceeds the preset local cleaning time threshold, the order of adjacent production tasks in the initial production task sorting combination is adjusted, and the initial control plan is regenerated.

Images