An intelligent production scheduling method for cigarette manufacturing based on industrial big data

Abstract

The present invention relates to the research field of intelligent production scheduling of cigarettes, and discloses an intelligent production scheduling method for cigarette manufacturing based on industrial big data, including the steps of collecting and integrating historical industrial big data, mining the similarity between different cigarette formulas, and outputting historical example information; initializing various parameters of the genetic algorithm; encoding the production sequence of cigarettes to generate an initial solution set to initialize the population of the genetic algorithm; decoding the initial solution set using a continuous decoding method or a discrete decoding method; evolving and updating the population using a genetic algorithm, and outputting the current optimal solution. The present invention is oriented to intelligent production scheduling and optimization of cigarette production and manufacturing, and integrates intelligent optimization algorithms, big data analysis technology, and real-time data processing technology. It uses intelligent optimization algorithms to calculate the best production scheduling plan for each production batch in real time, improves the efficiency and accuracy of the production plan, and ensures high optimization of the production process and rational use of resources.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent production scheduling of cigarettes, and in particular to an intelligent production scheduling method for cigarette manufacturing based on industrial big data. Background Art

[0002] Cigarette manufacturing is a complex industrial production process that includes many complex links such as raw material processing, silk making, and rolling. Traditionally, it still uses production scheduling based on manual experience or simple mathematical models, which leads to problems such as low efficiency. With the advancement of cloud computing, the Internet of Things, and mobile technology, the tobacco industry has accumulated rich data in actual production history, providing a good foundation for the intelligent transformation of cigarette manufacturing. However, at this stage, the production scheduling process of cigarette manufacturing still lacks a refined development model, and the development of big data is out of line with the production field. Summary of the invention

[0003] The purpose of the present invention is to improve the problem that the existing scheduling system has limited intelligence level, usually relies on manual experience to make decisions, lacks the support of big data analysis and artificial intelligence technology, and is difficult to achieve intelligent scheduling, and to provide a cigarette manufacturing intelligent production scheduling method based on industrial big data.

[0004] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0005] An intelligent production scheduling method for cigarette manufacturing based on industrial big data comprises the following steps:

[0006] Step 1: Collect and integrate historical industrial big data, mine the similarities between different cigarette formulas, and output historical example information;

[0007] Step 2, initializing various parameters of the genetic algorithm;

[0008] Step 3, encode the production sequence of cigarettes and generate an initial solution set to initialize the population of the genetic algorithm;

[0009] Step 4, decoding the initial solution set using a continuous decoding method or a discrete decoding method;

[0010] Step 5: Use genetic algorithm to evolve and update the population and output the current optimal solution.

[0011] Compared with the prior art, the present invention has the following beneficial effects:

[0012] The present invention establishes an integer programming model for multi-stage mixed production scheduling of cigarettes through big data analysis, intelligent mining and integrated modeling of cigarette production process data, and designs an improved genetic algorithm to solve this problem. In this model, the three parts of cigarette storage quality, overdue delay penalty and brand replacement cost are weighted based on the differentiated cigarette grade to form the objective function. In the improved genetic algorithm, the customized encoding and decoding strategies reduce the complexity while fully retaining the essential information of the problem. The reasonable design of genetic operation operators ensures the diversity and high convergence of the algorithm, and the introduction of the penalty mechanism avoids the emergence of infeasible solutions.

[0013] The present invention is mainly aimed at intelligent production scheduling and optimization of cigarette production and manufacturing. It integrates intelligent optimization algorithms, big data analysis technology and real-time data processing technology. It uses intelligent optimization algorithms to calculate the best production scheduling plan for each production batch in real time. The design of the algorithm comprehensively considers the multi-stage and replacement characteristics of cigarette production, improves the efficiency and accuracy of production plans, and ensures high optimization of the production process and rational use of resources.

[0014] The experimental results show that the improved genetic algorithm proposed in the present invention has a very significant optimization effect compared with the traditional MSA and manual production scheduling strategies, and is superior to other methods in multiple performance indicators, with high adaptability and robustness. The present invention provides an effective solution to the problem of intelligent production scheduling in cigarette manufacturing, and also provides new ideas for solving similar optimization problems, providing technical support for enterprises to further realize lean production and agile manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 Improved genetic algorithm flow chart for the method of the present invention;

[0017] Figure 2 This is the Gantt chart of the production scheduling results of the method of Example 2;

[0018] Figure 3 This is the Gantt chart of the production scheduling results of the traditional MSA algorithm in Example 2;

[0019] Figure 4 This is the Gantt chart of the manual production scheduling results of Example 2;

[0020] Figure 5 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and represented in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0022] It should be noted that similar numbers and letters represent similar items in the following figures, so once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance, or implying any such actual relationship or order between these entities or operations. In addition, the terms "connected", "connected", etc. can be directly connected between elements, or indirectly connected via other elements.

[0023] Embodiment 1:

[0024] The present invention is achieved through the following technical solutions: Figure 1 , Figure 5 As shown, a cigarette manufacturing intelligent production scheduling method based on industrial big data includes the following steps:

[0025] Step 1: Collect and integrate historical industrial big data, mine the similarities between different cigarette formulas, and output historical example information.

[0026] Collect industrial big data from the historical production process of cigarettes, including cigarette formula, tobacco shredder time, brand change time, processing time, etc.; cigarette formula refers to the different leaf group formulas used in different brands of cigarettes (such as Yunyan, Kuanzhai, Huanghelou, etc.), that is, they are made up of different proportions of tobacco leaves; brand change time refers to the time required for a machine to switch to manufacturing cigarette brand B after completing the manufacturing of cigarette brand A.

[0027] When integrating industrial big data, we divide it into batches, organize the number of each cigarette brand, the number of parallel machines producing this cigarette brand, processing time and other information by batch. And convert this information into data in a standardized format, such as calculating the formula similarity of cigarettes based on cigarette formula information to explore the degree of association between cigarette brands, thereby providing usable example information.

[0028] The case information focused on in this solution includes the following key objectives: maximizing storage quality TQ, minimizing overdue delay penalty or maximizing early completion reward TD, and minimizing brand change cost TC. Based on the different grades of cigarettes, a differentiated weight allocation strategy is adopted to distinguish the relative priorities of different sub-goals of each grade of cigarettes in the optimization process, using g i Indicates the grade of the i-th category cigarette, They represent the weight coefficients of the three optimization sub-goals TQ, TD, and TC of the i-th category of cigarettes, respectively. i represents the grade of the i-th category of cigarettes. The objective function Max z is as follows:

[0029] Max z=TQ+TD-TC

[0030] (1) Maximize storage quality TQ

[0031] The storage quality of cigarettes is mainly affected by the storage time of tobacco. ik-1 ,r ik-1 ], the storage quality of tobacco can remain stable for a period of time, including l ik-1 、r ik-1 They represent the shortest and longest storage time after process k-1 respectively. The tobacco needs to be stored for a corresponding time in this window before the next processing. As the storage time goes by, the quality of cigarettes shows an accelerated downward trend, showing an exponential decay characteristic. In order to characterize this quality change, a piecewise function of cigarette storage quality decay is constructed:

[0032]

[0033] Wherein, η(t) represents the storage quality corresponding to the storage time t; η0 represents the initial quality of tobacco, which is fixed at 100; θ is the mass attenuation coefficient; [l ik ,r ik ] represents the storage time window of the i-th type of cigarette after the k-th process, l ik Indicates the shortest storage time, r ik Indicates the longest storage time.

[0034] Based on the above-mentioned piecewise function of tobacco storage quality attenuation, the maximum storage quality TQ of each grade of cigarettes can be expressed as:

[0035]

[0036] Where t = a ik -b ik-1represents the storage time, that is, the time from the end of the previous process to the start of the next process; S1 represents the set of discrete production processes, S1 = {1, 2, ..., p}; S2 represents the set of continuous production processes, S2 = {p, p+1, ..., k}; M k Represents the set of machines for the kth process, M k ={1,2,...,h}; I represents the set of cigarette categories, and I = {1,2,...,i}; a ik represents the start time of processing of the i-th type of cigarettes in the k-th process; b ik-1 It indicates the completion time of the processing of the i-th category cigarettes in the k-1th process.

[0037] (2) Minimize the penalty for late completion or maximize the reward for early completion TD

[0038] The rewards for early completion or penalties for late completion of each grade of cigarettes are shown in the following formula:

[0039]

[0040] Among them, τ i represents the latest completion date of the i-th category of cigarettes, c i represents the actual completion time of the i-th type of cigarettes; β gi Indicates the weight coefficient for minimizing the penalty for overdue delay or maximizing the reward for early completion TD. TD can be positive or negative. When TD is negative, it indicates the penalty for overdue delay, and the penalty needs to be minimized; when TD is positive, it indicates early completion, and the reward needs to be maximized.

[0041] (3) Minimize the card change cost TC

[0042] The more similar the formulas of two types of finished cigarettes are, the less the cost and time of changing brands are, and the lower the complexity of machine switching and adjustment in the production process is. In order to characterize the similarity of cigarettes, the cosine similarity function is used to measure the similarity between two types of finished cigarettes. Suppose the formulas of the i-th and j-th cigarettes are i=(v1,v2,...,v n ) and j=(u1,u2,...,u n ), then the cosine similarity function sim(i,j) is as follows:

[0043]

[0044] Among them, v m represents the mth formula in the i-th category of cigarettes, u m Represents the mth formula in the jth category of cigarettes, m = 1, 2, ..., n.

[0045] Based on the actual production scenario, it is assumed that the switching cost required to produce two types of cigarettes with completely different formulas on machine h in the kth process is C h , the switching time is T h , then the brand switching cost matrix e for machine h switching from producing the i-th type of cigarettes to the j-th type of cigarettes ijk and the card change time matrix w ijk It can be expressed as:

[0046]

[0047] The minimum brand change cost TC in the total cigarette production process is expressed as:

[0048]

[0049] Where S represents the set of production processes, S = {1, 2, ..., k}; z ijhk Indicates whether the i-th category cigarettes take precedence over the j-th category cigarettes in the k-th process on machine h. If so, it is 1, otherwise it is 0.

[0050] Then the objective function is:

[0051]

[0052] Step 2: Initialize various parameters of the genetic algorithm.

[0053] Initialize various parameters of the genetic algorithm, including population size P, penalty coefficient λ, machine status Machines, job status Jobs, set the maximum number of iterations epoch_num, crossover probability Pc, mutation probability Pm, etc.

[0054] Step 3, encode the production sequence of cigarettes and generate an initial solution set to initialize the population of the genetic algorithm.

[0055] It is difficult for genetic algorithms to directly process the decision variables in the optimization problem space. These decision variables need to be converted into chromosome sequences composed of genes in a specific structural form, that is, encoding operations. This scheme uses natural number encoding to represent the chromosome sequence structure. Each number in the digital sequence {1,2,...,i} represents a specific type of cigarette, and the order of encoding represents the production order of cigarettes. For example, Figure 2 As shown, it is assumed that there are 3 processes (i.e. Stege1, Stege2, Stege3), each process has 11 machines, and each segment with a number represents a type of cigarette (i.e. cigarette brand). Figure 2 31 types of cigarettes are shown in the figure, and the horizontal axis represents the processing time.

[0056] The natural number encoding method has a concise and intuitive expression, which enables the algorithm to flexibly adapt to changing problem structures and complex constraints. In addition, the intuitive digital sequence not only facilitates the implementation of genetic operations, but also reduces the computational complexity of the operation process, thereby accelerating the convergence speed of the algorithm and improving the algorithm performance.

[0057] Step 4: Decode the initial solution set using a continuous decoding method or a discrete decoding method.

[0058] This solution combines the ideas of First Come First Service (FCFS) and Earlist Completed First (ECF), comprehensively considers the "earliest completion time" and "shortest processing time" of the machine, decodes in sequence according to the coding sequence, and dynamically allocates the optimal machine for each type of cigarette to ensure the efficiency and practicality of production scheduling. Taking into account the waiting time caused by the wire storage process, different machine allocation strategies are designed through different decoding methods for the continuous (no wire storage waiting time) and discrete (with wire storage waiting time) situations in the upstream and downstream stages to adapt to the complexity and diversity of cigarette production.

[0059] (I) Stage-by-stage decoding process

[0060] In the continuous production stage such as rolling to packaging, based on the FCFS concept, the type of cigarettes that are completed first in the upstream process is given priority and the machine of the next process is assigned to it, so as to ensure the smooth and efficient production process. In the initial stage, all types of cigarettes have the same production priority, and a type of cigarette is randomly selected for the first process. The specific decoding process is as follows:

[0061] Input: continuous production process set S2, card change time matrix w ijk , processing time matrix T hik , the ascending list of the completion time of the previous process k-1 CHS k-1 ;

[0062] Step P1: Traversing CHS k-1 , for all types of cigarettes i∈CHS k-1 , do the following:

[0063] 1) Calculate the completion time ls for each machine h in the current process k hk , machine processing time L hk , switch to the brand change time w of processing cigarette i hik , the processing time t of processing cigarette i hik ;

[0064] 2) Machine allocation based on ECF strategy:

[0065] a. Calculate the completion time b of each machine h processing cigarette i in the kth process at the current moment hik =ls hk +w hik +t hik ;

[0066] b.if count(min(b hik ))=1, select min(b hik ) corresponds to the earliest completed machine;

[0067] c.if count(min(b hik ))>1, select min(b hik ) corresponds to the earliest completed machine, min(L hk ) corresponds to the machine with the shortest processing time;

[0068] When neither db nor c is satisfied, a machine is randomly selected for processing;

[0069] 3) Update: The processing status and completion time ls of the kth process of cigarette i ik .

[0070] Step P2: Let k=k+1, if k∈S2, go to step P1, otherwise output the algorithm result.

[0071] Output: ascending list CHS of the completion time of each cigarette in process k k , processing time of each machine L hk , completion time ls hk .

[0072] For ease of understanding, as an example, see Figure 2 , when machine 1 (h = 1) in process Stage 2 (k = 2) processes cigarette 1, cigarette 8, cigarette 6 and cigarette 10 in sequence, the completion time ls hk is the time when cigarette 1 processing is completed, and the brand change time w hik The processing time t is the time it takes to convert the processed cigarette 1 into the processed cigarette 8 after the processing is completed. hik The time it takes to process cigarette 8, so the completion time b hik is the time when cigarette 8 processing is completed; the machine processing time L hk The time it takes for machine 1 to process cigarette 1, cigarette 8, cigarette 6, and cigarette 10. Then go back and look at the ascending list of completion time CHS for process Stage 1 k-1 In the middle, it is based on the completion time of each type of cigarette. hik Sort the table in ascending order. "if count(min(bhik ))=1" means that if the shortest completion time b appears in a process hik If there is a machine with a certain number of items, the machine will be selected to enter the next process to reasonably complete the production scheduling. hik ))>1" means that if there are multiple processes with the same shortest completion time b hik The machine with the shortest processing time min(L hk ) enters the next process.

[0073] (II) Stage Discrete Decoding Process

[0074] Taking into account the waiting time for shredded cigarettes, in discrete production stages such as shredded cigarettes to rolling and splicing, based on the FCFS concept, unlike the continuous type of upstream and downstream processes, the cigarette type that meets the shortest shredded cigarette storage time is given priority and the next process machine is assigned to it for processing. The specific decoding process is as follows:

[0075] Input: Discrete production process set S1, card change time matrix w ijk , processing time matrix T hik , process k-1 completion time ascending list CHS k-1 ;

[0076] Step S1: Traversing CHS k-1 , for all types of cigarettes i∈CHS k-1 , based on the silk storage time window [l ik-1 ,r ik-1 ]Calculate the available storage time of cigarette i [ls ik-1 +l ik-1 ,ls ik-1 +r ik-1 ], the available time of storage silk is arranged in ascending order to obtain the storage available ascending order list CHL k-1 ;

[0077] Step S2: Traversing CHL k-1 , for all types of cigarettes i∈CHL k-1 , do the following:

[0078] 1) Calculate the processing time L for each machine h in the current process k: hk , switch to the brand change time w of processing cigarette i hik , the processing time t of processing cigarette i hik ;

[0079] 2) Filter the available machine set Mh among the candidate machines to satisfy ls ik-1 +lik-1 ≤ls hk +w hik ≤ls ik-1 +r ik-1 ,ls hk is the completion time;

[0080] 3) If Then go to step 4), otherwise go to step 5);

[0081] 4) No machine meets the storage time window constraint, and cigarette i is assigned to min(ls hk +w hik -(ls ik-1 +r ik-1 )) corresponding to the machine, so that the degree of violation of the wire storage time window is minimized;

[0082] 5) Machine allocation based on ECF strategy:

[0083] a. Calculate the completion time b of each machine h processing cigarette i in the kth process at the current moment hik =ls hk +w hik +t hik ;

[0084] b.if count(min(b hik ))=1, select min(b hik ) corresponds to the earliest completed machine;

[0085] c.if count(min(b hik ))>1, select min(b hik ) corresponds to the earliest completed machine, min(L hk ) corresponds to the machine with the shortest processing time;

[0086] When neither db nor c is satisfied, a machine is randomly selected for processing;

[0087] 6) Update: The processing status and completion time ls of the kth process of cigarette i ik , processing time of each machine L hk , completion time ls hk ;

[0088] Step S3: Let k=k+1, if k∈S1, go to step S1, otherwise output the algorithm result.

[0089] Output: ascending list CHS of the completion time of each cigarette in process k k , processing time of each machine L hk , completion time ls hk .

[0090] For ease of understanding, as an example, assume that the storage time window of cigarette i is [l ik-1 ,r ik-1 ] is [1,2], indicating that its storage time window is 1 to 2 hours; the available storage time of cigarette i is [ls ik-1 +l ik-1 ,ls ik-1 +r ik-1 ls in ] ik-1 Indicates the time it takes to produce cigarette i, such as ls ik-1 is 9:00 a.m., then [ls ik-1 +l ik-1 ,ls ik-1 +r ik-1 ] is [10,11], which means the storage time is from 10 to 11 o'clock. Figure 2 In the process Stage 2, machine 1 processes cigarette 1, then ls ik-1 +l ik-1 ≤ls hk +w hik ≤ls ik-1 +r ik-1 If the completion time of processing cigarette 1 is ls ik-1 Plus the time w for converting into processed cigarettes 8 hik Within the available wire storage time, the machine belongs to the available machine set Mh.

[0091] Step 5: Use genetic algorithm to evolve and update the population and output the current optimal solution.

[0092] The step 5 specifically comprises the following steps:

[0093] Step 5-1, evaluate the fitness of each feasible solution in the initial solution set.

[0094] Based on the multi-stage and multi-constraint characteristics of the Multi-stage Permutation Flow Shop Scheduling Problem With Storage Delay (MPFFSPSD), the specific storage time window and resource limitation constraints will lead to a large number of infeasible solutions in the initial solution set. Therefore, a penalty number mechanism is introduced on the basis of the original objective function Max z to impose a penalty coefficient λ on the infeasible solution, and the fitness function f is:

[0095]

[0096] Step 5-2, use the roulette wheel method for parent selection.

[0097] Step 5-3: Perform chromosome exchange and gene mutation operations on the selected parental solutions, and record the iteration count counts.

[0098] Step 5-4: If counts < epoch_num, go to Step 5-1; otherwise, go to Step 5-5.

[0099] Step 5-5: Output the current optimal solution.

[0100] Example 2:

[0101] In this example, a technical comparison is made based on the above Example 1. Table 1 is a comparison table of the optimization target values of different algorithms. The problem scale represents the number of cigarette types. EGA represents the production scheduling algorithm of this solution, MSA represents the traditional continuous average algorithm, and Manual represents the manual production scheduling algorithm. As can be seen from Table 1, EGA has a more significant optimization effect compared to both MSA and the manual production scheduling strategy, with the average improvement of the optimization target being 15.13% and 17.13% respectively. Relatively speaking, the optimization effects of MSA and the manual production scheduling strategy are not obvious enough. This difference mainly stems from the fact that the EGA algorithm is more tailored to the problem characteristics in design. The multi-stage dynamic priority decoding strategy and the highly problem-adaptive evolutionary mechanism can ensure that the algorithm effectively understands and processes the core difficulties and key characteristics of the problem, thereby achieving more accurate and efficient optimization. In addition, the EGA algorithm designed in combination with the problem characteristics also shows higher applicability and robustness when facing changes in the problem scale.

[0102] Table 1 Comparison Table of Optimization Target Values of Different Algorithms

[0103]

[0104] To observe the specific differences between different production scheduling plans, a weekly production demand instance on May 22, 2023, is selected, and the number of cigarette types produced is 31. Compare the Gantt charts of applying EGA, MSA, and the manual production scheduling plan to the instance. Figure 2 This is the Gantt chart of the production scheduling result of the method of this solution. Figure 3 This is the Gantt chart of the production scheduling result of the traditional MSA algorithm. Figure 4 This is the Gantt chart of the manual production scheduling result. Among them, a long bar frame with different numerical numbers represents a type of cigarette. The x-axis is the processing time, and the numbers on the y-axis represent the machine numbers in each process (Stage1, Stage2, Stage3). To effectively measure the superiority of different production scheduling plans, compare the balance indicators of the production scheduling plans obtained by the three methods, as shown in Table 2.

[0105] Table 2 Comparison Table of Balance Indicators of Different Plans

[0106]

[0107] Depend on Figures 2 to 4 The scheduling Gantt chart and the comparison results of the balance index in Table 2 show that the scheduling scheme planned by EGA is superior to the other two algorithms in all balance indicators. Compared with the MSA algorithm, the variance of processing time is reduced by 53.6% on average, and the difference between the latest completion time and the earliest completion time is reduced by 38.7%. Compared with the manual scheduling strategy, EGA performs well in reducing processing time (average variance reduction of 65.9%), reducing idle time (average reduction of 54.3%), and shortening the difference in completion time (average reduction of 40.9%). The EGA algorithm shows the uniformity of machine load on the Gantt chart, avoiding the situation where some machines are overloaded and other machines are idle, reflecting the rationality of production resource allocation and the balance of machine utilization, thus reflecting better production efficiency. In contrast, the optimization results of MSA and manual scheduling show some unbalanced production of machines, which may weaken the smooth operation of the production line and the overall production efficiency. In traditional manual production scheduling practice, experience-driven decisions easily lead to unstable production rhythm and uneven resource allocation, especially in the stage after the wire storage waiting (Stage 2, Stage 3), the idleness of the machine is more obvious, indicating that the traditional manual method faces obvious challenges when dealing with discrete and continuous production modes. In general, compared with other methods, the improved genetic algorithm of this scheme can better balance the allocation of production tasks, optimize production plans, and has good scheduling optimization performance.

[0108] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An intelligent production scheduling method for cigarette manufacturing based on industrial big data, characterized by: It includes the following steps: Step 1: Collect and integrate historical industrial big data, mine the similarity between different cigarette formulas, and output historical example information; Step 2: Initialize the parameters of the genetic algorithm; Step 3: Encode the production sequence of cigarettes to generate an initial solution set to initialize the population of the genetic algorithm; Step 4: Decode the initial solution set using a continuous decoding method or a discrete decoding method; Step 5: Use the genetic algorithm to evolve and update the population and output the current optimal solution; The specific steps of Step 1 include the following steps: Construct the objective function Max z of the example information: Max z = TQ + TD - TC Where, TQ represents maximizing the storage quality, TD represents minimizing the overdue delay penalty or maximizing the early completion reward, and TC represents minimizing the brand change cost; Suppose the formulas of the i-th and j-th cigarettes are i=(v1,v2,...,v n ) and j=(u1,u2,...,u n ), then the cosine similarity function sim(i,j) is as follows: Among them, v m represents the mth formula in the i-th category of cigarettes, u m represents the mth formula in the jth category of cigarettes, m = 1, 2, ..., n; Assume that the switching cost required to produce two types of cigarettes with completely different formulas on machine h in the kth process is C h , the switching time is T h , then the brand switching cost matrix e for machine h switching from producing the i-th type of cigarettes to the j-th type of cigarettes ijk and the card change time matrix w ijk It is expressed as: Calculate the minimum brand change cost according to the brand change cost matrix.

2. According to the intelligent production scheduling method for cigarette manufacturing based on industrial big data described in claim 1, it is characterized in that: Maximizing the storage quality TQ is: Wherein, η(t) represents the storage quality corresponding to the storage time t; η0 represents the initial quality of the tobacco; θ is the mass attenuation coefficient; [l ik ,r ik ] represents the storage time window of the i-th type of cigarette after the k-th process, l ik Indicates the shortest storage time, r ik Indicates the longest storage time; Where t = a ik -b ik-1 Indicates the storage time, that is, the time from the end of the previous process to the start of the next process; a ik represents the start time of processing of the i-th type of cigarettes in the k-th process; b ik-1 represents the end time of the processing of the i-th type of cigarettes in the k-1th process; S1 represents the set of discrete production processes, S1 = {1, 2, ..., p}; M k Represents the set of machines for the kth process, M k ={1,2,...,h}; I represents the set of cigarette categories, and I = {1,2,...,i}; represents the weight coefficient for maximizing storage quality TQ; Minimizing the overdue delay penalty or maximizing the early completion reward TD is: Among them, τ i represents the latest completion date of the i-th category of cigarettes, c i represents the actual completion time of the i-th category of cigarettes; Represents the weight coefficient for minimizing the penalty for late completion or maximizing the reward for early completion TD; Then, the minimum brand change cost TC in the total cigarette production process is expressed as: Where S represents the set of production processes, S = {1, 2, ..., k}; z ijhk Indicates whether the i-th category cigarettes have priority over the j-th category cigarettes in the k-th process on machine h, if yes, it is 1, otherwise it is 0; Represents the weight coefficient for minimizing the card change cost TC; Then, the objective function is:

3. The method for intelligent production scheduling of cigarette manufacturing based on industrial big data according to claim 1 is characterized by: The initialized parameters in Step 2 include: population size P, penalty coefficient λ, machine status Machines, workpiece status Jobs, set the maximum number of iterations epoch_num, crossover probability Pc, and mutation probability Pm.

4. The method for intelligent production scheduling of cigarette manufacturing based on industrial big data according to claim 1 is characterized in that: In Step 4, the steps of decoding the initial solution set using a continuous decoding method include: Input: continuous production process set S2, card change time matrix w ijk , processing time matrix T hik , process k-1 completion time ascending list CHS k-1 ; Step P1: Traversing CHS k-1 , for all types of cigarettes i∈CHS k-1 , do the following: 1) Calculate the completion time ls for each machine h in the current process k hk , machine processing time L hk , switch to the brand change time w of processing cigarette i hik , the processing time t of processing cigarette i hik ; 2) Perform machine allocation based on the ECF strategy: a. Calculate the completion time b of each machine h processing cigarette i in the kth process at the current moment hik =ls hk +w hik +t hik ; b.if count(min(b hik ))=1, select min(b hik ) corresponds to the earliest completed machine; c.if count(min(b hik ))>1, select min(b hik ) corresponds to the earliest completed machine, min(L hk ) corresponds to the machine with the shortest processing time; d. When both b and c are not satisfied, randomly select a machine for processing; 3) Update: The processing status and completion time ls of the kth process of cigarette i ik ; Step P2: Let k = k + 1. If k ∈ S2, then go to Step P1, otherwise output the algorithm result; Output: ascending list CHS of the completion time of each cigarette in process k k , processing time of each machine L hk , completion time ls hk .

5. The method for intelligent production scheduling of cigarette manufacturing based on industrial big data according to claim 1 is characterized by: In Step 4, the steps of decoding the initial solution set using a discrete decoding method include: Input: Discrete production process set S1, card change time matrix w ijk , processing time matrix T hik , process k-1 completion time ascending list CHS k-1 ; Step S1: Traversing CHS k-1 , for all types of cigarettes i∈CHS k-1 , based on the silk storage time window [l ik-1 ,r ik-1 ]Calculate the available storage time of cigarette i [ls ik-1 +l ik-1 ,ls ik-1 +r ik-1 ], the available time of storage silk is arranged in ascending order to obtain the storage available ascending order list CHL k-1 ; Among them, l ik-1 、r ik-1 Respectively represent the shortest and longest storage time after step k-1; ls ik-1 Indicates the time when storage starts; Step S2: Traversing CHL k-1 , for all types of cigarettes i∈CHL k-1 , do the following: 1) Calculate the processing time L for each machine h in the current process k: hk , switch to the brand change time w of processing cigarette i hik , the processing time t of processing cigarette i hik ; 2) Filter the available machine set Mh among the candidate machines to satisfy ls ik-1 +l ik-1 ≤ls hk +w hik ≤ls ik-1 +r ik-1 ,ls hk is the completion time; 3) If Then go to step 4), otherwise go to step 5); 4) No machine meets the storage time window constraint, and cigarette i is assigned to min(ls hk +w hik -(ls ik-1 +r ik-1 )) corresponding to the machine, so that the degree of violation of the wire storage time window is minimized; 5) Perform machine allocation based on the ECF strategy: a. Calculate the completion time b of each machine h processing cigarette i in the kth process at the current moment hik =ls hk +w hik +t hik ; b.if count(min(b hik ))=1, select min(b hik ) corresponds to the earliest completed machine; c.if count(min(b hik ))>1, select min(b hik ) corresponds to the earliest completed machine, min(L hk ) corresponds to the machine with the shortest processing time; d. When both b and c are not satisfied, randomly select a machine for processing; 6) Update: The processing status and completion time ls of the kth process of cigarette i ik , processing time of each machine L hk , completion time ls hk ; Step S3: Let k = k + 1. If k ∈ S1, then go to Step S1, otherwise output the algorithm result; Output: ascending list CHS of the completion time of each cigarette in process k k , processing time of each machine L hk , completion time ls hk .

6. The method for intelligent production scheduling of cigarette manufacturing based on industrial big data according to claim 1 is characterized by: The specific steps of Step 5 include the following steps: Step 5-1: Evaluate the fitness of each feasible solution in the initial solution set; Step 5-2: Use the roulette wheel method for parent selection; Step 5-3: Perform chromosome exchange and gene mutation operations on the selected parent solutions, and record the number of iterations counts; Step 5-4: If counts < epoch_num, go to Step 5-1, otherwise go to Step 5-5; where epoch_num represents the maximum number of iterations; Step 5-5: Output the current optimal solution.

7. The method for intelligent production scheduling of cigarette manufacturing based on industrial big data according to claim 6 is characterized by: The specific steps of Step 5-1 include the following steps: Introduce a penalty number mechanism to impose a penalty coefficient on infeasible solutions based on the objective function Max z, then the fitness function is: Where, f is the fitness function; λ is the penalty coefficient.

Images