A workshop production scheduling control method

Abstract

The present application belongs to the technical field of production scheduling control, and particularly relates to a workshop production scheduling control method. The steps include: collecting relevant data; constructing a worker proficiency table, a machine precision table, and a process quality threshold table; constructing a multi-objective scheduling function containing product quality maximization and worker load balance objectives; and solving the function using an improved double-mapping adaptive chaotic differential evolution algorithm, which optimizes the solution through operations such as chaotic mapping initialization of parameters, dynamic adjustment of scaling factors and crossover rates, three-stage coding, archive management, and the like. The present application optimizes quality and efficiency, improves global search and local convergence capabilities, maintains solution set diversity, avoids local optimality, and improves scheduling scheme optimization quality and algorithm robustness.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of production scheduling control, and particularly relates to a workshop production scheduling control method. BACKGROUND

[0002] With the development of the textile industry in the direction of small batch, multi-variety and flexibility, the workshop production scheduling faces the demand for comprehensive optimization of multiple objectives such as product quality and worker load balance. Traditional scheduling methods are mostly based on heuristic rules or single-objective models, and it is difficult to balance quality and efficiency. Even if multi-objective evolutionary algorithms are introduced, fixed scaling factors and crossover rates are usually used, which makes it difficult to balance global exploration and local convergence during the search process and easily falls into local optimum. In addition, existing multi-objective algorithms only rely on simple metrics such as crowding degree or Euclidean distance to maintain population diversity, which cannot comprehensively reflect the distribution uniformity of the solution set in the multi-dimensional objective space, so that the Pareto front often lacks representativeness in some directions. Therefore, there is an urgent need for a production scheduling method that can effectively improve the production efficiency and product quality of the textile workshop. SUMMARY

[0003] The application proposes a workshop production scheduling control method to solve the technical problems in the background art.

[0004] To achieve the above purpose, the technical scheme adopted by the application comprises the following steps:

[0005] S1, collecting the process flow of the textile workshop, the machine running state, the historical operation record of the workers and the real-time fatigue monitoring data;

[0006] S2, based on the collected data, constructing a worker proficiency table representing the proficiency of the workers, a machine precision table reflecting the running stability of the equipment, and a process quality threshold table for the quality standard requirements of each process;

[0007] S3, based on the worker proficiency table, the machine precision table and the process quality threshold table, constructing a multi-objective scheduling function, the objectives of which include maximizing product quality and balancing worker load;

[0008] S4, using an improved double-mapping adaptive chaotic differential evolution algorithm to solve the multi-objective scheduling function, the implementation of which comprises:

[0009] S41, first, initializing the chaotic mapping parameters, and generating two mapping sequences using Logistic mapping and Tent mapping , , , wherein is the Logistic parameter, is the Tent parameter, and d represents the generation number;

[0010] S42, calculate the scaling factor and the crossover rate for the gth individual in the dth generation, the scaling factor is calculated as follows: , and the crossover rate is calculated as follows , where represents the preset boundary;

[0011] S43, then perform encoding and population initialization, adopt a three-section encoding structure including process sequence encoding, worker allocation encoding and equipment allocation encoding, and randomly generate a population , the size of the population is and evaluate the multi-objective value of each individual;

[0012] S44, then in the population evolution process, for each individual , randomly select from the current Pareto archive A, randomly select from the population of the current iteration , then generate a mutation vector, the calculation method is as follows ;

[0013] S45, according to the binary crossover rule, use to cross and to generate a trial vector ;

[0014] S46, incorporate the trial vector into the candidate set, perform fast non-dominated sorting and crowding comparison, and retain the first individuals from them to form a new population , if the population diversity index of 10 consecutive generations is less than the set diversity index threshold, reset the initial seed of the chaotic mapping;

[0015] S47, merge all newly generated non-dominated solutions in into the archive A, and remove the out-of-limit solutions in the order of minimum crowding priority, until the maximum number of iterations is reached, output all non-dominated solutions in the archive A as the final scheduling scheme.

[0016] As a preferred, the implementation of the worker proficiency table in step S2 includes:

[0017] First, collect the worker's historical operation records, which include the output per unit time , operation time , rework rate , operation time period t and fatigue index ;

[0018] Then, based on the time decay function The proficiency score of each operation is weighted to obtain the base proficiency score of worker i for process j , which is calculated as:

[0019] , wherein is a quality penalty factor;

[0020] When the fatigue index exceeds a set threshold, the current score is corrected according to to obtain the corrected proficiency score , wherein represents a fatigue index correction parameter;

[0021] The corrected proficiency score of each worker for all processes forms a proficiency score table.

[0022] As a preferred embodiment, the method for constructing the machine precision table in step S2 is:

[0023] The machine operating state is collected, including the operating error sequence of each machine m at different processes j , and the number of unplanned shutdowns ;

[0024] The error sequence is standardized, and the precision score function of each machine is defined as: , wherein is the upper limit of the normal error range, represents the historical average error of machine m at process j;

[0025] The precision score of all machines m for each process j forms a machine precision table.

[0026] As a preferred embodiment, the method for constructing the process quality threshold table in step S2 includes:

[0027] First, the quality detection score of each process j in the historical task is counted by machine learning, and the mean and standard deviation are calculated;

[0028] The lower limit threshold of quality is calculated according to the mean and standard deviation, and the calculation method is: , wherein represents the error coupling factor between upstream and downstream processes, which is calculated by counting the correlation coefficient of adjacent process quality indicators;

[0029] All processes and corresponding quality lower limit thresholds form a process quality threshold table.

[0030] As preferred, the method for constructing the multi-objective scheduling function in step S3 comprises:

[0031] Firstly, a product quality function is constructed: where I represents the number of people, J represents the number of processes, and M represents the number of machines, is a quality compliance factor term, which is 1 when the quality of a process is greater than or equal to the lower threshold of quality, otherwise 0;

[0032] Then, a worker load balance function is constructed to minimize the difference in work load between workers, and the calculation method is: where, represents the total work load of worker , which is defined as the sum of all process operation times it undertakes, where, is the operation time of worker in process j at time period t, represents the average work load of all workers;

[0033] Combining the product quality function and the worker load balance , a final function is constructed, and the final function is solved to find the maximum value of the final function .

[0034] As preferred, the implementation steps of the population diversity index in step S46 are:

[0035] Firstly, collect the complete multi-objective vectors of all solutions in the current Pareto archive A;

[0036] Map these multi-objective vectors to a discretized grid according to a preset division scheme, count the number of solutions in each cell and calculate the probability distribution, and calculate the information entropy of the distribution of the entire solution set in the target space;

[0037] Then, calculate the standard deviation of each target dimension in the archive separately, and aggregate the overall standard deviation through arithmetic mean;

[0038] Finally, the obtained information entropy and overall standard deviation are weighted and fused to obtain the population diversity index.

[0039] As preferred, the generation of the initial population in step S43 further comprises:

[0040] Pareto-guided sampling, which is used to preferentially extract non-dominated solutions from the historical archive A as parent seeds;

[0041] The disturbance local reorganization is used for respectively applying small random replacement and insertion left shift operator to the seed to generate new individuals to make up .

[0042] As preferred, in the step S44 From the archive A, the maximum principle of crowding degree is selected, Then, from the current population Randomly selected and meet .

[0043] Compared with the prior art, the application has the advantages and positive effects that:

[0044] 1. The multi-objective scheduling function based on worker proficiency, machine precision and process quality threshold is constructed, multiple objectives such as product quality maximization and worker load balance are comprehensively considered, so that the textile workshop production scheduling can optimize quality and efficiency at the same time. Compared with the traditional single-objective optimization method, the application can reasonably distribute the work load while meeting the quality requirements, avoiding the low efficiency problem caused by uneven load in the production process.

[0045] 2. The parameters are initialized by using chaotic mapping, and the scaling factor and the crossover rate are adjusted adaptively, which effectively improves the global search ability and local convergence, thereby improving the optimization quality of the scheduling scheme. Compared with the fixed parameter method of the traditional algorithm, the dynamic adjustment mechanism of the application makes the algorithm more flexible and efficient in the search process, and can better avoid local optimal solution.

[0046] 3. Through the non-dominated solution management of archive A and the minimum crowding degree priority elimination strategy, the application can maintain the diversity of the solution set during the evolution process, ensure that the final output scheduling scheme has good representativeness in the target space, and avoid the situation that the solution set is concentrated on a certain target. Compared with the traditional simple crowding degree or distance measurement method, the diversity maintenance mechanism of the application can comprehensively reflect the uniformity of the distribution of the solution set, improve the robustness and scheduling effect of the algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0048] Figure 1 It is a whole structure flow schematic diagram of a kind of workshop production scheduling control method. DETAILED DESCRIPTION

[0049] In order to enable the above-mentioned objects, features and advantages of the present application to be clearer, the present application will be further described below with reference to the drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0050] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, and therefore, the present application is not limited to the specific embodiments disclosed in the following description.

[0051] Embodiments, as the textile workshop develops towards high flexibility, small batch, multi-variety direction, production scheduling needs to meet the product quality stability, worker operation load balancing and other goals, while the existing scheduling scheme will appear quality in actual execution, worker load imbalance seriously affects the quality and efficiency of textile production problem. Therefore, the present application proposes a workshop production scheduling control method, by constructing a dynamic scheduling model and an improved optimization algorithm, realizing the collaborative optimization of product quality maximization and worker load balancing, the specific process is as shown in Figure 1 .

[0052] First, in order to achieve the goal of real-time grasp of workshop state, eliminate information island, collect all process flow, machine running state, worker historical operation record and real-time fatigue monitoring data in textile workshop, then based on the collected data, respectively build worker proficiency table representing worker operation proficiency, machine precision table reflecting equipment running stability, and process quality threshold table for quality standard requirements of each process. Specifically, the implementation of the worker proficiency table includes: collecting worker historical operation records, the worker historical operation records include unit time output , operation time , repair rate , operation time period t and fatigue index ; then based on the time decay function , the proficiency score of each operation is weighted, wherein represents the parameter for controlling the rate of time decay, the basic proficiency score of worker i to process j is obtained , the calculation method is: , wherein is a quality penalty factor; when the fatigue index exceeds the set threshold, the current score is corrected according to , to obtain the corrected proficiency score , wherein represents the fatigue index correction parameter; the corrected proficiency score of each worker to all processes The proficiency score table is constituted. The calculation of the fatigue index is that the fatigue data collected including heart rate data and blink frequency data are normalized, and then the fatigue index is obtained by summing the weighted values.

[0053] Then the machine precision table is constructed by collecting the machine running state, the machine running state including the running error sequence of each machine m at different processes j , the number of unplanned shutdowns ; the error sequence is normalized, and the precision score function of each machine is defined as: , wherein is the upper limit of the normal error range, , the historical average error of machine m at process j is represented; the precision scores of all machines m for each process j constitute the machine precision table. By quantifying the error performance of the machine at different processes, the precision is dynamically evaluated combined with historical data, so that the scheduling model can preferentially select high-precision equipment to process critical processes and reduce quality fluctuations.

[0054] Then the process quality threshold table is constructed, and the construction method includes: first, the quality detection score of each process j produced in the historical task is counted by machine learning, and the mean and standard deviation are calculated; the lower limit threshold of quality is calculated according to the mean and standard deviation, and the calculation method is: , wherein is the error coupling factor of the upstream and downstream processes, which is calculated by counting the correlation coefficient of the quality indicators between adjacent processes; all processes and the corresponding quality lower limit threshold constitute the process quality threshold table.

[0055] In order to achieve the dual goals of optimizing product quality and balancing worker load, the present application is based on the worker proficiency table, the machine precision table and the process quality threshold table to construct a multi-objective scheduling function, the goals including: maximizing product quality, balancing worker load. Specifically, the construction method of the multi-objective scheduling function includes: first, constructing the product quality function: , wherein I represents the number of workers, J represents the number of processes, and M represents the number of machines, is the quality compliance factor, which is 1 when the quality of a process is greater than or equal to the quality lower limit threshold, otherwise it is 0; then the worker load balancing function is constructed, which minimizes the difference in work load between workers, and the calculation method is: , wherein represents the total work load of the worker , which is defined as the sum of all process operation times it undertakes, , wherein is the operation time of the worker processing process j at time period t, representing the average work load of all workers; combining the product quality function and the worker load balance to build a final function , the final function solving the maximum value of the final function .

[0056] In order to achieve high quality and stability of the solved scheduling scheme, the improved double-mapping adaptive chaotic differential evolution algorithm is used to solve the multi-objective scheduling function. First, the parameters are initialized by chaotic mapping to enhance the randomness and global search ability of the optimization algorithm. Two random sequences are generated using Logistic and Tent mapping, which help the algorithm avoid falling into local optimal solution, thereby expanding the search range and improving the quality of the solution. Then the scaling factor and crossover rate of each individual are calculated, which will be dynamically adjusted according to the progress of the current search. The scaling factor controls the search step, while the crossover rate determines the mating probability between individuals, ensuring that the algorithm finds a balance between global exploration and local optimization. The specific implementation is first to initialize the parameters by chaotic mapping, and two mapping sequences are generated using Logistic mapping and Tent mapping wherein , wherein is the Logistic parameter, is the Tent parameter, and d represents the generation number. The scaling factor and crossover rate of the gth individual in the dth generation are calculated, the scaling factor is calculated as follows: and the crossover rate is calculated as follows wherein represents a preset boundary.

[0057] Then encoding and population initialization are performed, using a three-section encoding structure including process order encoding, worker allocation encoding and equipment allocation encoding, and a population is randomly generated, the size of the population is and the multi-objective value of each individual is evaluated. Specifically, during population initialization, the algorithm generates a group of individuals randomly, which are created according to the three-section encoding structure. Each individual represents a possible scheduling scheme, including the specific arrangement of process order, worker allocation and equipment allocation. The size of the population is preset and adjusted according to the complexity of the problem and the computing power. The multi-objective value of each individual is calculated, including product quality, worker load balance and other multiple objectives. By evaluating these objective values, reference can be provided for the next evolution, ensuring that the population can continuously approach the optimization target during the evolution process. Among them, the initial population The generation also includes Pareto guided sampling for preferentially extracting non-dominated solutions from the historical archive A as parent seeds, which can ensure the initial distribution of the population to be more diverse, thus improving the search efficiency of the algorithm. In order to enhance the diversity of the population, the algorithm also disturbs the seeds by local recombination methods, including small random replacement and insertion left shift operator, to generate new individuals to supplement the population . In this way, the population initialization stage not only covers a wide range of solution space, but also provides high-quality initial solutions for the subsequent evolution process.

[0058] Then in the population evolution process, for each individual , randomly select from the current Pareto archive A, , randomly select from the current iteration population , and then generate a mutation vector, wherein is selected from the archive A according to the principle of maximum crowding degree, is randomly selected from the current population and satisfies . The calculation method of the mutation vector is , in this way, new individuals are generated, which not only help to increase the diversity of the population, but also explore new solution space areas in the search process. Then according to the binary crossover rule, use to cross and to generate a trial vector , specifically, according to the binary crossover rule, if a randomly generated number is less than , an element is selected from the current individual , otherwise the corresponding element is selected from the mutation vector , so that the generated trial vector contains and partial characteristics, which may include part of the information of the current individual and part of the information after mutation, the purpose is to combine the advantages of the two to explore potential optimal solutions in the solution space. The crossover operation is used to combine individuals to generate new solutions and ensure that the diversity of the population is effectively maintained. Finally, the trial vector is integrated into the candidate set, and the trial vector represents a possible optimal solution, and then fast non-dominated sorting and crowding comparison are performed, from which the first individuals are retained to form a new population , if the population diversity index less than a set diversity index threshold, resetting the chaotic mapping initial seed. In this way, it is ensured that the elite individuals of the population can continue to participate in the next round of evolution, thereby promoting the algorithm to converge to a better solution. The population diversity index The implementation steps are as follows: first, collect the complete multi-objective vectors of all solutions in the current Pareto archive A; map these multi-objective vectors to a discretized grid according to a pre-set partition scheme, count the number of solutions in each cell and calculate the probability distribution, and calculate the information entropy of the entire solution set in the target space distribution; then calculate the standard deviation of each target dimension in the archive respectively, and aggregate the overall standard deviation by arithmetic average; finally, the information entropy and the overall standard deviation are weighted and fused to obtain the population diversity index.

[0059] Finally, all newly generated non-dominated solutions in are merged into the archive A, and the out-of-limit solutions are removed according to the principle of minimum crowding degree priority, until the maximum number of iterations is reached, and all non-dominated solutions in the archive A are output as the final scheduling scheme. Specifically, in each generation of evolution, the newly generated non-dominated solutions need to be added to the archive A. The archive A saves all high-quality solutions in the entire optimization process, and each solution performs well in multiple objectives and is not dominated by other solutions. Therefore, first, all non-dominated solutions generated in the current generation are collected and merged with the solutions in the archive A. After merging, the solutions in the archive A need to be non-dominated sorted, that is, each solution in the archive is checked to determine whether it is a non-dominated solution. If a solution is dominated by other solutions, it will be eliminated, otherwise it will be retained in the archive. The purpose of non-dominated sorting is to ensure that all solutions in the archive A perform well in the multi-objective space as much as possible. Then, the solutions in the archive A are compared in terms of crowding degree, that is, the diversity of each solution in the target space is measured by calculating its density. Solutions with lower crowding degree mean that they are more isolated in the target space and have higher representativeness and should be retained. Therefore, by comparing the crowding degrees of the solutions, those solutions that are more evenly distributed in the target space are preferentially retained. When the number of solutions in the archive A exceeds the predetermined upper limit, some solutions need to be removed according to the principle of minimum crowding degree priority to retain the most representative solutions. Finally, after non-dominated sorting and crowding degree removal, the solutions finally retained in the archive A are the optimal solution set of the multi-objective optimization problem. These solutions represent the optimal scheduling scheme under multiple objectives such as product quality and worker load balancing, and are output as the final scheduling scheme.

[0060] The above merely describes preferred embodiments of the present application, but is not intended to limit the present application to other forms, any person skilled in the art can make changes or modifications to the above disclosed technical contents into equivalent embodiments with equivalent changes, and apply to other fields, but any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application without departing from the technical solution content of the present application still belongs to the protection scope of the present application technical solution.

Claims

1. A method of controlling production scheduling in a plant, characterized by, The method comprises the following steps: S1, collecting the process flow of the textile workshop, the machine running state, the historical work record of the worker and the real-time fatigue monitoring data; S2, based on the collected data, a worker proficiency table representing the worker's operation proficiency, a machine precision table reflecting the stability of the equipment operation, and a process quality threshold table for the quality standard requirements of each process are constructed; S3, based on the worker proficiency table, the machine precision table and the process quality threshold table, a multi-objective scheduling function is constructed, the targets including product quality maximization and worker load balance; S4, an improved double-mapping adaptive chaotic differential evolution algorithm is used to solve the multi-objective scheduling function, and the implementation of the solution includes: S41, first, chaotic mapping parameter initialization, using Logistic mapping and Tent mapping to generate two mapping sequences wherein , wherein is a Logistic parameter, is a Tent parameter, and d represents algebra; S42, respectively calculate the scaling factor and the crossover rate for the gth individual in the dth generation, the scaling factor is calculated in the following manner: , and the crossover rate is calculated in the following manner , wherein is a preset boundary; S43. Next, encoding and population initialization are performed, using a three-stage encoding structure including process sequence encoding, worker allocation encoding, and equipment allocation encoding, and the population is randomly generated. The population size is And evaluate the multi-objective values ​​for each individual; S44, then in the process of population evolution, for each individual randomly selected from the current Pareto archive A , randomly selected from the population of the current iteration , randomly selected from the population of the current iteration , then generate a mutation vector, the calculation method is ; S45, according to the binary crossover rule, using will be and crossover to generate a trial vector ; S46, Experimental vector Merge into the candidate set, perform fast non-dominated sorting and crowding comparison, and retain the top candidates. Individuals constitute a new population. If the population diversity index for 10 consecutive generations If the value is less than the set diversity index threshold, the initial seed of the chaos mapping is reset. S47, will All newly generated non-dominated solutions are merged into archive A, and out-of-bounds solutions are eliminated in order of minimum congestion until the maximum number of iterations is reached. Then, all non-dominated solutions in archive A are output as the final scheduling scheme.

2. The method of claim 1, wherein, The implementation of the worker proficiency table in step S2 includes: First, a worker's history of work records including a unit time output amount , operation time , a repair rate , a work time period t, and a fatigue index are collected. Then the time decay function is applied The proficiency score for each job is weighted to obtain the base proficiency score of worker i for process j , which is calculated as wherein is a mass penalty factor; When fatigue index Exceeding the set limit When the threshold is reached, the current score is adjusted according to... Make corrections to obtain the revised proficiency score. ,in For fatigue index correction parameters; score each worker's proficiency at all processes construct a proficiency score table.

3. The method of claim 1, wherein, The construction method of the machine precision table in step S2 is: collecting machine operating states, the machine operating states comprising a sequence of operating errors of each machine m at different process j , number of unplanned stoppages ; The error sequence is normalized, and the precision score of each machine is defined as a score function: wherein, is the upper limit of the normal range of error, is the historical average error of machine m at process j; The precision score of each machine m for each process j is calculated A machine precision table is constructed.

4. The workshop production scheduling control method according to claim 1, characterized in that, The construction method of the process quality threshold table in step S2 includes: First, the quality detection score of each process j in the historical task is counted by machine learning, and the mean and standard deviation are calculated; The lower limit threshold of quality is calculated according to the mean value and the standard deviation, and the calculation manner is: wherein is an error coupling factor of the upstream and downstream processes, which is calculated by a correlation coefficient of the quality indexes between adjacent processes. all the procedures with the corresponding quality lower threshold value a table of procedure quality threshold values is constituted.

5. The method of claim 1, wherein, The construction method of the multi-objective scheduling function in step S3 includes: First, the product quality function is constructed: where I is the number of people, J is the number of processes, and M is the number of machines, is a quality compliance factor term, which is 1 when the quality of a process is greater than or equal to the lower threshold value of quality, and 0 otherwise. Then, a worker load balancing function is constructed to minimize the difference in workload among workers. The calculation method is as follows: ,in, Representative of workers The total workload is defined as the sum of the operation times of all processes it undertakes. ,in, For workers The operation time of process j is during time period t. This represents the average workload of all workers. combining a product quality function and a worker load balance function to construct a final function solving the final function for a maximum value .

6. The method of production scheduling control in a workshop according to claim 1, wherein, The population diversity index in the step S46 The implementation steps are as follows: First, collect the complete multi-objective vector of all solutions in the current Pareto archive A; Map these multi-objective vectors to the discretized grid according to the preset division scheme, count the number of solutions in each grid and calculate the probability distribution, and calculate the information entropy of the entire solution set in the target space distribution; Then, the standard deviation of each target dimension in the archive is calculated, and the overall standard deviation is aggregated by arithmetic mean; Finally, the information entropy and the overall standard deviation are weighted and fused to obtain the population diversity index.

7. The method of shop floor production scheduling control according to claim 1, wherein, The step S43 of generating the population further comprises: Pareto-guided sampling is used to preferentially extract non-dominated solutions from the historical archive A as parent seeds; perturbation local reorganization, for applying small random permutations and insertions of left shift operators to the above seeds respectively to generate new individuals to complement .

8. The method of claim 1, wherein, The step S44 From the archive A, select by the principle of maximum congestion, Then randomly select from the current iteration of the population And meet .

Images