Order acceptance and production scheduling integrated optimization method and system based on 3D printing
By constructing a hybrid integer planning model and heuristic algorithm, 3D printing order acceptance and production scheduling are optimized, and the delayed delivery problem caused by specific order periods in additive manufacturing is solved, which improves efficiency and benefits and reduces costs.
Patent Information
- Application Number
- CN202411099305.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-12-06
AI Technical Summary
Delayed delivery caused by orders in additive manufacturing has a specific construction period, resulting in disharmonious order acceptance and production scheduling, increasing the cost of manufacturers.
The integrated optimization method of order acceptance and production scheduling based on 3D printing is adopted, and by building a hybrid integer planning model and heuristic algorithm, net profit is maximized, order acceptance and production scheduling solutions are optimized, and delay costs are reduced.
It improves the benefits and efficiency of order reception and integrated scheduling in a stand-alone environment of additive manufacturing, reduces costs, and provides an effective solution for order reception and integrated scheduling in a stand-alone environment.
Smart Images

Figure CN119090191A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of 3D printing, and in particular relates to an integrated optimization method and system for order acceptance and production scheduling based on 3D printing. Background Art
[0002] As an emerging field in the manufacturing industry, the development and popularization of additive manufacturing (also known as 3D printing) has brought major changes to social production and life, providing a new processing mode for traditional manufacturing to cope with individual and complex manufacturing needs. However, additive manufacturing companies have always faced problems such as long working hours, high manufacturing costs, and limited equipment molding size. How to overcome the inherent complexity of the additive manufacturing model based on its production characteristics and provide a scientific and effective production scheduling optimization method has become a hot topic and difficulty that the industry and academia are jointly concerned about and researching.
[0003] At present, the research on production planning and scheduling problems in additive manufacturing is constantly developing, mainly focusing on minimizing production intervals, processing costs, and extension costs. The earliest research on production planning problems in additive manufacturing environments was proposed by Li (2017) et al., who studied how to assign parts to different groups of machines to minimize the average production cost per unit material volume for 3D printers with different specifications (i.e., production time and cost per unit material volume, processing time per unit height, setup time, maximum support area and height, etc.). Kucukkoc (2019) studied the problem of minimizing the maximum completion time scheduling for single and multiple additive manufacturing machines, established the corresponding mixed integer model, and obtained the optimal solutions for most problems through the CPLEX solver.
[0004] Faced with the need to solve large-scale additive manufacturing production scheduling in a realistic environment, many scholars have designed improved metaheuristic algorithms for applicable problems. Fera (2018) et al. provided a mathematical model for AM / SLM machine scheduling and used metaheuristic algorithms (such as genetic algorithms) to solve the problem. Subsequently, Fera (2020) et al. improved the additive manufacturing scheduling problem in previous literature and proposed a new metaheuristic algorithm based on the taboo search algorithm for solution. Ying (2022) et al. extended the iterative greedy algorithm to solve the additive manufacturing scheduling problem in a single-machine production environment and improved several calculation methods in the greedy algorithm. Kapadia (2022) et al. studied the order acceptance and scheduling problems of multiple jobs and subsequent processing in an additive manufacturing environment, and designed a random key-based genetic algorithm to select fully or partially accepted orders and generate high-quality production plans that meet all technical constraints. Aloui (2021) studied the production scheduling and typesetting problems with technical constraints and developed a heuristic method to solve large instances. Altekin and Bukchin (2022) studied the problem of production scheduling and layout with technical constraints, with the goal of minimizing the total delay of parts to be produced and maximizing the utilization of additive manufacturing equipment, and developed a heuristic method to solve large instances.
[0005] The technologies disclosed in the above literature all focus on the production planning and scheduling stage of enterprises. However, orders for additive manufacturing are extremely unique and customized. Under the constraints of a certain order duration, enterprises often need to formulate reasonable order acceptance and production scheduling plans to reduce the cost of delays and improve the overall profit of production. Li (2019) et al. considered the problem of dynamic order acceptance and scheduling in a powder bed fusion system. In order to maximize the average unit time profit throughout the entire production cycle, a strategy-based metaheuristic decision method was proposed to solve the problem, and the performance of different strategy sets was verified through comprehensive experiments. Kapadia (2022) et al. studied the problem of order acceptance and scheduling in multiple build chambers while considering operations such as support removal and surface polishing. In this problem, multiple orders can be assigned to one build chamber for processing and subsequent operations.
[0006] However, in the above-mentioned prior art, delayed delivery caused by the specific duration of additive manufacturing orders causes an inconsistency between order acceptance and production scheduling, thereby increasing the cost of the manufacturer. Summary of the invention
[0007] In order to solve the above problems existing in the prior art, the present invention discloses a method and system for integrated optimization of order acceptance and production scheduling based on 3D printing.
[0008] The present invention adopts the following technical solutions:
[0009] An integrated optimization method for order acceptance and production scheduling based on 3D printing, comprising the following steps:
[0010] S1. Obtain the maximum net profit corresponding to a 3D printing service provider, which consists of order revenue and costs;
[0011] S2. Construct a mixed integer programming model for order acceptance and production scheduling that conforms to the production characteristics of 3D printing;
[0012] S3. Solve the constructed mixed integer programming model through a heuristic algorithm and output the optimal order acceptance and production scheduling plan.
[0013] Preferably, in step S1, assume that a 3D printing service provider with one 3D printer receives an order list I at time zero, where i = 1...n, each order contains a part, and each part has a different profit s i , height h i , volume v i , area a i and due date d i . The printing time of the part is related to its height and volume; the parts are assigned to a set of jobs J, where j = 1...m; m < n, and are processed on the 3D printer in the form of jobs.
[0014] Preferably, in step S2, the mixed integer programming model is specifically as follows:
[0015] maxprofit = income - cost(1)
[0016]
[0017]
[0018] CT 0 = 0(7)
[0019]
[0020] Among them, equations (1) and (2) are objective functions, which represent maximizing profits, where profits consist of two parts: revenue and cost. Revenue is the sum of the benefits of all accepted parts; cost consists of production cost and extension cost; equation (3) indicates whether a part is accepted and assigned to a job; equation (4) indicates that the total area of parts produced in each job cannot exceed the maximum area of the machine; equation (5) represents the job utilization constraint, that is, the use is arranged in increasing order starting from job 1; equation (6) indicates that the production completion time of the current job is the sum of the production completion time of the previous job and the production time of the current job; equation (7) indicates that the production completion time of a virtual part is 0; equation (8) indicates the production time of each job; equation (9) indicates the extension cost of each job; equation (10) indicates the production cost of each job; equation (11) indicates that each part cannot exceed the maximum height of the machine; equations (12) and (13) represent 0-1 integer constraints;
[0021] Where i is the index of the part;
[0022] j is the index of the job;
[0023] h i is the height of the i-th part;
[0024] a i is the area of the ith part;
[0025] v i is the volume of the ith part;
[0026] d i is the duration of the i-th part;
[0027] s i is the selling price of the i-th part;
[0028] tc i is the unit postponement cost of the i-th part;
[0029] PT j is the production time of the jth job;
[0030] CT j is the completion time of the jth job;
[0031] TD j is the delay time of the jth job;
[0032] TC j is the delay cost of the jth job;
[0033] PC j is the production cost of the jth job;
[0034] VT is the time it takes the machine to form a unit volume;
[0035] HT is the time for the machine to spray a unit height;
[0036] ST is the machine’s setup time;
[0037] H is the maximum height of the machine production area;
[0038] MA is the largest area of the machine production area;
[0039] OC is the machine’s operating cost per unit time;
[0040] WC is the labor cost per unit time;
[0041] MC is the cost of consumable materials per unit volume;
[0042] profit is the total profit;
[0043] income is the income of all parts;
[0044] cost is the cost of all parts;
[0045] If the i-th part is assigned to the j-th job, X i,j =1; otherwise X i,j is 0;
[0046] If the i-th part is accepted, Y i =1; otherwise Y i is 0;
[0047] If the jth job has at least one part, Z j =1; otherwise Z j is 0.
[0048] Preferably, in step S3, the heuristic algorithm is specifically as follows:
[0049] S31.1, sorting the list of available parts according to different sorting rules;
[0050] S31.2, set an empty job list;
[0051] S31.3. Add a new job to the job list;
[0052] S31.4. Select a part according to the sorting order and arrange it into a job in the job list;
[0053] S31.5, calculate the current profit of the work list, and determine whether the current profit is greater than the previous profit. If so, proceed to step S31.6; if not, remove the part from the available parts list and proceed to step S31.7;
[0054] S31.6, determine whether the job meets the area constraint, if so, remove the part from the available parts list and fix the part in the job; if not, return to step S31.4;
[0055] S31.7, determine whether all parts in the available parts list are sorted, if so, proceed to step S31.8; if not, return to step S31.4;
[0056] S31.8. Determine whether the available parts list is empty. If so, end; if not, return to step S31.3.
[0057] The present invention also discloses an integrated optimization system for order acceptance and production scheduling based on 3D printing, which includes the following modules based on the above method:
[0058] Maximized net profit acquisition module: obtains the maximized net profit consisting of order revenue and cost corresponding to the 3D printing company;
[0059] Mixed integer programming model building module: builds a mixed integer programming model for order acceptance and production scheduling that meets the production characteristics of 3D printing;
[0060] Optimal solution output module: solves the constructed mixed integer programming model through a heuristic algorithm and outputs the optimal order acceptance and production scheduling solution.
[0061] The technical solution of the present invention reduces the cost, improves the benefits and efficiency of order reception and integrated scheduling in a stand-alone environment of additive manufacturing, establishes a mixed integer programming model for order reception and integrated scheduling in a stand-alone environment, designs a heuristic algorithm based on 5 different sorting rules, and proposes two meta-heuristic algorithms (improved genetic algorithm and improved simulated annealing algorithm) to solve the problem in comparison with the implementation examples. The effectiveness of the proposed algorithm was verified by designing small-scale and large-scale numerical experiments and useful inspirations were obtained. The experimental results show that the heuristic algorithm based on 5 different sorting rules of the present invention is close to the optimal solution in small-scale experiments, and has a greater advantage in the solution results compared with the improved genetic algorithm and the improved simulated annealing algorithm in large-scale experiments, and the solution speed is much lower than the two compared meta-heuristic algorithms. The present invention provides good support for order reception and integrated scheduling in a stand-alone environment of additive manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] Figure 1 It is a flow chart of the heuristic algorithm of the preferred embodiment.
[0063] Figure 2 It is the flow chart of improved simulated annealing algorithm.
[0064] Figure 3 It is the encoding of the improved simulated annealing algorithm solution.
[0065] Figure 4 It is the flowchart of improved genetic algorithm.
[0066] Figure 5 It is the operation diagram of the improved genetic algorithm solution.
[0067] Figure 6 This is a comparison chart of the effects of heuristic algorithms.
[0068] Figure 7 It is a block diagram of an integrated optimization system for order acceptance and production scheduling based on 3D printing according to a preferred embodiment of the present invention. DETAILED DESCRIPTION
[0069] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0070] Additive manufacturing (also known as 3D printing) is a new type of manufacturing method that can greatly meet customers' real needs for personalized products and timely delivery. Since additive manufacturing orders have a specific duration, the penalty for delayed delivery will increase costs. Manufacturers hope to reduce costs and increase profits by accepting orders and formulating reasonable production sequences. Based on this, the preferred embodiment of the present invention studies the problem of additive manufacturing order acceptance and scheduling in a single-machine environment, and the machine studied is preferably an SLM / DMLS type additive manufacturing machine.
[0071] The preferred embodiment of the present invention aims at the problem of accepting and scheduling production of additive manufacturing orders in a single-machine environment, and establishes a mixed integer programming model that conforms to the production characteristics of additive manufacturing; at the same time, it comprehensively uses different sorting rules (EDD, selling price, height, height-selling price ratio, construction period-selling price ratio) to construct an effective heuristic algorithm. In order to verify the effectiveness of the proposed embodiment method, it is compared with the Gurobi solver, genetic algorithm and simulated annealing algorithm. Through the results of small-scale and large-scale examples, it can be found that the heuristic algorithm proposed by the present invention has advantages in solution time and accuracy.
[0072] Specifically, this embodiment provides an integrated optimization method for order acceptance and production scheduling based on 3D printing, which comprises the following steps:
[0073] S1. Obtain the maximum net profit corresponding to the additive manufacturer, including order revenue and costs;
[0074] S2. Construct a mixed integer programming model for order acceptance and generation scheduling that meets the production characteristics of additive manufacturing;
[0075] S3. An effective heuristic algorithm is designed to solve the constructed mathematical model and output the optimal order acceptance and production scheduling plan.
[0076] Specifically, in step S1, assume that an additive manufacturing manufacturer with a 3D printer receives an order list I (i = 1... n) at time zero, and each order contains a part. Each part has a different revenue s i , height h i , volume v i , area a i and due date d i , and the printing time of the part is related to its height and volume. To process these parts, they need to be assigned to a set of jobs, J (j = 1... m; m < n), and processed in the form of jobs on an additive manufacturing machine. The enterprise needs to simultaneously determine the set of parts to be processed and the assignment plan to reduce costs and maximize the total net revenue.
[0077] This embodiment focuses on additive manufacturing manufacturers using SLM / DMLS technology. When each job is produced on the machine, the total area of the parts in the job cannot exceed the maximum processing area of the machine, and the maximum height of the part cannot be greater than the maximum height of the printing machine. Since only one additive manufacturing machine is considered for processing, therefore, the processing will be carried out in the order of job set J, and only need to focus on the two problems of how to assign parts to different jobs and which parts need to be excluded.
[0078] To better illustrate and conduct the problem research, the following assumptions need to be made:
[0079] (1) All parts are of the same material;
[0080] (2) The area of the part is the projected area;
[0081] (3) Each part is placed horizontally in the production area;
[0082] (4) Each part can be placed in the machine;
[0083] (5) Each part has a fixed placement direction;
[0084] (6) Each part does not overlap in the machine.
[0085] Specifically, in step S2:
[0086] To more clearly illustrate the mathematical model, the following is an explanation of the symbols and variables used in the mathematical model:
[0087] i is the index of the part;
[0088] j is the index of the job;
[0089] h i is the height of the i-th part;
[0090] a i is the area of the ith part;
[0091] v i is the volume of the ith part;
[0092] d i is the duration of the i-th part;
[0093] s i is the selling price of the i-th part;
[0094] tc i is the unit postponement cost of the i-th part;
[0095] PT j is the production time of the jth job;
[0096] CT j is the completion time of the jth job;
[0097] TD j is the delay time of the jth job;
[0098] TC j is the delay cost of the jth job;
[0099] PC j is the production cost of the jth job;
[0100] VT is the time it takes the machine to form a unit volume;
[0101] HT is the time for the machine to spray a unit height;
[0102] ST is the machine’s setup time;
[0103] H is the maximum height of the machine production area;
[0104] MA is the largest area of the machine production area;
[0105] OC is the machine’s operating cost per unit time;
[0106] WC is the labor cost per unit time;
[0107] MC is the cost of consumable materials per unit volume;
[0108] profit is the total profit;
[0109] income is the income of all parts;
[0110] cost is the cost of all parts;
[0111] Decision variables:
[0112] X i,j When it is 1, the i-th part is assigned to the j-th job; otherwise it is 0; Y i When it is 1, the i-th part is accepted; otherwise it is 0;
[0113] Z j When it is 1, the jth job has at least one part; otherwise it is 0. Overall model:
[0114] maxprofit=income-cost(1)
[0115]
[0116]
[0117] CT 0 =0(7)
[0118]
[0119] Formula (1) and Formula (2) are objective functions, which represent maximizing profit, where profit consists of two parts: revenue and cost. Revenue is the sum of the benefits of all accepted parts; cost consists of production cost and delay cost. Formula (3) indicates whether a part is accepted and assigned to a job. Formula (4) indicates that the sum of the area of parts produced in each job cannot exceed the maximum area of the machine. Formula (5) represents the job utilization constraint, that is, the use is arranged in increasing order starting from job 1. Formula (6) indicates that the production completion time of the current job is the sum of the production completion time of the previous job and the production time of the current job. Formula (7) indicates that the production completion time of a virtual part is 0. Formula (8) indicates the production time of each job. Formula (9) indicates the delay cost of each job. Formula (10) indicates the production cost of each job. Formula (11) indicates that each part cannot exceed the maximum height of the machine. Formulas (12) and (13) represent 0-1 integer constraints.
[0120] In step S3:
[0121] 1. The preferred embodiment of the present invention adopts a heuristic algorithm:
[0122] In order to better explain the designed heuristic algorithm, some terms used in the algorithm are first given below. The job list represents a collection of multiple jobs processed from front to back on a machine, and an empty job list means that no job has been assigned yet. A job represents a batch of parts processed on a machine at the same time, and an empty job means that no part has been assigned yet. The available parts list represents the set of assignable parts. If a part has been assigned to a job, it will be deleted from the available parts list. The temporary job list and temporary job represent the job list and job temporarily generated in a loop for receiving standard judgment. The temporary job list and temporary job are reset in each loop. The current profit represents the profit of the temporary job list in each loop, and the previous profit is initially 0. If the current profit is greater than the previous profit, the current profit is assigned to the previous profit.
[0123] The basic process of the algorithm is as follows Figure 1 As shown, the basic steps are as follows: First, sort the list of available parts according to different sorting rules and set an empty work list. Second, add an empty job to the work list. Third, add a part to the work list in the order of the available parts list and calculate the current profit of the work list. If the current profit is greater than the previous profit and the job meets the area restriction, the part is assigned to the job and deleted from the list of available parts; if the current profit is greater than the previous profit and the job does not meet the area restriction, the part is deleted from the job; if the current profit is less than the previous profit, the part is deleted from the job and the list of available parts. Repeat the above operations until all jobs in the job list are traversed. Finally, loop through the previous steps until the list of available parts is empty. If there are still unassigned parts at the end of the algorithm, these parts are rejected parts. The corresponding algorithm pseudo code is as follows:
[0124]
[0125]
[0126] One of the most critical steps in the constructed heuristic algorithm is how to sort the list of available parts. This is because different sorting rules will affect the final profit value. Therefore, five different sorting rules (EDD, selling price rule, height rule, height-to-sale price rule, and duration-to-sale price rule) are proposed based on the characteristics of the problem under study.
[0127] ·EDD: This rule is the earliest deadline rule. EDD can reduce late costs and thus increase profits.
[0128] Selling Price Rule (SPR): This rule sorts the parts by selling price from high to low. According to this rule, parts with high selling price can be selected first, thus obtaining higher profits.
[0129] Height Rule (HR): This rule sorts in ascending order of height. According to this rule, parts with lower production costs can be processed first, thus reducing production costs and increasing profits.
[0130] Height to Selling Price Rule (HTSPR): This rule sorts in ascending order by the ratio of height to selling price. According to this rule, parts with low height and high selling price can be selected first, thus achieving low production cost and high selling price, and obtaining higher profits.
[0131] ·Date-to-sale price rule (DDTSPR): This rule sorts in ascending order based on the ratio of delivery date to sales price. According to this rule, parts with low delay cost and high sales price can be selected first, thus obtaining higher profits. Taking five parts as an example, according to the above five sorting rules, the part parameters and the final sorting results are shown in the following table.
[0132] Table 1. Parts parameters
[0133]
[0134] Table 2. Sorting results
[0135]
[0136] 2. The meta-heuristic algorithm of the comparative embodiment is adopted, as follows:
[0137] Simulated annealing algorithm and genetic algorithm are typical algorithms in meta-heuristic algorithms, and these two algorithms represent the algorithms for individual optimization and group optimization respectively. Therefore, two meta-heuristic algorithms (simulated annealing algorithm and genetic algorithm) are introduced to solve the problem under study, and they are compared with the proposed heuristic algorithm in numerical experiments.
[0138] 2.1. Improved simulated annealing algorithm
[0139] According to the characteristics of the problem under study, the encoding method, perturbation rules and fitness calculation of the classic simulated annealing algorithm are modified. The following focuses on these three parts. The basic process of improving the simulated annealing algorithm is as follows: Figure 2 shown.
[0140] The improved simulated annealing algorithm is as follows:
[0141] S32.1. Randomly generate an initial solution w and calculate the fitness f(w) of the initial solution;
[0142] S32.2, generate a new solution w′ after perturbation and calculate the new fitness f(w′);
[0143] S32.3, let δ = f(w′)-f(w), determine whether δ≤0 holds, if so, accept the new solution according to the Metropolis criterion; if not, accept the new solution and update the fitness w = w′, f(w) = f(w′).
[0144] S32.4, determine whether the number of iterations reaches the preset number, if yes, proceed to step S32.5; if no, return to step S32.2;
[0145] S32.5. Determine whether the termination condition is met. If so, end; if not, lower the temperature and repeat the number of iterations, then return to step S32.2.
[0146] In the coding method, each part has two attributes, one is which job it is assigned to, and the other is whether it is accepted. These two attributes form a sublist. Therefore, each part has a sublist, and the sublists of all parts form a parts list. This list is the solution. This example gives a case with three parts and three operations, such as Figure 3 As shown. A solution has 3 parts and two attributes. Before the perturbation, the first sublist of the solution represents the first job, and the first part is assigned to the job when the job is accepted.
[0147] In the perturbation rule, a part is randomly selected and its two properties are randomly transformed. Figure 3 In , the second sublist is perturbed so that the second part changes from being non-acceptable to being acceptable and is put into the second job.
[0148] In the fitness calculation, fitness refers to the profit obtained from the solution. If the solution does not meet the area restriction, the fitness will be reduced by a maximum value.
[0149] 2.2 Improved Genetic Algorithm
[0150] Similarly, based on the characteristics of the problem being studied, the encoding method, crossover and mutation rules, and fitness calculation are changed on the basis of the classic genetic algorithm. The basic process of the improved genetic algorithm is as follows Figure 4 shown.
[0151] The improved genetic algorithm is as follows:
[0152] S33.1. Generate an initial population;
[0153] S33.2, determine whether the number of iterations reaches the preset number, if yes, output the result; if no, proceed to step S33.3;
[0154] S33.3. Calculate the fitness of each individual in the population, and after selection, crossover, and mutation operations, proceed to step S33.2.
[0155] In the encoding method, each part has two attributes, one is which job it is assigned to, and the other is whether it is accepted. These two attributes make up a gene, so each part represents a gene, and the parts make up a list of parts, which is considered a chromosome. This example gives a case with five parts and five operations, such as Figure 5 shown. Figure 5 There are two chromosomes in the genome, each with five parts, and two properties of each part make up a gene. The first gene of the first chromosome represents the first part assigned to the second job and accepted.
[0156] The crossover rule is to randomly select two genes from two chromosomes for crossover. Figure 5 In the case of the two chromosomes, the third and second genes were exchanged. The mutation rule is to randomly change two properties of a gene on a chromosome. Figure 5 The fourth gene on chromosome 1 in the gene has mutated.
[0157] The fitness is calculated by the profit obtained by the chromosome. If the chromosome does not meet the area limit, the fitness will be reduced by a maximum value.
[0158] The numerical experiments are as follows:
[0159] In order to verify the effect of the proposed heuristic algorithm (5 rules EDD, SPR, HR, HTSPR, DDTSPR) in the acceptance and scheduling problem of additive manufacturing orders in a single machine environment, it is compared with the planning solver Gurobi and two meta-heuristic algorithms. Two groups of experiments were designed, one group of experiments was a small-scale experiment, that is, the number of parts was 6, 8, 10, 12, 14, 16, 18, 20; the other group of experiments was a large-scale experiment, that is, the number of parts was 100, 200, 300, 400, 500, 600. In the small-scale experiment, the heuristic algorithm proposed in this embodiment is compared with the solver Gurobi, because the solver Gurobi can solve the exact optimal solution in the small-scale experiment, so it can well reflect the performance of the heuristic algorithm of this embodiment in a small scale. In large-scale experiments, the solver was unable to find the optimal solution within 3600 seconds due to the excessive number of parts. Therefore, the heuristic algorithm of this embodiment was compared with the simulated annealing algorithm and the genetic algorithm, which can also verify the performance of the heuristic algorithm of this embodiment under large-scale order demands.
[0160] The computer parameters of this experiment are configured as Apple M2 (8-core CPU, 10-core GPU). The algorithm code is compiled by Python 3.9, the environment is DataSpell 2023.2, and the problem is solved by calling the solver Gurobi through Python. The Gurobi version is version 10.0.3, and the maximum solution time of Gurobi is 3600 seconds.
[0161] 1. Experimental parameters
[0162] In the small-scale experiment, the parameters of the parts are derived from the parameters of 20 parts in the paper of Kucukkoc (2019). The small-scale experiment set up 8 groups of experiments (experiment numbers are P6, P8, P10, P12, P14, P16, P18, P20), and the required number of parts in each group of experiments are randomly selected from these 20 parts. The machine parameters are partially derived from the paper of Kucukkoc (2019) (VT=0.030864, HT=0.7, ST=1, MA=900, MC=0.05, OC=1, WC=0.5). In the large-scale experiment, 6 groups of experiments were set up (experiment numbers are P100, P200, P300, P400, P500, P600), and the volume v of the part in each group of experiments is 0. i Randomly generate in the range [70,1200], the area of the part is a i The height h of the part is randomly generated in the range [40,600] i Randomly generate the price of the part in the range [10,50] i v i *U[0.1,0.5], the construction period d of the part i Randomly generated in the range of [AVG_PC-AVG_PC*m / 2,AVG_PC+AVG_PC*m / 2], where AVG_PC is the average processing time of all parts, m is the duration parameter, which is a random number in the range [0,2]; delay penalty tc i For i *0.02. The parameter settings are shown in the table below. The small-scale parts price, construction period, and delay penalty are the same as those in the large-scale experiment. The machine parameters of the large-scale experiment are the same as those in the small-scale experiment.
[0163] Table 3. Experimental parameters
[0164]
[0165] 2. Small-scale experimental results
[0166] The performance of the heuristic algorithm proposed in this embodiment is compared with that of the Gurobi solver in a small-scale experiment.
[0167] Table 4. Small scale experimental results
[0168]
[0169] Among them, GAP BEST =(Gurobi profit-Heuristic Alorithm profit) / Gurobi profit.
[0170] As shown in the table above, 8 sets of experiments were conducted to show the comparison and gap between the two solution methods. Specifically, the first column is the experiment number, where the number indicates the number of parts in the experiment, the third and fourth columns are the solution results and CPU solution time of the Gurobi solver, the fourth to eighth columns are the solution results of the heuristic algorithm under the five sorting rules, and the ninth column is the gap between the maximum value of the five heuristic algorithm solution results and the Gurobi solution result.
[0171] According to the above table, it can be seen that the heuristic algorithm proposed in this embodiment has a relatively small difference between the solution results of the optimal sorting rule and the solution results of Gurobi under different sorting rules. In addition, in the experiment, the CPU solution time is a very important indicator. In the small-scale experiment, the solution time of the heuristic algorithm proposed in this embodiment is less than 1 second, while the CPU solution time of the Gurobi solver in solving the P16 group increases to 3600 seconds. Based on this, it can be inferred that the CPU solution time of the Gurobi solver will continue to increase in large-scale experiments, so the heuristic algorithm has a huge advantage in solution speed and the difference in solution results is small, while Gurobi is more difficult to get results in larger-scale experiments.
[0172] Among the 8 groups of experiments, several heuristic algorithms under 5 sorting rules have better performance. For example, the DDTSPR rule performs very well in the six groups of experiments P8, P10, P14, P16, and P20. Therefore, it can be found that the height and price of parts have a greater impact on the solution results in this heuristic algorithm.
[0173] 3. Large-scale experimental results
[0174] In order to better verify the advantages of the algorithm proposed in the preferred embodiment of the present invention, the heuristic algorithm under five sorting rules is compared with the meta-heuristic algorithm simulated annealing algorithm and the genetic algorithm using large-scale experiments. Hereinafter, HA represents the heuristic algorithm proposed in the present invention, SA represents the improved simulated annealing algorithm, and GA represents the improved genetic algorithm.
[0175] Table 5. Large-scale experimental results
[0176]
[0177] As shown in the table above, 6 groups of experiments were conducted. The first column is the experiment number, where the number represents the number of parts. The second column represents the optimal solution results of the heuristic algorithm under the five sorting rules. The third column represents the average CPU solution speed of this group of five sorting rules. The fourth, fifth, sixth and seventh columns are the solution results and CPU solution time of the simulated annealing algorithm and the genetic algorithm, respectively. Since the solution results of the simulated annealing algorithm and the genetic algorithm are volatile, the results in the table above are the arithmetic averages of the 10 results.
[0178] The last two columns of Table 5 show the gap between the two metaheuristic algorithms and the heuristic algorithms. In terms of the average profit of the solution, it can be noticed that the heuristic algorithm has better performance when the number of parts is large. Its maximum gap with the simulated annealing algorithm and the genetic algorithm is 31.35% and 25.68%, respectively. In terms of the average solution time, the maximum solution time of the heuristic algorithm is only 25.32 seconds, while both metaheuristic algorithms exceed 180 seconds. It can be seen that the heuristic algorithm has more advantages in solution results and solution time.
[0179] Figure 6 It is the maximum, minimum and average value of the improvement rate of the objective function between the heuristic algorithm and the two meta-heuristic algorithms in the 10 parallel experiments of each group of experiments. Among them, H-SA represents the improvement rate of the objective function of the heuristic algorithm compared with the improved simulated annealing algorithm, and H-GA represents the improvement rate of the objective function of the heuristic algorithm compared with the improved genetic algorithm. It can be seen from the figure that in some parallel experiments of the three groups of P100, P200 and P300, some gap values are less than 0, indicating that the solution results of the genetic algorithm are slightly higher than those of the heuristic algorithm. But overall, the solution results of the heuristic algorithm have a greater advantage, and the effect in some parallel experiments even exceeds 50%.
[0180] In summary, the present invention studies the acceptance and production scheduling problem of additive manufacturing orders in a single-machine environment and establishes a mixed integer programming model with additive manufacturing production characteristics. In order to solve this problem, five different sorting rules are designed: EDD, selling price, height, height-selling price ratio, and construction period-selling price ratio. An effective heuristic algorithm is constructed based on these rules. At the same time, two meta-heuristic algorithms are introduced, namely the improved simulated annealing algorithm and the genetic algorithm. In order to verify the effectiveness of the algorithm proposed in the present invention, two scales of numerical experiments are constructed. The Gurobi solver is used in the small-scale experiment to compare with the heuristic algorithm. The results show that the heuristic algorithm has a smaller average gap with Gurobi in the small-scale experiment, and the solution speed is also much lower than the Gurobi solver, which verifies the advantage of the heuristic algorithm under small-scale orders. In the large-scale experiment, the heuristic algorithm is compared with the simulated annealing algorithm and the genetic algorithm. The results show that the heuristic algorithm is optimal in each group of experiments, and it can be verified that the constructed heuristic algorithm also has certain advantages under large-scale order numbers.
[0181] like Figure 7 As shown, this embodiment discloses an integrated optimization system for order acceptance and production scheduling based on 3D printing, based on the above method embodiment, including the following modules:
[0182] Maximized net profit acquisition module: obtains the maximized net profit consisting of order revenue and cost corresponding to the 3D printing company;
[0183] Mixed integer programming model building module: builds a mixed integer programming model for order acceptance and production scheduling that meets the production characteristics of 3D printing;
[0184] Optimal solution output module: solves the constructed mixed integer programming model through a heuristic algorithm and outputs the optimal order acceptance and production scheduling solution.
[0185] For other contents of this embodiment, please refer to the above method embodiment.
[0186] In summary, the present invention establishes a mixed integer programming model with total profit maximization as the objective function, designs a heuristic algorithm based on 5 different sorting rules, and proposes two comparative meta-heuristic algorithms (improved genetic algorithm and improved simulated annealing algorithm) in the embodiment to solve the problem. Finally, the effectiveness of the proposed algorithm is verified by designing small-scale and large-scale numerical experiments and useful inspiration is obtained.
[0187] The implementation scheme described above is only a preferred solution of the present invention and does not limit the present invention in any form. There are other variations and modifications without exceeding the technical solution described in the claims.
Claims
1. An integrated optimization method for order acceptance and production scheduling based on 3D printing, characterized by Follow these steps: S1. Obtain the maximum net profit consisting of order revenue and cost corresponding to the 3D printing company; S2. Construct a mixed integer programming model for order acceptance and production scheduling that meets the production characteristics of 3D printing; S3. Solve the constructed mixed integer programming model through heuristic algorithm and output the optimal order acceptance and production scheduling plan.
2. The integrated optimization method for order acceptance and production scheduling based on 3D printing as claimed in claim 1, characterized in that: In step S1, assume that a 3D printing service provider with a 3D printer receives an order list I at time zero, where i = 1...n. Each order contains a part, and each part has a different selling price s i , height h i , volume v i , area a i and part processing time d i . The printing time of the part is related to its height and volume. The parts are assigned to a set of jobs J, where j = 1...m; m < n, and are processed on the 3D printer in the form of jobs.
3. The integrated optimization method for order acceptance and production scheduling based on 3D printing as claimed in claim 1, characterized in that: In step S2, the mixed integer programming model is as follows: maxprofit=income-cost(1) CT0=0(7) Among them, equations (1) and (2) are objective functions, which represent maximizing profits, where profits consist of two parts: revenue and cost. Revenue is the sum of the benefits of all accepted parts; cost consists of production cost and extension cost; equation (3) indicates whether a part is accepted and assigned to a job; equation (4) indicates that the total area of parts produced in each job cannot exceed the maximum area of the machine; equation (5) represents the job utilization constraint, that is, the use is arranged in increasing order starting from job 1; equation (6) indicates that the production completion time of the current job is the sum of the production completion time of the previous job and the production time of the current job; equation (7) indicates that the production completion time of a virtual part is 0; equation (8) indicates the production time of each job; equation (9) indicates the extension cost of each job; equation (10) indicates the production cost of each job; equation (11) indicates that each part cannot exceed the maximum height of the machine; equations (12) and (13) represent 0-1 integer constraints; Where i is the index of the part; j is the index of the job; h i is the height of the ith part; a i is the area of the ith part; v i is the volume of the ith part; d i is the duration of the i-th part; s i is the selling price of the i-th part; tc i is the unit postponement cost of the i-th part; PT j is the production time of the jth job; CT j is the completion time of the jth job; TD j is the delay time of the jth job; TC j is the delay cost of the jth job; PC j is the production cost of the jth job; VT is the time it takes the machine to form a unit volume; HT is the time for the machine to spray a unit height; ST is the machine’s setup time; H is the maximum height of the machine production area; MA is the largest area of the machine production area; OC is the machine’s operating cost per unit time; WC is the labor cost per unit time; MC is the cost of consumable materials per unit volume; profit is the total profit; income is the income of all parts; cost is the cost of all parts; If the i-th part is assigned to the j-th job, X i,j =1; otherwise X i,j is 0; If the i-th part is accepted, Y i =1; otherwise Y i is 0; If the jth job has at least one part, Z j =1; otherwise Z j is 0.
4. The integrated optimization method for order acceptance and production scheduling based on 3D printing as claimed in any one of claims 1 to 3, characterized in that: In step S3, the heuristic algorithm is as follows: S31.1, sorting the list of available parts according to different sorting rules; S31.2, set an empty job list; S31.
3. Add a new job to the job list; S31.
4. Select a part according to the sorting order and arrange it into a job in the job list; S31.5, calculate the current profit of the work list, and determine whether the current profit is greater than the previous profit. If so, proceed to step S31.6; if not, remove the part from the available parts list and proceed to step S31.7; S31.6, determine whether the job meets the area constraint, if so, remove the part from the available parts list and fix the part in the job; if not, return to step S31.4; S31.7, determine whether all parts in the available parts list are sorted, if so, proceed to step S31.8; if not, return to step S31.4; S31.
8. Determine whether the available parts list is empty. If so, end; if not, return to step S31.
3.
5. An integrated optimization system for order acceptance and production scheduling based on 3D printing, based on the method described in any one of claims 1 to 4, characterized in that Includes the following modules: Maximized net profit acquisition module: obtains the maximized net profit consisting of order revenue and cost corresponding to the 3D printing company; Mixed integer programming model building module: builds a mixed integer programming model for order acceptance and production scheduling that meets the production characteristics of 3D printing; Optimal solution output module: solves the constructed mixed integer programming model through a heuristic algorithm and outputs the optimal order acceptance and production scheduling solution.
Citation Information
Cited By
Steel member production scheduling optimization method and device and computer readable storage medium
CN120746181A
Multi-device parallel anti-sand grid printing scheduling optimization method
CN121787862A
