Full-process production scheduling method considering processing and assembling

Abstract

The invention provides a whole-process production scheduling method considering processing and assembling, which comprises the following steps: constructing a two-stage flexible flow shop scheduling model, scheduling processing tasks of all parts of a plurality of products on a parallel machine in the first stage, and scheduling assembling of the product parts and processing tasks of semi-finished products in the second stage; minimizing the maximum completion time and the total energy consumption of the machine is taken as a double-optimization target; a multi-target swarm intelligence algorithm is adopted to solve the model, population individuals represent a scheduling scheme through two-segment coding, the first segment of coding defines a process execution sequence, and the second segment of coding defines a machine selection result; in an iteration process, alternately executing a Thompson sampling strategy and a dedirectional sampling and generating strategy according to a preset probability, adaptively selecting a bottom layer optimization operator to realize global search, and constructing a guide solution set to realize local mining; and outputting a non-dominated solution set after iteration is ended, and obtaining a corresponding whole-process scheduling scheme after decoding.

Description

Technical Field

[0001] This invention belongs to the field of production scheduling technology for discrete manufacturing systems, and specifically relates to a production scheduling method that considers the entire process including processing and assembly. Background Technology

[0002] In the manufacturing of complex products such as automobile tires and prefabricated buildings, the production process can often be abstracted as an extended assembly shop scheduling problem. Compared with traditional scheduling problems, the core difficulty of this type of problem lies in its coverage of all processes from parts processing and intermediate assembly to final product processing, and the commencement of assembly operations strictly depends on the completion of all related parts processing tasks. To improve equipment utilization and production line throughput, actual production systems often configure multiple parallel machines with the same or similar functions at each stage, forming a multi-stage parallel machine operation environment. Although this flexible resource allocation is closer to the actual production layout, the strict process sequence constraints between stages and the multiple possibilities of task allocation among parallel machines significantly increase the decision-making dimensions of the scheduling problem and dramatically increase the complexity of resource coordination.

[0003] Existing research on optimization methods for assembly shop scheduling problems largely focuses on simplified scenarios. For example, some literature assumes the assembly stage is a single machine or only considers the parallelization of some processes, failing to fully reflect the actual production characteristics where parallel machines are equipped in multiple stages across the entire process. Regarding optimization objectives, most studies focus on minimizing the maximum completion time, which, while positive in improving production efficiency, struggles to meet the increasingly strong demand for green operations from manufacturing enterprises. As the manufacturing industry places greater emphasis on energy efficiency, simply shortening completion time may lead to equipment idling or uneven load distribution, resulting in energy waste. Therefore, systematically incorporating total machine energy consumption into scheduling objectives while optimizing time performance indicators has become a key challenge in achieving efficient and low-carbon production.

[0004] Although theories related to flexible flow shop scheduling are relatively mature, their models often fail to adequately consider the existence of assembly nodes and their blocking effects on the overall workflow. When applied to scheduling full-process operations including assembly, these models have limitations in expressing part completeness constraints and resource coordination between assembly and processing tasks. Furthermore, in a parallel machine environment, the diversity of part processing paths and the dependence of assembly tasks on the completion status of multiple parts further increase the difficulty of constructing accurate scheduling models and place higher demands on the search efficiency and solution quality of optimization algorithms. Therefore, there is an urgent need to develop novel scheduling models and solution methods that can comprehensively describe resource coordination throughout the entire process and simultaneously optimize time and energy consumption indicators to address the shortcomings of existing research in handling multi-objective assembly scheduling problems under full-process, multi-parallel machine configurations. Summary of the Invention

[0005] To address the shortcomings and deficiencies of existing technologies, this invention provides a production scheduling method and system that considers the entire process, including machining and assembly. It aims to solve the problems of high scheduling complexity and insufficient multi-objective coordination in a multi-parallel machine environment across all processes. The method first constructs a two-stage flexible flow shop scheduling model. The first stage coordinates the parallel machining tasks of all parts for multiple products. The second stage connects the assembly and semi-finished product machining tasks. It is stipulated that the initiation of machining for any product in the second stage requires the completion of machining of all its parts in the first stage. Each stage of the process is equipped with multiple selectable parallel machining machines, with the core optimization objective being to minimize the maximum completion time and total machine energy consumption.

[0006] To efficiently solve this model, this invention employs a multi-objective swarm intelligence algorithm that integrates the Thompson sampling strategy with a solution-oriented sampling and generation strategy. The algorithm represents the scheduling scheme through a two-stage encoding: the first stage defines the execution order of processes, and the second stage determines the machine selection result. During the iterative optimization process, the algorithm alternately executes two strategies with preset probabilities: the Thompson sampling strategy assigns independent prior reward distributions to each underlying optimization operator, dynamically adjusting the operator selection probability based on iterative feedback to achieve efficient global search; the solution-oriented sampling and generation strategy divides the search space, selects representative solutions, and generates a directional guidance solution set, enhancing local mining capabilities and balancing population convergence and diversity. After iteration terminates, a non-dominated solution set is output, which, after decoding, yields a full-process scheduling scheme that balances production efficiency and energy-saving requirements. This scheme adapts to the flexible configuration needs of complex production processes, effectively improving the comprehensiveness and practicality of scheduling decisions.

[0007] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0008] A production scheduling method considering the entire process including machining and assembly includes:

[0009] A two-stage flexible flow shop scheduling model is constructed. In the first stage, the processing tasks of all parts of multiple products are scheduled on parallel machines. In the second stage, the assembly and semi-finished product processing tasks of product parts are scheduled. The start time of processing of any product in the second stage is no earlier than the completion time of processing of all its parts in the first stage. Multiple optional parallel processing equipment are configured for each stage process, with the dual optimization objectives of minimizing the maximum completion time and the total energy consumption of the machines.

[0010] The model is solved using a multi-objective swarm intelligence algorithm. The individuals in the algorithm's population represent the scheduling scheme through a two-stage coding system. The first stage of coding defines the order of process execution, and the second stage of coding defines the machine selection result.

[0011] During the iteration process, the algorithm alternately executes the Thompson sampling strategy and the solution-oriented sampling and generation strategy according to a preset probability. The Thompson sampling strategy adaptively selects the underlying optimization operator to achieve global search, and the solution-oriented sampling and generation strategy constructs a guided solution set to achieve local mining, thus balancing the convergence and diversity of the population.

[0012] After the iteration terminates, the non-dominated solution set is output, and the corresponding full-process scheduling scheme is obtained after decoding.

[0013] Furthermore, the constraints of the scheduling model include: at any given time, only one process can be executed on the same machine; at any given time, only one process can be executed by one machine; each product must complete the processing of all parts in the first stage, the assembly and semi-finished product processing in the second stage in sequence, and cannot be executed in parallel across stages; different models of products follow the same process flow and the set of selectable machines for the same process is consistent; the processing time and energy consumption per unit time of different models of products on the same machine are different; in the first stage, the same functional parts of different products can be processed by different parallel machines with consistent process routes; in the second stage, the set of selectable machines for assembly and finished product processing processes is independently configured.

[0014] Furthermore, the specific definition of the dual optimization objectives is as follows: the maximum completion time is the latest time when all products complete the second stage of finished product processing; the total energy consumption of the machines is the sum of the idle energy consumption and processing energy consumption of all parallel machines during the scheduling cycle.

[0015] Furthermore, the Thompson sampling strategy assigns an independent beta distribution as the prior reward distribution to each underlying optimization operator, and dynamically updates the beta distribution parameters by feeding back the distance between the solution and the reference point during the iteration process, so as to adjust the selection probability of each operator; the underlying optimization operators include four differential evolution operators and one uniform mutation operator. The differential evolution operator achieves population evolution by adjusting the individual differences, and the uniform mutation operator achieves population evolution by adjusting random variables.

[0016] Furthermore, the implementation method of the solution-oriented sampling and generation strategy is as follows: the search space is divided into subspaces by the reference vector set, the individual closest to the ideal point in each subspace is selected as the representative solution, the oriented guidance solution is generated along the vector direction formed by the upper and lower bounds of the decision variable and the representative solution, and the guidance solution set is formed by non-dominated sorting.

[0017] Furthermore, the alternating execution of the two strategies is controlled by a preset probability c. In each generation of evolution, the Thompson sampling strategy is selected to be executed with probability c, and the solution-oriented sampling and generation strategy is selected to be executed with probability 1-c. The c is a preset constant or a parameter that is dynamically adjusted according to the size of the scheduling problem.

[0018] Furthermore, the total length of the two-segment code is adapted to the total number of part processes and the total number of assembly and semi-finished product processing processes. The first half is the process segment code and the second half is the machine segment code. The process segment code adopts a multi-dimensional coding rule that represents the product-part-process hierarchy, and the machine segment code matches the selectable machine for the corresponding process through interval division logic.

[0019] The encoding conversion adopts a sorting and matching method based on the ascending order of real number solutions to convert real number solutions into integer codes with practical scheduling significance; during decoding, a complete processing sequence, machine allocation and assembly processing plan are generated according to the process execution order and machine selection results.

[0020] Furthermore, the iteration termination condition is that the number of iterations reaches a preset value that fits the problem size, or that the objective value of the non-dominated solution set does not improve significantly for multiple consecutive generations; during the iteration process, the reference point and neighborhood are updated in real time, the reference point is the set of optimal values ​​of the objective function of the current non-dominated solution, and the neighborhood is constructed based on the multi-objective decomposition method.

[0021] Furthermore, the non-dominated solution set is obtained through a dynamic filtering strategy: based on the proportion of reference vectors of associated individuals, non-dominated sorting is selected when the proportion is less than 2 / 3, and a decomposition-based method is selected when the proportion is not less than 2 / 3; finally, a scheduling scheme that satisfies the dual optimization objectives is output through decoding.

[0022] And, a production scheduling system considering the entire process including machining and assembly, comprising:

[0023] The model building module is used to construct a two-stage flexible flow shop scheduling model. The first stage schedules the processing tasks of all parts of multiple products on parallel machines. The second stage schedules the assembly and semi-finished product processing tasks of product parts. The processing start time of any product in the second stage is no earlier than the processing completion time of all its parts in the first stage. Each stage of the process is equipped with multiple optional parallel processing machines. The model has the dual optimization objectives of minimizing the maximum completion time and the total energy consumption of the machines.

[0024] The algorithm execution module is used to solve the model using a multi-objective swarm intelligence algorithm. The individuals in the algorithm population represent the scheduling scheme through a two-stage encoding. The first stage of encoding defines the execution order of the process, and the second stage of encoding defines the machine selection result. During the iteration process, the algorithm alternately executes the Thompson sampling strategy and the solution-oriented sampling and generation strategy according to a preset probability. The Thompson sampling strategy adaptively selects the underlying optimization operator to achieve global search, and the solution-oriented sampling and generation strategy constructs a guided solution set to achieve local mining, thus balancing the convergence and diversity of the population.

[0025] The result output module is used to output the non-dominated solution set after the iteration terminates, and the corresponding full-process scheduling scheme is obtained after decoding.

[0026] And a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described above.

[0027] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0028] Compared with the prior art, the present invention and its preferred embodiments have at least the following beneficial effects:

[0029] By constructing a two-stage flexible flow shop scheduling model covering parts processing, assembly, and semi-finished product processing, and clarifying the parts matching constraints in both stages, the scheduling scheme better aligns with the actual production process's collaborative needs, effectively avoiding the scheduling disconnect problem caused by traditional schemes that only focus on certain stages and ignore process coupling. With minimizing maximum completion time and total machine energy consumption as dual optimization objectives, it breaks through the limitations of traditional single-objective scheduling, achieving a synergistic improvement in production efficiency and resource conservation, and adapting to the core demands of modern manufacturing for comprehensive benefits. A multi-objective swarm intelligence algorithm integrating Thompson sampling strategy and solution-oriented sampling and generation strategies balances global exploration and local mining capabilities by dynamically adjusting the selection probability of the underlying optimization operator and the synergistic effect of directional guided population search, improving the efficiency and stability of the optimization process and enabling rapid convergence to a high-quality non-dominated solution set. The two-stage encoding method is precisely adapted to the full-process scheduling requirements, ensuring the clarity and feasibility of process sequencing and machine allocation decisions. The decoded scheduling scheme can be directly applied to actual production configuration, reducing the complexity of technology implementation. The overall solution significantly improves the scientificity and practicality of scheduling decisions in a multi-parallel machine environment across the entire process through deep coupling of models, algorithms and coding, providing a more adaptable scheduling solution for complex discrete manufacturing systems. Attached Figure Description

[0030] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0031] Figure 1 This is a flowchart illustrating the algorithm solution process in an embodiment of the present invention.

[0032] Figure 2 This is a diagram illustrating the encoding conversion method according to an embodiment of the present invention;

[0033] Figure 3 This is a comparison chart of the target values ​​corresponding to the optimal solutions of each algorithm for problem T1-8 in this embodiment of the invention;

[0034] Figure 4 This is a schematic diagram of the algorithm solution results in an embodiment of the present invention;

[0035] Figure 5 A general flowchart is designed for embodiments of the present invention. Detailed Implementation

[0036] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:

[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0038] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0039] To address the poor optimization performance of existing scheduling methods due to their failure to consider subsequent processing constraints of assemblies and the limitations of their algorithm search performance, this embodiment provides a full-process production scheduling method that considers assemblies undergoing processing. The implementation process includes: constructing a full-process production scheduling model considering assemblies undergoing processing, which includes symbolic definitions, constraints, and multiple optimization objectives such as maximum completion time and machine energy consumption; designing a swarm intelligence algorithm that integrates Thompson sampling (TS) and solution-oriented sampling and generation (DS) strategies, and decomposing the multi-objective optimization problem into a single-objective optimization problem for solution. This invention can effectively handle full-process production scheduling requirements considering assemblies undergoing processing, while optimizing energy consumption during processing, resulting in a superior scheduling scheme that balances production efficiency and energy consumption.

[0040] like Figure 5 As shown, the implementation of this scheme specifically includes the following steps:

[0041] Step S1: Describe the full-process production scheduling problem, including machining and assembly, using symbols;

[0042] Step S2: Establish constraints that satisfy the problem;

[0043] Step S3: List the objective function expected in the actual production process according to the embodiment.

[0044] Step S4: Design a swarm intelligence algorithm that integrates Thompson sampling strategy and solution-oriented sampling and generation strategies;

[0045] Step S5: Update the solution set according to the Thompson sampling strategy and the solution-oriented sampling and generation strategy, complete the iterative optimization, and find the set of optimal solutions.

[0046] As a specific solution in this embodiment, step S1 includes the following:

[0047] A production workshop is equipped with m machines and needs to process N different models of products. Each model of product requires the processing and assembly of all J parts before it can enter the assembly stage. After assembly, the finished product still needs to undergo several subsequent processing steps. Different models of products follow the same technological process, and the equipment used in the same processes is consistent. However, the processing time and energy consumption of different models of products on the same equipment vary. Furthermore, some processes are equipped with multiple optional processing machines, constituting parallel processing resources.

[0048] The optimization objective is to determine the processing sequence and machine allocation (i.e., the solution) for each product and its parts to achieve the optimal performance of several indicators.

[0049] Compared to existing scheduling methods involving "processing-assembly", this invention emphasizes flexibility throughout the entire process, meaning that there are several machines available for each process, especially in the parts processing stage.

[0050] In this embodiment, step S2 first considers the following constraints:

[0051] Basic parameters: Pareto solution, a high-dimensional decision variable; The total number of products to be produced; Product number to be manufactured ; The number of parts for each product and the number of parts for each model are equal. Part number, ; Machine serial number; Total number of machines; A sufficiently large positive number; Phase 1 process number; Phase 2 process number; The energy consumption of each machine includes idle energy consumption and processing energy consumption;

[0052] Flexible assembly line scheduling parameters for parts processing: Phase 1 The first product The total number of processes for each part; No. The first product The first part The number of machines that can be selected for each process step; No. The first product The first part This process is done on the machine. Processing time; No. The first product The first part The start time of each processing step; No. The first product The first part The processing time for each step; No. The completion time of the first phase of this product is also the... Assembly start time for this product; No. The first product The first part The process is selected in the machine The value is 1 if the previous processing is performed, otherwise it is 0; Process Prior to process In the machine The value is 1 if the previous processing is performed, otherwise it is 0;

[0053] Flexible assembly line scheduling parameters for assembly and semi-finished product processing stages: No. The first product The total number of machines that can be selected for each process step; No. The first product This process is done on the machine. Processing time; No. The first product The start time of each processing step; The completion time for all products in Phase 2; No. The first product The process is selected in the machine The value is 1 if the previous processing is performed, otherwise it is 0; Process Prior to And adjacent to the process on the machine The value is 1 if the condition is met, otherwise it is 0.

[0054] In this embodiment, the production system scheduling model considered in step S2 includes multiple stages such as parts processing, assembly, and semi-finished product processing. In parts processing, homogeneous but heterogeneous parts for different products can be processed on parallel machines; in assembly and semi-finished product processing, each process also has multiple selectable machines. This resource selectivity characteristic makes the overall system a flexible pipeline scheduling problem. Constraints include:

[0055] Parts processing stage:

[0056] The sequence of processes for each part: ; ;

[0057] The completion time of a workpiece, i.e., the completion time of each workpiece, cannot exceed the total completion time: ;

[0058] At any given time, a single machine can only process one step at a time. ; ;

[0059] At any given time, the same process can only be processed by one machine. ;

[0060] Each machine operates in a cyclical manner: ; ;

[0061] This indicates that all parameter variables must be positive numbers: .

[0062] Constraints during the assembly and semi-finished product processing stages:

[0063] Each product must go through all stages, and each stage can only be processed on one machine: ;

[0064] Constraints on the order of different products in the same stage: ;

[0065] product and products All Phase 1 Processed on the machine and the product In products During pre-processing, constraints must be met; otherwise, the constraints are invalid. ;

[0066] The order in which products are processed on the same machine: ;

[0067] Define decision variables: ; ;

[0068] Define the start time of the workpiece in the second stage, indicating that machining can only proceed if the parts are complete: ;

[0069] The completion time of a product at each stage is determined by the completion time of the previous stage and the processing time of the current stage: .

[0070] Based on this, in this embodiment, step S3 includes the following: expressing the objective function of the maximum completion time. Total energy consumption of the machine :

[0071] ;

[0072] ;

[0073] because For the first The completion time of the first phase of this product is also the... Second phase assembly start time for this product This point in time connects the two phases of the flexible assembly line workshop. Let be the completion time of all products in stage 2. When all products are completed in stage 2, that is, the maximum completion time of the scheduling problem. For the energy consumption of each machine, therefore This represents the total energy consumption of all machines.

[0074] In this embodiment, step S4 can be broken down into the following steps:

[0075] Step S41: Define the parameters of the research question;

[0076] Step S42: Determine the individual encoding method in the swarm intelligence algorithm;

[0077] Step S43: Generate random numbers Each generation of evolution uses the Thompson sampling strategy with probability c and the solution-oriented sampling and generation strategy with probability (1-c) as the evolutionary operation of individuals in the swarm intelligence algorithm.

[0078] Step S44: Write the expression based on the Thompson sampling strategy;

[0079] Step S45: Write the expression based on the de-orientation sampling and generation strategy;

[0080] Step S46: Calculate the objective function value, update the reference point Z*, and update the neighborhood B based on Chebyshev decomposition.

[0081] In this embodiment, step S41 involves defining the population size Np, the Chebyshev decomposition replacement parameter nr, and the strategy selection parameter c; and initializing the Beta distribution parameters. and The learning rate is d; the DS policy parameters are Nw and Ns.

[0082] In this embodiment, the individual encoding in step S42 uses two-segment integer encoding, such as... Figure 2 As shown, the scheduling scheme is represented by two segments of real-number initialization codes, including part operations, part processing machines, and assembly and assembly semi-finished product processing machines. These codes are then converted to integer codes using the LOV method. Specifically, the scheme includes:

[0083] Let the number of operations for a part be... The number of assembly and assembly semi-finished product processes is: Total encoding length , among which the former Code the process segment, then Machine coding;

[0084] In process segment coding, the representation rule for processes is as follows: processing processes use a three-digit code, where the hundreds digit represents the product number, the tens digit represents the workpiece number in the part processing stage, the tens digit is represented by the character 'A' in the assembly and assembly semi-finished product processing stages, and the units digit represents the process number. For example, code 1-1-1 represents the first process of the first workpiece of the first product, and code 1-1-2 represents the second process of the first workpiece of the first product. Based on the known number of products, the number of parts in each product, and the number of processes for each part, firstly, the processes to be arranged are calculated and sorted in ascending order, then the real number codes are sorted in ascending order, and the process codes at corresponding positions are assigned values ​​corresponding to the real numbers. In machine segment coding, real number codes are used, and the elements of the machine segment correspond one-to-one with the workpiece segment, representing the processing machine selected for the corresponding process position. First, the upper limit Ub of the machine segment's independent variable X is selected, then the length J of the machine set that can be selected for each processing process is determined; then, the upper limit Ub of the independent variable is divided into J equal parts; finally, the interval corresponding to the machine segment's independent variable X is determined, the interval index is obtained, and the machine in the machine set is selected according to the index.

[0085] In this embodiment, step S44 can be broken down into the following steps:

[0086] Step S441: Select five underlying optimization operators, including four differential evolution (DE) operators and one uniform mutation (UM) operator, as shown below:

[0087] DE / rand / 1:

[0088] DE / rand / 2:

[0089] DE / current-to-rand / 1:

[0090] DE / current-to-rand / 2:

[0091] UM:

[0092] in, ; This represents a uniform distribution within the range [0,1], where U is the upper limit of the variable and L is the lower limit of the variable.

[0093] Step S442: Each underlying operator is assigned an independent prior reward distribution. Use shape parameters as and The prior distribution can be represented by the beta distribution: .in It is the gamma function. These are observations of a stochastic process.

[0094] Step S433: Assume the prior distribution is a uniform beta distribution. and use Let the posterior distribution after the t-th iteration of the algorithm be used as the current state. The Thompson sampling strategy selects the action with the highest probability of success at the t-th iteration. The underlying operator k makes : This is valid. The reward associated with the underlying operator k, at iteration t, will be used to update the next state if the obtained dissociation reference point is closer. and for: If the obtained dissociation reference point is further away, then the next state is updated accordingly. and for: .

[0095] Step S444: Based on the parameters given for each underlying operator and It can calculate the probability of each underlying operator being selected and perform underlying operator selection.

[0096] In this embodiment, step S45 can be broken down into the following steps:

[0097] Step S451: Select several representative solutions from the current parent population and perform targeted sampling, as shown below:

[0098] Using reference vector sets Divide the entire search space into several subspaces, and in each subspace, select the current parent population. Each individual in the array is assigned to the nearest reference vector. ,form A subpopulation. In each non-empty subspace, select the individual closest to the ideal point. As a representative solution.

[0099] Step S452: Determine the search direction, as shown below:

[0100] and , where L and U are the real lower and upper bound vectors of the decision variables, respectively.

[0101] Step S453: Generate a guiding solution for directional sampling, as shown below:

[0102] Along direction Randomly generated One solution: Along the direction Randomly generate N s One solution Solution for exceeding the boundary and It will be projected back to the feasible region. This is obtained by merging samples from all subspaces. For each solution, sort all solutions by non-dominated order and select non-dominated solutions to store in the guiding solution set S. G .

[0103] Step S454: Guiding the solution set S G Guided dual reproduction, the first stage of which involves reproduction using guided solutions: the current parent population The individuals and the randomly selected guided solution set S G The first stage involves crossover to generate offspring from the guided solution, followed by mutation; the second stage uses reproduction without the guided solution: in the parent population... Internally, a new population is generated by randomly selecting two individuals for crossover and then mutating the resulting offspring. .

[0104] Step S455: Perform complementary environmental selection to update the population. :

[0105] Calculate the number of reference vectors Associated with individuals (Ns), if the number of occupied vectors is small (Ns < Non-dominated sorting is used to prioritize improving convergence; if many vectors are occupied (Ns≥...), If the diversity is not maintained, a decomposition-based approach is used to better preserve the diversity.

[0106] In this embodiment, step S46 can be broken down into the following steps:

[0107] Step S461: Calculate the objective function value;

[0108] Step S462: Update reference point Z*;

[0109] Step S463: Update neighborhood B based on Chebyshev decomposition.

[0110] Step S5 includes the following:

[0111] When t < tmax, t = t + 1, and the steps starting from S43 designed above are executed.

[0112] Otherwise, exit the loop and output the final solution set. After the iteration process is complete, start from the final population. Extract all non-dominated solutions to form the final non-dominated solution set. .

[0113] Using the obtained final non-dominated solution set The scheduling scheme is obtained by decoding the code of each individual and presented in the form of a Gantt chart. Users can select the best scheme from all solutions based on the different emphases on the two optimization objectives.

[0114] Based on the above design, the solution process of the swarm intelligence algorithm (MOEA / D-DT) that integrates Thompson sampling strategy and solution-oriented sampling and generation strategy obtained in this invention is as follows: Figure 1 As shown.

[0115] Pareto solution using the algorithm of this invention The solution employs a two-stage integer encoding: the first stage represents the part's process code, and the second stage represents the machine selection code. The manufacturing process scheduling for each product comprises three parts: multiple part processing stages, part machine selection, and assembly and assembly of semi-finished products machine selection. With numerous variables involved in the decision-making process, this is a high-dimensional decision variable optimization problem. Since the MOEA / D-DT algorithm uses real numbers for iteration, this implementation uses a separate conversion method for the two-stage encoding before merging them into a Pareto solution. The sorting generated by the method of converting real number solutions to integer solutions naturally satisfies the process constraint of "consistent workpiece processing order on all machines." The specific conversion rules for the two-stage conversion method are as follows:

[0116] 1) Process Segment Encoding Conversion Method: Based on the known number of products, the number of parts in each product, and the number of processes for each part. First, the processes of all products need to be sorted according to the part order and its associated processes to obtain a complete product process sequence. The core of this method is to arrange the real-number solutions generated by the algorithm in ascending order and then systematically match them with the above process sequence to convert them into integer solutions with practical scheduling significance. During the matching process, a product is randomly selected, and each digit of the ascending sequence corresponds from left to right to each digit of the product's process sequence. Then, the next product is selected, and the uncorresponding positions in the ascending sequence are matched from left to right to each digit of the newly selected product's process sequence. In this way, the ascending sequence is matched with all product process sequences. For example... Figure 2 As shown, yellow represents the scheduling scheme for product T1, and green represents the scheduling scheme for product T2. The process sequence to be sorted is [1 1 2 3 3 4] in the second row, where the yellow [1 1 2] corresponds to product T1. The first occurrence of 1 indicates process 1 for part 1, and the second occurrence indicates process 2 for part 1. The numbers 3 and 4 correspond to processes for product T2. The real number solution for the process part generated by the MOEA / D-DT algorithm iteration is [0.58 0.47 0.05 0.32 0.64 0.99]. Sort it in ascending order to obtain the ascending sequence [0.05 0.32 0.47 0.58 0.64 0.99]. The ascending sequence is matched one by one with the process sequence of all products, and the integer solution is [3 2 1 1 3 4]. This means that the process 1 of part 1 of product T2 is processed first, then the process 1 of part 2 of product T1 is processed, then the process 1 of part 1 of product T1 is processed, then the process 2 of part 1 of product T1 is processed, then the process 2 of part 1 of product T2 is processed, and then the process 1 of part 2 of product T2 is processed.

[0117] 2) Machine segment encoding conversion method: Real number encoding is used, with each element of the machine segment corresponding one-to-one with a workpiece segment, representing the processing machine selected for the corresponding position and operation. Each element... When product T i The number of available machines for the k-th operation of the j-th workpiece is When the time is right, the interval [0,1) can be divided into equal parts. Take the lower boundary as the first interval, and extend the other intervals sequentially, for example [0, ) represents the first interval, [ , The second interval is ) and so on. If the elements in the machine layer Falling in the first ( When there are ) intervals, then product T i The k-th process for the j-th workpiece is selected as the first... Taiwanese processing machines. For example... Figure 2 As shown, the real solution for the machine part is 0.41, and the number of usable machines is 3. Divide the interval [0,1) into 3 equal parts, [0, ) represents the first interval, [ , ) represents the second interval, [ , If the interval is the third interval, and the real solution 0.41 falls in the second interval, then the second processing machine is selected for the process 1 of part 1 of product T2.

[0118] To intuitively demonstrate the effectiveness of the MOEA / D-DT algorithm of this invention in solving problems, such as Figure 3 As shown, for small-scale engineering problems T1-T8, each comparative algorithm is solved once, yielding a set of target values ​​corresponding to the optimal solutions for each algorithm. It is evident that, compared to the comparative algorithms, the MOEA / D-DT algorithm significantly reduces manufacturing cycle time and energy consumption, resulting in a superior overall performance of the generated scheduling scheme.

[0119] It should be noted that the above MOEA / D-DT algorithm should be understood as the optimal solution scheme of this invention, rather than the only limitation on the solution of the model of this invention. The above comparative experiments have proved that other similar algorithms can also be used to solve the model constructed by this invention. At the same time, the design of the optimal solution scheme and the adaptation work such as coding in the solution process are also part of the creative work of this invention.

[0120] like Figure 4 As shown, if the manufacturing cycle is taken as the priority target, the scheduling (manufacturing cycle 587.7 min, energy consumption 1840.91 kW) based on the MOEA / D-DT algorithm is studied. At 30.6 min, all parts of tire T7 are assembled first, and then assembly and semi-finished product processing begin first. This avoids the need for all parts of all models to be processed before assembly and semi-finished product processing, which is required by traditional scheduling, and effectively improves production efficiency.

[0121] As can be seen from the above, compared with similar existing technologies, the core differences and advantages of this invention are reflected in: 1. Considering the machine selection at each part processing stage; 2. In the part processing stage, different machines can be selected for processing parts with the same function in different products, but the process route for each part is the same, and its production scheduling belongs to the flexible flow shop scheduling problem. In the assembly and semi-finished product processing stages, different machines can be selected for processing each assembly or semi-finished product, and the production scheduling of each product has the characteristics of a flexible flow shop. Therefore, the full-process production scheduling problem of "processing-assembly-processing" is divided into a two-stage flexible flow shop scheduling problem; 3. Providing an optimal solution scheme adapted to the above new model.

[0122] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0123] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0124] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0125] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0126] This invention is not limited to the preferred embodiment described above. Anyone inspired by this invention can derive other forms of production scheduling methods that consider the entire process, including processing and assembly. All equivalent variations and modifications made within the scope of the claims of this invention should be included within the scope of this invention.

Claims

1. A production scheduling method considering the entire process including machining and assembly, characterized in that, include: A two-stage flexible flow shop scheduling model is constructed. In the first stage, the processing tasks of all parts of multiple products on parallel machines are scheduled. In the second stage, the assembly of product parts and the processing of semi-finished products after assembly are scheduled. The processing start time of any product in the second stage is no earlier than the processing completion time of all its parts in the first stage. Multiple optional parallel processing machines are configured for each stage process, with the dual optimization objectives of minimizing the maximum completion time and the total energy consumption of the machines. The model is solved using a multi-objective swarm intelligence algorithm. The individuals in the algorithm's population represent the scheduling scheme through a two-stage coding system. The first stage of coding defines the order of process execution, and the second stage of coding defines the machine selection result. During the iteration process, the algorithm alternately executes the Thompson sampling strategy and the solution-oriented sampling and generation strategy according to a preset probability. The Thompson sampling strategy adaptively selects the underlying optimization operator to achieve global search, and the solution-oriented sampling and generation strategy constructs a guided solution set to achieve local mining, thus balancing the convergence and diversity of the population. After the iteration terminates, the non-dominated solution set is output, and the corresponding full-process scheduling scheme is obtained after decoding.

2. The production scheduling method considering the entire process including processing and assembly, as described in claim 1, is characterized in that: The constraints of the scheduling model include: at any given time, only one process can be executed on the same machine; at any given time, only one process can be executed on the same machine; each product must complete the processing of all parts in the first stage, the assembly and semi-finished product processing in the second stage in sequence, and cannot be executed in parallel across stages; different models of products follow the same process flow and the set of selectable machines for the same process is consistent; the processing time and energy consumption per unit time of different models of products on the same machine are different; in the first stage, the same functional parts of different products can be processed on different parallel machines with the same process route; in the second stage, the set of selectable machines for the assembly processing process and the assembly semi-finished product processing process are independently configured.

3. The production scheduling method considering the entire process including processing and assembly, as described in claim 1, is characterized in that: The specific definitions of the dual optimization objectives are: the maximum completion time is the latest time for all products to complete the second stage of finished product processing; the total energy consumption of the machines is the sum of the idle energy consumption and processing energy consumption of all parallel machines during the scheduling cycle.

4. The production scheduling method considering the entire process including processing and assembly, as described in claim 1, is characterized in that: The Thompson sampling strategy assigns an independent beta distribution as the prior reward distribution to each underlying optimization operator. The beta distribution parameters are dynamically updated by feedback from the distance between the solution and the reference point during the iteration process to adjust the selection probability of each operator. The underlying optimization operators include four differential evolution operators and one uniform mutation operator. The differential evolution operator achieves population evolution by adjusting individual differences, while the uniform mutation operator achieves population evolution by adjusting random variables.

5. A production scheduling method considering the entire process including machining and assembly, as described in claim 1, characterized in that: The implementation method of the solution orientation sampling and generation strategy is as follows: the search space is divided into subspaces by the reference vector set, the individual closest to the ideal point in each subspace is selected as the representative solution, the orientation guidance solution is generated along the vector direction formed by the upper and lower bounds of the decision variable and the representative solution, and the guidance solution set is formed by non-dominated sorting.

6. The production scheduling method considering the entire process including processing and assembly, as described in claim 1, is characterized in that: The alternating execution of the two strategies is controlled by a preset probability c. In each generation of evolution, the Thompson sampling strategy is selected to be executed with probability c, and the solution-oriented sampling and generation strategy is selected to be executed with probability 1-c. c is a preset constant or a parameter that is dynamically adjusted according to the size of the scheduling problem.

7. A production scheduling method considering all processes including machining and assembly, as described in claim 1, characterized in that: The total length of the two-segment code is adapted to the total number of part processes and the total number of assembly and semi-finished product processing processes. The first half is the process segment code and the second half is the machine segment code. The process segment code adopts a multi-dimensional coding rule that represents the product-part-process hierarchy, and the machine segment code matches the selectable machine for the corresponding process through interval division logic. The encoding conversion adopts a sorting and matching method based on the ascending order of real number solutions to convert real number solutions into integer codes with practical scheduling significance; during decoding, a complete processing sequence, machine allocation and assembly and semi-finished product processing plan are generated according to the process execution order and machine selection results.

8. A production scheduling method considering the entire process including processing and assembly, as described in claim 1, characterized in that: The iteration termination condition is that the number of iterations reaches a preset value that fits the problem size, or the objective value of the non-dominated solution set does not improve significantly for several consecutive generations. During the iteration process, the reference point and neighborhood are updated in real time. The reference point is the set of optimal objective function values ​​of the current non-dominated solution, and the neighborhood is constructed based on the multi-objective decomposition method.

9. A production scheduling method considering the entire process including machining and assembly, as described in claim 1, characterized in that: The non-dominated solution set is obtained through a dynamic filtering strategy: based on the proportion of reference vectors of associated individuals, non-dominated sorting is selected when the proportion is less than 2 / 3, and a decomposition-based method is selected when the proportion is not less than 2 / 3; finally, a scheduling scheme that satisfies the dual optimization objectives is output through decoding.

10. A production scheduling system considering the entire process including machining and assembly, characterized in that, include: The model building module is used to construct a two-stage flexible flow shop scheduling model. In the first stage, the processing tasks of all parts of multiple products are scheduled on parallel machines. In the second stage, the assembly and semi-finished product processing tasks of product parts are scheduled. The start time of processing of any product in the second stage is no earlier than the completion time of processing of all its parts in the first stage. Each stage of the process is configured with multiple optional parallel processing equipment. The model has the dual optimization objectives of minimizing the maximum completion time and the total energy consumption of the machines. The algorithm execution module is used to solve the model using a multi-objective swarm intelligence algorithm. The individuals in the algorithm population represent the scheduling scheme through a two-stage encoding. The first stage of encoding defines the execution order of the process, and the second stage of encoding defines the machine selection result. During the iteration process, the algorithm alternately executes the Thompson sampling strategy and the solution-oriented sampling and generation strategy according to a preset probability. The Thompson sampling strategy adaptively selects the underlying optimization operator to achieve global search, and the solution-oriented sampling and generation strategy constructs a guided solution set to achieve local mining, thus balancing the convergence and diversity of the population. The result output module is used to output the non-dominated solution set after the iteration terminates, and the corresponding full-process scheduling scheme is obtained after decoding.

Images