Multi-stage hybrid flow shop scheduling optimization method and system based on improved MPRVNS

Abstract

The invention discloses a multi-stage hybrid flow shop scheduling optimization method and system based on an improved MPRVNS, and relates to the field of shop scheduling optimization, and the scheduling optimization method comprises the steps: initializing the input parameters of an algorithm; calculating the complexity of each order; orders are dynamically distributed to machines with light loads in all stages; an improved MPRVNS algorithm is used to optimize the manufacturing cycle of the order. According to the improved MPRVNS-based multi-stage hybrid flow shop scheduling optimization method and system, a machine load peak value is accurately predicted through a dynamic processing time model, an order complexity sorting rule is combined, high-complexity order delay accumulation is avoided, the length of a critical path is shortened, the number of iterations is reduced through adaptive neighborhood search of MPRVNS, and the optimization efficiency of the multi-stage hybrid flow shop scheduling optimization method and system is improved. Meanwhile, high-quality gene segments of parents and individuals are fused, invalid search is reduced, a machine allocation path of an order is fixed, the scheduling scheme adjustment frequency is reduced, and computational logic is simplified.

Description

Technical Field

[0001] The present invention relates to the field of workshop scheduling optimization, and particularly to a multi-stage hybrid flow shop scheduling optimization method and system based on improved MPRVNS. Background Art

[0002] The order scheduling of a multi-stage hybrid flow shop is a typical workshop scheduling problem. Particularly for the multi-stage proportional hybrid flow shop scheduling problem with deteriorating jobs, it aims to minimize the manufacturing cycle by optimizing order allocation, job sequencing, and machine scheduling strategies. This technology can be applied to complex production scheduling scenarios in manufacturing, such as electronic assembly, automobile manufacturing, and other fields that require multi-stage collaborative processing. In the "proportional flow shop scheduling method" proposed by Shakhlevich et al. (1998), the problem of minimizing the total weighted completion time of a single-machine multi-stage proportional flow shop was studied. Its technical solution is based on a permutation schedule with a fixed processing order, assuming that the processing order is the same for each stage, and the job sequencing is optimized through dynamic programming. In the classic variable neighborhood search (VNS), the iterative process depends on modifying a single solution and updating its neighborhood during the search process (Mladenovic and Hansen, 1997). VNS has been applied to many combinatorial optimization problems, such as the flow shop scheduling problem (Zhao et al., 2017), the vehicle routing problem (Ferreira et al., 2018). The "Hybrid Variable Neighborhood Search" (HVNS) proposed by Li et al. (2014) for the hybrid flow shop scheduling problem combines the local search strategies of variable neighborhood search (VNS) and genetic algorithms to optimize the scheduling scheme for multi-stage multi-machine, and its technical measures include random key coding, neighborhood structure design, and adaptive search mechanism.

[0003] However, the above technical solutions have the following problems:

[0004] (1) Failure to consider dynamic processing time and order removal time: This technology assumes that the actual processing time of the workpiece is a fixed value and does not incorporate the dynamic impact of machine running time on the processing time (i.e., the model), resulting in the scheduling scheme being unable to adapt to the actual scenario where the processing time deteriorates gradually with machine operation. At the same time, it does not consider the removal time after order processing, ignoring the preparation time required for equipment to switch orders in actual production, thus reducing the practicality of the scheduling results;

[0005] (2) Single-machine environment limitation: This method is only designed for a single-machine multi-stage flow shop and has not been extended to the multi-machine parallel scenario (i.e., a hybrid flow shop with multiple machines in each stage), unable to make full use of the multi-machine collaborative ability to shorten the manufacturing cycle, restricting its application in large-scale production;

[0006] (3) Unoptimized deteriorating job characteristics: The algorithm does not design a dedicated optimization strategy for deteriorating jobs (i.e., jobs whose processing time increases dynamically with the machine running time), resulting in a significant decrease in the efficiency and accuracy of the scheduling scheme under the dynamic processing time model;

[0007] (4) Lack of order grouping constraint handling: The hard constraints of order grouping processing (such as non-interruptibility of orders and the requirement for continuous processing of jobs within the same order) are not introduced, resulting in the need for frequent adjustment of the scheme when the algorithm involves multi-order collaborative scheduling, with high computational complexity and difficulty in ensuring feasibility;

[0008] (5) Insufficient computational efficiency: In the scenario of multi-stage and multi-machine, its neighborhood search strategy does not combine the problem characteristics (such as the processing consistency of the proportional flow shop), resulting in an overly large search space, slow algorithm convergence speed, and difficulty in meeting the real-time scheduling requirements. Summary of the Invention

[0009] The purpose of the present invention is to provide a multi-stage hybrid flow shop scheduling optimization method and system based on improved MPRVNS, which overcomes the deficiencies of the prior art and solves the problems existing in the above-mentioned prior art.

[0010] To solve the above problems, the technical solutions adopted by the present invention are as follows:

[0011] The present invention provides a multi-stage hybrid flow shop scheduling optimization method based on improved MPRVNS, and the scheduling optimization method includes:

[0012] S1: Initialize the input parameters of the algorithm;

[0013] The input parameters include: the number of working stages m of the workshop, the number of machines q in each working stage, the number of orders U to be processed, the processing time p i , the order removal time s, and the machine start time t0 = 0;

[0014] S2: Calculate the complexity of each order;

[0015] The complexity calculation formula of the order:

[0016] S3: Dynamically allocate orders to the machines with lighter loads in each stage;

[0017] According to the order complexity calculated and quantified in S2, allocate the orders to the machines with lighter loads in each stage dynamically, and the same order is fixed to be processed by the machine with the same number in all stages;

[0018] S4: Use the improved MPRVNS algorithm to optimize the manufacturing cycle of the orders;

[0019] The specific optimization steps of the improved MPRVNS algorithm are as follows:

[0020] S41: Adopt random key coding, map order allocation and sorting to a real number vector, generate a random key vector, and then generate a feasible scheduling plan through decoding;

[0021] S42: Select parent individuals for parametric crossover, retain the allocation strategy of high-complexity orders, and randomly adjust the machine allocation of low-complexity orders;

[0022] S43: Design a dedicated neighborhood structure for different stages, accelerate convergence by combining local search, and recalculate the manufacturing cycle;

[0023] The calculation formula of the manufacturing cycle is:

[0024]

[0025] Preferably, the processing time p in S1 i Adopt a dynamic processing time model: Dynamically associate the actual processing time of the workpiece with the machine running time, and at the same time embed the order removal time s as a fixed constraint.

[0026] Preferably, the workpiece sorting and order scheduling rules in S3 are as follows:

[0027] (1) There is no constraint on the workpieces within an order: Exchanging workpieces within the same order does not affect the total manufacturing cycle. Therefore, random order processing is adopted to reduce the computational complexity;

[0028] (2) Non-increasing sorting between orders: Arrange the orders in non-increasing order of , and give priority to processing orders with high processing complexity to minimize the critical path delay.

[0029] Preferably, in S42, the initial solution is input into the designed neighborhood structure for search to obtain the global optimal solution; the global optimal solution is the optimal manufacturing cycle when the order is completed.

[0030] The present invention also provides a multi-stage hybrid flow shop scheduling optimization system based on the improved MPRVNS. The scheduling optimization system includes:

[0031] An input parameter initialization module for initializing the input parameters of the algorithm;

[0032] The input parameters include: the working stage m of the workshop, the number q of machines in each working stage, the number U of orders to be processed, the processing time p i , the order removal time s, and the machine start time t0 = 0;

[0033] An order complexity calculation module that calculates the complexity of each order;

[0034] The complexity calculation formula for the order:

[0035] An order dynamic allocation module that dynamically allocates orders to machines with lighter loads in each stage;

[0036] According to the quantified order complexity calculated in the order complexity calculation module, the orders are dynamically allocated to the machines with lighter loads in each stage, and the same order is fixed to be processed by the machine with the same number in all stages;

[0037] A manufacturing cycle optimization module that optimizes the manufacturing cycle of orders using an improved MPRVNS algorithm;

[0038] The specific optimization steps of the manufacturing cycle optimization module are as follows:

[0039] S41: Adopt random key encoding, map order allocation and sorting to a real number vector, generate a random key vector, and then generate a feasible scheduling plan through decoding;

[0040] S42: Select parent individuals for parametric crossover, retain the allocation strategy of high-complexity orders, and randomly adjust the machine allocation of low-complexity orders;

[0041] S43: Design a dedicated neighborhood structure for different stages, combine local search to accelerate convergence, and recalculate the manufacturing cycle;

[0042] The calculation formula for the manufacturing cycle is:

[0043]

[0044] Preferably, the processing time p in the input parameter initialization module i Adopt a dynamic processing time model: Dynamically associate the actual processing time of the workpiece with the machine running time, and at the same time embed the order removal time s as a fixed constraint.

[0045] Preferably, the workpiece sorting and order scheduling rules in the order dynamic allocation module are as follows:

[0046] (1) There is no constraint on workpieces within an order: the exchange of workpieces within the same order does not affect the total manufacturing cycle, so a random order is adopted for processing to reduce the computational complexity;

[0047] (2) Non-increasing sorting between orders: Sort the orders in non-increasing order of ∏ Ji∈Ok (1 + p i ) to give priority to processing orders with high processing complexity in order to minimize the critical path delay.

[0048] Preferably, in S42, the initial solution is input into the designed neighborhood structure for search to obtain the global optimal solution; the global optimal solution is the optimal manufacturing cycle when the order is completed.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention accurately predicts the peak value of machine load through a dynamic processing time model, combines the order complexity sorting rule, avoids the accumulation of high-complexity order delays, shortens the critical path length, reduces the number of iterations through the adaptive neighborhood search of MPRVNS, and at the same time integrates high-quality gene fragments of parental individuals, reduces ineffective searches, fixes the machine allocation path of orders, reduces the adjustment frequency of the scheduling scheme, and simplifies the calculation logic.

[0051] The present invention reduces the load difference and improves resource utilization rate through a dynamic allocation strategy, allocates orders according to the real-time load, avoids single-machine overload, takes ∏Dk as an index, preferentially allocates high-complexity orders to idle machines, reduces the scheduling steps of workpieces within the order, reduces the need for manual intervention, and shortens the manufacturing cycle.

[0052] The present invention reduces the equipment idle time and overall energy consumption through a multi-machine cooperation strategy, accurately schedules to shorten the workpiece waiting time, reduces the material loss rate during the processing, and at the same time, in the scenario of random fluctuations in processing time, the manufacturing cycle volatility of the MPRVNS algorithm is significantly better than the traditional method, supports large-scale workpiece scheduling scenarios, and the calculation time increases linearly, meeting the requirements of the intelligent manufacturing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a schematic diagram of the overall process structure of the multi-stage hybrid flow shop scheduling optimization method and system based on improved MPRVNS;

[0054] Figure 2 is a schematic diagram of the main process structure of MPRVNS in the multi-stage hybrid flow shop scheduling optimization method and system based on improved MPRVNS. DETAILED DESCRIPTION OF THE INVENTION

[0055] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0056] Please refer to Figure 1 —2, the present invention provides a multi-stage hybrid flow shop scheduling optimization method based on improved MPRVNS, and the scheduling optimization method includes:

[0057] S1: Initialize the input parameters of the algorithm;

[0058] The input parameters include: the working stage m of the workshop, the number q of machines in each working stage, the number U of orders to be processed, and the processing time p i , the order removal time s, and the machine startup time t0 = 0;

[0059] The processing time p in S1 i Adopt a dynamic processing time model: Dynamically associate the actual processing time of the workpiece with the machine running time, and at the same time embed the order removal time s as a fixed constraint;

[0060] S2: Calculate the complexity of each order;

[0061] The complexity calculation formula of the order:

[0062] S3: Dynamically allocate orders to the machines with lighter loads in each stage;

[0063] According to the order complexity calculated and quantified in S2, allocate the orders to the machines with lighter loads in each stage dynamically. The same order is fixed to be processed by the machine with the same number in all stages;

[0064] The workpiece sorting and order scheduling rules in S3 are as follows:

[0065] (1) There is no constraint on the workpieces within an order: The exchange of workpieces within the same order does not affect the total manufacturing cycle. Therefore, a random order is adopted for processing to reduce the computational complexity;

[0066] (2) Non-increasing sorting between orders: Arrange the orders in non-increasing order of , and give priority to processing the orders with high processing complexity to minimize the critical path delay.

[0067] S4: Use the improved MPRVNS algorithm to optimize the manufacturing cycle of the orders;

[0068] The specific optimization steps of the improved MPRVNS algorithm are as follows:

[0069] S41: Adopt random key coding, map the order allocation and sorting to a real number vector, generate a random key vector, and then generate a feasible scheduling scheme through decoding;

[0070] S42: Select parent individuals for parametric crossover, retain the allocation strategy of high-complexity orders, and randomly adjust the machine allocation of low-complexity orders;

[0071] S43: Design a dedicated neighborhood structure for different stages, combine local search to accelerate convergence, and recalculate the manufacturing cycle;

[0072] The calculation formula of the manufacturing cycle is:

[0073]

[0074] In S42, the initial solution is input into the designed neighborhood structure for search to obtain the global optimal solution; the global optimal solution is the optimal manufacturing cycle when the order is completed.

[0075] This embodiment also provides a multi-stage hybrid flow shop scheduling optimization system based on improved MPRVNS. The scheduling optimization system includes:

[0076] An input parameter initialization module for initializing the input parameters of the algorithm;

[0077] The input parameters include: the number of working stages m of the workshop, the number of machines q in each working stage, the number of orders U to be processed, the processing time p i , the order removal time s, and the machine start time t0 = 0;

[0078] The processing time p in the input parameter initialization module i Adopts a dynamic processing time model: Dynamically associates the actual processing time of the workpiece with the machine running time, and at the same time embeds the order removal time s as a fixed constraint;

[0079] An order complexity calculation module for calculating the complexity of each order;

[0080] The calculation formula for the complexity of the order:

[0081] An order dynamic allocation module for dynamically allocating orders to the machines with lighter loads in each stage;

[0082] According to the quantified order complexity calculated in the order complexity calculation module, the orders are dynamically allocated to the machines with lighter loads in each stage, and the same order is fixed to be processed by the machine with the same number in all stages;

[0083] The workpiece sorting and order scheduling rules in the order dynamic allocation module are as follows:

[0084] (1) There is no constraint on the workpieces within an order: the exchange of workpieces within the same order does not affect the total manufacturing cycle, so a random order is adopted for processing to reduce the computational complexity;

[0085] (2) Non-increasing sorting between orders: Arrange the orders in non-increasing order of , and give priority to processing orders with high processing complexity to minimize the critical path delay;

[0086] A manufacturing cycle optimization module for optimizing the manufacturing cycle of orders using the improved MPRVNS algorithm;

[0087] The specific optimization steps of the manufacturing cycle optimization module are as follows:

[0088] S41: Adopt random key coding, map order allocation and sorting into a real number vector, generate a random key vector, and then generate a feasible scheduling plan through decoding;

[0089] S42: Select parent individuals for parametric crossover, retain the allocation strategy of high-complexity orders, and randomly adjust the machine allocation of low-complexity orders;

[0090] S43: Design a dedicated neighborhood structure for different stages, accelerate convergence by combining local search, and recalculate the manufacturing cycle;

[0091] The calculation formula for the manufacturing cycle is:

[0092]

[0093] In S42, the initial solution is input into the designed neighborhood structure for search to obtain the global optimal solution; the global optimal solution is the optimal manufacturing cycle when the order is completed.

[0094] The present invention accurately predicts the peak value of machine load through a dynamic processing time model, combines the order complexity sorting rule, avoids the cumulative delay of high-complexity orders, shortens the critical path length, reduces the number of iterations through the adaptive neighborhood search of MPRVNS, and at the same time integrates high-quality gene fragments of parent individuals, reduces ineffective search, fixes the machine allocation path of orders, reduces the adjustment frequency of the scheduling plan, and simplifies the calculation logic.

[0095] The present invention adopts a dynamic allocation strategy to reduce the load difference, improve resource utilization rate, allocate orders according to the real-time load, avoid single-machine overload, take ∏Dk as an index, preferentially allocate high-complexity orders to idle machines, reduce the workpiece scheduling steps within the order, reduce the need for manual intervention, and shorten the manufacturing cycle.

[0096] The present invention reduces the equipment idle time through a multi-machine cooperation strategy, reduces the overall energy consumption, accurately schedules to shorten the workpiece waiting time, reduces the material loss rate during the processing, and at the same time, in the scenario of random fluctuations in processing time, the manufacturing cycle volatility of the MPRVNS algorithm is significantly better than the traditional method, supports large-scale workpiece scheduling scenarios, and the calculation time increases linearly, meeting the requirements of the intelligent manufacturing system.

[0097] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An optimization method for multi-stage hybrid flow shop scheduling based on improved MPRVNS, characterized in that The described scheduling optimization method includes: S1: Initialize the input parameters of the algorithm; The input parameters include: the working stage m of the workshop, the number q of machines in each working stage, the number U of orders to be processed, the processing time p i , the order removal time s, and the machine startup time t0 = 0; S2: Calculate the complexity of each order; The complexity calculation formula for the said order: S3: Dynamically allocate orders to the machines with lighter loads in each stage; According to the quantified order complexity calculated in S2, allocate the orders to the machines with lighter loads in each stage dynamically. The same order is fixed to be processed by the machine with the same number in all stages; S4: Use the improved MPRVNS algorithm to optimize the manufacturing cycle of the orders; The specific optimization steps of the improved MPRVNS algorithm are as follows: S41: Adopt random key encoding, map order allocation and sorting into a real number vector, generate a random key vector, and then generate a feasible scheduling plan through decoding; S42: Select parent individuals for parametric crossover, retain the allocation strategy of high-complexity orders, and randomly adjust the machine allocation of low-complexity orders; S43: Design a dedicated neighborhood structure for different stages, combine local search to accelerate convergence, and recalculate the manufacturing cycle; The calculation formula of the manufacturing cycle is:

2. The multi-stage hybrid flow shop scheduling optimization method based on improved MPRVNS according to claim 1, wherein The processing time p in S1 i Adopt a dynamic processing time model: Dynamically associate the actual processing time of the workpiece with the machine running time, and at the same time embed the order removal time s as a fixed constraint.

3. The multi-stage hybrid flow shop scheduling optimization method based on the improved MPRVNS according to claim 1, characterized in that, The workpiece sorting and order scheduling rules in S3 are as follows: (1) There is no constraint on workpieces within an order: The exchange of workpieces within the same order does not affect the total manufacturing cycle. Therefore, a random order is adopted for processing to reduce the computational complexity; (2) Non-increasing sorting between orders: Arrange the orders in non-increasing order according to to minimize the critical path delay by giving priority to processing orders with high processing complexity.

4. The multi-stage hybrid flow shop scheduling optimization method based on improved MPRVNS according to claim 1, characterized in that In S42, the initial solution is input into the designed neighborhood structure for searching to obtain the global optimal solution; The global optimal solution is the optimal manufacturing cycle when the order is completed.

5. The multi-stage hybrid flow shop scheduling optimization system based on improved MPRVNS is characterized in that, The described scheduling optimization system includes: An input parameter initialization module that initializes the input parameters of the algorithm; The input parameters include: the working stage m of the workshop, the number q of machines in each working stage, the number U of orders to be processed, and the processing time p i , the order removal time s, and the machine startup time t0 = 0; An order complexity calculation module that calculates the complexity of each order; Complexity calculation formula for the said order: An order dynamic allocation module that dynamically allocates orders to the machines with lighter loads in each stage; According to the quantified order complexity calculated by the order complexity calculation module, allocate the orders to the machines with lighter loads in each stage dynamically. The same order is fixed to be processed by the machine with the same number in all stages; A manufacturing cycle optimization module that uses the improved MPRVNS algorithm to optimize the manufacturing cycle of the orders; The specific optimization steps of the manufacturing cycle optimization module are as follows: S41: Adopt random key encoding, map order allocation and sorting into a real number vector, generate a random key vector, and then generate a feasible scheduling plan through decoding; S42: Select parent individuals for parametric crossover, retain the allocation strategy of high-complexity orders, and randomly adjust the machine allocation of low-complexity orders; S43: Design a dedicated neighborhood structure for different stages, combine local search to accelerate convergence, and recalculate the manufacturing cycle; The calculation formula of the manufacturing cycle is:

6. The multi-stage hybrid flow shop scheduling optimization system based on improved MPRVNS according to claim 5, characterized in that: The processing time p in the input parameter initialization module i Adopt a dynamic processing time model: Dynamically associate the actual processing time of the workpiece with the machine running time, and at the same time embed the order removal time s as a fixed constraint.

7. The multi-stage hybrid flow shop scheduling optimization system based on the improved MPRVNS according to claim 5, characterized in that: The workpiece sorting and order scheduling rules in the order dynamic allocation module are as follows: (1) There is no constraint on workpieces within an order: The exchange of workpieces within the same order does not affect the total manufacturing cycle. Therefore, a random order is adopted for processing to reduce the computational complexity; (2) Non-increasing sorting between orders: Sort orders in non-increasing order according to , and give priority to processing orders with high processing complexity to minimize the critical path delay.

8. The multi-stage hybrid flow shop scheduling optimization system based on improved MPRVNS according to claim 5, characterized in that: In S42, the initial solution is input into the designed neighborhood structure for searching to obtain the global optimal solution; The global optimal solution is the optimal manufacturing cycle when the order is completed.

Images