A method and system for generating workshop scheduling scheme based on local search

Abstract

The present invention belongs to the field of workshop scheduling technology and specifically discloses a workshop scheduling solution generation method and system based on local search, which includes: performing AL operations on a processing process sequence to generate an initial workshop scheduling Gantt chart; cyclically performing IL operations and IR operations on the initial workshop scheduling Gantt chart until a steady state is reached, thereby obtaining an IL Gantt chart and an IR Gantt chart; transforming the IL Gantt chart or the IR Gantt chart to obtain a domain solution, and judging whether the domain solution is feasible based on process constraints to obtain several feasible solutions; determining the head and tail lengths based on the transformed IL Gantt chart and the IR Gantt chart, performing an approximate evaluation of feasible solutions based on the head length and tail lengths, screening out the current solution, and performing the next round of optimization until the optimal solution is obtained as the workshop scheduling solution. The present invention simultaneously utilizes data under the constraints of both processes, thereby improving local search capabilities, significantly reducing unnecessary idle time, and obtaining a more optimal workshop scheduling Gantt chart.

Description

Technical Field

[0001] The present invention belongs to the technical field of workshop scheduling, and more specifically, relates to a method and system for generating a workshop scheduling plan based on local search. Background Art

[0002] Scheduling is a form of resource allocation. Shop scheduling involves rationally allocating resources within a production workshop to achieve or optimize one or more production metrics, thereby reducing production costs and improving enterprise profitability. To meet the demands of actual production, shop scheduling problems incorporate a wide range of production metrics, such as maximum completion time, energy consumption, lead times / tardiness, and their composites. However, the most common and often the most critical metric is maximum completion time. This is because maximum completion time directly determines production costs and is positively correlated with many other metrics, such as machine energy consumption, employee work hours, and tardiness. These metrics, which are positively correlated with maximum completion time, are known as regularity metrics. Furthermore, in the study of shop scheduling problems, where the optimization objective is to minimize maximum completion time, numerous properties of the problem have been discovered, resulting in the development of numerous sophisticated designs. These insights have provided valuable insights for the study of other metrics, offering valuable insights.

[0003] The job shop scheduling problem, with the optimization objective of minimizing the maximum completion time, is generally described as follows: there are n workpieces, each with m processes, and the required machines and processing time for each process are known. The task is to arrange the processing sequence and processing time of the processes on each machine so that the last process to be completed on all machines has the earliest completion time.

[0004] The job shop scheduling problem, whose optimization objective is to minimize the maximum completion time, is an NP-hard problem. For large-scale problems, finding an optimal solution within a finite time is difficult. Therefore, scholars often use approximate algorithms to quickly find a better solution to meet actual production needs. The most widely used approximate algorithms in enterprises are scheduling rules, including FCFS, SPT, and EDD. These rules are not only extremely fast and generate stable solutions, but also mathematically interpretable, making them popular among enterprises.

[0005] The solutions generated by these rules are sequences of processing steps. For practical use, these sequences need to be converted into usable shop scheduling Gantt charts. The same sequence will generate different shop scheduling Gantt charts under different shop scheduling conditions. Existing shop scheduling types include semi-active scheduling, active scheduling, and full-active scheduling (Zhang Chaoyong, Guan Zailin, Liu Qiong, Shao Xinyu, Li Peigen. A new scheduling type and its application in job shop scheduling [J]. Journal of Mechanical Engineering, 2008(10):24-31). The problem with these scheduling types is that the quality of the generated shop scheduling Gantt charts is low, and for large-scale and complex problems, the shop scheduling Gantt charts contain a lot of idle time. Summary of the Invention

[0006] In response to the above-mentioned defects or improvement needs of the prior art, the present invention provides a method and system for generating a workshop scheduling plan based on local search, the purpose of which is to significantly reduce unnecessary idle time and obtain a better workshop scheduling Gantt chart.

[0007] To achieve the above-mentioned object, according to a first aspect of the present invention, a method for generating a workshop scheduling plan based on local search is proposed, comprising the following steps:

[0008] S1. Perform AL operations on the processing sequence to generate the initial workshop scheduling Gantt chart;

[0009] S2. Perform IL and IR operations on the initial shop scheduling Gantt chart cycle to adjust the critical path until a steady state is reached, thereby obtaining an IL Gantt chart and an IR Gantt chart;

[0010] S3. Transform the IL Gantt chart or the IR Gantt chart to obtain multiple domain solutions, and determine whether the domain solutions are feasible based on process constraints to obtain several feasible solutions;

[0011] S4. Determine the head length and tail length based on the transformed IL Gantt chart and IR Gantt chart, perform an approximate evaluation of feasible solutions based on the head length and tail length, and select the feasible solution with the shortest maximum completion time as the current solution;

[0012] S5. Use the current solution as the initial workshop scheduling Gantt chart and repeat steps S2 to S4. When the maximum completion time corresponding to the current solution does not change over several generations, stop the loop, and then use the solution with the smallest maximum completion time among all generations of solutions as the final solution, and use this final solution as the workshop scheduling plan.

[0013] As a further preferred embodiment, the AL operation is as follows: each process starts processing as early as possible without changing the processing order of the processes.

[0014] As a further preference, the IL operation is: each process starts processing as early as possible without delaying other processes; the IR operation is: each process starts processing as late as possible without advancing other processes.

[0015] As a further preferred embodiment, in step S3, the specific method for determining the domain solution is:

[0016] The IL Gantt chart or IR Gantt chart is transformed. During the transformation, the basic neighborhood action is to move a single process forward or backward. The process whose sum of head length and tail length is less than the maximum completion time is not moved. Only the processes on the critical path are moved, so as to obtain the domain solution.

[0017] As a further preferred embodiment, the head length and the tail length are determined based on the IL Gantt chart and the IR Gantt chart after the transformation operation, specifically as follows:

[0018] For a certain process of a certain workpiece, the head length refers to the start time of the process in the IL Gantt chart, and the tail length refers to the maximum completion time of all processes in the IR Gantt chart minus the start time of the process.

[0019] As a further preferred embodiment, the critical path refers to a process sequence in a Gantt chart that determines the maximum completion time.

[0020] According to a second aspect of the present invention, a system for generating a workshop scheduling plan based on local search is provided, comprising a processor configured to execute the above-mentioned method for generating a workshop scheduling plan based on local search.

[0021] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for generating a workshop scheduling solution based on local search is implemented.

[0022] In general, the above technical solutions conceived by the present invention have the following technical advantages compared with the existing technology:

[0023] 1. This invention proposes a new type of shop scheduling. Compared with existing shop scheduling types, this invention uses cyclic IL and IR operations to destroy the critical path as much as possible, fully utilizes the idle time in the shop scheduling Gantt chart, significantly reduces unnecessary idle time, and realizes the rapid and accurate acquisition of the Gantt chart with the shortest total processing time from a process sequence.

[0024] 2. The present invention generates two Gantt charts with the same total processing time but different process processing sequences through IL and IR operations, and the data of these two Gantt charts can be simultaneously applied in the neighborhood structure. Using this method, more neighborhood solutions generated by changes in a single process can be obtained, thereby improving the search effect.

[0025] 3. The present invention performs two optimization searches on the neighborhood solutions: first, the neighborhood solutions whose neighborhood actions are processes where the sum of the moving head length and tail length is less than the maximum completion time are deleted, thereby improving the neighborhood search efficiency; then, the remaining neighborhood solutions are approximately evaluated using the new head and tail length definitions, and the evaluation results are smaller than those under the original definitions. This essentially takes into account the utilization of idle time in the next step of adjustment, and is more conducive to finding neighborhood solutions that are truly conducive to continuous improvement.

[0026] 4. For large-scale workshop scheduling problems (generally referring to problems with a total number of processes exceeding 500), the idle time in the workshop scheduling Gantt chart corresponding to a certain process sequence under semi-active scheduling, active scheduling, and full-active scheduling is often relatively large. This method can be used to continuously optimize the problem multiple times, thereby quickly obtaining a better solution without iteration, saving iterative search time, and thus ensuring the efficiency and effectiveness of the solution.

[0027] 5. The present invention is applicable to various workshop scheduling problems, including job shop scheduling problems, flow shop scheduling problems, open shop scheduling problems, etc., and is a method with universal application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a structural diagram of a method for generating a workshop scheduling solution based on local search according to an embodiment of the present invention;

[0029] Figure 2 It is the relationship diagram between various workshop scheduling and optimal scheduling;

[0030] Figure 3 A semi-active scheduling Gantt chart constructed for an embodiment of the present invention;

[0031] Figure 4 An activity scheduling Gantt chart constructed for an embodiment of the present invention;

[0032] Figure 5 A full activity scheduling tail length Gantt chart constructed for an embodiment of the present invention;

[0033] Figure 6 A Gantt chart of the full activity scheduling head length constructed for an embodiment of the present invention;

[0034] Figure 7 A Gantt chart of fully active scheduling header lengths constructed for an embodiment of the present invention;

[0035] Figure 8 A tail length Gantt chart of a complete activity schedule constructed for an embodiment of the present invention;

[0036] Figure 9 A Gantt chart of IR tail length for a complete activity schedule constructed for an embodiment of the present invention;

[0037] Figure 10 A Gantt chart of IR header length for a fully active schedule constructed for an embodiment of the present invention;

[0038] Figure 11 A Gantt chart of IL header length for a complete activity schedule constructed for an embodiment of the present invention;

[0039] Figure 12 A Gantt chart of IL tail length for a complete activity schedule constructed for an embodiment of the present invention;

[0040] Figure 13 This is a schematic diagram of an approximate evaluation of a process shift according to an embodiment of the present invention;

[0041] Figure 14 A schematic diagram of an approximate evaluation of process advancement constructed for an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0043] To facilitate understanding of the present invention, the following terms are first defined.

[0044] Gantt chart: specifies the processing start time and processing completion time of each process.

[0045] Critical path: The sequence of operations in the Gantt chart that determines the maximum completion time.

[0046] AL (Arrange from left) operation: Each process starts processing as early as possible without changing the processing order of the processes.

[0047] AR (Arrange from right) operation: Each process starts processing as late as possible without changing the processing order of the processes.

[0048] AL and AR operations cannot change the maximum completion time, but can obtain semi-active scheduling and assist in calculating the head and tail lengths.

[0049] IL (Insert to the left) operation: Each process starts processing as early as possible without delaying other processes.

[0050] IR (Insert to the left) operation: Each process starts processing as late as possible without advancing other processes.

[0051] IL and IR operations may reduce the maximum completion time, but cannot increase it.

[0052] Steady State: The Shop Scheduling Gantt chart can no longer reduce the maximum completion time through IL and IR.

[0053] Steady-state judgment conditions: the maximum completion time does not decrease after executing IL (or IR) for a feasible solution, and the maximum completion time still does not decrease after continuing to execute IR (or IL).

[0054] Based on this, the various types of workshop scheduling are analyzed as follows:

[0055] Semi-active scheduling: AL operations are performed on the processing sequence.

[0056] Activity Scheduling: Execute IL operations on a sequence of processing operations.

[0057] Full activity scheduling: Execute IL+IR+AL operations on the processing sequence.

[0058] Fully active scheduling: IL and IR operations are performed cyclically on the processing operation sequence until the shop scheduling Gantt chart enters a steady state.

[0059] The embodiment of the present invention provides a method for generating a workshop scheduling plan based on local search, such as Figure 1 As shown, the following steps are included:

[0060] S1. Obtain the processing sequence from the scheduling rules, and perform AL operations on the processing sequence to generate the initial workshop scheduling Gantt chart.

[0061] S2. Perform IL and IR operations on the initial workshop scheduling Gantt chart in turn to adjust the critical path, and continuously generate new and better workshop scheduling Gantt charts until a steady state is reached, and two Gantt charts with equal maximum completion times are obtained. The last one obtained by the IL operation is the IL Gantt chart, and the one obtained by the IR operation is the IR Gantt chart.

[0062] Specifically, semi-active scheduling and active scheduling can only produce one Gantt chart, while full-active scheduling can produce two Gantt charts for process processing sequences, but unconditionally discards the IL Gantt chart and retains the IR Gantt chart. The full-active scheduling used in the present invention can produce two Gantt charts with equal maximum completion times and different process processing sequences. The present invention redefines the head and tail lengths based on the IL Gantt chart and the IR Gantt chart. Compared to defining the head and tail lengths separately on a Gantt chart, under the new definition of the present invention, the head and tail lengths of processes on non-critical paths remain unchanged or are smaller, as follows:

[0063] Head Length: The classic definition of the head length of process u is the duration from time 0 to the start time of process u under AL operation. In this invention, the head length of process u is the duration from time 0 to the start time of process u under IL operation after reaching steady state; that is, the start time of the process in the IL Gantt chart.

[0064] Tail length: The classic definition of the tail length of process u is the duration from the start time of process u to the completion time of all processes under AR operation. In this invention, the tail length of process u is the duration from the start time of process u to the completion time of all processes under IR operation after reaching steady state; that is, the final completion time of all processes in the IR Gantt chart minus the start time of the process.

[0065] S3. In the local search, the basic neighborhood action is to move a single process forward or backward, and transform the IL Gantt chart or the IR Gantt chart (either one can be transformed) to obtain multiple domain solutions; and the feasibility of the domain solution is judged by the head and tail lengths under the constraints of the processing order of the two processes. Because the head and tail lengths are smaller under the new definition, more neighborhood solutions can be judged as feasible solutions.

[0066] Furthermore, when transforming the Gantt chart, the processes whose sum of head length and tail length is less than the maximum completion time are not moved, but only the processes on the critical path are moved to improve the movement efficiency.

[0067] S4. Based on the transformed IL Gantt chart and IR Gantt chart, the head and tail lengths are re-determined. The quality of the feasible solutions is approximately evaluated based on the new head and tail lengths, and the most promising feasible solution is selected as the current solution. This evaluation method takes into account the use of idle time during the next adjustment, which helps to find the truly optimal neighborhood solution.

[0068] Specifically, when redetermining the head and tail lengths, if the IL Gantt chart was previously transformed, the head length is determined by the transformed IL Gantt chart, and the tail length is determined by the untransformed IR Gantt chart; if the IR Gantt chart was transformed, the head length is determined by the untransformed IL Gantt chart, and the tail length is determined by the transformed IR Gantt chart.

[0069] S5. Use the current solution as the initial workshop scheduling Gantt chart and repeat steps S2 to S4. When the maximum completion time corresponding to the current solution does not change over several generations, stop the loop, and then use the solution with the smallest maximum completion time among all generations of solutions as the final solution, and use this final solution as the workshop scheduling plan.

[0070] To make the present invention more clear, the present invention is described below with reference to two groups of specific embodiments.

[0071] Example 1

[0072] Figure 2This paper demonstrates the relationship between existing shop floor scheduling and optimal scheduling. Existing shop floor scheduling types include semi-active scheduling, active scheduling, fully active scheduling, and the fully active scheduling proposed in this invention. Active scheduling is always semi-active scheduling, fully active scheduling is always active scheduling, fully active scheduling is always fully active scheduling, and optimal scheduling is always fully active scheduling.

[0073] Figure 3 The following figure shows a Gantt chart for semi-active scheduling, generated by AL operations. In this chart, processes for the same workpiece are represented by the same color, and processes processed by the same machine are placed in the same row. Processes for the same workpiece cannot be processed simultaneously, and a machine cannot process multiple workpieces simultaneously. The maximum completion time for this Gantt chart is 60.

[0074] Figure 4 Shown Figure 3 The Gantt chart is generated by IL operation under activity scheduling. The maximum completion time of the new Gantt chart is 58, which is the same as Figure 3 Compared to the previous period, it has decreased by 2.

[0075] Figure 5 Shown Figure 4 The Gantt chart is generated by IR operation to generate the tail length Gantt chart under full activity scheduling. The maximum completion time of the new Gantt chart is 56, which is the same as Figure 4 Compared to the previous period, it has decreased by 2.

[0076] Figure 6 Shown Figure 5 The Gantt chart is generated by AL operation to generate the head length Gantt chart under full activity scheduling. The maximum completion time of the new Gantt chart is 56, which is the same as Figure 5 Compared to that, there is no change.

[0077] Figure 7 Shown Figure 5 or Figure 6 The Gantt chart is generated by IL operation. The maximum completion time of the new Gantt chart is 47, which is the same as Figure 5 or Figure 6 Compared to the previous period, it decreased by 9.

[0078] Figure 8 Shown Figure 7 The Gantt chart is generated by IL operation. The maximum completion time of the new Gantt chart is 47, which is the same as Figure 7 Compared to that, there is no change.

[0079] After two consecutive IL and IR operations, the maximum completion time does not change. Therefore, the workshop scheduling Gantt chart reaches a steady state. Figure 7 and Figure 8 Gantt chart for complete activity scheduling. Figure 7 Gantt chart for head length, Figure 8Gantt chart for tail length.

[0080] It can be seen that the present invention has a significant improvement over the traditional workshop scheduling type, and the quality of the solution can generally be improved.

[0081] Example 2

[0082] Figure 9 The IR tail length Gantt chart of a process sequence obtained through complete activity scheduling is shown below. Figure 10 Gantt chart for IR header length, Figure 11 Gantt chart for IL header length, Figure 12 This is the IL tail length Gantt chart. The process sequence constraints of the IL Gantt chart and the IR Gantt chart are different.

[0083] As can be seen from the figure, the IL header length is shorter than the IR header length, and the IL tail length is longer than the IR tail length. Therefore, compared to using only the IR header and tail lengths, or only the IL header and tail lengths, using both the IL header and tail lengths minimizes both the header and tail lengths. Neighborhood solution feasibility determination method: JP(u) represents the predecessor of process u, and JS(u) represents the successor of process u.

[0084] (1) If the tail length of JS(u) is not longer than the tail length of v, u can be moved to v and then processed;

[0085] (2) If the head length of JP(v) is not longer than the head length of u, v can be moved to the front of u for processing.

[0086] Because the head and tail lengths are smaller under the new definition, more neighborhood solutions can be determined as feasible solutions.

[0087] Figure 13 and Figure 14 To approximate the quality of the neighborhood solution, Figure 13 To move the process on the critical path backward, Figure 14 Move the process on the critical path forward.

[0088] (1) If u is moved to v for processing: Assume that there are processes o1, o2, ... o between u and v k :

[0089] 1) Calculate u, o1, o2, ... o in the neighborhood solution respectively k , the estimated tail length of v:

[0090] The tail length of u is estimated as follows: u's processing time + max{JS(u)'s tail length, MS(v)'s tail length},

[0091] o kTo o1, the estimated tail length of v is: the estimated tail length of the subsequent process of the machine + its own processing time;

[0092] 2) Calculate u, o1, o2, ... o in the neighborhood solution respectively k , the estimated length of v's header:

[0093] The estimated head length of o1 is: the head length of JP(o1) + the processing time of JP(o1),

[0094] o2, ...o k , the head length estimation of v,u is: max{processing time of the machine predecessor process + head length of the machine predecessor process, head length of JP(w) + processing time of JP(w)}, w is the current process;

[0095] 3) Maximum completion time estimation of the neighborhood solution: for u, o1, o2, ... o k , v, add up the head length and tail length estimates of each process, and take the largest one as the estimate of the neighborhood solution.

[0096] (2) If v is moved to the processing before u: Assume that there are processes o1, o2, ... o between u and v k :

[0097] 1) Calculate u, o1, o2, ... o respectively k , the estimated length of v's header:

[0098] The estimated length of v's header is: max{header length of JP(v),header length of MP(u)},

[0099] u, o1 to o k The estimated head length is: the estimated head length of the machine's leading process + the processing time of the machine's leading process;

[0100] 2) Calculate u, o1, o2, ... o respectively k , the estimated tail length of v:

[0101] o k The tail length of JS(o k )’s tail length,

[0102] o k-1 , ..., the head length of o1, v, u is estimated as: processing time of w + max{tail length of the subsequent process of the machine, tail length of JS(w)}, where w is the current process;

[0103] 3) Maximum completion time estimation of the neighborhood solution: for u, o1, o2, ... o k , v, add up the head length and tail length estimates of each process, and take the largest one as the estimate of the neighborhood solution.

[0104] In order to verify the practical application effect of the present invention, two simulation experiments are carried out below.

[0105] Experiment 1 used 80 examples from the DMU standard test case for job shop scheduling problems to conduct simulation tests under different job shop scheduling conditions. For each example, 100 randomly generated processing sequence sequences were used. Gantt charts were generated using semi-active scheduling, active scheduling, full-active scheduling, and the present invention, and the average maximum completion time was calculated. The results of semi-active scheduling for each example were then used as the denominator, and the data were normalized for easy comparison, as shown in Table 1.

[0106] From the above experiments, it can be seen that in all 80 examples of the Dmu set, the maximum completion time of the Gantt chart obtained by the present invention is smaller than the maximum completion time obtained by other shop scheduling types.

[0107] Experiment 2 uses the Dmu01-80 benchmark test case for the job shop scheduling problem under full-activity scheduling. In the experiment, each benchmark case is randomly decoded 100 times, and the number of neighborhood solutions under the new and old definitions of head and tail lengths is compared, as shown in Table 2.

[0108] From the above experiments, it can be seen that in all 80 examples of the Dmu set, the number of neighborhood solutions obtained by the present invention is greater than the number of neighborhood solutions under the original definition, and the neighborhood scale is significantly expanded.

[0109] Table 1 Results of Dmu01-80 example

[0110]

[0111] Table 2 New and old head and tail lengths

[0112]

[0113] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating a workshop scheduling plan based on local search, characterized in that: The steps include: S1. Perform AL operation on the processing sequence to generate the initial workshop scheduling Gantt chart; AL operation is: without changing the processing order of the processes, each process starts processing as early as possible; S2. Perform IL and IR operations on the initial shop scheduling Gantt chart to adjust the critical path until a steady state is reached, resulting in IL and IR Gantt charts. IL operations are: each process starts processing as early as possible without delaying other processes; IR operations are: each process starts processing as late as possible without advancing other processes. S3. Transform the IL Gantt chart or the IR Gantt chart. During the transformation, the basic neighborhood action is to move a single process forward or backward. The process whose sum of the head length and tail length is less than the maximum completion time is not moved. Only the processes on the critical path are moved. Thus, multiple neighborhood solutions are obtained. The feasibility of the neighborhood solutions is determined based on the process constraints, and several feasible solutions are obtained. S4. Determine the head length and tail length based on the transformed IL Gantt chart and IR Gantt chart, perform an approximate evaluation of feasible solutions based on the head length and tail length, and select the feasible solution with the shortest maximum completion time as the current solution; S5. Use the current solution as the initial workshop scheduling Gantt chart and repeat steps S2 to S4. When the maximum completion time corresponding to the current solution does not change over several generations, stop the loop, and then use the solution with the smallest maximum completion time among all generations of solutions as the final solution, and use this final solution as the workshop scheduling plan.

2. The method for generating a workshop scheduling plan based on local search according to claim 1, wherein: The head length and tail length are determined based on the IL Gantt chart and IR Gantt chart after the transformation operation, specifically: For a certain process of a certain workpiece, the head length refers to the start time of the process in the IL Gantt chart, and the tail length refers to the maximum completion time of all processes in the IR Gantt chart minus the start time of the process.

3. The method for generating a workshop scheduling plan based on local search according to claim 1 or 2, characterized in that: The critical path refers to the process sequence in the Gantt chart that determines the maximum completion time.

4. A workshop scheduling plan generation system based on local search, characterized in that: The method comprises a processor configured to execute the method for generating a workshop scheduling solution based on local search according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a workshop scheduling plan based on local search according to any one of claims 1 to 3 is implemented.

Images