Workshop dynamic scheduling method considering equipment occupation under emergency order insertion

Abstract

The invention discloses a workshop dynamic scheduling method considering equipment occupation under emergency order insertion, which comprises the following steps: receiving a trigger event in a production system, and obtaining workshop production real-time state data according to the trigger event; according to the real-time state data, a dynamic scheduling decision engine is activated, and a global scheduling problem is decomposed into deterministic sub-problems at the current decision moment; according to the deterministic sub-problem, identifying the state of a work-in-process and generating an unfinished process set, and re-bringing the unfinished process set into a rearrangement resource pool; selecting a rearrangement strategy based on a preset influence threshold according to the unfinished process set; according to the rearrangement strategy, a planning model is constructed, and the planning model aims at minimizing the maximum completion time and minimizing the order insertion change degree; according to the planning model, an improved genetic simulated annealing algorithm is adopted for solving, and a dynamic scheduling scheme is output and issued to a production execution layer. According to the invention, the disturbance of order insertion on the original production system can be minimized.

Description

Technical Field

[0001] This invention belongs to the field of dynamic scheduling technology, and particularly relates to a workshop dynamic scheduling method that considers equipment occupancy under emergency order insertion. It is especially suitable for single-objective dynamic rescheduling that comprehensively considers the real-time equipment occupancy status and overall production efficiency under the disturbance of emergency order insertion. Background Technology

[0002] Against the backdrop of modern manufacturing's transformation towards intelligence and flexibility, production scheduling systems, as the core of operations management, directly determine an enterprise's production efficiency and market competitiveness through their dynamic response capabilities and optimization levels. Traditional static scheduling models often fail when faced with common workshop disturbances such as emergency order insertions and equipment failures due to a lack of real-time adjustment mechanisms, leading to a series of problems such as reduced equipment utilization, delayed order delivery, and excessively high production costs. Emergency order insertions, in particular, are highly dynamic and high-priority disturbances that require not only rapid integration into the existing production plan but also careful consideration of real-time equipment occupancy status to avoid resource conflicts and ensure overall production efficiency. Therefore, researching dynamic scheduling methods that consider equipment occupancy under emergency order insertions has significant theoretical value and is a pressing practical need.

[0003] As the manufacturing industry shifts towards a multi-variety, small-batch, customized model, the frequent insertion of urgent orders has become the norm in workshop production. These dynamic disturbances not only disrupt pre-defined static scheduling plans but also, due to the limited equipment resources within the workshop, trigger complex resource competition and conflicts. Traditional scheduling methods often lack the ability to perceive and dynamically adjust equipment occupancy status in real time, leading to problems such as delivery delays, low equipment utilization, and poor production stability when faced with order insertions. Therefore, various uncertainties and disturbances are prevalent within manufacturing systems, such as the insertion of urgent orders, sudden equipment failures, and process delays. This causes static scheduling schemes based on idealized assumptions to quickly deviate from expectations in actual implementation, resulting in production disorder, decreased resource utilization efficiency, and order delivery delays. Without effective dynamic response and optimization capabilities, the production system will struggle to adapt to the dynamic changes in the real environment. Therefore, a dynamic workshop scheduling method that considers equipment occupancy under the condition of urgent order insertions is urgently needed. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a dynamic workshop scheduling method that considers equipment occupancy during emergency order insertion. This method can minimize the disruption to the original production system caused by order insertion while ensuring production efficiency.

[0005] To achieve the above objectives, the present invention provides a workshop dynamic scheduling method considering equipment occupancy under emergency order insertion, comprising:

[0006] Receive trigger events from the production system and obtain real-time production status data of the workshop based on the trigger events;

[0007] Based on the real-time status data, the dynamic scheduling decision engine is activated to decompose the global scheduling problem into deterministic sub-problems at the current decision moment;

[0008] Based on the deterministic subproblem, identify the work-in-process status and generate a set of incomplete processes, then re-include the set of incomplete processes into the rearrangement resource pool;

[0009] Based on the set of incomplete processes, a rearrangement strategy is selected based on a preset impact threshold.

[0010] Based on the rearrangement strategy, a planning model is constructed, the planning model aiming to minimize the maximum completion time and the degree of change of insertion orders;

[0011] Based on the planning model, an improved genetic simulated annealing algorithm is used to solve the problem, outputting a dynamic scheduling scheme and sending it to the production execution layer.

[0012] Optionally, the triggering events include: order events, task events, processing resource events, and system events;

[0013] The order events include emergency orders arriving, order cancellations, or order changes;

[0014] The task events include task processing start, task completion, or task processing interruption;

[0015] The processing resource events include equipment idle, equipment processing, or equipment malfunction;

[0016] The system events are triggered at fixed time intervals.

[0017] Optionally, the real-time production status data in the workshop includes: equipment occupancy status data, task pool status data, and system performance status data;

[0018] The equipment occupancy status data includes the equipment's current processing task, consumed working hours, estimated completion time, and a waiting queue of tasks.

[0019] The task pool status data includes a set of tasks being processed and a set of tasks available for scheduling. The set of tasks being processed identifies the preemption status of each task, while the set of tasks available for scheduling includes newly arrived tasks and released regular tasks.

[0020] The system performance status data includes equipment utilization, work-in-process inventory, and system throughput.

[0021] Optionally, the preemption status identification of the set of tasks being processed includes:

[0022] Get the identifier of the task currently being processed;

[0023] Based on the task identifier, determine whether each task is in a preemptible state;

[0024] Based on the judgment results, a subset of preemptible tasks and a subset of non-preemptible tasks are generated.

[0025] Optionally, based on the deterministic subproblem, identifying the work-in-process status and generating a set of incomplete processes, and then re-incorporating the set of incomplete processes into the rearrangement resource pool includes:

[0026] Based on the aforementioned deterministic sub-problem, detect task interruption events caused by emergency order insertion or equipment failure;

[0027] Based on the task interruption event, extract the processed information and the incomplete information;

[0028] Based on the incomplete information, the remaining information is marked as an incomplete work order and its process constraints are retained;

[0029] Based on the incomplete work orders, determine whether they affect the order delivery date. If they do, the corresponding tasks will be included in the resource reordering pool.

[0030] Optionally, based on the set of incomplete processes, the reordering strategy selected according to a preset impact threshold includes:

[0031] Based on the set of incomplete processes, calculate the comprehensive impact value of emergency order insertion on the original plan. The comprehensive impact value is composed of a weighted average of the processing time variation and the fixture change time variation.

[0032] Based on the comparison between the comprehensive impact value and the preset impact threshold, a rescheduling strategy is selected: when the comprehensive impact value is not greater than the threshold, a right-shift rescheduling is performed; when the comprehensive impact value is greater than the threshold, the resource pool is entered and a full rescheduling is performed.

[0033] Optionally, constructing the planning model based on the rearrangement strategy includes:

[0034] Based on the rearrangement strategy, constraints and objective functions are established, and the planning model is obtained based on the constraints and objective functions.

[0035] The constraints include:

[0036] Process sequence constraints ensure that different processes for the same workpiece are processed in the order of the predetermined process path.

[0037] Equipment selection constraints ensure that only one available piece of equipment is selected for each process and each operation.

[0038] Equipment time constraints ensure that the same equipment processes only one operation at a time.

[0039] Process time constraints ensure that the completion time of a process equals the start time plus the processing time.

[0040] Optionally, based on the planning model, an improved genetic simulated annealing algorithm is used to solve the problem, outputting a dynamic scheduling scheme and distributing it to the production execution layer, including:

[0041] Based on the task priority hierarchical information of the planning model, generate hierarchical initial population, sequential initial population, and random initial population;

[0042] Based on the initial population, if the time has not exceeded the preset time, the crossover rate and mutation rate are adaptively adjusted according to the current population diversity index.

[0043] Based on the crossover rate and mutation rate, and combined with the problem characteristics, a crossover operator is selected, including hierarchical crossover or sequentially preserved crossover.

[0044] Based on the crossover rate and mutation rate, and combined with the problem characteristics, a mutation operator is selected, including hierarchical mutation or sequential sensing mutation.

[0045] Based on the crossover and mutation operators, simulated annealing local search is performed on elite individuals, and the optimal solution is accepted according to the Metropolis criterion to obtain the crossover and mutation success rate;

[0046] The crossover rate and mutation rate are adjusted based on the crossover and mutation success rate feedback until the termination condition is met. A dynamic scheduling scheme is then output and sent to the production execution layer.

[0047] Compared with the prior art, the present invention has the following advantages and technical effects:

[0048] The scheduling method proposed in this invention can produce significant comprehensive positive effects. Technically, this method effectively solves the scheduling optimization problem under dynamic disturbances such as emergency order insertion by integrating the global search of genetic algorithms with the local escape capability of simulated annealing mechanisms and introducing an optimization model oriented towards real-time equipment occupancy. Economically, it can directly improve equipment utilization, shorten delivery cycles, and reduce inventory and delay costs, thereby saving operating expenses and tapping into production capacity potential. In terms of management, this invention promotes the transformation of production decision-making from experience-driven to data-intelligent driven, enhancing enterprises' agility and resilience in responding to market changes. Overall, this invention provides a core engine for high-end manufacturing execution systems, possesses broad industrial application prospects, and is of significant value in promoting the intelligent upgrading of manufacturing, achieving cost reduction and efficiency improvement, and green and sustainable development. Attached Figure Description

[0049] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0050] Figure 1 This is a flowchart of a workshop dynamic scheduling method considering equipment occupancy under emergency order insertion according to an embodiment of the present invention;

[0051] Figure 2 This is a flowchart of work-in-process management in the emergency order insertion scenario of this invention.

[0052] Figure 3 This is a flowchart of the emergency order insertion and reordering strategy according to an embodiment of the present invention;

[0053] Figure 4 This is an overall flowchart of the improved genetic simulated annealing algorithm according to an embodiment of the present invention. Detailed Implementation

[0054] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0055] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0056] This embodiment proposes a dynamic workshop scheduling method that considers equipment occupancy under emergency order insertion, such as... Figure 1 As shown, the specific steps include:

[0057] Receive trigger events from the production system and obtain real-time production status data of the workshop based on the trigger events;

[0058] Based on real-time status data, the dynamic scheduling decision engine is activated to decompose the global scheduling problem into deterministic sub-problems at the current decision moment;

[0059] Based on the deterministic problem, identify the status of work-in-process and generate a set of incomplete processes, and then re-include the set of incomplete processes into the rearranged resource pool;

[0060] Based on the set of incomplete processes, a rearrangement strategy is selected based on a preset impact threshold.

[0061] Based on the rearrangement strategy, a planning model is constructed, with the goal of minimizing the maximum completion time and minimizing the degree of change in the insertion order.

[0062] Based on the planning model, an improved genetic simulated annealing algorithm is used to solve the problem, outputting a dynamic scheduling scheme and sending it to the production execution layer.

[0063] With the increasing demand for personalized products and services, manufacturing enterprises face frequent emergency order insertions and tight equipment resources, while traditional scheduling methods lack flexibility. This embodiment addresses the workshop scheduling problem under the disturbance of emergency order insertions by constructing a dynamic scheduling optimization method that considers equipment occupancy status to achieve a dual-objective optimization of minimizing the maximum completion time and the degree of order insertion changes. An integrated method combining a dynamic scheduling framework, mathematical model, and improved algorithm is proposed, constructing an event-driven dynamic scheduling mechanism, establishing a mixed integer programming model, and designing a multi-objective deep reinforcement learning algorithm (such as MO-D3QN). Multiple simulation experiments show that this method performs excellently in three test instances of different scales, reducing the disturbance of the original plan by order insertions while ensuring equipment utilization. It converges quickly and has good solution set distribution, effectively coping with the uncertainty of emergency order insertions, improving the real-time performance and robustness of workshop scheduling, and providing a feasible solution for intelligent manufacturing system scheduling.

[0064] To address the challenge of dynamic scheduling in workshops characterized by frequent emergency order insertions and limited equipment resources, this embodiment proposes a dynamic scheduling method integrating multi-objective optimization and deep reinforcement learning. By constructing a mixed-integer programming model that considers equipment occupancy status, and with the dual objectives of minimizing maximum completion time and the degree of order insertion changes, an improved algorithm based on MO-D3QN is designed. Intelligent decision-making is achieved using multi-dimensional state representation, a scheduling rule base, and a continuous reward mechanism. Experiments show that this method effectively balances equipment utilization and the impact of order disturbances, outperforming comparative algorithms in three scale test instances. It provides a real-time and robust solution for workshop scheduling under emergency order insertion conditions.

[0065] More specifically, the proposed dynamic scheduling framework employs a closed-loop control logic of "perception-decision-execution," the core of which lies in decomposing the complex global dynamic scheduling problem into a series of event-triggered local deterministic optimization problems. The overall structure of the framework is as follows: Figure 1As shown, the system mainly consists of three modules: the event perception and status monitoring layer, the dynamic scheduling decision engine, and the scheduling scheme execution and feedback layer. The event perception and status monitoring layer (top layer) acts as the system's "sensors," responsible for receiving various production events (such as emergency order insertions) and collecting real-time status data to provide a basis for decision-making. The dynamic scheduling decision engine (middle layer) acts as the system's "brain" and is the core of the framework. It first decomposes complex problems, then constructs mathematical models, and finally solves the problem through a hybrid adaptive solution strategy. This strategy innovatively combines: metaheuristic algorithms for deep search and optimization to ensure the quality of the solution; and multi-objective deep reinforcement learning for intelligent guidance, adaptively selecting the best scheduling rules based on the state. These two work together to generate high-quality scheduling schemes. The scheduling scheme execution and feedback layer (bottom layer) acts as the system's "execution and learning organ," responsible for executing scheduling schemes and feeding back the actual execution results to the learning mechanism, enabling the system to continuously optimize and form a virtuous cycle of "perception-decision-execution-learning." The knowledge base and rule base serve as a supporting platform, providing domain knowledge and scheduling rules for the algorithms in the decision engine, forming the foundation of system intelligence.

[0066] Furthermore, triggering events include: order events, task events, processing resource events, and system events;

[0067] Order events include urgent orders arriving, order cancellations, or order changes;

[0068] Task events include task processing start, task completion, or task processing interruption;

[0069] Processing resource events include equipment idle, equipment in operation, or equipment malfunction;

[0070] System events are triggered at fixed time intervals.

[0071] Specifically, this section serves as the interface between the system framework and the workshop environment, responsible for capturing various events related to drive scheduling and monitoring the system status in real time. The events triggered by emergency order insertions mainly fall into the following four categories:

[0072] (1) The order event model mainly includes three states: emergency order arrival, order cancellation or change, and the modeling is as follows:

[0073] ;

[0074] In the formula, EOID is the identification code for the emergency order; V represents the status of the emergency order, a1 indicates order arrival, a2 indicates order cancellation, a3 indicates order modification, and T1 is the status collection time. The data collection frequency is twice per second, and the collected data will be stored in a temporary database.

[0075] (2) Task events include three states: task processing start, normal completion, and processing interruption. The modeling is as follows:

[0076] ;

[0077] In the formula, TID represents the task identification code; W represents the task status, w1 is the start of processing, w2 is the normal completion, w3 is the interruption of processing, and T2 is the status acquisition time.

[0078] (3) Processing resource events mainly include three states: equipment idle, equipment processing, and equipment failure. The modeling is as follows:

[0079] ;

[0080] In the formula, MID represents the machine tool's identification code; S represents the machine tool's power on / off status; OID represents the identification code of the task being processed; P represents the processing information; B represents the machine tool's fault information; and T3 is the status acquisition time.

[0081] (4) System events: Rescheduling triggered based on a fixed time period.

[0082] ;

[0083] In the formula, SID represents the system rescheduling identification code. T is the rescheduling interval. The time of the last rescheduling trigger. This is the time when the rescheduling was triggered.

[0084] Furthermore, the real-time production status data in the workshop includes: equipment occupancy status data, task pool status data, and system performance status data;

[0085] Equipment occupancy status data includes the current processing task of the equipment, the time consumed, the estimated completion time, and the list of tasks in the waiting queue.

[0086] The task pool status data includes the set of tasks being processed and the set of tasks waiting to be scheduled. The set of tasks being processed identifies the preemption status of each task, while the set of tasks waiting to be scheduled contains newly arrived inserts and released regular tasks.

[0087] System performance status data includes equipment utilization, work-in-process inventory, and system throughput.

[0088] Specifically, establishing a real-time production status monitoring model is the cornerstone and prerequisite for the effective operation of the entire dynamic scheduling framework. Its core function is to transform the complex physical production system into digital dynamic information that can be understood and processed by the computing system. At each decision-making moment, the system captures the global state of the scheduling system, which mainly includes equipment occupancy status, task pool status, and system performance status.

[0089] The equipment occupancy status model can accurately describe the real-time status of each piece of equipment, including the current processing task, the time consumed, the estimated completion time, the waiting queue length, and the task list. It is the core of reflecting "equipment occupancy".

[0090] ;

[0091] In the formula, Assign a device occupancy status number. For the current processing task, For the time already spent, For the estimated completion time, This specifies the length of the waiting queue and the task list. This refers to the status acquisition time.

[0092] Establishing a task pool state model aims to provide a precise decision-making basis for dynamic scheduling. It structurally classifies tasks, accurately manages process dependencies and state transitions (such as preemption and release) between tasks, and ensures scheduling feasibility. Simultaneously, it provides core inputs for optimization algorithms and scheduling rules, serving as a key hub for shifting from passive response to globally intelligent decision-making.

[0093] The processing task set indicates whether each task is in a preemptible state.

[0094] ;

[0095] In the formula, The task number is the one currently being processed. In order to seize the initiative, This indicates that it can be seized. It cannot be seized.

[0096] The set of schedulable waiting tasks includes newly arrived inserts and released regular tasks.

[0097] ;

[0098] In the formula, This is the number of the schedulable waiting task. C represents the newly arrived insert, and C represents the released regular task.

[0099] Furthermore, the preemption state identification of the processing task set includes:

[0100] Get the identifier of the task currently being processed;

[0101] Based on the task identifier, determine whether each task is in a preemptible state;

[0102] Based on the judgment results, a subset of preemptible tasks and a subset of non-preemptible tasks are generated.

[0103] Specifically, the emergency order insertion scheduling decision-making mechanism is the "brain" of the framework, activated after an event is triggered, responsible for generating new scheduling instructions. The dynamic scheduling decision-making mechanism proposed in this paper combines event triggering with periodic driving, performing rolling optimization and rescheduling based on the system's real-time operating status. Its core significance lies in transforming the production system from passively responding to disturbances to an optimized system with proactive adaptive capabilities, thereby continuously ensuring production efficiency, delivery time, and system stability in a constantly changing environment, effectively avoiding operational disorder and profit loss caused by rigid planning. As a key intelligent decision-making hub connecting "perception" and "execution," this mechanism constructs a closed-loop control structure of "perception → decision → execution → feedback," enabling the system to continuously learn and self-evolve, ultimately achieving a profound transformation towards an intelligent production system.

[0104] Problem decomposition and subproblem generation:

[0105] Drawing on the approximation concept from stochastic dynamic programming, the original problem is approximately decomposed into a deterministic scheduling subproblem under the current state. The decision-making scope of this subproblem is limited to: newly arrived urgent order insertion tasks, the set of waiting tasks that currently satisfy the preceding constraints, and the set of work-in-process tasks on non-critical paths. The research focuses on work-in-process management, urgent order insertion processing logic, large-volume order splitting strategies, and parallel process scheduling in dynamic scheduling scenarios, which greatly reduces the problem size and improves decision-making efficiency.

[0106] Work-in-progress management in emergency order scenarios:

[0107] like Figure 2 As shown, the system needs to support changes in the status of work-in-process due to disturbances such as emergency order insertions and equipment failures during the production process. When an order is interrupted, the system needs to identify the processed and unfinished parts, and reschedule the set of unfinished processes as the "unfinished work orders" set. If the original order is interrupted, its remaining processes will be marked and retained, and subsequent scheduling can be based on this set. If the interruption affects the order delivery date, the system needs to include the affected tasks in the rescheduling scope. If it does not affect the delivery date, the original plan can be retained.

[0108] Furthermore, based on the deterministic sub-problem, the work-in-process status is identified and a set of incomplete processes is generated. This set of incomplete processes is then reintegrated into the rearranged resource pool, including:

[0109] Based on the nature of the problem, detect task interruption events caused by emergency order insertion or equipment failure;

[0110] Based on the task interruption event, extract the processed and unfinished information;

[0111] Based on the incomplete information, mark the remaining information as incomplete work orders and retain their process constraints;

[0112] Based on the incomplete work orders, determine whether they affect the order delivery date. If they do, add the corresponding tasks to the resource pool for rearrangement.

[0113] Furthermore, based on the set of incomplete processes, the reordering strategy selected according to a preset impact threshold includes:

[0114] Based on the set of incomplete processes, calculate the comprehensive impact value of emergency orders on the original plan. The comprehensive impact value is a weighted average of the changes in processing time and the changes in fixture change time.

[0115] Based on the comparison between the comprehensive impact value and the preset impact threshold, a rearrangement strategy is selected: when the comprehensive impact value is not greater than the threshold, a right-shift rescheduling is performed; when the comprehensive impact value is greater than the threshold, the resource pool is rearranged and a full rescheduling is performed.

[0116] Specifically, the strategies for emergency order insertion and reordering are as follows:

[0117] like Figure 3 As shown, urgent orders have the highest priority, and all their processes will be assigned the highest priority and scheduled for production first. However, rescheduling urgent orders will lead to the replacement of production line fixtures, increasing production support time and extending the production cycle. Therefore, when urgent orders conflict with the processes of the original orders, the scheduling system handles them according to the following two strategies: Strategy 1: Without considering the cost of changing production and the stability of the existing plan, the affected processes are directly rescheduled into available equipment resources, which is suitable for urgent orders that are sensitive to delivery dates; Strategy 2: Under the premise of ensuring the stability of the original production line, the system minimizes the interference with the original plan and only reschedules when necessary. An impact threshold for urgent orders is set. If the threshold is not exceeded, a right-shift rescheduling method is used to ensure production continuity; if the threshold is exceeded, a full rescheduling method is used. The threshold σ is derived from the following, considering both processing time and changeover costs. This scheme is suitable for scenarios with high requirements for production line continuity. In addition, the system needs to record the interference nodes of the urgent orders on the original tasks, calculate the start time of the affected processes and shift them backward. If the shift exceeds the delivery date, a rescheduling is triggered.

[0118] ;

[0119] In the formula, N k The number of product types processed by manufacturing unit k; N iT represents the number of operations contained in part i; ｉk A represents the processing time of the i-th product in manufacturing unit k after an emergency order is placed; ｉk C represents the planned processing time for the i-th product in manufacturing unit k; ｉk W represents the fixture changeover time for the i-th product in manufacturing unit k after an emergency order. ｉk Let be the planned fixture changeover time for the i-th product in manufacturing unit k.

[0120] Furthermore, based on the rearrangement strategy, the planning model is constructed as follows:

[0121] Based on the rearrangement strategy, constraints and objective functions are established, and a planning model is obtained based on the constraints and objective functions.

[0122] The constraints include:

[0123] Process sequence constraints ensure that different processes for the same workpiece are processed in the order of the predetermined process path.

[0124] Equipment selection constraints ensure that only one available piece of equipment is selected for each process and each operation.

[0125] Equipment time constraints ensure that the same equipment processes only one operation at a time.

[0126] Process time constraints ensure that the completion time of a process equals the start time plus the processing time.

[0127] Specifically, the problem description and mathematical model are as follows:

[0128] The workshop scheduling problem considering equipment occupancy under emergency order insertion can be described as follows: At time 0, there are m parts {J0, J1, J2, ... J...} m} on n devices {M0, M1, M2, ..., M n The part i is awaiting processing and contains l processes {O}. i0 O i1 O i2 , ..., O il The affected workpiece processes and machine tools can only be put back into production after the anomaly is resolved. Optimization is performed using maximum completion time and carbon emissions as the objective function.

[0129] ;

[0130] In the formula, n ｉjk Let f_i be the quantity of the j-th process required to be processed in manufacturing unit k for the i-th product; jk t represents the defect rate of the j-th process of the i-th product in manufacturing unit k; ｉjk d represents the unit processing time required for the j-th process of the i-th product in manufacturing unit k;ｉjk The unit production preparation time required for the j-th process of the i-th product in manufacturing unit k; C k (t) represents the change in fixture change time of manufacturing unit k caused by real-time changes in the working conditions of the manufacturing workshop;

[0131] After a disturbance event occurs in the workshop, and following rescheduling, each workpiece must still be processed sequentially according to the given processing route and its own processing requirements. Specific requirements are as follows:

[0132] a. There is a strict processing sequence for different processes of the same workpiece, that is, the next process can only be processed after the previous process of the same workpiece is completed.

[0133] ;

[0134] b. The process itself must meet processing constraints, that is, each process can only be processed by one piece of equipment at a time.

[0135] ;

[0136] c. The time constraints between each process on the processing equipment must be met, that is, the same equipment can only process one process at a time;

[0137] ;

[0138] d. The process itself must meet time constraints, that is, the completion time of the process is the sum of the start time and the completion time of the process.

[0139] ;

[0140] In the formula, i, j, and k represent the workpiece number, process number, and equipment number, respectively, and M... ij For process O ij Optional device set, θ ijk For decision variables 0 / 1, if process O ij When processing is performed on device Mk, the decision variable is 1; otherwise, it is 0. CT ij ST ij and PT ij Representing process O ij The processing end time, start time, and processing time of CT ijk and ST ijk These represent process O respectively. ij In device M k The processing end time and start time.

[0141] Furthermore, based on the planning model, an improved genetic simulated annealing algorithm is used to solve the problem, outputting a dynamic scheduling scheme which is then distributed to the production execution layer, including:

[0142] Based on the task priority hierarchical information of the planning model, generate hierarchical initial population, sequential initial population, and random initial population;

[0143] Based on the initial population, if the time has not exceeded the preset time, the crossover rate and mutation rate are adaptively adjusted according to the current population diversity index.

[0144] Based on the crossover rate and mutation rate, and combined with the characteristics of the problem, the crossover operator is selected. The crossover operator includes hierarchical crossover or sequentially preserved crossover.

[0145] Based on the crossover rate and mutation rate, and combined with the characteristics of the problem, the mutation operator is selected. The mutation operators include hierarchical mutation or sequential sensing mutation.

[0146] Among them, the adaptive crossover rate / mutation rate is the "macro-policy controller": it answers "whether" and "what probability" to perform crossover or mutation operations on the population.

[0147] For example, when the algorithm gets stuck in a local optimum, the adaptive mechanism reduces the crossover rate (reduces mixing with similar parents) while increasing the mutation rate (increases random perturbations to escape the trap). This is a macroscopic, search-state-based strategic adjustment.

[0148] Crossover / mutation operators are "micro-tactical executors": they answer the question of "how to specifically perform" crossover or mutation. Algorithms may select the most suitable operator from a library of operators such as hierarchical crossover, sequential crossover, and traditional two-point crossover, based on the characteristics of the problem, to actually generate offspring.

[0149] Relationship: Adaptive parameters determine the extent to which these operators are triggered, while the operators themselves determine the specific behavioral effects after triggering. The former is like "fine-tuning the accelerator and steering wheel," while the latter is like "choosing which specific path to take."

[0150] Based on the crossover and mutation operators, simulated annealing local search is performed on elite individuals, and the optimal solution is accepted according to the Metropolis criterion to obtain the crossover and mutation success rate;

[0151] Among them, the crossover and mutation operators (exploration and innovation engine):

[0152] Target: Ordinary individuals (non-elites) in the main population.

[0153] Objective: To be responsible for global exploration and maintaining population diversity. Through recombination (crossover) and perturbation (mutation), to explore new and unknown regions in the solution space and discover new potential solutions.

[0154] Role: This is the part of the algorithm that dares to "take risks," responsible for finding new possibilities.

[0155] Elite Individuals (The Cornerstone of Development and Quality Assurance):

[0156] Subject: One or several individuals with the highest fitness selected from the current population.

[0157] Objective: To ensure convergence and preserve historical best results. These results do not participate in crossover or mutation; they are directly "protected" and replicated to the next generation.

[0158] Role: This is the "conservative" and "robust" part of the algorithm, ensuring that the search process does not regress.

[0159] Elite individuals serve as "immune objects" and "result containers" for crossover and mutation operations. Operators are responsible for mining new ore (new solutions) in the "veins" (solution space), while elite individuals are responsible for storing the "largest nugget" (historical best solution) that has been found and ensuring that it is not accidentally destroyed or discarded by subsequent mining activities (crossover and mutation).

[0160] The crossover and mutation rates are adjusted based on the success rate of crossover and mutation until the termination condition is met. A dynamic scheduling scheme is then output and sent to the production execution layer.

[0161] Specifically, this embodiment proposes an improved genetic algorithm that integrates simulated annealing and multiple adaptive strategies, aiming to efficiently solve the dynamic scheduling problem of workshops under emergency order insertion. This algorithm significantly enhances global exploration capability and local development depth through priority-level initialization, dynamic parameter adjustment, and simulated annealing local search. Its core process is as follows, and the overall framework is as follows: Figure 4 As shown.

[0162] Step 1. Algorithm Initialization:

[0163] 1. Parameter settings: Read basic parameters such as population size `N`, maximum number of generations `G_max`, initial crossover rate `P_c0`, and initial mutation rate `P_m0`.

[0164] 2. Priority stratification: If there are predefined sorting rules for the problem, the tasks are prioritized according to the rules to guide the generation of the initial population.

[0165] 3. Population Initialization: A hybrid initialization strategy is employed to balance population quality and diversity. This includes:

[0166] Hierarchical initialization: Construct high-quality initial solutions based on task priority hierarchies.

[0167] Sequential initialization: Generate regular sequences using random or heuristic methods.

[0168] Diverse initialization: Introducing completely random individuals ensures that the solution space is fully explored.

[0169] Step 2. Genetic main cycle:

[0170] While (if `G_max` is not reached and timeout has not occurred) do;

[0171] 1. Fitness Assessment: For each individual in the population, feasible scheduling schemes are generated using forward scheduling, backward scheduling, or a hybrid scheduling strategy, and its fitness value `Fitness` is calculated. The fitness function comprehensively considers multiple optimization objectives such as weighted delay, equipment utilization, and replacement cost.

[0172] 2. Elite Record: Compare and update the current population with the historical global optimal solution.

[0173] 3. Diversity Monitoring and Parameter Adaptation: Calculate the diversity index of the current population and dynamically adjust the crossover rate `P_c` and mutation rate `P_m` accordingly. When diversity decreases, appropriately increase `P_m` to introduce new genes.

[0174] 4. Selection operation: The tournament selection method is adopted, and the superior parent individuals are selected from the current population based on fitness.

[0175] 5. Crossover operation: Adaptively select the crossover operator based on whether the problem has priority hierarchy:

[0176] If a hierarchical structure exists, a hierarchical crossover is used to preserve the superior hierarchical structure of the parent generation.

[0177] If it is a normal sequence, then the order is maintained by crossing over to inherit the relative processing order of the parent generation.

[0178] 6. Mutation operation: Similarly, the mutation operator is adaptively selected according to the characteristics of the problem, including hierarchical mutation, sequential sensing mutation or traditional mutation, in order to maintain population diversity.

[0179] 7. Population Update and Elite Enhancement:

[0180] Implement an elite retention strategy, directly preserving the best individuals of the present generation for the next.

[0181] For the retained elite individuals, simulated annealing algorithm is introduced to perform local fine search in order to further improve the quality of the solution and enhance the local exploitation capability of the algorithm.

[0182] 8. Parameter feedback adjustment: Based on the success rate of crossover and mutation operations in this round, fine-tune `P_c` and `P_m` to achieve adaptive optimization of parameters.

[0183] 9. Premature convergence handling: If the fitness of the population stagnates for several consecutive generations, it is determined to be premature convergence. In this case, the population size is dynamically expanded and the maximum number of generations is increased to escape the local optimum and restart the global exploration.

[0184] Step 3. Output the results:

[0185] After the algorithm terminates, the recorded global optimal solution is output as the final scheduling scheme.

[0186] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for dynamic scheduling of a job shop with emergency insertions considering equipment occupation, characterized in that, The method comprises the following steps: receiving a trigger event in a production system, and obtaining workshop production real-time state data according to the trigger event; activating a dynamic scheduling decision engine according to the real-time state data, and decomposing a global scheduling problem into a deterministic sub-problem at a current decision time; identifying a work-in-process state and generating an unfinished process set according to the deterministic sub-problem, and re-introducing the unfinished process set into a rearrangement resource pool; selecting a rearrangement strategy based on a preset influence threshold according to the unfinished process set; constructing a planning model according to the rearrangement strategy, the planning model aiming to minimize the maximum completion time and minimize the degree of order change; solving the planning model by using an improved genetic simulated annealing algorithm, outputting a dynamic scheduling scheme, and delivering the dynamic scheduling scheme to a production execution layer.

2. The method according to claim 1, wherein, The trigger event includes an order event, a task event, a processing resource event, and a system event; The order event includes arrival of an emergency order, order cancellation, or order change; The task event includes task processing start, task normal completion, or task processing interruption; The processing resource event includes device idle, device processing, or device failure; The system event is triggered based on a fixed time interval.

3. The method of claim 2, wherein the method further comprises: The workshop production real-time state data includes device occupation state data, task pool state data, and system performance state data; The device occupation state data includes a current processing task of a device, consumed working hours, a predicted completion time, and a waiting queue task list; The task pool state data includes a processing task set and a schedulable waiting task set, wherein the processing task set identifies a preemption state of each task, and the schedulable waiting task set includes a newly arrived order and a released regular task; The system performance state data includes device utilization, work-in-process quantity, and system throughput.

4. The method of claim 3, wherein the method further comprises: The preemption state identification of the processing task set includes: obtaining a task identifier of a task currently being processed; judging whether each task is in a pre-emptible state according to the task identifier; generating a pre-emptible task subset and a non-pre-emptible task subset according to the judgment result.

5. The method of claim 1, wherein, The identification of the work-in-process state and the generation of the unfinished process set according to the deterministic sub-problem, and the re-introduction of the unfinished process set into the rearrangement resource pool include: detecting a task interruption event caused by an emergency order or device failure according to the deterministic sub-problem; extracting processed information and unfinished information according to the task interruption event; labeling remaining information as an unfinished work order and retaining process constraints thereof according to the unfinished information; judging whether the unfinished work order affects an order delivery period, and introducing a corresponding task into the rearrangement resource pool if the unfinished work order affects the order delivery period.

6. The method of claim 1, wherein, The selection of the rearrangement strategy based on the preset influence threshold according to the unfinished process set includes: calculating a comprehensive influence value of an emergency order on an original plan according to the unfinished process set, the comprehensive influence value being composed of a processing time variation and a jig replacement time variation; selecting a rearrangement strategy according to a comparison result of the comprehensive influence value and a preset influence threshold: when the comprehensive influence value is not greater than the threshold, right-shift rescheduling is performed; and when the comprehensive influence value is greater than the threshold, a complete rescheduling is performed in the rearrangement resource pool.

7. The method of claim 1, wherein, According to the rearrangement strategy, the planning model is constructed, which comprises: According to the rearrangement strategy, the constraint conditions and the objective function are established, and the planning model is obtained based on the constraint conditions and the objective function; The constraint conditions comprise: Process sequence constraint, ensuring that different processes of the same workpiece are processed in a predetermined process path sequence; Device selection constraint, ensuring that only one available device is selected for each process at each time; Device time constraint, ensuring that only one process is processed on the same device at the same time; Process time constraint, ensuring that the completion time of the process is equal to the start time plus the processing time.

8. The method of claim 1, wherein, According to the planning model, an improved genetic simulated annealing algorithm is used for solving, and a dynamic scheduling scheme is output and sent to the production execution layer, which comprises: According to the task priority layering information of the planning model, a layered initial population, a sequential initial population and a random initial population are generated; Based on the initial population, when the time does not exceed the preset time, the diversity index of the current population is used to adaptively adjust the crossover rate and the mutation rate; According to the crossover rate and the mutation rate, a crossover operator is selected according to the characteristics of the problem, which comprises layered crossover or sequence preserving crossover; According to the crossover rate and the mutation rate, a mutation operator is selected according to the characteristics of the problem, which comprises layered mutation or sequence-aware mutation; Based on the crossover operator and the mutation operator, the simulated annealing local search is performed on the elite individuals, the optimized solution is accepted according to the Metropolis criterion, the crossover mutation success rate is obtained; The crossover rate and the mutation rate are adjusted according to the crossover mutation success rate feedback until the termination condition is met, and the dynamic scheduling scheme is output and sent to the production execution layer.

Images