Method and device for dynamic scheduling optimization for single-piece small-batch flexible manufacturing workshops

Abstract

The application relates to a dynamic scheduling optimization method and device for a single-piece small-batch flexible manufacturing workshop, and the method comprises the following steps: establishing a cellular automaton network model according to an actual processing route of a workpiece, and obtaining an initial state of each cell; according to initial processable order selection of each station, all possible first-step actions are traversed, and different first-step actions are executed: whether all orders are completed is judged, if yes, the process is ended, and a final scheduling solution is obtained; whether the production state is changed is judged, if yes, corresponding processing is executed based on the state change type, if not, the next step work is obtained according to the current situation of each station; the states of all cells are updated based on a cellular basic state evolution function; final scheduling solutions corresponding to all first-step actions are obtained, an optimal solution is selected, and the production process of the workshop is controlled based on the optimal solution. Compared with the prior art, the application has the advantages of good robustness, reduced influence of dynamic events on the workshop scheduling, and actual production demand of the factory.

Description

Technical Field

[0001] This invention relates to the field of manufacturing workshop scheduling, and in particular to a dynamic scheduling optimization method and apparatus for flexible manufacturing workshops with single-piece and small-batch production. Background Technology

[0002] With increasing market competition and product diversification, traditional production methods are no longer adequate for modern production demands, making multi-variety, small-batch production the mainstream approach. Simultaneously, flexible manufacturing models, distinct from traditional large-scale mass production, are continuously evolving. Existing research on manufacturing workshop scheduling methods largely stems from idealized, static production, solely pursuing the lowest cost and fastest speed. This disconnects production and logistics scheduling, neglecting dynamic and uncertain events in actual production, such as workshop malfunctions, machine failures, and urgent orders, which can cause significant disruptions. Consequently, even production plans with perfect data and optimal objectives cannot match actual production activities and ultimately cannot be implemented effectively. Furthermore, in actual project implementation, centralized scheduling proves difficult to consider multiple constraints such as order delays, just-in-time orders, and worker capabilities. Therefore, the resulting scheduling is not only difficult to respond to in real time but may even be unenforceable.

[0003] Faced with frequent production disruptions, scholars have largely considered centralized scheduling from both static and dynamic perspectives. Static scheduling often employs predictive scheduling and robust scheduling to improve the anti-interference capability of the scheduling scheme. However, this approach comes at the cost of some production efficiency, and in customized workshops with flexible scheduling and ever-changing environments, the real-time performance of scheduling and decision-making is not well addressed, resulting in limited application effectiveness. Dynamic scheduling often uses rescheduling methods, driven by events or cycles. However, these methods suffer from slow anomaly response, poor real-time decision-making, and low system stability. Especially for the scheduling problem in flexible manufacturing workshops with multi-variety, small-batch characteristics, the problem is more complex due to its multi-objective and multi-disturbance nature, making it difficult for algorithms to simultaneously satisfy both speed and optimality.

[0004] Chinese patent application CN113761732A discloses a method for modeling and optimizing flexible scheduling of a class of multi-disturbance workshops based on reinforcement learning, including the following steps: analyzing and inducing disturbance factors of multi-disturbance workshop production scheduling; abstracting the scheduling problem of a class of multi-disturbance workshops based on the idea of cellular automata modeling, inducing the characteristics and operation mechanism of the model abstraction, completing the establishment of a double-layer cellular automaton scheduling model, and constructing the double-layer cellular space of the cellular automaton scheduling model; optimizing the evolution rules of the cellular automaton scheduling model of the multi-disturbance workshop based on the idea of reinforcement learning algorithm; and finally establishing a simulation model system. The main scheduling goal of this method only considers minimizing the completion time of all workpieces and maximizing the average utilization rate of the same group of equipment. Although this method can achieve dynamic monitoring and good robustness, this method separates production and logistics scheduling, and is only suitable for manufacturing scenarios with single manufacturing unit function and fixed process route, and is not suitable for flexible workshop scheduling scenarios with multiple manufacturing unit functions.

[0005] Chinese patent application CN113570134A discloses a cellular automaton collaborative scheduling method for large equipment production and transportation system, which includes the following steps: constructing a grid model of production scheduling cellular automaton and setting evolution rules; optimizing the evolution rules of the production scheduling cellular automaton model; constructing a grid model of transportation scheduling cellular automaton according to the actual transportation operation status and setting evolution rules; optimizing the transportation scheduling evolution rules by means of genetic algorithm; and designing a large-scale equipment manufacturing and transportation collaborative scheduling simulation system interface. In the invention, the workpiece particles on the work station trigger the modification task after completion, and the transportation returns to the specified original position to wait for the next task when there is no other task instruction after completing the transportation, which is passive and leads to low work efficiency.

[0006] The document "Large Part Flexible Job Shop Scheduling Algorithm Based on Cellular Automaton and Improved GA" proposes a hybrid scheduling algorithm combining cellular automaton and improved genetic algorithm for large part flexible job shop scheduling problem. According to the optimization objectives of minimizing total processing time, high load rate of each station, and high load balance rate of each station in the same station group, an improved genetic algorithm optimization model for a single static scheduling unit after discretization is established, and the optimization process is illustrated with examples. However, the article uses periodic rescheduling in the design of the model rescheduling mechanism. Although periodic rescheduling can ensure a certain stability of production, it is not fast and sensitive enough to handle unexpected events.

[0007] The current prior art generally does not consider the complex processing route situation and the overall factory scheduling situation under the condition of multi-variety and small batch. If a dynamic scheduling method of multi-agent decentralized decision is considered, the total scheduling problem is subdivided to each production agent, and each agent autonomously selects a scheduling strategy according to its own operating characteristics, is responsive, and can optimally express the differences between different agents, but how to build a production multi-agent autonomous decision framework and utilize the loose coupling relationship between agents to collaboratively optimize the overall strategy is the difficulty of the scheduling method. In summary, the existing main technical difficulties of the dynamic scheduling decentralized decision method for single-piece and small batch flexible manufacturing workshops include: 1) multi-agent simulation model construction under complex factory environment; 2) construction of multi-agent dynamic scheduling decision relationship; 3) fast optimization of scheduling scheme. SUMMARY

[0008] The purpose of the present application is to overcome the defects of the prior art and provide a dynamic scheduling optimization method and device for single-piece and small batch flexible manufacturing workshops, which has good robustness, reduces the influence of dynamic events on workshop scheduling results, and meets the actual production needs of the factory.

[0009] The purpose of the present application can be achieved by the following technical solutions:

[0010] A dynamic scheduling optimization method for single-piece and small batch flexible manufacturing workshops, characterized in that it comprises the following steps:

[0011] According to the actual processing route of the workpiece, a cellular network model is established to obtain the initial state of each cell, and the cells of the cellular network model include buffer cells, workstation cells, AGV logistics vehicle cells and workpiece cells;

[0012] According to the initial processable order selection of each workstation, all possible first-step actions are traversed, and the following steps are performed with different first-step actions:

[0013] 1) Determine whether all orders are completed, if yes, end, obtain a scheduling final solution, if no, perform step 2);

[0014] 2) Determine whether the production state is changed, if yes, perform corresponding processing based on the state change type, if no, perform step 3), the state change type includes process interruption, emergency order addition and process order process completion;

[0015] 3) Each workstation obtains the next step work according to the current situation;

[0016] 4) Update the state of all cells based on the basic state evolution function of the cell, and return to step 1);

[0017] Obtain the scheduling final solution corresponding to all first-step actions, select the optimal solution, and control the workshop production process based on the optimal solution.

[0018] Further, the cellular automaton network model is a complex cellular automaton network with mobile particles constructed in the form of a reference Petri net, and each cellular process is a mobile particle.

[0019] Further, when the state change type is a process interruption, the corresponding processing includes:

[0020] redefining all related cell states and raising the priority of the unfinished order to the highest;

[0021] setting the current station state to failure;

[0022] When the state change type is an emergency order, the corresponding processing includes:

[0023] raising the priority of the emergency order to the highest;

[0024] When the state change type is that a processing order has completed a process, the corresponding processing is to start a logistics distribution process, specifically including:

[0025] performing behavior selection optimization of the AGV logistics vehicle based on the loading condition of the AGV logistics vehicle, the optimization target, and the pre-constructed behavior candidate set until the transportation of all workpieces is completed.

[0026] Further, in the logistics distribution process, the workshop transportation route of the AGV logistics vehicle involved in multi-order transportation is obtained based on genetic algorithm optimization.

[0027] Further, the next step of each station according to the current situation is to:

[0028] Each workpiece selects and optimizes behaviors through an optimization algorithm according to the optimization target and the pre-constructed behavior candidate set.

[0029] Further, the cellular basic state evolution function is used to define the rule of change of the cellular state over time, which is determined based on the current cellular state, related cellular state, and decision behavior.

[0030] Further, the cellular state is constructed based on various parameters of each type of cell under different processes, behaviors, and times.

[0031] Further, the buffer cell includes raw material storage, station line edge inventory cell, workshop storage cell, and finished product storage cell.

[0032] Further, a periodic order polling method is used to determine whether the production state changes.

[0033] The application also provides a dynamic scheduling optimization device for a single-piece small-batch flexible manufacturing workshop, comprising:

[0034] A dynamic monitoring module is used for collecting real-time state data of the workshop.

[0035] A dynamic scheduling optimization module is used for controlling the production process of the workshop according to the real-time state data of the workshop by using the dynamic scheduling optimization method as described above.

[0036] Compared with the prior art, the application has the following beneficial effects:

[0037] 1) The application defines a complex cellular network with mobile particles for a large flexible manufacturing workshop, which can effectively abstract and model the complex processing environment, and defines AGV logistics vehicles used for logistics as logistics cells, so that the logistics scheduling scheme can be considered at the same time as the production scheduling optimization, and the overall scheduling optimization of the workshop production and logistics is realized.

[0038] 2) The application uses a multi-agent decentralized decision method for workshop scheduling, which subdivides the total scheduling problem to each production agent, and each production agent autonomously selects a scheduling strategy according to its own operating characteristics, which is responsive and can better express the differences between different agents.

[0039] 3) The application enables the production scheduling decision to be made quickly when facing a variable production environment, effectively solves the influence of dynamic events on the workshop scheduling result, has good robustness, and has very strong resistance to external factors.

[0040] 4) The application considers multiple evolution rules and can select the optimal rule according to the actual factory to meet the actual production needs of the factory. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 It is a schematic diagram of the overall architecture of the application.

[0042] Figure 2 It is a flowchart of the design steps of the dynamic scheduling optimization module of the application.

[0043] Figure 3 It is a schematic diagram of the logical relationship between cells in the dynamic scheduling optimization module of the application.

[0044] Figure 4 It is a flowchart of the overall behavior selection of the dynamic scheduling optimization module of the application.

[0045] Figure 5 It is a flowchart of different processing processes for different emergency situations in Step 202 of the application.

[0046] Figure 6A flow chart is selected for the logistics distribution behavior of the present application.

[0047] Figure 7 A Gantt chart is selected for the work station of the embodiment of the present application. DETAILED DESCRIPTION

[0048] The present application will be described in detail below in combination with the drawings and specific embodiments. The embodiments are implemented on the premise of the technical solution of the present application, and detailed implementation and specific operation processes are given, but the protection scope of the present application is not limited to the following embodiments.

[0049] The embodiment provides a dynamic scheduling optimization method for a single-piece small-batch flexible manufacturing workshop, and comprises the following steps:

[0050] According to the actual processing route of the workpiece, a cellular network model is established, and the initial state of each cell is obtained, wherein the cells of the cellular network model comprise buffer cells, work station cells, AGV logistics vehicle cells and workpiece cells;

[0051] According to the initial processable order selection of each work station, all possible first-step actions are traversed, and the following steps are executed with different first-step actions: 1) judging whether all orders are completed, if yes, ending, obtaining a scheduling final solution, if not, executing step 2); 2) judging whether the production state is changed in a periodic order polling mode, if yes, executing corresponding processing based on the state change type, if not, executing step 3), wherein the state change type comprises process interruption, emergency order joining and process order process completion; 3) each work station obtains the next step work according to the current situation; 4) updating the state of all cells based on the basic state evolution function of the cell, and returning to step 1);

[0052] The scheduling final solution corresponding to all first-step actions is obtained, the optimal solution is selected, and the production process of the workshop is controlled based on the optimal solution.

[0053] In the above method, the cellular network model is a complex cellular network with mobile particles constructed in reference to the Petri net form. The complex cellular network models the work stations and transportation tools (such as AGV) in the workshop as “cells”, and models the orders to be processed as a kind of “mobile particles”, so that the orders processed by the work stations and transportation tools can be simulated through the particles contained in the cells.

[0054] In the method, when the state change type is a process interruption, the corresponding processing includes redefining all related cell states, and raising the priority of an unfinished order to the highest; and setting the current station state as a fault. When the state change type is an emergency order joining, the corresponding processing includes raising the priority of the emergency order to the highest. When the state change type is that a processing order has a process completion, the corresponding processing is to start a logistics distribution process, and specifically includes behavior selection optimization of an AGV logistics vehicle based on loading conditions of the AGV logistics vehicle, an optimization target and a pre-constructed behavior candidate set, until transportation of all workpieces is completed.

[0055] In the logistics distribution process of the method, a workshop transportation route of an AGV logistics vehicle involved in multi-order transportation is obtained based on a genetic algorithm optimization.

[0056] In the method, the next step work of each station is obtained according to current conditions, and specifically includes behavior selection optimization of each workpiece based on an optimization target and a pre-constructed behavior candidate set through an optimization algorithm.

[0057] In the method, a cell basic state evolution function is used to define a rule of change of the cell state over time, and the cell basic state evolution function is determined based on a current cell state, related cell states and a decision behavior. The cell state is obtained based on a plurality of parameters of each type of cell under different processes, behaviors and times.

[0058] In the method, the buffer cell includes a raw material warehouse, a station line edge inventory cell, a workshop storage cell and a finished product warehouse.

[0059] If the method is implemented in the form of a software function unit and sold or used as an independent product, the method can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the present application or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various storage medium that can store program codes.

[0060] Based on the dynamic scheduling optimization method, the embodiment further provides an optimization scheduling device for a single-piece small-batch flexible manufacturing workshop, as shown in Figure 1As shown, the device includes a dynamic scheduling optimization module A and a dynamic monitoring module B, which are connected to the factory workshop system.

[0061] 1) The dynamic scheduling optimization module includes a warehouse module A1, a manufacturing module A2, a transportation module A3, a periodic order polling module A4, and a global observer A5.

[0062] The warehouse module A1 includes four types of warehouses: raw material warehouse A1-1, workstation line edge inventory A1-2, workshop storage A1-3, and finished product warehouse A1-4. The raw material warehouse is mainly responsible for controlling the entry of orders, making decisions such as entering the factory for processing for orders with tight delivery dates, delaying processing for orders with later delivery dates, etc.; the workstation line edge inventory is used to store orders that can be processed by the current workstation and orders that have been processed but have not been sent away by the AGV logistics vehicle; the workshop storage is used to store orders that cannot be stored in the line edge inventory or orders that can be processed by the line edge workstation; the finished product warehouse is used to store finished product orders. The four types of warehouses mentioned above have a certain storage space, so the storage capacity and cost of the workshop need to be considered.

[0063] Each cell in the manufacturing module A2 corresponds to each workstation in actual processing, mainly completes various processing tasks of orders, and has a high autonomous selection capability. A workstation includes one or more devices and related auxiliary tools and operators. Due to the fact that the production process needs to consider the process skills of personnel, the changeover time of processing centers with different processing capabilities, and the bias of processing personnel towards the same type of order processing, which are easily ignored factors in optimization calculation. In this device, the optimization indicators such as production capacity utilization rate and order delay rate are considered, as well as the above-mentioned production constraints, so that the results obtained are suitable for actual production needs.

[0064] The transportation module A3 is a logistics transportation tool, responsible for selecting and transporting completed orders in the workshop to complete the flow of workpieces. The transportation module A3 includes multiple AGV logistics vehicles, each AGV logistics vehicle has a fixed carrying volume and can transport multiple orders at a time, so it involves the problem of path planning of the trolley and the autonomous decision problem of whether to transport the completed order.

[0065] The periodic order polling module A4 filters all orders that have not entered the workshop for processing according to the actual needs of the factory at a certain time, selects orders that meet the processing requirements according to the raw material warehouse situation, workshop work-in-process inventory situation, and order delivery date urgency, and enters the workshop processing flow.

[0066] The global observer A5 is equivalent to the overall management of the workshop, which organizes all hardware availability, real-time position of the trolley, processing state of all orders, storage state of the storage unit, and remaining available processing time of the workstation according to the sensor signals collected by the dynamic monitoring module B.

[0067] 2) Dynamic monitoring module B includes sensor module B1, UWB positioning module B2 and production real-time management module B3.

[0068] Sensor module B1 is used for monitoring the state of the manufacturing station and AGV logistics vehicle, generally including temperature, humidity and vibration sensors, and the availability of the station and logistics vehicle is determined by the monitoring signal.

[0069] UWB positioning module B2 is used for positioning the work-in-process, product and AGV logistics vehicle, including positioning base station and positioning tag. There is one positioning tag on each AGV logistics vehicle, and one positioning tag on the tray of each large or small part. The tag transmits pulses at a certain frequency and continuously measures the distance with the base station of known position, thereby determining the position of each object.

[0070] Production real-time management module B3 is used for real-time feedback of worker processing and temporary conditions of the workshop.

[0071] The dynamic scheduling optimization module in the device realizes the dynamic scheduling decentralized decision optimization of the single-piece small-batch flexible manufacturing workshop by using the above dynamic scheduling optimization method.

[0072] The above dynamic scheduling optimization method is realized based on a pre-established cellular automaton network model. The construction and evolution rule design process of the cellular automaton network model is as shown in Figure 2 , including the following steps:

[0073] Step 101: Establishing a cellular automaton network model

[0074] For the complex customized order production situation of the workshop, the structure of the workshop scheduling optimization module needs to be designed. This embodiment refers to the form of Petri net to design a complex cellular automaton network with mobile particles. According to the definition of the actual processing workpiece process route, combined with each module in Figure 1 , the cellular logic relationship as shown in Figure 3 is formed.

[0075] In Figure 3The first order of the bid marked No. 1 is taken as an example, the starting processing station corresponds to "station cell C", the process route is "station C-> station B-> station D", and the final processing is completed, so in the complex cell network, for the case of order No. 1, station B is the downstream machine of station C; relatively, station C is the upstream machine of station B. The order first meets the condition that it can be processed at the current time, is sent out from "raw material warehouse A1-1", and then enters the corresponding first process line side warehouse C for waiting; after being processed by station cell C, it is re-placed in the corresponding line side warehouse for temporary storage, and if the line side warehouse is full, it is temporarily stored in "workshop storage A1-3"; the AGV logistics vehicle polls in the workshop, and under the condition that the transportation capacity is not exceeded, according to the part position given by the UWB positioning system B2 and the order part state given by the B3 production management system, it autonomously selects the appropriate order part to transport to the downstream station, and the AGV logistics vehicle can support one vehicle multiple order path planning to select the appropriate route; in this way, all processing tasks of order No. 1 are completed, and finally it is sent to "finished product warehouse A1-4".

[0076] It should be noted that according to the different product process routes in different orders, the logical relationship between the cells is also different. A workshop processes multiple orders at the same time, so multiple logical relationships are formed in the "dynamic scheduling optimization module A" at the same time, such as Figure 3 Order No. 2 (denoted by 2) in the above represents another kind of logical relationship.

[0077] Step 102: Set the cell basic state evolution function

[0078] The workshop scheduling optimization module is composed of various types of cell machines in each sub-module, and the state and evolution rules of different cells are different due to their different functions, but the overall implementation logic is as shown in Figure 3 , which is based on the cell state function , wherein represents the state of cell i at time t, represents the state of all cell groups N related to cell i at time t. The evolution function defines the rule of change of cell state with time, contains the current state and related cell state and decision behavior, and describes the relationship between the cell and other cells in the cell network designed in step 1.

[0079] 1) The state of any buffer cell (station line side warehouse cell) Y1 at time t+1

[0080]

[0081] Wherein f is the local state transition rule, i.e. the job scheduling rule; the state of the buffer cell at time t+1 is determined by the buffer cell All the work cell upstream of t and its corresponding buffer cell The state set at t and all the work cell downstream of t and its corresponding buffer cell The state set at t determines.

[0082] 2) The state of any work cell Y2 at t+1 is determined by:

[0083]

[0084] Work cell The state at t+1 Buffer cell of its upstream process The state at t Buffer cell of its downstream process The state at t Work cell and its buffer cell The state at t determines.

[0085] 3) The state of any AGV logistics cell Y3 at t+1 is determined by:

[0086]

[0087] The state of AGV logistics cell Y3 at t+1 The state at t of the buffer cell set in the workshop The state at t and the state of Y3 at t determines.

[0088] Step 103: Set the basic state function of the cell

[0089] The state function of each type of cell should be determined according to its service object and function, and a variety of parameters that can represent the state of a certain type of cell at different processing, behavior, and time should be described. The essence is a mathematical description of the actual processing elements corresponding to the cell, which is a definition made after analyzing its functional elements, i.e. the definition of each in Step 102.

[0090] 1) The state of buffer cell Y1 at t can be represented as:

[0091]

[0092] The total capacity of the work-in-process space of the buffer cell, static attribute; occupied capacity in the buffer cell, dynamic attribute, unoccupied capacity in the buffer cell, dynamic attribute, queue length, dynamic attribute, indicating the number of workpieces waiting for processing in the buffer cell.

[0093] 2) The state of the workstation cell Y2 at time t can be represented as:

[0094]

[0095] T is the total time available for processing in a day, i.e., the shift; s s is the workstation busy state, dynamic attribute, s s ∈{0, 1, 2}, 0 idle, 1 busy, 2 fault; s co The total processing capacity of the workstation that has been occupied, i.e., the load of the workstation, dynamic attribute, equal to the processed and processing workpieces, corresponding to the workpieces waiting for processing in the queue of the buffer cell, the sum of the processing capacity values required by the three parts of the parts.

[0096] s cl The remaining processing capacity of the workstation cell, dynamic attribute, s cl = T - s co .

[0097] 3) The state of the AGV logistics vehicle cell Y3 at time t can be represented as:

[0098]

[0099] v is the running speed of the AGV logistics vehicle cell; s as is the AGV logistics vehicle busy state, dynamic attribute, s as ∈{0, 1, 2}, 0 idle, 1 busy, 2 fault; s an is the current position of the AGV logistics vehicle; s as is the position to be reached by the next action of the AGV logistics vehicle; s av is the carrying volume possessed by the AGV logistics vehicle.

[0100] 4) The state of the workpiece particle Y4 at time t can be represented as:

[0101]

[0102] p t is the total number of processes required by the particle, which is a static attribute; p fThe number of processes that the particle has completed, dynamic attribute; np is the next process number of the particle, dynamic attribute; s n The space required by the particle, static attribute, corresponding to the buffer area The s of the AGV logistics vehicle cell av ; dp is the workpiece processing priority, static attribute, determined by the delivery period of the workpiece and the importance of the order user, the closer the delivery period, the higher the priority; endt is the delivery time of the particle, static attribute; t a The time to reach the cell, divided into the time to reach the buffer cell, i.e. the time to start queuing, and the time to reach the workstation cell, i.e. the time to start processing, dynamic attribute.

[0103] Step 104: Determine the behavior of the cellular machine

[0104] This step is to define the behavior according to the current state, and multiple heuristic methods can be defined as different actions of the cellular machine for workstation order production and AGV logistics vehicle transportation. Without loss of generality, the following methods are determined in this embodiment:

[0105] The set of behaviors of the workstation cell machine:

[0106] (1) Behavior a1, select parts for processing according to first come first served (FCFS);

[0107] (2) Behavior a2, select parts for processing according to shortest processing time first (SPF);

[0108] (3) Behavior a3, select parts for processing according to longest processing time first (LPF);

[0109] (4) Behavior a4, select parts for processing according to minimum delivery period (EPF);

[0110] (5) Behavior a5, select parts for processing according to priority (PF);

[0111] (6) Behavior a6, do not select any job processing.

[0112] The set of behaviors of the AGV logistics vehicle:

[0113] (1) Behavior a1, select parts according to first come first served (FCFS);

[0114] (2) Behavior a2, select parts according to shortest delivery time first (SPF);

[0115] (3) Behavior a3, select parts according to longest delivery time first (LPF);

[0116] (4) Behavior a4, select parts according to minimum delivery period (EPF);

[0117] (5) Behavior a5: Select parts based on priority (PF);

[0118] (6) Behavior a6: Do not select any parts delivery.

[0119] Step 105: Determine the self-organizing evolution rules of the model

[0120] This step defines how to select the behavior and the evolution rules of the model, which together determine the evolution function f described in Step 102.

[0121] 1.5.1 Setting Model Evolution Rules

[0122] Based on the characteristics of dynamic scheduling in flexible work workshops, the self-organizing evolution rules of this model are summarized into five rules, as shown in Table 1.

[0123] Table 1 Model Evolution Rules

[0124]

[0125] The evolutionary process specifically includes:

[0126] Order Selection: Order Selection Rule R sb→bs The raw material storage cellular machine selects the part particles to be processed based on their priority, minimum processing time, and delivery time. However, since the factory will assign dedicated logistics personnel for such situations, logistics considerations are not required.

[0127] Workpiece selection: Workpiece selection rule R bs→s The next processing order for a workstation is determined by combining workstation cells with a greedy algorithm and the processing priority of the workpieces. First, the workpieces are sorted according to their priority, with higher priority workpieces listed first. Then, the order is processed by t... a Sort, t a Larger items are processed first; finally, based on the current processing status of the workstation, if the selected order can be inserted into the current production schedule to increase the workstation's capacity, it will be moved to the first position in the queue for processing; otherwise, the first workpiece in the current queue will be selected for processing according to the FCFS (First Come, First Served) principle.

[0128] Transportation Selection: Transportation Particle Selection Rule R bs→a The transport particle is selected based on the FCFS (First-Come, First-Served) principle, firstly according to t a Sort, t a Larger items are listed first; then the priorities of the workpieces are sorted, for example, t. aThe same priority is arranged in front; Finally, the order time is sorted according to the time.

[0129] Processing task trigger: processing task trigger rule R ls Two conditions need to be met: 1) target station cell idle s s =0; 2) the workpiece to be processed is located at the front of the queue, and when the conditions are met, the station enters the target station, the target station is busy s s =1, the length of the buffer cell queue lq=lq-1, the start processing time is generated, and t q is updated at the same time. After processing, the workpiece is placed in the current process and is placed in the online inventory to wait for AGV logistics vehicle transportation.

[0130] Transportation task trigger: transportation task trigger rule R la Three conditions need to be met: 1) target AGV logistics vehicle idle s os =0; 2) the remaining transportation space of the target AGV logistics vehicle is greater than the space required by the workpiece; 3) the workpiece to be transported is located at the front of the queue, and when the conditions are met, the AGV logistics vehicle starts transportation, and if the AGV logistics vehicle has no transportation space, it is placed in busy s as =1, the remaining transportation space needs to be reduced by the space occupied by the current transportation order, and the workpiece is transported to the next processing station line of the workpiece particle, and the particle t a , line lq=lq+1; if the processing is completed, the workpiece is transported to the finished product warehouse.

[0131] 1.5.2 Cell behavior selection

[0132] In terms of behavior selection, various optimization objectives can be selected for optimization according to actual problems, such as linear programming, genetic algorithm, particle swarm algorithm, etc. This patent takes order waiting time and station replacement number as optimization objectives as an example, and introduces a kind of example of using greedy algorithm for algorithm solving. This example considers two optimization objectives, i.e. order waiting time and minimum station replacement number. The two objectives are described as follows:

[0133] (1) Order waiting time

[0134] This objective considers the work-in-process inventory situation and measures the part waiting time.

[0135] First, define the state function μ k (t) for part state, i.e.

[0136]

[0137] r k represents the order waiting time at the kth decision moment:

[0138]

[0139] t k With t k-1 k and k-1 decision time, r k Indicates the shape of the part processed at time t k The length of time waiting for processing of parts i.

[0140] (2) Minimize the number of station change

[0141] This objective is to consider the processing workload of workers, measured by the number of station change.

[0142] First define the change state θ k (t) of the station cell before and after the two decisions, that is

[0143]

[0144] Whether to change the decision made at time t and time t-1 leads to the shape of the processed parts, if not, change is needed.

[0145] r k Indicates the total number of changes at the kth decision time:

[0146]

[0147] Where n represents the number of parts, t k With t k-1 k and k-1 decision time, r k Indicates the shape of the part processed at time t k Total number of changes.

[0148] After the construction of the cellular automaton network model and its evolution rules, the dynamic scheduling optimization module can realize the decision optimization of dynamic scheduling according to the real-time sensing signal, and the overall process is as Figure 4 shown, including the following steps:

[0149] Step 201: According to the actual situation of the factory, write the initial state of the cellular automaton, select according to the initial processable order of each station, and traverse all possible first-step actions.

[0150] Step 202: Determine whether all orders in the system are completed, if there are orders not completed, determine whether the current state is changed; State change has three cases of process interruption caused by sudden reasons such as station failure or worker leave Step 31, emergency order joining Step 32 and process order process completion Step 33, as Figure 5 shown.

[0151] Process interruption Step 31 specifically includes:

[0152] Step 311: Inform the system of the interruption reason, re-define all relevant cell states, and raise the priority of the unfinished order to the highest level. When the work station can process normally, the order is processed first.

[0153] Step 312: Set the current work station state to failure, and suspend processing of all orders in the line side warehouse.

[0154] The emergency order joining Step 32 specifically includes:

[0155] Step 321: Send the workpiece to the corresponding first process work station line side area and place it in the highest position of the waiting queue.

[0156] As shown in Figure 6 , the processing order has a process completion Step 33, which specifically includes:

[0157] Step 331: Enter the logistics distribution process;

[0158] Step 3311: Determine whether the logistics trolley has transportation space. If not, the workpiece remains in the current state and enters the waiting state.

[0159] Step 3312: If the order processing is completed and waiting for transportation in the line side warehouse, select the order for transportation according to the greedy strategy and the current loading condition.

[0160] Step 3313: For multi-order transportation, the selection of the workshop transportation route is involved. This patent uses genetic algorithm for route optimization.

[0161] Step 3314: Determine whether there are untransported orders in the current workshop. If there are, enter Step 2-3011. If not, enter Step 3315.

[0162] Step 3315: The trolley enters the waiting state.

[0163] Step 203: According to the selection result of Step 202, different emergency situations are handled.

[0164] Step 204: Each work station performs the next step of work according to the current situation and the behavior selection optimization. In this step, there are various optimization strategies to choose from, such as genetic algorithm, reinforcement learning, etc. In this patent, only the greedy algorithm is used as an example of a method of behavior selection.

[0165] Step 205: Advance the simulation to t+1 time, determine the system state St+1 according to the action selection result, and update all cell states in the system.

[0166] Step 206: According to the final solution caused by different first-step selection, select an optimal solution.

[0167] The dynamic scheduling optimization method and device for the single-piece small-batch flexible manufacturing workshop can realize optimization of dynamic scheduling decentralized decision-making. The hardware module dynamically learns about real-time environmental changes under the condition of real-time monitoring of the factory, and the software module analyzes the real-time data of the factory to obtain a dynamic scheduling result. The software module implementation includes the following steps: 1) constructing a complex cell body network model for a single-piece small-batch flexible manufacturing workshop; 2) establishing basic state evolution functions and basic state functions of each cell in combination with actual factory needs; and 3) determining cell behavior and evolution methods. Ultimately, the production scheduling decision can be quickly made when facing a variable production environment, with good robustness and dynamic scheduling.

[0168] Embodiment

[0169] Taking 8 batches of orders of 8 stations and 1 AGV logistics vehicle of an oil cylinder manufacturer as an example, the feasibility and effectiveness of the above algorithm are verified. Table 2 is the parameter index of the order, and Table 3 is the parameter index of each station.

[0170] Table 2 Processing parameter index of each order

[0171]

[0172]

[0173] Table 3 Parameter index of each station

[0174] Station 1 Station 2 Station 3 Station 4 Station 5 Station 6 Station 7 Station 8 Station 1 0 1 2 3 4 5 6 7 Station 2 0 1 2 3 4 5 6 Station 3 0 1 2 3 4 5 Station 4 0 1 2 3 4 Station 5 0 1 2 3 Station 6 0 1 2 Station 7 0 1 Station 8 0

[0175] Table 4 Shift table of each station

[0176] Station 1 Station 2 Station 3 Station 4 Station 5 Station 6 Station 7 Station 8 Shift 1 3 2 3 2 3 2 2

[0177] Based on the construction of the state attributes of each station cell according to Step 101-Step 105, a table label is written as shown in Table 5. Then, the dynamic scheduling optimization is realized based on the following steps:

[0178] S201: First, generate the initial state of each cell unit according to the production order. As shown in the example, the state table should be as follows, where T represents the total processing time of each station cell per day, s s represents the current cell state, s co represents the current cell processing time, s cl represents the order sequence that has been processed or is being processed by the current cell, and Cs represents the buffer queue of the current cell corresponding to the buffer cell, i.e., the order sequence that can be processed.

[0179] Table 5 Initial state attributes of each cell machine

[0180]

[0181] S202: For the initialization state selection, the first action should have 6 choices, as follows:

[0182] Table 6: State attribute after the first step decision of each cellular automaton case 1

[0183]

[0184] Table 7: State attribute after the first step decision of each cellular automaton case 2

[0185]

[0186] The subsequent steps will continue to expand with the optimal case 2 as an example.

[0187] S203: After 50 minutes, the process is completed, and the cellular state is as shown in the following table, where * indicates that the order has been processed and enters S204, but has not yet been transported to the next process by the AGV.

[0188] Table 8: State attribute after the first state change of each cellular automaton

[0189]

[0190] S204 and S205 are performed simultaneously at this point:

[0191] In S205, each station selects the next processing order for processing according to the greedy algorithm.

[0192] In S204, the current AGV logistics vehicle selects workpiece 6 and workpiece 5 for transportation according to its idle condition and the total processing time required. According to the genetic algorithm, the transportation path should be station 1 to station 5 to station 7. Assuming the vehicle speed is 1 m / min, the time for workpiece 5 to arrive at station 5 should be the 54th minute of the start of operation, and the time for workpiece 6 to arrive at station 7 should be the 56th minute of the start of operation.

[0193] S206: According to the action selection, the simulation is advanced to t+1, and all cellular states in the system are updated.

[0194] Table 9: State attribute after the first state change of each cellular automaton

[0195]

[0196] Continue to advance the system simulation according to the above process, and when the second system state changes, the cellular state should be as shown in the following table.

[0197] Table 10: State attribute after the second state change of each cellular automaton

[0198]

[0199] In the example, an AGV logistics vehicle is used for transportation.

[0200] S207: After all orders are processed, there are multiple solutions, and after the optimal solution is selected, the system Gantt chart should be as shown in Figure 7

[0201] The preferred embodiments of the present application have been described in detail. It should be understood that modifications and variations can be made by those skilled in the art without creating spurious logic based on the concept of the present application. Therefore, any technical solutions obtained by logical analysis, reasoning or limited experiments based on the concept of the present application in the prior art should be within the scope of protection defined by the claims.​

Claims

1. A dynamic scheduling optimization method for single-piece small-lot flexible manufacturing job shop, characterized in that, The method comprises the following steps: According to the process route of the actual processed workpiece, a cellular network model is established, and the initial state of each cell is obtained, wherein the cells of the cellular network model include buffer cells, station cells, AGV logistics vehicle cells and workpiece cells; According to the initial processable order selection of each station, all possible first-step actions are traversed, and the following steps are performed with different first-step actions: 1) Determine whether all orders are completed, if yes, end and obtain a final scheduling solution, if not, perform step 2); 2) Determine whether the production state changes, if yes, perform corresponding processing based on the state change type, if not, perform step 3), wherein the state change type includes process interruption, emergency order addition and process order process completion, When the state change type is process interruption, the corresponding processing includes: Re-defining the state of all related cells, and raising the priority of the unfinished order to the highest; Setting the current station state to fault; When the state change type is emergency order addition, the corresponding processing includes: Raising the priority of the emergency order to the highest; When the state change type is process order process completion, the corresponding processing is to start the logistics distribution process, specifically including: Based on the loading condition of the AGV logistics vehicle, the optimization target and the pre-constructed behavior candidate set, the behavior selection optimization of the AGV logistics vehicle is performed until the transportation of all workpieces is completed, and the workshop transportation route of the AGV logistics vehicle involved in multi-order transportation is obtained based on genetic algorithm optimization; 3) Each station obtains the next step work according to the current situation; 4) Update the state of all cells based on the cell basic state evolution function, and return to step 1); Obtain the final scheduling solution corresponding to all first-step actions, select the optimal solution, and control the workshop production process based on the optimal solution.

2. The dynamic scheduling optimization method for single-piece small-lot flexible manufacturing cells according to claim 1, wherein, The cellular network model is a complex cellular network with moving particles constructed in reference to Petri net form, and the process orders processed by each cell are moving particles.

3. The method of claim 1, wherein, The next step work obtained by each station according to the current situation is specifically: Each workpiece selects and optimizes the behavior based on the optimization target and the pre-constructed behavior candidate set through an optimization algorithm.

4. The method of claim 1, wherein, The cell basic state evolution function is used to define the law of change of the cell state with time, and the cell basic state evolution function is determined based on the current cell state, the related cell state and the decision behavior.

5. The dynamic scheduling optimization method for single-piece- lot flexible manufacturing cells according to claim 4, characterized in that, The cell state is constructed and obtained based on various parameters of each type of cell under different processing, behavior and time.

6. The method of claim 1, wherein, The buffer cells include raw material storage, station line edge inventory cells, workshop storage cells and finished product storage cells.

7. The method of claim 1, wherein, The periodic order polling method is used to determine whether the production state changes.

8. A device for dynamic scheduling optimization for single piece small lot flexible manufacturing job shop, characterized in that, It comprises: A dynamic monitoring module for collecting real-time state data of the workshop; A dynamic scheduling optimization module for controlling the workshop production process by using the dynamic scheduling optimization method of any one of claims 1-7 based on the real-time state data of the workshop.

Images