Intelligent manufacturing system dynamic scheduling method

Abstract

The application discloses a kind of intelligent manufacturing system dynamic scheduling method, specifically relates to the technical field of intelligent scheduling;Through the scheduling performance evaluation mechanism of fusion task switching interference index and target function deviation index, the scheduling health score prediction is realized using polynomial regression model, and the scheduling performance anomaly is accurately identified;In abnormal condition, the system can dynamically adjust the scheduling target function structure and algorithm disturbance parameters according to different index sources, effectively enhance the ability to jump out of local optimum, improve the global optimization effect;Through the closed-loop scheduling execution and feedback mechanism, the robustness, adaptability and execution feasibility of the scheduling system are significantly improved, to ensure that high-optimal task is completed in time and improve overall production efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent scheduling, and particularly relates to a dynamic scheduling method for an intelligent manufacturing system. BACKGROUND

[0002] Dynamic scheduling of an intelligent manufacturing system refers to a process in which, in an intelligent manufacturing environment, the system automatically optimizes and adjusts the allocation of production tasks according to real-time data and changes in production resources (such as device status, order changes, process adjustments, etc.). It integrates technologies such as the Internet of Things, artificial intelligence, and big data analysis to achieve flexible response and efficient collaboration in manufacturing processes, thereby improving production efficiency, resource utilization, and the ability to respond to unexpected events.

[0003] The prior art has the following shortcomings:

[0004] In complex real-time multi-objective scheduling, heuristic or local search algorithms often have difficulty breaking out of existing task ordering patterns when the search space dimension is high. For example, on an electronic product assembly line, to reduce the number of line changes, the system constantly prioritizes the same type of product, falling into a local optimal solution, resulting in high-priority orders being continuously delayed, which seriously affects delivery dates. The root cause is that the scheduling algorithm lacks an effective "breakout" mechanism, such as the perturbation strategy in simulated annealing or high-intensity mutation in genetic algorithms, making it difficult to globally optimize under multi-objective trade-offs, thereby limiting the adaptability and robustness of the scheduling system. SUMMARY

[0005] The purpose of the present application is to provide a dynamic scheduling method for an intelligent manufacturing system to address the shortcomings in the background art.

[0006] To achieve the above purpose, the present application provides the following technical solution: a dynamic scheduling method for an intelligent manufacturing system, comprising:

[0007] Collecting real-time state data and task information in the manufacturing system;

[0008] Generating a task switching disturbance index after analyzing the degree of resource disturbance caused by frequent task switching, and generating a target function deviation index after analyzing the degree of dynamic deviation between the current scheduling result and the theoretical global optimum;

[0009] Inputting the task switching disturbance index and the target function deviation index into a pre-constructed scheduling performance anomaly prediction model, and the output of the model is a scheduling health score value;

[0010] Comparing the scheduling health score value with a preset threshold value, if the scheduling health score value is lower than the preset threshold value, it is determined that the scheduling has a performance anomaly;

[0011] For scheduling performance abnormal situation, dynamically select the corresponding scheduling optimization strategy, including whether to introduce task switching penalty term in the scheduling objective function according to the task switching interference index, and adjusting the disturbance parameter in the scheduling algorithm according to the objective function offset index;

[0012] The optimized scheduling algorithm is executed, the final scheduling scheme is output, and the scheduling result is sent to the manufacturing execution system for production task execution.

[0013] Preferably, the state data includes device operating state, task progress, resource occupancy rate and material distribution state; the task information includes order parameters, process path, delivery requirements and personnel scheduling information.

[0014] Preferably, the generation method of the task switching interference index is:

[0015] For each resource r∈R, R represents a set of all resources participating in the current scheduling process, a fixed length sliding time window W is set, and the unit time interval is Δt; the task execution sequence of the resource in the window is recorded as: ; wherein represents the task number executed by the resource in the ith time slice, and n represents the total time slice; the sequence is traversed to record whether the adjacent tasks change: ; wherein, the cumulative switching number ; the comprehensive switching intensity is calculated , and the expression is: ;

[0016] The resource disturbance weight is defined, the switching interference index of a single resource is calculated, and the expression is: ; wherein is the maximum disturbance weight in the window, is the switching interference index of a single resource; the interference indexes of all resources are summarized to form the overall task switching interference index, and the expression is: ; wherein, is the task switching interference index.

[0017] Preferably, the generation method of the objective function offset index is:

[0018] The optimization objectives of the scheduling system are set to m, and ; wherein: x represents a scheduling solution; represents the value of the ith objective function; the reference Pareto front set is constructed, and the expression is: ; z is the total number of objective functions; the minimum distance between the current solution and the Pareto front is calculated, and for the current scheduling solution The objective function value of which is , calculate the distance to each point in the Pareto frontier set: ; in the formula, indicates the Euclidean distance; take the minimum distance as the offset distance; the objective function offset index is obtained after normalizing the offset distance , the expression is: ; wherein: is the upper limit of the reference distance.

[0019] Preferably, the task switching disturbance index and the objective function offset index are input into a pre-constructed scheduling performance anomaly prediction model, and the output of the model is a scheduling health score value, including:

[0020] The task switching disturbance index and the objective function offset index are converted into a comprehensive feature vector, and the comprehensive feature vector is taken as the input of a machine learning model. The machine learning model takes the scheduling health score value label predicted by each set of comprehensive feature vectors as the prediction target, minimizes the sum of prediction errors of all scheduling health score value labels as the training target, and trains the machine learning model until the sum of prediction errors converges to stop model training. The scheduling health score value is determined according to the model output result, wherein the machine learning model is a polynomial regression model.

[0021] Preferably, the scheduling health score value is compared with a preset threshold value. If the scheduling health score value is lower than the preset threshold value, it is determined that the scheduling performance is abnormal, including:

[0022] The obtained scheduling health score value is compared with a preset score threshold value. If the scheduling health score value is greater than or equal to the score threshold value, it is determined that the scheduling performance is normal; if the scheduling health score value is less than the preset score threshold value, it is determined that the scheduling performance is abnormal.

[0023] Preferably, the optimization strategy for the task switching disturbance index TSDD includes: modifying the objective function in the scheduling optimization process, and introducing a task switching penalty term as an additional optimization target; the optimized objective function is represented as: ; in the formula, is the original scheduling objective function; TSDD is the task switching disturbance index calculated by the current system; α is a penalty weight coefficient, , wherein is the reference threshold value of TSDD, is an adjustment coefficient.

[0024] Preferably, the optimization strategy for the objective function offset index OFDR includes: adjusting the temperature attenuation factor in the simulated annealing algorithm, and the expression is: If the OFDR is greater than its reference threshold, the disturbance at the beginning of the annealing process is increased: ; is the current temperature parameter, and λ is the temperature decay factor, represents the default decay rate, is the adjustment coefficient, is the reference threshold of the OFDR;

[0025] The mutation probability in the genetic algorithm is adjusted, and the expression is: ; represents the current mutation probability, is the default mutation probability; is the adjustment coefficient;

[0026] When the OFDR is greater than its reference threshold, the mutation probability is automatically increased to enhance the ability of the search to jump out of the local optimum; when the OFDR is less than its reference threshold, the mutation probability gradually falls back.

[0027] In the above technical solution, the technical effects and advantages provided by the present application are:

[0028] When the OFDR is greater than its reference threshold, the mutation probability is automatically increased to enhance the ability of the search to jump out of the local optimum; when the OFDR is less than its reference threshold, the mutation probability gradually falls back.

[0029] In the above technical solution, the technical effects and advantages provided by the present application are:

[0030] 1、The present application introduces two innovative parameters, task switching disturbance index and objective function offset index, for real-time dynamic evaluation of scheduling state, effectively compensating for the technical shortcomings of existing scheduling systems in complex multi-objective environments, such as being easily trapped in local optimum and lacking stability control. With the help of sliding window modeling, Pareto frontier distance measurement, and machine learning prediction model, the system can realize quantitative evaluation of scheduling state, and through scheduling health score value, it can identify potential performance abnormalities in advance, significantly improving the adaptability and intelligence level of the scheduling system.

[0031] 2、The present application dynamically selects optimization strategy based on scheduling abnormality source, not only improves scheduling stability by introducing task switching penalty term, but also adaptively adjusts search disturbance parameters for scheduling results deviating from the optimal, effectively improving the ability of scheduling algorithm to jump out of local optimum, thus realizing efficient, stable, and executable scheduling results. The overall scheme builds a "perception-evaluation-optimization-execution-feedback" closed-loop scheduling control system, which has strong robustness, real-time performance, and industrial applicability, and can be widely applied to various intelligent manufacturing scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0033] Figure 1 This is a mind map of the method of the present invention. DETAILED DESCRIPTION

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0035] For examples, see Figure 1 As shown, the dynamic scheduling method of an intelligent manufacturing system described in this embodiment includes:

[0036] Collect real-time status data and task information in manufacturing systems;

[0037] The task switching interference index is generated by analyzing the resource disturbance degree caused by frequent task switching, and the objective function offset index is generated by analyzing the dynamic deviation degree between the current scheduling result and the theoretical global optimal.

[0038] Inputting the task switching interference index and the objective function offset index into a pre-built scheduling performance anomaly prediction model, the output of the model being a scheduling health score value;

[0039] Comparing the scheduling health score with a preset threshold, if the scheduling health score is lower than the preset threshold, determining that the scheduling has performance anomalies;

[0040] For abnormal scheduling performance, the corresponding scheduling optimization strategy is dynamically selected, including whether to introduce a task switching penalty term into the scheduling objective function based on the task switching interference index, and adjusting the disturbance parameters in the scheduling algorithm based on the objective function offset index;

[0041] Execute the optimized scheduling algorithm, output the final scheduling plan, and send the scheduling results to the manufacturing execution system for production task execution.

[0042] In the present application, the first step is to collect real-time state data and task information in the manufacturing system, which is used to provide accurate and comprehensive input data basis for the subsequent scheduling algorithm, and to ensure the real-time and effectiveness of the scheduling calculation.

[0043] In the present embodiment, the manufacturing system is a typical discrete multi-station assembly production line, involving multiple CNC machines, industrial robots, AGVs, production process sites and auxiliary resources.

[0044] To achieve accurate collection of real-time state and task information, the present embodiment uses the following specific technical means:

[0045] An industrial Internet of Things (IIoT) architecture is used to collect equipment operating status through intelligent sensors and data collection terminals (such as PLCs and edge gateways) deployed at key nodes of the production line. The state information includes but is not limited to: equipment operating status such as "standby", "processing", "abnormal", "maintenance"; current task number and task phase identifier; spindle load, current, voltage, temperature and other working condition data; fault warning information (such as abnormal vibration frequency, insufficient lubricating oil pressure); all collected data is uploaded to the manufacturing execution system (MES) through the OPC UA protocol and processed synchronously through time stamping, ensuring that the scheduling system receives high consistency and low latency state data.

[0046] The system dynamically calculates resource occupancy based on the current task schedule and device availability, including: the scheduled use time of a single device within a specified time window; process station occupancy and idle status; worker scheduling status and skill matching degree; AGV path congestion and call frequency; resource occupancy is updated in matrix form in real time for the scheduling engine to use for resource conflict determination and load balancing optimization.

[0047] Order information is real-time interfaced from the enterprise ERP system, with specific parameters including: order number, customer level and delivery requirement; product specifications, BOM list and process path; urgency (priority) marking; order insertion and change record; the ERP and MES systems are connected through API interface, and the order information is input into the scheduling engine after standardized processing to form a task decomposition list.

[0048] The scheduling system monitors the execution progress of each task in real time through data linkage with the MES, combined with RFID tracking, barcode scanning, station reporting and other means, including: the current task process node; completed and to-be-completed task phases; actual start and end time compared with planned time; delay identification and cause analysis (such as material shortage, downtime); task progress is dynamically presented in Gantt chart structure to determine task lag risk and whether to trigger the rescheduling mechanism.

[0049] All collected data are pre-processed and uniformly modeled through an industrial big data platform, using data standardization, redundancy filtering, time synchronization and other processing methods, and finally stored in JSON format or database table structure for calling by the scheduling system.

[0050] Meanwhile, to support algorithm efficiency, the system caches core data items in an in-memory database (such as Redis) and sets a data refresh period (e.g., updating every 10 seconds) to ensure that the scheduling algorithm can obtain the latest system snapshot for calculation.

[0051] In the above manner, the embodiment realizes high-precision collection and unified management of real-time state data and task information in a manufacturing system, not only providing basic data input required for scheduling, but also providing key support for subsequent anomaly identification, parameter calculation (such as TSDD and OFDR) and scheduling optimization, embodying high adaptability and practicality of the application in the industrial intelligent scheduling scenario.

[0052] After analyzing the degree of resource disturbance caused by frequent task switching, a task switching interference index is generated, a sliding time window is established on each resource (such as a machine tool, an operator, and a robot), the frequency, intensity and complexity of task switching within the window are counted, and the disturbance degree to system stability is calculated in combination with the sensitive weight of the resource, thereby forming the task switching interference index.

[0053] The specific generation method is as follows:

[0054] For each resource r∈R, R represents a set of all resources participating in the current scheduling process, containing multiple resource units. A sliding time window W of fixed length is set, and the unit time interval is Δt; the task execution sequence of the resource within the window is recorded as: ; wherein represents the task number executed by the resource in the i-th time slice, and n represents the total time slice; the sequence is traversed to record whether the adjacent tasks change: ; wherein, the cumulative switching number ; each task switching may have different disturbance complexity, and the task switching intensity function is defined as:

[0055] =β1*process switching level+β2*material replacement complexity+β3*equipment adjustment time; the comprehensive switching intensity is calculated, and the expression is: ;

[0056] Each type of resource has different tolerance to switching, and the resource disturbance weight is defined, such as a high-precision CNC machine tool: = 1.0; Automatic handling robot: = 0.6; General assembly station: = 0.3;

[0057] The target function offset index is generated after analyzing the dynamic deviation degree between the current scheduling result and the theoretical global optimum, and the generation method is:

[0058] The switching interference index of a single resource is calculated, and the expression is: ; wherein is the maximum disturbance weight in the window (for normalization), is the switching interference index of a single resource; The interference indexes of all resources are summarized to form the overall task switching interference index, and the expression is: ; wherein, is the task switching interference index. The closer the interference index is to 1, the poorer the stability under the current scheduling, and the system needs to be adjusted.

[0059] The target function offset index is generated after analyzing the dynamic deviation degree between the current scheduling result and the theoretical global optimum, and the generation method is:

[0060] The optimization goal of the scheduling system is set to m (such as m = 2, the most common is total completion time Makespan and energy consumption Energy), and is defined ; wherein: x represents a scheduling solution; represents the value of the i-th objective function. The reference Pareto front set is constructed, and the expression is: ; z is the total number of objective functions; The set is obtained in the following ways: using evolutionary algorithms (such as NSGA-II, MOEA / D) to run in the same task scenario, saving non-dominated solutions; Based on the known "scheduling optimal solution" in the historical data, the ideal scheduling boundary is constructed; Simulation model to obtain ideal scheduling boundary.

[0061] The minimum distance between the current solution and the Pareto front is calculated, and for the current scheduling solution , the objective function value is , and the distance from each point in the Pareto front set to the current scheduling solution is calculated: ; wherein, represents the Euclidean distance; Take the minimum distance as the offset distance;

[0062] The target function offset index is obtained after normalizing the offset distance, and the expression is: ; wherein: For reference distance upper limit, the maximum possible objective function difference of the current problem or the maximum distance of two points in the Pareto boundary point set is taken; OFDR represents the objective function offset index, ranging from 0 to 1, and the closer to 1, the more serious the offset.

[0063] The task switching disturbance index and the objective function offset index are input into a pre-constructed scheduling performance anomaly prediction model, and the output of the model is a scheduling health score value, including:

[0064] The task switching disturbance index and the objective function offset index are converted into a comprehensive feature vector, the comprehensive feature vector is taken as the input of a machine learning model, the machine learning model takes the scheduling health score value label predicted from each set of comprehensive feature vectors as the prediction target, minimizes the sum of prediction errors of all scheduling health score value labels as the training target, and trains the machine learning model until the sum of prediction errors converges to stop model training. The scheduling health score value is determined according to the model output result, wherein the machine learning model is a polynomial regression model.

[0065] The scheduling health score value is compared with a preset threshold value, and if the scheduling health score value is lower than the preset threshold value, it is determined that the scheduling performance is abnormal;

[0066] The score threshold Hth is set in (0, 1) to judge the scheduling performance state. The obtained scheduling health score value is compared with the preset score threshold, and if the scheduling health score value is greater than or equal to the score threshold, it is determined that the scheduling performance is normal; if the scheduling health score value is less than the preset score threshold, it is determined that the scheduling performance is abnormal.

[0067] For the scheduling performance abnormality, a corresponding scheduling optimization strategy is dynamically selected, including judging whether to introduce a task switching penalty term in the scheduling objective function according to the task switching disturbance index, and adjusting the disturbance parameter in the scheduling algorithm according to the objective function offset index;

[0068] The optimization strategy for the task switching disturbance index TSDD is used to reduce the system disturbance caused by task switching and improve the stability of scheduling.

[0069] In the scheduling optimization process, the objective function is modified, and a task switching penalty term is introduced as an additional optimization target.

[0070] The optimized objective function is represented as: ; in the formula, is the original scheduling objective function (such as total completion time, total delay time, etc.); TSDD is the task switching disturbance index currently calculated by the system; and a is a penalty weight coefficient for adjusting the importance of TSDD in the objective function, which can be dynamically adjusted.

[0071] If TSDD is greater than its reference threshold, then α is increased appropriately to strengthen the stability orientation; the following can be set wherein TSDD is the reference threshold for TSDD, is the adjustment coefficient.

[0072] The optimization strategy for the objective function offset index OFDR is used to enhance the ability of the scheduling algorithm to jump out of the local optimum and improve the global search capability.

[0073] The disturbance control parameter in the scheduling algorithm is dynamically adjusted, which is applicable to heuristic search algorithms (such as genetic algorithm, simulated annealing, particle swarm optimization, etc.).

[0074] The temperature decay factor in the simulated annealing algorithm is adjusted, and the expression is: If OFDR is greater than its reference threshold, then the disturbance at the initial stage of the annealing process is increased: ; is the current temperature parameter, λ is the temperature decay factor, represents the default decay rate (such as 0.95), is the adjustment coefficient, which determines the disturbance intensity response speed, is the reference threshold for OFDR.

[0075] The mutation probability in the genetic algorithm is adjusted, and the expression is: ; represents the current mutation probability, is the default mutation probability (such as 0.05); is the adjustment coefficient.

[0076] When OFDR is greater than its reference threshold, the system automatically increases the mutation probability to enhance the ability of the search to jump out of the local optimum; when OFDR is less than its reference threshold, the mutation probability can gradually fall back.

[0077] The optimized scheduling algorithm is executed, and the final scheduling scheme is output, and the scheduling result is sent to the manufacturing execution system for production task execution.

[0078] The system re-invokes the scheduling engine to perform scheduling calculation based on the updated scheduling objective function or algorithm disturbance parameter. At this time, the scheduling algorithm has integrated the task switching penalty term or enhanced the global search capability, making it more suitable for the current state of resource allocation and optimization objectives.

[0079] The scheduling algorithm takes the real-time task list, device available state, resource constraints, and optimization strategy as input data to perform task allocation and sequencing calculation. The scheduling engine aims to find the global optimal or near-optimal solution, and re-searches the optimal scheduling path in the solution space according to the updated objective function.

[0080] During the execution process, the system also needs to monitor the convergence state of the scheduling calculation process and the resource load balancing situation in real time to avoid new problems such as resource overload or scheduling imbalance. After the scheduling solution is completed, the system will automatically generate a new set of task scheduling scheme.

[0081] The results output by the scheduling engine include the allocation of resources for each task, the planned start and end times, task priority ranking, and critical path identification. All scheduling information will be packaged into a unified data structure for subsequent system integration.

[0082] In addition, the system can generate scheduling visualization graphs such as task Gantt charts, resource utilization charts, and conflict alert charts in real time for operators to view and verify the rationality of the scheduling results.

[0083] The scheduling scheme will also indicate whether there are high-risk task nodes to be observed, such as tasks in resource bottlenecks, near the critical line of the delivery period, or easily affected by disturbances. The system will mark and prompt these tasks to facilitate subsequent monitoring in the running phase.

[0084] After the scheduling scheme is confirmed, the system will send the task allocation results to the manufacturing execution system (MES) through a standard interface. This process is usually achieved through message queues, API interfaces, or industrial bus communication protocols such as OPC UA and MQTT to interface with the MES.

[0085] After the manufacturing execution system receives the task plan, it triggers the scene control logic and assigns the task to specific equipment or operating positions. For example, automated equipment will receive the machining path, process parameters, and task execution time, and operators will receive task scheduling and material distribution instructions through terminals.

[0086] During this process, the system ensures that the scheduling results are consistent with the actual resource state in the MES, avoiding problems such as data inconsistency, repeated scheduling, or resource conflicts. At the same time, the scheduling engine retains all scheduling records and result data for subsequent traceability analysis and scheduling model optimization.

[0087] The manufacturing execution system begins to operate according to the new scheduling plan, and the system continuously monitors the task execution state and actual feedback data. If there are changes in resource state, urgent order insertion, or task abnormalities during the execution process, the system will trigger the scheduling adaptive process again based on real-time feedback, forming a closed-loop control mechanism.

[0088] Through the above steps, the system realizes seamless connection from optimization strategy generation to scheduling scheme output and execution, ensuring that the scheduling results are not only theoretically optimal but also can be efficiently and stably implemented in the field. The entire process emphasizes real-time, closed-loop, and operability, and is the core support for the dynamic scheduling capability of intelligent manufacturing systems.

[0089] The above formulas are all dimensionless values calculated, and the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0090] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD) or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0091] It should be understood that the term "and / or" herein only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone, where A and B can be singular or plural. In addition, the character " / " herein generally represents that the associated objects before and after are an "or" relationship, but can also represent an "and / or" relationship, which can be understood in combination with the preceding and following texts. Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0092] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A dynamic scheduling method for intelligent manufacturing system, characterized in that: The method comprises the following steps: Collecting real-time state data and task information in a manufacturing system; Analyzing the degree of resource disturbance caused by frequent task switching to generate a task switching disturbance index, and analyzing the dynamic deviation degree between the current scheduling result and the theoretical global optimum to generate a target function deviation index; Wherein, the generation method of task switching interference index is: for each resource r∈R, R represents the set composed of all resources participating in the current scheduling process, set a fixed length sliding time window W, the unit time interval is Δt; The task execution sequence of the resource in the window is recorded as: ; Wherein represents the task number executed by the resource in the ith time slice, n represents the total time slice; The sequence is traversed, and whether the adjacent tasks change is recorded: ; Wherein, the cumulative switching number ; The comprehensive switching strength is calculated, and the expression is: ; Wherein, represents the task switching strength between task in the task switching process, which is calculated by weighted summation of process switching level, material replacement complexity and equipment adjustment time, the resource disturbance weight is defined, the switching interference index of single resource is calculated, and the expression is: ; Wherein is the maximum disturbance weight in the window, is the switching interference index of single resource; The interference indexes of all resources are summarized to form the overall task switching interference index; The generation method of the objective function offset index is as follows: setting the optimization objective of the scheduling system as m, defining ; wherein x represents a scheduling solution; represents the value of the i-th objective function; a reference Pareto front set is constructed , and the expression is as follows: ; z is the total number of objective functions; the minimum distance between the current solution and the Pareto front is calculated, and for the current scheduling solution , the objective function value is , the distance from the current solution to each point in the Pareto front set is calculated: ; in the formula, , the Euclidean distance is represented; the minimum distance is taken as the offset distance; the objective function offset index is obtained after the offset distance is normalized, and the expression is as follows: ; wherein: is the upper limit of the reference distance; Inputting the task switching disturbance index and the target function deviation index into a pre-constructed scheduling performance anomaly prediction model, and the output of the model is a scheduling health score value; Comparing the scheduling health score value with a preset threshold value, and if the scheduling health score value is lower than the preset threshold value, it is determined that the scheduling has performance anomaly; For the scheduling performance anomaly, dynamically selecting a corresponding scheduling optimization strategy, including determining whether to introduce a task switching penalty term in the scheduling target function according to the task switching disturbance index, and adjusting the disturbance parameter in the scheduling algorithm according to the target function deviation index; Specifically include: for the optimization strategy of task switching disturbance index TSDD, including: in the scheduling optimization process, modify the objective function, introduce the task switching penalty term as an additional optimization goal; the optimized objective function is expressed as: ; In the formula, TSDD is the original scheduling objective function; TSDD is the task switching disturbance index calculated by the current system; alpha is the penalty weight coefficient, , wherein TSDD is the reference threshold of TSDD, is the adjustment coefficient; The optimization strategy for the objective function deviation ratio (OFDR) includes adjusting the temperature attenuation factor in the simulated annealing algorithm, expressed as: ; and increasing the disturbance at the initial stage of the annealing process if the OFDR is greater than a reference threshold value thereof: ; is a current temperature parameter, λ is a temperature attenuation factor, represents a default attenuation rate, is an adjustment coefficient, is a reference threshold value of the OFDR; Adjust the mutation probability in genetic algorithm, the expression is: ; represents the current mutation probability, is the default mutation probability; is the adjustment coefficient; When the OFDR is greater than its reference threshold value, the mutation probability is automatically increased to enhance the ability to jump out of the local optimum; when the OFDR is less than its reference threshold value, the mutation probability gradually falls back; Executing the optimized scheduling algorithm to output a final scheduling scheme, and sending the scheduling result to a manufacturing execution system for production task execution.

2. The dynamic scheduling method of intelligent manufacturing system according to claim 1, characterized in that: The state data includes device running state, task progress, resource occupancy rate and material distribution state; the task information includes order parameters, process path, delivery requirements and personnel scheduling information.

3. The dynamic scheduling method of intelligent manufacturing system according to claim 1, characterized in that: The task switching disturbance index and the target function deviation index are input into a pre-constructed scheduling performance anomaly prediction model, and the output of the model is a scheduling health score value, which comprises: The task switching disturbance index and the target function deviation index are converted into a comprehensive feature vector, and the comprehensive feature vector is used as the input of a machine learning model. The machine learning model takes the prediction of the scheduling health score value label as the prediction target, minimizes the sum of prediction errors of all scheduling health score value labels as the training target, and trains the machine learning model until the sum of prediction errors converges to stop model training. The scheduling health score value is determined according to the model output result, wherein the machine learning model is a polynomial regression model.

4. The dynamic scheduling method of an intelligent manufacturing system according to claim 3, characterized in that: Comparing the scheduling health score value with a preset threshold value, if the scheduling health score value is lower than the preset threshold value, it is determined that the scheduling has performance anomaly, which comprises: Comparing the obtained scheduling health score value with a preset score threshold value, if the scheduling health score value is greater than or equal to the score threshold value, it is determined that the scheduling performance is normal; if the scheduling health score value is less than the preset score threshold value, it is determined that the scheduling has performance anomaly.

Images