Intelligent Scheduling and Dispatching System and Method Based on Digital Twin

Abstract

This application relates to the field of intelligent production scheduling and dispatching technology. Specifically, it discloses an intelligent production scheduling and dispatching system and method based on digital twin. After arranging the set of orders to be scheduled into two sequences of orders to be scheduled by adopting different sorting strategies, it uses a resource-time matrix to perform production scheduling simulation on these two sequences of orders to be scheduled respectively. By simulating the production process under different production scheduling methods, corresponding simulation logs are generated, and then based on the comparison between the simulation logs, an optimized production scheduling method is determined for corresponding scheduling and management. In this way, potential problems can be discovered in advance through virtual simulation in digital twin technology, the production scheduling method can be optimized in a timely manner, and the production plan can be adjusted in a timely manner to avoid losses in actual production.

Description

Technical Field

[0001] This application relates to the technical field of intelligent production scheduling, and more specifically, to an intelligent production scheduling system and method based on digital twin. Background Art

[0002] In modern manufacturing, the manufacturing process in a workshop mainly consists of machining, thermal processing, and inspection. These processing processes include multiple procedures, and each procedure in turn includes multiple production lines. Therefore, production scheduling is one of the core links in production management. The production scheduling goal is to sort order tasks to meet production requirements for order demands within a certain period of time. Effective production scheduling can not only improve production efficiency, reduce resource waste, but also ensure on-time delivery of products and improve customer satisfaction. With the intensification of market competition, enterprises are facing increasingly high production efficiency requirements and shorter product delivery cycles. Therefore, how to scientifically and reasonably arrange production plans has become an important manifestation of an enterprise's competitiveness.

[0003] However, traditional production scheduling methods mostly rely on the experience and intuition of production management personnel. This method may be relatively effective when facing simple and repetitive production tasks, but in dealing with complex and changeable production environments, it is prone to decision-making errors, resulting in low production efficiency. Moreover, traditional production scheduling methods usually also adopt some fixed rules or algorithms, and this way is difficult to adapt to market changes and quickly respond to customer needs. Especially in the multi-variety and small-batch production mode, fixed rules are difficult to meet diverse needs, easily causing resource waste and production delays. In addition, traditional production scheduling methods often cannot make full use of existing resources. Whether it is the way of manual experience or fixed rule algorithms, it will result in a low level of intelligence in the production management and scheduling system, leading to longer idle time and changeover time of equipment, resulting in low overall production efficiency.

[0004] Therefore, an optimized production scheduling system is desired. Summary of the Invention

[0005] This application provides an intelligent production scheduling system and method based on digital twin, which can discover potential problems in advance through virtual simulation in digital twin technology, optimize the production scheduling method in a timely manner, and adjust the production plan in a timely manner to avoid losses in actual production.

[0006] In a first aspect, there is provided an intelligent production scheduling and dispatching system based on digital twin, including: a set acquisition module for to-be-scheduled orders, configured to acquire a set of to-be-scheduled orders; a set arrangement module for to-be-scheduled orders, configured to arrange the set of to-be-scheduled orders in a first manner to obtain a sequence of first to-be-scheduled orders and arrange the set of to-be-scheduled orders in a second manner to obtain a sequence of second to-be-scheduled orders; a production scheduling module for to-be-scheduled order sequences, configured to set production scheduling time and perform production scheduling on the sequence of first to-be-scheduled orders and the sequence of second to-be-scheduled orders based on a resource-time matrix to obtain a simulation log of the first production scheduling method and a simulation log of the second production scheduling method; a simulation log display module for production scheduling methods, configured to display the simulation log of the first production scheduling method and the simulation log of the second production scheduling method; a comparison and decision-making module for simulation logs of production scheduling methods, configured to determine whether to arrange the set of to-be-scheduled orders in the first manner or the second manner based on a comparison between the simulation log of the first production scheduling method and the simulation log of the second production scheduling method.

[0007] In a second aspect, there is provided an intelligent production scheduling and dispatching method based on digital twin, including: acquiring a set of to-be-scheduled orders; arranging the set of to-be-scheduled orders in a first manner to obtain a sequence of first to-be-scheduled orders and arranging the set of to-be-scheduled orders in a second manner to obtain a sequence of second to-be-scheduled orders; setting production scheduling time and performing production scheduling on the sequence of first to-be-scheduled orders and the sequence of second to-be-scheduled orders based on a resource-time matrix to obtain a simulation log of the first production scheduling method and a simulation log of the second production scheduling method; displaying the simulation log of the first production scheduling method and the simulation log of the second production scheduling method; determining whether to arrange the set of to-be-scheduled orders in the first manner or the second manner based on a comparison between the simulation log of the first production scheduling method and the simulation log of the second production scheduling method.

[0008] The intelligent production scheduling and dispatching system and method based on digital twin provided by this application arrange the set of to-be-scheduled orders into two sequences of to-be-scheduled orders by adopting different sorting strategies, then perform production scheduling simulations on these two sequences of to-be-scheduled orders respectively using a resource-time matrix, generate corresponding simulation logs by simulating the production process under different production scheduling methods, and then determine the optimized production scheduling method based on the comparison between the simulation logs for corresponding scheduling and management. In this way, potential problems can be discovered in advance through virtual simulation in digital twin technology, the production scheduling method can be optimized in a timely manner, and the production plan can be adjusted in a timely manner to avoid losses in actual production. Description of the Drawings

[0009] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of this application and do not limit this application.

[0010] Figure 1 It is a schematic block diagram of the digital-twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application.

[0011] Figure 2 It is a schematic block diagram of the production scheduling module for the order sequence to be scheduled in the digital-twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application.

[0012] Figure 3 It is a schematic block diagram of the comparison and decision-making module for the emission mode simulation log in the digital-twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application.

[0013] Figure 4 It is a schematic diagram of the data flow of the comparison and decision-making module for the emission mode simulation log in the digital-twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application.

[0014] Figure 5 It is a schematic flowchart of the digital-twin-based intelligent production scheduling and dispatching method according to an embodiment of the present application. Detailed implementation manners

[0015] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts also belong to the scope of protection of the present application.

[0016] Digital twin technology is an advanced technology that has emerged in recent years. It realizes seamless connection between the physical world and the digital world by constructing a virtual model of a physical entity. In the manufacturing industry, digital twin can be used to simulate the working state of a production line, predict equipment failures, optimize production processes, etc. Applying digital twin technology to the production scheduling and dispatching management process in industrial production can reflect the situation on the production site in real time, discover potential problems in advance through virtual simulation, and automatically optimize the dispatching of production methods, so as to make full use of equipment resources and reduce waste and losses in the actual production process.

[0017] Based on this, in the technical solution of the present application, a digital-twin-based intelligent production scheduling and dispatching system is proposed, which can use digital twin technology and artificial intelligence technology to solve the problems existing in traditional production scheduling methods. Specifically, in the technical solution of the present application, Figure 1 It is a schematic block diagram of the digital-twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application. As Figure 1As shown, the digital-twin-based intelligent production scheduling and dispatching system 100 includes: an order set to-be-scheduled acquisition module 110, configured to acquire a set of orders to be scheduled; an order set to-be-scheduled arrangement module 120, configured to arrange the set of orders to be scheduled in a first manner to obtain a first sequence of orders to be scheduled and arrange the set of orders to be scheduled in a second manner to obtain a second sequence of orders to be scheduled; an order sequence production scheduling module 130, configured to set production scheduling time and perform production scheduling on the first sequence of orders to be scheduled and the second sequence of orders to be scheduled based on a resource-time matrix to obtain a first simulation log of the production scheduling manner and a second simulation log of the production scheduling manner; a production scheduling manner simulation log display module 140, configured to display the first simulation log of the production scheduling manner and the second simulation log of the production scheduling manner; and a production scheduling manner simulation log comparison and decision-making module 150, configured to determine whether to arrange the set of orders to be scheduled in the first manner or the second manner based on a comparison between the first simulation log of the production scheduling manner and the second simulation log of the production scheduling manner.

[0018] Correspondingly, after arranging the set of orders to be scheduled into two sequences of orders to be scheduled by adopting different sorting strategies, the above intelligent production scheduling and dispatching system respectively performs production scheduling simulation on the two sequences of orders to be scheduled by using a resource-time matrix. By simulating the production process under different production scheduling manners, corresponding simulation logs are generated, and then an optimized production scheduling manner is determined based on the comparison between the simulation logs for corresponding scheduling and management. In this way, potential problems can be discovered in advance through virtual simulation in digital-twin technology, the production scheduling manner can be optimized in a timely manner, and the production plan can be adjusted in a timely manner to avoid losses in actual production.

[0019] In the above digital-twin-based intelligent production scheduling and dispatching system, the order set to-be-scheduled acquisition module 110 is configured to acquire a set of orders to be scheduled. It should be understood that acquiring a set of orders to be scheduled is the basis of the entire production scheduling and dispatching process. Only by clarifying the set of orders to be scheduled can it be known which products need to be produced within a specific time period, and then production resources can be reasonably arranged. By acquiring the order set, the specific requirements of each order (such as bill of materials, processes, equipment, etc.) can be understood, so as to reasonably allocate production resources and avoid waste of resources. Based on the order set, a detailed production plan can be formulated, including the start time, end time, required equipment, etc. of each order, to ensure the smooth progress of the production process. By acquiring the set of orders to be scheduled, potential problems such as insufficient equipment and material shortage can be discovered before production scheduling, and the production plan can be adjusted in a timely manner to avoid losses in actual production.

[0020] Optionally, in an embodiment of the present application, obtaining a set of orders to be scheduled includes: collecting information on all orders to be scheduled from the sales department or the customer management system. The information on the orders to be scheduled includes order numbers, product models, quantities, and delivery dates. Organize the collected order information into a unified format for subsequent processing. Here, spreadsheets, databases, or other data management tools can be used to store and manage this order information. Verify the collected order information to ensure the accuracy and integrity of the data. For example, check for duplicate orders, whether the order information is complete, and whether the delivery date is reasonable, etc.

[0021] Further, in another embodiment of the present application, obtaining a set of orders to be scheduled includes: obtaining the production processes of the orders to be scheduled, the processing time of each process, and the delivery time. Here, the manufacturing process in the workshop mainly consists of three parts: machining - heat treatment - inspection, that is, the production processes include: machining, heat treatment, inspection. Among them, machining and inspection each include multiple production lines, and heat treatment includes multiple processes such as A, B, C, D, E, etc. Each process includes multiple production lines, and different products have different processes in the heat treatment part. The scheduling goal is to sort the order tasks according to the order requirements within a period of time to meet the production needs.

[0022] In the above intelligent scheduling and dispatching system based on digital twin, the module 120 for arranging the set of orders to be scheduled is used to arrange the set of orders to be scheduled in a first manner to obtain a sequence of the first orders to be scheduled and arrange the set of orders to be scheduled in a second manner to obtain a sequence of the second orders to be scheduled. It should be understood that through different arrangement methods, the advantages and disadvantages of the scheduling plan can be evaluated from multiple perspectives to ensure that the finally selected plan is the optimal one. At the same time, a single arrangement method may ignore some potential problems. Through multiple arrangement methods, these problems can be discovered and avoided in advance. Different arrangement methods can cope with different production environments and market demands, improving the flexibility and adaptability of production scheduling. By comparing the simulation results of different arrangement methods, decisions can be made more scientifically, the best scheduling method can be selected, and production efficiency and resource utilization rate can be improved.

[0023] Optionally, in an embodiment of the present application, the set of orders to be scheduled is arranged to obtain a sequence of the first orders to be scheduled by adopting a forward scheduling strategy for all processes of the machining and heat treatment parts and the semi-finished products, that is, starting from the production start time, the processing time of each order is gradually arranged according to the processing sequence of the orders. At the same time, the set of orders to be scheduled is arranged to obtain a sequence of the second orders to be scheduled by adopting a backward scheduling strategy for the last inspection process in the order. That is, starting from the delivery date of the order, the processing time of each order is arranged in reverse to ensure that the order can be delivered on time. By means of the backward scheduling strategy, it can be ensured that the order is delivered on time and the delivery delay time is minimized as much as possible. It should be noted that other arrangements can also be adopted. For example, arranging according to the urgency of the order (such as the priority required by the customer). Arranging according to the production difficulty of the order (such as the complexity of the required equipment and the complexity of the process). Only in the technical solution of the present application, the above two scheduling methods are used for illustration, but those skilled in the art should know that the technical solution of the present application is not limited to this, but covers all possible scheduling schemes.

[0024] In the above intelligent scheduling and dispatching system based on digital twin, the order sequence scheduling module 130 is configured to set a scheduling time and schedule the sequence of the first orders to be scheduled and the sequence of the second orders to be scheduled based on the resource-time matrix to obtain a simulation log of the first scheduling method and a simulation log of the second scheduling method. It should be understood that by setting a scheduling time and simulating different scheduling methods based on the resource-time matrix, the effects of different scheduling methods can be evaluated and the optimal scheduling scheme can be selected. More specifically, before scheduling, a clear scheduling time period needs to be set, which is usually from the current time to a period in the future (such as one week, one month, etc.). By setting the scheduling time, it can be ensured that the scheduling scheme is effective within a specific time range. At the same time, it is ensured that all resources (such as equipment, manpower, materials, etc.) are available within this time period.

[0025] Optionally, in an embodiment of the present application, Figure 2 is a schematic block diagram of the order sequence scheduling module in the intelligent scheduling and dispatching system based on digital twin according to the embodiment of the present application. As Figure 2As shown, the production scheduling module 130 for the to-be-scheduled order sequence includes: a resource-time matrix initialization unit 131 for initializing the resource-time matrix, where the abscissa of the resource-time matrix represents the resources required for production, and the ordinate of the resource-time matrix represents the time axis; an order information extraction unit 132 for extracting the order information of the current order from the sequence of the first to-be-scheduled orders; an order detailed information parsing unit 133 for extracting the bill of materials, processes, equipment, and processing time of each process required for the current order from the order information of the current order; and an order process scheduling unit 134 for finding the earliest start time of the earliest process that satisfies the current order in the resource-time matrix based on the processing order of the processes of the current order, recording the processing time of the earliest process, and occupying the time period in the resource-time matrix.

[0026] More specifically, the production scheduling module for the to-be-scheduled order sequence includes: all information initialized before production scheduling, including equipment, product BOM, each product process, etc. Set the production scheduling time period, and then initialize a resource-time matrix, where the abscissa represents the resources required for production (machinery and equipment, manpower, materials, etc.), and the ordinate represents the time axis. In this scenario, minutes are used as the minimum unit. For the sequence of the first to-be-scheduled orders, according to the index in the sequence, sequentially grab the current order information, calculate the BOM, processes, equipment required for the current order, and the processing time of each process, etc. Then, according to the processing order of the processes, find the earliest start time that satisfies the current process processing in the resource-time matrix, record the processing time of this process, and record (occupy) this time period in the resource-time matrix.

[0027] Further, when scheduling the sequence of the first order to be scheduled, the following rules are also set for constraint. The rules include: First, the transfer rule is set with the order as the unit, and transfer is carried out every time an order is completed. For example, in an example of this application, the machining - heat treatment dle (Detailed Line Execution) is 15 minutes, the in - process transfer within heat treatment dle is 20 minutes, and the heat treatment dle - inspection is 1 hour. Second, if the types of processed products before and after the selected processing equipment time period are different, the change - over time of this production line needs to be vacated. Third, if the process after machining is A, overall consideration needs to be taken in the calculation. After calculating the start - end time of process A, it is necessary to verify whether the total time from machining to A is less than 48 hours. If it is less, no modification is required. If it is greater than 48 hours, the previous machining process needs to be modified, and the start time of the machining process needs to be postponed until the 48 - hour constraint is met. Fourth, if multiple devices can be selected during production, select the device that can start production earliest. Fifth, the start time of each process needs to be after the end time of the previous process plus the transfer time.

[0028] More specifically, the module for scheduling the sequence of orders to be scheduled further includes: extracting the order information of the current order from the set of orders to be scheduled, including order number, product model, quantity, delivery date, required bill of materials, processes, equipment, and processing time of each process, etc. According to the order information, calculate the required bill of materials, processes, equipment, and processing time of each process for the current order. Starting from the delivery date, inversely find the latest end time of each inspection process, record the processing time of this process, and occupy this period of time in the resource - time matrix.

[0029] Further, when scheduling the sequence of the second order to be scheduled, the following rules are also set for constraint. The rules include: First, delivery - date constraint, under the condition of ensuring on - time delivery, the end time should be as close to the delivery date as possible. Second, earliest completion time, if on - time delivery cannot be guaranteed, select the time that can be completed earliest. Third, transfer rule: with the order as the unit, transfer is carried out every time an order is completed. The machining - heat treatment is 15 minutes, the in - process transfer within heat treatment is 20 minutes, and the heat treatment - inspection is 1 hour. Fourth, change - over time: if the types of processed products before and after the selected processing equipment time period are different, the change - over time of this production line needs to be vacated.

[0030] Finally, after completing the scheduling, the model will respectively output different resource - time matrices for the sequences of the first order to be scheduled and the second order to be scheduled, as well as information such as the start time, end time, and occupied machines of each process of each order. According to the resource - time matrix, the utilization rate and change - over times of each device can be calculated; according to the end time, the delay time of each order can be calculated, and a Gantt chart can be drawn.

[0031] In the above intelligent production scheduling and dispatching system based on digital twin, the emission mode simulation log display module 140 is used to display the first emission mode simulation log and the second emission mode simulation log. It should be understood that by displaying the simulation log, the effects of different production scheduling methods can be intuitively evaluated, and the optimal production scheduling plan can be selected. At the same time, by displaying the simulation log, potential problems such as insufficient resources, equipment conflicts, and production bottlenecks can be discovered before actual production, and the production plan can be adjusted in time to avoid losses in actual production. By displaying the simulation log, resource allocation can be better optimized to ensure the smooth progress of the production process. The resource-time matrix can intuitively show the usage of resources, helping managers reasonably arrange production resources and reduce the idle time of equipment.

[0032] Optionally, in an embodiment of the present application, a Gantt chart is used to display the start time, end time, and occupied machine of each process of each order. The Gantt chart can intuitively show the production progress of the order and the resource usage. A bar chart or a line chart is used to display the utilization rate of each device. A bar chart or a line chart is used to display the number of changeovers in the thermal processing part. A bar chart or a line chart is used to display the total delay delivery time of all orders.

[0033] In the above intelligent production scheduling and dispatching system based on digital twin, the emission mode simulation log comparison and decision-making module 150 is used to determine whether to use the first mode or the second mode to arrange the set of orders to be scheduled based on the comparison between the first emission mode simulation log and the second emission mode simulation log. It should be understood that in the above intelligent production scheduling and dispatching system based on digital twin, it is crucial to compare the first emission mode simulation log and the second emission mode simulation log to determine which production scheduling method to use. This is because by comparing the simulation logs of the two production scheduling methods, the utilization rate of each device under different production scheduling methods can be evaluated. Selecting a production scheduling method with a higher utilization rate can maximize the use of existing resources and improve production efficiency.

[0034] Specifically, when comparing the first emission mode simulation log and the second emission mode simulation log, the technical concept of the present application is to extract the sequence distribution of equipment utilization rates from the simulation logs of different emission modes, and use data processing and analysis algorithms based on artificial intelligence and deep learning at the backend to analyze the sequence distribution data of the equipment utilization rates of these different emission modes, so as to capture the collaborative interaction and fusion characterization information between different equipment utilization rates, thereby optimizing the production scheduling method to determine which production scheduling method to use for production processing. In this way, the production scheduling and dispatching of the processing process can be realized in a more intelligent manner, thus ensuring the reliability of the production process and the scientific nature of decision-making, and being able to make full use of the existing equipment resources for product manufacturing.

[0035] Optionally, in an embodiment of the present application, Figure 3 is a schematic block diagram of an emission mode simulation log comparison and decision-making module in the digital twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application. Figure 4 is a schematic diagram of data flow of an emission mode simulation log comparison and decision-making module in the digital twin-based intelligent production scheduling and dispatching system according to an embodiment of the present application. As Figure 3 and Figure 4 shown, the emission mode simulation log comparison and decision-making module 150 includes: a first equipment utilization rate determination unit 151, configured to determine the utilization rate of each mechanical equipment based on the first emission mode simulation log to obtain a sequence distribution of the first equipment utilization rate; a second equipment utilization rate determination unit 152, configured to determine the utilization rate of each mechanical equipment based on the second emission mode simulation log to obtain a sequence distribution of the second equipment utilization rate; an equipment utilization rate sequence encoding unit 153, configured to perform sequence encoding on the sequence distribution of the first equipment utilization rate and the sequence distribution of the second equipment utilization rate respectively to obtain an implicit correlation feature vector between the first equipment utilization rates and an implicit correlation feature vector between the second equipment utilization rates; an equipment utilization rate collaborative interaction compensation and fusion unit 154, configured to perform feature interaction and fusion based on common features on the implicit correlation feature vector between the first equipment utilization rates and the implicit correlation feature vector between the second equipment utilization rates to obtain a representation of equipment utilization rate collaborative interaction compensation and fusion; a production scheduling mode optimization unit 155, configured to optimize the production scheduling mode based on the representation of equipment utilization rate collaborative interaction compensation and fusion to determine an optimization result, where the optimization result is used to represent adopting the first mode or the second mode.

[0036] In the above digital twin-based intelligent production scheduling and dispatching system, the first equipment utilization rate determination unit 151 and the second equipment utilization rate determination unit 152 are configured to determine the utilization rate of each mechanical equipment based on the first emission mode simulation log to obtain a sequence distribution of the first equipment utilization rate, and determine the utilization rate of each mechanical equipment based on the second emission mode simulation log to obtain a sequence distribution of the second equipment utilization rate. It should be understood that the equipment utilization rate is an important indicator for measuring production efficiency and resource utilization. By determining the utilization rate of each mechanical equipment, the effects of different production scheduling modes can be evaluated, and the optimal production scheduling plan can be selected.

[0037] Optionally, in an embodiment of the present application, the calculation method of the utilization rate of each machining equipment is: single-machine machining equipment utilization rate = actual production time of the machining equipment / (last production time of the machining equipment - production scheduling start time).

[0038] In the above intelligent production scheduling and dispatching system based on digital twin, the equipment utilization rate sequence encoding unit 153 is configured to perform sequence encoding on the sequence distribution of the first equipment utilization rate and the sequence distribution of the second equipment utilization rate respectively to obtain an implicit correlation feature vector between the first equipment utilization rates and an implicit correlation feature vector between the second equipment utilization rates. It should be understood that considering that when scheduling and dispatching processing generation tasks, there are mutual correlation relationships and complex correlation patterns among the utilization rates of various mechanical equipment in different emission modes. Based on this, in the technical solution of this application, the sequence distribution of the first equipment utilization rate and the sequence distribution of the second equipment utilization rate are further input into a sequence encoder based on a bidirectional LSTM model for feature mining, so as to extract the implicit correlation feature information between the equipment utilization rates in different ways of the two respectively, thereby obtaining an implicit correlation feature vector between the first equipment utilization rates and an implicit correlation feature vector between the second equipment utilization rates.

[0039] Optionally, in an embodiment of this application, the equipment utilization rate sequence encoding unit is configured to: input the sequence distribution of the first equipment utilization rate and the sequence distribution of the second equipment utilization rate into a sequence encoder based on a bidirectional LSTM model to obtain an implicit correlation feature vector between the first equipment utilization rates and an implicit correlation feature vector between the second equipment utilization rates.

[0040] In the above intelligent production scheduling and dispatching system based on digital twins, the device utilization collaborative interaction compensation and fusion unit 154 is used to perform feature interaction and fusion based on common features on the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates to obtain a device utilization collaborative interaction compensation and fusion representation. It should be understood that considering that the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates respectively contain the implicit correlation feature information of the device utilization rates of different production scheduling methods. When performing feature interaction and fusion and comparative analysis on the features of these two, each data source often carries unique information, and there is also a certain amount of redundant or overlapping information. Traditional feature fusion methods may ignore these unique key implicit correlation feature information of device utilization rates, resulting in an incomplete feature representation. In addition, there may be misalignment of the device utilization rate features between different feature sources, which will further affect the effect of feature fusion. Also, considering that although there are differences in the device utilization rates under different production scheduling methods, there are also some common features, and these common features reflect some basic laws and patterns of the device utilization rates under different production scheduling methods. Therefore, if you want to improve the accuracy of the subsequent production scheduling method optimization results, you can use the common features of the device utilization rates under these different production scheduling methods as the basis to compensate and optimize the expression of the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates respectively. Based on this, in the technical solution of this application, the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates are further subjected to feature interaction and fusion based on common features to obtain the device utilization collaborative interaction compensation and fusion representation. In particular, the feature interaction and fusion method based on common features can perform feature compensation and fusion in multi-modal or multi-perspective feature representations. By extracting the common features between the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates, and based on this, compensating for the unique information existing in the implicit correlation features between the device utilization rates under each production scheduling method, thereby realizing a richer and more comprehensive feature representation. In this way, not only can the common information of the device utilization rates under these two production scheduling methods be retained, but also the unique information in each feature source can be effectively supplemented, thereby improving the effect of feature fusion, enhancing the performance of the machine learning model, and providing a basis for subsequent production scheduling method optimization and dispatching.

[0041] Optionally, in an embodiment of the present application, the device utilization collaborative interaction compensation fusion unit includes: a common feature extraction secondary subunit, configured to input the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates into a common feature extraction network to obtain a common feature representation vector between the device utilization rate features; a device utilization rate feature probability processing secondary subunit, configured to input the implicit correlation feature vector between the first device utilization rates, the implicit correlation feature vector between the second device utilization rates, and the common feature representation vector between the device utilization rate features into a vector probability unit based on the Sigmoid function to obtain a probability-based implicit correlation feature vector between the first device utilization rates, a probability-based implicit correlation feature vector between the second device utilization rates, and a probability-based common feature representation vector between the device utilization rate features; a first information fine-grained compensation calculation secondary subunit, configured to calculate a first device utilization rate information fine-grained compensation vector of the probability-based implicit correlation feature vector between the first device utilization rates relative to the probability-based common feature representation vector between the device utilization rate features; a second information fine-grained compensation calculation secondary subunit, configured to calculate a second device utilization rate information fine-grained compensation vector of the probability-based implicit correlation feature vector between the second device utilization rates relative to the probability-based common feature representation vector between the device utilization rate features; a first fine-grained compensation secondary subunit, configured to input the first device utilization rate information fine-grained compensation vector and the implicit correlation feature vector between the first device utilization rates into a fine-grained compensation module to obtain a first device utilization rate fine-grained feature compensation vector; a second fine-grained compensation secondary subunit, configured to input the second device utilization rate information fine-grained compensation vector and the implicit correlation feature vector between the second device utilization rates into the fine-grained compensation module to obtain a second device utilization rate fine-grained feature compensation vector; a vector concatenation secondary subunit, configured to concatenate the first device utilization rate fine-grained feature compensation vector, the second device utilization rate fine-grained feature compensation vector, and the common feature representation vector between the device utilization rate features to obtain a device utilization collaborative interaction compensation fusion representation vector as the device utilization collaborative interaction compensation fusion representation.

[0042] Optionally, in an embodiment of the present application, the common feature extraction secondary subunit is configured to: calculate the position-wise sum between the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates to obtain a device utilization rate correlation fusion feature vector; multiply the device utilization rate correlation fusion feature vector by a modulation weight matrix and then perform a position-wise addition with a modulation bias vector, and input the obtained modulated device utilization rate correlation fusion feature vector into a tanh function for activation processing to obtain the common feature representation vector between the device utilization rate features.

[0043] Optionally, in an embodiment of the present application, the first information fine-grained compensation calculation secondary subunit is configured to: calculate the position-wise division between the implicit association feature vector between the probabilized first device utilization rates and the common feature representation vector between the probabilized device utilization rate features to obtain a first device utilization rate common feature interaction feature vector; after calculating the base-2 logarithmic function values of the absolute values of the feature values at each position in the first device utilization rate common feature interaction feature vector, multiply the obtained first device utilization rate common feature interaction logarithmic representation vector and the implicit association feature vector between the probabilized first device utilization rates position-wise to obtain a first device utilization rate information optimized representation vector; calculate the natural exponential function values with the feature values at each position in the first device utilization rate information optimized representation vector as the exponential powers with the natural constant e as the base, and then input the obtained feature vector into the softmax function for processing to obtain the first device utilization rate information fine-grained compensation vector.

[0044] Optionally, in an embodiment of the present application, the first fine-grained compensation secondary subunit is configured to: calculate the position-wise dot product between the first device utilization rate information fine-grained compensation vector and the implicit association feature vector between the first device utilization rates to obtain the first device utilization rate fine-grained feature compensation vector.

[0045] In summary, in the embodiment of the present application, the following feature interaction fusion formula is used to perform feature interaction fusion based on common features on the implicit association feature vector between the first device utilization rates and the implicit association feature vector between the second device utilization rates to obtain the device utilization rate collaborative interaction compensation fusion representation; where the feature interaction fusion formula is: ; where and are respectively the implicit association feature vector between the first device utilization rates and the implicit association feature vector between the second device utilization rates, and are respectively the modulation weight matrix and the modulation bias vector, is function, is the common feature representation vector between the device utilization rate features, is function, , and are respectively the common feature representation vector between the probabilized device utilization rate features, the implicit association feature vector between the probabilized first device utilization rates, and the implicit association feature vector between the probabilized second device utilization rates, , and They are respectively the position feature values in the common feature representation vector between the probabilistic device utilization rate features, the implicit correlation feature vector between the probabilistic first device utilization rates, and the implicit correlation feature vector between the probabilistic second device utilization rates. is the value of the natural exponential function with the natural constant e as the base. is a function. is the position feature value in the fine-grained compensation vector of the first device utilization rate information. is the position feature value in the fine-grained compensation vector of the second device utilization rate information. and are respectively the fine-grained compensation vector of the first device utilization rate information and the fine-grained compensation vector of the second device utilization rate information. is dot product by position. represents addition by position. and are respectively the fine-grained feature compensation vector of the first device utilization rate and the fine-grained feature compensation vector of the second device utilization rate. is a concatenation operation. is the collaborative interaction compensation fusion representation vector of the device utilization rate.

[0046] Specifically, the feature interaction and fusion process based on the common feature includes the steps: First, through the common feature extraction network, a neural network architecture such as a fully connected layer is used to learn the common features and internal relationships between the implicit correlation feature vector between the first device utilization rates and the implicit correlation feature vector between the second device utilization rates. These common features reflect some basic laws and patterns of the device utilization rates under different scheduling methods. Retaining the common information helps to ensure the consistency of the final feature representation, avoid decision-making biases caused by feature loss, and provide a basis for subsequent feature compensation and scheduling optimization tasks.

[0047] Next, the implicit correlation feature vector between the first device utilization rates, the implicit correlation feature vector between the second device utilization rates, and the common feature representation vector between the device utilization rate features are input into the vector probability unit based on the Sigmoid function. The Sigmoid function is an activation function that can map real values to the interval (0,1). In this step, the role of the Sigmoid function is to normalize each feature vector, so that each position feature value in each feature vector is converted into a probability form, thus better measuring the importance of the features. That is, the Sigmoid function is used to probabilize the feature vector. The probabilized feature vector can more intuitively reflect which device utilization rate correlation features are significant and which are secondary, and helps to more comprehensively describe the correlation patterns of the utilization rates among various mechanical devices.

[0048] Then, considering that the variation patterns of equipment utilization rates may be different under different production scheduling methods, and each method has its unique information. Therefore, further calculate the first equipment utilization rate information fine-grained compensation vector of the implied correlation feature vector between the probabilistic first equipment utilization rates with respect to the common feature representation vector between the probabilistic equipment utilization rate features. This process finds the unique information about the implied correlation feature vector between the first equipment utilization rates in addition to the common features by calculating the difference between the two probabilistic feature vectors, and this information can contribute to subsequent production scheduling and scheduling tasks. At the same time, calculate the second equipment utilization rate information fine-grained compensation vector of the implied correlation feature vector between the probabilistic second equipment utilization rates with respect to the common feature representation vector between the probabilistic equipment utilization rate features. Similar to the former, the goal is to identify the information in the implied correlation feature vector between the second equipment utilization rates that is not fully covered by the common features. In this way, it can be ensured that as much information as possible about the implied correlation features and unique information of the equipment utilization rates for different production scheduling tasks is retained in the final feature fusion process.

[0049] Subsequently, input the first equipment utilization rate information fine-grained compensation vector and the implied correlation feature vector between the first equipment utilization rates into the fine-grained compensation module to obtain the first equipment utilization rate fine-grained feature compensation vector. And input the second equipment utilization rate information fine-grained compensation vector and the implied correlation feature vector between the second equipment utilization rates into the fine-grained compensation module to obtain the second equipment utilization rate fine-grained feature compensation vector. The fine-grained compensation module is usually a mechanism designed to strengthen or supplement specific features. Here, the module receives the original implied correlation feature vector between the equipment utilization rates and its corresponding compensation vector, and in some way, such as multiplication operation, adds the compensation vector to the original feature vector to generate a new feature vector containing more detailed information, ensuring that the final feature representation retains both the common information between the two and the unique information, guaranteeing the consistency and integrity of the final feature fusion.

[0050] Finally, cascade the first equipment utilization rate fine-grained feature compensation vector, the second equipment utilization rate fine-grained feature compensation vector, and the common feature representation vector between the equipment utilization rate features to obtain the equipment utilization rate collaborative interaction compensation fusion representation vector as the equipment utilization rate collaborative interaction compensation fusion representation. This step creates a comprehensive feature representation by combining all the compensated feature vectors. In particular, the cascading operation may be a simple vector concatenation or a more complex fusion mechanism, such as an attention mechanism, etc., to ensure that the final feature representation retains both the common features and the unique information in the two input feature vectors. This comprehensive feature representation can more comprehensively reflect the changes in equipment utilization rates under different production scheduling methods, improve the effect of feature fusion, and provide a basis for subsequent production scheduling tasks.

[0051] In the above intelligent production scheduling and dispatching system based on digital twin, the production scheduling method optimization unit 155 is configured to optimize the production scheduling method based on the collaborative interaction compensation fusion representation of the equipment utilization rate to determine an optimization result, and the optimization result is used to represent adopting a first method or a second method. Optionally, in an embodiment of the present application, the production scheduling method optimization unit is configured to: input the collaborative interaction compensation fusion representation vector of the equipment utilization rate into a production scheduling method optimization module based on a classifier to obtain an optimization result, and the optimization result is used to represent adopting a first method or a second method. That is, two collaborative interaction compensation fusion feature representations of the equipment utilization rate are used for classification processing to optimize the production scheduling method. Further, in the technical solution of the present application, not only the two production scheduling methods listed above need to be considered, but also other possible production scheduling methods (such as the production difficulty of the order, the value of the order, the resource requirements of the order, etc.) need to be considered. By comparing all possible production scheduling methods in the above manner, it is determined which production scheduling method is optimal for production and processing. In this way, the production scheduling and dispatching of the processing process can be realized in a more intelligent manner, thereby ensuring the reliability of the production process and the scientific nature of the decision-making, and being able to make full use of the existing equipment resources for product manufacturing.

[0052] Optionally, in an embodiment of the present application, inputting the collaborative interaction compensation fusion representation vector of the equipment utilization rate into a production scheduling method optimization module based on a classifier to obtain an optimization result includes: using the fully connected layer of the classifier to perform fully connected encoding on the collaborative interaction compensation fusion representation vector of the equipment utilization rate to obtain a fully connected collaborative interaction compensation fusion representation vector of the equipment utilization rate; inputting the fully connected collaborative interaction compensation fusion representation vector of the equipment utilization rate into the Softmax classification function of the classifier to obtain the probability values of the collaborative interaction compensation fusion representation vector of the equipment utilization rate belonging to each classification label, where the classification labels include a first classification label for representing adopting the first method and a second classification label for representing adopting the second method; and determining the classification label corresponding to the largest of the probability values as the optimization result.

[0053] When the implicit correlation feature vector between the first equipment utilization rates and the implicit correlation feature vector between the second equipment utilization rates respectively represent the short-range-long-range bidirectional time series correlation features of the first equipment utilization rate and the second equipment utilization rate, during the feature-intermediate compensation interaction with the common feature as the mask, the sparsity between the common mask features corresponding to the multi-dimensional correlation differences of the time series features under the source time series distribution will cause the lack of local compensation interaction between the features, that is, it will cause the interaction distribution representation of the collaborative interaction compensation fusion representation vector of the equipment utilization rate to be simply repeated in the local interaction domain, affecting its semantic logical dependence and reducing the accuracy of the classification result.

[0054] Therefore, preferably, when the collaborative interaction compensation fusion representation vector of the device utilization rate is input into the scheduling method optimization module based on a classifier to obtain an optimization result, the collaborative interaction compensation fusion representation vector of the device utilization rate is optimized, specifically including: determining a collaborative interaction compensation fusion correlation response matrix and a collaborative interaction compensation fusion distance response matrix of the device utilization rate based on the square root of the correlation value and the L2 distance value of the eigenvalue of the collaborative interaction compensation fusion representation vector of the device utilization rate, that is: ; where and respectively represent the th eigenvalue and the th eigenvalue of the collaborative interaction compensation fusion representation vector of the device utilization rate, represents the collaborative interaction compensation fusion representation vector of the device utilization rate, represents the set of real numbers, represents the length of the collaborative interaction compensation fusion representation vector of the device utilization rate, represents the eigenvalue at the position in the collaborative interaction compensation fusion correlation response matrix of the device utilization rate, represents the eigenvalue at the position in the collaborative interaction compensation fusion distance response matrix of the device utilization rate.

[0055] Multiply the collaborative interaction compensation fusion representation vector of the device utilization rate with the collaborative interaction compensation fusion correlation response matrix of the device utilization rate to obtain a collaborative interaction compensation fusion correlation response vector of the device utilization rate.

[0056] Multiply the collaborative interaction compensation fusion distance response matrix of the device utilization rate with the transposed vector of the collaborative interaction compensation fusion representation vector of the device utilization rate to obtain a collaborative interaction compensation fusion distance response vector of the device utilization rate.

[0057] After performing a dot product on the collaborative interaction compensation fusion correlation response matrix and the collaborative interaction compensation fusion distance response matrix of the device utilization rate, multiply the result with the transposed vector of the collaborative interaction compensation fusion representation vector of the device utilization rate to obtain a collaborative interaction compensation fusion bias vector of the device utilization rate.

[0058] Perform a dot sum on the collaborative interaction compensation fusion correlation response vector, the collaborative interaction compensation fusion distance response vector, and the collaborative interaction compensation fusion bias vector of the device utilization rate to obtain an optimized collaborative interaction compensation fusion characterization vector of the device utilization rate.

[0059] Here, the optimized collaborative interaction compensation fusion characterization vector of the device utilization rate Expressed as: wherein, and : wherein, represents the device utilization collaborative interaction compensation fusion correlation response matrix, represents the device utilization collaborative interaction compensation fusion distance response matrix, represents the device utilization collaborative interaction compensation fusion representation vector, represents the set of real numbers, represents the length of the device utilization collaborative interaction compensation fusion representation vector, represents matrix multiplication, represents element-wise multiplication, represents element-wise addition, represents the transpose of a vector, represents the device utilization collaborative interaction compensation fusion correlation response vector, represents the device utilization collaborative interaction compensation fusion distance response vector, represents the device utilization collaborative interaction compensation fusion logic bias vector, represents the optimized device utilization collaborative interaction compensation fusion representation vector.

[0060] That is to say, by using the device utilization collaborative interaction compensation fusion representation vector as its statistical no-reference distribution response framework based on the self-correlation matrix and the self-distance matrix, avoiding simple repetition of feature distributions through the construction of a retrieval-enhanced reverse response for the device utilization collaborative interaction compensation fusion representation vector, and avoiding its surface combinatorial mapping by ensuring the internal mapping logic of the device utilization collaborative interaction compensation fusion representation vector based on the retrieval-response context, so as to realize the semantic logic dependence mapping of the device utilization collaborative interaction compensation fusion representation vector while maintaining the intuitive response relevance, thereby improving the accuracy of the optimization result obtained by inputting the device utilization collaborative interaction compensation fusion representation vector into the classifier-based production scheduling optimization module. In this way, the production scheduling and dispatching of the processing process can be realized in a more intelligent manner, thus ensuring the reliability of the production process and the scientificity of decision-making.

[0061] Optionally, in another embodiment of the present application, an optimization solver black box optimization framework can also be used to optimize the production scheduling method of the device utilization collaborative interaction compensation fusion representation vector to determine the optimization result. Among them, the algorithm framework adopts Bayesian optimization based on the Gaussian regression process, and an approximate solution of a complex objective function can be obtained with fewer evaluation times. This software adopts advanced black box optimization technology and can optimize various industrial production problems without data modeling, greatly promoting the cost reduction and efficiency improvement of enterprises.

[0062] The basic optimization process of the optimization solver includes: The surrogate model is a machine learning model trained with data from the task database and is used to approximate the simulation process; the acquisition function is used to provide the search direction; for sequential optimization, local search mainly adopts the neighborhood search algorithm. Since it takes a large amount of time to calculate the objective function for one simulation, a surrogate model is used to approximate the simulation. After generating an objective function for each simulation, the objective function and the corresponding solution are updated to the database, and then the surrogate model is updated. Based on the updated surrogate model, local search can solve for a new solution with a higher probability of being optimal. After the new solution is simulated, the true objective function is obtained. By continuously iterating in this way, good optimization results can be obtained with fewer iterations.

[0063] In summary, the intelligent production scheduling and dispatching system based on digital twin according to the embodiments of the present application is described. After arranging the set of orders to be scheduled into two sequences of orders to be scheduled by adopting different sorting strategies, production scheduling simulations are respectively performed on these two sequences of orders to be scheduled by using the resource-time matrix. By simulating the production process under different production scheduling methods, corresponding simulation logs are generated, and then based on the comparison between the simulation logs, the optimized production scheduling method is determined for corresponding scheduling and management. In this way, potential problems can be discovered in advance through virtual simulation in digital twin technology, the production scheduling method can be optimized in a timely manner, and the production plan can be adjusted in a timely manner to avoid losses in actual production.

[0064] Figure 5 This is a schematic flowchart of the intelligent production scheduling and dispatching method based on digital twin according to the embodiments of the present application. As Figure 5 shown, the intelligent production scheduling and dispatching method based on digital twin includes: S1, obtaining a set of orders to be scheduled; S2, arranging the set of orders to be scheduled in a first manner to obtain a sequence of the first orders to be scheduled and arranging the set of orders to be scheduled in a second manner to obtain a sequence of the second orders to be scheduled; S3, setting the production scheduling time and performing production scheduling on the sequence of the first orders to be scheduled and the sequence of the second orders to be scheduled based on the resource-time matrix to obtain a first production scheduling method simulation log and a second production scheduling method simulation log; S4, displaying the first production scheduling method simulation log and the second production scheduling method simulation log; S5, based on the comparison between the first production scheduling method simulation log and the second production scheduling method simulation log, determining whether to arrange the set of orders to be scheduled in the first manner or the second manner.

[0065] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned intelligent production scheduling and dispatching method based on digital twin have been described above with reference to Figures 1 to 4It has been introduced in detail in the description of the intelligent production scheduling and dispatching system based on digital twin, and therefore, its repeated description will be omitted.

[0066] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and easy understanding, rather than limitations. The above details do not limit the present invention to necessarily adopt the above specific details for implementation.

[0067] In the above embodiments, the descriptions of each embodiment have their own focuses. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the module division is only a logical function division, and there can be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0068] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0069] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any associated drawing marks in the claims should not be regarded as limiting the claimed rights.

Claims

1. An intelligent production scheduling and dispatching system based on digital twins, characterized in that: include: A module for obtaining a set of orders to be scheduled, used to obtain a set of orders to be scheduled; A module for arranging a set of orders to be arranged, configured to arrange the set of orders to be arranged in a first manner to obtain a first sequence of orders to be arranged and to arrange the set of orders to be arranged in a second manner to obtain a second sequence of orders to be arranged; A production scheduling module for a sequence of pending orders, used for setting a production scheduling time and scheduling the sequence of the first pending orders and the sequence of the second pending orders based on a resource-time matrix to obtain a first emission mode simulation log and a second emission mode simulation log; An emission mode simulation log display module, used for displaying the first emission mode simulation log and the second emission mode simulation log; An emission mode simulation log comparison decision module, used for determining whether to arrange the set of to-be-arranged orders in a first mode or a second mode based on a comparison between the first emission mode simulation log and the second emission mode simulation log; Among them, the emission mode simulation log comparison decision module includes: a first equipment utilization determination unit, which is used to determine the utilization of each mechanical equipment based on the first emission mode simulation log to obtain the sequence distribution of the first equipment utilization; a second equipment utilization determination unit, which is used to determine the utilization of each mechanical equipment based on the second emission mode simulation log to obtain the sequence distribution of the second equipment utilization; an equipment utilization sequence encoding unit, which is used to sequence encode the sequence distribution of the first equipment utilization and the sequence distribution of the second equipment utilization respectively to obtain the implicit correlation feature vector between the first equipment utilization and the implicit correlation feature vector between the second equipment utilization; an equipment utilization collaborative interaction compensation fusion unit, which is used to perform feature interaction fusion based on common features on the implicit correlation feature vector between the first equipment utilization and the implicit correlation feature vector between the second equipment utilization to obtain the equipment utilization collaborative interaction compensation fusion representation; a production scheduling optimization unit, which is used to optimize the production scheduling based on the equipment utilization collaborative interaction compensation fusion representation to determine the optimization result, and the optimization result is used to indicate whether to adopt the first method or the second method.

2. The intelligent production scheduling and dispatching system based on digital twin according to claim 1 is characterized in that: The production scheduling module for the sequence of pending orders includes: a resource-time matrix initialization unit, used to initialize the resource-time matrix, wherein the horizontal axis of the resource-time matrix represents the resources required for production, and the vertical axis of the resource-time matrix represents the time axis; an order information extraction unit, used to extract the order information of the current order from the sequence of the first pending orders; an order detailed information parsing unit, used to extract the bill of materials, processes, equipment and processing time of each process required for the current order from the order information of the current order; an order process scheduling unit, used to find the earliest start time of the earliest process that satisfies the current order in the resource-time matrix based on the processing sequence of the processes of the current order, record the processing time of the earliest process, and occupy this period of time in the resource-time matrix.

3. The intelligent production scheduling and dispatching system based on digital twin according to claim 2 is characterized in that: The equipment utilization sequence encoding unit is used to: input the sequence distribution of the first equipment utilization and the sequence distribution of the second equipment utilization into a sequence encoder based on a bidirectional LSTM model to obtain an implicit correlation feature vector between the first equipment utilizations and an implicit correlation feature vector between the second equipment utilizations.

4. The intelligent production scheduling and dispatching system based on digital twin according to claim 3 is characterized in that: The equipment utilization collaborative interaction compensation fusion unit includes: a common feature extraction secondary subunit, which is used to input the implicit correlation feature vector between the first equipment utilizations and the implicit correlation feature vector between the second equipment utilizations into a common feature extraction network to obtain a common feature representation vector between equipment utilization features; an equipment utilization feature probabilistic processing secondary subunit, which is used to input the implicit correlation feature vector between the first equipment utilizations, the implicit correlation feature vector between the second equipment utilizations and the common feature representation vector between the equipment utilization features into a vector probabilization unit based on a Sigmoid function to obtain a probabilistic first equipment utilization implicit correlation feature vector, a probabilistic second equipment utilization implicit correlation feature vector and a probabilistic common feature representation vector between equipment utilization features; a first information fine-grained compensation calculation secondary subunit, which is used to calculate the first equipment utilization information fine-grained compensation vector of the probabilistic first equipment utilization implicit correlation feature vector relative to the probabilistic common feature representation vector between equipment utilization features; and a second information fine-grained processing secondary subunit. A compensation calculation secondary subunit is used to calculate the second device utilization information fine-grained compensation vector of the implicit correlation feature vector between the probabilistic second device utilizations relative to the common feature representation vector between the probabilistic device utilization features; a first fine-grained compensation secondary subunit is used to input the first device utilization information fine-grained compensation vector and the first device utilization implicit correlation feature vector into the fine-grained compensation module to obtain the first device utilization fine-grained feature compensation vector; a second fine-grained compensation secondary subunit is used to input the second device utilization information fine-grained compensation vector and the second device utilization implicit correlation feature vector into the fine-grained compensation module to obtain the second device utilization fine-grained feature compensation vector; a vector cascade secondary subunit is used to cascade the first device utilization fine-grained feature compensation vector, the second device utilization fine-grained feature compensation vector and the common feature representation vector between the device utilization features to obtain the device utilization collaborative interaction compensation fusion representation vector as the device utilization collaborative interaction compensation fusion representation.

5. The intelligent production scheduling and dispatching system based on digital twin according to claim 4 is characterized in that: The common feature extraction secondary subunit is used to: calculate the positional sum between the implicit correlation feature vector between the first device utilizations and the implicit correlation feature vector between the second device utilizations to obtain a device utilization correlation fusion feature vector; multiply the device utilization correlation fusion feature vector with the modulation weight matrix and then add it with the modulation bias vector in position, and input the obtained modulated device utilization correlation fusion feature vector into a tanh function for activation processing to obtain a common feature representation vector between the device utilization features.

6. The intelligent production scheduling and dispatching system based on digital twin according to claim 5 is characterized in that: The first information fine-grained compensation calculation secondary subunit is used to: calculate the positional division between the implicit correlation feature vector between the probabilistic first device utilizations and the common feature representation vector between the probabilistic device utilization features to obtain the first device utilization common feature interaction feature vector; After calculating the logarithmic function value with base 2 of the absolute value of the feature value at each position in the common feature interaction feature vector of the first device utilization, the obtained first device utilization common feature interaction logarithmic representation vector and the probabilistic first device utilization implicit correlation feature vector are multiplied by position to obtain the first device utilization information optimization representation vector; after calculating the natural exponential function value with base e as the natural constant and the feature value at each position in the first device utilization information optimization representation vector as the exponential power, the obtained feature vector is input into the softmax function for processing to obtain the first device utilization information fine-grained compensation vector.

7. The intelligent production scheduling and dispatching system based on digital twin according to claim 6 is characterized in that: The first fine-grained compensation secondary subunit is used to calculate the first device utilization information fine-grained compensation vector and the first device utilization implicit association feature vector by point-by-point multiplication to obtain the first device utilization fine-grained feature compensation vector.

8. The intelligent production scheduling and dispatching system based on digital twin according to claim 7 is characterized in that: The production scheduling optimization unit is used to: input the equipment utilization collaborative interaction compensation fusion representation vector into the classifier-based production scheduling optimization module to obtain an optimization result, and the optimization result is used to indicate whether the first method or the second method is adopted.

9. An intelligent production scheduling and dispatching method based on digital twins, using the intelligent production scheduling and dispatching system based on digital twins according to claim 1, characterized in that: include: Get the collection of pending orders; Arranging the set of to-be-arranged orders in a first manner to obtain a first sequence of to-be-arranged orders and arranging the set of to-be-arranged orders in a second manner to obtain a second sequence of to-be-arranged orders; Set the scheduling time and schedule the first sequence of orders to be scheduled and the second sequence of orders to be scheduled based on the resource-time matrix to obtain a first emission mode simulation log and a second emission mode simulation log; display the first emission mode simulation log and the second emission mode simulation log; based on the comparison between the first emission mode simulation log and the second emission mode simulation log, determine whether to use the first method or the second method to arrange the set of orders to be scheduled.

Images