A method for collaborative scheduling of production-logistics of construction machines

Abstract

This invention relates to the field of production scheduling technology, and more particularly to a collaborative scheduling method for engineering machinery production and logistics, comprising: constructing a production logistics diagram; reading relationship records from the production flow diagram through a scheduling program, executing corresponding scheduling measures based on the relationship records and generating production urgency and logistics accessibility, calculating collaborative scheduling intensity, constructing a collaborative intensity matrix and a time series prediction model, calculating predicted scheduling intensity, fusing the predicted scheduling intensity with the collaborative scheduling intensity to generate a final predicted scheduling intensity, and writing the final predicted scheduling intensity into the predicted scheduling matrix by the scheduling server; using the predicted scheduling matrix as an index, the scheduling server locates the field entities corresponding to the relationship records one by one, calculates the executability, calculates the final execution score based on the resource executability, sorts the final execution scores to generate an execution sequence, and organizes the execution sequences into a set of scheduling schemes after conflict resolution.

Description

Technical Field

[0001] This invention relates to the field of production scheduling technology, and more particularly to a collaborative scheduling method for the production and logistics of engineering machinery. Background Technology

[0002] With the rapid development of the modern construction machinery manufacturing industry, the coordinated scheduling of production and logistics has received increasing attention. In the production process of construction machinery, efficient coordination between production resources and logistics resources is crucial to ensuring production capacity and resource utilization. However, current production-logistics scheduling systems face a series of challenges, especially in the manufacturing of large-scale construction machinery, where the coupling relationships between resources such as production equipment, warehousing facilities, and transportation vehicles are highly complex, and production tasks and logistics demands are often dynamically changing.

[0003] Most existing scheduling systems rely on traditional optimization methods, typically using linear or heuristic algorithms to solve scheduling problems. These methods often exhibit low flexibility and efficiency when facing highly complex and dynamically changing production environments. For example, traditional scheduling systems often assume that production processes and logistics routes are fixed, making them difficult to cope with sudden events or changes in the real-time environment, such as equipment failures, raw material shortages, and traffic congestion. This results in the system's inability to respond quickly and frequently requires manual intervention to adjust scheduling strategies.

[0004] Furthermore, existing methods have limited ability to coordinate multiple resources, often failing to achieve global optimization in dynamic production environments, resulting in suboptimal resource utilization and scheduling efficiency. Most existing optimization methods are based on the scheduling of single resources, neglecting the interdependencies between resources and failing to effectively coordinate the scheduling of production and logistics resources, leading to reduced system efficiency. To address these issues, a new scheduling optimization method is urgently needed to overcome the shortcomings of existing systems, better cope with complex production and logistics environments, improve resource utilization, and enhance scheduling efficiency. Summary of the Invention

[0005] To address the above problems, this invention provides a collaborative scheduling method for engineering machinery production and logistics.

[0006] To achieve the above objectives, this invention proposes a collaborative scheduling method for engineering machinery production and logistics, comprising:

[0007] The production equipment nodes, warehousing nodes, and transportation equipment nodes are unified into a node set. An edge set of the node set is constructed. The node set and edge set are written into the same graph structure to generate a production logistics graph.

[0008] The scheduler reads the relationship records of the production process diagram, executes corresponding scheduling measures based on the relationship records, and generates production urgency and logistics accessibility for each relationship record. Based on the production urgency and logistics accessibility, the coordinated scheduling intensity corresponding to each relationship record is calculated by combining the coordinated function, and a coordinated intensity matrix is ​​constructed based on the coordinated scheduling intensity.

[0009] A time series prediction model is constructed. The predicted scheduling intensity is calculated for each relation record in the coordination intensity matrix using the time series prediction model. The predicted scheduling intensity is then fused with the coordination scheduling intensity to generate the final predicted scheduling intensity. The scheduling server writes the final predicted scheduling intensity corresponding to all relation records into the prediction scheduling matrix.

[0010] The scheduling server uses the predicted scheduling matrix as an index to locate the field entities corresponding to the relationship records one by one, calculates the resource executability of the field entities, calculates the final execution score based on the resource executability, sorts the final execution scores to generate execution sequences, resolves conflicts in the execution sequences, and organizes the conflict-resolved execution sequences into a scheduling scheme set.

[0011] In some embodiments, the edge set includes records of process connection relationships, material transfer relationships, and transportation resource occupancy relationships in the manufacturing scenario.

[0012] In some embodiments, the execution of corresponding scheduling measures based on the relationship record specifically includes:

[0013] If the relationship record represents two production equipment nodes, the program extracts the completion progress of the preceding process, the release status of the subsequent process, and the input status of the target equipment corresponding to the relationship record from the order process table, process progress table, and equipment availability status table in the MES.

[0014] If the relationship records are warehouse nodes and production equipment nodes, the program extracts material delivery request information, the existence of accessible vehicles, and the availability of origin and destination receiving and dispatching conditions from the transportation task table, vehicle location table, and warehouse status table.

[0015] If the relationship record is between a production equipment node and a transportation equipment node, the program extracts the connectivity of the transportation equipment within the current scheduling cycle from the vehicle task queue and the task binding table.

[0016] In some embodiments, the parameters of the cooperative function include logistics accessibility, a balance coefficient between the comprehensive term and the cooperative term, and a difference penalty coefficient.

[0017] In some embodiments, the time series prediction model includes a sequence embedding layer, a gated recurrent unit network, and a linear mapping layer.

[0018] In some embodiments, the final predicted scheduling intensity is generated by a prediction function, the parameters of which include the collaborative scheduling intensity of the current period, the predicted scheduling intensity generated by the time series prediction model, the actual scheduling intensity of the historical period, the prediction fusion coefficient, and the fluctuation suppression coefficient.

[0019] In some embodiments, the resource executability is calculated and generated by an execution function, the parameters of which include the resource availability ratio, the resource acceptance ratio, and the connection feasibility ratio of the link corresponding to the relationship record.

[0020] In some embodiments, the final execution score is generated by calculating a scoring function, the parameters of which include predicted scheduling intensity, resource executability, resource occupancy intensity, and resource occupancy penalty coefficient.

[0021] In some embodiments, the scheduling scheme set includes production equipment number, warehouse node number, transportation equipment number, and execution sequence issuance order.

[0022] In some embodiments, before unifying the production equipment nodes, warehousing nodes, and transportation equipment nodes into a node set, the method further includes:

[0023] The scheduler extracts the devices that have been included in the scheduling scope and maps these devices to production equipment nodes.

[0024] The location of the line-side warehouse, buffer area, and entrance / exit of the vertical warehouse is extracted by the scheduler, and the location is mapped to a warehouse node.

[0025] The scheduling program extracts the transportation equipment numbers of AGVs, forklifts, and handling vehicles, and maps these transportation equipment numbers to transportation equipment nodes.

[0026] The beneficial effects of this invention are as follows:

[0027] This invention, by collecting real-time operational data from production and logistics systems, enables the system to rapidly adjust its scheduling strategy in the event of dynamic disturbances, ensuring seamless integration of production and logistics. Secondly, by introducing a collaborative optimization mechanism for production and logistics resources, this invention fully considers the interdependencies between resources during multi-resource scheduling, optimizing production task allocation and logistics route selection. This results in closer coordination between production and logistics transportation, reducing idle time and improving overall resource utilization efficiency. Simultaneously, the scheduling method of this invention adaptively optimizes the matching of production and logistics routes in real-time adjustments, providing a more efficient and flexible solution for scheduling between production equipment and transportation vehicles. These innovations directly address key issues in existing technologies, such as insufficient multi-resource coordination, slow dynamic scheduling response, and weak global optimization capabilities, thereby significantly improving the efficiency and reliability of production-logistics collaborative scheduling. Attached Figure Description

[0028] Figure 1 This is a flowchart of a collaborative scheduling method for engineering machinery production and logistics in a specific embodiment of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] refer to Figure 1 As shown in the embodiment of this application, a collaborative scheduling method for engineering machinery production and logistics is proposed, including:

[0031] S1: Unify the production equipment nodes, warehousing nodes, and transportation equipment nodes into a node set, construct the edge set of the node set, and write the node set and edge set into the same graph structure to generate a production logistics graph, specifically including:

[0032] In the construction machinery manufacturing scenario, production equipment, storage locations, and transportation equipment are managed by equipment control systems, storage management systems, and logistics scheduling systems, respectively. If the original data organization methods of each system are directly used, the scheduling program can only see scattered equipment status, inventory locations, and transportation tasks, and cannot simultaneously identify the linkage relationships between the three types of constraints on the same object: "which process must be done first," "which material must be delivered first," and "which transportation equipment is already occupied." Therefore, we first organize the original on-site operating data into a unified production-logistics diagram. The raw data is aggregated by the workshop data acquisition server according to the scheduling cycle: production-side data comes from the control interfaces of equipment such as welding robot controllers, assembly station PLCs, and CNC machining unit controllers, which continuously output equipment numbers, current process numbers, equipment occupancy status, and corresponding order numbers; logistics-side data comes from the AGV scheduling server, forklift task terminals, and automated warehouse control systems, which store transportation equipment numbers, current location numbers, transportation task numbers, and target location numbers; task-side data comes from the MES system, which records order numbers, process sequences, optional equipment for each process, and material delivery requests. After aligning the above records with their completion times, the acquisition server writes them into the scheduling database. The scheduling program directly reads these records within the current scheduling cycle and generates nodes and relationships in the graph accordingly. The purpose of this processing is to transform the data, which was originally scattered according to system boundaries, into a data structure organized according to scheduling objects, enabling subsequent steps to simultaneously perform production and logistics analysis around the same graph object.

[0033] The first construction step in the graph is to unify the node set. This uses the basic expression of set union in discrete mathematics, derived from the definition in set theory that "the overall set is obtained by the union of several subsets." Based on this, the three types of entities in the manufacturing site are mapped to three subsets, and then a unified node set is formed through union. Specifically, the scheduler extracts the equipment numbers of devices already within the scheduling range from the equipment control system and maps each device to a production equipment node; it extracts the location numbers of line-side warehouses, buffer areas, and automated warehouse entrances / exits from the warehouse management system and maps each location to a warehouse node; it extracts the numbers of transportation equipment such as AGVs, forklifts, and pallet trucks from the logistics scheduling system and maps each transportation equipment to a transportation equipment node. After mapping, the unified node set is generated using the following formula:

[0034] ;

[0035] In the formula, This is the set of all nodes within the current scheduling period, and its value is generated by the scheduler from the database. This is a set of production equipment nodes, whose values ​​come from a list of equipment numbers within the scheduling range of the equipment control system. This is a set of warehouse nodes, whose values ​​come from the storage location number and buffer zone number in the warehouse management system; This is a set of transportation equipment nodes, whose values ​​are derived from the list of vehicle numbers in the logistics scheduling system.

[0036] This further specifies that the three types of elements correspond to equipment, warehousing, and transportation equipment, respectively, thus transforming the general set definition into a node generation rule oriented towards the manufacturing site. Since both sides of the formula represent set objects—the left side being a unified set of nodes and the right side being the result of merging three node subsets—the structural meaning remains consistent. Taking a chassis assembly workshop as an example, if a scheduling cycle reads 25 production equipment units (welding, machining, final assembly, etc.), 8 locations (line-side buffer area and automated warehouse entrance / exit), and 12 AGVs and forklifts, then... There are 25 nodes. There are 8 nodes. There are 12 nodes, after merging It contains a total of 45 nodes. This 45 is not a manually set number, but rather the actual merged result of deduplicating the three lists of numbers in the database within the current period. If a transportation device happens to be offline and not within the scheduling scope of this period, its number will not be included. It will not enter the final stage. .

[0037] After unifying the nodes, it is necessary to establish the edges between them, thus transforming isolated nodes into a graph structure usable for scheduling analysis. Here, the ordered binary representation from graph theory is used, its original form derived from the classical graph theory definition that "a graph consists of a set of nodes and a set of edges." Based on this definition, the edge set... The records are broken down into three categories: process connection relationships, material transfer relationships, and transportation resource occupancy relationships within the manufacturing scenario, and then uniformly written into the same graph structure. The corresponding expressions are as follows:

[0038] ;

[0039] In the formula, This is a production-logistics graph, whose values ​​are determined by the set of nodes and the set of relationships generated within the current scheduling cycle. This is the unified set of nodes already obtained in the previous formula; This is a set of relationships between nodes, and its value is generated by three parts of data: The first part comes from the process sequence records in the MES system, used to establish production relationship edges. For example, order O1 includes three processes: "material cutting - welding - assembly." The scheduler establishes a sequential connection relationship between the corresponding equipment nodes based on the process route and equipment allocation results of order O1. The second part comes from the transportation task records in the logistics scheduling system, used to establish logistics relationship edges. For example, if material M1 is transported from warehouse W3 to assembly station P8, the program establishes a material transfer relationship between the corresponding node of W3 and the corresponding node of P8. The third part comes from the binding records of transportation equipment and tasks, used to establish transportation resource occupancy relationships. For example, if transportation task T6 has been assigned to AGV-07, the program associates the AGV-07 node with the start and end link corresponding to T6, thus reflecting in the graph that the transportation equipment has been occupied by this link. The logical relationship between this formula and the previous formula is clear: the previous formula first gives "what objects are in the graph," and the second formula gives "how these objects are related." Only by obtaining the first formula can the relationship be established. Only then can it continue to generate Thus forming a complete In its derivation, it can be understood that the general graph definition in classical graph theory is first retained as the outer structure, and then the specific business relationships in the manufacturing site are written into the edge set. Therefore, it is not a simple copy of the graph theory expression, but a scenario-based rewrite from a general graph to a production-logistics collaboration graph. Since both sides of the equation represent graph structure objects, with the left side being the graph model and the right side being the paired structure of the node set and the relation set, their meanings are consistent.

[0040] Substituting the above rules into a specific embodiment provides a more intuitive understanding of the generation process of this step. Assume that within a certain scheduling cycle, the database contains three core tables: an equipment status table, a storage location table, and a transportation equipment table. The equipment status table filters out 25 equipment numbers within the scheduling range, corresponding to welding units, machining units, and assembly stations, respectively; the storage location table filters out 8 storage location and buffer area numbers; and the transportation equipment table filters out 12 AGV and forklift numbers. The scheduling program first generates a node set according to the above formula, therefore the final total number of nodes is 45. The program then reads the order process records from the MES system and finds that there are 10 orders participating in scheduling in the current period. Order O1 forms 2 process connections, order O2 forms 3 process connections, and the remaining 8 orders form a total of 15 process connections, resulting in a total of 20 production relationship edges. Next, it reads the logistics task table and finds that there are 18 material transfer tasks to be executed in the current period, resulting in a total of 18 logistics relationship edges. Finally, it reads the transportation equipment task binding table and finds that 12 transportation tasks have been clearly assigned to corresponding vehicles, resulting in a total of 12 transportation resource occupancy relationships. At this point, the edge set... The total number of relationship records is 50. After the scheduler writes 45 nodes and 50 relationships into the graph database or in-memory graph structure, the production-logistics graph for this cycle is obtained. Taking order O1 as an example, if its process route is "frame welding → hydraulic pump installation", and the hydraulic pump needs to be delivered from the automated warehouse exit to the assembly station, and the transport vehicle is AGV-07, then the program will write one production relationship edge between the frame welding equipment node and the hydraulic pump installation equipment node, one logistics relationship edge between the automated warehouse exit node and the hydraulic pump installation station node, and one transportation occupancy relationship between the AGV-07 node and this transportation link. In this way, subsequent scheduling steps will read... At that time, not only can we see the process relationships between equipment, but we can also directly see the material transfer link that the process depends on and the transportation equipment currently undertaking the link, thus providing complete input for subsequent collaborative scheduling.

[0041] After the above processing, the output result is a production-logistics diagram. The graph contains a set of nodes. and relation set ,in Used to characterize production equipment, storage locations, and transportation equipment. This is used to characterize the relationship between process sequence, material transfer, and transportation resource usage. This is how it is obtained. The operational information that was originally scattered across multiple business systems during the manufacturing process of engineering machinery has been organized into a unified graph structure. Subsequent steps can directly carry out collaborative scheduling calculations around this graph without having to repeatedly parse the original records in the production and logistics systems.

[0042] S2: The scheduler reads the relationship records of the production flow diagram, executes corresponding scheduling measures based on the relationship records, and generates production urgency and logistics accessibility for each relationship record. Based on the production urgency and logistics accessibility, and combined with the coordination function, the coordination scheduling intensity corresponding to each relationship record is calculated. Based on the coordination scheduling intensity, a coordination intensity matrix is ​​constructed, specifically including:

[0043] S1 forms the production-logistics diagram The production equipment, storage locations, transportation equipment, and their inter-process connections, material transfer relationships, and transportation occupancy relationships within the current scheduling cycle have been organized into the same graph structure. Therefore, S2 directly uses... Using the set of relations in the graph as input, calculate the cooperative scheduling strength for each relation and write the results into the scheduling cooperation matrix. This calculation addresses a core issue in the engineering machinery manufacturing scenario: whether a production chain is worth prioritizing depends on whether two conditions are met simultaneously: first, whether the corresponding subsequent processes have entered a state of urgency; and second, whether the logistics chain supporting that process is ready for execution.

[0044] S2 calculates each edge of the graph relationship given by S1, no longer reverting to the original system's independent scheduling, but instead allowing each edge in the graph to simultaneously carry both production and logistics state information. In practice, the scheduler scans sequentially... The program retrieves relationship records from the MES (Manufacturing Execution System). If the relationship record represents two production equipment nodes, the program extracts the completion progress of the preceding process, whether the subsequent process has been released, and whether the target equipment is available from the order process table, process progress table, and equipment availability status table. If the relationship record represents a warehouse node and a production node, the program extracts the material's delivery request, the availability of accessible vehicles, and whether the origin and destination meet the receiving and dispatching conditions from the logistics scheduling system's transportation task table, vehicle location table, and warehouse status table. If the relationship record represents a production equipment node and a transportation equipment node, the program extracts whether the transportation equipment can continue the link within the current scheduling cycle from the vehicle task queue and task binding table. After completing these reads, the program generates two proportional quantities for each relationship, where... Indicates the urgency of production. This indicates the accessibility of logistics. The following values ​​are directly obtained from three on-site records: the completion rate of the preceding process, the release status of the subsequent process, and the availability status of the target equipment. These three values ​​are averaged with equal weight. For example, if the completion rate of the preceding process is 0.9, the release status of the subsequent process is 1, and the availability status of the target equipment is 1, then... Values ; Similarly, this is directly calculated using three records: valid delivery request flag, available vehicle ratio, and start / end node release flag. These three are averaged with equal weight. For example, if a valid delivery request is 1, one of two nearby vehicles can be accessed (vehicle ratio 0.5), and both start and end nodes have been released (1), then... Values Both of these quantities are derived line by line from the graph relations in S1, therefore the graph... Each edge in S2 is reassigned a computable state of "whether production is urgent" and "whether logistics are smooth".

[0045] In the construction machinery production-logistics scenario, simply using linear weighting will still result in a high score for the relationship of "extremely urgent production but significantly lagging logistics," causing equipment to wait for materials in advance. While simply using a product term can amplify the relationship of "both being high simultaneously," it doesn't differentiate sufficiently for relationships of "basically feasible but slightly deviating." Therefore, this paper first retains the main body of the weighted summation, then adds a product enhancement term to highlight the truly synergistic relationship, and simultaneously adds a difference penalty term to suppress significant imbalances in production and logistics. This yields the synergistic function, whose expression is:

[0046] ;

[0047] In the formula, Represents a node With nodes The relationship between them represents the strength of their coordinated scheduling within the current scheduling cycle, and its value is ultimately written into the scheduling coordination matrix. ; This indicates the production urgency of the relationship, and its value is generated from the order process table, process schedule table, and equipment availability status table in the MES. This represents the logistic reachability of the relationship, and its value is generated from the corresponding transportation task table, vehicle location table, and warehouse location status table in the logistics scheduling system. This represents the balance coefficient between the comprehensive and synergistic terms, and its value is given by the scheduling system configuration file. This represents the difference penalty coefficient, the value of which is given by the scheduling system configuration file.

[0048] The derivation process of this formula is clear and continuous: first use... This represents the average schedulable level of production and logistics, and then... Strengthen the synergy of "both being relatively high", and finally use Subtract from relationships that are clearly unbalanced; due to , , and Both sides of the formula use proportional quantities, maintaining the same evaluation scale. This formula is suitable for engineering machinery scenarios because materials such as chassis, booms, and hydraulic modules are large in volume and have long transfer links. The relationships that can be prioritized must simultaneously satisfy "production window has arrived" and "logistics conditions have been met," rather than only one side being prominent.

[0049] Substituting the above calculation rules into specific relationships yields directly executable scheduling results. For example, in a certain scheduling cycle, the graph generated by S1... There exists a relationship from the automated warehouse exit to the final assembly station, which serves the hydraulic module installation for order O7. The scheduling program reads from the MES records that the completion rate of the preceding frame positioning process for hydraulic module installation is 0.9, the installation station has been released, and the target assembly station is ready for deployment. Therefore, the production urgency of this relationship is [not specified]. The value is 0.967; further data from the logistics dispatch system shows that a delivery request has been generated for this module, one of the two available vehicles near the automated warehouse is currently idle, and the origin and destination nodes are ready for dispatch and pickup, therefore the logistics reachability is [value missing]. It is 0.833. If the configuration parameter is taken as... , The coordination strength of this relationship is calculated using the following formula:

[0050] ;

[0051] This result indicates that this relationship possesses both high production urgency and logistical accessibility, and therefore should be prioritized in the current scheduling cycle. Looking at another relationship, if its corresponding subsequent welding process is nearing release, it makes... The value reached 0.900, but currently available transport vehicles have not yet returned to the depot, resulting in... If it is only 0.400, then under the same parameters, we can obtain... Although these relationships are already urgent in the production plan, their coordination intensity is actively suppressed due to insufficient logistics access conditions, thereby avoiding long waiting times for equipment to receive materials. In the debugging record, after replaying and calculating 45 nodes and 50 relationships for a certain assembly shift, 3 of the relationships that were originally ranked in the top ten based solely on production urgency were demoted out of the top ten due to excessively low logistics accessibility, while the other 4 relationships that were both high in production and logistics entered the top ten. The equipment waiting phenomenon corresponding to these relationships was significantly reduced in subsequent operations, indicating that the coordination intensity calculation can further transform the graph structure in S1 into an execution-oriented priority structure.

[0052] After completing all relation calculations, the scheduler proceeds according to the diagram. The node number in the table will be used to identify each relationship. Write the scheduling coordination matrix If the diagram With 45 nodes and 50 relationships, this step will yield 50 valid collaboration strength values, which will then be filled into the corresponding positions in the matrix. This results in... This corresponds strictly one-to-one with the graph structure in S1: every edge in the graph has a comparable cooperative strength in the matrix; edges that do not exist in the graph remain empty in the matrix. Thus, S2 completes the transformation from "relationships exist" to "relationships are comparable, sortable, and directly usable for scheduling," and subsequent steps can directly revolve around the matrix. The high-value relationships in the model generate the collaborative execution order of production tasks and logistics tasks.

[0053] S3: Construct a time series prediction model. Calculate the predicted scheduling intensity for each relation record in the coordination strength matrix using the time series prediction model. Merge the predicted scheduling intensity with the coordinated scheduling intensity to generate the final predicted scheduling intensity. The scheduling server writes the final predicted scheduling intensity corresponding to all relation records into the prediction scheduling matrix. Specifically, this includes:

[0054] S1 has completed the production-logistics diagram. This diagram structure uniformly describes the production equipment nodes, warehousing nodes, transportation equipment nodes, and the process connections, material transfer relationships, and transportation occupancy relationships between them; based on this, S2 calculates the collaborative scheduling matrix for each relationship. Matrix elements Represents a node With nodes The relationship between them is the intensity of coordinated scheduling in the current scheduling cycle. In actual operation, engineering machinery manufacturing workshops experience significant time fluctuations. For example, the welding time of large structural components is affected by material batches, AGV transportation time is affected by path congestion, and assembly cycle time is affected by differences in manual operation. These fluctuations cause the coordination intensity to show a trend change between consecutive scheduling cycles. Therefore, in S3, the matrix output by S2 is used. Using the same input object as the basis, and combining the coordination strength sequence of the same relationship in historical cycles, the current coordination strength is predicted and dynamically corrected to form a predictive scheduling matrix. This approach originates from the autoregressive concept in time series forecasting theory, its original form stemming from the autoregressive model in statistics, which uses historical data to predict future values. This solution builds upon this by applying scenario-based extrapolation, transforming the forecast object from a single time series into a production-logistics graph. The sequence of collaborative strengths for each relationship is used to achieve self-supervised learning of the changing trends of scheduling relationships.

[0055] In practice, the scheduling server will update the current matrix at the end of each scheduling cycle. Stored in the scheduling database, thus forming a time-ordered sequence of coordination matrices in the database, for example... , , and the current periodic matrix Since these matrices all originate from the same production-logistics diagram The set of relations, therefore any position in the matrix... The meaning of remains consistent, that is, it always corresponds to the relationship. The coordination strength. The scheduling server first extracts the relationship. Historical intensity sequences in continuous periods, for example , , These historical values ​​are all derived from the calculation results of S2, therefore their numerical range is consistent with... To maintain consistency, a time series prediction model was then constructed, consisting of three parts: the first part is a sequence embedding layer, which maps the historical intensity sequence of length 3 to a feature vector; the second part is a gated recurrent unit (GRU) network, which contains two layers of GRU structure, each with a hidden state dimension of 32, used to learn the variation of cooperative intensity over time; the third part is a linear mapping layer, which maps the hidden state of the GRU at the last time step to the predicted intensity. The network automatically constructs training samples using historical scheduling data, such as sequences. As input, the actual value Training is conducted using historical scheduling records as the supervised target, thus falling under the category of self-supervised learning. The training process automatically generates a large number of samples using historical scheduling records, enabling the model to learn the evolutionary patterns of the collaborative relationship between production and logistics over time.

[0056] The objective function used in model training is derived from the error minimization principle in time series forecasting. A trend consistency constraint is introduced to ensure that the predicted direction aligns with the actual direction of change. Classical autoregressive models typically focus only on the error between predicted and true values. This approach adds a differencing trend term to stabilize the predicted sequence in the direction of change, making it more suitable for engineering machinery production and logistics scenarios. The corresponding loss expression is as follows:

[0057] ;

[0058] In the formula, Representing relations The prediction loss is used to train the time series prediction network; The model predicts the first Periodic synergy strength; This represents the actual collaboration strength recorded in the database, and the value comes from S2 during the period. The matrix calculated in ; This represents the actual collaborative strength in the previous period, and its value comes from historical matrix records. This represents the trend consistency weighting coefficient, set by the scheduling system configuration file. The first term of the formula comes from the absolute error in the classical regression model, used to control the deviation between the predicted and actual values; the second term comes from the difference sequence constraint, comparing the predicted changes... Compared with the actual change The differences in these variables allow the model to learn trends in the strength of the relationship. Since all variables are dimensionless proportional quantities, both sides of the formula maintain a consistent dimensional scale.

[0059] After completing network training and entering the online operation phase, the scheduling server uses the model to optimize the current matrix. The predicted scheduling intensity is calculated for each relation. The predicted scheduling strength is then fused with the current coordination strength to obtain the final predicted scheduling strength. This fusion method originates from the state-trend fusion concept in predictive control theory. Its basic idea is to superimpose the predicted trend onto the current observations. This scheme adds a fluctuation suppression term to this concept to adapt to the stability requirements of the high-value material transportation chain in engineering machinery production scenarios. The corresponding prediction function calculation formula is as follows:

[0060] ;

[0061] In the formula, Represents a node With nodes The predicted scheduling strength of the relationship between them is written into the predicted scheduling matrix. ; The current periodic coordinated scheduling strength is represented by the S2 output matrix. ; This represents the next-cycle predicted scheduling intensity output by the time series prediction model. and It represents the true coordination strength between two historical periods, and its value comes from the historical matrix sequence in the database; This represents the predicted fusion coefficient, which is given by the scheduling system configuration file and is used to balance the influence ratio of the current state and future trends. This represents the fluctuation suppression coefficient, used to suppress drastic fluctuations in the coordination strength of a relationship. The first two terms of the formula come from the state and prediction superposition structure in predictive control, while the third term is a fluctuation suppression term designed for production-logistics coordinated scheduling scenarios. When the fluctuation range of a certain relationship is large in a continuous period, the system will automatically reduce its scheduling priority to avoid frequent switching between production equipment and transportation equipment.

[0062] The calculation process can be illustrated through specific examples. Assume a production-logistics diagram in an assembly workshop. There is a relationship between them. The synergy strength of these three cycles is as follows: , , Time series models are based on historical sequences. Predicting the intensity of the next cycle If the system parameters are taken as follows: , The predicted scheduling intensity is calculated as follows:

[0063] ;

[0064] in Substituting into the calculation, we get:

[0065] ;

[0066] This result indicates that the relationship will maintain a high degree of synergy in future cycles, therefore its priority is slightly higher than its current synergy strength. If the strength of another relationship in the last three cycles is... Then the fluctuation difference Even if the predicted value is high, the fluctuation suppression term will significantly reduce the final strength, thereby preventing the scheduling system from relying too much on unstable relationships.

[0067] After completing the above calculations, the scheduling server will allocate all relationships... Write to the prediction scheduling matrix If the production-logistics diagram generated by S1 If the dataset contains 45 nodes and 50 relationships, this step will generate 50 predicted scheduling intensity values, each associated with a set of relationships. This corresponds to a relationship in the matrix. The resulting matrix... It includes three types of information: current collaborative status, historical change trends, and fluctuation stability. This enables subsequent scheduling decisions to respond in advance to changes in logistics conditions while maintaining stable production rhythm, thereby achieving more stable collaborative scheduling between production equipment and logistics resources.

[0068] Step 4: The scheduling server uses the predicted scheduling matrix as an index to locate the field entities corresponding to the relationship records one by one, calculates the resource executability of the field entities, calculates the final execution score based on the resource executability, sorts the final execution scores to generate execution sequences, and resolves conflicts in the execution sequences. The scheduling server then organizes the conflict-resolved execution sequences into a scheduling scheme set, specifically including:

[0069] S3 has formed a predictive scheduling matrix. Each element in the matrix Both correspond to production-logistics diagrams China-US relations The predicted scheduling intensity within the current scheduling window allows S4 to directly complete the final scheduling execution and resource allocation along the same relation index. The key task here is not to re-evaluate the importance of the relation, but rather to transform the matrix... The relationship strength already possessing trend information is transformed into an executable task sequence, and before issuing instructions, the availability status of equipment, warehouse receiving and dispatching status, and transportation equipment accessibility status are incorporated into a unified execution judgment. For engineering machinery manufacturing scenarios, this step has direct engineering significance: even if the same relationship achieves a high predictive scheduling strength in S3, it cannot be executed immediately if the corresponding equipment is already occupied, the line-side buffer is temporarily unavailable for receiving and dispatching, or the available AGV has not yet returned to the field; conversely, a relationship with slightly lower predictive strength but complete resource conditions is often more suitable for priority execution within the current window. Based on this characteristic, this step first... For each relationship, the resource executability is calculated, and then the final execution score is generated by combining the resource usage. Production tasks and logistics tasks are then issued one by one according to the score order.

[0070] The construction of resource feasibility originates from the feasibility evaluation concept in operations research. Its original form typically combines multiple normalized constraints into a single feasibility index. In the engineering machinery production-logistics scenario, this concept is rewritten as a unified description of three conditions: "equipment can be deployed, warehousing can handle receipts and dispatches, and transportation can be accessed." The scheduling server operates according to a matrix... Index-based locating relationship The associated on-site entities: If the relationship corresponds to a process to be executed, the system reads whether the target equipment is currently idle from the equipment status table of the MES system, whether the corresponding material start and end nodes have the conditions for receiving and dispatching from the node status table of the warehouse management system, and whether there is transportation equipment that can be connected to the link in this window from the vehicle status table and task team list of the logistics scheduling system. If the relationship corresponds to a replenishment or transfer link, the system also reads the status of the processing end equipment, the warehouse end status, and the transportation end status according to the nodes mapped by the relationship. To ensure that these three types of constraints participate in subsequent sorting on the same scale, the scheduling server normalizes them into three proportional quantities and calculates the resource executability through an execution function. The specific expression of the execution function is as follows:

[0071] ;

[0072] In the formula, Representing relations The resource executability is determined by the three types of execution conditions within the current window. This indicates the proportion of resources that can be invested corresponding to the starting node of the relationship. Its value is generated by reading the status table corresponding to the node type. When the starting node is a production equipment node, it is taken from the equipment status table of the MES system. When the starting node is a warehouse node, it is taken from the node status table of the warehouse management system. When the starting node is a transportation equipment node, it is taken from the vehicle status table of the logistics scheduling system. This indicates the acceptable proportion of resources corresponding to the target node. Its value is also generated by reading the status table corresponding to the node type. When the target node is a production equipment node, it is taken from the equipment status table of the MES system. When the target node is a warehouse node, it is taken from the node status table of the warehouse management system. When the target node is a transportation equipment node, it is taken from the vehicle status table of the logistics scheduling system. Representing relations The feasible proportion of the corresponding link is generated jointly by the vehicle status table, task team list, and node release record in the logistics scheduling system and the warehouse management system. For example, if there are two candidate vehicles near the required path and one of them can be accessed within the current window, the proportion is 0.5; if both can be accessed, the proportion is 1. The derivation process of this formula is to first convert the three types of on-site constraints into proportional quantities of the same scale, and then use equal weighted averaging to form a unified executability. Since all three terms are dimensionless proportional values, the result on the right side is also a dimensionless proportional value, consistent with the left side. Taking a certain relationship of "automatic warehouse exit - final assembly station" as an example, if the automatic warehouse exit is currently allowed to exit and the corresponding buffer position still has the release conditions, the available proportion of the resources corresponding to the starting node is taken as 0.5. If the equipment corresponding to the target assembly station is idle and allows material reception, then the acceptable proportion of the resources corresponding to the target node is taken as follows: If one of the two nearby AGVs can be connected, then Therefore, the resource executability of this relationship can be obtained as follows: This result directly reflects that the relationship is currently well-suited for implementation, but certain limitations still exist on the transportation side.

[0073] After obtaining the resource feasibility, the scheduling server continues to generate the final execution score. This score originates from the revenue-feasibility joint scoring model in operations research, the basic idea of ​​which is to multiply the task value by its execution feasibility and then handle resource congestion through a penalty term. In this scenario, the task value is determined by the predicted scheduling intensity output by S3. This indicates that the feasibility of implementation is obtained from the above formula. Resource congestion is characterized by the degree to which related resources are allocated within the current window. Considering that large equipment, line-side buffers, and AGV vehicles are scarce resources in engineering machinery workshops, if too many relationships are continuously allocated to a resource, although subsequent relationships are theoretically executable, a waiting chain will form on-site. Therefore, an occupancy penalty term is added to the scoring function to obtain the final execution score. The expression of the scoring function is:

[0074] ;

[0075] In the formula, Representing relations The final execution score is used to generate the execution order; This represents the predicted scheduling strength obtained from S3, and its value comes from the predicted scheduling matrix. ; This indicates the resource executability calculated using the previous formula; Representing relations The resource occupancy intensity is calculated by the scheduling server in the current window. For example, the number of tasks assigned to the device involved in the relationship, the number of send and receive operations occupied by the warehouse node, and the number of transportation links bound to the transportation device are normalized and merged. This represents the resource consumption penalty coefficient, the value of which is given by the scheduling system configuration file. The derivation of this formula can be divided into two steps: first using... Retaining the relationship of "high predictive strength and ease of execution," the trend information of S3 and the executable information of the current window are simultaneously effective within a single product term; then subtract... This pushes down relationships with excessively concentrated resources, thereby freeing up scheduling opportunities for other equally executable relationships with more balanced resource loads. Because... , , and Both use proportional quantities, with the right side maintaining a dimensionless scoring system consistent with the left side. Taking two relationships in a certain assembly shift as an example, the predicted scheduling intensity of relationship one is... The resource executability is calculated using the previous formula. The equipment and vehicles involved in this relationship have already performed a task once within this window, and this has been normalized to obtain... If taken Then the final execution score of this relationship is The predicted scheduling intensity for relation two is slightly lower, which is... However, the resources have high executability, which is... Furthermore, the associated resources are currently under relatively light load. Under the same parameters, we get This result aligns perfectly with the execution logic on-site: although the predicted intensity of the second relationship is slightly lower, it should be prioritized for implementation due to its more complete resource conditions and more balanced resource load.

[0076] After the scheduling server completes all relation scoring, it proceeds according to... Candidate execution sequences are generated from highest to lowest score, and conflict resolution and instruction issuance are performed sequentially. Specifically: the relationship with the highest score is first selected and converted into an actual execution record; if the relationship corresponds to a process connection between equipment, the next process start instruction is written into the MES system, the corresponding material transfer task is written into the logistics scheduling system, and the corresponding inbound / outbound node is locked in the warehouse management system; if the relationship corresponds to a material replenishment relationship from warehouse to workstation, a vehicle task is generated in the logistics scheduling system, and the corresponding subsequent process is marked as "start after material arrival" in the MES system. For each execution record issued, the scheduling server synchronously updates the resource occupancy intensity. The system relies on statistical tables, and then re-examines whether subsequent candidate relationships conflict with already assigned relationships in terms of equipment, storage nodes, or vehicles. Conflicting relationships will be retained for the next round of sorting, and their scores will automatically decrease in the next round based on the updated occupancy records. Taking a replay shift as an example, the graph constructed by S1... The matrix output by S3 contains 45 nodes and 50 relationships. It contains 50 valid predicted intensity values. This step calculates each of these 50 relationships one by one. and Subsequently, the first round of sorting selected 16 high-scoring relationships as the first batch of execution relationships, including 6 equipment process connection tasks, 7 material transfer tasks, and 3 execution tasks corresponding to transportation resource occupation relationships. As these tasks were issued, the relevant equipment, buffer slots, and AGV occupation records were written back to the scheduling server in real time. The second round of sorting selected another 12 non-conflicting relationships to continue execution. Shift playback results showed that after adopting this execution scoring method, the tasks that were originally concentrated on two assembly machines and three AGVs were more evenly distributed across four assembly machines and five AGVs, significantly shortening the equipment waiting chain and significantly alleviating the queuing phenomenon for transportation tasks.

[0077] Finally, the scheduling server organizes the execution records that have been officially issued after conflict resolution within this window into a set of scheduling schemes. .gather Each record in the matrix is ​​related to the matrix. One of the high-scoring relationships corresponds to a set that includes the production equipment number, warehouse node number, transportation equipment number, and task distribution order required for actual execution. This set follows the graph along S1. The matrix of S2 The matrix of S3 This process, derived layer by layer, forms a complete closed loop from relationship structure, coordination strength, trend prediction to final execution. For collaborative scheduling of construction machinery production and logistics, this approach ensures that the final result is not a static sorting outcome, but a real-time scheduling scheme that takes into account trends, feasibility, and resource balance. This allows equipment startup, material replenishment, and vehicle access to be unified into a single execution sequence.

[0078] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A collaborative scheduling method for engineering machinery production and logistics, characterized in that, include: The production equipment nodes, warehousing nodes, and transportation equipment nodes are unified into a node set. An edge set of the node set is constructed. The node set and edge set are written into the same graph structure to generate a production logistics graph. The scheduler reads the relationship records of the production process diagram, executes corresponding scheduling measures based on the relationship records, and generates production urgency and logistics accessibility for each relationship record. Based on the production urgency and logistics accessibility, the coordinated scheduling intensity corresponding to each relationship record is calculated by combining the coordinated function, and a coordinated intensity matrix is ​​constructed based on the coordinated scheduling intensity. A time series prediction model is constructed. The predicted scheduling intensity is calculated for each relation record in the coordination intensity matrix using the time series prediction model. The predicted scheduling intensity is then fused with the coordination scheduling intensity to generate the final predicted scheduling intensity. The scheduling server writes the final predicted scheduling intensity corresponding to all relation records into the prediction scheduling matrix. The scheduling server uses the predicted scheduling matrix as an index to locate the field entities corresponding to the relationship records one by one, calculates the resource executability of the field entities, calculates the final execution score based on the resource executability, sorts the final execution scores to generate execution sequences, resolves conflicts in the execution sequences, and organizes the conflict-resolved execution sequences into a scheduling scheme set.

2. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The edge set includes records of process connection relationships, material transfer relationships, and transportation resource occupancy relationships in the manufacturing scenario.

3. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The execution of corresponding scheduling measures based on the relationship records specifically includes: If the relationship record represents two production equipment nodes, the program extracts the completion progress of the preceding process, the release status of the subsequent process, and the input status of the target equipment corresponding to the relationship record from the order process table, process progress table, and equipment availability status table in the MES. If the relationship records are warehouse nodes and production equipment nodes, the program extracts material delivery request information, the existence of accessible vehicles, and the availability of origin and destination receiving and dispatching conditions from the transportation task table, vehicle location table, and warehouse status table. If the relationship record is between a production equipment node and a transportation equipment node, the program extracts the connectivity of the transportation equipment within the current scheduling cycle from the vehicle task queue and the task binding table.

4. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The parameters of the cooperative function include logistics accessibility, the balance coefficient between the comprehensive term and the cooperative term, and the difference penalty coefficient.

5. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The time series prediction model includes a sequence embedding layer, a gated recurrent unit network, and a linear mapping layer.

6. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The final predicted scheduling intensity is generated by a prediction function, the parameters of which include the collaborative scheduling intensity of the current period, the predicted scheduling intensity generated by the time series prediction model, the actual scheduling intensity of the historical period, the prediction fusion coefficient, and the fluctuation suppression coefficient.

7. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The resource executability is calculated and generated by an execution function. The parameters of the execution function include the resource availability ratio, the resource acceptance ratio, and the connection feasibility ratio of the corresponding link of the relationship record.

8. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The final execution score is generated by a scoring function, the parameters of which include predicted scheduling intensity, resource executability, resource occupancy intensity, and resource occupancy penalty coefficient.

9. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, The scheduling scheme set includes production equipment number, warehouse node number, transportation equipment number, and execution sequence issuance order.

10. The collaborative scheduling method for engineering machinery production and logistics according to claim 1, characterized in that, Before unifying production equipment nodes, warehousing nodes, and transportation equipment nodes into a node set, the process also includes: The scheduler extracts the devices that have been included in the scheduling scope and maps these devices to production equipment nodes. The location of the line-side warehouse, buffer area, and entrance / exit of the vertical warehouse is extracted by the scheduler, and the location is mapped to a warehouse node. The scheduling program extracts the transportation equipment numbers of AGVs, forklifts, and transport vehicles, and maps these transportation equipment numbers to transportation equipment nodes.

Images