Plate-type small home production dispatching system and method

Abstract

The invention relates to the technical field of furniture intelligent manufacturing and production dispatching, and discloses a plate-type small home production dispatching system and method. According to the system, product composition and production requirements are extracted through a work order deep analysis module. And the dynamic resource coupling module performs matching operation on the analysis data and the production resource real-time data pool to generate a resource coupling scheme. And the process chain generation module traverses the process rule base based on the scheme, and dynamically constructs a customized virtual process chain. And the state prediction module performs simulation deduction in combination with historical data and outputs process prediction. And the dispatching optimization module iteratively optimizes the instruction sequence according to the instruction sequence to generate an optimized dispatching instruction set. And the execution tracking module issues an instruction, collects feedback data and updates the resource pool in real time. According to the system, the dynamic and accurate matching of the production resources and the process paths is realized, and the resource utilization efficiency and the flexible response capability of the production system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent manufacturing and production scheduling of furniture, in particular to a production dispatching system and method for panel-based small furniture. BACKGROUND

[0002] In the production of panel-based custom furniture, the traditional production dispatching mode relies on fixed process routes and static resource allocation plans. The system usually decomposes the process according to the preset standard process path bound to the product category, and then allocates tasks according to experience or a simple resource state table. In this mode, the process path is rigid and cannot be dynamically adjusted according to the subtle differences of orders or the real-time load of the production line, and the resource matching is often based on single-dimensional, non-real-time information, ignoring the coupling relationship and coordination possibilities of equipment, materials, and personnel in space and time.

[0003] The prior art has defects. The static resource matching method cannot perceive the real-time conditions of the production site, such as temporary failures of an edge banding machine, actual arrival conditions of a batch of panels, or accumulated fatigue of an operator at a workstation, which can easily cause the dispatching instructions to deviate from the actual situation on site, resulting in resource conflicts, process waiting, or bottleneck effects, making it difficult to continuously optimize the overall utilization efficiency of resources. At the same time, the fixed process route cannot adapt to the production characteristics of panel-based small furniture, which is characterized by "small batch, multiple categories, and fast response". Different orders, even with differences in panels, hardware, and process details, are still forced to follow the same process, restricting the flexibility of production and making it difficult to achieve personalized production flow that is optimal in cost and efficiency for individual orders.

[0004] There is a need for a dispatching technology that can deeply integrate real-time production data and dynamically plan the optimal production path for each order to solve the flexibility and efficiency problems caused by static resource planning and fixed process routes. SUMMARY

[0005] The present application aims to provide a production dispatching system and method for panel-based small furniture to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides a production dispatching system for panel-based small furniture, which comprises:

[0007] a work order deep analysis module for structurally deconstructing the original production work order, extracting product composition elements and production demand elements within the work order, and forming initialization deconstruction data;

[0008] a dynamic resource coupling module for obtaining a production resource real-time data pool, matching and coupling the initialization deconstruction data with the data in the production resource real-time data pool, and generating a resource coupling scheme;

[0009] a process chain generation module configured to traverse a preset process rule library based on the resource coupling scheme to perform logical routing and node concatenation, and to construct a customized virtual process chain for the current work order;

[0010] a state prediction module configured to import historical production efficiency data streams, combine node attributes of the customized virtual process chain to perform simulation deduction, and output a production process prediction state;

[0011] a dispatch optimization module configured to receive the production process prediction state, and perform iterative adjustment and recombination on a dispatch instruction sequence according to a preset optimization objective function to generate an optimized dispatch instruction set;

[0012] an execution tracking module configured to issue the optimized dispatch instruction set to a physical production line, collect execution trace data fed back by the production line, and return the execution trace data to the dynamic resource coupling module to update the production resource real-time data pool.

[0013] Preferably, the work order deep analysis module performs structural decomposition on an original production work order, including:

[0014] reading a document format and data coding of the original production work order, performing standardization preprocessing on the original production work order, and obtaining a standardized work order data stream;

[0015] calling a natural semantic analysis unit to perform key entity and relationship extraction on the standardized work order data stream, and identifying a board type entity, a hardware fitting entity, a processing technology entity, and an order constraint relationship;

[0016] performing correlation expansion and hierarchical decomposition on the identified board type entity, hardware fitting entity, and processing technology entity according to a preset product knowledge graph, and obtaining an atomic-level production task list;

[0017] integrating and packaging the atomic-level production task list and the order constraint relationship to form the initial decomposition data.

[0018] Preferably, the dynamic resource coupling module acquires a production resource real-time data pool, including:

[0019] connecting data interfaces of all processing equipment in a workshop, inventory interfaces of a material storage system, and on-duty state interfaces of a personnel management system to establish multi-dimensional data acquisition channels;

[0020] periodically polling current work load data of the equipment, real-time inventory and location data of the materials, skill level and available working hour data of the personnel through the multi-dimensional data acquisition channels;

[0021] timestamp alignment and data cleaning of the current workload data of the device acquired by polling, the real-time inventory and location data of the material, the skill level and available working hour data of the personnel, and importing into a unified resource state cache area;

[0022] Organizing and managing the data in the resource state cache area according to resource type, resource identifier, and resource state attribute, and constructing a dynamically updated production resource real-time data pool.

[0023] Preferably, the dynamic resource coupling module matches and couples the initialization disassembly data with the data in the production resource real-time data pool, including:

[0024] Parsing the ability demand attribute and quantity demand attribute of the atomic-level production task list from the initialization disassembly data;

[0025] Retrieving a candidate resource set that meets the ability demand attribute from the production resource real-time data pool, and obtaining the real-time state attribute of each resource unit in the candidate resource set;

[0026] Constructing a coupling degree evaluation model, the input of which includes the quantity demand attribute, the real-time state attribute, and a resource conversion cost parameter;

[0027] Using the coupling degree evaluation model to comprehensively score and sort the candidate resource set, and selecting the resource combination scheme with the highest score as the resource coupling scheme.

[0028] Preferably, the process chain generation module traverses a preset process rule library for logical pathfinding and node concatenation, including:

[0029] Loading the process rule library, which stores the rules of the order-dependent relationship, mutual exclusion relationship, and parallel execution condition between different processing operations;

[0030] Taking the starting processing resource in the resource coupling scheme as the root node, performing a depth-first search according to the order-dependent relationship rule to explore all possible process paths that can reach the end point;

[0031] Using the mutual exclusion relationship rule and the parallel execution condition rule to perform feasibility filtering and efficiency evaluation on all possible process paths that can reach the end point;

[0032] Selecting the one with the optimal evaluation from the process paths that pass the feasibility filtering, and encapsulating all the processing operation nodes contained therein in order to form the customized virtual process chain.

[0033] Preferably, the state prediction module simulates the nodes of the customized virtual process chain in combination with the node attributes, including:

[0034] A state transition model is configured for each processing operation node in the customized virtual process chain, which defines the probability distribution of the completion state of the processing operation node over time under the input of specific resource conditions;

[0035] The historical production efficiency data stream is used as a training sample to calibrate the parameters in the state transition model, so that the probability distribution output by the state transition model is consistent with the historical actual situation;

[0036] Starting from the current system time, the resource coupling scheme is used as the initial input, and the state transition model of each node is triggered in turn according to the node order of the customized virtual process chain;

[0037] Collect the simulation results of all node state transitions to generate a complete simulation timeline with time stamp and state confidence as the production process prediction state.

[0038] Preferably, the dispatch optimization module iteratively adjusts and recombines the dispatch instruction sequence according to a preset optimization objective function, including:

[0039] According to the customized virtual process chain and the resource coupling scheme, an initial dispatch instruction sequence is generated, which contains the resource identifier and the planned start time of each atomic-level production task;

[0040] Set the optimization objective function, which at least contains three optimization dimensions of total completion time, resource load balancing degree and process switching cost;

[0041] In the constraint space defined by the production process prediction state, randomly or heuristically perturb the task order, resource allocation and time arrangement in the initial dispatch instruction sequence to generate a series of new candidate dispatch instruction sequences;

[0042] Evaluate each new candidate dispatch instruction sequence using the optimization objective function, retain the sequence with better evaluation results, and use it as the basis for a new round of perturbation and evaluation. After multiple iterations, the sequence with the best evaluation result is determined as the optimized dispatch instruction set.

[0043] Preferably, the execution tracking module sends the optimized dispatch instruction set to the physical production line, including:

[0044] Convert the optimized dispatch instruction set into control codes recognizable by target production equipment and job instruction sheets readable by target operators.

[0045] distributing the control code to the controller of the corresponding target production equipment through an industrial internet of things gateway;

[0046] pushing the work instruction to the display device of the corresponding target operator through a workshop information terminal;

[0047] establishing an instruction execution confirmation loop to receive confirmation feedback signals from the controller of the target production equipment and the display device of the target operator.

[0048] Preferably, the execution tracking module collects execution trace data fed back by the production line, including:

[0049] During the execution of the control code, the device operation log, processing progress signal and abnormal alarm information reported by the target production equipment controller are listened to;

[0050] During the execution of the work instruction, the process start time, process end time and finished product quality inspection result manually entered by the target operator through the workshop information terminal are received;

[0051] The device operation log, processing progress signal, abnormal alarm information listened to and the process start time, process end time and finished product quality inspection result received are associated and integrated to generate structured execution trace data.

[0052] Preferably, the present application further comprises a plate type small household production dispatching method, which comprises all the modules and method processes of the plate type small household production dispatching system as described above.

[0053] Compared with the prior art, the present application has the following beneficial effects:

[0054] By constructing a dynamically updated production resource real-time data pool, and performing multi-dimensional matching and coupling operation between the parsed work order demand and the resource coupling scheme considering multiple constraint conditions such as device state, material flow, personnel load, etc., a panoramic and instant resource information is obtained, which ensures the high synchronization of the dispatching instruction and the actual situation of the production site, avoids the resource idling, process blocking and production conflict caused by information lag or one-sidedness, and improves the overall utilization rate of various resources of the production line and the smoothness of collaborative operation.

[0055] Based on the preset process rule base containing rich process logic and rules, a customized virtual process chain is dynamically constructed by logical routing and node concatenation according to the specific attributes of the current work order and the matched resource scheme. This technology makes the process path no longer rigidly bound to the product type, but is dynamically generated according to real-time demand and conditions. Each order can obtain an optimal or feasible process sequence tailored for it, and the system can automatically avoid current resource bottlenecks and select the most suitable alternative process or equipment, thereby enhancing the inherent flexibility of the production system to respond to order diversity and on-site uncertainty, realizing the personalization and precision of the production process, and shortening the overall through time of the order. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A timing diagram for the plate small household production dispatching system described in the present application;

[0057] Figure 2 A flowchart for the structured deconstruction of the work order depth analysis module;

[0058] Figure 3 A flowchart for the dynamic resource coupling module to obtain the real-time data pool of production resources;

[0059] Figure 4 A calibration effect diagram for the normal distribution fitting of the cutting process time-consuming state transition model;

[0060] Figure 5 A resource load balancing thermal map for each process of plate household production dispatching. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0062] Please refer to Figure 1The application provides a plate small household production dispatching system, which comprises a work order deep analysis module, a dynamic resource coupling module, a process chain generation module, a state prediction module and a dispatching optimization module.

[0063] Embodiment 1: refer to Figure 2 The work order deep analysis module performs structured decomposition on the original production work order, reads the document format and data coding of the original production work order, performs standardized preprocessing on the original production work order, and obtains standardized work order data stream. The natural semantic analysis unit is called to perform key entity and relationship extraction on the standardized work order data stream, and the plate type entity, hardware accessory entity, processing technology entity and order constraint relationship are identified. According to the preset product knowledge graph, the plate type entity, hardware accessory entity and processing technology entity are associated and expanded and hierarchically decomposed, and the atomic-level production task list is obtained. The atomic-level production task list and the order constraint relationship are integrated and packaged to form the initialization decomposition data.

[0064] In specific implementation, the work order deep analysis module performs structured decomposition on the original production work order. The original production work order can be a structured data file from an enterprise resource planning system, such as a JSON document containing product number, plate list, hardware list, process requirement and delivery date field. It can also be a semi-structured document generated by a sales system, such as a PDF file containing product schematic diagram, text description and specification parameters. The work order deep analysis module first reads the document format and data coding of the original production work order. For JSON document format, the work order deep analysis module directly analyzes the internal data structure. For PDF file format, the work order deep analysis module calls an optical character recognition engine to extract text content, and performs coding unification and field missingness check on the extracted text content, completes the standardized preprocessing of the original production work order, and outputs a standardized work order data stream with unified fields and unified coding format.

[0065] In a specific implementation, the identified board type entity, hardware accessory entity, and processing technology entity are associated, expanded, and hierarchically decomposed according to a preset product knowledge graph. The product knowledge graph stores the association and subordination relationship between the constituent elements of the board-based home furnishing product. For example, when the product entity "bedside cabinet" is identified, the product knowledge graph indicates that it is composed of "side plate", "bottom plate", "back plate", "drawer panel", "guide rail", "handle", and other sub-components. For the "side plate" component, the product knowledge graph further indicates that its production tasks include "cutting", "edge sealing", and "drilling", thereby decomposing the high-level product description into specific, non-divisible atomic-level processing actions, i.e., an atomic-level production task list. Optionally, the atomic-level production task list and order constraint relationship are integrated and packaged, and the work order depth analysis module creates a structured data object. This data object includes two main parts. The first part is the atomic-level production task list, and each record in the list defines the material specifications, process type, precision requirement, and quantity required for an atomic task. The second part is a set of order constraint relationships, which records all the constraint conditions extracted from the work order, such as delivery time constraints, special process constraints, or customer-specified accessory constraints.

[0066] Embodiment 2: refer to Figure 3 The dynamic resource coupling module acquires a production resource real-time data pool, connects the data interfaces of all processing equipment in the workshop, the inventory interfaces of the material storage system, and the on-duty state interfaces of the personnel management system, and establishes multi-dimensional data acquisition channels. Through the multi-dimensional data acquisition channels, the current work load data of the equipment, the real-time inventory and location data of the materials, and the skill level and available working hour data of the personnel are periodically polled and acquired. The current work load data of the equipment, the real-time inventory and location data of the materials, and the skill level and available working hour data of the personnel polled are time-stamped and aligned and data cleaned, and are imported into a unified resource state cache area. The data in the resource state cache area is organized and managed according to resource type, resource identifier, and resource state attribute, and a dynamically updated production resource real-time data pool is constructed. The dynamic resource coupling module matches and couples the data in the production resource real-time data pool with the initialized deconstructed data, and parses the ability requirement attribute and quantity requirement attribute of the atomic-level production task list from the initialized deconstructed data. A candidate resource set that meets the ability requirement attribute is retrieved from the production resource real-time data pool, and the real-time state attribute of each resource unit in the candidate resource set is acquired. A coupling degree evaluation model is constructed, and the inputs of the coupling degree evaluation model include the quantity requirement attribute, the real-time state attribute, and the resource conversion cost parameter. The coupling degree evaluation model is used to comprehensively score and sort the candidate resource set, and the resource combination scheme with the highest score is selected as the resource coupling scheme.

[0067] In a specific implementation, the dynamic resource coupling module acquires the production resource real-time data pool. The dynamic resource coupling module connects the data interfaces of all processing equipment in the workshop, the inventory interfaces of the material storage system, and the on-duty state interfaces of the personnel management system through industrial Ethernet and standard application programming interfaces, establishes a multi-dimensional data acquisition channel covering the dimensions of equipment, materials, and personnel. The data interface of the processing equipment provides the running state, current workload, and planned maintenance time window data of the equipment. The inventory interface of the material storage system provides real-time inventory, storage location, and material code data of plates, edge strips, and hardware. The on-duty state interface of the personnel management system provides team, skill level, current task state, and remaining available working hour data of the operating personnel. Through the multi-dimensional data acquisition channel, the dynamic resource coupling module automatically polls the current workload data of the equipment, the real-time inventory and location data of the materials, and the skill level and available working hour data of the personnel at a fixed time period to form the original resource state flow.

[0068] In some embodiments, the current workload data of the equipment, the real-time inventory and location data of the materials, and the skill level and available working hour data of the personnel polled are time-stamped and cleaned. The dynamic resource coupling module has a data preprocessing engine built-in. The data preprocessing engine stamps a uniform system timestamp on each batch of data polled from different interfaces and eliminates abnormal values obviously beyond the reasonable range, such as records with equipment load rate exceeding 100% or records with negative material inventory, and then merges the cleaned data into a unified resource state cache area in memory according to resource categories. The data in the resource state cache area is organized and managed according to resource type, resource identifier, and resource state attribute. The resource type includes “numerical control cutting machine”, “intelligent edge banding machine”, “warehouse shelf A01”, and “operating worker Zhang San”. The resource identifier is a unique identification code. The resource state attribute is a set of key-value pairs changing over time, thereby constructing a dynamically updated production resource real-time data pool. Any record update of the production resource real-time data pool will overwrite the old record to reflect the latest state.

[0069] In practical implementation, the dynamic resource coupling module matches and couples the initialized deconstructed data with the data in the real-time production resource data pool. The module first parses the atomic-level production task list from the initialized deconstructed data, extracting the resource capacity and quantity requirements for each atomic-level production task. Optionally, a coupling degree evaluation model is constructed to comprehensively score the candidate resource set. This model considers quantity requirement matching degree, real-time status quality, and resource conversion cost parameters. Quantity requirement matching degree assesses the candidate resource's ability to meet the task's quantity requirements; real-time status quality is based on the resource's idleness or load level; and resource conversion cost parameters include the preparation time or material handling distance required for switching equipment for the task. In essence, the process of using the coupling degree evaluation model to comprehensively score and rank the candidate resource set involves calculating the coupling degree evaluation value for each candidate resource or resource combination. The coupling degree evaluation model can be expressed as:

[0070] ;

[0071] in: This represents the coupling degree evaluation value for a specific candidate resource unit. These are preset weighting coefficients. It is a matching function used to evaluate the current available capacity of a resource. With the number of tasks required The degree of matching, It is a state evaluation function used to quantify the real-time state attributes of resources. The degree of superiority or inferiority, This represents the switching cost required to switch resources to perform the current task. The Dynamic Resource Coupling module calculates the cost of switching all candidate resources or combinations of resources. Value, and based on The values ​​are sorted in descending order, and the resource combination scheme with the highest score is selected and marked as the resource coupling scheme generated in this operation.

[0072] Example 3: The process chain generation module traverses a pre-defined process rule library for logical pathfinding and node concatenation. It loads the process rule library, which stores rules regarding the sequential dependencies, mutual exclusions, and parallel execution conditions between different processing operations. Using the initial processing resource in the resource coupling scheme as the root node, a depth-first search is performed based on the sequential dependency rules to explore all possible process paths leading to the endpoint. The mutual exclusion and parallel execution rules are used to perform feasibility filtering and efficiency evaluation on all possible process paths leading to the endpoint. The optimal process path is selected from those that pass the feasibility filtering, and all its processing operation nodes are sequentially encapsulated to form a customized virtual process chain.

[0073] In a specific implementation, the process chain generation module traverses the preset process rule library for logical routing and node concatenation. The process chain generation module loads the process rule library from a storage medium. The process rule library is stored in a form of a graph structure or a relational data table, which records rich logical relationships between different processing operation nodes, including a precedence relationship rule that processing operation A must be completed before processing operation B, a mutual exclusion relationship rule that processing operation C and processing operation D cannot be processed by the same device, and a parallel execution condition rule that processing operation E and processing operation F can be simultaneously performed after the material is prepared. The process chain generation module receives the resource coupling scheme from the dynamic resource coupling module. The resource coupling scheme specifies the starting processing resource required for executing the work order, for example, a numerical control cutting machine. The process chain generation module takes the starting processing resource in the resource coupling scheme as the root node of the entire process path search.

[0074] In some embodiments, the starting processing resource in the resource coupling scheme is taken as the root node, and a depth-first search is performed according to the precedence relationship rule in the process rule library to explore all possible process paths that can reach the end point. The depth-first search algorithm starts from the root node, explores all subsequent feasible processing operation nodes, recursively explores the subsequent nodes of each new node reached, and stops until the end node representing task completion or a node that cannot continue is reached. This process generates one or more search trees representing all possible process sequences. For example, for a plate that needs to be cut, edged, and drilled, the search may find different sequences such as "cutting → edging → drilling" and "cutting → drilling → edging". Each path is a complete sequence of processing steps.

[0075] In a specific implementation, the mutual exclusion relationship rule and the parallel execution condition rule are used to filter the feasibility of all possible process paths that can reach the end point and evaluate the efficiency. The feasibility filtering directly eliminates invalid paths according to the rules. For example, if the process rule library defines that "laser edging" and "manual trimming" are mutually exclusive due to device conflicts, any path containing both nodes will be filtered out. The efficiency evaluation performs quantitative analysis on the filtered paths, and the evaluation dimensions can include estimated total processing time, device switching times, and cache waiting time. From the process paths that pass the feasibility filtering, the one with the best evaluation is selected. The selection criteria can be the shortest estimated total processing time, and the evaluation process can be formally expressed. For a path , the evaluation value is

[0076] ;

[0077] wherein: represents the evaluation value of the process path , and indicates that all processing operation nodes in the path are considered. the corresponding values of the nodes are summed up, representing the execution node the standard man-hour of the processing operation corresponding to the node, representing the switching time from the preceding node to the node the preparation or switching time required. The process chain generation module calculates the value of each candidate path, and selects the path with the minimum value as the optimal path.

[0078] Optionally, all processing operation nodes contained in the selected optimal path are encapsulated in sequence to form a customized virtual process chain. The encapsulation process assigns each node in the path a unique serial number, records the processing operation description, input and output materials, bound resources and logical relationship between nodes corresponding to the node, and finally generates a structured data object, which represents the complete production process blueprint from start to finish customized for the current work order. It can be understood that the customized virtual process chain is not fixed and unchangeable, and its construction strictly depends on the real-time logic of the specific resource coupling scheme and process rule library of this work order. Therefore, for different work orders or the same work order under different resource states, the generated customized virtual process chain may be different.

[0079] In the embodiment 4, the state prediction module simulates the state transition of each node in the customized virtual process chain based on the node attributes of the customized virtual process chain, and configures a state transition model for each node in the customized virtual process chain. The state transition model defines the probability distribution of the completion state of the node changing with time under the input of specific resource conditions. The historical production efficiency data stream is used as a training sample to calibrate the parameters in the state transition model, so that the probability distribution output by the state transition model is consistent with the historical actual situation. From the current system time, the resource coupling scheme is used as the initial input, and the state transition model of each node is triggered in sequence according to the node order of the customized virtual process chain. The simulation results of the state transition of all nodes are collected to generate a complete simulation timeline with time stamp and state confidence as the predicted state of the production process. The dispatch optimization module adjusts and recombines the dispatch instruction sequence according to the preset optimization objective function, generates an initial dispatch instruction sequence based on the customized virtual process chain and the resource coupling scheme, and the initial dispatch instruction sequence includes the resource identifier and the planned start time of each atomic production task. The optimization objective function is set, and the optimization objective function includes at least three optimization dimensions of total completion time, resource load balancing degree and process switching cost. Within the constraint space defined by the predicted state of the production process, the task order, resource allocation and time arrangement in the initial dispatch instruction sequence are randomly or heuristically disturbed to generate a series of new candidate dispatch instruction sequences. The optimization objective function is used to evaluate each new candidate dispatch instruction sequence, and the sequence with better evaluation result is retained, and a new round of disturbance and evaluation is performed based on the sequence. After multiple iterations, the sequence with the best evaluation result is determined as the optimized dispatch instruction set.

[0080] In a specific implementation, the state prediction module conducts simulation inference in combination with the node attributes of the customized virtual process chain, the state prediction module configures a state transition model for each processing operation node in the customized virtual process chain, the state transition model defines the probability distribution of the completion state of the processing operation node over time under the input of specific resource conditions. For processing operation nodes such as "numerical control cutting", the state transition model may describe that under the given plate type, equipment model and tool state conditions, the time required to complete the cutting task conforms to a probability distribution, for example, a normal distribution with a mean of 5 minutes and a standard deviation of 0.5 minutes. The state prediction module imports a historical production efficiency data stream, which is a structured data set containing a large number of completed production task records extracted from the manufacturing execution system database, each record in the data set records the task type, the resources used, the actual start time, the actual end time and any abnormal interruption information, the state prediction module takes these historical records as training samples to calibrate the distribution parameters in the state transition model, so that the probability distribution output by the state transition model is consistent with the actual situation, for example, the average time and variance of the "numerical control cutting" task under the same resource conditions are calculated using historical data, and the mean and standard deviation parameters of the state transition model are updated.

[0081] In some embodiments, referring to Table 1, the structure of the historical production efficiency data stream can be as shown in Table 1, Table 1 provides data for calibrating the state transition model, and shows the data of some fields in the historical production efficiency data stream, which is used to learn the time consumption rule of different tasks under specific resource conditions.

[0082] Table 1: Historical production efficiency data table

[0083] In a specific implementation, starting from the current system time, the state transition models of each node are triggered in sequence according to the node order of the customized virtual process chain, with the resource coupling scheme as the initial input. The simulation engine of the state prediction module first generates a random actual time consumption for the first node of the process chain "cutting", which is sampled from the probability distribution defined by the calibrated state transition model corresponding to the "cutting" operation and the specific cutting machine, for example, 5.2 minutes is sampled from a normal distribution with a mean of 5 minutes and a standard deviation of 0.5 minutes. The simulation engine records the simulation start time of the "cutting" node as the current system time T, and the simulation completion time as T+5.2 minutes. Subsequently, the simulation engine takes the simulation completion time of the first node as the simulation start time of the second node "edge sealing", and again samples the random time consumption of the "edge sealing" operation from its corresponding calibrated state transition model. This iteration is performed until the simulation of the last node of the process chain is completed. The state prediction module collects the simulation results of all node state transitions, generates a complete simulation timeline with timestamps and state confidence, and this simulation timeline is used as the production process prediction state. The prediction state not only includes the estimated time point of each node, but also includes the uncertainty range based on the probability distribution. The core of the state transition model can be expressed as the completion time In the formula , represents the completion time random variable generated for a certain processing operation node in the simulation deduction, and the symbol indicates "subject to … distribution", represents a parameterized probability distribution, such as a normal distribution or a gamma distribution, represents the parameter set of the probability distribution, and the parameter set is obtained through the calibration of historical production performance data.

[0084] Optionally, the dispatch optimization module iteratively adjusts and reorganizes the dispatch instruction sequence according to a preset optimization objective function. The dispatch optimization module generates an initial dispatch instruction sequence based on the customized virtual process chain and resource coupling scheme. The initial dispatch instruction sequence is a list, and each instruction in the list specifies the resource identifier and the planned start time of an atomic production task. The planned start time is usually estimated based on the default order of the process chain and the theoretical capacity of the resource. The dispatch optimization module sets the optimization objective function, which includes at least three optimization dimensions: total completion time, resource load balancing degree, and process switching cost. The total completion time refers to the time from the start of the first task to the end of the last task. The resource load balancing degree measures the difference in workload between different devices or workers. The process switching cost measures the additional time consumption caused by device tool changing, mold changing, or line changing due to changes in task order. The specific form of the optimization objective function can be the weighted sum of the minimum total completion time, resource load balancing degree, and process switching cost.

[0085] In some embodiments, the task order, resource allocation, and time arrangement in the initial dispatch instruction sequence are randomly or heuristically disturbed within the constraint space defined by the production process prediction state, generating a series of new candidate dispatch instruction sequences. The constraint space is defined by the production process prediction state. For example, if the prediction state indicates that a certain device may be in a high load state during a certain time period, the optimization process will try to avoid scheduling new tasks during that time period. The disturbance operation includes swapping the execution order of two tasks that do not violate the process dependency relationship, reassigning a task to another idle resource with the same capability, or slightly advancing or postponing the task start time within the allowed time window. It can be understood that each new candidate dispatch instruction sequence is evaluated using the optimization objective function. The evaluation process calculates the total completion time, resource load balancing degree, and process switching cost of each candidate sequence, and calculates a comprehensive evaluation value by substituting the optimization objective function. The dispatch optimization module retains the sequence with better evaluation results. The dispatch optimization module performs a new round of disturbance and evaluation based on the sequence with better evaluation results. After multiple iterations, the sequence with the best evaluation result is determined as the optimized dispatch instruction set. Compared with the initial dispatch instruction sequence, the optimized dispatch instruction set usually has significant improvements in the comprehensive objective.

[0086] Referring to Figure 4In the parameter calibration process of the cutting process time-consuming state transition model, the probability distribution fitting depends on the statistical analysis of the historical production performance data stream. In specific operation, the actual time-consuming distribution is presented in the form of a histogram, corresponding to the historical actual time-consuming data of the cutting process under specific resource conditions; the fitted normal distribution curve is the probability distribution of the state transition model calibrated based on the historical data, in which the distribution parameters μ = 299.3 (mean) and σ = 24.4 (standard deviation). The matching effect of the two intuitively reflects the calibration accuracy of the state transition model: the frequency distribution of the histogram is basically consistent with the probability density trend of the normal curve, indicating that the model can effectively describe the probability law of the time-consuming of the cutting process changing with time. During the parameter configuration process, the mean and standard deviation of the normal distribution are calculated based on the historical production performance data stream (such as task records containing actual start / end time stamps), ensuring the consistency of the model output and the actual production situation.

[0087] In embodiment 5, the execution tracking module transmits the optimized dispatch instruction set to the physical production line, and converts the optimized dispatch instruction set into control codes recognizable by the target production equipment and job instruction sheets readable by the target operators. The control codes are distributed to the controllers of the corresponding target production equipment through the industrial Internet of Things gateway. The job instruction sheets are pushed to the display devices of the corresponding target operators through the workshop information terminal. An instruction execution confirmation loop is established to receive confirmation feedback signals from the controllers of the target production equipment and the display devices of the target operators. The execution tracking module collects the execution trace data fed back by the production line. During the execution of the control codes, the device running logs, processing progress signals and abnormal alarm information reported by the controllers of the target production equipment are listened to. During the execution of the job instruction sheets, the process start time, process end time and finished product quality inspection results manually entered by the target operators through the workshop information terminal are received. The listened device running logs, processing progress signals and abnormal alarm information, and the received process start time, process end time and finished product quality inspection results are associated and integrated to generate structured execution trace data.

[0088] In a specific implementation, the execution tracking module dispatches the optimized dispatching instruction set to the physical production line, and the instruction conversion unit embedded in the execution tracking module converts the structured optimized dispatching instruction set into two types of executable instructions, one is the control code recognizable by the target production equipment, and the other is the job instruction sheet readable by the target operator. For the control code recognizable by the target production equipment, the instruction conversion unit translates the processing steps, parameters and sequence into machine instructions according to the model and controller type of the target production equipment, such as converting the cutting path and speed of the plate into G code files that can be parsed by the numerical control system. For the job instruction sheet readable by the target operator, the instruction conversion unit assembles task information, material information and quality requirements into structured documents or graphical interfaces, such as generating HTML5 pages or PDF documents containing process steps, drawings, precautions and quality inspection standards.

[0089] In some embodiments, the control code is distributed to the controller of the corresponding target production equipment through an industrial Internet of Things gateway, which is connected to the execution tracking module through the internal network of the workshop and maintains a mapping table of device network addresses and resource identifiers. The execution tracking module sends the control code together with its corresponding target production equipment resource identifier to the industrial Internet of Things gateway, which finds the network address of the target production equipment controller according to the mapping table, transmits the control code file to the target controller through ModbusTCP, OPCUA or a specific device manufacturer's protocol, and triggers the controller to load and execute the new processing program. The job instruction sheet is pushed to the display device of the corresponding target operator through the workshop information terminal, which can be a fixed workstation touch screen computer or a mobile handheld device. The execution tracking module binds the job instruction sheet data with the target operator's account and pushes the data to the workshop information terminal logged in by the operator through the workshop wireless network or message queue. The job instruction sheet is presented to the target operator in a clear and interactive form on the display device.

[0090] In a specific implementation, a command execution confirmation loop is established to receive confirmation feedback signals from the controller of the target production device and the display device of the target operator, which is a feedback mechanism for ensuring that the command has been successfully received and is ready for execution. For the delivery of control codes, the controller of the target production device will return a "ready to receive" status code to the industrial Internet of Things gateway after successfully receiving and verifying the control codes, and the industrial Internet of Things gateway will forward the status code back to the execution tracking module. For the delivery of job instruction books, the target operator needs to manually click the "confirm receipt" button after checking the job instruction book on the workshop information terminal, and the workshop information terminal will generate a confirmation message and send it back to the execution tracking module. The execution tracking module only marks the status of the task as "dispatched - pending execution" in the system after receiving confirmation feedback signals from both the controller and the operator for a specific task, thereby completing the closed loop of command delivery. The confirmation feedback signal can be formally defined as a state synchronization, and its logic can be represented as:

[0091] ;

[0092] wherein: represents the confirmation signal from the controller of the target production device, represents the confirmation signal from the display device of the target operator, and the symbol represents the logical "and" operation, i.e. both need to be present at the same time, and the symbol represents "trigger state transition", represents the update of the task status to a new state "dispatched - pending execution".

[0093] Optionally, the execution tracking module collects the execution trace data feedback from the production line. During the execution of the control code, the execution tracking module listens to the device operation logs, processing progress signals and abnormal alarm information reported by the target production device controller through the industrial Internet of Things gateway. The device operation logs include detailed operation parameters such as spindle start-stop, speed, power curve, etc. The processing progress signal is usually reported in the form of completion percentage or specific process code. The abnormal alarm information includes error code and description. These data are transmitted to the execution tracking module in real time or quasi-real time. During the execution of the job instruction book, the process start time, process end time and finished quality inspection results manually entered by the target operator through the workshop information terminal are received. The target operator clicks the "start" button when starting a process, and clicks the "end" button and fills in or selects the quality inspection results when ending the process. These interactive actions generate time stamps and result data through the workshop information terminal and send them to the execution tracking module.

[0094] It can be understood that the execution tracking module relies on a unified dispatch task identifier to realize the association of data. The data stream from the equipment and the personnel carries the dispatch task identifier, and the data integration engine of the execution tracking module merges, sorts and time stamp aligns the discrete data records belonging to the same production task according to the identifier, and fills into a predefined structured data template, so as to generate a complete structured execution trace data which describes the actual execution process of a single task, and the execution trace data contains the key event and state records in the whole life cycle from the beginning to the end of the task.

[0095] Referring to Figure 5 With the process as the horizontal dimension and the resource type as the vertical dimension, the cutting, edge sealing, drilling, polishing and assembly five processes are intuitively presented in the distribution state of the three resource indicators of equipment load, personnel working hours and material usage rate. Specifically, the drilling process is at a high level in the three indicators of equipment load (90.0%), personnel working hours (8.2 hours) and material usage rate (92.0%), and is the resource high-load process in the current production process; and the equipment load (65.0%), personnel working hours (5.5 hours) and material usage rate (75.0%) of the polishing process are relatively low, and there is a certain redundancy space in resource utilization. The heat map realizes the visualization and quantification of the resource state through the color gradient (dark color represents high load / high usage rate, and light color represents low load / low usage rate), which can provide direct reference basis for the resource load balancing objective function of the dispatch optimization module.

[0096] It should be noted that, in this document, the terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0097] Although the embodiments of the present application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A dispatching system for panel furniture production, characterized in that, The system includes: The work order deep analysis module is used to deconstruct the original production work order, extract the product components and production demand elements in the work order, and form the initial deconstruction data. The dynamic resource coupling module is used to obtain the real-time data pool of production resources, match and couple the initial deconstructed data with the data in the real-time data pool of production resources, and generate a resource coupling scheme. The process chain generation module is used to construct a customized virtual process chain for the current work order by traversing a preset process rule library to perform logical pathfinding and node concatenation based on the resource coupling scheme. The status prediction module is used to import historical production efficiency data streams, combine them with the node attributes of the customized virtual process chain to perform simulation and deduction, and output the predicted status of the production process. The dispatch optimization module is used to receive the predicted status of the production process and iteratively adjust and reorganize the dispatch instruction sequence according to the preset optimization objective function to generate an optimized dispatch instruction set. The execution tracking module is used to send the optimized dispatch instruction set to the physical production line, collect the execution trace data fed back by the production line, and send the execution trace data back to the dynamic resource coupling module to update the real-time production resource data pool.

2. The dispatching system for panel furniture production as described in claim 1, characterized in that, The work order deep parsing module performs structured deconstruction on the original production work orders, including: The document format and data encoding of the original production work order are read, and the original production work order is preprocessed in a standardized manner to obtain a standardized work order data stream. The natural language processing unit is invoked to extract key entities and relationships from the standardized work order data stream, identifying entities such as sheet material type, hardware accessories, processing technology, and order constraint relationships; Based on the preset product knowledge graph, the identified sheet material type entities, hardware accessory entities, and processing technology entities are associated, expanded, and hierarchically decomposed to obtain an atomic-level production task list; The atomic-level production task list and the order constraint relationship are integrated and encapsulated to form the initialization deconstruction data.

3. The dispatching system for panel furniture production as described in claim 2, characterized in that, The dynamic resource coupling module acquires a real-time data pool of production resources, including: Establish a multi-dimensional data collection channel by connecting the data interfaces of all processing equipment in the workshop, the inventory interface of the material storage system, and the on-duty status interface of the personnel management system; The system periodically polls through the multi-dimensional data acquisition channel to obtain the equipment's current workload data, the real-time inventory and location data of materials, and the skill level and available working hours data of personnel. The current workload data of the equipment, the real-time inventory and location data of the materials, and the skill level and available working hours data of the personnel obtained by polling are timestamped and cleaned, and then merged into a unified resource status cache area. The data in the resource status cache is organized and managed according to resource type, resource identifier, and resource status attributes to construct a dynamically updated real-time data pool for production resources.

4. The dispatching system for panel furniture production as described in claim 3, characterized in that, The dynamic resource coupling module performs matching and coupling operations between the initialization deconstruction data and the data in the real-time production resource data pool, including: The resource capacity requirement attributes and quantity requirement attributes of the atomic-level production task list are parsed from the initial deconstructed data. Retrieve a set of candidate resources that meet the capability requirement attributes from the real-time data pool of production resources, and obtain the real-time status attributes of each resource unit in the set of candidate resources; Construct a coupling degree evaluation model, the input of which includes the quantity demand attribute, the real-time status attribute, and the resource conversion cost parameter; The coupling degree evaluation model is used to comprehensively score and rank the candidate resource set, and the resource combination scheme with the highest score is selected and marked as the resource coupling scheme.

5. A dispatching system for panel furniture production as described in claim 1, characterized in that, The process chain generation module traverses a preset process rule base to perform logical pathfinding and node concatenation, including: Load the process rule library, which stores rules on the sequential dependency relationship, mutual exclusion relationship, and parallel execution condition rules between different processing operations; Using the initial processing resource in the resource coupling scheme as the root node, a depth-first search is performed according to the sequential dependency rules to explore all possible process paths that can reach the destination. The mutual exclusion rules and parallel execution condition rules are used to perform feasibility filtering and efficiency evaluation on all possible process paths that may reach the destination. The optimal process path is selected from those that pass the feasibility filter, and all its processing operation nodes are encapsulated sequentially to form the customized virtual process chain.

6. A dispatching system for panel furniture production as described in claim 5, characterized in that, The state prediction module performs simulation and deduction based on the node attributes of the customized virtual process chain, including: Configure a state transition model for each processing operation node in the customized virtual process chain. The state transition model defines the probability distribution of the completion state of the processing operation node changing over time under specific resource input conditions. The historical production efficiency data stream is used as a training sample to calibrate the parameters in the state transition model so that the probability distribution output by the state transition model matches the actual historical situation. Starting from the current system time, with the resource coupling scheme as the initial input, the state transition model of each node is triggered sequentially according to the node order of the customized virtual process chain; Collect the simulation results of state transitions of all nodes and generate a complete simulation timeline with timestamps and state confidence scores, which serves as the predicted state of the production process.

7. A dispatching system for panel furniture production as described in claim 6, characterized in that, The dispatch optimization module iteratively adjusts and reorganizes the dispatch instruction sequence according to a preset optimization objective function, including: Based on the customized virtual process chain and the resource coupling scheme, an initial dispatch instruction sequence is generated, which includes the resource identifier and planned start time allocated to each atomic-level production task. Define an optimization objective function, which includes at least three optimization dimensions: total completion time, resource load balancing, and process changeover cost. Within the constraint space defined by the predicted state of the production process, the task order, resource allocation, and time arrangement in the initial dispatch instruction sequence are randomly or heuristically perturbed to generate a series of new candidate dispatch instruction sequences. Each new candidate dispatch instruction sequence is evaluated using the aforementioned optimization objective function. The sequence with the better evaluation result is retained, and a new round of perturbation and evaluation is performed based on this. After multiple rounds of iteration, the sequence with the best evaluation result is determined as the optimized dispatch instruction set.

8. A dispatching system for panel furniture production as described in claim 7, characterized in that, The execution tracking module sends the optimized dispatch instruction set to the physical production line, including: The optimized dispatch instruction set is converted into control codes recognizable by the target production equipment and work instructions readable by the target operators; The control code is distributed to the controller of the corresponding target production equipment via an industrial IoT gateway; The work instruction is pushed to the display device of the corresponding target operator through the workshop information terminal; Establish an instruction execution confirmation loop to receive confirmation feedback signals from the controller of the target production equipment and the display device of the target operator.

9. A dispatching system for panel furniture production as described in claim 8, characterized in that, The execution tracking module collects execution trace data fed back from the production line, including: During the execution of the control code, the system monitors the equipment operation logs, processing progress signals, and abnormal alarm information reported by the target production equipment controller. During the execution of the work instruction, the system receives the process start time, process end time, and completion quality inspection results manually entered by the target operator through the workshop information terminal. The monitored equipment operation logs, processing progress signals, abnormal alarm information, and received process start time, process end time, and completion quality inspection results are correlated and integrated to generate structured execution trace data.

10. A method for dispatching work in the production of panel furniture, characterized in that, It includes all the modules and method processes of the panel furniture production dispatch system as described in any one of claims 1 to 9.

Images