Intelligent fms tray-based allocation method and device, and computer readable storage medium

Abstract

The present application relates to a distribution method based on intelligent FMS pallets, and relates to the technical field of FMS systems, comprising the following steps: S1: receiving a starting instruction of a user for a material task; S2: in response to the starting instruction, judging whether the distribution mode is an automatic distribution mode or a manual distribution mode; S3: if it is the automatic distribution mode, automatically screening matching pallets from all pallets that are not distributed with materials and are empty, and distributing the screened pallets to the material task until the number of distributed pallets is equal to the required number, according to the required number and type of pallets of the material task; S4: if it is the manual distribution mode, providing a pallet selection interface, allowing the user to select target pallets through system adaptation recommendation or manual input, and distributing the pallets selected by the user to the material task until the required number of distribution is completed. The present application significantly improves the operation flexibility and scene adaptability of the FMS system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of FMS systems, and in particular to a distribution method and device based on intelligent FMS pallets and a computer readable storage medium. BACKGROUND

[0002] A flexible manufacturing system (FMS) is a highly automated production system that integrates numerical control machine tools, automated material handling equipment (such as AGV), computer control systems, and pallet storage systems, and is widely used in multi-variety, small-batch production scenarios in the fields of automobiles, aerospace, precision machinery, etc. In an FMS, pallets serve as the core carriers for carrying workpieces and facilitating the transfer between processes, and their distribution efficiency and accuracy directly affect the system's running rhythm, equipment utilization, and task response speed. Pallet distribution equipment or modules are usually integrated into the scheduling control system of an FMS, responsible for distributing appropriate idle pallets to material tasks to be executed according to production task requirements, and are the key link connecting production planning and physical execution. With the development of intelligent manufacturing, pallet distribution has evolved from early manual scheduling to rule-based automatic distribution, but its intelligence, flexibility, and fault tolerance still face serious challenges.

[0003] Existing FMS scheduling systems usually adopt a fixed distribution mode, i.e., the system is preset to "full automatic" or "full manual" mode at initialization, lacking the ability to dynamically switch according to task type, user permission, or production state. The root cause lies in the single flow control logic of the system architecture, without designing a mode judgment module and branch scheduling engine. This design stems from the early system's preference for "deterministic control", believing that the higher the automation level, the better, ignoring the necessity of human-machine collaboration. If this problem is not addressed, when an emergency order or equipment failure occurs, operators cannot manually specify the optimal pallet and must wait for the automatic system to be re-planned, causing a response delay; in addition, facing modern production demands of multi-variety, small-batch, and high-change, a single mode cannot balance efficiency and flexibility. SUMMARY

[0004] The present application aims to address the shortcomings of the prior art and provides a distribution method, device, and computer readable storage medium based on intelligent FMS pallets, which can effectively solve the above problems.

[0005] To achieve the above requirements, the technical solution adopted by the present application to solve its technical problems is as follows:

[0006] A distribution method based on intelligent FMS pallets is provided, comprising the following steps:

[0007] S1: receiving a start instruction of a material task by a user;

[0008] S2: in response to the start instruction, determining whether the allocation mode is an automatic allocation mode or a manual allocation mode;

[0009] S3: if the allocation mode is the automatic allocation mode, automatically filtering matching trays from all unallocated and empty trays according to the required tray quantity and tray type of the material task, and allocating the filtered trays to the material task until the allocated tray quantity equals the required quantity;

[0010] S4: if the allocation mode is the manual allocation mode, providing a tray selection interface to allow the user to select target trays through system adaptive recommendation or manual input, and allocating the user-selected trays to the material task until the required quantity of allocation is completed;

[0011] S5: wherein the system adaptive recommendation mode is to automatically list available tray numbers matching the tray type of the current material task in the drop-down box of the tray selection interface for the user to directly select.

[0012] S6: before or at the same time as S2, the system constructs a mode recommendation engine based on the current system state, which analyzes the availability of the required tray type, the load balancing degree of the production line, whether there is a high-priority urgent task insertion, and recommends a better allocation mode to the user based on preset rules or a lightweight model, and the user can choose to accept the recommendation or insist on the original instruction;

[0013] S7: in the manual allocation mode, when the user rejects the tray recommended by the system and selects another tray, or after making a selection in conflict negotiation, the system records this operation as a "human feedback sample" and tracks the final execution effect of the decision for subsequent optimization of the recommendation strategy and learning model;

[0014] S8: in the manual allocation mode, when the tray number manually input by the user is type-matched but already occupied, the system starts a conflict negotiation mechanism, providing an "wait for allocation" option, while adding a "seize" option (requires high-level permission) and a "reservation" function, and the user can reserve the priority use of the tray;

[0015] S9: in the conflict negotiation interface of S8, the system quantifies and displays the conflict cost of "waiting for the tray will cause task delay X minutes" or "selecting alternative solution Y will increase logistics cost Z yuan", to assist the user in making a cost-effective decision;

[0016] S10: In the multi-objective optimization algorithm in the system, the "fixture compatibility cost" is explicitly defined as a composite cost, including fixture type matching degree, installation adjustment time cost and estimated wear cost, which is calculated through historical data and equipment database; The "tray historical use deviation degree" is explicitly defined as an index for measuring the load balancing of similar trays, and the Gini coefficient or standard deviation is used to calculate the cumulative working time distribution of all similar trays.

[0017] Preferably, the manual allocation mode further includes a cancel allocation function. During task execution, the user selects the allocated tray number in the tray list and cancels the allocation state of the tray by right-clicking to release the tray for use by other tasks.

[0018] Preferably, in the automatic allocation process, if the system does not find an empty tray that meets the type requirement in the current available trays, it continues to monitor the tray state until a tray that meets the requirement becomes idle and completes the allocation.

[0019] Preferably, the manual input mode is that in the tray selection interface, the user does not select the drop-down list content recommended by the system, but directly inputs the target tray number in the input box. After the system verifies that the tray number corresponds to a tray type matching and the state is empty, it allocates it to the current material task.

[0020] Preferably, in the automatic allocation mode, the system further constructs a tray usage frequency model based on historical task execution data and predicts the future release time window in combination with the current production line load state.

[0021] When the number of empty trays of the required type is insufficient, the system preferentially sends a dispatch preparation instruction to the station where the task that will soon complete processing and matches the tray type is located, to start the unloading and tray return process in advance, thereby shortening the overall waiting time.

[0022] Preferably, in the manual allocation mode, if the tray number input by the user is type-matched but already occupied, the system does not directly refuse, but starts a conflict negotiation mechanism: analyzes the estimated release time of the occupied tray, and provides the user with two options of "wait for allocation" or "recommend alternative";

[0023] If "recommend alternative" is selected, the system intelligently filters and recommends the optimal alternative from other available trays based on material size, fixture compatibility and shortest path principle, and completes the allocation after user confirmation.

[0024] Preferably, in the automatic allocation mode, the system uses a machine learning model based on reinforcement learning to dynamically optimize the tray scheduling strategy: the model takes tray state, task priority, station load, and historical scheduling efficiency as input features, and outputs the optimal allocation decision.

[0025] The model continuously updates parameters through online learning, so that the system continuously improves resource utilization and task response speed in a complex dynamic environment.

[0026] Preferably, in the manual allocation mode, when the user selects "recommended replacement" or the system automatically recommends, the system starts a multi-objective optimization algorithm to construct a target function:

[0027]

[0028] wherein, : transportation time from the current handling starting point to the target station (unit: seconds);

[0029] : clamp compatibility cost, defined as 1-compatibility, the higher the compatibility, the smaller;

[0030] : tray historical use deviation degree, reflecting the load balancing degree.

[0031] is an adjustable weight coefficient;

[0032] The system solves the optimization problem to select the tray with the best comprehensive score from the candidate tray set as the recommended result.

[0033] The intelligent FMS tray allocation device based on the above model comprises:

[0034] A task receiving module is configured to receive a material task and an allocation mode instruction initiated by a user.

[0035] A mode determining module is configured to determine whether to use an automatic allocation mode or a manual allocation mode for the current task.

[0036] An automatic allocation module is configured to automatically match and allocate idle trays according to the tray quantity and type required by the task.

[0037] A manual allocation module is configured to provide a graphical interface to support system adaptation recommendation selection or user manual input of a tray number for allocation.

[0038] A state management module is configured to monitor the allocation state and type information of each tray in real time to ensure the accuracy and safety of the allocation process.

[0039] A prediction scheduling module is configured to construct a tray use frequency model based on historical task data, and in combination with line load prediction and pre-release time window, send a pre-preparation instruction to the target station when resources are insufficient.

[0040] A conflict negotiation module is used to analyze the expected release time of the manually input tray when it is occupied, and provide a "wait for allocation" or "recommended alternative" option, wherein the recommended alternative function is based on the size of the material, clamp compatibility and the shortest principle of carrying path for intelligent screening.

[0041] The manual allocation module is also integrated with a tray de-allocation function, allowing users to operate through a graphical interface to release the binding relationship between the allocated tray and the task.

[0042] A computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the allocation method based on the intelligent FMS tray.

[0043] The beneficial effects of the present application are:

[0044] The allocation method and device based on the intelligent FMS tray, through the design of the automatic / manual dual-mode allocation mechanism, enables the system to dynamically switch the tray allocation strategy according to user instructions, realizing flexible and configurable resource scheduling functions. Compared with the rigid allocation mode of the prior art which only supports single automatic or manual mode, the operation flexibility and scene adaptability of the FMS system are significantly improved, especially in dynamic production environments such as emergency order insertion and task changes, which can effectively avoid scheduling interruptions caused by mode locking.

[0045] Through the design of the system adaptation recommendation mode, in the manual allocation mode, the drop-down box of the tray selection interface can automatically list the available tray numbers matching the current material task tray type, realizing intelligent pre-filtering based on task requirements. Compared with the prior art which requires users to query tray status one by one, the manual search time is greatly reduced, the operation efficiency is improved, and the risk of human error selection is reduced.

[0046] Through the design of the de-allocation function, users can release the allocated tray through right-click operation during task execution, realizing dynamic resource unbinding and re-allocation. Compared with the prior art in which the tray is locked to the task end after allocation, the turnover rate and scheduling flexibility of the tray resource are significantly improved, especially suitable for abnormal scenarios such as production plan adjustment and equipment failure, effectively avoiding resource idling and task blocking. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the present application will be further described below with reference to the drawings and embodiments. The drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor:

[0048] Figure 1System flowchart of the present application;

[0049] Figure 2 FMS system tray allocation module flowchart of the present application;

[0050] Figure 3 Flowchart of the smart FMS tray-based allocation method of the present application. DETAILED DESCRIPTION

[0051] The terms "first", "second", "third", and "fourth" and the like in the description and claims of the present application and the accompanying drawings, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. The terms "include" and "have" and their derivatives, mean "comprising" or "including", and are not limiting. For example, a process, method, object, or apparatus that "comprises" or "has" a list of steps or elements, does not necessarily comprise or have only those steps or elements but can include additional steps or elements not expressly listed or inherent to such process, method, object, or apparatus.

[0052] Reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "in one embodiment" or "an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is expressly understood that any of the embodiments described herein can be incorporated into any other embodiment.

[0053] "Multiple" means two or more. "And / or" describes an associative relationship between associated objects, indicating that there can be three relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. The character " / " generally represents an "or" relationship between the associated objects before and after it.

[0054] Furthermore, the terms "upper", "lower", "left", "right", "top", "bottom", "vertical", and the like, which indicate the orientation of the device or apparatus described in the present application, are based on the attitude position of the device or apparatus in normal use.

[0055] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0056] The embodiments of the present application disclose a smart FMS tray-based allocation method as shown in Figures 1 to 3 The smart FMS tray-based allocation method comprises the following steps:

[0057] S1: receiving a start instruction of a material task by a user;

[0058] S2: in response to the start instruction, determining whether the allocation mode is an automatic allocation mode or a manual allocation mode;

[0059] S3: if the allocation mode is the automatic allocation mode, automatically filtering matching trays from all trays that are not allocated with materials and are empty, and allocating the filtered trays to the material task until the number of allocated trays equals the required number of trays according to the required number and type of trays of the material task;

[0060] S4: if the allocation mode is the manual allocation mode, providing a tray selection interface to allow the user to select target trays through system adaptive recommendation or manual input, and allocating the selected trays to the material task until the required number of trays is allocated;

[0061] S5: the system adaptive recommendation mode is that the available tray numbers matching the tray type of the current material task are automatically listed in the drop-down box of the tray selection interface for the user to directly select;

[0062] S6: before or at the same time as S2, the system constructs a mode recommendation engine based on the current system state, the engine analyzes the availability of the required tray type, the load balancing degree of the production line, whether there is a high-priority emergency task insertion, and recommends a better allocation mode to the user based on preset rules or a lightweight model, the user can choose to accept the recommendation or insist on the original instruction;

[0063] S7: in the manual allocation mode, when the user rejects the tray recommended by the system and selects another tray, or after making a selection in conflict negotiation, the system records the operation as a "human feedback sample" and tracks the final execution effect of the decision for subsequent optimization of the recommendation strategy and learning model;

[0064] S8: in the manual allocation mode, when the tray number manually input by the user is type-matched but has been occupied, the system starts a conflict negotiation mechanism, provides an "wait for allocation" option, increases a "seize" option (requires high-level permission) and a "reservation" function, and the user can reserve the priority use of the tray;

[0065] S9: in the conflict negotiation interface of S8, the system quantifies and displays the conflict cost of "waiting for the tray will cause a task delay of X minutes" or "selecting an alternative solution Y will increase the logistics cost of Z yuan", to assist the user in making a cost-effective decision;

[0066] S10: In the multi-objective optimization algorithm in the system, the "fixture compatibility cost" is explicitly defined as a composite cost including fixture type matching degree, installation adjustment time cost and estimated wear cost, which is calculated through historical data and equipment database; the "tray historical use deviation degree" is explicitly defined as an index for measuring the load balance of similar trays, and the Gini coefficient or standard deviation is used to calculate the cumulative working time distribution of all similar trays. By defining the "fixture compatibility cost" as a composite cost function containing type matching degree, installation adjustment time and estimated wear cost, and defining the "tray historical use deviation degree" as a statistical index calculated by Gini coefficient or standard deviation, the fuzzy concept in the multi-objective optimization algorithm is converted into precise and quantifiable technical parameters, realizing the function of building a high-precision and interpretable intelligent optimization model. Compared with the existing technology, the optimization target definition is fuzzy, depends on experience and parameter adjustment, and the effect is unstable. The scientificity, stability and reproducibility of the optimization result are significantly improved, and the effect of fine and intelligent scheduling is truly realized. This concrete mathematical modeling enables the system to accurately measure the comprehensive cost of different scheduling schemes and make decisions that are more in line with actual production efficiency.

[0067] The core in S6 is to build a "mode recommendation engine" which aims to break the complete separation of "man-machine mode" in traditional systems, and to provide decision support for users by actively sensing the environment, and to realize man-machine intelligent cooperation. Specifically: (1) The system queries the database in real time, calculates the percentage of the number of currently idle and type-matched trays in the total number of trays of this type. For example, if there are 10 trays of a certain type, and 3 are currently idle, the availability rate is 30%. When the availability rate is below the threshold (such as 20%), it indicates that the resources are tight. (2) The system obtains the current task queue length and the estimated completion time of each processing station (such as CNC machine tool, detection station). Calculate the standard deviation or variance of the task queue length of all stations. The smaller the standard deviation, the more balanced the load. If the standard deviation is too large, it indicates that there is a bottleneck in the production line. (3) The system receives instructions from the upper MES (Manufacturing Execution System), and identifies new tasks with "urgent" and "high priority" labels. When the user starts a material task on the operation interface, the system displays a prominent icon (such as a light bulb) or text prompt "System suggestion: manual allocation" next to the "allocation mode" selection box. The user can click on the prompt to view the recommended reasons (such as "There are only 2 heavy-duty trays left, and there is an urgent order"). The user can choose to "accept the suggestion" or "ignore the suggestion and execute the original plan", and the system records this interaction. By designing a mode recommendation engine module and executing the mode recommendation step after receiving user instructions and before determining the allocation mode, the system can actively sense and analyze the system status such as tray availability, production line load balancing, and the presence of high-priority urgent tasks, and can actively recommend the optimal allocation mode to the user. Compared with the existing technology, which is completely passive in responding to user instructions and cannot provide decision assistance, the system significantly improves the efficiency of human-machine collaborative decision-making and the level of system intelligence. Specifically, when the system detects resource shortages or bottlenecks, it will actively recommend manual allocation to guide operators to conduct fine scheduling; when resources are sufficient, it will recommend automatic allocation to improve efficiency. This active and context-based recommendation mechanism effectively avoids mode selection errors caused by insufficient user experience or incomplete information, allowing the system to always maintain an optimal operating state in a dynamic production environment. By designing a mode recommendation engine module to analyze system status and recommend allocation modes to users in real time, the system not only provides decision recommendations (expected functions), but also generates a system health diagnosis report during its operation, achieving a passive monitoring to active warning transition function, and achieving the unexpected effect of exposing system bottlenecks and potential risks before the user is aware of them. For example, when the engine frequently recommends manual mode due to "low tray availability", this recommendation behavior itself is a strong system resource warning signal, prompting administrators to check tray flow efficiency or increase tray quantity.Further, the recommended result of the module can trigger the "prediction scheduling module" in linkage, for example, when the system continues to recommend the manual mode and the reason is that a certain type of tray is in short supply, the prediction priority of the type of tray is automatically raised, and the "pre-preparation instruction" is actively sent to the relevant station, forming a "recommendation-early warning-active intervention" closed loop.

[0068] In S7-S8, a higher level of "human-machine cooperation" and "resource scheduling" mechanism is constituted, which promotes the operator from a passive executor to a decision participant. Specifically: (1) when the system recommends tray A for the user in manual mode, but the user finally selects tray B, or in conflict negotiation, the system recommends alternative X, and the user selects Y; (2) the system creates a "human feedback sample" record, including: task ID, timestamp, system original recommendation, user actual selection, user identity (permission level), and short reason filled by the user when selecting; (3) the system will track the subsequent execution of the task, such as task completion time, whether an exception occurs, actual logistics path time consumption, etc.; (4) these samples and their effect data will be used as valuable training data for subsequent optimization of "system adaptive recommendation" algorithm and "reinforcement learning model", especially for correcting the cognitive bias of the system, such as "short path but actual congestion" or "type matching but fixture adjustment time-consuming". Through the design of collecting the user's rejection of the system's recommendation or the selection made in conflict negotiation as "human feedback sample" and tracking its execution effect, the system can convert the on-site decision-making experience of the operator into learnable data, realizing the function of building a "human feedback reinforcement learning (RLHF)" closed loop. Compared with the limitations of existing machine learning models that rely only on historical task data and are difficult to integrate human expert experience, the system scheduling strategy can evolve from human wisdom, significantly improving decision-making accuracy and adaptability in complex scenarios. At the same time, through the design of quantifying and displaying the expected time delay and logistics cost of the waiting or alternative scheme in the conflict negotiation interface, the abstract resource conflict is converted into specific economic and time indicators, realizing the function of assisting users in making scientific decisions based on cost-effectiveness. Compared with the existing technology of providing simple options and relying on user experience for rough processing, the effect of significantly reducing decision-making risk and optimizing overall production cost and schedule management is achieved.

[0069] In S8, (1) when the user selects "preempt", the system first verifies the user's authority. If the authority is sufficient, the system sends a high-priority notification (such as a pop-up window, email) to the task owner currently occupying the tray, informing that the tray will be urgently requisitioned. At the same time, the system finds an alternative tray for the original task or suspends it. This function is only used in extremely urgent situations and requires strict authority control; (2) when the user selects "wait for allocation" and clicks "reserve", the system adds the "future idle time point" of the tray and the current task ID to a "reservation queue". When the tray is released, the system will immediately interrupt other possible allocation processes and prioritize the allocation of the tray to the task that has been reserved, and send a notification to the original task owner "The tray P101 you reserved has been successfully allocated". By adding the "preempt" (high-level authority required) and "reserve" functions to the conflict negotiation mechanism, users can not only choose to "wait" or "replace" when facing an occupied tray, but also actively intervene and lock future resources. Compared with the existing technology, which lacks advanced resource coordination methods and easily leads to emergency task blocking, the system greatly enhances the scheduling flexibility and response capability in emergency situations. The "preempt" function with high-level authority allows breaking the regular scheduling in extremely urgent situations to ensure the priority execution of critical tasks; while the "reserve" function avoids disordered competition for the same tray by establishing resource usage expectations, achieving orderly and predictable resource usage, and effectively improving the system stability in high-concurrency scenarios.

[0070] In the conflict negotiation interface of step S8, the system quantifies and displays the conflict cost of "waiting for the tray will cause task delay X minutes" or "selecting alternative solution Y will increase logistics cost Z yuan", assisting users in making cost-effective decisions.

[0071] Specifically, the system runs on an industrial control computer or a server, receives a task starting instruction from an operator or a higher-level MES system, and the instruction contains a task ID, a material type, a required number of pallets, a pallet type, and an allocation mode (automatic or manual). When the system receives the starting instruction, it first parses the "allocation mode" field and enters the corresponding branch process: (1) If it is an automatic allocation mode: the system calls the database query module to retrieve all pallet records with a status of "idle" and not bound to a task; then it filters according to the "pallet type" required by the task to form a candidate pallet set; finally, it selects pallets from the candidate set in turn according to a preset strategy (such as shortest transportation path priority or FIFO) until the number of allocations meets the task requirements. After the allocation is completed, the system updates the pallet status to "allocated" and binds it to the task ID, while recording the timestamp.

[0072] In this embodiment, the manual allocation mode also includes a cancel allocation function. During the execution of a task, the user selects a pallet number that has been allocated in the pallet list and cancels the allocation status of the pallet by operation, releasing the pallet for use by other tasks. Through the collaborative design of the continuous monitoring mechanism and the prediction scheduling module in the automatic allocation mode, the system not only passively waits when it cannot find available pallets, but also actively calls the prediction scheduling module to analyze the station where the pallet will be released soon and sends a "preparation instruction", realizing the mode transition function from passive polling to active intervention, achieving the effect of significantly shortening the pallet waiting time (the average reduction is 35%) and reducing the AGV empty running rate. This effect is much better than the independent effect of a single "waiting" or "prediction" mechanism. Through the reverse triggering design of the cancel allocation function and the prediction scheduling module, when the user manually releases the pallet and the behavior is captured by the state management module, not only is the local state updated, but also the prediction scheduling module is notified of the "resource early release", so as to adjust the priority of the original preparation instruction or cancel redundant scheduling, realizing the feedback loop function between manual intervention and system prediction, achieving the energy-saving effect of avoiding excessive scheduling and saving energy and equipment wear and tear. This linkage mechanism enables the system to have the ability of "self-adaptive correction", which is much better than the traditional one-way scheduling logic.

[0073] Specifically, the system identifies the allocated pallets in the graphical interface with a specific visual style (such as a red border). The user can right-click on the pallet number to trigger a context menu, and the menu contains a "cancel allocation" option. After selection, the system pops up a confirmation dialog box to prevent misoperation;

[0074] After confirmation, the system performs the following operations:

[0075] The "task binding field" of the tray is set to null;

[0076] Its status is updated to "idle";

[0077] Operation logs (including operation time, operator, tray number) are recorded;

[0078] Interface refresh is triggered, and the tray reappears in the available tray list. This function is particularly suitable for scenarios such as task changes, urgent orders, or misallocation, effectively improving tray utilization.

[0079] In this embodiment, during the automatic allocation process, if the system does not find an empty tray that meets the type requirements in the current available trays, it continues to monitor the tray status until a tray that meets the requirements becomes idle and completes the allocation.

[0080] Specifically, when the system cannot immediately find an idle tray that meets the type and quantity requirements in automatic allocation mode, an asynchronous state monitoring mechanism is started. The system places the current task in the "waiting queue" and starts a background monitoring process. This process polls the tray database at fixed time intervals (such as every 5 seconds) to detect whether there is a tray that meets the task requirements.

[0081] The monitoring conditions include: (1) the tray type matches the task requirements; (2) the tray state becomes idle; (3) the tray quantity meets the task requirements; once the conditions are met, the system immediately selects the optimal one (such as the closest) from the candidate trays to complete the allocation, and updates the task status to "allocated", while notifying the user.

[0082] In this embodiment, the manual input method is: in the tray selection interface, the user does not select the drop-down list content recommended by the system, but directly inputs the target tray number in the input box. After the system verifies that the tray number corresponds to a tray type that matches and is empty, it allocates it to the current material task.

[0083] Specifically, in manual allocation mode, the user can directly input the tray number in the input box, and the system needs to verify the legality of the input. The verification process is as follows:

[0084] (1) The system obtains the tray number input by the user;

[0085] (2) Query the database to confirm whether the tray corresponding to the number exists;

[0086] (3) If it exists, further verify whether its "tray type" is consistent with the current task requirements;

[0087] (4) At the same time, check if its "state" is idle (i.e. not occupied by other tasks);

[0088] (5) Only when the type matches and the state is idle, the system allows allocation and updates the database record. If either condition is not met, the system returns an error prompt to the user, such as "tray does not exist", "type mismatch" or "tray is occupied". This mechanism ensures the accuracy of manual operation and system safety.

[0089] In the automatic allocation mode in this embodiment, the system further constructs a tray usage frequency model based on historical task execution data, and predicts the future pre-release time window in combination with the current production line load state;

[0090] When the number of empty trays of the required type is insufficient, the system preferentially sends a scheduling preparation instruction to the station where the task that will be completed soon and the tray type matches, to start the unloading and tray return process in advance, so as to shorten the overall waiting time.

[0091] Through the design of the prediction scheduling mechanism, the system constructs a tray usage frequency model based on historical task data, and predicts the future pre-release time window in combination with the current production line load, and sends a scheduling preparation instruction to the station where the task will be completed soon when resources are insufficient, to realize active intervention and pre-return function. Compared with the passive waiting tray release mode in the prior art, the overall waiting time is significantly shortened, and the production line coordination efficiency is improved.

[0092] Specifically, the system automatically identifies the required tray type according to the material characteristics, realizes intelligent matching, and the material characteristics include: size, weight, shape, processing accuracy requirement, etc. The system has a built-in tray type mapping rule library, and the mapping relationship is shown in the following table:

[0093] Material characteristics Recommended tray type Large size, heavy weight Heavy duty load tray Round, easy to roll Round tray Precision parts, need to be positioned Precision jig tray Ordinary square workpiece Square tray

[0094] The system analyzes the characteristic parameters in the material task, calls the rule engine for matching, and automatically determines the required tray type without manual intervention. Through the design of the continuous monitoring mechanism, if there is no available tray during the automatic allocation process, the system can continuously poll the tray state until a tray that meets the requirements is released, realizing asynchronous waiting and automatic capture function. Compared with the processing method of directly reporting an error or interrupting allocation in the prior art, the system robustness and task completion rate are improved.

[0095] In the manual allocation mode in this embodiment, if the tray number input by the user is type-matched but occupied, the system does not directly refuse, but starts a conflict negotiation mechanism: analyzes the expected release time of the occupied tray, and provides the user with two options of "wait for allocation" or "recommend alternative";

[0096] If the "recommended replacement" is selected, the system intelligently filters and recommends the optimal replacement solution from other available pallets based on material size, fixture compatibility, and shortest path principle, and completes the allocation after user confirmation.

[0097] Through the design of manual input verification mechanism, users can directly input the pallet number, and the system completes the allocation after verifying that the type matches and the state is empty, realizing the function of quick and direct allocation. Compared with the cumbersome process of selecting through menus in the prior art, the operation efficiency of advanced users is improved, and the data integrity and system security are guaranteed through triple verification (existence, type, and state).

[0098] Through the linkage design of manual input and conflict negotiation mechanism, the user's input of an occupied pallet number not only triggers an error prompt, but also activates the conflict negotiation module, which in turn starts the estimated release time calculation and replacement recommendation process, realizing the function of converting user "error operation" into system intelligent decision trigger event, and achieving the dual effects of improving system fault tolerance and human-machine collaboration efficiency. This design converts the traditional input behavior considered as "abnormal" into a valuable scheduling signal, which has significant non-obviousness.

[0099] Specifically, the system constructs a pallet usage frequency model based on historical task execution data:

[0100]

[0101] Wherein:

[0102] : the average usage frequency of a certain type of pallet in the time period ;

[0103] : the usage duration of the task ;

[0104] : the total number of tasks in the statistical period.

[0105] Combined with the current production line load (such as the number of tasks at the station, the equipment operation rate), the pallet release situation in the future time is predicted;

[0106] When the required pallets are insufficient, a "scheduling preparation instruction" is sent to the station that will complete the processing, prompting early unloading and shortening the return time.

[0107] The dynamic generation mechanism of the system adaptation recommended drop-down box is synchronized in real time with the state management module, so that the drop-down box content is automatically refreshed with the tray state change, not only providing initial recommendations, but also continuously updating the available tray list during user operation, realizing the functional evolution from a static recommendation interface to a dynamic resource dashboard, and achieving the effect of avoiding user selection of invalid trays and reducing repeated operations, which is particularly significant in a multi-user concurrent operation scenario.

[0108] In the automatic allocation mode, the system uses a machine learning model based on reinforcement learning to dynamically optimize the tray scheduling strategy: the model takes the tray state, task priority, workstation load, and historical scheduling efficiency as input features, and outputs the optimal allocation decision.

[0109] The model continuously updates the parameters through online learning, enabling the system to continuously improve resource utilization and task response speed in complex dynamic environments.

[0110] In the reward function of the reinforcement learning machine learning model, an item of "consistency with expert operation" is added as a reward or penalty. This "expert operation" is derived from the "human feedback sample". By adding the "consistency with expert operation" reward or penalty item in the reward function of the reinforcement learning model, the optimization goal of the model is no longer limited to direct indicators such as task waiting time and utilization, but also includes the decision preferences of human experts, achieving a deep integration of scheduling strategies and human experience. Compared with the potential risk of "counterintuitive" or "unsafe" decisions that may be made by reinforcement learning models in existing technologies, this design effectively guides model learning and inherits the tacit knowledge of human experts, improving the acceptability and safety of scheduling decisions. This design ensures that the system maintains consistent behavior patterns with experienced operators while pursuing efficiency, accelerating the application of intelligent systems in real production environments.

[0111] Through the design of the reinforcement learning scheduling optimization model, the system takes the tray state, task priority, workstation load, and historical scheduling efficiency as input in the automatic allocation mode to dynamically optimize allocation decisions and continuously update model parameters through online learning, achieving self-adaptive and self-evolution intelligent scheduling functions. Compared with the static scheduling method based on fixed rules in existing technologies, the system significantly improves resource utilization and task response speed in complex dynamic environments.

[0112] The system introduces a reinforcement learning model in the automatic allocation mode to dynamically optimize scheduling strategies. The model structure is a deep Q network (DQN) or PPO algorithm, and its input state s includes:

[0113] Tray state matrix

[0114] Task priority vector

[0115] Workstation load vector

[0116] Historical scheduling efficiency

[0117] Action space Select a pallet from the candidate pallet set for allocation. Reward function. :

[0118]

[0119] in:

[0120] Task wait time variation (the smaller the better);

[0121] Increased pallet utilization;

[0122] Conflict probability ;

[0123] Weighting coefficients, satisfying .

[0124] The model continuously updates its parameters by interacting with the environment, achieving long-term performance optimization.

[0125] Specifically, when a user manually enters the number of a currently occupied tray, the system initiates a conflict negotiation mechanism. The system first calls the task scheduling module to obtain the estimated release time of the tray, calculated as follows:

[0126]

[0127] in:

[0128] : Expected release time;

[0129] Current time;

[0130] The sum of the remaining processing times for the tasks currently bound to this tray.

[0131] Offer users two options:

[0132] "Waiting for assignment": The system is in Automatic time allocation;

[0133] "Recommended replacement": the system starts the replacement recommendation.

[0134] If the user selects "recommended replacement", the system filters the candidate pallets based on the following three dimensions: (1) material size compatibility: pallet size ≥ material size; (2) fixture compatibility: fixture type matching degree (0-1); (3) shortest carrying path: the shortest distance from the current carrying starting point to the target station.

[0135] In this embodiment, in the manual allocation mode, when the user selects "recommended replacement" or the system automatically recommends, the system starts a multi-objective optimization algorithm to construct the objective function:

[0136]

[0137] Wherein, : transportation time from the current carrying starting point to the target station (unit: seconds);

[0138] : fixture compatibility cost, defined as 1-compatibility, the higher the compatibility, the smaller the cost;

[0139] : pallet historical use deviation, reflecting the load balancing degree.

[0140] is a tunable weight coefficient;

[0141] The system solves this optimization problem to filter out the best pallet with the best comprehensive score from the candidate pallet set as the recommended result;

[0142] Wherein, the fixture compatibility cost and the pallet historical use deviation are defined and calculated by S10. By explicitly using the composite cost and statistical indicators in claim 1 S10 as the optimization target in the multi-objective optimization algorithm, the entire optimization solving process is based on a solid and accurate mathematical foundation, realizing the paradigm upgrade function from fuzzy rules to accurate calculation. Compared with the existing technology, the optimization algorithm has a large deviation due to the inaccuracy of input parameters, which greatly improves the accuracy and reliability of resource allocation decisions, ensuring the long-term stable and efficient operation of the system. This is not only an improvement of the algorithm, but also a deepening of the entire intelligent scheduling methodology.

[0143] The intelligent FMS pallet allocation device is a distributed software system running on an industrial server, with each module communicating through a message bus, including:

[0144] Task receiving module, used to receive user-initiated material tasks and allocation mode instructions; ​

[0145] A mode determination module is configured to determine whether to adopt an automatic allocation mode or a manual allocation mode for the current task.

[0146] An automatic allocation module is configured to automatically match and allocate the idle trays according to the tray quantity and type required by the task.

[0147] A manual allocation module is configured to provide a graphical interface to support system adaptation recommendation selection or user manual input of tray numbers for allocation.

[0148] A state management module is configured to monitor the allocation state and type information of each tray in real time to ensure the accuracy and safety of the allocation process.

[0149] A prediction scheduling module is configured to construct a tray usage frequency model based on historical task data, and in combination with production line load prediction and pre-release time window, send a pre-preparation instruction to the target station when resources are insufficient.

[0150] A conflict negotiation module is configured to analyze the expected release time of a manually input tray when it is occupied, and provide a "wait for allocation" or "recommended replacement" option, wherein the recommended replacement function is based on the material size, clamp compatibility, and shortest carrying path principle for intelligent screening.

[0151] A mode recommendation engine module is configured to analyze the availability of the required tray type, production line load balancing, and whether there is a high-priority emergency task insertion based on the current system state after receiving a start instruction, and provide allocation mode recommendations to the user.

[0152] A human feedback collection module is configured to collect user feedback samples in the manual allocation mode, such as the user's rejection of system recommendations or choices made in conflict negotiation, and track the execution effect.

[0153] A conflict cost quantification module is configured to calculate and display the expected time delay and logistics cost of the waiting or replacement scheme during conflict negotiation. By quantifying and displaying the waiting time and replacement cost in the conflict negotiation interface, the user decision-making process is highly transparent (expected function), which unexpectedly becomes a real-time classroom for production management knowledge training for operators, achieving the implicit training function of improving the cost awareness and decision-making ability of operators, and achieving the unexpected effect of improving the production management level of the entire team in daily operations.

[0154] The manual allocation module further integrates a tray de-allocation function, allowing users to operate through a graphical interface to release the binding relationship between the allocated tray and the task.

[0155] Specifically, by collecting users' "corrective samples" in manual mode and tracking the design of the execution effect, this module not only provides training data for the reinforcement learning model (expected function), but also becomes a gold standard for evaluating and optimizing the performance of the "system adaptation recommendation" algorithm. The database of "human feedback samples" collected by it itself realizes the dual role of serving model training and algorithm verification, and achieves the effect of continuously iterating and optimizing the recommendation accuracy without the need to build an additional test set. In addition, when the system detects that the "feedback sample" of a senior operator consistently leads to better execution results, this information can trigger the "conflict negotiation module" to prioritize the historical selection preferences of this "expert" when recommending alternative solutions to other users, realizing the cross-user sharing and inheritance of "expert experience".

[0156] By defining "fixture compatibility cost" as a composite function including estimated wear cost, and defining "tray historical usage deviation degree" as a statistical indicator (such as Gini coefficient) based on cumulative working time, the multi-objective optimization not only optimizes the current task (expected function), but also automatically realizes the whole life cycle health management of the tray assets, realizing the dual role of deep integration of scheduling optimization and equipment maintenance, achieving the far-reaching effect of prolonging the service life of the tray, reducing equipment maintenance cost, and preventing sudden failure. More importantly, this concrete and data-based optimization target provides high-quality and interpretable feedback signals for the online learning of the "reinforcement learning model" triggered in conjunction. The model not only learns "which tray is good", but also understands "why it is good" (because of low wear cost or because of load balancing), so as to learn more essential and robust scheduling strategies, rather than simple pattern matching.

[0157] By adding the "reservation" function in conflict negotiation, users can lock future resources (expected function), which unexpectedly builds a lightweight "tray futures market", realizes the function of predicting future tray demand through user reservation behavior, and achieves the unexpected effect of providing more accurate and forward-looking demand signals for the "predictive scheduling module". The system's prediction no longer relies solely on historical data and work orders, but also integrates users' "market expectations" (i.e. reservation behavior), greatly improving the accuracy of the prediction model. The "occupation" function, when executed, will trigger the "state management module" and "cancel allocation function" in conjunction, automatically releasing the resources of the original binding task of the occupied tray, forming a complete and automated emergency response chain.

[0158] A computer readable storage medium, which stores a computer program, the program is executed by a processor to realize the allocation method based on the intelligent FMS tray, the medium (such as SSD, U disk, optical disk) stores the computer program, and when the processor loads and executes the program, the complete tray allocation function is realized.

[0159] It should be understood that, for those skilled in the art, improvements or changes can be made according to the above description, and all these improvements and changes shall belong to the protection scope of the appended claims of the present application.

Claims

1. A distribution method based on an intelligent FMS tray, characterized in that, Includes the following steps: S1: Receive the user's start command for the material task; S2: In response to the startup command, determine whether the allocation mode is automatic allocation mode or manual allocation mode; S3: If it is an automatic allocation mode, then according to the number of pallets and pallet type required by the material task, automatically filter matching pallets from all unallocated materials and empty pallets, and allocate the filtered pallets to the material task until the number of allocated pallets equals the required number. S4: If it is a manual allocation mode, a pallet selection interface is provided, allowing users to select target pallets through system adaptation recommendations or manual input, and assigning the user-selected pallets to the material task until the required number of pallets are allocated; S5: The system adaptation recommendation method is as follows: the available pallet numbers that match the current material task pallet type are automatically listed in the drop-down box of the pallet selection interface for the user to select directly; S6: Before or simultaneously with S2, the system builds a pattern recommendation engine based on the current system status. This engine analyzes the availability of the required tray type, the production line load balancing, and whether there are high-priority emergency tasks being inserted. Based on preset rules or a lightweight model, it recommends a better allocation pattern to the user. The user can choose to accept the recommendation or stick to the original instructions. S7: In the manual allocation mode, when a user rejects the system-recommended tray and selects another tray, or after making a choice during conflict negotiation, the system records the operation as a "human feedback sample" and tracks the final execution effect of the decision for subsequent optimization of recommendation strategies and learning models.

2. The allocation method based on intelligent FMS tray according to claim 1, characterized in that: The manual allocation mode also includes a cancellation function. During task execution, the user can select an allocated tray number from the tray list and cancel the allocation status of the tray, releasing the tray for use by other tasks.

3. The allocation method based on intelligent FMS tray according to claim 1, characterized in that: During the automatic allocation process, if the system does not find an empty pallet that meets the type requirements in the currently available pallets, it will continuously monitor the pallet status until a pallet that meets the requirements becomes idle and the allocation is completed.

4. The allocation method based on intelligent FMS tray according to claim 2, characterized in that: The manual input method is as follows: In the pallet selection interface, the user does not select the drop-down list recommended by the system, but directly enters the target pallet number in the input box. After the system verifies that the pallet type corresponding to the number matches and the status is empty, it assigns it to the current material task.

5. The allocation method based on intelligent FMS tray according to claim 3, characterized in that: In the automatic allocation mode, the system further constructs a pallet usage frequency model based on historical task execution data, and predicts the future pre-release time window by combining the current production line load status; When there is insufficient quantity of empty pallets of the required type, the system prioritizes sending scheduling preparation instructions to the workstations where tasks with matching pallet types are about to be completed, thus initiating the unloading and pallet return process in advance to shorten the overall waiting time.

6. The allocation method based on intelligent FMS tray according to claim 1, characterized in that: The method further includes: S8: In the manual allocation mode, when the tray number manually entered by the user is of the same type but is already occupied, the system starts the conflict negotiation mechanism. In addition to providing the "wait for allocation" option, it adds the "preemption" option and the "reservation" function, allowing the user to reserve the priority right to use the tray. S9: In the conflict negotiation interface of S8, the system quantifies and displays the conflict costs of "waiting for this pallet is expected to cause a task delay of X minutes" or "choosing alternative Y will increase logistics costs by Z yuan", assisting users in making cost-benefit-based decisions. S10: In the multi-objective optimization algorithm of the system, "fixture compatibility cost" is explicitly defined as a composite cost, including fixture type matching degree, installation and adjustment time cost and estimated wear cost, which is calculated through historical data and equipment database; "pallet historical usage deviation" is explicitly defined as an indicator to measure the load balance of similar pallets, and the cumulative working time distribution of all similar pallets is calculated using the Gini coefficient or standard deviation. In the manual allocation mode, if the user-input tray number is of the same type but is already occupied, the system does not reject it directly, but instead initiates a conflict negotiation mechanism: analyzes the estimated release time of the occupied tray and provides the user with two options: "wait for allocation" or "recommend alternative". If "Recommended Alternative" is selected, the system will intelligently filter and recommend the best alternative from other available pallets based on material size, fixture compatibility, and shortest path principles. The allocation will be completed after user confirmation.

7. The allocation method based on intelligent FMS tray according to claim 5, characterized in that: In the automatic allocation mode, the system uses a reinforcement learning-based machine learning model to dynamically optimize the pallet scheduling strategy: the model takes pallet status, task priority, workstation load, and historical scheduling efficiency as input features and outputs the optimal allocation decision. The model continuously updates its parameters through online learning, enabling the system to continuously improve resource utilization and task response speed in complex and dynamic environments. In particular, the reward function of the reinforcement learning machine learning model is supplemented with a reward or penalty for "consistency with expert actions", which is derived from the aforementioned "human feedback samples".

8. The allocation method based on intelligent FMS tray according to claim 6, characterized in that: In the manual allocation mode, when the user selects "recommended alternative" or the system makes an automatic recommendation, the system starts a multi-objective optimization algorithm to construct the objective function: ,in, : Transportation time from the current handling point to the target workstation (unit: seconds); Fixture compatibility cost, defined as 1 - compatibility degree; the higher the compatibility degree, the higher the cost. The smaller; Historical usage deviation of the pallet reflects the degree of load balance; These are adjustable weighting coefficients; The system solves the optimization problem and selects the pallet with the best comprehensive score from the candidate pallet set as the recommendation result. The fixture compatibility cost and the pallet historical usage deviation are defined and calculated using the methods described in S10. By defining the fixture compatibility cost as a composite cost function that includes type matching degree, installation and adjustment time and estimated wear cost, and defining the pallet historical usage deviation as a statistical index calculated using the Gini coefficient or standard deviation, the fuzzy concepts in the multi-objective optimization algorithm are transformed into precise and quantifiable technical parameters.

9. A dispensing device based on an intelligent FMS tray, characterized in that, include: The task receiving module is used to receive material tasks and allocation mode instructions initiated by the user. The mode determination module is used to determine whether the current task is assigned in automatic or manual mode. The automatic allocation module is used to automatically match and allocate tasks from available trays based on the number and type of trays required by the task. The manual allocation module provides a graphical interface to support system adaptation recommendations or allows users to manually enter the tray number for allocation; The status management module is used to monitor the allocation status and type information of each tray in real time to ensure the accuracy and security of the allocation process. The predictive scheduling module is used to build a pallet usage frequency model based on historical task data, and combine it with the production line load to predict the pre-release time window, and send pre-preparation instructions to the target workstation when resources are insufficient. The conflict negotiation module is used to analyze the expected release time when a manually entered pallet is occupied and provides "wait for allocation" or "recommended alternative" options. The recommended alternative function intelligently filters based on material size, fixture compatibility and the shortest handling path. The pattern recommendation engine module is used to analyze the availability of the required tray type, production line load balancing, and whether there are high-priority urgent tasks to be inserted based on the current system status after receiving the start command, and to provide the user with an allocation pattern recommendation. The human feedback collection module is used to collect user feedback samples in manual allocation mode, such as when users reject system recommendations or make choices during conflict negotiation, and to track the effectiveness of these choices. The conflict cost quantification module is used to calculate and display the expected time delays and logistical costs of waiting or alternative solutions during conflict negotiation. The manual allocation module also integrates a tray unallocation function, allowing users to unbind the allocated tray from the task through a graphical interface.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the allocation method based on the intelligent FMS tray as described in any one of claims 1 to 8.

Images