Dynamic storage method, device, equipment and medium for semiconductor materials

Abstract

The application relates to a dynamic storage method, device, equipment and medium for semiconductor materials. The method quantifies attribute information of the materials into activity values representing carrying priorities through an activity prediction model, allocates storage areas according to the activity values of the materials, ensures that high-activity materials are preferentially matched to storage areas convenient for rapid access, low-activity materials are stored in storage areas far away and with low access efficiency, the predicted carrying time of each candidate position to a target machine is obtained, a target position is determined from the candidate positions based on at least the predicted carrying time corresponding to each candidate position, and the time consumption of single carrying is further compressed; the allocation of storage positions and the priority matching of material carrying are realized, so that high-priority materials have faster flow efficiency, the average transmission time of high-activity material delivery is greatly shortened, the carrying amount of storage areas far away and with low access efficiency is reduced, and material congestion is alleviated.

Description

Technical Field

[0001] This application relates to the field of semiconductor manufacturing technology, and in particular to a method, apparatus, device, and medium for dynamic storage of semiconductor materials. Background Technology

[0002] In semiconductor wafer manufacturing, wafers are typically carried in front-opening unified pods (FOUPs) and moved between different process stations via an automated material handling system (AMHS). Once the FOUP has completed its processing at the current station, it needs to be moved to a storage area near the next station for temporary storage, waiting to be loaded onto the next station when it becomes available.

[0003] Currently, the next target machine for the FOUP is usually determined by the Manufacturing Execution System (MES) or Real-Time Dispatching (RTD). The Material Control System (MCS) determines whether there are still empty spaces in the adjacent storage locations of the FOUP's current machine or the next target machine based on the list of adjacent storage locations for each machine. If so, the FOUP is stored there.

[0004] However, this method of allocating adjacent storage locations requires binding each machine to an adjacent storage location, resulting in a large maintenance workload, unreasonable path planning, and easy material congestion. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, device, or medium for dynamic storage of semiconductor materials to address the aforementioned technical problems.

[0006] In a first aspect, this application provides a method for dynamic storage of semiconductor materials, the method comprising:

[0007] Obtain the attribute information of the material, which includes at least the material type, inventory status, holding status, and target machine.

[0008] The material's attribute information is input into the activity prediction model to predict the material's activity value at the target machine. The material's activity value represents the material's handling priority.

[0009] Based on the activity value of the material, determine the target storage area that matches the material, and select candidate storage locations within the target storage area;

[0010] Obtain the predicted handling time from each of the candidate warehouses to the target machine;

[0011] The target warehouse is determined from the candidate warehouses based at least on the predicted handling time corresponding to each candidate warehouse.

[0012] In one embodiment, the step of inputting the property information of the material into an activity prediction model to predict the activity value of the material at the target machine includes:

[0013] Obtain historical material handling records, which include at least the material type, inventory status, holding status of the material being handled, the machine identifier of the target machine on which the material being handled is loaded, and a tag value representing the handling time.

[0014] Construct an initial activity prediction model;

[0015] Using the material type, inventory status, holding status, and machine identifier from the historical material handling records as input features, and the label value as the prediction target, the initial activity prediction model is trained using the mean squared error loss function to obtain the activity prediction model.

[0016] In one embodiment, determining the target storage area matching the material based on the material's activity value includes:

[0017] Query the total available amount of storage slots that are currently available in the first storage area;

[0018] Obtain the sorting position of the material in the current materials to be stored, sorted by activity value from high to low;

[0019] If the sorting position of the material is within the available total quantity, then the target storage area matched with the material is determined to be the first storage area; if the sorting position is outside the available total quantity, then the target storage area matched with the material is determined to be the second storage area.

[0020] In one embodiment, determining the target storage area matching the material based on the activity value of the material further includes:

[0021] Based on the available total quantity, a rank interval is determined as an activity buffer in the activity value ranking of the current materials to be stored;

[0022] For the material whose activity value falls into the activity buffer, the target storage area matched by the material in the previous scheduling cycle is taken as the target storage area of ​​the material in the current scheduling cycle.

[0023] The first storage area is a high-frequency turnover area, and the second storage area is a low-frequency reserve area.

[0024] In one embodiment, it further includes:

[0025] Based on the warehouse location distribution data, machine distribution data, and track distribution data of semiconductor manufacturing plants, a database of logistics node relationships is constructed.

[0026] The step of obtaining the predicted handling time from each of the candidate warehouses to the target machine includes: querying the logistics node association database to obtain the predicted handling time from each of the candidate warehouses to the target machine;

[0027] The logistics node association database includes the mapping relationship between each storage location and each machine in the semiconductor manufacturing plant. The mapping relationship includes at least the shortest handling distance, travel time, retrieval time, and queuing time. The predicted handling time is the sum of the travel time, the retrieval time, and the queuing time.

[0028] In one embodiment, constructing the logistics node association database includes:

[0029] Each of the aforementioned storage compartments and each of the aforementioned machines serves as a logistics node, and each track unit serves as a connection line linking the logistics nodes, thereby generating a logistics node network.

[0030] Based on the logistics node network, each of the storage wards is taken as the starting node and each of the machines is taken as the target node. The shortest path algorithm is used to calculate the shortest handling distance and travel time from the starting node to the target node. The starting node, the target node, the shortest handling distance and the travel time are recorded as a mapping relationship and written into the logistics node association database.

[0031] In one embodiment, constructing the logistics node association database further includes:

[0032] Based on historical material handling records, obtain the instruction dispatch timestamp, the start pickup timestamp, and the vehicle arrival timestamp for each storage location;

[0033] The time interval between the start pickup timestamp and the instruction dispatch timestamp shall be used as the pickup time of the storage location;

[0034] The time between the vehicle's arrival time at the starting point and the start time of pickup is taken as the queuing time for the storage space.

[0035] The pickup time and queuing time are written into the mapping relationship corresponding to the storage location in the logistics node association database.

[0036] Secondly, this application also provides a dynamic storage device for semiconductor materials, the device comprising:

[0037] The acquisition module is used to acquire the attribute information of the material, which includes at least the material type, inventory status, holding status, and target machine.

[0038] The activity assessment module is used to input the attribute information of the material into the activity prediction model to predict the activity value of the material at the target machine, and the activity value represents the handling priority of the material.

[0039] The matching module is used to determine the target storage area for matching the material based on the activity value of the material, and the storage location of the target storage area is used as a candidate storage location;

[0040] The transport time determination module is used to obtain the predicted transport time between each of the candidate warehouses and the target machine;

[0041] The warehouse location screening module is used to determine the target warehouse location from the candidate warehouse locations based on the predicted handling time of each candidate warehouse location.

[0042] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the dynamic storage method for semiconductor materials described in the first aspect.

[0043] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the dynamic storage method for semiconductor materials described in the first aspect.

[0044] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the dynamic storage method for semiconductor materials described in the first aspect.

[0045] The aforementioned dynamic storage method, apparatus, equipment, and medium for semiconductor materials have the following unexpected effects: By using an activity prediction model, the material's attribute information is quantified into activity values ​​characterizing handling priority. Storage areas are allocated based on these activity values, ensuring that highly active materials are preferentially matched to storage areas that facilitate rapid access, while less active materials are stored in storage areas that are farther away and have lower inflow / outflow efficiency. By obtaining the predicted handling time from each candidate storage location to the target machine, the target storage location is determined from the candidate storage locations based at least on the predicted handling time corresponding to each candidate storage location, further reducing the time consumed in a single handling operation. The allocation of storage locations is matched with the handling priority of materials, ensuring that high-priority materials have faster turnover efficiency, significantly shortening the average delivery time of highly active materials, reducing the handling volume in storage areas that are farther away and have lower inflow / outflow efficiency, and alleviating material congestion. Furthermore, the dynamic storage method for semiconductor materials in this embodiment does not require binding storage areas to each machine individually. When a new machine comes online or the storage device changes, only the factory layout data needs to be updated to continue executing the dynamic storage method for semiconductor materials, reducing the application and maintenance difficulty of the dynamic storage method. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating a method for dynamically storing semiconductor materials in one embodiment;

[0048] Figure 2 This is a schematic diagram of a process for predicting the activity value of a material in one embodiment;

[0049] Figure 3 This is a flowchart illustrating the process of determining the target storage area for material matching in one embodiment;

[0050] Figure 4 This is a flowchart illustrating the construction of a logistics node association database in one embodiment;

[0051] Figure 5 This is a flowchart illustrating the steps involved in constructing a logistics node association database in one embodiment.

[0052] Figure 6 This is a flowchart illustrating the process of updating the logistics node association database in one embodiment;

[0053] Figure 7 This is a block diagram of a dynamic storage device for semiconductor materials in one embodiment;

[0054] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0057] Currently, the industry standard for allocating adjacent storage locations requires each process machine to have a corresponding list of adjacent storage locations and a priority order based on the factory layout and material handling routes. When a new machine comes online, the location of an existing machine changes, or storage equipment is added, removed, or replanned, the storage location settings for all affected machines need to be checked and updated one by one. Any omissions or improper settings will result in material handling instructions failing to execute or unreasonable material handling routes. Related solutions fill storage areas according to a fixed priority order, without distinguishing the urgency of material storage. This leads to the high-bay temporary storage areas, which are convenient to access and close to the machines, being occupied by low-urgency materials for extended periods. Meanwhile, high-urgency materials that will soon be available at the next machine and need to be delivered quickly are stored in distant wafer storage bins with low access efficiency because there are no available spaces in the high-bay temporary storage areas. This not only prolongs the handling time of high-urgency materials but also reduces overall production efficiency.

[0058] Wafer storage silos typically have only a limited number of input / output ports, and materials must queue inside the silo to enter or exit through these ports. When a large amount of material is stored in the same wafer storage silo, materials that need to be retrieved later will be stuck inside the silo for a long time due to port queuing, preventing them from being delivered to the target equipment in a timely manner and further exacerbating handling delays.

[0059] In summary, the relevant technologies lack the ability to quantitatively assess the urgency of material handling and cannot dynamically adjust storage locations based on real-time production status and handling resources.

[0060] In one exemplary embodiment, a method for dynamic storage of semiconductor materials is provided. This method can be applied to the material control system of a semiconductor factory. The material control system can be, but is not limited to, various industrial computers, laptops, smartphones, tablets, etc. Figure 1 As shown, the dynamic storage method for semiconductor materials in this embodiment includes the following steps S101-S105.

[0061] Step S101: Obtain the attribute information of the material. The attribute information of the material includes at least the material type, inventory status, holding status, and target machine.

[0062] In this embodiment, the material control system receives a handling request for the material to be stored and parses the material's attribute information from the handling request. The material is a wafer transfer box or a wafer carried in a wafer transfer box. The material's attribute information includes at least the material type, inventory status, holding status, and target machine.

[0063] For example, the material type can be a product, engineering sample, or test sample; the inventory status can be waiting, shipped, or held; the holding status can be held or not held; and the target machine is the identifier of the process machine where the material to be stored will be loaded in the next step.

[0064] Step S102: Input the material's attribute information into the activity prediction model to predict the material's activity value at the target machine. The material's activity value represents the material's handling priority.

[0065] The material control system inputs material attribute information into the activity prediction model, which is a pre-trained deep neural network (DNN) model. The input features of the activity prediction model include the material type, inventory status, and holding status, all processed by one-hot encoding, as well as the target machine identifier mapped into a low-dimensional dense vector through an embedding layer. The output of the activity prediction model is the activity value of the material to be stored. The activity value is a numerical value, such as a value between 0 and 1, and it characterizes the urgency with which the material to be stored must be prioritized for handling in the automated material handling system.

[0066] Step S103: Based on the activity value of the material, determine the target storage area for matching the material, and select candidate storage locations within the target storage area.

[0067] In this embodiment, the material control system determines the target storage area for the materials to be stored based on their activity values. Materials with high activity values ​​are prioritized for handling due to their high urgency and are matched with high-frequency turnover storage areas. Materials with low activity values ​​are prioritized for handling due to their low urgency and are matched with high-frequency turnover, low-frequency storage areas. After determining the target storage area, the material control system queries the current occupancy status of each storage location within the target storage area and selects available storage locations as candidate locations.

[0068] Step S104: Obtain the predicted handling time from each candidate warehouse to the target machine.

[0069] The material control system queries a pre-built and real-time updated logistics node relationship database to obtain the predicted handling time from each candidate storage location to the target machine. This database stores the mapping relationships between each storage location and machine within the semiconductor manufacturing plant. Each mapping relationship includes the shortest handling distance, travel time, retrieval time, and queuing time. The predicted handling time is the sum of travel time, retrieval time, and queuing time.

[0070] Step S105: Determine the target warehouse from the candidate warehouses based at least on the predicted handling time corresponding to each candidate warehouse.

[0071] The material control system comprehensively considers the material's activity value, the current occupancy status of each candidate storage location, and the predicted handling time for each candidate location to determine the target storage location. Materials with higher activity values ​​are assigned higher priority to candidate storage locations with shorter predicted handling times and currently available status. After determining the target storage location, the material control system generates a material handling instruction, directing an overhead shuttle (OHT) to transport the material to the target storage location. In this way, by quantifying the urgency of materials into a comparable activity value through an activity prediction model, obtaining the predicted handling time from each candidate storage location to the target storage location through a logistics node relationship database, and dynamically selecting storage locations by combining activity values ​​and handling times, the system achieves intelligent and automated semiconductor material storage. This effectively solves the problems of cumbersome manual storage area maintenance, blockage of high-urgency materials, and congestion at storage device exits.

[0072] The aforementioned dynamic storage method for semiconductor materials quantifies the material's attribute information into activity values ​​that characterize handling priority through an activity prediction model. Storage areas are allocated based on these activity values, ensuring that highly active materials are preferentially matched to storage areas that facilitate rapid access, while less active materials are stored in storage areas that are farther away and have lower inflow / outflow efficiency. By obtaining the predicted handling time from each candidate storage location to the target machine, the target storage location is determined from the candidate storage locations based at least on the predicted handling time corresponding to each candidate storage location, further reducing the time spent on a single handling operation. This achieves matching of storage location allocation with material handling priority, ensuring that high-priority materials have faster turnover efficiency, significantly shortening the average delivery time for highly active materials, reducing the handling volume in storage areas that are farther away and have lower inflow / outflow efficiency, and alleviating material congestion. Furthermore, this embodiment of the dynamic storage method for semiconductor materials does not require binding storage areas to each machine individually. When a new machine comes online or the storage device changes, only the factory layout data needs to be updated to continue executing the dynamic storage method for semiconductor materials, reducing the application and maintenance difficulty of the dynamic storage method.

[0073] In some embodiments, step S105 determines the target storage location from the candidate storage locations based at least on the predicted handling time corresponding to each candidate storage location, including: determining the target storage location from the candidate storage locations based on the activity value of the material, the current occupancy status of each candidate storage location, and the predicted handling time corresponding to each candidate storage location; wherein, materials with higher activity values ​​are assigned to target storage locations with shorter predicted handling times and currently available status with higher priority.

[0074] The material control system acquires the activity value of the material, the current occupancy status of each candidate storage location, and the predicted handling time for each candidate storage location, and then determines the target storage location from the candidate storage locations based on this information.

[0075] The selection rule for target storage locations is that materials with higher activity values ​​are given higher priority when assigned to candidate storage locations with shorter predicted handling times and currently available status. For materials with lower activity values ​​(such as materials awaiting processing that are ranked lower), they can be assigned to candidate storage locations with slightly longer predicted handling times but currently available status, while the storage location with the shortest handling time is reserved for materials with higher activity values.

[0076] In addition, currently occupied storage slots will be removed from the candidate list and will not be allocated. If all candidate slots are unavailable, the system can expand the search scope to include other storage slots near the target machine and re-execute the selection.

[0077] In this way, the method achieves optimized allocation of storage resources on a global scale, ensuring that high-quality storage locations with short handling times prioritize materials with the highest handling priority, thus balancing the timeliness and fairness of allocation.

[0078] In some embodiments, refer to Figure 2 Step S102 inputs the material's attribute information into the activity prediction model to predict the material's activity value at the target machine, including steps S1021-S1023.

[0079] Step S1021: Obtain historical material handling records. Historical material handling records include at least the material type, inventory status, holding status of the material being handled, the machine identifier of the target machine on which the material is loaded, and a tag value representing the handling time.

[0080] Retrieve historical material handling records from the material control system database for a specific historical period (e.g., the last three months or six months) as the training dataset. Each historical material handling record includes at least: the material type (LOT_TYPE) of the material being handled, its inventory status (LOT_INV_STATE), its holding status (LOT_HOLD_STATE), the machine identifier (EQ) of the target machine on which the material is loaded, and a label value representing the handling time.

[0081] Missing and outlier values ​​were removed. For each machine, the number of handling instructions under different material types, inventory statuses, and hold status combinations was statistically analyzed daily. The changing trends of the number of handling instructions for each machine over historical periods (e.g., the past three months or six months) were analyzed to identify and remove abnormal data caused by machine malfunctions, preventative maintenance, etc. To make the model more closely resemble actual production conditions, the data for each machine under each status combination was sorted from largest to smallest by the number of handling instructions, and the top 80% was extracted as the model input data, while the bottom 20% of data with the smallest number of handling instructions were filtered out.

[0082] Then, the categorical variables are coded. Unique thermal encoding is used for the material type, inventory status, and holding status of the transported materials; label encoding is used for the machine identifier of the target machine to which the transported materials are loaded. The number of transport instructions is normalized and scaled to the [0, 1] interval.

[0083] The preprocessed data was divided into training and validation sets at a ratio of 80% and 20%, respectively. To ensure consistency in data distribution, stratified sampling was performed based on machine identifiers to ensure that the data proportions for each machine in the training and validation sets were approximately the same.

[0084] Step S1022: Construct an initial activity prediction model.

[0085] Construct a deep neural network model that includes an embedding layer, multiple hidden layers, and an output layer.

[0086] The embedding layer maps the tagged machine identifier to a low-dimensional dense vector, with the initial value of the embedding layer's output dimension set to 10. Then, the low-dimensional dense vector output by the embedding layer is merged with the material type, inventory status, and holding status processed by unique thermal encoding, and used together as the input for the subsequent hidden layer.

[0087] Next, three hidden layers are constructed sequentially: the first hidden layer has 128 neurons and uses the ReLU activation function; the second hidden layer has 64 neurons, uses the ReLU activation function with added L2 regularization; and the third hidden layer has 32 neurons, uses the ReLU activation function with added L2 regularization. A dropout layer is added after each hidden layer, with an initial dropout probability of 0.3, used to randomly discard some neuron outputs during training to prevent overfitting. Finally, the output layer has one neuron using a linear activation function to output the predicted value.

[0088] Mean squared error (MSE) is used as the loss function of the model to measure the difference between the model's predicted values ​​and the true values. The Adam optimizer is used for parameter optimization. The MSE loss function is expressed as:

[0089] .

[0090] In the formula, n is the number of samples. This represents the true value of the i-th sample (i.e., the actual number of handling instructions). Let be the model's prediction for the i-th sample. During training, the model aims to minimize the mean squared error and uses the Adam optimizer to update the network parameters. The Adam optimizer is insensitive to the initial learning rate and can adaptively adjust the learning step size of each parameter, exhibiting stable performance across various tasks. Data is loaded in batches using a DataLoader, with each batch size set to 64 to improve training efficiency. The training process is iteratively performed over multiple epochs, updating the model parameters multiple times within each epoch until the loss function converges.

[0091] A Bayesian optimization algorithm is used to search for the optimal hyperparameter combination. The search range for each hyperparameter is defined as follows: embedding layer output dimension is [10, 20, 50, 100], dropout probability is [0.2, 0.3, 0.4], training batch size is [32, 64, 128], and training iterations are [50, 100]. Within the search range, the optimal hyperparameter combination that minimizes the validation set loss is found through Bayesian optimization, and the model is retrained based on this optimal combination to obtain the initial liveness prediction model.

[0092] Step S1023: Using the material type, inventory status, holding status and machine identification in the historical material handling records as input features, the label value as the prediction target, and the mean squared error loss function, the initial activity prediction model is trained to obtain the activity prediction model.

[0093] Using material type, inventory status, holding status, and machine identifier from historical material handling records (training set) as input features, an initial activity prediction model is trained. This model predicts all combinations of each machine with various material types, inventory statuses, and holding statuses, obtaining the true value Yij for each combination. Then, for each combination, its predicted value is divided by the sum of all predicted values ​​ΣY, yielding the activity value P = Yij / ΣY. The activity value P reflects the relative urgency of material handling on the target machine. A higher P value indicates a greater need for priority handling and rapid access to the material.

[0094] Once trained, the activity prediction model can be deployed. In practical applications, when a storage decision needs to be made for a certain material, the material type, inventory status, holding status, and target machine are input into the activity prediction model. The model then outputs the corresponding activity value for subsequent storage and sorting.

[0095] In some embodiments, step S102 inputs the material attribute information into the activity prediction model to predict the activity value of the material at the target machine, and further includes step S1024: every second preset time interval, the activity prediction model is trained and updated using the newly added historical material handling records within the second preset time interval.

[0096] To ensure the activity prediction model can continuously adapt to changes in production status, the system retrieves newly added historical material handling records from the database every second preset time interval (e.g., weekly or monthly). The field format of these newly added records is consistent with the initial training data in step S1021, including the material type, inventory status, holding status, target machine identifier, and a label value representing the handling time. The activity prediction model is retrained and updated using these newly added historical material handling records from the second preset time interval.

[0097] By updating regularly, the activity prediction model can capture the evolution of handling time patterns caused by changes in machine status and product structure, thus maintaining the accuracy of activity prediction.

[0098] In some embodiments, after step S1024 trains and updates the activity prediction model, step S1025 is performed: based on the updated activity prediction model, the activity value of the current material to be stored is re-predicted, and the activity value of each material to be stored is updated.

[0099] After the model is updated, the material control system iterates through all materials waiting to be allocated storage locations, obtains the attribute information of each material (material type, inventory status, holding status, target machine), inputs it into the updated activity prediction model, re-predicts the activity value of each material, and updates the corresponding record in the material control system with the newly predicted activity value.

[0100] In this way, the activity values ​​of all materials to be stored are recalibrated based on the latest production data and model parameters, ensuring that when sorting and distributing by activity value, the activity data that best reflects the current urgency of production is used.

[0101] In some embodiments, refer to Figure 3 Step S103 determines the target storage area for the material based on the material's activity value, including steps S1031-S1033.

[0102] Step S1031: Query the total available amount of storage slots that are currently available in the first storage area.

[0103] In this embodiment, after obtaining the activity value of the material to be stored, the material control system needs to determine the type of target storage area where the material should be stored.

[0104] In some embodiments, the first storage area is a high-frequency turnover area, corresponding to the overhead temporary storage area OHB, which is convenient to access and has a short handling time; the second storage area is a low-frequency reserve area, corresponding to the wafer storage bin STK, which has a large storage capacity but relatively limited import and export.

[0105] The material control system queries the storage area status database to obtain the total number of storage locations in the first storage area that are currently available (i.e., not occupied, not under maintenance, or locked). For example, the query result shows that there are currently 70 available storage locations in the first storage area.

[0106] Step S1032: Obtain the sorting position of the material in the current materials to be stored, sorted by activity value from high to low.

[0107] The material control system acquires the activity values ​​of all materials to be stored and sorts them from highest to lowest activity values. After sorting, the system determines the position of each material in the sorting results.

[0108] Step S1033: If the sorting position of the material is within the available total quantity, then the target storage area for matching the material is determined to be the first storage area; if the sorting position is outside the available total quantity, then the target storage area for matching the material is determined to be the second storage area.

[0109] The material control system compares the material's ranking with the total available storage slots in the first storage area. If the ranking is less than or equal to the available total, it indicates that the material has a high handling priority and should receive priority access, thus being assigned to the first storage area. If the ranking is greater than the available total, it is assigned to the second storage area.

[0110] For example, if the current item's sort position is 45th, and the total available quantity in the first storage area is 70, then 45 is within the 70, and therefore this item is matched to the first storage area. If the current item's sort position is 95th, then this item is matched to the second storage area.

[0111] After determining the target storage area, the material control system queries the current occupancy status of each storage location within the target storage area, eliminates occupied locations, and identifies the currently available storage locations as candidate locations for further selection of storage locations in subsequent steps.

[0112] This embodiment achieves automatic diversion by dynamically comparing the activity value of the material with the available capacity of the first storage area, thus prioritizing the storage of highly active materials in the high-frequency turnover area and storing low-activity materials in the low-frequency storage area. This ensures that limited rapid access resources always serve the most urgent handling tasks.

[0113] In some embodiments, refer to Figure 3 Step S103 determines the target storage area for material matching based on the activity value of the material. Steps S1034-S1035 are also executed after step S1032.

[0114] Step S1034: Based on the available total quantity, determine a rank interval in the current activity value ranking of the materials to be stored as an activity buffer.

[0115] Based on the total available quantity in the first storage area, the material control system determines a range of positions as an activity buffer in the activity value ranking of the materials to be stored, using the cutting position corresponding to the total available quantity in the first storage area as a benchmark.

[0116] For example, if the current available total in the first storage area is 70, the cutting position is the 70th position. The material control system uses the 70th position as a reference and expands upwards and downwards by a preset number of positions (e.g., 3 positions each). The position range between the 68th and 72nd positions is defined as the activity buffer. Materials falling into the position range of the activity buffer have activity values ​​in the boundary area between high and low activity.

[0117] Step S1035: For materials whose activity values ​​fall into the activity buffer, the target storage area matched by the material in the previous scheduling cycle is used as the target storage area of ​​the material in the current scheduling cycle.

[0118] The material control system checks whether the current material to be allocated falls within the active buffer after sorting. If it does, it queries the target storage area record that the material was matched in the previous scheduling cycle and maintains the matching result unchanged, that is, it uses the target storage area of ​​the previous cycle as the target storage area of ​​the current cycle.

[0119] For example, a material was assigned to the low-frequency reserve area (second storage area) in the previous scheduling cycle due to its low activity value. In the current cycle, its activity value has increased slightly, and its ranking has risen from 73rd to 71st, falling right into the activity buffer zone. At this time, the system maintains the matching result of the previous cycle and still assigns the engineering piece to the low-frequency reserve area, rather than immediately switching it to the high-frequency turnover area due to the slight increase in ranking.

[0120] Conversely, if the ranking of a material is outside the active buffer (e.g., significantly increased to the 50th position, or significantly decreased to the 80th position), the target storage area is determined directly according to the rules of step S1033, without being restricted by the buffer.

[0121] Thus, by introducing an active buffer zone, this embodiment effectively avoids frequent switching of material storage areas due to slight fluctuations in production cycle time, reduces unnecessary cross-area handling, improves the stability of the storage strategy, and ensures that materials that are truly urgent or not so urgent can still have their storage areas adjusted in a timely manner.

[0122] In some embodiments, refer to Figure 4 The system also includes step S100: constructing a logistics node relationship database based on the warehouse distribution data, machine distribution data, and track distribution data of the semiconductor manufacturing plant. This database can be constructed offline during system initialization and dynamically updated in real-time as the plant layout changes or equipment is added or removed.

[0123] Step S104 obtains the predicted handling time from each candidate warehouse to the target machine, including: querying the logistics node association database to obtain the predicted handling time from each candidate warehouse to the target machine; wherein, the logistics node association database includes the mapping relationship between each storage warehouse and each machine in the semiconductor manufacturing plant, and the mapping relationship includes at least the shortest handling distance, travel time, picking time and queuing time; the predicted handling time is the sum of travel time, picking time and queuing time.

[0124] In this embodiment, the material control system needs to obtain the predicted handling time from each candidate storage location to the target machine in order to select the optimal target storage location. The material control system queries a pre-built and real-time updated logistics node relationship database to obtain the predicted handling time from each candidate storage location to the target machine. The logistics node relationship database stores the mapping relationships between each storage location and each machine within the semiconductor manufacturing plant. Each mapping relationship starts with a storage location as the starting node and ends with a machine as the target node, and includes at least the shortest handling distance, travel time, retrieval time, and queuing time.

[0125] By querying the logistics node relationship database, the estimated transportation time from each candidate warehouse to the target machine can be obtained automatically, eliminating the need for manual setting of transportation routes or estimation of transportation time, thus achieving automation and precision in the selection of storage warehouses.

[0126] In some embodiments, refer to Figure 5 Step S100 involves constructing a logistics node association database, including steps S1001-S1002.

[0127] Step S1001: Using each storage location and each machine as a logistics node, and each track unit as a connection to the logistics nodes, a logistics node network is generated.

[0128] The material control system can obtain warehouse location data, machine location data, and track location data of the semiconductor manufacturing plant from the factory's manufacturing execution system or layout management system. Alternatively, third-party clients can upload the warehouse location data, machine location data, and track location data of the semiconductor manufacturing plant.

[0129] The storage location distribution data includes the identifier of each storage location and its offset distance on the corresponding track unit. Storage locations include each temporary storage position in the overhead buffer (OHB) and each storage slot in the wafer storage unit (STK). The equipment distribution data includes the identifier of each process equipment and its offset distance on the corresponding track unit. The track distribution data includes the physical length, travel speed, and travel direction of each track unit, as well as the connection relationships between the track units.

[0130] The material control system uses each storage location and each machine as a logistics node, and each track unit as a connection to the logistics nodes to generate a logistics node network.

[0131] Step S1002: Based on the logistics node network, each storage location is taken as the starting node and each machine is taken as the target node. The shortest path algorithm is used to calculate the shortest handling distance and travel time from the starting node to the target node. The starting node, target node, shortest handling distance and travel time are written into the logistics node association database as a mapping relationship.

[0132] The material control system uses each storage location as a starting node and each machine as a target node. For each pair of starting and target nodes, a shortest path algorithm (such as Dijkstra's algorithm) is used to calculate the shortest handling distance and travel time from the starting node to the target node. The material control system records the calculation results of each pair of starting and target nodes as a mapping relationship and writes it into the logistics node association database. Each mapping relationship includes at least: starting node identifier, target node identifier, shortest handling distance, and travel time. Picking time and queuing time are obtained and added in subsequent steps based on historical handling records.

[0133] Thus, by pre-building a database of logistics node relationships, a foundation is provided for real-time querying of the predicted handling time from each candidate warehouse to the target machine.

[0134] In some embodiments, step S1002 employs a shortest path algorithm to calculate the shortest transport distance from the starting node to the target node, including steps S1002-1 to S1002-2.

[0135] Step S1002-1: Determine whether the starting node and the target node are located in the same orbital unit.

[0136] The material control system obtains the track unit identifiers of the starting node (storage bin) and the target node (machine) from the track distribution data, and determines whether the two belong to the same track unit.

[0137] Step S1002-2: If the starting node and the target node are located in the same track unit, determine the transport path type based on the travel direction from the starting node to the target node.

[0138] If the starting node and the target node are located in the same track unit, the transport path type is further determined based on the position of the starting node relative to the target node in the direction of travel within the track unit. Since shuttle tracks are usually unidirectional, the relative positions of the starting node and the target node on the same track unit determine whether the shuttle can reach the target node by traveling directly in the same direction or by circling the entire track loop.

[0139] When the transport path type is unidirectional movement, the shortest transport distance is determined based on the difference in offset distance between the starting node and the target node within the track unit. When the transport path type is detour movement, the shortest transport distance is determined by subtracting the distance traveled by the starting node along the track unit to the end point of the track unit from the total path distance output by the shortest path algorithm.

[0140] When the starting node is upstream of the target node in the direction of travel on the track unit, the transport path type is unidirectional movement. In this case, the shuttle can travel directly from the starting node to the target node in the direction of travel. The shortest transport distance is the difference between the offset distance of the starting node and the offset distance of the target node on the track unit. The travel time is the shortest transport distance divided by the travel speed of the track unit.

[0141] When the starting node is downstream of the target node in the direction of travel of the track unit, the transport path type is detour. In this case, since the track travels in one direction, the shuttle cannot travel backwards. It must start from the starting node, travel along the direction of travel to the end of the track unit, enter the subsequent track unit, detour around the entire loop, and then enter from the starting point of the track unit containing the target node, finally arriving at the target node. The total path distance is output using the shortest path algorithm. However, the distance traveled from the starting node along the direction of travel to the end of its track unit is a repetitive part in the actual transport and needs to be deducted. Therefore, the shortest transport distance is the total path distance output by the shortest path algorithm, minus the distance traveled from the starting node along the direction of travel to the end of its track unit. The travel time is calculated based on the deducted shortest transport distance and the travel speed of the corresponding track segment.

[0142] In this way, when constructing the logistics node relationship database, it can calculate the shortest handling distance and travel time that conform to the actual driving path of the handling vehicle based on two handling scenarios: same-direction movement and detour movement. This ensures that the predicted handling time obtained based on the logistics node relationship database can truly reflect the physical handling cost.

[0143] In some embodiments, refer to Figure 5 Step S100 constructs a logistics node association database. After completing steps S1001 and S1002, the following steps S1003-S1006 are also executed.

[0144] Step S1003: Based on historical material handling records, obtain the instruction dispatch timestamp, start pickup timestamp, and vehicle arrival timestamp for each storage location.

[0145] The material control system extracts the timestamp information of handling instructions related to each storage location from historical material handling records. Furthermore, for each historical handling instruction involving a storage location as the starting node, the system obtains the instruction dispatch timestamp, the start time of pickup timestamp, and the vehicle arrival timestamp.

[0146] The instruction dispatch timestamp is the moment when the material control system generates and issues the handling instruction to the handling system.

[0147] The start pickup timestamp is the moment when the handling system responds to the instruction and begins to execute the pickup action.

[0148] The vehicle arrival time stamp is the moment when the transport trolley arrives at the input / output port or pickup location of the storage compartment.

[0149] Step S1004: The time between the start pickup timestamp and the instruction dispatch timestamp is used as the pickup time for the storage location.

[0150] For each storage location, the material control system uses the time between the start pickup timestamp and the instruction dispatch timestamp as the pickup time for that storage location.

[0151] Step S1005: The time between the vehicle's arrival time at the starting point and the start time of pickup is used as the queuing time for the storage space.

[0152] For each storage location, the material control system uses the time between the vehicle's arrival at the starting point and the start of the pickup as the queuing time for that storage location. Queuing time characterizes the time it takes for the transport vehicle to respond to a pickup instruction and actually arrive at the storage location. Queuing time is mainly affected by factors such as the busyness of the storage location's input / output ports and the status of rail traffic.

[0153] Step S1006: Write the pickup time and queuing time into the mapping relationship between the logistics node association database and the storage location.

[0154] The material control system writes the picking time and queuing time into the mapping relationship between the logistics node association database and the storage location.

[0155] At this point, each mapping relationship in the logistics node association database includes complete handling cost information: shortest handling distance, travel time, pickup time, and queuing time. Among them, pickup time and queuing time are dynamic attributes based on historical handling records, making the predicted handling time closer to the actual handling scenario and providing a more accurate decision-making basis for subsequent warehouse location selection.

[0156] In some embodiments, refer to Figure 6 Step S100 constructs a logistics node association database. After step S1006, the following step S1007 is also executed: Every first preset time interval, based on the newly added historical material handling records within the first preset time interval, the new pickup time and new queuing time of each storage location are recalculated, and the new pickup time and new queuing time are updated to the mapping relationship of the corresponding storage location in the logistics node association database as the pickup time and queuing time of the storage location.

[0157] The system retrieves newly added historical material handling records from the database every preset time period (e.g., two weeks or one month). These records include the latest instruction dispatch timestamp, pickup start timestamp, and vehicle arrival timestamp for each storage location. Based on these new records, the system recalculates the new pickup time and queuing time for each storage location. The calculated new pickup and queuing times are then updated in the logistics node association database, replacing the existing pickup and queuing times in each mapping relationship related to the corresponding storage location.

[0158] In this way, the dynamic time attributes of each warehouse in the logistics node association database are always refreshed based on the actual handling data in the most recent period, which can accurately reflect the current production rhythm and logistics busyness.

[0159] The dynamic storage method for semiconductor materials in this application has the following unexpected effects: By using an activity prediction model, the attribute information of the materials is quantified into activity values ​​that characterize handling priority. Storage areas are allocated based on the activity values ​​of the materials, ensuring that highly active materials are preferentially matched to storage areas that facilitate rapid access, while less active materials are stored in storage areas that are farther away and have low access efficiency. By obtaining the predicted handling time from each candidate storage location to the target machine, the target storage location is determined from the candidate storage locations based at least on the predicted handling time corresponding to each candidate storage location, further reducing the time consumed in a single handling operation. The allocation of storage locations is matched with the handling priority of the materials, ensuring that high-priority materials have faster turnover efficiency, significantly shortening the average delivery time of highly active materials, reducing the handling volume in storage areas that are farther away and have low access efficiency, and alleviating material congestion. Furthermore, the dynamic storage method for semiconductor materials in this embodiment does not require binding storage areas to each machine individually. When a new machine comes online or the storage device changes, only the factory layout data needs to be updated to continue executing the dynamic storage method for semiconductor materials, reducing the application and maintenance difficulty of the dynamic storage method.

[0160] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0161] Based on the same inventive concept, this application also provides a dynamic storage device for semiconductor materials to implement the dynamic storage method for semiconductor materials described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the dynamic storage device for semiconductor materials provided below can be found in the limitations of the dynamic storage method for semiconductor materials described above, and will not be repeated here.

[0162] According to an exemplary embodiment, this embodiment provides a dynamic storage device for semiconductor materials, referring to... Figure 7 The device includes an acquisition module 201, an activity assessment module 202, a matching module 203, a handling time determination module 204, and a storage location screening module 205.

[0163] The acquisition module 201 is used to acquire the attribute information of the material. The attribute information of the material includes at least the material type, inventory status, holding status and target machine.

[0164] The activity assessment module 202 is used to input the material's attribute information into the activity prediction model to predict the material's activity value at the target machine. The activity value represents the material's handling priority.

[0165] The matching module 203 is used to determine the target storage area for material matching based on the activity value of the material, and the storage location of the target storage area is used as a candidate storage location.

[0166] The transport time determination module 204 is used to obtain the predicted transport time between each candidate warehouse and the target machine.

[0167] The warehouse screening module 205 is used to determine the target warehouse from the candidate warehouses based on the predicted handling time of each candidate warehouse.

[0168] In some embodiments, the activity evaluation module 202 includes a historical material acquisition unit, a construction unit, and a training unit.

[0169] The historical material acquisition unit is used to acquire historical material handling records. The historical material handling records include at least the material type, inventory status, holding status of the material being handled, the machine identifier of the target machine on which the material is loaded, and a tag value representing the handling time.

[0170] Building blocks are used to construct the initial activity prediction model.

[0171] The training unit is used to train the initial activity prediction model by taking the material type, inventory status, holding status and machine identification in the historical material handling records as input features, the label value as the prediction target, and the mean squared error loss function, so as to obtain the activity prediction model.

[0172] In some embodiments, the matching module 203 includes a query unit, a sorting unit, and a matching unit.

[0173] The first query unit is used to query the total available amount of storage slots that are currently available in the first storage area.

[0174] The sorting unit is used to obtain the sorting position of the material in the current materials to be stored, sorted by activity value from high to low.

[0175] The matching unit is used to determine the target storage area for material matching as the first storage area if the sorting position of the material is within the available total quantity; and to determine the target storage area for material matching as the second storage area if the sorting position is outside the available total quantity.

[0176] In some embodiments, the matching module 203 further includes a buffer setting unit, which is used to determine a rank interval as an activity buffer based on the available total quantity in the activity value sorting of the current material to be stored.

[0177] The matching unit is also used to use the target storage area matched by the material in the previous scheduling cycle as the target storage area of ​​the material in the current scheduling cycle for materials whose activity value falls into the activity buffer.

[0178] In some embodiments, the first storage area is a high-frequency turnover area, and the second storage area is a low-frequency reserve area.

[0179] In some embodiments, the handling time determination module 204 is used to query the logistics node association database to obtain the predicted handling time from each candidate warehouse to the target machine; wherein, the logistics node association database includes the mapping relationship between each storage warehouse and each machine in the semiconductor manufacturing plant, and the mapping relationship includes at least the shortest handling distance, travel time, picking time and queuing time; the predicted handling time is the sum of travel time, picking time and queuing time.

[0180] In some embodiments, the dynamic storage device for semiconductor materials further includes a library construction module for constructing a logistics node association library based on warehouse distribution data, machine distribution data, and track distribution data of the semiconductor manufacturing plant.

[0181] In some embodiments, the library building module includes a generation unit and a writing unit.

[0182] The network generation unit is used to generate a logistics node network, with each storage location and machine as a logistics node and each track unit as a connection connecting the logistics nodes.

[0183] The path calculation unit is used to calculate the shortest handling distance and travel time from the starting node to the target node based on the logistics node network, with each storage location as the starting node and each machine as the target node, using the shortest path algorithm. The starting node, target node, shortest handling distance and travel time are then written into the logistics node association database as a mapping relationship.

[0184] In some embodiments, the path calculation unit is configured to: determine whether the starting node and the target node are located in the same track unit; if the starting node and the target node are located in the same track unit, determine the transport path type based on the travel direction from the starting node to the target node; when the transport path type is unidirectional movement, determine the shortest transport distance based on the difference in the offset distance between the starting node and the target node in the track unit; when the transport path type is detour movement, determine the shortest transport distance by subtracting the distance traveled by the starting node along the track unit to the end point of the track unit from the total path distance output by the shortest path algorithm.

[0185] In some embodiments, the library building module further includes a timestamp acquisition unit, a first statistics unit, a second statistics unit, and a writing unit.

[0186] The timestamp acquisition unit is used to acquire the instruction dispatch timestamp, the start pickup timestamp, and the vehicle arrival timestamp for each storage location based on historical material handling records.

[0187] The first statistical unit is used to calculate the time elapsed between the start pickup timestamp and the instruction dispatch timestamp as the pickup time for the storage location.

[0188] The second statistical unit is used to calculate the time between the vehicle's arrival time at the starting point and the start time of pickup as the queuing time for the storage space.

[0189] The write unit is used to write the pickup time and queuing time into the mapping relationship between the logistics node association database and the storage location.

[0190] In some embodiments, the library building module further includes a first update unit.

[0191] The first update unit is used to recalculate the new pickup time and new queuing time of each storage location based on the historical material handling records added within the first preset time interval at each first preset time interval. The new pickup time and new queuing time are then updated to the mapping relationship of the corresponding storage location in the logistics node association database as the pickup time and queuing time of the storage location.

[0192] In some embodiments, the storage location screening module 205 is used to determine the target storage location from the candidate storage locations based on the activity value of the material, the current occupancy status of each candidate storage location, and the predicted handling time corresponding to each candidate storage location; wherein, the material with the higher activity value is assigned a higher priority to the target storage location with a shorter predicted handling time and currently in an available state.

[0193] In some embodiments, the dynamic storage device for semiconductor materials further includes a second update unit for training and updating the activity prediction model at intervals of a second preset time period using newly added historical material handling records within the second preset time period.

[0194] In some embodiments, the dynamic storage device for semiconductor materials further includes an activity refresh unit for re-predicting the activity value of the material to be stored based on an updated activity prediction model, and updating the activity value of each material to be stored.

[0195] In some embodiments, the material is a wafer transport box or a wafer carried in a wafer transport box.

[0196] Each module in the aforementioned dynamic storage device for semiconductor materials can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0197] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores dynamic storage data of semiconductor materials. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a dynamic storage method for semiconductor materials.

[0198] Those skilled in the art will understand that Figure 8 The structure shown is only a block diagram of a part of the structure related to the present application and does not constitute a limitation on the computer device on which the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0199] In some embodiments, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described dynamic storage method for semiconductor materials.

[0200] In some embodiments, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described dynamic storage method for semiconductor materials.

[0201] In some embodiments, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the above-described dynamic storage method for semiconductor materials.

[0202] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0203] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0204] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0205] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A dynamic storage method of semiconductor materials, characterized by, The method includes: Obtain the attribute information of the material, which includes at least the material type, inventory status, holding status, and target machine. The material's attribute information is input into the activity prediction model to predict the material's activity value at the target machine. The material's activity value represents the material's handling priority. Based on the activity value of the material, determine the target storage area that matches the material, and select candidate storage locations within the target storage area; Obtain the predicted handling time from each of the candidate warehouses to the target machine; The target warehouse is determined from the candidate warehouses based at least on the predicted handling time corresponding to each candidate warehouse.

2. The dynamic storage method of semiconductor materials according to claim 1, wherein, The step of inputting the property information of the material into the activity prediction model to predict the activity value of the material on the target machine includes: Obtain historical material handling records, which include at least the material type, inventory status, holding status of the material being handled, the machine identifier of the target machine on which the material being handled is loaded, and a tag value representing the handling time. Construct an initial activity prediction model; Using the material type, inventory status, holding status, and machine identifier from the historical material handling records as input features, and the label value as the prediction target, the initial activity prediction model is trained using the mean squared error loss function to obtain the activity prediction model.

3. The dynamic storage method for semiconductor materials according to claim 1, characterized in that, The step of determining the target storage area matching the material based on the activity value of the material includes: Query the total available amount of storage slots that are currently available in the first storage area; Obtain the sorting position of the material in the current materials to be stored, sorted by activity value from high to low; If the sorting position of the material is within the available total quantity, then the target storage area matched with the material is determined to be the first storage area; if the sorting position is outside the available total quantity, then the target storage area matched with the material is determined to be the second storage area.

4. The dynamic storage method for semiconductor materials according to claim 3, characterized in that, The step of determining the target storage area matching the material based on the activity value of the material further includes: Based on the available total quantity, a rank interval is determined as an activity buffer in the activity value ranking of the current materials to be stored; For the material whose activity value falls into the activity buffer, the target storage area matched by the material in the previous scheduling cycle is taken as the target storage area of ​​the material in the current scheduling cycle. The first storage area is a high-frequency turnover area, and the second storage area is a low-frequency reserve area.

5. The dynamic storage method for semiconductor materials according to claim 1, characterized in that, Also includes: Based on the warehouse location distribution data, machine distribution data, and track distribution data of semiconductor manufacturing plants, a database of logistics node relationships is constructed. The step of obtaining the predicted handling time from each of the candidate warehouses to the target machine includes: querying the logistics node association database to obtain the predicted handling time from each of the candidate warehouses to the target machine; The logistics node association database includes the mapping relationship between each storage location and each machine in the semiconductor manufacturing plant. The mapping relationship includes at least the shortest handling distance, travel time, retrieval time, and queuing time. The predicted handling time is the sum of the travel time, the retrieval time, and the queuing time.

6. The dynamic storage method for semiconductor materials according to claim 5, characterized in that, The construction of the logistics node association database includes: Each of the aforementioned storage compartments and each of the aforementioned machines serves as a logistics node, and each track unit serves as a connection line linking the logistics nodes, thereby generating a logistics node network. Based on the logistics node network, each of the storage wards is taken as the starting node and each of the machines is taken as the target node. The shortest path algorithm is used to calculate the shortest handling distance and travel time from the starting node to the target node. The starting node, the target node, the shortest handling distance and the travel time are recorded as a mapping relationship and written into the logistics node association database.

7. The dynamic storage method for semiconductor materials according to claim 6, characterized in that, The construction of the logistics node association database also includes: Based on historical material handling records, obtain the instruction dispatch timestamp, the start pickup timestamp, and the vehicle arrival timestamp for each storage location; The time interval between the start pickup timestamp and the instruction dispatch timestamp shall be used as the pickup time of the storage location; The time between the vehicle's arrival time at the starting point and the start time of pickup is taken as the queuing time for the storage space. The pickup time and queuing time are written into the mapping relationship corresponding to the storage location in the logistics node association database.

8. A dynamic storage device for semiconductor materials, characterized in that, The device includes: The acquisition module is used to acquire the attribute information of the material, which includes at least the material type, inventory status, holding status, and target machine. The activity assessment module is used to input the attribute information of the material into the activity prediction model to predict the activity value of the material at the target machine, and the activity value represents the handling priority of the material. The matching module is used to determine the target storage area for matching the material based on the activity value of the material, and the storage location of the target storage area is used as a candidate storage location; The transport time determination module is used to obtain the predicted transport time between each of the candidate warehouses and the target machine; The warehouse location screening module is used to determine the target warehouse location from the candidate warehouse locations based on the predicted handling time of each candidate warehouse location.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the dynamic storage method for semiconductor materials according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the dynamic storage method for semiconductor materials according to any one of claims 1 to 7.

Images