A control method and system for mold machining

Abstract

The application discloses a control method and system for mold processing, comprising: inputting a modeling file of a mold to be processed into a position category identification model and a time length prediction model after processing, obtaining position category data and processing time length data of the mold to be processed; and assigning a processing device to the mold to be processed according to the position category data, the processing time length data, a priority of the mold to be processed and operation information of the processing device. The position category of the mold to be processed and the processing time length are automatically identified and predicted by using the position category identification model and the time length prediction model, so that the consumption of time and manpower in processing scheduling of the mold is reduced; and the processing device is assigned to the mold to be processed according to the position category data, the processing time length data, the priority of the mold to be processed and the operation information of the processing device, so that the mold processing can be conveniently and quickly scheduled.

Description

Technical Field

[0001] This application relates to the field of smart factories, and more particularly to a control method and system for mold processing. Background Technology

[0002] For processing orders, factories typically rely on experienced technicians to deduce the multiple processing positions for each mold to be processed in the order, the processing type at each position, and the processing time for each position. Only then can the factory randomly schedule the processing of parts across different machines based on these deduced processing types and times.

[0003] Because the determination of processing location and type relies heavily on skilled workers, they cannot focus on specific processing tasks. In workshop scheduling, workers need to estimate processing time and obtain machine downtime to manually determine the machine for processing a particular part. When there are priority processing types, temporary machine malfunctions, or parts needing to be processed earlier or later, workers must reschedule processing. When there are too many orders and mold processing needs to be rescheduled, the computational workload for rescheduling becomes extremely large, consuming significant time and manpower.

[0004] In summary, there is a need for a control method and system for mold processing that can reduce the time and manpower consumption of mold processing scheduling and can conveniently and quickly schedule mold processing. Summary of the Invention

[0005] To address the above problems, this application proposes a control method and system for mold processing.

[0006] On the one hand, this application proposes a control method for mold processing, including:

[0007] After processing the modeling file of the mold to be processed, it is input into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed.

[0008] Based on the processing location category data, the processing time data, the priority of the mold to be processed, and the operating information of the processing equipment, one or more processing equipment are assigned to the mold to be processed.

[0009] Preferably, the step of processing the modeling file of the mold to be processed and inputting it into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed includes:

[0010] Multiple two-dimensional views are generated based on the modeling file of the mold to be processed, wherein the modeling file is a three-dimensional modeling file;

[0011] The topological features of the mold to be processed are determined based on the 3D modeling file;

[0012] Multiple two-dimensional views and the topological features are input into the position category recognition model to obtain the position category data of the mold to be processed. The position category data includes the position category data of selective processing technology and the position category data of non-selective processing technology.

[0013] The 3D modeling file of the mold to be processed is converted into a 3D matrix;

[0014] The three-dimensional matrix and the topological features are input into the time prediction model to obtain the processing time data of the mold to be processed.

[0015] Preferably, the step of allocating processing equipment to the mold to be processed based on the type of the location to be processed, the processing time data, the priority of the mold to be processed, and the operating information of the processing equipment includes:

[0016] Obtain operational information from the processing equipment to identify equipment with and without abnormalities.

[0017] Based on the priority information of the abnormal processing equipment and the mold to be processed, determine whether there are any unrecorded abnormal processing equipment and whether the mold to be processed has a priority.

[0018] If there are any unrecorded abnormal processing devices and the mold to be processed has no priority, then the current time point is taken as the abnormal time point; determine all processing tasks that have not started processing on the processing devices at the abnormal time point; obtain the processing position category data and processing time data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; using a greedy algorithm, assign the processing devices without abnormalities to the processing tasks and the mold to be processed based on the processing devices without abnormalities, the processing position category data and processing time data of the processing tasks, and the processing position category data and processing time data of the mold to be processed;

[0019] If there are no unrecorded abnormal processing devices and the mold to be processed has a priority, then the current time is taken as the abnormal time point; identify all processing tasks that have not started processing on any of the non-abnormal processing devices at the abnormal time point; obtain the processing position category data and processing duration data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; determine the processing order based on the processing position category data of the mold to be processed; assign non-abnormal processing devices to the mold to be processed based on the processing order, the running information, and the processing duration data; use a greedy algorithm to assign processing devices to the processing tasks based on the non-abnormal processing devices and the processing position category data and processing duration data of the processing tasks.

[0020] If there are any unrecorded abnormal processing devices and the mold to be processed has a priority, then the current time is taken as the abnormal time point; identify all processing tasks that have not started processing on the processing devices at the abnormal time point; obtain the processing position category data and processing duration data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; determine the processing order based on the processing position category data of the mold to be processed; assign the abnormal-free processing device to the mold to be processed based on the processing order, the running information, and the processing duration data; use a greedy algorithm to assign the abnormal-free processing device to the processing task based on the abnormal-free processing device and the processing position category data and processing duration data of the processing task.

[0021] Preferably, after determining whether there are any unrecorded abnormal processing devices and whether the mold to be processed has a priority based on the priority information of the abnormal processing equipment and the mold to be processed, the method further includes:

[0022] If there are no unrecorded abnormal processing devices and the mold to be processed has no priority, then the processing sequence is determined according to the processing position category data of the mold to be processed, and the abnormal processing device is assigned to the mold to be processed according to the processing sequence, the running information and the processing time data.

[0023] Preferably, before the modeling file of the mold to be processed is processed and input into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed, the following steps are included:

[0024] Generate location category datasets and duration datasets using historical orders;

[0025] The topology feature extraction model is trained using the modeling files in the historical orders to obtain the trained topology feature extraction model.

[0026] The trained topological feature extraction model is used to extract topological features from the modeling file of the mold to be processed;

[0027] The location category recognition model is trained using the location category dataset and the topological features to obtain the trained location category recognition model.

[0028] The duration prediction model is trained using the duration dataset and the topological features to obtain the trained duration prediction model.

[0029] Preferably, the step of generating location category datasets and duration datasets using historical orders includes:

[0030] Obtain the mold modeling file, processing location identifier, processing type of processing location, and processing time corresponding to the processing type of mold from historical orders. The modeling file is a three-dimensional modeling file.

[0031] Generate multiple 2D views for each of the 3D modeling files;

[0032] Based on the processing location identifier and the processing type of the processing location, mark the processing location for all two-dimensional views and record the processing type corresponding to the processing location to obtain a location category dataset;

[0033] Based on the processing time corresponding to the processing type of the mold, record the required processing types and the total processing time for each processing type for each 3D modeling file to obtain a time dataset.

[0034] Preferably, training the location category recognition model using the location category dataset and the topological features to obtain the trained location category recognition model includes:

[0035] Each of the two-dimensional views is converted into a two-dimensional matrix, resulting in multiple two-dimensional matrices;

[0036] Based on the location category dataset, each processing location in each two-dimensional matrix is ​​normalized to obtain multiple normalized location data, wherein each normalized location data includes its corresponding processing type;

[0037] The normalized location data and the topological features are used to train an artificial intelligence recognition model to obtain the trained location category recognition model.

[0038] Preferably, the step of training the duration prediction model using the duration dataset and the topological features to obtain the trained duration prediction model includes:

[0039] Convert the 3D modeling file of the duration dataset into a 3D matrix;

[0040] The total processing time is normalized to obtain the normalized total processing time data for each processing type;

[0041] The deep convolutional neural network is trained using the topological features, the three-dimensional matrix, and the normalized total processing time data to obtain the trained time prediction model.

[0042] Secondly, this application proposes a control system for mold processing, including: a mold identification module and an equipment allocation module;

[0043] The mold recognition module is used to process the modeling file of the mold to be processed and input it into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed.

[0044] The equipment allocation module is used to allocate processing equipment to the mold to be processed based on the processing location category data, the processing time data, the priority of the mold to be processed, and the operation information of the processing equipment.

[0045] Preferably, the location category recognition model includes:

[0046] The identification model unit is used to extract mapping features from multiple two-dimensional views generated from the modeling file of the mold to be processed; send the mapping features to the feature fusion unit; perform target detection on the fused features to obtain the features to be identified; send the features to be identified to the multilayer feedforward neural network unit; determine the processing position category data of the non-selective processing technology of the mold to be processed based on the features to be identified, and send it to the data output unit.

[0047] The feature fusion unit is used to extract and fuse features from the mapping features and the topological features of the mold to be processed, to obtain the fused features of the mold to be processed, and send them to the recognition model unit.

[0048] A multi-layer feedforward neural network unit is used to determine the processing position category data of the selective processing technology of the mold to be processed based on the features to be identified, and send it to the data output unit;

[0049] The data output unit determines the processing position category data of the mold to be processed based on the processing position category data of the non-selective processing technology and the processing position category data of the selective processing technology.

[0050] The advantages of this application are: by using a location category recognition model and a duration prediction model to automatically identify and predict the processing location category and processing duration of the mold to be processed, the time and manpower consumption of mold processing scheduling can be reduced; based on the processing location category data, processing duration data, priority of the mold to be processed and the operation information of the processing equipment, one or more processing equipment can be allocated to the mold to be processed, which can facilitate and quickly schedule mold processing. Attached Figure Description

[0051] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0052] Figure 1 This is a schematic diagram of the steps of a control method for mold processing provided in this application;

[0053] Figure 2 This is a schematic diagram illustrating the steps of changing the priority or operating information of a control method for mold processing provided in this application;

[0054] Figure 3 This is a schematic diagram of the processing position category data of a control method for mold processing provided in this application;

[0055] Figure 4 This is a schematic diagram of the data set generation process of a control method for mold processing provided in this application;

[0056] Figure 5 This is a schematic flowchart illustrating the identification of the mold to be processed in a control method for mold processing provided in this application;

[0057] Figure 6 This is a flowchart illustrating the process of acquiring abnormal information in a control method for mold processing provided in this application;

[0058] Figure 7 This is a schematic diagram of the processing equipment allocation in a control method for mold processing provided in this application;

[0059] Figure 8 This is a flowchart illustrating the training position category recognition model for a control method used in mold processing provided in this application;

[0060] Figure 9 This is a flowchart illustrating a training time prediction model for a control method used in mold processing, as provided in this application.

[0061] Figure 10 This is a schematic diagram of a control system for mold processing provided in this application;

[0062] Figure 11 This is a schematic diagram of a mold identification module in a control system for mold processing provided in this application;

[0063] Figure 12 This is a schematic diagram of one embodiment of a control system for mold processing provided in this application;

[0064] Figure 13 This is a schematic diagram of another embodiment of a control system for mold processing provided in this application. Detailed Implementation

[0065] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0066] According to an embodiment of this application, a control method for mold processing is proposed, such as... Figure 1 As shown, it includes:

[0067] S101, after processing the modeling file of the mold to be processed, input it into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed;

[0068] S102, based on the data of the location to be processed, the processing time data, the priority of the mold to be processed, and the operating information of the processing equipment, assign one or more processing equipment to the mold to be processed.

[0069] like Figure 2 As shown, after allocating processing equipment to the mold to be processed based on the processing location category data, processing time data, priority of the mold to be processed, and operation information of the processing equipment, the process further includes: S103, saving the processing location category data, processing time data, and operation information of the mold to be processed.

[0070] like Figure 2 As shown, after saving the processing position category data, processing time data and operation information of the mold to be processed, the method further includes: S104, if the priority of the mold to be processed and / or the operation information of the processing equipment changes, then S102 is executed, and one or more processing equipment are assigned to the mold to be processed according to the processing position category data, processing time data, priority of the mold to be processed and operation information of the processing equipment.

[0071] The operational information of the processing equipment includes: the operating status of the processing equipment, the tasks awaiting processing, and the tasks currently being processed (the processing tasks being executed). The operating status of the processing equipment includes: running, not running, and abnormal operation.

[0072] The modeling file of the mold to be processed is processed and then input into the position category recognition model and the time prediction model to obtain the processing position category data and processing time data of the mold to be processed. This includes: generating multiple two-dimensional views based on the modeling file of the mold to be processed, and the modeling file is a three-dimensional modeling file; determining the topological features of the mold to be processed based on the three-dimensional modeling file; inputting the multiple two-dimensional views and topological features into the position category recognition model to obtain the processing position category data of the mold to be processed, which includes processing position category data for selective processing processes and processing position category data for non-selective processing processes; converting the three-dimensional modeling file of the mold to be processed into a three-dimensional matrix; and inputting the three-dimensional matrix and topological features into the time prediction model to obtain the processing time data of the mold to be processed.

[0073] The processing time data for the mold to be processed includes all the processing types that need to be processed, as well as the total time required to complete each processing type. The processing position type data for the mold to be processed includes six two-dimensional views of the mold to be processed, with the processing position and processing type marked.

[0074] like Figure 3 The diagram shows the data for the processing location categories. The boxed areas represent the processing locations of the mold, and boxes with the same letter indicate those using the same processing category. Figure 3 The four boxes labeled A indicate locations that use the same processing method, while the box labeled B indicates a location that uses a different processing method than A. Specifically, different colored boxes can be used to represent different processing categories (processing methods).

[0075] Based on the data of the location category to be processed, processing time data, priority of the mold to be processed, and operating information of the processing equipment, processing equipment is assigned to the mold to be processed. This includes: obtaining the operating information of the processing equipment to determine which processing equipment is normal and which is abnormal; based on the priority information of the abnormal processing equipment and the mold to be processed, determining whether there are any unrecorded abnormal processing equipment and whether the mold to be processed has a priority; if there are unrecorded abnormal processing equipment and the mold to be processed has no priority, then the current time point is taken as the abnormal time point; identifying all processing tasks that have not started processing on all processing equipment at the abnormal time point; obtaining the processing equipment to be processed... The process involves: identifying the task's pending processing location category and processing time; clearing all pending processing tasks assigned to all processing equipment that did not begin processing at the abnormal time point; using a greedy algorithm, assigning processing equipment without abnormalities to the pending tasks and molds based on the data of processing equipment without abnormalities, the pending processing location category and processing time of the pending tasks, and the pending processing location category and processing time of the molds; if there are no unrecorded processing equipment with abnormalities and the molds have priority, then the current time is taken as the abnormal time point; identifying all pending processing tasks that did not begin processing at the abnormal time point using processing equipment without abnormalities. Obtain the processing location category data and processing time data for the tasks to be processed; clear all tasks assigned to all processing equipment that did not start processing at the abnormal time point; determine the processing sequence based on the processing location category data of the mold to be processed, and assign processing equipment without abnormalities to the mold to be processed based on the processing sequence, running information, and processing time data; use a greedy algorithm to assign processing equipment to the tasks to be processed based on the processing equipment without abnormalities, and the processing location category data and processing time data of the tasks to be processed; if there is an unrecorded processing equipment with abnormalities and the mold to be processed has a priority, then the current time is used as the abnormal time. At normal time points; identify all processing tasks that have not started processing on all processing equipment at abnormal time points; obtain the processing location category data and processing time data of the processing tasks; clear all processing tasks assigned to all processing equipment that have not started processing at abnormal time points; determine the processing sequence based on the processing location category data of the mold to be processed, and assign processing equipment without abnormalities to the mold to be processed based on the processing sequence, running information, and processing time data; use a greedy algorithm to assign processing equipment without abnormalities to the processing tasks based on the processing equipment without abnormalities and the processing location category data and processing time data of the processing tasks.

[0076] Because the mold currently undergoing processing (the mold in processing) may require multiple types of processing, necessitating the sequential use of multiple processing devices, and there may be no available processing devices before proceeding to the next type of processing, the mold in processing may need to wait for a period of time after completing the previous type of processing before starting the next. Therefore, not only do the molds already assigned processing devices have one or more pending processing tasks, but the mold in processing may also have one or more pending processing tasks. Furthermore, the pending processing tasks correspond to priority information for both the mold in processing and the molds already assigned processing devices. Therefore, pending processing tasks include: all pending processing tasks for the molds already assigned processing devices and all pending processing tasks for the molds in processing. Priority information includes: priority information for all pending processing tasks and priority information for the molds not yet assigned processing devices. Since there may be more currently functioning processing devices capable of performing processing tasks without abnormalities than the previously saved number of such devices (i.e., newly added processing devices capable of performing processing tasks normally, such as those that have completed maintenance, repair, or overhaul), the processing devices used to assign pending processing tasks and molds are those that were functioning normally at the time of the abnormal processing point.

[0077] After determining whether there are any unrecorded abnormal processing devices and whether the mold to be processed has a priority based on the priority information of abnormal processing equipment and mold to be processed, the process further includes: if there are no unrecorded abnormal processing devices and the mold to be processed has no priority, then the processing sequence is determined based on the processing position category data of the mold to be processed, and abnormal processing equipment is assigned to the mold to be processed based on the processing sequence, operation information and processing time data.

[0078] Preferably, the priority information for the molds to be processed includes: mold priority (mold priority within the order), order priority, and pre-processing priority. If the priority information for the mold to be processed is pre-processing priority or order priority, then the mold to be processed has priority information. If the priority for the mold to be processed is mold priority, then the mold to be processed has no priority information. However, when the same order includes multiple molds to be processed, the molds with mold priority will be assigned to processing equipment first, compared to those without mold priority. Order priority means that the entire order has a higher priority than orders without priority; that is, all molds to be processed within the entire order have the same priority, and this priority makes the priority of all molds to be processed in the entire order higher than that of molds to be processed in orders without priority or in the molds being processed.

[0079] Common methods for machining molds include turning, milling, electrical discharge machining (EDM), planing, grinding, and electro-hydraulic (EDM) machining. These processes have a specific sequence, and the mold to be machined can complete all the necessary machining operations on a single machine tool. For example, suppose a mold requires turning and EDM machining on six faces, with the machining sequence being turning first, followed by EDM. In this case, the mold needs to first complete all the turning operations on all six faces on a machine tool capable of turning, and then complete all the EDM machining operations on the remaining six faces on a machine tool capable of EDM.

[0080] For cases where there are no unrecorded abnormal processing equipment and the mold to be processed has no priority, the specific processing equipment allocation method is as follows: First, based on the processing position category data of the mold to be processed, determine the processing sequence. Based on the processing sequence, select the first processing equipment for the category that the mold needs to process first (first processing category). Among multiple processing equipment capable of performing the same type of processing, select the first usable equipment as the first processing equipment. After assigning the first processing equipment to the mold to be processed, determine the completion time for the mold to complete the first processing category on the first processing equipment based on the processing time data. Based on the completion time, using the same processing equipment selection method, select the second processing equipment for the next category to be processed (second processing category) of the mold to be processed, and determine the completion time again. This process continues until all processing equipment corresponding to the required processing categories of the mold to be processed is assigned, thus completing the scheduling of processing equipment for this mold.

[0081] Before processing the modeling file of the mold to be processed and inputting it into the location category recognition model and the duration prediction model to obtain the processing location category data and processing duration data of the mold to be processed, the process includes: generating a location category dataset and a duration dataset using historical orders; training a topology feature extraction model using the modeling file from historical orders to obtain a trained topology feature extraction model; extracting topology features from the modeling file of the mold to be processed using the trained topology feature extraction model; training a location category recognition model using the location category dataset and topology features to obtain a trained location category recognition model; and training a duration prediction model using the duration dataset and topology features to obtain a trained duration prediction model.

[0082] The topology feature extraction model is a model trained using an existing 3D model and based on the SDM-NET model to extract geometric topology features from the modeling file (3D modeling file) of the mold to be processed.

[0083] like Figure 4As shown, a location category dataset and a duration dataset are generated using historical orders. This includes: obtaining the mold modeling file, processing location identifier, processing type of the processing location, and processing duration corresponding to the processing type of the mold from historical orders; the modeling file is a 3D modeling file; generating multiple 2D views for each 3D modeling file; marking the processing location for all 2D views according to the processing location identifier and processing type of the processing location, and recording the processing type corresponding to the processing location to obtain the location category dataset; and recording the required processing types and the total processing duration for each processing type for each 3D modeling file according to the processing duration corresponding to the processing type of the mold to obtain the duration dataset.

[0084] To generate a location category dataset, historical orders can be used. Specifically, PyVista can be used to generate six 2D views from the saved 3D modeling file. Labelme can then be used to mark the processing locations in the six 2D views and record the processing type of each location, resulting in the location category dataset. The processing types corresponding to each 3D modeling file and the total processing time for each processing type can be saved in a CSV file.

[0085] The location category recognition model is trained using a location category dataset and topological features. The process includes: converting each two-dimensional view into a two-dimensional matrix, resulting in multiple two-dimensional matrices; normalizing each processing position in each two-dimensional matrix based on the location category dataset, resulting in multiple normalized location data sets, where each normalized location data set includes its corresponding processing type; and training an artificial intelligence recognition model using the normalized location data and topological features. The training of the location category recognition model is then described.

[0086] The position category recognition model preferably includes a YOLOv7 object detection model. The bottleneck network in the position category recognition model can extract mapping features from multiple two-dimensional views. After concatenating the mapping features and topological features, convolution and pooling are performed to scale the increased-dimensional features after concatenation to the required input feature dimension suitable for the remaining YOLO source models, resulting in fused features. These fused features are input to the Neck side of the YOLO model for object detection, yielding features to be recognized. These features are then input to the Head side (detection head) and two multi-layer feedforward neural networks of the position category recognition model. The Head side determines the processing position category data for the non-selective processing technology of the mold based on the input features, while the two multi-layer feedforward neural networks determine the processing position category data for the selective processing technology of the mold based on the features. Finally, the processing position category data of the mold is determined based on the processing position category data for both the non-selective and selective processing technologies. The Head side and the two multi-layer feedforward neural networks process the input features in parallel.

[0087] Two multilayer feedforward neural networks are used: a first multilayer feedforward neural network and a second multilayer feedforward neural network. The first multilayer feedforward neural network determines whether the mold to be processed includes selective processing techniques based on the input features, and whether such selective processing techniques need to be executed. The result is represented by a one-hot vector indicating whether selective processing techniques exist and whether they need to be executed. The second multilayer feedforward neural network, when the mold to be processed includes selective processing techniques, determines the processing location and type of the selective processing techniques based on the input features. Ultimately, whether the processing location and type of the selective processing techniques need to be processed can be determined by the value of the one-hot vector. For the molds to be processed, some molds may require drilling, while others do not; "drilling" is a selective processing technique. However, due to molding issues, some scrap materials need secondary processing; that is, all original molds in a factory must be trimmed. "Trimming" is a necessary processing technique.

[0088] The training of a duration prediction model using a duration dataset and topological features includes: converting the 3D modeling file of the duration dataset into a 3D matrix; normalizing the total processing time to obtain normalized total processing time data for each processing type; and training a deep convolutional neural network using topological features, the 3D matrix, and the normalized total processing time data to obtain the trained duration prediction model.

[0089] The embodiments of this application will be further described below.

[0090] like Figure 5 As shown, firstly, the 3D modeling file of the mold to be processed in the order is obtained. Six corresponding 2D views are generated using this 3D modeling file, i.e., the six views of the mold to be processed. Preferably, PyVista can be used to generate the six 2D views of the 3D modeling file. The mold in this 3D modeling file is converted into a 3D matrix. The 3D matrix and topological features are input into the duration prediction model to obtain the processing duration data of the mold to be processed. The six 2D views are input into the location category recognition model to obtain the processing location category data of the mold to be processed. Since the method of this application uses a multi-threaded approach, not only can the steps of generating the six 2D views and converting the 3D matrix be performed simultaneously, but also the steps of inputting the 3D matrix into the duration prediction model and inputting the six 2D views into the location category recognition model can be performed simultaneously. After obtaining the processing location category data and processing duration data, these data are used for scheduling and storage, respectively.

[0091] like Figure 6 As shown, scheduling requires consideration of the processing location category data, processing time data, priority of the mold to be processed, and the operating information of the processing equipment. First, it's necessary to obtain the operating information of all processing equipment to determine which equipment is operating normally and capable of performing processing tasks, and which may be undergoing maintenance, inspection, repair, or awaiting repair. Therefore, although some equipment may be operating normally, they cannot currently perform processing tasks. Equipment currently capable of performing processing tasks is designated as "no-abnormal" equipment, and equipment currently unable to perform processing tasks is designated as the first "abnormal" equipment. The operating information of the processing equipment can be obtained from the equipment itself or manually entered. The operating information of all previously saved processing equipment is retrieved, and equipment recorded as unable to perform processing tasks is designated as the second "abnormal" equipment. By comparing the first and second "abnormal" equipment, it's determined whether there are any unrecorded abnormal processing equipment. Second, the priority information of the mold to be processed needs to be determined. The priority of the mold to be processed can be set before the modeling file of the mold is input into the position category recognition model and the duration prediction model, or it can be manually modified after the processing equipment has been assigned to the mold. After determining the operating information of all current processing equipment and the priority of the mold to be processed, it can be used together with the position category data and processing duration data as judgment conditions to assign processing equipment to the mold.

[0092] like Figure 7 As shown, processing equipment is assigned to the mold to be processed based on whether there are any unrecorded abnormal processing equipment and whether the mold to be processed has priority information.

[0093] If there are unrecorded abnormal processing devices, but the mold to be processed has no priority, then the current time point is taken as the abnormal time point. All processing tasks assigned to all processing devices that did not start processing at this abnormal time point are identified, and all processing tasks that started after the abnormal time point are cleared. The processing location category data, processing time data, and priority information of all processing tasks are obtained. Using a greedy algorithm, based on all currently normal processing devices, the priority information of all processing tasks, the priority information of the mold to be processed, the processing location category data and processing time data of the processing tasks, and the processing location category data and processing time data of the mold to be processed, the processing devices that were not abnormal at the abnormal time point are assigned to the processing tasks and the molds to be processed.

[0094] If there are no unrecorded abnormal processing devices, but the mold to be processed has a priority, then the current time is taken as the abnormal time point. All processing tasks assigned to non-abnormal processing devices that did not start processing at this abnormal time point are identified, and all processing tasks that started after the abnormal time point on these non-abnormal processing devices are cleared. Based on the order of the processing categories of the molds to be processed, non-abnormal processing devices are assigned to the molds. The processing location category data, processing time data, and priority information of all processing tasks are obtained. Using a greedy algorithm, based on all non-abnormal processing devices, the priority information of all processing tasks, the priority information of the molds to be processed, the processing location category data and processing time data of the processing tasks, and the processing location category data and processing time data of the molds to be processed, non-abnormal processing devices that were not abnormal at the abnormal time point are assigned to the processing tasks and molds.

[0095] If any unrecorded abnormal processing equipment exists, and the mold to be processed has a priority, then the current time is taken as the abnormal time point. All processing tasks assigned to all processing equipment that did not start processing at this abnormal time point are identified, and all processing tasks that started after the abnormal time point are cleared. Based on the order of the processing categories of the molds to be processed, abnormal-free processing equipment is assigned to each mold. The processing location category data, processing time data, and priority information of all processing tasks are obtained. Using a greedy algorithm, based on all currently abnormal-free processing equipment, the priority information of all processing tasks, the priority information of the molds to be processed, the processing location category data and processing time data of the processing tasks, and the processing location category data and processing time data of the molds to be processed, abnormal-free processing equipment that was not abnormal at the abnormal time point is assigned to the processing tasks and molds.

[0096] If there are no unrecorded abnormal processing equipment and the mold to be processed has no priority, then the processing sequence is determined based on the processing position category data of the mold to be processed, and the abnormal processing equipment is assigned to the mold to be processed based on the processing sequence, operation information and processing time data.

[0097] When it is necessary to allocate processing equipment to multiple molds to be processed at the same time, preferably, the priority of allocating processing equipment to the molds to be processed is: processing sequence priority > order priority > mold priority.

[0098] After completing the allocation of processing equipment for the mold to be processed, save the processing location category data, processing time data, and processing equipment operation information of the mold to be processed.

[0099] Below, taking the example of a mold to be processed that needs to undergo milling and planing, we will further explain the allocation of processing equipment for the mold to be processed in the case that there are no unrecorded abnormal processing equipment and the mold to be processed has no priority.

[0100] After processing the modeling file of the mold to be processed, it was input into the position category recognition model and the time prediction model respectively to obtain the position category data and processing time data of the mold to be processed, as shown in Table 1. The processing sequence of the mold to be processed is milling first, followed by planing, and the total time corresponding to each type of processing is 3 minutes and 51 seconds and 2 minutes and 16 seconds respectively.

[0101] A (Electrical Spark) B (Milling) C (planing) D (turning) - 03:51 02:16 -

[0102] Table 1

[0103] First, based on the processing sequence, milling of the mold to be processed is designated as the first processing category, and planing as the second processing category. Second, based on the current time, all processing equipment capable of milling is identified from among all non-abnormal processing equipment as candidate equipment for the first processing type. Among the candidate equipment for the first processing type, the equipment that can execute the first processing category fastest is selected as the first processing equipment. That is, assuming the current time for obtaining the processing position category data and processing time data of the mold to be processed is 9:33 AM, if there are only two milling processing equipment, X1 and X2, among the candidate equipment for the first processing type, and their completion times for the assigned processing tasks are X1 at 10:58 AM and X2 at 11:18 AM respectively, then X1 is selected as the first processing equipment for the first processing category of the mold to be processed. The total milling processing time of 3 minutes and 51 seconds is added to X1's task completion time, and the task completion time of the first processing equipment is updated, resulting in a first completion time of 11:02 AM. Subsequently, based on the first completion time, all processing equipment capable of planing is identified from all non-abnormal processing equipment as candidate equipment for the second processing type. If there are only three planing processing equipment (B1, B2, and B3) among the candidate equipment for the second processing type, and their completion times for the assigned processing tasks are B1 at 10:40 AM, B2 at 11:10 AM, and B3 at 1:12 PM respectively, then B2 is selected as the second processing equipment for the second processing category of the mold to be processed. The total planing processing time of 2 minutes and 16 seconds is added to B2's task completion time, and the task completion time of the second processing equipment is updated, resulting in a second completion time of 11:13 AM, thus completing the assignment of the processing task for the mold to be processed.

[0104] All identical molds of the same type corresponding to each modeling file are processed using the same assigned processing equipment. That is, if there are 10 identical molds of the same type corresponding to the same modeling file that require milling followed by planing, the time in the final predicted processing time data is the total time for all 10 molds to be processed on the same processing equipment.

[0105] As shown in Table 2 below, taking an order containing two molds to be processed as an example, we will further explain the allocation of processing equipment for the molds to be processed that have mold priority, assuming there are no unrecorded abnormal processing equipment and the molds to be processed have no priority.

[0106] Turning Grinding and processing Mold A 01:40 03:00 Mold B 01:40 -

[0107] Table 2

[0108] As shown in Table 2, mold A (Mold A) requires both turning and grinding, while mold B (Mold B) only requires turning. Assuming the processing order is turning first, then grinding, and that mold A has mold priority while mold B does not (i.e., within this order, mold A has a higher priority than mold B), then processing equipment needs to be assigned first for turning of mold A, then for turning of mold B, and finally for grinding of mold A. Specifically, all currently available turning equipment is designated as the first processing type candidate equipment, and all currently available grinding equipment is designated as the second processing type candidate equipment. First, from the first processing type candidate equipment, the equipment that can execute the first processing type fastest is selected and assigned to mold A, refreshing the task completion time of mold A's first processing equipment. Second, from the first processing type candidate equipment, the equipment that can execute the first processing type fastest is selected and assigned to mold B, refreshing the task completion time of mold B's first processing equipment. Finally, based on the task completion time of the first processing equipment of mold A after the update, select the alternative equipment that can execute the second processing category the fastest from the alternative equipment of the second processing type, assign it to mold A, update the task completion time of the second processing equipment of mold A, and complete the allocation of processing equipment for mold A and mold B in this order.

[0109] Below, as Figure 8 As shown, the training of the location category recognition model of this application is further explained.

[0110] First, each two-dimensional view in the location category dataset is read and converted into a two-dimensional matrix. Since the two-dimensional matrix is ​​an RGB matrix, each two-dimensional view is converted into three two-dimensional matrices: matrix R, matrix G, and matrix B. Second, the processing locations and their corresponding processing types in the location category dataset are read and normalized to obtain multiple normalized location data. Then, the hyperparameters required for the AI ​​recognition model to be trained are adjusted, and the normalized location data and topological features are used to train the AI ​​recognition model. The hyperparameters include: number of training iterations, batch size, model input image size, and optimizer type. Preferably, the AI ​​recognition model is trained on a Graphics Processing Unit (GPU). The normalized labeled processing locations and their corresponding processing types are used as the model output to obtain the Intersection over Union (IoU). If the cross-ratio of the AI ​​recognition model to be trained is greater than the recognition threshold, then the training of the AI ​​recognition model to be trained is completed, and a trained location category recognition model is obtained. If the cross-ratio of the AI ​​recognition model to be trained is less than or equal to the recognition threshold, then the hyperparameters required for the AI ​​recognition model to be trained are adjusted, and the AI ​​recognition model to be trained continues to be trained until the cross-ratio is less than the recognition threshold, and a trained location category recognition model is obtained. Preferably, the recognition threshold is 90%. Preferably, the AI ​​recognition model to be trained uses the YOLOv7 object detection algorithm.

[0111] Below, as Figure 9 As shown, the training of the duration prediction model of this application is further explained.

[0112] First, all 3D modeling files in the time-based dataset are read, and each 3D modeling file is converted into a 3D matrix, resulting in multiple 3D matrices. Second, all total processing times in the time-based dataset are read, and extreme value normalization is used to map each total processing time to the range of 0 to 1, resulting in multiple total processing time data. Then, by adjusting the hyperparameters required for training the deep convolutional neural network (DCNN), the DCNN is trained using the 3D matrices, the normalized total processing time data, and topological features. The hyperparameters include: training iterations, batch size, optimizer type, etc. Preferably, the DCNN is trained on an image processor. The predicted total processing time is used as the output of the DCNN. If the loss between the predicted total processing time and the actual time is less than or equal to the prediction threshold, the deep convolutional neural network to be trained is completed, and a trained duration prediction model is obtained. If the loss between the predicted total processing time and the actual time is greater than the prediction threshold, the hyperparameters required for training the deep convolutional neural network are adjusted, and the deep convolutional neural network to be trained continues to be trained until the loss between the predicted total processing time and the actual time is less than the prediction threshold, and a trained duration prediction model is obtained. Preferably, the prediction threshold is 0.01.

[0113] The embodiments of this application also include an equipment information module and an order information module for modifying processing equipment information and order information. The equipment information module is used for storing, retrieving, adding, deleting, individually modifying, and batch modifying information and operating information of the processing equipment. The order information module is used for storing, retrieving, adding, and deleting basic information of orders and / or molds to be processed and molds in process (molds currently being processed) within an order. The order information module also includes an interactive visualization interface for 3D molds, used to modify the information of a single mold. The equipment information module and order information module use Vue, ElementUI, and ECharts to complete the human-computer interaction page and front-end and back-end logic interaction, realizing the modification of processing equipment information and order information.

[0114] The implementation method of this application adopts a message queue approach to decouple the information stored in the equipment information module and order information module from the synchronization between the location category recognition model and the duration prediction model. It adopts a multi-threaded approach to link the location category recognition model and the duration prediction model, thereby realizing the recognition of the processing location category data and the prediction of the processing duration data of each batch of molds to be processed by the location category recognition model and the duration prediction model, and saving this information.

[0115] Specifically, the implementation of this application includes three types of threads. The first type of thread drives the position category recognition model, the second type drives the duration prediction model, and the third type saves data such as the position category data to be processed, processing duration data, the priority of the mold to be processed, the operating information of the processing equipment, and equipment information and order information to the database. Each thread in the first and second types drives only one model. Each time, when the position category recognition model and / or duration prediction model are needed to identify the position category data to be processed and / or predict the processing duration data, the modeling file of the mold to be processed, after processing (these processes can also be added to the position category recognition model and duration prediction model), is sent to the "position category recognition" message queue and the "duration prediction" message queue, respectively. The first type of thread retrieves the mold information to be processed for position category recognition from the "position category recognition" message queue, such as multiple two-dimensional views; the second type of thread retrieves the mold information to be processed for duration prediction from the "duration prediction" message queue, such as the three-dimensional matrix of the mold to be processed. Then, the processing position categories of the mold to be processed are identified, and the processing time of the mold to be processed is predicted, obtaining the processing position category data and processing time data of the mold to be processed. The position category identification model sends the processing position category data to the "information storage" message queue, and the time prediction model sends the processing time data to the "information storage" message queue. The third type of thread retrieves the information from the "information storage" message queue to complete the data storage.

[0116] Existing technologies primarily focus on mold production. Since mold production uses methods like injection molding and casting, similar to 3D printing, to create a new mold, a single machine can complete the production of an entire mold. However, the embodiments of this application focus on mold processing. Unlike mold production, mold processing often requires multiple processing equipment to perform various processing techniques on the mold. Therefore, it necessitates the use of many different processing devices, requiring optimization of the multi-device processing scheme for the same mold to improve efficiency. The embodiments of this application provide an efficient optimization scheme, thereby solving the problem of optimizing the production scheme for processing a single mold using multiple devices. Specifically, by obtaining the processing position category data of the mold to be processed through a position category recognition model, the processing type of the mold can be determined. Since each processing type corresponds to different processing equipment, the number of different processing devices to be allocated to the mold can be determined based on the identified processing type. Since the processing sequence is known, different processing devices can be allocated to the mold according to the processing sequence. Furthermore, since the processing time can be obtained through a time prediction model, it is also known when each processing device will be able to complete its assigned processing task. This allows for the efficient allocation of different processing devices to the mold to be processed, thereby improving the processing efficiency of the mold.

[0117] Secondly, according to the embodiments of this application, a control system for mold processing is also proposed, such as... Figure 10 As shown, it includes: a mold identification module 101 and an equipment allocation module 102;

[0118] The mold recognition module 101 is used to process the modeling file of the mold to be processed and input it into the position category recognition model 110 and the duration prediction model 120 respectively to obtain the processing position category data and processing duration data of the mold to be processed.

[0119] The equipment allocation module 102 is used to allocate processing equipment to the mold to be processed based on the processing location category data, processing time data, priority of the mold to be processed, and operation information of the processing equipment.

[0120] The mold identification module 101 is also used to store data on the processing location category and processing time of the mold to be processed. The equipment allocation module 102 is also used to store operating information.

[0121] like Figure 11 As shown, the mold recognition module 101 also includes:

[0122] Topology feature extraction model 130 is used to extract the topology features of the mold to be processed from the 3D modeling file of the mold to be processed, and send them to the location category recognition model 110 and the duration prediction model 120;

[0123] The modeling file processing unit 140 is used to generate multiple two-dimensional views based on the modeling file of the mold to be processed and send them to the position category recognition model 110. The modeling file is a three-dimensional modeling file. The three-dimensional modeling file of the mold to be processed is converted into a three-dimensional matrix and sent to the duration prediction model 120.

[0124] Location category recognition model 110 includes:

[0125] The identification model unit 111 is used to extract mapping features from multiple two-dimensional views generated from the modeling file of the mold to be processed; send the mapping features to the feature fusion unit; perform target detection on the fused features to obtain the features to be identified; send the features to be identified to the multilayer feedforward neural network unit; determine the processing position category data of the non-selective processing technology of the mold to be processed based on the features to be identified, and send it to the data output unit.

[0126] The feature fusion unit 112 is used to extract and fuse features from the mapping features and the topological features of the mold to be processed, so as to obtain the fused features of the mold to be processed and send them to the recognition model unit.

[0127] The multilayer feedforward neural network unit 113 is used to determine the processing position category data of the selective processing technology of the mold to be processed based on the features to be identified, and send it to the data output unit.

[0128] Data output unit 114 determines the processing position category data of the mold to be processed based on the processing position category data of non-selective processing technology and the processing position category data of selective processing technology.

[0129] The recognition model unit 111 is preferably a target detection model based on the YOLOv7 model. The YOLOv7 model of the recognition model unit 111 includes: an input side, a bottleneck network, a neck side, and a head side. The output of the bottleneck network is connected to the feature fusion unit 112, which sends the mapped features to the feature fusion unit 112. The feature fusion unit 112 performs feature extraction and feature fusion on the mapped features and the topological features of the mold to be processed, and obtains the fused features of the mold to be processed, which are then sent to the Neck side. The Neck side performs target detection on the fused features to obtain the features to be identified. The features to be identified are then sent to the multilayer feedforward neural network unit 113 and the Head side. The Head side determines the processing position category data of the non-selective processing technology of the mold to be processed based on the features to be identified, and sends it to the data output unit 114. The multilayer feedforward neural network unit 113 determines the processing position category data of the selective processing technology of the mold to be processed based on the features to be identified, and sends it to the data output unit 114. The data output unit 114 determines the processing position category data of the mold to be processed based on the processing position category data of the non-selective processing technology and the processing position category data of the selective processing technology.

[0130] The multilayer feedforward neural network unit 113 includes: a first multilayer feedforward neural network and a second multilayer feedforward neural network.

[0131] like Figure 12As shown, the implementation of this application further includes: an exception handling module 103, used to acquire the operating information of the processing equipment, determine the processing equipment without exception and the processing equipment with exception; based on the priority information of the processing equipment with exception and the mold to be processed, determine whether there are any unrecorded processing equipment with exception and whether the mold to be processed has priority; if there are any unrecorded processing equipment with exception and the mold to be processed has no priority, then the current time point is taken as the exception time point; determine all processing tasks that have not started processing at the exception time point; acquire the processing location category data and processing time data of the processing tasks; clear... This involves identifying all pending tasks assigned to all processing equipment that did not begin processing at the abnormal time point; using a greedy algorithm, assigning processing equipment without abnormalities to the pending tasks and molds based on the data of processing equipment without abnormalities, the pending task's location category, processing time, and the pending mold's location category and processing time; if no abnormal processing equipment is recorded and the pending mold has priority, then the current time is considered the abnormal time point; identifying all pending tasks that did not begin processing on processing equipment without abnormalities at the abnormal time point; and obtaining the pending task's location category. Identify and process time data; clear all pending tasks assigned to all processing equipment that did not start processing at the abnormal time point; determine the processing sequence based on the pending position category data of the pending mold; assign processing equipment without abnormalities to the pending mold based on the processing sequence, running information, and processing time data; use a greedy algorithm to assign processing equipment to the pending tasks based on processing equipment without abnormalities, and the pending position category data and processing time data of the pending tasks; if there is an unrecorded processing equipment with abnormalities and the pending mold has a priority, then the current time is taken as the abnormal time point; determine... All pending processing tasks that did not start processing on all processing equipment at the abnormal time point; obtain the pending processing location category data and processing time data of the pending processing tasks; clear all pending processing tasks assigned to all processing equipment that did not start processing at the abnormal time point; determine the processing sequence based on the pending processing location category data of the pending mold, and assign the pending mold to processing equipment without abnormalities based on the processing sequence, running information, and processing time data; use a greedy algorithm to assign the pending processing tasks to processing equipment without abnormalities based on the pending processing location category data and processing time data of the pending processing tasks.

[0132] The front end of the exception handling module 103 includes: a priority information modification page for the mold to be processed and an operation information (fault information) modification page for the processing equipment. The exception handling module 103 is also used to assign a non-abnormal processing equipment to the processing task and the mold to be processed based on the information input from the front end, the adjustment of the priority information of the mold to be processed, and / or the adjustment of the operation information of the processing equipment.

[0133] The exception handling module 103 uses a heuristic greedy algorithm to complete the scheduling calculation of processing equipment in the workshop. When the priority information of the mold to be processed changes or there is an unrecorded abnormal processing equipment, the current time point will be regarded as the abnormal time point. All molds to be processed and processing tasks that started processing after the abnormal time point will be rescheduled in batches to avoid affecting the recent processing schedule.

[0134] The equipment allocation module 102 is specifically used to determine the processing sequence based on the processing position category data of the mold to be processed if there are no unrecorded abnormal processing equipment and the mold to be processed has no priority. Based on the processing sequence, running information and processing time data, it allocates abnormal processing equipment to the mold to be processed.

[0135] like Figure 13 As shown, the embodiments of this application further include:

[0136] The order information module 104 is used for saving and maintaining basic order information, such as the storage, retrieval, addition, and deletion functions of basic information on orders and / or molds to be processed and molds in processing (molds being processed) within orders; and the position category recognition model and duration prediction model in the linkage mold recognition module 101, which improve the information of molds to be processed in the order based on the processing position category data and processing duration data of the molds to be processed; the front end of the order information module 104 includes: order information retrieval, management, and addition pages;

[0137] The equipment information module 105 is used to save and maintain the basic information of a single processing equipment and adjust the working time of batch equipment, such as storing, retrieving, adding, deleting, modifying individual and batch information of processing equipment; the front end of the equipment information module 105 includes: equipment information retrieval, management, addition and overall management pages;

[0138] The scheduling visualization module 106 is used to visualize workshop scheduling information, such as the tasks in progress and tasks to be processed by processing equipment, using Gantt charts to make the schedule information easy to understand. The front end of the scheduling visualization module 106 includes a schedule visualization page based on category and order retrieval.

[0139] The front-end pages of the mold recognition module 101, equipment allocation module 102, exception handling module 103, order information module 104, equipment information module 105, and scheduling visualization module 106 all include UI pages. Specifically, the front-end pages are developed using Vue, ElementUI, and ECharts tools, enabling users to achieve convenient and highly readable interaction through the front-end pages of the system proposed in this application, and further realizing human-computer interaction pages and front-end and back-end logic interaction.

[0140] The backend of the mold identification module 101, equipment allocation module 102, exception handling module 103, order information module 104, equipment information module 105, and scheduling visualization module 106, developed using Flask and MariaDB, enables long-term storage and efficient data interaction operations for order information, mold information to be processed, processing equipment information, and processing information, such as recording, modification, and deletion functions, as well as recording and modification functions for process scheduling schedule information.

[0141] The implementation method of this application uses multi-threading and message queues to link and decouple the position category recognition model and duration prediction model in the mold recognition module 101. After the order is saved, the information improvement function is triggered, so that the system reads the order information in the database and uses the processing position category data and processing duration data output by the position category recognition model and duration prediction model to improve the processing information of each mold to be processed in the order.

[0142] The exception handling module 103 uses a greedy algorithm to generate a scheduling schedule based on the processing information of each mold to be processed, such as the processing location category data and processing time data, as well as the processing constraints of the factory, such as the priority of the mold to be processed and the operation information of the processing equipment. The module also saves the final result to the database. At the same time, it deletes all processing tasks of all processing equipment after the abnormal time point and reassigns all processing tasks on the processing equipment, including the non-abnormal (non-fault) period of the abnormal processing equipment.

[0143] In this application, compared to manually recording order details and scheduling of molds in Excel spreadsheets, the method automatically identifies and predicts the processing location category and processing time of the molds using recognition and prediction models. This reduces the time and manpower consumption of mold processing scheduling. Based on the processing location category data, processing time data, mold priority, and equipment operation information, processing equipment is allocated to the molds, enabling convenient and quick scheduling of mold processing. A heuristic greedy algorithm is used to perform shop floor scheduling calculations. In the event of changes in mold priority information or machine malfunctions, all processing tasks starting after the abnormal time point are recalculated in batches, thus not affecting recent processing schedules. Compared to the problems of manually recording data in Excel spreadsheets during actual processing, such as voluminous and complex data, difficult spreadsheet management, limited data modification and sharing methods, low data visibility, extremely low data error tolerance, high demand for human resources, and high dependence on manual labor, the implementation method of this application uses multi-threading and message queues to link and decouple the position category recognition model and duration prediction model in the mold recognition module. After the order is saved, the information improvement function is triggered, enabling the system to read the order information from the database and use the processing position category data and processing duration data output by the position category recognition model and duration prediction model to improve the processing information of each mold to be processed in the order. By automatically linking and updating the processing information of each mold to be processed in the order, and improving the processing position category data and processing duration data of each mold in the processing order, the reliance on technical staff can be reduced. The backend of each module developed based on Flask and MariaDB can realize convenient functions for saving, modifying, and deleting order data and its processing information, making the originally voluminous and complex data operation methods simpler and faster. The anomaly handling module and equipment allocation module, through a linkage location category recognition model and duration prediction model, schedule processing equipment in the workshop, automatically completing the scheduling of processing steps for the molds to be processed, thus eliminating the need for manual calculation of processing step allocation. By adding a priority information modification page for the molds to be processed and an operating information (fault information) modification page for the processing equipment to the anomaly handling module, the maintenance of processing information and the ability to handle anomalies can be facilitated, thereby making the implementation method of this application highly available. By adding a UI interface to the front end of each module, human-computer interaction functions are completed through a web page, realizing more convenient human-computer interaction and highly visualizing data. This helps factory (workshop) employees to retrieve their order information and mold processing information more simply and intuitively, significantly reducing the human resource consumption for managing order information in a real production environment, and enabling production work to proceed efficiently.

[0144] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A control method for mold processing, characterized in that, include: After processing the modeling file of the mold to be processed, it is input into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed. Based on the processing location category data, the processing time data, the priority of the mold to be processed, and the operating information of the processing equipment, one or more processing equipment are assigned to the mold to be processed; The step of allocating processing equipment to the mold to be processed based on the type of the location to be processed, the processing time data, the priority of the mold to be processed, and the operating information of the processing equipment includes: Obtain operational information from the processing equipment to identify equipment with and without abnormalities. Based on the priority information of the abnormal processing equipment and the mold to be processed, determine whether there are any unrecorded abnormal processing equipment and whether the mold to be processed has a priority. If there are any unrecorded abnormal processing devices and the mold to be processed has no priority, then the current time point is taken as the abnormal time point; determine all processing tasks that have not started processing on the processing devices at the abnormal time point; obtain the processing position category data and processing time data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; using a greedy algorithm, assign the processing devices without abnormalities to the processing tasks and the mold to be processed based on the processing devices without abnormalities, the processing position category data and processing time data of the processing tasks, and the processing position category data and processing time data of the mold to be processed; If there are no unrecorded abnormal processing devices and the mold to be processed has a priority, then the current time is taken as the abnormal time point; identify all processing tasks that have not started processing on any of the non-abnormal processing devices at the abnormal time point; obtain the processing position category data and processing duration data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; determine the processing order based on the processing position category data of the mold to be processed; assign non-abnormal processing devices to the mold to be processed based on the processing order, the running information, and the processing duration data; use a greedy algorithm to assign processing devices to the processing tasks based on the non-abnormal processing devices and the processing position category data and processing duration data of the processing tasks. If there are any unrecorded abnormal processing devices and the mold to be processed has a priority, then the current time is taken as the abnormal time point; identify all processing tasks that have not started processing on the processing devices at the abnormal time point; obtain the processing position category data and processing duration data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; determine the processing order based on the processing position category data of the mold to be processed; assign the abnormal-free processing device to the mold to be processed based on the processing order, the running information, and the processing duration data; use a greedy algorithm to assign the abnormal-free processing device to the processing task based on the abnormal-free processing device and the processing position category data and processing duration data of the processing task.

2. The method as described in claim 1, characterized in that, The process involves processing the modeling file of the mold to be processed and then inputting it into the position category recognition model and the time prediction model, respectively, to obtain the processing position category data and processing time data of the mold to be processed, including: Multiple two-dimensional views are generated based on the modeling file of the mold to be processed, wherein the modeling file is a three-dimensional modeling file; The topological features of the mold to be processed are determined based on the 3D modeling file; Multiple two-dimensional views and the topological features are input into the position category recognition model to obtain the position category data of the mold to be processed. The position category data includes the position category data of selective processing technology and the position category data of non-selective processing technology. The 3D modeling file of the mold to be processed is converted into a 3D matrix; The three-dimensional matrix and the topological features are input into the time prediction model to obtain the processing time data of the mold to be processed.

3. The method as described in claim 1, characterized in that, After determining whether there are any unrecorded abnormal processing devices and whether the mold to be processed has a priority based on the priority information of the abnormal processing equipment and the mold to be processed, the method further includes: If there are no unrecorded abnormal processing devices and the mold to be processed has no priority, then the processing sequence is determined according to the processing position category data of the mold to be processed, and the abnormal processing device is assigned to the mold to be processed according to the processing sequence, the running information and the processing time data.

4. The method as described in claim 1, characterized in that, Before the modeling file of the mold to be processed is processed and input into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed, the following steps are included: Generate location category datasets and duration datasets using historical orders; The topology feature extraction model is trained using the modeling files in the historical orders to obtain the trained topology feature extraction model. The trained topological feature extraction model is used to extract topological features from the modeling file of the mold to be processed; The location category recognition model is trained using the location category dataset and the topological features to obtain the trained location category recognition model. The duration prediction model is trained using the duration dataset and the topological features to obtain the trained duration prediction model.

5. The method as described in claim 4, characterized in that, The method of generating location category datasets and duration datasets using historical orders includes: Obtain the mold modeling file, processing location identifier, processing type of processing location, and processing time corresponding to the processing type of mold from historical orders. The modeling file is a three-dimensional modeling file. Generate multiple 2D views for each of the 3D modeling files; Based on the processing location identifier and the processing type of the processing location, mark the processing location for all two-dimensional views and record the processing type corresponding to the processing location to obtain a location category dataset; Based on the processing time corresponding to the processing type of the mold, record the required processing types and the total processing time for each processing type for each 3D modeling file to obtain a time dataset.

6. The method as described in claim 5, characterized in that, The step of training a location category recognition model using the location category dataset and the topological features to obtain the trained location category recognition model includes: Each of the two-dimensional views is converted into a two-dimensional matrix, resulting in multiple two-dimensional matrices; Based on the location category dataset, each processing location in each two-dimensional matrix is ​​normalized to obtain multiple normalized location data, wherein each normalized location data includes its corresponding processing type; The normalized location data and the topological features are used to train an artificial intelligence recognition model to obtain the trained location category recognition model.

7. The method as described in claim 5, characterized in that, The step of training a duration prediction model using the duration dataset and the topological features to obtain the trained duration prediction model includes: Convert the 3D modeling file of the duration dataset into a 3D matrix; The total processing time is normalized to obtain the normalized total processing time data for each processing type; The deep convolutional neural network is trained using the topological features, the three-dimensional matrix, and the normalized total processing time data to obtain the trained time prediction model.

8. A control system for mold processing, characterized in that, include: Mold recognition module and equipment allocation module; The mold recognition module is used to process the modeling file of the mold to be processed and input it into the position category recognition model and the duration prediction model respectively to obtain the processing position category data and processing duration data of the mold to be processed. The equipment allocation module is used to allocate processing equipment to the mold to be processed based on the processing location category data, the processing time data, the priority of the mold to be processed, and the operation information of the processing equipment. The step of allocating processing equipment to the mold to be processed based on the type of the location to be processed, the processing time data, the priority of the mold to be processed, and the operating information of the processing equipment includes: Obtain operational information from the processing equipment to identify equipment with and without abnormalities. Based on the priority information of the abnormal processing equipment and the mold to be processed, determine whether there are any unrecorded abnormal processing equipment and whether the mold to be processed has a priority. If there are any unrecorded abnormal processing devices and the mold to be processed has no priority, then the current time point is taken as the abnormal time point; determine all processing tasks that have not started processing on the processing devices at the abnormal time point; obtain the processing position category data and processing time data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; using a greedy algorithm, assign the processing devices without abnormalities to the processing tasks and the mold to be processed based on the processing devices without abnormalities, the processing position category data and processing time data of the processing tasks, and the processing position category data and processing time data of the mold to be processed; If there are no unrecorded abnormal processing devices and the mold to be processed has a priority, then the current time is taken as the abnormal time point; identify all processing tasks that have not started processing on any of the non-abnormal processing devices at the abnormal time point; obtain the processing position category data and processing duration data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; determine the processing order based on the processing position category data of the mold to be processed; assign non-abnormal processing devices to the mold to be processed based on the processing order, the running information, and the processing duration data; use a greedy algorithm to assign processing devices to the processing tasks based on the non-abnormal processing devices and the processing position category data and processing duration data of the processing tasks. If there are any unrecorded abnormal processing devices and the mold to be processed has a priority, then the current time is taken as the abnormal time point; identify all processing tasks that have not started processing on the processing devices at the abnormal time point; obtain the processing position category data and processing duration data of the processing tasks; clear all processing tasks assigned to all processing devices that have not started processing at the abnormal time point; determine the processing order based on the processing position category data of the mold to be processed; assign the abnormal-free processing device to the mold to be processed based on the processing order, the running information, and the processing duration data; use a greedy algorithm to assign the abnormal-free processing device to the processing task based on the abnormal-free processing device and the processing position category data and processing duration data of the processing task.

9. The system as described in claim 8, characterized in that, The location category recognition model includes: The identification model unit is used to extract mapping features from multiple two-dimensional views generated from the modeling file of the mold to be processed; send the mapping features to the feature fusion unit; perform target detection on the fused features to obtain the features to be identified; send the features to be identified to the multilayer feedforward neural network unit; determine the processing position category data of the non-selective processing technology of the mold to be processed based on the features to be identified, and send it to the data output unit. The feature fusion unit is used to extract and fuse features from the mapping features and the topological features of the mold to be processed, to obtain the fused features of the mold to be processed, and send them to the recognition model unit. A multi-layer feedforward neural network unit is used to determine the processing position category data of the selective processing technology of the mold to be processed based on the features to be identified, and send it to the data output unit; The data output unit determines the processing position category data of the mold to be processed based on the processing position category data of the non-selective processing technology and the processing position category data of the selective processing technology.

Images