A method and system for shaping medicine columns

Abstract

This invention discloses a method and system for shaping a drug column. The method includes: acquiring an image of the drug column to be shaped as a sample image; modeling a three-dimensional model based on the sample image and acquiring the three-dimensional coordinates of the three-dimensional model as initial coordinates; obtaining the three-dimensional coordinates of the shaped drug column as target coordinates based on the initial coordinates and processing dimension data; initializing a process parameter package, which includes several processing parameters, including: depth of cut, cutting tip rotation speed, cutting head movement speed, and cutting head offset angle; calculating morphological change parameters based on the initial coordinates and target coordinates, and selecting the corresponding processing parameters from the process parameter package as target processing parameters; determining processing compensation parameters based on the processing cutting head dimension data, including: axial wear compensation value and radial wear compensation value of the cutting head; and completing the shaping process of the drug column to be shaped based on the target processing parameters and processing compensation parameters.

Description

Technical Field

[0001] This invention relates to the field of drug column shaping technology, specifically a method and system for drug column shaping. Background Technology

[0002] Existing propellant cartridge shaping processes typically rely on manual experience or pre-set fixed procedures, which presents key technical bottlenecks: On the one hand, due to the differences in the initial geometry of different propellant cartridges, fixed-parameter processing modes are difficult to dynamically adapt to the shaping needs of various complex shapes, easily leading to discrepancies between the shaped cartridge size and the expected size; on the other hand, during the processing, tool wear affects the cutting performance of the tool tip, and traditional methods lack online detection and real-time compensation mechanisms for tool wear, causing the processing accuracy to gradually deteriorate as the processing flow progresses, ultimately resulting in the shaped cartridge size not meeting the shaping specifications. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for shaping drug cartridges.

[0004] The objective of this invention is mainly achieved through the following technical solutions:

[0005] On one hand, this application provides a method for shaping a drug column, comprising:

[0006] Obtain an image of the drug column to be shaped as a sample image;

[0007] A 3D model is formed based on the sample image, and the 3D coordinates of the 3D model are obtained as the initial coordinates.

[0008] The three-dimensional coordinates of the shaped propellant column are obtained by matching the initial coordinates and processing size data, and used as the target coordinates.

[0009] Initialize the process parameter package, which includes several machining process parameters. Each machining process parameter includes: depth of cut, tool tip speed, tool head movement speed, and tool head offset angle.

[0010] The shape change parameters are calculated based on the initial coordinates and the target coordinates, and the corresponding processing parameters are selected from the process parameter package as the target processing parameters.

[0011] The machining compensation parameters are determined based on the machining head size data. The machining compensation parameters include: the axial wear compensation value and the radial wear compensation value of the cutting head.

[0012] The shaping process of the drug column to be shaped is completed based on the target processing parameters and processing compensation parameters;

[0013] The process parameter package is obtained through a neural network model, and the process includes:

[0014] Obtain the initial and target coordinates aligned to the timestamp in the historical processing and shaping process table, and obtain the processing parameters corresponding to the initial and target coordinates;

[0015] The morphological change parameters of the propellant grain are calculated using the initial and target coordinates, and the formulas include:

[0016] ;

[0017] in, Represents morphological change parameters, ( , , ) represents the coordinates of the i-th point in the initial coordinate system. , , () represents the coordinates of the i-th point in the target coordinate system;

[0018] A neural network model is constructed, and the morphological change parameters and their corresponding processing parameters are used as sample data to train the neural network model. The input features of the neural network model are the morphological change parameters, and the output features of the neural network model are the processing parameters.

[0019] The neural network model outputs the qualified processing parameters corresponding to the predicted different morphological change parameters as the optimal process parameters.

[0020] The timestamps, morphological change parameters, and their corresponding optimal process parameters are stored in the process parameter package according to their association relationships.

[0021] Furthermore, the training steps for a neural network model include:

[0022] The morphological change parameters are used as input features X, and the processing parameters are used as labels Y;

[0023] The input features X and labels Y are assigned to the input layer of the neural network model for standardization and normalization.

[0024] The hidden layer receives the input features X processed by each input layer and uses the weighted input calculation formula to calculate the weighted input of each hidden layer.

[0025] The weighted input of each hidden layer is calculated using the first activation function to obtain the output value of each hidden layer neuron.

[0026] The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer.

[0027] The processing parameters corresponding to each morphological change parameter are predicted by the second activation function, and the predicted processing parameters are evaluated by the loss function until the evaluation is qualified.

[0028] The output layer will output the qualified processing parameters as the final result value.

[0029] Furthermore, the formula for the first activation function includes:

[0030] ;

[0031] in, This represents the output value of the i-th neuron in the hidden layer. This represents the weighted input to the i-th neuron in the hidden layer.

[0032] Furthermore, the formula for the second activation function includes:

[0033] ;

[0034] in, This represents the processing parameters predicted by the output layer. This represents the weighted input of the output layer.

[0035] Furthermore, the steps for determining machining compensation parameters based on the tool head size data include:

[0036] Obtain the front view, side view, and top view of the machining head before it is worn as reference drawings;

[0037] Collect the current front view, side view, and top view of the machining head as actual measurement drawings;

[0038] Compare the measured image with the reference image to calculate the wear amount of the cutting tip profile and the wear amount of the cutting head side.

[0039] Wear compensation is calculated based on the wear amount of the cutting tip profile to obtain the radial wear compensation value of the cutting tip;

[0040] Wear compensation is calculated based on the amount of wear on the side of the cutter head to obtain the axial wear compensation value of the cutter head.

[0041] Furthermore, the formula for calculating wear compensation based on the wear amount of the tool tip profile includes:

[0042] ;

[0043] in, This indicates the radial wear compensation value of the cutting head. Indicates the radial compensation coefficient. This indicates the amount of wear on the radius of the blade tip arc, as measured by image recognition. This indicates the radial compensation correction amount.

[0044] Furthermore, the formula for calculating wear compensation based on the wear amount on the side of the cutter head includes:

[0045] ;

[0046] in, This indicates the axial wear compensation value of the tool tip. Indicates the axial compensation coefficient. This indicates the amount of wear on the cutter head, measured through image recognition. This indicates the axial compensation correction amount.

[0047] On the other hand, this application also provides a system for shaping drug cartridges, comprising:

[0048] The acquisition unit is configured to acquire an image of the drug column to be shaped as a sample image;

[0049] The modeling unit is configured to perform 3D modeling based on the sample image to form a three-dimensional model, and obtain the three-dimensional coordinates of the three-dimensional model as the initial coordinates;

[0050] The matching unit is configured to match the three-dimensional coordinates of the shaped propellant column based on the initial coordinates and processing size data as the target coordinates;

[0051] The prediction unit is configured to predict the processing parameters corresponding to the morphological change parameters of the drug column through a neural network model.

[0052] The initialization unit is configured to initialize the process parameter package;

[0053] The compensation parameter determination unit is configured to determine machining compensation parameters based on the machining head size data.

[0054] The parameter selection unit is configured to select the corresponding processing parameters as the target processing parameters from the process parameter package based on the calculated morphological change parameters.

[0055] The processing execution unit is configured to complete the shaping of the drug column to be shaped according to the target processing parameters and processing compensation parameters.

[0056] In summary, the present invention has the following advantages compared with the prior art:

[0057] The core benefits of this solution lie in its precise resolution of two key issues in traditional pharmacopoeia shaping: poor adaptability due to reliance on fixed parameters and human experience, and piece-by-piece degradation of machining accuracy caused by tool wear. A neural network model trained on historical data intelligently predicts and dynamically adapts optimal machining parameters based on the morphological changes between the initial and final shapes of the pharmacopoeia, overcoming the difficulty of fixed programs in handling variations in pharmacopoeia shape and ensuring the accuracy of the machining strategy from the outset. Simultaneously, the vision module integrated into the system quantitatively identifies tool wear and automatically calculates compensation values ​​by acquiring images of the machining tool and comparing them with a reference image. This enables real-time online compensation for tool wear, effectively suppressing accuracy drift caused by tool wear. The synergistic effect of these two core technologies constructs a closed-loop intelligent machining system from intelligent parameter decision-making to real-time compensation during execution, significantly reducing reliance on operator experience and substantially improving the dimensional consistency, pass rate, and production efficiency of pharmacopoeia shaping. Attached Figure Description

[0058] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0059] Figure 1 The flowchart of the drug column shaping method provided in this application;

[0060] Figure 2 A schematic diagram of the system structure for drug column shaping provided in this application. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0062] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0063] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0064] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0065] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0066] Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of the present invention.

[0067] In the following description, suffixes such as "module," "part," "component," or "unit" are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, they can be used interchangeably.

[0068] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0069] Firstly, please refer to the following: Figure 1 This application provides a method for shaping a drug delivery column, including:

[0070] Obtain an image of the drug column to be shaped as a sample image;

[0071] A 3D model is formed based on the sample image, and the 3D coordinates of the 3D model are obtained as the initial coordinates.

[0072] The three-dimensional coordinates of the shaped propellant column are obtained by matching the initial coordinates and processing size data, and used as the target coordinates.

[0073] Initialize the process parameter package, which includes several machining process parameters, wherein each machining process parameter includes: depth of cut, tool tip rotation speed, tool head movement speed, and tool head offset angle;

[0074] The shape change parameters are calculated based on the initial coordinates and the target coordinates, and the corresponding processing parameters are selected from the process parameter package as the target processing parameters.

[0075] The machining compensation parameters are determined based on the machining head size data. The machining compensation parameters include: the axial wear compensation value and the radial wear compensation value of the cutting head.

[0076] The shaping process of the drug column to be shaped is completed according to the target processing parameters and processing compensation parameters.

[0077] The process parameter package is obtained through a neural network model, and the process includes:

[0078] Obtain the initial and target coordinates aligned to the timestamp in the historical processing and shaping process table, and obtain the processing parameters corresponding to the initial and target coordinates;

[0079] The morphological change parameters of the propellant grain are calculated using the initial and target coordinates, and the formulas include:

[0080] ;

[0081] in, Represents morphological change parameters, ( , , ) represents the coordinates of the i-th point in the initial coordinate system. , , () represents the coordinates of the i-th point in the target coordinate system;

[0082] A neural network model is constructed, and the morphological change parameters and their corresponding processing parameters are used as sample data to train the neural network model. The input features of the neural network model are the morphological change parameters, and the output features of the neural network model are the processing parameters.

[0083] The neural network model outputs the qualified processing parameters corresponding to the predicted different morphological change parameters as the optimal process parameters.

[0084] The timestamps, morphological change parameters, and their corresponding optimal process parameters are stored in the process parameter package according to their association relationships.

[0085] In this embodiment, the core process is as follows: First, high-resolution industrial cameras are used to acquire images of the drug column to be shaped from multiple fixed angles, obtaining high-definition images from multiple angles as sample images. Then, computer vision algorithms (SFM or Structure from Motion) are used to process the sample images, generating a high-precision 3D model of the drug column through 3D point cloud reconstruction technology, and extracting the 3D spatial coordinates of each feature point on the surface of the model as initial coordinates. Subsequently, the operator clamps and fixes the drug column to be shaped on the drug column shaping machine, and then accesses the processing and shaping database of the shaping machine's control system. The main table of this database stores the initial coordinate samples of various types of drug columns in the past shaping processing history, as well as the corresponding 3D coordinates of the shaped drug column after shaping (the mapping relationship between the initial coordinates and the 3D coordinates of the shaped drug column in the main table is one-to-many). Then, the system performs a preliminary matching query on the database using the currently acquired initial coordinates, retrieving multiple similar historical shaping target coordinates. Based on the specific processing dimensions (diameter and height of the shaped propellant grain) as a filtering condition, it selects the three-dimensional coordinates that best match the current processing target from these historical coordinates as the target coordinates for this processing. Next, the system initializes the process parameter package and calculates the overall morphological change parameters of the propellant grain using the current initial coordinates and the selected target coordinates. The system then uses these morphological change parameters to retrieve the corresponding processing parameters from the process parameter package as the target processing parameters for this processing.

[0086] Next, the system acquires image data of the current machining head through the vision module integrated on the shaping machine. Using image processing algorithms, it performs a precise comparison between the current image data and a pre-stored reference image (i.e., an image of the machining head in its standard state before use). By calculating contour differences and dimensional changes, it accurately calculates the axial and radial wear of the machining head and obtains machining compensation parameters accordingly. Finally, the CNC system loads the target machining parameters and machining compensation parameters to generate the final machining path instructions, driving the shaping machine to complete the high-precision, adaptive shaping of the medicament cells to be shaped.

[0087] Throughout the shaping process, the cutting trajectory is based on the target machining parameters, and machining compensation parameters are dynamically applied to correct the tool path in real time. Specifically, the motion controller generates an initial shaping path based on the depth of cut, tool tip speed, tool head movement speed, and tool head offset angle from the target machining parameters; simultaneously, the axial wear compensation value and radial wear compensation value of the tool head are superimposed on the shaping path coordinates of the tool head to achieve precise compensation for tool wear; during machining, the servo system precisely controls the spindle rotation based on the tool tip speed parameters and monitors the tool position in real time through a 3D scanner to ensure that the shaping process is executed with high precision according to the compensated trajectory, ultimately keeping the error between the actual machined dimensions of the propellant and the target coordinates within the allowable range.

[0088] In this embodiment, the system first accesses the processing and shaping database, querying the historical processing and shaping process table using initial and target coordinates. This table stores a large number of successful processing case data sets from past processing history. Each data set strictly includes the initial coordinates before processing, the target coordinates after processing, and the processing parameters used in that processing, sorted by timestamp. Subsequently, the system calls the morphology change parameter calculation module to calculate the morphology change quantification value (i.e., morphology change parameter) corresponding to each data set in batches using the initial and target coordinates. After calculation, the system uses these morphology change parameters as model input features and their corresponding processing parameters as output labels, together forming a training sample set. This sample set is fed into a multilayer perceptron neural network for training. During training, the model continuously adjusts its internal weights through optimization algorithms to learn the mapping relationship between "morphology change" and "optimal processing parameters." After training, the neural network model has predictive capabilities. For any new morphology change parameter input, it can output a set of qualified processing parameters with a high probability of achieving that morphology change. This set of parameters is identified by the system as the optimal processing parameters. Finally, the system stores the historical processing timestamps corresponding to the initial and target coordinates, as well as the morphological change parameters and optimal process parameters corresponding to those timestamps, into a process parameter package according to their associations, forming a continuously growing and optimized process parameter knowledge base. Subsequently, the optimal process parameters corresponding to the process parameter package can be retrieved simply by using the timestamps and morphological change parameters as unique indexes.

[0089] In one possible implementation, the training steps of the neural network model include:

[0090] The morphological change parameters are used as input features X, and the processing parameters are used as labels Y;

[0091] The input features X and labels Y are assigned to the input layer of the neural network model for standardization and normalization.

[0092] The hidden layer receives the input feature X processed by each of the input layers and calculates the weighted input of each hidden layer using the weighted input calculation formula;

[0093] The weighted input of each hidden layer is calculated using the first activation function to obtain the output value of each hidden layer neuron.

[0094] The output layer receives the output value of each hidden layer neuron and uses the weighted input calculation formula to calculate the weighted input of each output layer.

[0095] The processing parameters corresponding to each morphological change parameter are predicted by the second activation function, and the predicted processing parameters are evaluated by the loss function until the evaluation is qualified.

[0096] The output layer will output the qualified processing parameters as the final result value.

[0097] In this embodiment, during the model training phase, the input layer first performs data preprocessing on the input features and output labels, including standardization and normalization, to eliminate dimensional differences and accelerate model convergence. The preprocessed data is then fed into the hidden layer. Each neuron in the hidden layer receives the output values ​​of all input layer neurons, calculates its weighted input, and performs a nonlinear transformation using a first activation function to obtain the neuron's output value. The final output layer of the model receives the output of the last hidden layer, calculates its weighted input, and generates the final predicted value, i.e., the predicted processing parameters, using a second activation function. Subsequently, the difference between the predicted value and the true label is calculated using a loss function, and the weights and bias parameters in the network are iteratively updated using a backpropagation algorithm and an optimizer to minimize the loss function. This entire process is iterated until the model's prediction accuracy reaches the preset convergence criterion, at which point training is complete. After passing the evaluation, the model is deployed for actual prediction.

[0098] In one possible implementation, the formula for the first activation function includes:

[0099] ;

[0100] in, This represents the output value of the i-th neuron in the hidden layer. This represents the weighted input to the i-th neuron in the hidden layer.

[0101] In one possible implementation, the formula for the second activation function includes:

[0102] ;

[0103] in, This represents the processing parameters predicted by the output layer. This represents the weighted input of the output layer.

[0104] In one possible implementation, the steps for determining machining compensation parameters based on the tool head size data include:

[0105] Obtain the front view, side view, and top view of the machining head before it is worn as reference drawings;

[0106] The current front view, side view, and top view of the machining head are collected as actual measurement drawings;

[0107] The measured image is compared with the reference image to calculate the wear amount of the blade tip contour and the wear amount of the blade side.

[0108] Wear compensation is calculated based on the wear amount of the cutting tip profile to obtain the radial wear compensation value of the cutting head;

[0109] Wear compensation is calculated based on the wear amount on the side of the cutter head to obtain the axial wear compensation value of the cutter head.

[0110] In this embodiment, when the cutting head is in brand new condition, a CCD camera is used to capture clear images of the cutting head from three orthogonal directions: its front, side, and bottom. After rigorous calibration, these images are stored as reference images. Before each machining task, the system again uses the vision module to capture images of the current cutting head's front, side, and bottom surfaces as measured images. Subsequently, an image registration algorithm is used to precisely align the measured images with the reference images. Then, image processing and computer vision algorithms are used to calculate the wear amount of the cutting tip's arc contour and the wear amount of the cutting head's side length. Based on the calculated cutting tip contour wear amount, the radial wear compensation calculation formula is substituted, taking into account the tool material and cutting characteristics, to calculate an accurate radial wear compensation value for the cutting head. Similarly, based on the wear amount of the cutting head's side, the axial wear compensation calculation formula is substituted to calculate an accurate axial wear compensation value for the cutting head. These compensation values ​​are incorporated in real time into the subsequent CNC machining code generation to dynamically compensate the tool path, thereby ensuring that the dimensional accuracy of the machined propellant cartridges is not affected by tool wear.

[0111] In one possible implementation, the formula for calculating wear compensation based on the profile wear of the cutting tip includes:

[0112] ;

[0113] in, This indicates the radial wear compensation value of the cutting head. Indicates the radial compensation coefficient. This indicates the amount of wear on the radius of the blade tip arc, as measured by image recognition. This indicates the radial compensation correction amount.

[0114] In one possible implementation, the formula for calculating wear compensation based on the wear amount on the side of the cutter head includes:

[0115] ;

[0116] in, This indicates the axial wear compensation value of the tool tip. Indicates the axial compensation coefficient. This indicates the amount of wear on the cutter head, measured through image recognition. This indicates the axial compensation correction amount.

[0117] Secondly, please refer to the following: Figure 2 This application provides a system based on the above-described drug cartridge shaping method, comprising:

[0118] The acquisition unit is configured to acquire an image of the drug column to be shaped as a sample image;

[0119] The modeling unit is configured to perform 3D modeling based on the sample image to form a three-dimensional model, and to obtain the three-dimensional coordinates of the three-dimensional model as initial coordinates;

[0120] The matching unit is configured to match the three-dimensional coordinates of the shaped propellant column based on the initial coordinates and processing size data as the target coordinates;

[0121] The prediction unit is configured to predict the processing parameters corresponding to the morphological change parameters of the drug column through a neural network model.

[0122] The initialization unit is configured to initialize the process parameter package;

[0123] The compensation parameter determination unit is configured to determine machining compensation parameters based on the machining head size data.

[0124] The parameter selection unit is configured to select the corresponding processing parameters as the target processing parameters from the process parameter package based on the calculated morphological change parameters.

[0125] The processing execution unit is configured to complete the shaping process of the drug column to be shaped according to the target processing parameters and processing compensation parameters.

[0126] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of bullet sizing, characterized by, The method comprises: acquiring an image of a to-be-shaped medicine column as a sample image; performing 3D modeling according to the sample image to form a three-dimensional model, and acquiring three-dimensional coordinates of the three-dimensional model as initial coordinates; matching the initial coordinates and machining size data to obtain three-dimensional coordinates of the shaped medicine column as target coordinates; initializing a process parameter package, wherein each machining process parameter in the process parameter package comprises a depth of tool descent, a tool tip rotation speed, a tool head movement speed, and a tool head deflection angle; calculating a shape change parameter according to the initial coordinates and the target coordinates, and selecting corresponding machining process parameters in the process parameter package as target machining parameters; determining machining compensation parameters according to machining tool head size data, wherein the machining compensation parameters comprise a tool head axial wear compensation value and a tool head radial wear compensation value; completing the shaping machining of the to-be-shaped medicine column according to the target machining parameters and the machining compensation parameters; the process parameter package is obtained through a neural network model, and the process comprises: acquiring initial coordinates and target coordinates aligned by time stamps in a historical machining shaping process table, and acquiring machining process parameters corresponding to the initial coordinates and the target coordinates; calculating a shape change parameter of the medicine column using the initial coordinates and the target coordinates, wherein the formula comprises: ； wherein, denotes a morphing parameter, , , ) denotes the coordinates of the i-th point in the initial coordinates, , , ) denotes the coordinates of the i-th point in the target coordinates; constructing a neural network model, and assigning the shape change parameter and the corresponding machining process parameter as sample data to the neural network model for training, wherein the input feature of the neural network model is the shape change parameter, and the output feature of the neural network model is the machining process parameter; the neural network model outputs qualified machining process parameters corresponding to different predicted shape change parameters as optimal process parameters; storing the time stamps, the shape change parameters, and the corresponding optimal process parameters in the process parameter package according to the correlation.

2. A method of bullet sizing as defined in claim 1, wherein, the training steps of the neural network model comprise: taking the shape change parameter as an input feature X and the machining process parameter as a label Y; assigning the input feature X and the label Y to the input layer of the neural network model for standardization and normalization processing; each hidden layer receives the processed input feature X of each input layer, and calculates the weighted input of each hidden layer using a weighted input calculation formula; each hidden layer calculates the weighted input using a first activation function to obtain the output value of each hidden layer neuron; each output layer receives the output value of each hidden layer neuron, and calculates the weighted input of each output layer using a weighted input calculation formula; the second activation function is used to predict the machining process parameter corresponding to each shape change parameter, and a loss function is used to evaluate the predicted machining process parameter until the evaluation is qualified; the output layer outputs the qualified machining process parameter as a final result value.

3. A method of bullet sizing as defined in claim 2, wherein, the formula of the first activation function comprises: ； wherein, represents the output value of the i-th neuron of the hidden layer, represents the weighted input of the i-th neuron of the hidden layer.

4. A method of bullet sizing as defined in claim 2, wherein, the formula of the second activation function comprises: ； wherein, represents an output layer predicted process parameter, represents a weighted input of the output layer.

5. The method of sizing a propellant grain according to claim 1 wherein, the step of determining the machining compensation parameters according to the tool head size data comprises: acquiring a front view, a side view, and a top view of the machining tool head before wear as a reference image; acquiring a front view, a side view, and a top view of the machining tool head as a measured image; The measured image is compared with the reference image, and the profile wear of the tool tip and the wear of the side surface of the tool head are calculated; According to the profile wear of the tool tip, the wear compensation calculation is carried out to obtain the radial wear compensation value of the tool head; According to the wear of the side surface of the tool head, the wear compensation calculation is carried out to obtain the axial wear compensation value of the tool head.

6. A method of bullet sizing as defined in claim 5, wherein, The formula for wear compensation calculation according to the profile wear of the tool tip includes: ； wherein, represents a tool head radial wear compensation value, represents a radial compensation coefficient, represents a wear amount of the tool tip circular arc radius measured by image recognition, represents a radial compensation correction amount.

7. A method of sizing a propellant grain as defined in claim 5, wherein, The formula for wear compensation calculation according to the wear of the side surface of the tool head includes: ； wherein, represents an axial wear compensation value of the tool head, represents an axial compensation coefficient, represents a wear amount of the tool head length measured by image recognition, represents an axial compensation correction amount.

8. A system for performing the method of sizing a drug rod according to any one of claims 1 to 7, characterized in that Including: An acquisition unit is configured to acquire an image of a to-be-shaped drug column as a sample image; A modeling unit is configured to perform 3D modeling according to the sample image to form a three-dimensional model, and acquire three-dimensional coordinates of the three-dimensional model as initial coordinates; A matching unit is configured to match the initial coordinates and the processing size data to obtain three-dimensional coordinates of the shaped drug column as target coordinates; A prediction unit is configured to predict the processing process parameters corresponding to the morphological change parameters of the drug column through a neural network model; An initialization unit is configured to initialize the process parameter package; A compensation parameter determination unit is configured to determine the processing compensation parameters according to the processing tool head size data; A parameter selection unit is configured to select the corresponding processing process parameters as target processing parameters in the process parameter package according to the calculated morphological change parameters; A processing execution unit is configured to complete the shaping processing of the to-be-shaped drug column according to the target processing parameters and the processing compensation parameters.

Images