Tool wear monitoring method and detection device based on digital twin driving

Abstract

This invention relates to a tool wear monitoring method and detection device driven by digital twins. The purpose of this invention is to monitor and predict the tool wear state during CNC milling processes, achieving three-dimensional visual monitoring of the tool wear state. It includes a workpiece transfer mechanism, a depth camera bracket, a depth camera, a CCD camera, and an end effector rotation mechanism, an X-axis motion mechanism, a Y-axis motion mechanism, and a Z-axis motion mechanism. The method includes: constructing a digital space for the CNC milling process; establishing a connection between the physical space and the digital space; determining the position of the workpiece milling surface based on the acquired data; the digital twin system controlling the three-axis motion mechanism to drive the CCD camera to acquire images of the workpiece milling surface at the viewpoint; constructing a tool monitoring model based on SVM; and acquiring image data in real time. This invention belongs to the field of intelligent manufacturing technology.

Description

Technical Field

[0001] This invention relates to a tool wear monitoring method and detection device, specifically to a tool wear monitoring method and detection device based on digital twin drive, and belongs to the field of intelligent manufacturing technology. Background Technology

[0002] With the emergence of technologies such as big data, the Internet of Things, and cloud computing, digital twins have gradually been proposed to promote the transformation of the manufacturing industry. As an advanced technology for implementing smart manufacturing and Industry 4.0, digital twins use a data-driven approach to describe and model the attributes, characteristics, behaviors, and performance of physical entities, and are gradually being applied in manufacturing workshops, ships, vehicles, and other fields. In recent years, digital twins have gradually developed in the industrial sector, providing new solutions for the monitoring and optimization of mechanical products.

[0003] In the field of mechanical manufacturing, cutting tools, acting as the "teeth" of machine tools, have a significant impact on the stability of the entire machining process. During cutting, cutting tools experience wear and breakage due to thermal coupling. Severe tool wear can affect workpiece machining quality and efficiency, and may even lead to machine tool downtime, disrupting the normal operation of the entire machining system. Tool wear accounts for 20% of total machine tool downtime. Therefore, timely and accurate monitoring of tool wear is a crucial prerequisite for ensuring the quality and safety of both workpieces and machine tools.

[0004] In conventional machining processes, tool condition monitoring is typically divided into direct and indirect monitoring. Direct monitoring involves directly visually measuring the wear condition of the cutting tool; indirect monitoring measures the tool wear using machining process signals. Since the side areas of the tool are difficult to measure during direct monitoring, online capture of process signals is used to monitor tool wear. During milling, the tool and workpiece make periodic contact. Due to the nature of this complex process mechanism, periodic process signals are obtained, leading to unpredictable and premature tool wear. Furthermore, the sensors used to measure these signals are in direct contact with the machine tool, which can hinder machining. To overcome these problems, this invention proposes a digital twin-driven, non-tactile, low-cost, and high-speed tool wear monitoring method. Summary of the Invention

[0005] The purpose of this invention is to provide a tool wear monitoring method and detection device based on digital twin drive, which can monitor and predict the tool wear state during CNC machine tool milling process, and realize three-dimensional visualization monitoring of tool wear state.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] The tool wear detection device based on digital twin drive includes a workpiece conveying mechanism, a depth camera bracket, a depth camera, a CCD camera, an end effector rotation mechanism, an X-axis motion mechanism, two Y-axis motion mechanisms, and two Z-axis motion mechanisms. The depth camera bracket is mounted on the conveying mechanism frame, and the depth camera is mounted on the depth camera bracket above the workpiece conveying mechanism. The two Z-axis motion mechanisms are symmetrically mounted on the frames on both sides of the workpiece conveying mechanism. Each Z-axis motion mechanism has a Y-axis motion mechanism slidably mounted on it, and the two Y-axis motion mechanisms are arranged vertically relative to each other. The X-axis motion mechanism is slidably mounted on the two Y-axis motion mechanisms. The CCD camera is mounted on the X-axis motion mechanism through the end effector rotation mechanism. The workpiece is placed on the conveyor belt of the workpiece conveying mechanism.

[0008] A digital twin-driven method for monitoring tool wear, comprising the following steps:

[0009] Step 1: Construct a virtual space for the CNC machine tool milling process and perform real-time milling simulation;

[0010] Step 2: Establish a connection between the physical space and the virtual space, and obtain multi-data from the physical space;

[0011] Step 3: Determine the position of the workpiece milling surface based on the acquired data to obtain the viewpoint during image acquisition;

[0012] Step 4: The digital twin system controls the three-axis motion mechanism to drive the CCD camera to acquire images of the milled surface of the workpiece at the viewpoint;

[0013] Step 5: Construct a tool monitoring model based on SVM, and build a dataset to train and test the model, so that the model has high recognition accuracy and generalization ability;

[0014] Step 6: Collect image data in real time, input it into the model to obtain the tool wear value, and issue a warning to replace the tool when the monitored tool wear value exceeds the tool wear threshold.

[0015] Furthermore, the virtual space construction in step one includes:

[0016] 3D modeling: In industrial modeling software, a 3D model in virtual space is created based on the actual CNC machine tool and testing equipment, and the model is optimized in 3D modeling software;

[0017] Scene setup: Import the optimized model into 3D visualization software to build the scene;

[0018] Behavioral logic model construction: After the scene is built, the specific behaviors and logical relationships of the virtual model in the virtual space are given; based on the data collected in real time by external sensors and machine tool communication protocols, the variables bound to the associated model and the preset behavioral actions, the CNC machine tool driven by multiple data performs milling simulation, builds the parent-child relationship between the model tree structure, and improves the construction of the machine tool virtual space;

[0019] Human-computer interaction: The system interface is developed in 3D visualization software to realize 3D visualization monitoring of tool wear status.

[0020] Furthermore, the establishment of the connection between physical space and virtual space in step two includes:

[0021] Register an OPC UA server in the CNC system;

[0022] A TCP / IP network layer communication connection is established between the physical space and the virtual space via a Socket interface;

[0023] After establishing communication between the machine tools and testing devices in the physical space and the virtual space, the machine tool operating status information and workpiece milling surface image data are obtained by accessing the OPC UA server. The virtual space receives multi-data information from the physical space and stores it in the MySQL database.

[0024] Furthermore, static information includes machine tool attribute data, tool attribute data, workpiece attribute data, and process attribute data; among which, machine tool attribute data includes equipment model and system model, tool attribute data includes tool name, tool material, and tool parameters, workpiece attribute data includes workpiece name, workpiece material, and workpiece size, and process attribute data includes clamping method and machining parameters; machining parameters include cutting speed, feed rate, and depth of cut; dynamic information is machine tool status data.

[0025] Furthermore, based on the scene point cloud acquired by the depth camera, the scene point cloud is denoised and segmented to obtain the target point cloud, the workpiece model is point cloudified to obtain the source point cloud, and the source point cloud and the target point cloud are registered to obtain the pose of the workpiece in the scene.

[0026] The position of the milled surface on the detection device is obtained by combining the workpiece coordinate system and the position of the milled surface extracted from the NC program of the CNC system with the workpiece pose.

[0027] Furthermore, the viewpoint for image acquisition is determined based on the workpiece model, the position of the milled surface, and the optical parameters of the camera.

[0028] The digital twin system controls a three-axis motion mechanism to drive a CCD camera to acquire image data of the milled surface of the workpiece.

[0029] Furthermore, an SVM machine learning model is constructed. The gray-level co-occurrence matrix is ​​used to extract image contrast and second diagonal moments from the acquired images. These two extracted features are input into the SVM model to predict tool wear status. A tool wear dataset is constructed, and the dataset is randomly divided into training and test sets for each wear status at a 3:1 ratio. Feature extraction and feature selection are performed on the training and test sets respectively to determine the feature vector set. After normalization, a grid search method is used to find all points within the range for parameter optimization, obtaining as many hyperparameter combinations as possible. This process is repeated three times through cross-validation to find the penalty factor and kernel parameter hyperparameter combination that maximizes the classification accuracy of the dataset. Finally, the tool monitoring model is obtained.

[0030] Furthermore, the wear value of the cutting tool is obtained, and a warning is issued based on the threshold to remind the user to change the tool.

[0031] The beneficial effects of this invention compared to the prior art are:

[0032] Most current methods for monitoring tool wear involve acquiring and analyzing signals during the machine tool machining process. This approach not only interferes with workpiece machining but also makes it difficult to comprehensively analyze tool wear. This invention constructs a digital twin system for tool detection by establishing a virtual space within the physical space of the milling process. This system can indirectly monitor tool wear by acquiring image data of the workpiece surface. Sensor data acquired in the physical space is input into a machine learning model in the virtual space to obtain the current tool wear detection value, which is then displayed in a 3D visualization interface.

[0033] Compared to traditional methods, the combination of digital twin technology and machine learning models fully leverages the advantages of digital twins in terms of virtual-real integration and interaction, while the machine learning models are capable of predictive analysis. Therefore, this invention can significantly improve the ability to analyze tool wear conditions during milling, effectively increasing tool utilization, machining efficiency, and machining quality.

[0034] The purpose of this invention is to provide a digital twin-driven tool wear monitoring method that can monitor tool wear values ​​in real time during milling. By using digital twin technology to analyze image data collected during milling, the tool wear status can be obtained, thereby improving workpiece machining quality, machine tool efficiency, and tool utilization. Attached Figure Description

[0035] Figure 1 This is a flowchart of the tool wear monitoring method for milling processes according to the present invention.

[0036] Figure 2 This is a flowchart illustrating the construction process of the digital twin virtual space of this invention.

[0037] Figure 3 This is a schematic diagram of the communication between the physical space and the virtual space of the digital twin system of this invention.

[0038] Figure 4 This is a schematic diagram of the detection device of the present invention.

[0039] Figure 5 This is a flowchart of the process for acquiring images of the milled surface when the workpiece is located on the production line.

[0040] Figure 6 This is a flowchart illustrating the construction of the tool monitoring machine learning model and the tool monitoring data processing of this invention. Detailed Implementation

[0041] Specific implementation method one: Combining Figure 4 This embodiment describes a digital twin-driven tool wear detection device, which includes a workpiece conveying mechanism 1, a depth camera bracket 3, a depth camera 4, a CCD camera 5, an end effector rotation mechanism 6, an X-axis motion mechanism 8, two Y-axis motion mechanisms 7, and two Z-axis motion mechanisms 9. The depth camera bracket 3 is mounted on the frame of the conveying mechanism 1, and the depth camera 4 is mounted on the depth camera bracket 3 above the workpiece conveying mechanism 1. The two Z-axis motion mechanisms 9 are symmetrically mounted on the frames on both sides of the workpiece conveying mechanism 1. Each Z-axis motion mechanism 9 has a Y-axis motion mechanism 7 slidably mounted on it, and the two Y-axis motion mechanisms 7 are arranged vertically relative to each other. The X-axis motion mechanism 8 is slidably mounted on the two Y-axis motion mechanisms 7. The CCD camera 5 is mounted on the X-axis motion mechanism 8 through the end effector rotation mechanism 6, and the workpiece 2 is placed on the conveyor belt of the workpiece conveying mechanism 1.

[0042] The depth camera 4 is fixed at the top of the depth camera bracket 3, and its projection position on the workpiece conveyor mechanism 1 is exactly at the center of the workpiece conveyor belt; the three-axis linkage mechanism is fixed on the sliding mechanism on both sides of the workpiece conveyor mechanism 1, which can realize three-axis movement of XYZ on the conveyor belt; the CCD camera 5 is located at the end of the rotating shaft at the end of the three-axis linkage mechanism; the servo motor controlled by the digital twin system drives the movement and rotation of all devices.

[0043] Specific Implementation Method Two: Combining Figures 1-6 This embodiment describes a digital twin-driven tool wear monitoring method, which includes the following steps:

[0044] Step 1: Construct a virtual space for the CNC machine tool milling process and perform real-time milling simulation;

[0045] Step 2: Establish a connection between the physical space and the virtual space, and obtain multi-data from the physical space;

[0046] Step 3: Determine the position of the workpiece milling surface based on the acquired data to obtain the viewpoint during image acquisition;

[0047] Step 4: The digital twin system controls the three-axis motion mechanism to drive the CCD camera to acquire images of the milled surface of the workpiece at the viewpoint; the three-axis motion mechanism consists of X-axis motion mechanism 8, Y-axis motion mechanism 7, and Z-axis motion mechanism 9;

[0048] Step 5: Construct a tool monitoring model based on SVM, and build a dataset to train and test the model, so that the model has high recognition accuracy and generalization ability;

[0049] Step 6: Collect image data in real time, input it into the model to obtain the tool wear value, and issue a warning to replace the tool when the monitored tool wear value exceeds the tool wear threshold.

[0050] Specific implementation method three: Combining Figure 2 This embodiment describes a digital twin-driven tool wear monitoring method. Step one, virtual space construction, includes:

[0051] 3D Modeling: In industrial modeling software, a 3D model in virtual space is created based on the actual CNC machine tool and testing equipment, and the model is optimized in 3D modeling software; based on the model drawings provided by the equipment manufacturer and the actual measured dimensions of the equipment, a 3D model is drawn in SolidWorks at a 1:1 scale and saved as an STL file; the drawn 3D model is imported into 3ds Max for rendering, and the model is optimized by reducing the number of polygons in the software; finally, the rendered model is saved as an FBX file.

[0052] Scene setup: Import the optimized model into 3D visualization software to build the scene; import the rendered model into Unity3D software, set the model with the same materials as the physical space; for models with low requirements for detail rendering such as auxiliary facilities, floors, and walls, use textures and GameObject components in Unity3D to create them; design an aesthetically pleasing UGUI interactive interface; and use C# language to write program scripts to implement system functions.

[0053] Behavioral logic model construction: After completing the scene construction, specific behaviors and logical relationships are assigned to the virtual models in the virtual space. Based on real-time data collected from external sensors and machine tool communication protocols, variables bound to the associated models, and preset behavioral actions, multi-data-driven CNC machine tools perform milling simulations, constructing parent-child relationships between model tree structures to improve the construction of the machine tool virtual space. The SetParent function is used to establish parent-child relationships between related models, such as the parent-child relationship between a three-axis linkage mechanism and a camera. The motion of the virtual model includes two categories: translation and rotation. The Translate and Rotate functions are used to implement the translation of the transmission mechanism and the rotation of the spindle. By defining rules such as the model's translation limits and rotation speed adjustment, the construction of the machine tool digital twin space is further improved.

[0054] Human-Computer Interaction: A system interface is developed using 3D visualization software to achieve 3D visual monitoring of tool wear status. The software interface is developed in Unity. Users enter their account and password to access the main system interface, and move the mouse to click buttons to activate corresponding functions. In scene roaming mode, the "W", "A", "S", and "D" keyboard buttons control the movement of the screen within the window, the mouse wheel controls zooming, and the right mouse button controls rotation. The interactive interface displays the status information of the machine tool, cutting tool, and workpiece, as well as milling warning information. Other components and connection methods are the same as in Specific Implementation Method Two.

[0055] Specific implementation method four: Combination Figure 3 This embodiment describes a digital twin-driven tool wear monitoring method. Step two, establishing the connection between the physical and virtual spaces, includes:

[0056] Registering an OPC UA server in the CNC system is fundamental for data transmission at the application layer and a prerequisite for remote access to applications. The OPC UA server standardizes data from different devices and systems, enabling data interoperability and sharing. As a data communication intermediary, the OPC UA server standardizes data from different devices and systems, ensuring data interoperability and sharing. This standardized data format and protocol guarantee the reliability and security of data transmission and reduce the difficulty of system integration and maintenance.

[0057] A TCP / IP network layer communication connection is established between the physical space and the virtual space through the Socket interface; the API is used to establish a connection between the CNC kernel and the application. The API consists of many COM objects, which can export functions that communicate with the CNC kernel. All CNC variables can be accessed through the API by connecting to any interface based on the OPCUA transport protocol.

[0058] The tool detection device establishes TCP / IP communication with the server through a host computer interface to control the motion device and collect motion data.

[0059] After the machine tool and testing device in the physical space and virtual space establish communication, they obtain machine tool operating status information and workpiece milling surface image data by accessing the OPC UA server. The virtual space receives multi-data information from the physical space and stores it in a MySQL database. Other components and connection methods are the same as in Specific Implementation Method Two.

[0060] Specific Implementation Method Five: Combining Figures 5-6 This embodiment describes a digital twin-driven tool wear monitoring method. Static information includes machine tool attribute data, tool attribute data, workpiece attribute data, and process attribute data. The machine tool attribute data includes the equipment model and system model; the tool attribute data includes the tool name, tool material, and tool parameters; the workpiece attribute data includes the workpiece name, workpiece material, and workpiece dimensions; and the process attribute data includes the clamping method and machining parameters, including cutting speed, feed rate, and depth of cut.

[0061] The dynamic information is machine tool status data. Other components and connection methods are the same as in Specific Implementation Method Two.

[0062] Specific Implementation Method Six: Combination Figure 5 This embodiment describes a digital twin-driven tool wear monitoring method. It acquires scene point clouds using a depth camera, performs noise reduction and segmentation on the scene point clouds to obtain a target point cloud, performs point cloudification on the workpiece model to obtain a source point cloud, and registers the source point cloud and the target point cloud to obtain the workpiece's pose in the scene.

[0063] The position of the milled surface on the detection device is obtained by extracting the workpiece coordinate system and the position of the milled surface from the NC program of the CNC system and combining them with the workpiece pose.

[0064] The position of the final milled surface is extracted from the NC program. Then, the pose of the workpiece on the conveyor belt is estimated based on the depth camera to obtain the viewpoint. The three-axis linkage mechanism drives the CCD camera to the designated position for image acquisition.

[0065] The digital twin system, through communication with the CNC system, extracts the workpiece coordinate system during the machining process, the position of the milled surface when the machining is completed, and the model of the workpiece after machining from the NC program;

[0066] A 3D point cloud of the physical scene is obtained using a depth camera. Then, a pass-through filter is used to extract the region of interest. Finally, a statistical filter is used to remove outliers from the extracted region.

[0067] The filtered scene point cloud is the region where the object to be pose estimated is located. The point cloud in this region still contains background clusters, object clusters, and large noise clusters. The segmentation algorithm combining RANSAC clustering and Euclidean clustering is applied to first estimate the planar background in the scene point cloud and remove it. Then, the remaining points are clustered with a threshold set according to Euclidean distance. The resulting clustered points are the target object point cloud.

[0068] The workpiece model is processed into a point cloud to obtain the source point cloud. The normal angle constraint and the number of neighborhood constraints are applied to extract key points from the scene target point cloud and the object template point cloud. Feature descriptions are then constructed on the extracted key points and coarse registration with random sampling consistency is performed. The normal consistency property is used to filter the corresponding points for further fine registration, so that the accurate result of the target object pose can be calculated iteratively.

[0069] The target pose estimation results are combined with the workpiece coordinate system and the position of the workpiece milled surface to obtain the precise position of the milled surface on the workpiece conveyor belt; then, the viewpoint of the acquired image is determined based on the average normal vector of the model at the milling position and the optical parameters of the CCD camera.

[0070] The Z-axis motion mechanism maintains the same speed as the workpiece conveyor belt, which is acquired by the PLC. The X-axis motion mechanism reaches the position of the workpiece milling surface relative to the X-axis of the workpiece conveyor belt, based on the pose estimation. The angle between the Y-axis and the end-rotation mechanism is determined according to the normal vector of the milling surface and optical parameters such as the depth of field of the CCD camera. This controls the position and orientation of the camera during acquisition, completing the data acquisition of the milling surface image. Other components and connection methods are the same as in Specific Implementation Method Five.

[0071] Specific implementation method seven: Combining Figure 6 This embodiment describes a digital twin-driven tool wear monitoring method, which determines the viewpoint for image acquisition based on the workpiece model, the position of the milled surface, and the optical parameters of the camera.

[0072] The digital twin system controls a three-axis motion mechanism to drive a CCD camera to acquire image data of the milled surface of the workpiece. Other components and connections are the same as in Specific Implementation Method Two.

[0073] Specific implementation method eight: Combination Figure 6This embodiment describes a digital twin-driven tool wear monitoring method. It constructs an SVM machine learning model, using a gray-level co-occurrence matrix to extract image contrast and second diagonal moments from acquired images. These two extracted features are input into the SVM model to predict tool wear status. A tool wear dataset is constructed, and the dataset is randomly divided into training and test sets for each wear status at a 3:1 ratio. Feature extraction and feature selection are performed on the training and test sets respectively to determine the feature vector set. After normalization, a grid search method is used to find all points within the range for parameter optimization, obtaining as many hyperparameter combinations as possible. This process is repeated three times through cross-validation to find the penalty factor and kernel parameter hyperparameter combination that maximizes the classification accuracy of the dataset. Finally, the tool monitoring model is obtained. Features are extracted from images acquired by a CCD camera and input into an established SVM machine learning model for tool wear monitoring. Images of the workpiece milling surface during each pass of ball end mills, flat end mills, and end mills are acquired by a CCD camera. Simultaneously, the wear of the cutter's flank face after each pass is measured and recorded using a microscope to establish a dataset for training and testing. The average tool wear and wear state for each pass are determined based on the actual conditions of the milling process.

[0074] Data collected from the detection device is input into the model to identify the tool wear state. Based on the model's output values, tool wear monitoring, visualization, and alarm processing are completed. The tool wear values ​​are uploaded to a MySQL database in real time and interfaced with a digital twin space built in Unity3D via SQL statements. Simultaneously, the physical space is mapped in real time in the virtual space, jointly achieving a visualized mapping of tool wear monitoring.

[0075] In summary, this invention provides a tool monitoring method and detection device driven by digital twins, which can manage tool status online based on historical and real-time machining data, and promptly transmit tool change warning information, thereby improving tool utilization and machine tool processing efficiency, and realizing three-dimensional visual monitoring of tool wear status. Its composition and connection method are the same as in specific embodiment two.

[0076] Specific Implementation Method Nine: Combining Figure 1 and Figure 3 This embodiment describes a digital twin-driven tool wear monitoring method. It obtains the tool wear value and determines whether to issue a warning based on a threshold to remind the user to replace the tool. Other components and connections are the same as in Embodiment Two.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent substitutions, and improvements made to the above embodiments without departing from the scope of the present invention, based on the technical essence of the present invention and within the spirit and principles of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A digital twin-driven method for monitoring tool wear, characterized in that: It is achieved using a tool wear detection device, which includes a workpiece conveying mechanism (1), a depth camera bracket (3), a depth camera (4), a CCD camera (5), an end-rotating mechanism (6), an X-axis motion mechanism (8), two Y-axis motion mechanisms (7), and two Z-axis motion mechanisms (9). The depth camera bracket (3) is mounted on the frame of the conveying mechanism (1), the depth camera (4) is mounted on the depth camera bracket (3) above the workpiece conveying mechanism (1), the two Z-axis motion mechanisms (9) are symmetrically mounted on the frames on both sides of the workpiece conveying mechanism (1), each Z-axis motion mechanism (9) has a Y-axis motion mechanism (7) slidably mounted on it, the two Y-axis motion mechanisms (7) are vertically arranged relative to each other, the X-axis motion mechanism (8) is slidably mounted on the two Y-axis motion mechanisms (7), the CCD camera (5) is mounted on the X-axis motion mechanism (8) through the end-rotating mechanism (6), and the workpiece (2) is set on the conveyor belt of the workpiece conveying mechanism (1). The method includes the following steps: Step 1: Construct a virtual space for the CNC machine tool milling process and perform real-time milling simulation; Step 2: Establish a connection between the physical space and the virtual space, and obtain multi-dimensional data from the physical space. This multi-dimensional data includes both static and dynamic information from the physical space. Step 3: Obtain scene point cloud from depth camera, perform noise reduction and segmentation on scene point cloud to obtain target point cloud, perform point cloudification on workpiece model to obtain source point cloud, and register source point cloud and target point cloud to obtain workpiece pose in scene. The position of the milled surface on the detection device is obtained by combining the workpiece coordinate system and the position of the milled surface with the workpiece pose from the NC program of the CNC system. The viewpoint during image acquisition is solved based on the workpiece model, the position of the milled surface and the optical parameters of the camera. Step 4: The digital twin system controls the three-axis motion mechanism to drive the CCD camera to acquire images of the milled surface of the workpiece at the viewpoint; Step 5: Construct a tool monitoring model based on SVM, and build a dataset to train and test the model, so that the model has high recognition accuracy and generalization ability; Step 6: Collect image data in real time, input it into the model to obtain the tool wear value, and issue a warning to replace the tool when the monitored tool wear value exceeds the tool wear threshold.

2. The method for monitoring tool wear driven by digital twin according to claim 1, characterized in that: Step one, virtual space construction, includes: 3D modeling: In industrial modeling software, a 3D model in virtual space is created based on the actual CNC machine tool and testing equipment, and the model is optimized in 3D modeling software; Scene setup: Import the optimized model into 3D visualization software to build the scene; Behavioral logic model construction: After the scene is built, the specific behaviors and logical relationships of the virtual model in the virtual space are given; based on the data collected in real time by external sensors and machine tool communication protocols, the variables bound to the associated model and the preset behavioral actions, the CNC machine tool driven by multiple data performs milling simulation, builds the parent-child relationship between the model tree structure, and improves the construction of the machine tool virtual space; Human-computer interaction: The system interface is developed in 3D visualization software to realize 3D visualization monitoring of tool wear status.

3. The method for monitoring tool wear driven by digital twin according to claim 1, characterized in that: Step two, which involves establishing a connection between the physical and virtual spaces, includes: Register an OPC UA server in the CNC system; A TCP / IP network layer communication connection is established between the physical space and the virtual space through the Socket interface; After establishing communication between the machine tools and testing devices in the physical space and the virtual space, the machine tool operating status information and workpiece milling surface image data are obtained by accessing the OPC UA server. The virtual space receives multi-data information from the physical space and stores it in the MySQL database.

4. The method for monitoring tool wear driven by digital twin according to claim 1, characterized in that: Static information includes machine tool attribute data, tool attribute data, workpiece attribute data, and process attribute data. Among them, machine tool attribute data includes equipment model and system model; tool attribute data includes tool name, tool material, and tool parameters; workpiece attribute data includes workpiece name, workpiece material, and workpiece dimensions; and process attribute data includes clamping method and machining parameters. Machining parameters include cutting speed, feed rate, and depth of cut. Dynamic information is machine tool status data.

5. The method for monitoring tool wear driven by digital twin according to claim 1, characterized in that: An SVM machine learning model is constructed. The gray-level co-occurrence matrix is ​​used to extract image contrast and second diagonal moments from the acquired images. These two extracted features are input into the SVM model to predict tool wear status. A tool wear dataset is constructed, and the dataset is randomly divided into training and test sets for each wear status at a 3:1 ratio. Feature extraction and feature selection are performed on the training and test sets respectively to determine the feature vector set. After normalization, a grid search method is used to find all points within the range for parameter optimization, obtaining as many hyperparameter combinations as possible. This process is repeated three times through cross-validation to find the penalty factor and kernel parameter hyperparameter combination that maximizes the classification accuracy of the dataset. Finally, the tool monitoring model is obtained.

6. The method for monitoring tool wear driven by digital twin according to claim 1, characterized in that: The wear value of the cutting tool is obtained, and a warning is issued based on the threshold to remind the user to change the tool.

Images