Hard alloy sorting system and method

Abstract

The invention discloses a hard alloy sorting system and a hard alloy sorting method. The hard alloy sorting system comprises a multi-joint robot body, the conveying mechanism is used for conveying hard alloy workpieces to be sorted; the multi-modal detection device is used for detecting the hard alloy workpiece and comprises a hyperspectral imaging unit used for acquiring surface spectral information, a three-dimensional contour measurement unit used for acquiring three-dimensional geometric information and an eddy current detection unit used for acquiring surface conductivity information; the control unit is used for receiving and fusing the surface spectrum information, the three-dimensional geometric information and the surface conductivity information, identifying the workpiece category of the hard alloy workpiece and generating a corresponding sorting control instruction; and the end effector is used for grabbing and sorting the hard alloy workpieces of the corresponding workpiece types according to the sorting control instruction. According to the method, accurate and efficient recognition and sorting of the hard alloy workpieces are achieved by obtaining and fusing the multi-mode heterogeneous information, and the recognition precision and sorting efficiency of the complex mixed hard alloy are improved.

Description

Technical Field

[0001] This invention relates to the field of waste resource utilization technology, and in particular to a cemented carbide sorting system and method. Background Technology

[0002] Tungsten carbide (WC) is a high-performance composite material with tungsten carbide (WC) as the hard phase and metals such as cobalt (Co) or nickel (Ni) as the binder phase. Due to its extremely high hardness, wear resistance, and good corrosion resistance, it is widely used in the manufacture of various cutting tools, such as drills, end mills, and CNC inserts. To improve cutting performance and service life, these tools are often coated with one or more wear-resistant coatings such as TiN, TiCN, or Al2O3 using chemical vapor deposition (CVD) or physical vapor deposition (PVD) processes. In the field of resource recycling, the recycling and reuse of waste cemented carbide has significant economic and environmental value.

[0003] However, coated and uncoated cemented carbides differ fundamentally in chemical composition, and their subsequent metallurgical recycling processes (such as coating removal and dissolution extraction parameters) are also drastically different. Mixing the two not only reduces metal recovery and product purity but also increases process complexity and processing costs. Therefore, precise sorting of the coating state during the recycling pretreatment stage is a crucial step in improving recycling efficiency and product quality.

[0004] Currently, automated sorting equipment based on traditional RGB cameras has emerged in the market. However, these devices can only identify the color and macroscopic shape of the workpiece. In actual recycling scenarios, the surface conditions of waste cemented carbide workpieces are extremely complex: the coating may be severely worn or partially peeled off due to long-term use; the surface may be covered with oil, cutting fluid, or other contaminants; the metal substrate or coating itself may produce strong specular reflections. These factors can severely interfere with or even completely obscure feature extraction based on traditional machine vision, leading to a sharp decline in the accuracy and robustness of the recognition algorithm. This makes it impossible to meet the reliability requirements of industrial-grade sorting and reduces the recognition accuracy and sorting efficiency for complex mixed cemented carbide (such as drill bits and cutting tools). Summary of the Invention

[0005] In view of this, it is necessary to provide a cemented carbide sorting system and method to solve the technical problem that the existing cemented carbide sorting system has low identification accuracy and sorting efficiency for complex mixed cemented carbide.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides a cemented carbide sorting system, comprising: Multi-jointed robot body; A conveying mechanism is located within the working area of ​​the multi-joint robot body and is used to convey cemented carbide workpieces to be sorted. A multimodal detection device, mounted on a multi-joint robot body, is used to detect the cemented carbide workpiece. It includes at least a hyperspectral imaging unit for acquiring surface spectral information, a three-dimensional contour measurement unit for acquiring three-dimensional geometric information, and an eddy current detection unit for acquiring surface electrical conductivity information. The control unit is communicatively connected to the multimodal detection device and the multi-joint robot body, and is used to receive and fuse the surface spectral information, three-dimensional geometric information and surface electrical conductivity information, identify the workpiece category of the cemented carbide workpiece, and generate corresponding sorting control instructions; An end effector, installed at the operating end of the multi-joint robot body, is used to grasp and sort cemented carbide workpieces of the corresponding workpiece category according to the sorting control command.

[0007] In one possible implementation, the hyperspectral imaging unit includes a hyperspectral camera, the response wavelength of which includes the visible light band and the near-infrared band, for acquiring the characteristic spectral reflectance curve of the coating on the surface of the cemented carbide workpiece, thereby obtaining the surface spectral information.

[0008] In one possible implementation, the three-dimensional contour measurement unit is a three-dimensional structured light profiler, used to acquire three-dimensional point cloud data of the cemented carbide workpiece, and to obtain the three-dimensional geometric information by extracting features from the three-dimensional point cloud data.

[0009] In one possible implementation, the eddy current detection unit is integrated into the clamping surface of the end effector, and the eddy current detection unit includes an eddy current detection probe for detecting the eddy current response signal on the surface of the cemented carbide workpiece to obtain surface conductivity information.

[0010] In one possible implementation, the control unit includes a workpiece recognition module and a motion control module; The workpiece recognition module is used to fuse surface spectral information, three-dimensional geometric information and surface electrical conductivity information to identify the workpiece category of the cemented carbide workpiece and obtain the recognition result; A motion control module is used to generate the sorting control command based on the recognition result.

[0011] In one possible implementation, the workpiece recognition module includes a pre-trained deep learning model.

[0012] In one possible implementation, the workpiece category includes at least coated cemented carbide, uncoated cemented carbide, and CNC inserts.

[0013] In a second aspect, the present invention also provides a cemented carbide sorting method, applied to the cemented carbide sorting method described in any one of the first aspects, comprising: The cemented carbide workpieces to be sorted are continuously conveyed by a conveying mechanism; When the cemented carbide workpiece enters the inspection station, the surface spectral information, three-dimensional geometric information and surface electrical conductivity information of the cemented carbide workpiece are acquired simultaneously through the multi-modal inspection device. By fusing surface spectral information, three-dimensional geometric information, and surface electrical conductivity information, the control unit identifies the workpiece category of the cemented carbide workpiece and generates sorting control instructions. The sorting control command is executed to grasp and sort the cemented carbide workpieces of the corresponding workpiece category.

[0014] In one possible implementation, the workpiece category includes at least coated cemented carbide, uncoated cemented carbide, and CNC cutting tools; the step of identifying the workpiece category of the cemented carbide workpiece by fusing surface spectral information, three-dimensional geometric information, and surface electrical conductivity information by the control unit includes: When the surface spectral information indicates the presence of a target coating and the surface conductivity information indicates insulation, the corresponding cemented carbide workpiece is classified as coated cemented carbide. When the surface spectral information does not indicate the presence of a target coating and the surface conductivity information indicates conductivity, the corresponding cemented carbide workpiece is classified as an uncoated cemented carbide. When the three-dimensional geometric information characterizes the workpiece as having a regular sheet-like structure, the corresponding cemented carbide workpiece is classified as a CNC cutting tool.

[0015] The beneficial effects of this invention are: The cemented carbide sorting system provided by this invention includes a multi-joint robot body capable of efficiently handling cemented carbide workpieces of various shapes, improving the flexibility of sorting execution; a conveying mechanism, disposed within the working area of ​​the multi-joint robot body, for conveying cemented carbide workpieces to be sorted; and a multi-modal detection device, mounted on the multi-joint robot body, for detecting cemented carbide workpieces, including at least a hyperspectral imaging unit for acquiring surface spectral information, a three-dimensional contour measurement unit for acquiring three-dimensional geometric information, and an eddy current detection unit for acquiring surface conductivity information, thereby acquiring three types of heterogeneous information: hyperspectral imaging based on chemical properties, three-dimensional contour measurement based on geometric properties, and eddy current detection based on physical conductivity properties, thus enabling identification of cemented carbide workpieces from multiple dimensions. This invention effectively overcomes the problem of drastic drop in recognition rate of traditional vision systems under complex surface conditions such as coating wear, contamination, and glare. The control unit, communicating with the multimodal detection device and the multi-joint robot body, receives and fuses surface spectral information, three-dimensional geometric information, and surface conductivity information to identify the workpiece category of cemented carbide workpieces and generate corresponding sorting control commands. This achieves complementarity and mutual verification between multimodal heterogeneous information, effectively overcoming interference caused by complex working conditions such as wear, contamination, and glare on the surface of cemented carbide workpieces, and improving the accuracy and robustness of the system in identifying mixed cemented carbide workpieces in real industrial environments. The end effector, installed at the operating end of the multi-joint robot body, is used to grasp and sort cemented carbide workpieces of the corresponding workpiece category according to the sorting control commands. By acquiring and fusing multimodal heterogeneous information, this invention achieves accurate and efficient identification and sorting of cemented carbide workpieces, improving the identification accuracy and sorting efficiency for complex mixed cemented carbide workpieces. Attached Figure Description

[0016] Figure 1 A schematic diagram of an embodiment of the cemented carbide sorting system provided by the present invention; Figure 2 This is a schematic diagram of another embodiment of the cemented carbide sorting system provided by the present invention; Figure 3 This is a schematic flowchart of an embodiment of the cemented carbide sorting method provided by the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] In the description of the embodiments of the present invention, unless otherwise stated, "a plurality of" means two or more.

[0019] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] This invention provides a cemented carbide sorting system and method, which will be described below.

[0022] Figure 1 A schematic diagram of an embodiment of the cemented carbide sorting system provided by the present invention is shown below. Figure 1 As shown, the cemented carbide sorting system includes: Multi-joint robot body 10; The conveying mechanism 20 is located within the working area of ​​the multi-joint robot body and is used to convey the cemented carbide workpieces to be sorted. The multimodal detection device 30 is mounted on the multi-joint robot body 10 and is used to detect the cemented carbide workpiece. It includes at least a hyperspectral imaging unit 31 for acquiring surface spectral information, a three-dimensional contour measurement unit 32 for acquiring three-dimensional geometric information, and an eddy current detection unit 33 for acquiring surface electrical conductivity information. The control unit 40 is communicatively connected to the multimodal detection device 30 and the multi-joint robot body 10, and is used to receive and fuse the surface spectral information, three-dimensional geometric information and surface electrical conductivity information, identify the workpiece category of the cemented carbide workpiece, and generate corresponding sorting control instructions. An end effector 50 is installed at the operating end of the multi-joint robot body 10 and is used to grasp and sort cemented carbide workpieces of the corresponding workpiece category according to the sorting control command.

[0023] The carbide workpieces to be sorted include drill bits and CNC inserts. Drill bits and CNC inserts can be distinguished according to their geometry. Drill bits are divided into coated drill bits and uncoated drill bits.

[0024] The multi-jointed robot body 10 can be a multi-degree-of-freedom humanoid robot.

[0025] The inventors discovered that currently, the mechanisms for sorting cemented carbide workpieces are usually sorting machines based on fixed robotic arms with simple grippers or pneumatic suction cups, and their operation modes are often relatively simple. Therefore, this embodiment adopts a robot-based sorting system, which can efficiently process cemented carbide workpieces of various shapes and improve the flexibility of sorting execution.

[0026] The conveying mechanism 20 can be a conveyor belt running at a constant speed, located in the working area directly in front of the multi-joint robot body 10. The conveying mechanism 20 is used to transport the cemented carbide workpieces to be sorted, i.e., mixed cemented carbide workpieces (not shown in the figure), to the inspection station in front of the multi-joint robot body 10. The speed of the conveyor belt can be adjusted according to the sorting speed.

[0027] A multimodal inspection device 30, mounted on a multi-joint robot body, is used to simultaneously inspect cemented carbide workpieces located at the inspection station. The multimodal inspection device 30 includes at least a hyperspectral imaging unit 31 for acquiring surface spectral information, a three-dimensional contour measurement unit 32 for acquiring three-dimensional geometric information, and an eddy current detection unit 33 for acquiring surface conductivity information. The hyperspectral imaging unit 31 can be a hyperspectral camera, with its lens facing the cemented carbide workpiece on the conveyor belt, for scanning the surface of the cemented carbide workpiece and acquiring surface spectral information.

[0028] The 3D contour measurement unit 32 is installed adjacent to the hyperspectral imaging unit. Encoded light spots are projected onto the cemented carbide workpiece, and the deformed light spots are captured by a camera. The 3D point cloud data of the cemented carbide workpiece is calculated in real time, thereby obtaining high-precision 3D geometric information, including the size and volume of the workpiece, as well as key features (such as the helical groove angle, tip shape, flatness and corners of the drill bit).

[0029] The eddy current detection unit 33, which can be an eddy current detection probe, can be encapsulated and integrated into the fingertip or the inner side of the gripping surface of the end effector 50. Before or during the robot's gripping operation, the probe can be close to the surface of the cemented carbide workpiece. By emitting a high-frequency electromagnetic field and detecting changes in the eddy current field induced on the surface of the cemented carbide workpiece, surface conductivity information is generated. Since the cemented carbide substrate (WC-Co) has a certain degree of conductivity, while certain coatings (such as TiN, Al2O3) are insulators, this surface conductivity information can directly reflect the conductivity of the area below the probe.

[0030] Understandably, in this embodiment, since the system acquires three types of heterogeneous information—hyperspectral imaging based on chemical properties, three-dimensional contour measurement based on geometric properties, and eddy current detection based on physical conductivity properties—it can identify cemented carbide workpieces from multiple dimensions. This effectively overcomes the problem of a sharp drop in recognition rate of traditional vision under complex surface conditions such as coating wear, contamination, and reflection. It is beneficial to improve the ability to distinguish surface conditions under conditions of consistent geometry, thereby achieving reliable and accurate differentiation between "coated drill bits" and "uncoated drill bits." It can penetrate surface interference and accurately identify the material surface coating state of cemented carbide workpieces.

[0031] In this embodiment, the control unit 40 can be an industrial computer or a high-performance embedded controller, which is connected to the multimodal detection device 30 and the multi-joint robot body 10 via a cable. The control unit 40 internally runs specialized data fusion and classification software. This software receives and synchronizes three types of multimodal information with timestamp alignment in real time: surface spectral information, three-dimensional geometric information, and surface conductivity information. It then performs fusion analysis on the three types of information to determine the workpiece category. For example, the fusion analysis logic is as follows: if the three-dimensional geometric information represents a regular polygonal sheet structure, then the cemented carbide workpiece is determined to be a CNC cutting tool; if the three-dimensional geometric information represents a drill bit with spiral grooves and a tapered head, then the cemented carbide workpiece is determined to be a drill bit. When the control unit identifies a "high-confidence TiN coating feature" from the surface spectral information, and the surface conductivity information indicates that the area is "insulated," then the surface of the cemented carbide workpiece is determined to have a coating; if the surface spectral information shows "exposed substrate features" and the surface conductivity information indicates "conductivity," then it is determined to be uncoated. For cemented carbide workpieces with clearly regular, thin sheet shapes, even if the coating condition is complex to determine, they are preferentially classified as "CNC cutting tools" to identify the workpiece category. Then, the control unit 40 plans the robot's grasping path based on the workpiece category (e.g., "coated drill bit") and the precise workpiece coordinates provided by the 3D profilometer, and generates sorting control commands containing target position, posture, and clamping force parameters. Understandably, by fusing multimodal information through the control unit 40, the complementary and mutually corroborating information between the multimodal heterogeneous information is achieved. This effectively overcomes the interference caused by complex working conditions such as wear, contamination, and reflection on the surface of cemented carbide workpieces, improving the accuracy and system robustness of identifying mixed cemented carbide workpieces in real industrial environments.

[0032] The end effector 50 can be an adaptive bionic gripper with force-sensing feedback, mounted on the last joint (operating end) of the multi-joint robot body 10 via a flange. This gripper has multiple phalanges that can move independently or collaboratively, with pressure sensors integrated into the fingertips. Upon receiving instructions from the control unit 40, it automatically adjusts the finger angle, envelope posture, and gripping force based on the type of workpiece to be grasped and its corresponding three-dimensional geometry. This enables precise and efficient identification and sorting of cemented carbide workpieces, improving the identification accuracy and sorting efficiency for complex mixed cemented carbide materials.

[0033] In summary, the cemented carbide sorting system provided by this invention includes a multi-joint robot body capable of efficiently handling cemented carbide workpieces of various shapes, improving the flexibility of sorting execution; a conveying mechanism, disposed within the working area of ​​the multi-joint robot body, for conveying cemented carbide workpieces to be sorted; and a multi-modal detection device, mounted on the multi-joint robot body, for detecting cemented carbide workpieces, including at least a hyperspectral imaging unit for acquiring surface spectral information, a three-dimensional contour measurement unit for acquiring three-dimensional geometric information, and an eddy current detection unit for acquiring surface conductivity information. This allows for the acquisition of three types of heterogeneous information: hyperspectral imaging based on chemical properties, three-dimensional contour measurement based on geometric properties, and eddy current detection based on physical conductivity properties. This enables the identification of cemented carbide workpieces from multiple dimensions, effectively overcoming the limitations of traditional vision methods in areas such as coating wear, etc. The system addresses the issue of drastically reduced recognition rates under complex surface conditions such as contamination and reflection. The control unit, communicating with the multimodal detection device and the multi-joint robot body, receives and fuses surface spectral information, three-dimensional geometric information, and surface conductivity information to identify the workpiece category of cemented carbide workpieces and generate corresponding sorting control commands. This achieves complementarity and mutual verification between multimodal heterogeneous information, effectively overcoming interference from complex working conditions such as wear, contamination, and reflection on the surface of cemented carbide workpieces, improving the accuracy and robustness of identifying mixed cemented carbide workpieces in real industrial environments. The end effector, installed at the operating end of the multi-joint robot body, executes the gripping and sorting of cemented carbide workpieces of the corresponding workpiece category according to the sorting control commands, achieving precise and efficient sorting of cemented carbide workpieces and improving the recognition accuracy and sorting efficiency of complex mixed cemented carbide workpieces.

[0034] In some embodiments of the present invention, the hyperspectral imaging unit includes a hyperspectral camera, the response wavelength of which includes the visible light band and the near-infrared band, for acquiring the characteristic spectral reflectance curve of the coating on the surface of the cemented carbide workpiece, thereby obtaining the surface spectral information.

[0035] The hyperspectral camera operates in a band that covers the visible light (e.g., 400-780nm) to near-infrared (e.g., 780-1700nm) range, and can capture the surface spectral characteristics of cemented carbide workpieces in the visible and near-infrared bands, i.e., surface spectral information, which is used to identify the coating state of cemented carbide workpieces.

[0036] Specifically, a hyperspectral camera acquires continuous spectral data for each pixel of a cemented carbide workpiece, forming surface spectral information. This surface spectral information can effectively distinguish the unique spectral characteristics of coating materials such as TiN and Al2O3 from those of the WC-Co substrate. Because coating materials (such as TiN and Al2O3) and the exposed cemented carbide substrate have unique spectral reflectance curves at different wavelengths, accurate identification of the coating condition can be achieved, overcoming the problem of low recognition rate of traditional RGB cameras under coating wear or reflective conditions.

[0037] In some embodiments of the present invention, the three-dimensional contour measurement unit is a three-dimensional structured light profiler, used to acquire three-dimensional point cloud data of cemented carbide workpieces, and to obtain the three-dimensional geometric information by extracting features from the three-dimensional point cloud data.

[0038] Specifically, the 3D contour measurement unit is used to acquire 3D point cloud data of cemented carbide workpieces, accurately measuring the workpiece's contour dimensions and geometry. By performing point cloud registration and feature extraction on the 3D point cloud data, it is possible to distinguish the helical groove structure of the drill bit, the tip angle, and the specific planar geometric features of the CNC insert. Understandably, the high-density point cloud provided by the 3D structured light profilometer can accurately capture and quantify complex and minute features such as drill bit helical grooves and insert cutting edges.

[0039] It is worth noting that the three-dimensional geometric information provides the workpiece with six degrees of freedom (position and orientation) information in space, ensuring a high success rate of grasping and operational safety.

[0040] In some embodiments of the present invention, the eddy current detection unit is integrated on the clamping surface of the end effector, and the eddy current detection unit includes an eddy current detection probe for detecting the eddy current response signal on the surface of the cemented carbide workpiece to obtain surface conductivity information.

[0041] Specifically, the eddy current detection unit includes an eddy current detection probe integrated into the gripping surface of an end effector, such as the fingertip of a robot gripper, used to detect the electrical conductivity characteristics of the surface of a cemented carbide workpiece. The coating is typically a non-conductive layer, while the cemented carbide substrate has a certain degree of conductivity. By observing the difference in eddy current response signals, the presence of a coating can be further verified, and this can be corroborated with surface spectral information. The detection process is as follows: the eddy current detection probe emits a high-frequency alternating electromagnetic field onto the surface of the cemented carbide workpiece. If the workpiece surface is a conductive cemented carbide substrate, eddy currents will be induced on the surface. These eddy currents generate a reverse magnetic field, causing a significant change in the impedance of the probe coil, producing a "strong response" signal. If the workpiece surface is completely covered by a non-conductive coating, almost no eddy currents can be formed, and the signal change detected by the probe is weak, i.e., a "weak response" signal. The eddy current response signal is converted into surface conductivity information that can be used for qualitative judgment. The eddy current detection unit is integrated into the clamping surface, avoiding fluctuations in the detection signal caused by changes in workpiece posture, vibration, or air gap, ensuring the consistency of surface conductivity information acquisition. At the same time, the end effector can automatically adapt to the shape of the workpiece during gripping, and the probe integrated into the clamping surface also moves to a suitable detection position (such as the cylindrical surface of a drill bit or the plane of a cutting tool), solving the detection blind zone problem that may exist in fixed-mount probes and improving the detection coverage of irregular and complex-shaped workpieces.

[0042] In some embodiments of the present invention, such as Figure 2 The diagram shown is a structural schematic of another embodiment of the cemented carbide sorting system; the control unit 40 includes a workpiece identification module 41 and a motion control module 42. The workpiece identification module 41 is used to fuse surface spectral information, three-dimensional geometric information and surface electrical conductivity information to identify the workpiece category of the cemented carbide workpiece and obtain the identification result; The motion control module 42 is used to generate the sorting control command based on the recognition result.

[0043] Specifically, the workpiece recognition module 41 may include a machine vision submodule and a material analysis submodule. The machine vision submodule is used to analyze the three-dimensional geometric information to obtain a first analysis result, which is divided into whether the cemented carbide workpiece is a drill bit or a CNC cutting tool. Then, the first analysis result, surface spectral information and surface conductivity information are fused. The fusion process may be as follows: fuse the surface spectral information and surface conductivity information to obtain a second analysis result of whether the cemented carbide workpiece has a coating. Then, the second analysis result and the first analysis result are fused to determine the workpiece category, which is the recognition result of the corresponding cemented carbide workpiece.

[0044] Based on the recognition results, the motion control module 42 can also perform motion planning by combining the corresponding three-dimensional geometric information, calculating the optimal collision-free path for the multi-joint robot body 10 from its current position to the gripping point and then to the target collection container. Simultaneously, it generates adaptive control parameters based on the workpiece type and three-dimensional geometric features (e.g., envelope gripping for slender drill bits, light pinch gripping for thin cutting blades, and setting a safe gripping force threshold), encapsulating these paths and parameters into specific sorting control instructions. These instructions are transmitted in real-time via the communication network to the drive system of the multi-joint robot body 10 and the force controller of the end effector 50, driving the robot to complete precise and compliant gripping-movement-placement actions.

[0045] It is worth noting that the workpiece identification module is also used to compare and verify surface conductivity information with surface spectral information to determine the surface coating state of the cemented carbide workpiece. The comparison and verification method is as follows: if the surface spectral information indicates the presence of a target coating and the surface conductivity information indicates insulation, then a coating is determined to exist on the workpiece surface; if the surface spectral information does not indicate the presence of a target coating and the surface conductivity information indicates conductivity, then the workpiece surface is determined to be an exposed cemented carbide substrate. This achieves the ability to directly identify the coating material composition using the surface spectral information and the ability to directly detect the coating's insulating physical properties using the surface conductivity information, cross-validating each other to improve the robustness of coating state determination under conditions of coating wear, contamination, or reflectivity.

[0046] In some embodiments of the present invention, the workpiece recognition module includes a pre-trained deep learning model.

[0047] Specifically, the pre-trained deep learning model can be a multi-branch feature fusion neural network. This network has three independent feature extraction branches, each corresponding to one of three heterogeneous input data streams: Spectral feature branch: Receives surface spectral information from the hyperspectral imaging unit. This branch typically contains a one-dimensional convolutional layer (1D-CNN) or a fully connected layer to extract deep abstract features related to the coating chemical composition and matrix material (WC-Co) from the surface spectral information, i.e., the coating chemical feature vector.

[0048] Geometric Feature Branch: Receives 3D geometric information from the 3D contour measurement unit. This branch uses a point cloud processing network (such as PointNet or the improved PointNet++ structure) to operate on the disordered point cloud, learn and extract 3D shape feature vectors that can characterize the macroscopic category (such as drill bit, cutting tool) and detailed geometric structure (such as spiral groove, tool tip angle, flatness) of the workpiece.

[0049] Conductivity Feature Branch: Receives surface conductivity information (e.g., a set of numerical sequences or images characterizing the probe response intensity) from the eddy current detection unit. This branch can employ a one-dimensional convolution or a simple fully connected network to extract surface physical property feature vectors from the electrical signal that can distinguish between "conductive" and "insulating" states.

[0050] The feature vectors extracted from the three branches (coating chemical features, 3D shape features, and surface physical property features) are concatenated or fused through an attention fusion module to form a unified joint feature representation containing multi-dimensional information. This joint feature is then fed into the subsequent fully connected classification layer. During training, the model learns the complex, non-linear mapping relationship between these joint features and the final workpiece category (e.g., "coated drill bit," "uncoated drill bit," "CNC insert"). The decision logic is implicitly encoded in the model parameters: for example, by learning from a large number of samples, the model has mastered the rule that "when the 'coating chemical features' strongly indicate TiN and the 'surface physical property features' simultaneously indicate insulation, the output probability of the 'coated drill bit' category should be greatly increased." This fusion judgment is learned by the model itself based on data.

[0051] Model Training and Deployment: For the training phase: A large-scale, pre-labeled dataset can be used to train the model offline. This dataset covers cemented carbide workpiece samples with various wear states, contamination levels, and placement orientations. Each sample contains synchronously acquired surface spectral information, 3D geometric information, surface conductivity information, and a true class label confirmed by experts. The training process adjusts the network's weight parameters using the backpropagation algorithm to minimize the error between the model's predicted class and the true label.

[0052] Deployment Phase: The best-performing model, after training and validation, is deployed to the workpiece recognition module of control unit 40. During actual sorting, the three types of sensor data collected by the system in real time undergo the same preprocessing and are then input into the model. The model will directly output the category probability distribution of each workpiece within milliseconds, and the category with the highest probability is taken as the recognition result.

[0053] Understandably, in this embodiment, by fusing multimodal data through a pre-trained deep learning model, the ability to automatically extract and fuse the most essential features from multimodal data is achieved, thereby improving the recognition accuracy and decision consistency under complex conditions such as coating wear and contamination.

[0054] In some embodiments of the present invention, the workpiece category includes at least coated cemented carbide, uncoated cemented carbide, and CNC cutting tools.

[0055] Specifically, the workpiece categories in this embodiment include at least coated cemented carbide, uncoated cemented carbide, and CNC cutting tools, as shown in Table 1, which is a logical lookup table for workpiece category identification: Table 1. Logical Comparison Table for Workpiece Category Identification

[0056] In some embodiments of the present invention, the end effector is a force-controlled bionic gripper.

[0057] The end effector is an adaptive bionic gripper with force feedback functionality. For example, a bionic hand often has two or three "fingers" that can move independently or in coupling. Each finger contains multiple bionic phalanges, driven by built-in micromotors or pneumatic artificial muscles, capable of simulating the bending, opening, and lateral swinging movements of human fingers, thereby achieving various complex grasping envelopes. Furthermore, the bionic gripper integrates a multi-dimensional force / torque sensor and a tactile sensor array. The force / torque sensors are mounted on the wrist of the gripper or the base of each finger to measure and provide real-time feedback on the magnitude of the clamping force applied to the workpiece during grasping, as well as any lateral forces or torques that may be present. The tactile sensors can be distributed on the gripping surfaces inside the fingers to sense the pressure distribution and sliding tendency at the contact points with the workpiece.

[0058] Specifically, the force-controlled bionic gripper performs the grasping and sorting of cemented carbide workpieces of the corresponding workpiece categories. By utilizing force feedback and control, it ensures that the clamping force is always limited within the safe threshold, fundamentally eliminating the risk of chipping, micro-cracks, or coating peeling of the precision cutting edge of cemented carbide due to excessive clamping. This protects the cutting edge of the cemented carbide workpiece and improves sorting safety and efficiency.

[0059] like Figure 3 As shown, the present invention also provides a sorting method based on cemented carbide. Figure 3 This is a schematic flowchart of an embodiment of the cemented carbide-based sorting method provided by the present invention. The method includes: S301. The cemented carbide workpieces to be sorted are continuously conveyed through the conveying mechanism; S302. When the cemented carbide workpiece enters the inspection station, the surface spectral information, three-dimensional geometric information and surface electrical conductivity information of the cemented carbide workpiece are acquired simultaneously through the multi-modal detection device. S303. By fusing surface spectral information, three-dimensional geometric information and surface electrical conductivity information through the control unit, the workpiece category of the cemented carbide workpiece is identified and sorting control instructions are generated. S304. Execute the sorting control command to grasp and sort the cemented carbide workpieces of the corresponding workpiece category.

[0060] It should be noted that in the embodiments of the cemented carbide sorting method described above, please refer to the corresponding description in the cemented carbide sorting system above, and the beneficial effects that can be achieved should also be referred to the beneficial effects described above, which will not be repeated here.

[0061] For embodiments based on the cemented carbide sorting method, please refer to the steps performed by the corresponding processing unit in the above text. For other embodiments based on the cemented carbide sorting method that are not described, please refer to the corresponding content mentioned above, which will not be repeated here.

[0062] In some embodiments of the present invention, the workpiece category includes at least coated cemented carbide, uncoated cemented carbide, and CNC cutting tools; the step of identifying the workpiece category of the cemented carbide workpiece by fusing surface spectral information, three-dimensional geometric information, and surface conductivity information by the control unit includes: when the surface spectral information indicates the presence of a target coating and the surface conductivity information indicates insulation, classifying the corresponding cemented carbide workpiece as coated cemented carbide; when the surface spectral information does not indicate the presence of a target coating and the surface conductivity information indicates conductivity, classifying the corresponding cemented carbide workpiece as uncoated cemented carbide; and when the three-dimensional geometric information indicates that the workpiece has a regular sheet-like structure, classifying the corresponding cemented carbide workpiece as a CNC cutting tool.

[0063] In this embodiment, the target coating refers to a specific coating on a cemented carbide workpiece, such as TiN or Al2O3.

[0064] Specifically, the control unit is equipped with a pre-trained decision tree model or a neural network module embedded with specific rules to perform the following classification process: For coating condition determination and classification: The control unit simultaneously analyzes the surface spectral information obtained by the hyperspectral imaging unit and the surface conductivity information obtained by the eddy current detection unit. Specifically, when the surface spectral information indicates the presence of a target coating, and the surface conductivity information simultaneously indicates insulation (weak eddy current response signal), the control unit classifies the corresponding workpiece as coated cemented carbide, utilizing dual evidence of chemical identification (spectral) and physical verification (conductivity). When the surface spectral information does not indicate the presence of a target coating, and the surface conductivity information simultaneously indicates conductivity (strong eddy current response signal), the control unit classifies the corresponding workpiece as uncoated cemented carbide.

[0065] For geometric type determination and classification: The control unit analyzes the three-dimensional geometric information acquired by the three-dimensional contour measurement unit in parallel. When the workpiece is identified from the three-dimensional geometric information as having a regular polygonal sheet-like structure (e.g., square, triangular, and with a thickness-to-area ratio that matches the characteristics of a cutting tool), the control unit 40 will preferentially classify it as a CNC cutting tool, regardless of its surface spectrum and electrical conductivity information (it may or may not have a coating). This rule uses geometry as the primary classification criterion, enabling rapid and reliable identification of cutting tool-type workpieces.

[0066] By combining the above three rules, the control unit can quickly and accurately classify various types of cemented carbide workpieces that appear in a mixture online, providing precise instructions for subsequent robot grasping and sorting.

[0067] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0068] The cemented carbide sorting system and method provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A cemented carbide sorting system, characterized in that, include: Multi-jointed robot body; A conveying mechanism is located within the working area of ​​the multi-joint robot body and is used to convey cemented carbide workpieces to be sorted. A multimodal detection device, mounted on a multi-joint robot body, is used to detect the cemented carbide workpiece. It includes at least a hyperspectral imaging unit for acquiring surface spectral information, a three-dimensional contour measurement unit for acquiring three-dimensional geometric information, and an eddy current detection unit for acquiring surface electrical conductivity information. The control unit is communicatively connected to the multimodal detection device and the multi-joint robot body, and is used to receive and fuse the surface spectral information, three-dimensional geometric information and surface electrical conductivity information, identify the workpiece category of the cemented carbide workpiece, and generate corresponding sorting control instructions; An end effector, installed at the operating end of the multi-joint robot body, is used to grasp and sort cemented carbide workpieces of the corresponding workpiece category according to the sorting control command.

2. The cemented carbide sorting system according to claim 1, characterized in that, The hyperspectral imaging unit includes a hyperspectral camera, whose response wavelengths include the visible light band and the near-infrared band, and is used to acquire the characteristic spectral reflectance curves of the surface coating of the cemented carbide workpiece to obtain the surface spectral information.

3. The cemented carbide sorting system according to claim 1, characterized in that, The three-dimensional contour measurement unit is a three-dimensional structured light profiler, used to acquire three-dimensional point cloud data of cemented carbide workpieces, and to obtain the three-dimensional geometric information by extracting features from the three-dimensional point cloud data.

4. The cemented carbide sorting system according to claim 1, characterized in that, The eddy current detection unit is integrated on the clamping surface of the end effector. The eddy current detection unit includes an eddy current detection probe for detecting the eddy current response signal on the surface of the cemented carbide workpiece to obtain surface conductivity information.

5. The cemented carbide sorting system according to claim 1, characterized in that, The control unit includes a workpiece identification module and a motion control module; The workpiece recognition module is used to fuse surface spectral information, three-dimensional geometric information and surface electrical conductivity information to identify the workpiece category of the cemented carbide workpiece and obtain the recognition result; A motion control module is used to generate the sorting control command based on the recognition result.

6. The cemented carbide sorting system according to claim 5, characterized in that, The workpiece recognition module includes a pre-trained deep learning model.

7. The cemented carbide sorting system according to claim 1, characterized in that, The workpiece categories include at least coated cemented carbide, uncoated cemented carbide, and CNC cutting tools.

8. The cemented carbide sorting system according to claim 1, characterized in that, The end effector is a force-controlled bionic gripper.

9. A method for sorting cemented carbide using the cemented carbide sorting system as described in any one of claims 1-8, characterized in that, include: The cemented carbide workpieces to be sorted are continuously conveyed by a conveying mechanism; When the cemented carbide workpiece enters the inspection station, the surface spectral information, three-dimensional geometric information and surface electrical conductivity information of the cemented carbide workpiece are acquired simultaneously through the multi-modal inspection device. By fusing surface spectral information, three-dimensional geometric information, and surface electrical conductivity information, the control unit identifies the workpiece category of the cemented carbide workpiece and generates sorting control instructions. The sorting control command is executed to grasp and sort the cemented carbide workpieces of the corresponding workpiece category.

10. The cemented carbide sorting method according to claim 9, characterized in that, The workpiece categories include at least coated cemented carbide, uncoated cemented carbide, and CNC cutting tools; The step of identifying the workpiece category of the cemented carbide workpiece by fusing surface spectral information, three-dimensional geometric information, and surface electrical conductivity information through the control unit includes: When the surface spectral information indicates the presence of a target coating and the surface conductivity information indicates insulation, the corresponding cemented carbide workpiece is classified as coated cemented carbide. When the surface spectral information does not indicate the presence of a target coating and the surface conductivity information indicates conductivity, the corresponding cemented carbide workpiece is classified as an uncoated cemented carbide. When the three-dimensional geometric information characterizes the workpiece as having a regular sheet-like structure, the corresponding cemented carbide workpiece is classified as a CNC cutting tool.

Images