A pump valve body full sequence processing method based on conjugate double stations and visual servo and five-face machining center

Abstract

The present application relates to a kind of based on conjugate double position and visual servoing pump valve body full sequence processing method and five surface machining center.The method includes: by industrial camera, pump valve body blank real-time image is acquired, input pre-training convolutional neural network, output instance segmentation mask and reference point pixel coordinate;Based on PnP algorithm, workpiece pose is solved in real time;Through visual servo control, control signal is generated, and vertical turning module and horizontal machining module are driven along the correction path and parallelly cooperated cutting;Real-time acquisition multi-dimensional force sensing data, combined with processing error compensation model, dynamically compensates servo shaft displacement;By the phase conversion of precision indexing table, until completing full sequence processing.Five surface machining center includes base, precision indexing table and the vertical turning module and horizontal machining module arranged around, is driven to realize parallel cutting by same numerical control system.The present application improves processing precision and efficiency, realizes full sequence automatic processing.

Description

Technical Field

[0001] This invention relates to the field of machining technology, and in particular to a method for full-sequence machining of pump and valve bodies based on conjugate dual-station and vision servo, and a five-sided machining center. Background Technology

[0002] In modern manufacturing, composite machining technology has received widespread attention and application to improve production efficiency and processing accuracy. However, traditional composite machining centers are limited by their structural design and functional integration, typically employing a single spindle or possessing only limited composite machining capabilities, making it difficult to meet the overall demand for multi-process, multi-angle, and high-efficiency machining of complex workpieces. Especially when dealing with large or irregularly shaped workpieces, it is often necessary to integrate multiple processes such as turning, milling, and drilling within the same machining flow, and to operate from different directions. For example, complex multi-faceted parts such as pump and valve bodies often have multiple ports, cavities, and flow channels in different directions, requiring high-precision hole machining, planar milling, thread turning, and sealing surface finishing on multiple spatial surfaces.

[0003] Currently, traditional equipment for machining pump and valve bodies generally follows a production model of multiple clamping, multi-machine collaboration, or frequent transfer between processes. However, due to the irregular shape and large casting tolerances of pump and valve body blanks, it is difficult to achieve high-precision initial positioning during clamping. This leads to deviations between the tool path and the actual workpiece surface during subsequent machining, directly affecting machining accuracy and surface quality. Furthermore, machining errors caused by tool wear, machine tool thermal deformation, and cutting force fluctuations are difficult to monitor and compensate for in real time. Existing technologies largely rely on post-processing inspection and manual correction, failing to achieve dynamic error closed-loop control during machining, thus limiting product yield.

[0004] Furthermore, the lack of online detection and adaptive adjustment capabilities for the real-time position and orientation of workpieces makes it difficult to achieve highly flexible and fully automated continuous processing. This not only significantly increases production cycle and cost, but also introduces cumulative errors due to repeated positioning, which seriously affects the final processing accuracy. Moreover, each process is usually executed serially, making it impossible to achieve true parallel processing, which restricts further improvement in overall efficiency.

[0005] Therefore, the applicant hereby proposes a method for full-sequence machining of pump and valve bodies based on conjugate dual-station and vision servo, as well as a five-sided machining center. Summary of the Invention

[0006] This invention proposes a method for full-sequence machining of pump and valve bodies based on conjugate dual-station machining, as well as a five-sided machining center, aiming to solve the technical problems of limited machining accuracy, low machining efficiency, low degree of automation, and difficulty in real-time compensation of machining errors in the prior art.

[0007] To achieve the above objectives, this invention proposes a method for the full-sequence machining of pump valve bodies based on conjugate dual-station and vision servo, comprising the following steps:

[0008] S1. Obtain a real-time two-dimensional image of the pump valve body blank and input it into a pre-trained convolutional neural network model.

[0009] S2. The model synchronously outputs the instance segmentation mask of the pump valve body and the pixel coordinates of multiple predefined machining reference points; and based on the pixel coordinates, known camera intrinsic parameters and the three-dimensional coordinates of each reference point, the PnP algorithm is used to calculate the current pose of the workpiece relative to the camera in real time.

[0010] S3. Compare the current pose with the desired pose, calculate the pose error, and generate a control signal based on the visual servo control law to drive the CNC system to control the vertical turning module and the horizontal machining module to perform parallel and collaborative cutting of the workpiece along the correction path.

[0011] S4. During the cutting process, multi-dimensional force sensing data is collected in real time, and combined with the real-time pose, the real-time position correction of the CNC axis is dynamically calculated through the preset machining error compensation model, and displacement compensation is performed on the corresponding servo axis.

[0012] S5. Control the precision indexing worktable to perform indexing conversion to the next phase, and repeat steps S10 to S50 until the entire sequence machining is completed.

[0013] Preferably, in step S10, a two-dimensional image of the pump valve body blank clamped on the precision indexing worktable is acquired in real time using an industrial camera fixed to the machine tool.

[0014] The acquisition trigger of the industrial camera is synchronized with the machine tool spindle movement or PLC signal.

[0015] Preferably, in step S20, the pre-trained convolutional neural network model is a model based on MaskR-CNN or YOLO-Pose architecture.

[0016] Preferably, the PnP algorithm minimizes the reprojection error function through iterative optimization, and the reprojection error function is:

[0017] ;

[0018] in, Let be the pixel coordinates of the i-th keypoint, K be the camera intrinsic parameter matrix, R be the rotation matrix, and t be the translation vector. Let i be the three-dimensional coordinates of the i-th key point in the workpiece coordinate system. The Project function represents the projection transformation from the three-dimensional point to the image pixel coordinates.

[0019] Preferably, in step S30, the visual servo control law adopts a position-based control strategy or an image-based visual servo control strategy.

[0020] When using an image-based vision servo control strategy, the control signals are the motion speed commands for each axis of the machine tool. And the speed commands for each axis of the machine tool The image feature error e satisfies the following relationship: or ;in, Scale gain matrix, This is the gain coefficient. This is the pseudo-inverse of the estimated Jacobian matrix of the image;

[0021] When a position-based control strategy is adopted, the control signal is the position command of each axis of the machine tool. By calculating the pose error between the current pose and the desired pose, the position compensation information of each axis ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ] is obtained through inverse kinematics.

[0022] Preferably, in step S30, the dynamic correction of the CNC machining code specifically involves:

[0023] Based on the pose compensation information ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ], the subsequent CNC machining code to be executed is dynamically corrected, and a machining path that matches the actual pose of the workpiece and is used for parallel and collaborative cutting by the vertical turning module and the horizontal machining module is planned.

[0024] Through coordinate transformation formula The tool position coordinates (X, Y, Z) in subsequent G-code instructions are transformed in real time online. The rotation matrix constructed for Euler angles.

[0025] As a preferred option, it also includes:

[0026] Before and after key processes, dimensional verification and compensation are performed using in-machine measurement probes integrated into the machine tool; according to processing requirements, automatic tool changing is performed using the corresponding disc tool magazine; automatic workpiece loading and unloading is performed using a robot-assisted module; and chips are continuously removed using an automatic chip removal device.

[0027] As a preferred embodiment, a five-sided machining center includes a base, wherein a precision indexing worktable for clamping workpieces and performing angular indexing is mounted in the center of the base, and a plurality of independent machining modules are arranged around the periphery of the precision indexing worktable, characterized in that;

[0028] The machining module includes a vertical turning module and a horizontal machining module; the vertical turning module includes a column fixed on the base, a vertical slide that slides vertically along the column, and a turning ram disposed on the vertical slide and reciprocating in the horizontal or vertical direction; the horizontal machining module includes a crossbeam spanning the base, a horizontal slide that slides horizontally along the crossbeam, and a horizontal spindle box disposed on the horizontal slide; the vertical turning module and the horizontal machining module are driven by the same CNC system, and the cutting tools on the turning ram and the cutting tools on the horizontal spindle box are configured to cut in parallel on the same workpiece on the precision indexing worktable.

[0029] Preferably, the CNC system is a three-channel control system, which uses a spatial interference verification algorithm to achieve synchronous motion control between the vertical turning module and the horizontal machining module.

[0030] The beneficial effects of the technical solution of this invention are as follows:

[0031] By directly learning and extracting deep features from images using deep convolutional neural networks—features that are insensitive to changes in illumination, local occlusion, and surface contamination—this method replaces traditional, easily failing threshold segmentation and manual feature extraction. This avoids feature extraction errors and pose calculation deviations caused by uneven illumination or workpiece surface defects, as seen in previous techniques. Furthermore, the use of strict PnP geometric constraints ensures the mathematical accuracy of pose calculation, improving the accuracy and robustness of pose recognition for pump and valve body machining.

[0032] During the cutting process, multi-dimensional force sensing data is acquired in real time and combined with real-time pose. The position of the CNC axis is dynamically calculated and compensated through a machining error compensation model, effectively suppressing machining errors caused by factors such as tool wear and changes in cutting force, thus ensuring machining quality and accuracy. The conjugate dual-station design, combined with the parallel collaborative cutting of the vertical turning module and the horizontal machining module, significantly shortens the machining cycle and improves production efficiency.

[0033] The introduction of a vision servo system can adapt to the randomness of the blank, eliminating the need for precise pre-positioning, thus improving the system's flexibility. It is suitable for processing pump and valve bodies of various varieties and in small batches, and the entire processing process is highly automated, reducing manual intervention, lowering the labor intensity of operators, and avoiding human error.

[0034] By arranging multiple independent machining modules, particularly the vertical turning module and the horizontal machining module, around the precision indexing worktable and driving them all through the same CNC system, the tools on the turning ram and the tools on the horizontal spindle box can cut in parallel on the same workpiece, greatly shortening the machining cycle and improving production efficiency. Furthermore, the workpiece can complete multiple machining operations, including vertical turning, indexing drilling, and horizontal milling, in a single setup, eliminating the cumulative errors in the reference datum caused by multiple setups in traditional sequential machining and greatly ensuring the spatial positional accuracy of the polyhedron.

[0035] Modular design allows the machining center to be flexibly configured with different types of machining modules according to the processing requirements of the workpiece, adapting to the processing of various complex workpieces and improving the flexibility and adaptability of the equipment. Furthermore, using a single CNC system to drive all machining modules simplifies the programming and operation of the equipment, reducing the skill requirements for operators. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the structure of a five-sided machining center in one embodiment of this application;

[0037] Figure 2 This is a side view of a five-sided machining center in one embodiment of this application;

[0038] Figure 3 This is a schematic diagram of the terminal equipment structure of the pump valve body full-sequence machining method based on conjugate dual-station and visual servo in one embodiment of this application.

[0039] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0040] The solutions in 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0042] It should also be noted that when a component is described as "fixed to" or "set on" another component, it can be directly on the other component or there may be an intervening component present. When a component is described as "connected to" another component, it can be directly connected to the other component or there may be an intervening component present.

[0043] Furthermore, descriptions using terms such as "first" and "second" in this invention are for descriptive purposes only (e.g., to distinguish identical or similar elements) and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" and "second" may explicitly or implicitly include at least one of those features. Additionally, technical solutions from different embodiments can be combined with each other, but only if they are feasible for those skilled in the art. If a combination of technical solutions is contradictory or impossible to implement, such a combination should be considered nonexistent and not within the scope of protection claimed by this invention.

[0044] This invention proposes a method for full-sequence machining of pump valve bodies based on conjugate dual-station and vision servo, comprising the following steps:

[0045] S1. Acquire real-time two-dimensional images of the pump / valve body blank and input them into a pre-trained convolutional neural network model. Specifically, in this step, an industrial camera fixed to the machine tool is used to acquire real-time two-dimensional images of the pump / valve body blank clamped on a precision indexing table. The acquisition trigger of the industrial camera is synchronized with the machine tool spindle movement or PLC signal to ensure the coordination and consistency between image acquisition and the processing. The acquired real-time two-dimensional images are input into the pre-trained convolutional neural network model for processing.

[0046] S2. The model synchronously outputs the instance segmentation mask of the pump valve body and the pixel coordinates of multiple predefined machining reference points; and based on the pixel coordinates, known camera intrinsic parameters and the three-dimensional coordinates of each reference point, the PnP algorithm is used to calculate the current pose of the workpiece relative to the camera in real time.

[0047] The pre-trained convolutional neural network model can be based on Mask R-CNN or YOLO-Pose architectures. These models can efficiently and accurately identify target objects in images and perform instance segmentation, while outputting the pixel coordinates of key points. The instance segmentation mask output by the model is used to accurately identify the contour of the pump valve body blank, while the pixel coordinates of multiple predefined machining reference points provide key information for subsequent pose calculation.

[0048] In an optional embodiment of the present invention, the pre-trained convolutional neural network model may be a model based on the YOLO-Pose architecture. This model employs a single-stage detection architecture and performs the following operations:

[0049] This framework unifies object detection and keypoint localization within a dense prediction framework. The backbone network employs a CSPDarknet53 architecture with 53 convolutional layers, utilizing cross-stage partial connections (CSP) to reduce computation and enhance gradient flow. The neck network employs a path aggregation network (PANet) to strengthen the feature pyramid, achieving better multi-scale feature fusion. The output head contains three parallel branches: one predicts bounding boxes (center coordinates, width, height, and confidence score), one predicts instance segmentation masks, and the last branch directly regresses the keypoint pixel coordinates of eight predefined processing reference points.

[0050] The model directly outputs the absolute pixel coordinates of each keypoint, rather than a heatmap. For each predicted region in the image, the model outputs a vector of dimension (4+1+1+16), where 4 values ​​represent bounding box information (center point x, y, width w, height h), 1 value represents the target confidence, 1 value represents the class probability (pump / valve body), and 16 values ​​represent the (x, y) coordinates of 8 keypoints. For example, for a bolt hole on a flange, the model directly outputs its center point coordinates (u_i, v_i) = (512.34, 256.78), without the need for post-processing to find heatmap extrema.

[0051] Furthermore, based on these pixel coordinates, pre-calibrated camera intrinsic parameters (including focal length, principal point coordinates, and distortion coefficients), and the known 3D coordinates of each reference point in the workpiece coordinate system, the current pose of the pump valve body blank relative to the industrial camera is calculated in real time using the PnP (Perspective-n-Point) algorithm. The current pose includes at least a rotation matrix R and a translation vector t, which together describe the spatial position and orientation of the workpiece in the camera coordinate system.

[0052] In a preferred embodiment, the PnP algorithm solves for the pose by iteratively optimizing and minimizing a reprojection error function. The reprojection error function is defined as:

[0053] ;

[0054] in, Let be the pixel coordinates of the i-th keypoint, K be the camera intrinsic parameter matrix, R be the rotation matrix, and t be the translation vector. Let R be the three-dimensional coordinates of the i-th key point in the workpiece coordinate system. The Project function represents the projection transformation from the three-dimensional point to the image pixel coordinates. Thus, by minimizing this error function, the optimal rotation matrix R and translation vector t can be obtained.

[0055] In another optional embodiment of the present invention, the pre-trained convolutional neural network model adopts a model based on the Mask R-CNN architecture. The model is pre-trained and data augmented on a dataset of pump valve body images containing various lighting conditions, different angles, and surfaces with oil stains or scratches to improve the model's generalization and robustness. The recognition and pose information extraction of the pump valve body are achieved through the following process:

[0056] The received RGB image is first scaled to 1024×1024 pixels and normalized to the range [0, 1]. The processed image is then fed into the model for inference. The backbone feature extraction network of this model adopts a ResNet-101-FPN structure, containing 101 convolutional layers, capable of extracting multi-scale depth features of the image. The Region Proposal Network (RPN) generates approximately 1000 candidate regions, and then the RoIAlign layer aligns the features of these regions and feeds them into two parallel branches: one branch predicts the category and bounding box offset of the candidate regions, and the other branch predicts the instance segmentation mask and keypoint heatmap.

[0057] The model synchronously outputs an instance segmentation mask of the pump valve body contour and the keypoint pixel coordinates of predefined machining reference points. The instance segmentation mask is a binary image where white pixels (value 1) represent the pump valve body area and black pixels (value 0) represent the background. Keypoint detection is performed by locating extreme points on a heatmap. These keypoints include the center of bolt holes on the pump valve body flange, the center of the end face, and the edge of the sealing groove, totaling eight feature points. For example, for a bolt hole on the flange, the model outputs the keypoint pixel coordinates as follows: =(512.34, 256.78), its accuracy can reach the sub-pixel level.

[0058] Using the keypoint pixel coordinates, the known camera intrinsic matrix K, and the three-dimensional coordinates of these keypoints in the workpiece coordinate system... The pose of the pump valve body relative to the camera is solved using the Perspective-n-Point (PnP) algorithm. Specifically, the reprojection error function is iteratively solved using optimization algorithms such as the Levenberg-Marquardt algorithm to obtain the optimal rotation matrix R and translation vector t.

[0059]

[0060] Among them, the camera intrinsic parameter matrix K was obtained in advance through Zhang Zhengyou calibration method, etc.; three-dimensional coordinates Extracted from the pump / valve body CAD model; the Project function represents the transformation of projecting 3D points in the world coordinate system onto image pixel coordinates, and its expression is:

[0061]

[0062] Optimize until the reprojection error is less than a preset threshold (e.g., 0.1 pixels) or reaches the maximum number of iterations (e.g., 50 times). Finally, calculate the pose with high translation and rotation accuracy.

[0063] S3. Compare the current pose with the desired pose, calculate the pose error, and generate a control signal based on the visual servo control law to drive the CNC system to control the vertical turning module and the horizontal machining module to perform parallel and collaborative cutting of the workpiece along the correction path.

[0064] In a preferred embodiment, the calculated current pose is compared with a preset desired pose to obtain a pose error. This pose error is input to the visual servo control module. The visual servo control law can employ a position-based control strategy or an image-based visual servo control strategy.

[0065] When using an image-based vision servo control strategy, the control signals are the motion speed commands for each axis of the machine tool. And the speed commands for each axis of the machine tool The image feature error e satisfies the following relationship: or ;in, Scale gain matrix, This is the gain factor (e.g., 0.5). It is the pseudo-inverse of the image Jacobian matrix estimation; the control quantity is directly calculated by feeding back the image error e, without the need for precise solution and dependence on the 3D pose of the workpiece, thus significantly reducing the sensitivity to camera calibration accuracy and robot model accuracy, and enhancing the robustness of the system.

[0066] For example, the control process is driven by comparing the currently detected keypoint coordinates with the preset desired keypoint coordinates. Let the desired keypoint pixel coordinates obtained from the CAD model or offline teaching be... The coordinates of the key points detected in the current frame are Then the image feature error ee is defined as a vector consisting of the coordinate deviations of all key points:

[0067] The error vector ee has a dimension of 16×1, and the motion speed commands for each axis of the machine tool are... The basic control law is generated based on the image error calculation: ;

[0068] For a single feature point Its image Jacobian matrix can be approximately represented as:

[0069]

[0070] In the formula, f is the camera focal length, and Z is the approximate depth information of the feature point relative to the camera (provided by the PnP solution). For multiple feature points, the total Jacobian matrix can be obtained by stacking the points. The obtained, its pseudo-inverse It can be calculated online or estimated in advance using methods such as singular value decomposition.

[0071] In summary, the final generated control signal is a 6×1 vector, used to represent the instantaneous motion velocity of the camera (or machine tool) in the X, Y, Z translational directions and the rotational directions around the X, Y, Z axes.

[0072] When a position-based control strategy is adopted, the control signal is the position command of each axis of the machine tool. By calculating the pose error between the current pose and the desired pose, the position compensation information of each axis ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ] is obtained through inverse kinematics.

[0073] Where Δx, Δy, and Δz are translational deviations, and Δα, Δβ, and Δγ are rotational deviations about the X, Y, and Z axes, respectively.

[0074] Furthermore, based on the pose compensation information ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ], the subsequent CNC machining code to be executed is dynamically corrected, and a machining path matching the actual pose of the workpiece is planned for parallel and collaborative cutting by the vertical turning module and the horizontal machining module.

[0075] Through coordinate transformation formula For tool position coordinates in subsequent G-code instructions Real-time online transformation is performed, among which, A rotation matrix constructed using Euler angles is used to ensure that the toolpath accurately follows the actual workpiece pose, thereby achieving high-precision machining. Driven by the CNC system, the vertical turning module and horizontal machining module perform parallel and coordinated cutting of the workpiece along the corrected path, significantly improving machining efficiency.

[0076] S4. During the cutting process, cutting force, torque, and other data are collected in real time using multi-dimensional force sensors. These multi-dimensional force sensor data, along with the real-time pose calculated in step S2, are input into the preset machining error compensation model.

[0077] The machining error compensation model can be established based on methods such as machine learning, finite element analysis, or empirical formulas. It is used to predict and compensate for machining errors caused by factors such as tool wear, machine tool thermal deformation, and changes in cutting force. The model dynamically calculates the real-time position correction of the CNC axes, and the CNC system performs displacement compensation on the corresponding servo axes, thereby achieving real-time dynamic compensation for machining errors and ensuring machining accuracy.

[0078] S5. After completing the machining of the current phase, the CNC system controls the precision indexing table to perform precise indexing, rotating the pump valve body blank to the next machining phase. Subsequently, steps S10 to S50 are repeated, that is, image acquisition, pose calculation, path correction, parallel collaborative cutting and error compensation are performed again until all machining surfaces and features of the pump valve body are completed, realizing full-sequence machining.

[0079] In a preferred embodiment, dimensional verification and compensation can be performed before and after critical processes using in-machine measurement probes integrated into the machine tool; automatic tool changing is performed using a corresponding disc tool magazine according to processing requirements; automatic workpiece loading and unloading is performed using a robot-assisted module; and chips are continuously removed using an automatic chip removal device. This further improves processing accuracy and ensures that critical dimensions meet requirements.

[0080] In a preferred embodiment, in step S30, the dynamic correction of the CNC machining code specifically involves:

[0081] Based on the pose compensation information ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ], the subsequent CNC machining code to be executed is dynamically corrected, and a machining path that matches the actual pose of the workpiece and is used for parallel and collaborative cutting by the vertical turning module and the horizontal machining module is planned.

[0082] Through coordinate transformation formula For tool position coordinates in subsequent G-code instructions Real-time online transformation is performed, among which, This represents the translational error of the workpiece in the X, Y, and Z directions. Represents the rotational error of the workpiece about the X, Y, and Z axes (e.g., Euler angles or quaternions). The rotation matrix is ​​constructed using Euler angles. Using the ZYX Euler angle order, this rotation matrix is ​​obtained by multiplying the three fundamental rotation matrices:

[0083] .

[0084] in, ;

[0085] ;

[0086]

[0087] The CNC system dynamically corrects subsequent G-code instructions based on this pose compensation information. For example, suppose the original G-code instruction is "G01X100.0Y50.0Z-10.0F1000", and the current pose compensation information is Δx=0.1, Δy=-0.05, Δz=0.02, Δα=0.001, Δβ=-0.002, Δγ=0.0005. Then, the module will calculate the new coordinate values ​​(X', Y', Z') in real time and correct the instruction online to "G01X100.098Y49.952Z-9.978F1000". This correction is performed in real time in a streamlined manner with extremely low latency, thus ensuring machining accuracy.

[0088] This real-time online coordinate transformation ensures that the actual movement trajectory of the tool can accurately compensate for the initial installation error and real-time positional deviation of the workpiece, thereby achieving high-precision machining.

[0089] On the other hand, after receiving the corrected G-code instructions, the vertical turning module and the horizontal machining module work together to cut the pump and valve body blank. This parallel collaborative cutting not only improves machining efficiency, but also ensures machining quality due to real-time pose correction.

[0090] See Figures 2-3 This invention proposes a vertical-horizontal composite machining center, comprising: a base 1, which is made of high-strength cast iron or welded structure to provide a foundation for high-precision machining; and a precision indexing worktable 2 installed in the middle of the base 1.

[0091] The precision indexing table 2 is used to clamp the workpiece to be processed, thereby achieving high-precision angular indexing. Furthermore, through the rotation and positioning of the precision indexing table 2, different processing surfaces of the workpiece can be oriented towards different processing modules, thus enabling multi-faceted processing.

[0092] Multiple independent machining modules are arranged around the precision indexing table 2. This gives the machining center a high degree of flexibility, allowing different types and numbers of machining modules to be configured according to actual machining needs.

[0093] Furthermore, the machining module includes at least a vertical turning module and a horizontal machining module.

[0094] The vertical turning module is mainly used for turning workpieces. It includes:

[0095] The column 3 is fixed to the base 1, providing vertical support and guidance.

[0096] The vertical slide 4 slides vertically along the column 3 and is driven by a transmission mechanism such as a servo motor and a ball screw to realize the feed motion of the turning tool in the vertical direction.

[0097] The turning ram 5, mounted on the vertical slide table 4, is capable of reciprocating in either a horizontal or vertical direction. A tool clamping mechanism is integrated within the turning ram 5 for mounting turning tools. Its reciprocating motion enables cutting feed and tool retraction during turning operations.

[0098] The horizontal machining module is mainly used for horizontal machining of workpieces, such as milling and drilling. It includes:

[0099] The crossbeam 6 spans the base 1, providing horizontal support and guidance.

[0100] The horizontal slide table 7 slides horizontally along the crossbeam 6 and is driven by a transmission mechanism such as a servo motor and a ball screw to realize the feed motion of the horizontal spindle box 8 in the horizontal direction.

[0101] The horizontal spindle box 8 is mounted on the horizontal slide table 7 and integrates a high-speed spindle and a tool clamping mechanism inside. It is used to mount milling tools, drill bits, etc., so as to realize the high-speed rotation and feed motion of the tools.

[0102] In this embodiment, both the vertical turning module and the horizontal machining module are driven by the same CNC system. This unified CNC system enables precise synchronization and collaborative work between the modules. More importantly, the cutting tools on the turning slide 5 and the horizontal spindle box 8 are configured to perform parallel cutting on the same workpiece on the precision indexing table 2. This means that during machining, the vertical turning module and the horizontal machining module can simultaneously process different parts of the workpiece, significantly improving machining efficiency. For example, when one face of the workpiece is being turned, its adjacent faces can be simultaneously milled or drilled, achieving true multi-station collaborative parallel machining.

[0103] In another preferred embodiment, the CNC system is a three-channel control system. The three-channel control system can simultaneously and independently control at least three motion axes or modules, thereby achieving more complex coordinated motion. Through a spatial interference verification algorithm, the CNC system can monitor the spatial positions of the vertical turning module and the horizontal machining module in parallel machining processes in real time, and perform interference prediction and avoidance to ensure that the two modules do not collide when cutting the same workpiece in parallel, thus achieving safe and efficient synchronous motion control.

[0104] In a preferred embodiment, the precision indexing table 2 integrates a direct-drive torque motor 21. The direct-drive torque motor 21 directly drives the table rotation, eliminating the need for traditional gear transmission mechanisms. This offers advantages such as fast response speed, high positioning accuracy, zero backlash, and good transmission rigidity, ensuring high precision and dynamic performance of the precision indexing table 2 during the indexing process. Furthermore, the bottom of the precision indexing table 2 is equipped with a hydraulic locking mechanism 22 for rigid locking after indexing. When the precision indexing table 2 rotates to a predetermined angle and is positioned, the hydraulic locking mechanism 22 quickly and rigidly locks the table, eliminating any minute gaps and vibrations, thereby providing extremely stable support during processing and further ensuring processing accuracy.

[0105] In another preferred embodiment, the horizontal spindle box 8 is driven by a servo motor in conjunction with a synchronous belt. The servo motor provides precise speed and position control, while the synchronous belt drive offers advantages such as smooth transmission, low noise, no lubrication required, and convenient maintenance, ensuring high-speed, high-precision rotation of the spindle of the horizontal spindle box 8. Furthermore, the side of the horizontal spindle box 8 is equipped with a first disc tool magazine 9 for automatically changing milling tools. The first disc tool magazine 9 can store various milling tools, drill bits, etc., and the required tools can be quickly and accurately changed during machining via an automatic tool changer (not shown), reducing manual intervention and improving automation and machining efficiency.

[0106] In another preferred embodiment, a second disc tool magazine 10 for automatically changing turning tools is disposed on the side of the turning ram 5. The second disc tool magazine 10 is the same as the first disc tool magazine 9, used to store various turning tools, and realizes rapid automatic changing of turning tools through an automatic tool changer (not shown), further improving the automation level and machining flexibility of the turning module.

[0107] In another preferred embodiment, a robot-assisted module 11 is further provided on the edge of the base 1. The end effector of the robot-assisted module 11, such as a robotic gripper, is located above the precision indexing table 2 for automatic loading and unloading of workpieces. The introduction of the robot-assisted module 11 realizes a fully automated process from workpiece blank to finished product, without the need for manual intervention, greatly improving production efficiency, reducing labor intensity, and minimizing errors and safety risks that may be caused by manual operation.

[0108] In another preferred embodiment, an automatic chip removal device is provided below the base 1. The automatic chip removal device includes scraper type, chain plate type, or magnetic chip conveyor, etc., used to automatically collect and discharge the chips generated during processing. The chip collection port of the automatic chip removal device extends below the precision indexing table 2.

[0109] In this embodiment, it is ensured that during the workpiece processing, regardless of the direction from which the chips are generated, they can effectively fall into the chip collection port and be discharged in time, avoiding chip accumulation that may affect processing accuracy or cause equipment failure, and keeping the working area clean.

[0110] Furthermore, this application embodiment also provides a terminal device, the internal structure of which can be as follows: Figure 3 As shown, the terminal device includes a processor, memory, communication interface, and database connected via a system bus. The processor provides computing and control capabilities. The terminal device's memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The terminal device's database stores data called by the computer programs. The terminal device's communication interface is used for data communication with external terminals. The terminal device's input device receives signals from external devices. When the computer program is executed by the processor, it implements a pump valve body full-sequence machining method based on conjugate dual-station and visual servoing as described in the above embodiment.

[0111] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the solution of this application, and does not constitute a limitation on the terminal device to which the solution of this application is applied.

[0112] Furthermore, this application also proposes a readable storage medium comprising a computer program that, when executed by a processor, implements the steps of the pump valve body full-sequence machining method based on conjugate dual-station and visual servoing as described in the above embodiments. It is understood that the readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0113] In summary, the pump and valve body full-sequence processing method, apparatus, terminal equipment, and readable storage medium provided in this application embodiment combine systematic image preprocessing, hierarchical target recognition, multi-dimensional feature extraction, and component correlation map analysis to construct an efficient, accurate, and reliable automated urine analysis solution.

[0114] Those skilled in the art will understand that implementing all or part of the processes in the above embodiments of the pump valve body full-sequence machining method based on conjugate dual-station and visual servoing can be accomplished by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above-described pump valve body full-sequence machining method based on conjugate dual-station and visual servoing. Any references to memory, storage, databases, or other media used in this application and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Dual-speed SDRAM (SSRSDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus Direct RAM (RDRAM), Direct Memory Bus Dynamic RAM (DRDRAM), and Memory Bus Dynamic RAM (RDRAM), etc.

[0115] The above description is only a part or preferred embodiment of the present invention. Neither the text nor the drawings should limit the scope of protection of the present invention. All equivalent structural transformations made using the content of the present invention specification and drawings under the overall concept of the present invention, or direct / indirect applications in other related technical fields, are included within the scope of protection of the present invention.

Claims

1. A method for full-sequence machining of pump valve bodies based on conjugate dual-station and vision servo, characterized in that, Includes the following steps: S1. Obtain a real-time two-dimensional image of the pump valve body blank and input it into a pre-trained convolutional neural network model. S2. The model synchronously outputs the instance segmentation mask of the pump valve body and the pixel coordinates of multiple predefined machining reference points; and based on the pixel coordinates, known camera intrinsic parameters and the three-dimensional coordinates of each reference point, the PnP algorithm is used to calculate the current pose of the workpiece relative to the camera in real time. S3. Compare the current pose with the desired pose, calculate the pose error, and generate a control signal based on the visual servo control law to drive the CNC system to control the vertical turning module and the horizontal machining module to perform parallel and collaborative cutting of the workpiece along the correction path. S4. During the cutting process, multi-dimensional force sensing data is collected in real time, and combined with the real-time pose, the real-time position correction of the CNC axis is dynamically calculated through the preset machining error compensation model, and displacement compensation is performed on the corresponding servo axis. S5. Control the precision indexing worktable to perform indexing conversion to the next phase, and repeat steps S10 to S50 until the entire sequence machining is completed.

2. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 1, characterized in that, In step S10, a two-dimensional image of the pump valve body blank clamped on the precision indexing worktable is acquired in real time by an industrial camera fixed to the machine tool. The acquisition trigger of the industrial camera is synchronized with the machine tool spindle movement or PLC signal.

3. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 1, characterized in that, In step S20, the pre-trained convolutional neural network model is a model based on MaskR-CNN or YOLO-Pose architecture.

4. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 2, characterized in that, In step S20, the current pose includes at least a rotation matrix and a translation vector.

5. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 3, characterized in that, The PnP algorithm minimizes the reprojection error function through iterative optimization. The reprojection error function is: ； in, Let be the pixel coordinates of the i-th keypoint, K be the camera intrinsic parameter matrix, R be the rotation matrix, and t be the translation vector. Let i be the three-dimensional coordinates of the i-th key point in the workpiece coordinate system. The Project function represents the projection transformation from the three-dimensional point to the image pixel coordinates.

6. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 5, characterized in that, In step S30, the visual servo control law adopts a position-based control strategy or an image-based visual servo control strategy. When using an image-based vision servo control strategy, the control signals are the motion speed commands for each axis of the machine tool. And the speed commands for each axis of the machine tool The image feature error e satisfies the following relationship: or ;in, Scale gain matrix, This is the gain coefficient. This is the pseudo-inverse of the estimated Jacobian matrix of the image; When a position-based control strategy is adopted, the control signal is the position command of each axis of the machine tool. By calculating the pose error between the current pose and the desired pose, the position compensation information of each axis ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ] is obtained through inverse kinematics.

7. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 1, characterized in that, In step S30, the dynamic correction of the CNC machining code specifically involves: Based on the pose compensation information ΔP=[Δx, Δy, Δz, Δα, Δβ, Δγ], the subsequent CNC machining code to be executed is dynamically corrected, and a machining path that matches the actual pose of the workpiece and is used for parallel and collaborative cutting by the vertical turning module and the horizontal machining module is planned. Through coordinate transformation formula The tool position coordinates (X, Y, Z) in subsequent G-code instructions are transformed in real time online. The rotation matrix constructed for Euler angles.

8. The pump valve body full-sequence machining method based on conjugate dual-station and vision servo according to claim 1, characterized in that, Also includes: Before and after key processes, dimensional verification and compensation are performed using in-machine measurement probes integrated into the machine tool; according to processing requirements, automatic tool changing is performed using the corresponding disc tool magazine; automatic workpiece loading and unloading is performed using a robot-assisted module; and chips are continuously removed using an automatic chip removal device.

9. A five-sided machining center, comprising a base (1), wherein a precision indexing worktable (2) for clamping workpieces and performing angular indexing is mounted in the middle of the base (1), and a plurality of independent machining modules are arranged around the periphery of the precision indexing worktable (2). Its characteristics are: The machining module includes a vertical turning module and a horizontal machining module; the vertical turning module includes a column (3) fixed on the base (1), a vertical slide (4) that slides vertically along the column (3), and a turning ram (5) that is set on the vertical slide (4) and reciprocates in the horizontal or vertical direction; the horizontal machining module includes a crossbeam (6) spanning the base (1), a horizontal slide (7) that slides horizontally along the crossbeam (6), and a horizontal spindle box (8) set on the horizontal slide (7); the vertical turning module and the horizontal machining module are driven by the same CNC system, and the cutting tools on the turning ram (5) and the cutting tools on the horizontal spindle box (8) are configured to cut in parallel on the same workpiece on the precision indexing table (2).

10. A five-sided machining center according to claim 1, characterized in that, The CNC system is a three-channel control system that uses a spatial interference verification algorithm to achieve synchronous motion control between the vertical turning module and the horizontal machining module.

Images