A deep learning-based automatic double-end face milling control method and system

Abstract

The application relates to the technical field of numerical control machining equipment, in particular to an automatic double-end face milling control method and system based on deep learning, which comprises the following steps: based on an end face image of a profile to be machined, a lowest point in a profile section, an upper step boundary and left and right end machining postures are identified through a deep learning model, a common machining reference of double-end face milling is established by taking the lowest point in the profile section as the reference, and a double-end same-reference profile positioning result is obtained; based on the double-end same-reference profile positioning result, left and right end machining angles, left and right end step machining sizes and a preset saw blade shaft to rotating shaft center compensation relationship, a left end angle cutting track, a right end angle cutting track and corresponding step compensation cutting tracks are respectively generated. The application solves the length error, step misplacement and angle deviation problems caused by the non-uniform reference in double-end machining of long profiles.

Description

Technical Field

[0001] This invention relates to the field of CNC machining equipment technology, and in particular to an automated dual-head end milling control method and system based on deep learning. Background Technology

[0002] Double-head end milling machines are commonly used for end-to-end cutting, angle cutting, and step surface processing of long workpieces such as wood profiles and stepped profiles. In existing processing methods, the profile length, the processing dimensions of the steps at both ends, and the cutting angle usually rely on manual measurement, data entry, and machine adjustment. Especially when the angles at both ends of the profile are different, the step heights are different, or there is a slight offset at the end face, the left and right processing heads are prone to perform processing according to their respective zero points, resulting in inconsistent processing references at both ends, which in turn leads to length errors, step misalignment, or angle deviations.

[0003] Although some CNC equipment can calculate the saw blade feed position based on the input angle, it mainly relies on static parameter conversion and fails to combine the actual posture of the profile end face, the inner lowest point reference, the compensation relationship between the saw blade axis and the center of the rotation axis, and the length closure relationship between the left and right machine heads for linkage control.

[0004] If the intersection between the angle cutting trajectory and the step compensation trajectory is insufficient, residual steps are likely to be generated; if the intersection is excessive, it may cause overcutting and uneven cut surface. Therefore, there is an urgent need for an automated dual-head end milling control method that is based on deep learning to identify the end face reference and combines dual-end reference positioning, dual-end linkage trajectory generation and intersection state correction to improve machining accuracy and batch machining stability. Summary of the Invention

[0005] This invention provides an automated dual-head end milling control method and system based on deep learning.

[0006] A deep learning-based automated dual-head end milling control method includes the following steps: Based on the end face image of the profile to be processed, the lowest point inside the profile section, the upper step boundary, and the processing posture of the left and right ends are identified by a deep learning model. The lowest point inside the profile section is used to establish a common processing datum for double-head end milling, and the positioning result of the profile with the same datum at both ends is obtained. Based on the positioning results of the double-end reference profile, the processing angles of the left and right ends, the processing dimensions of the steps at the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis, the cutting trajectory of the left end angle, the cutting trajectory of the right end angle, and the corresponding step compensation cutting trajectory are generated respectively. The double-end linkage processing trajectory is generated according to the length closure relationship between the cutting trajectory of the left end angle and the cutting trajectory of the right end angle. Based on the dual-end linkage processing trajectory control, the left and right processing heads simultaneously clamp the profile and perform end cutting, angle cutting and step compensation cutting. During the processing, the feed endpoint of the corresponding processing head is corrected according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory, so as to obtain the profile processing result with consistent double-end step surfaces.

[0007] Furthermore, the end face image of the profile to be processed is acquired after entering the double-head end milling area. The end face image is then input into a pre-trained deep learning model. The deep learning model identifies the outer contour, inner lower boundary, upper step boundary, and end orientation of the profile in the end face image. The end face contour recognition result is represented as follows: ;in, This indicates the end face contour recognition result. This refers to a pre-trained deep learning model. This represents an image of the end face of the profile to be processed. Indicates the outer contour of the profile. Indicates the inner lower boundary. Indicates the bottom boundary of the end face. Indicates the boundary of the upper step. Indicates the direction of the end; Based on the end face contour recognition results, the intersection position of the inner lower boundary and the bottom boundary of the end face is determined, and the intersection position is taken as the lowest point inside the profile section. At the same time, the step removal area is determined according to the upper step boundary, and the processing posture of the left and right ends is determined according to the end orientation. The profile end face reference recognition result is represented as follows: ;in, This indicates the profile end face reference identification result. Indicates the lowest point inside the profile section. Indicates the area to be cut off from the step. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

[0008] Preferably, based on the profile end face reference identification results, the lowest point inside the profile section is converted into a common machining reference point for the dual-head end milling machine, and the step removal area and the machining postures of the left and right ends are mapped to the coordinate system of the left machining head and the coordinate system of the right machining head, respectively. The resulting positioning result of the profile with the same reference at both ends is expressed as follows: ;in, This indicates the positioning results of the profiles at both ends using the same reference. This indicates the common machining reference point in the equipment's machining coordinate system. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

[0009] As a further technical solution of the present invention, based on the positioning results of the double-end same-reference profile, the step removal area in the left machining head coordinate system, the step removal area in the right machining head coordinate system, the left-end machining posture, and the right-end machining posture are read respectively, and the left-end machining angle, the right-end machining angle, the left-end step machining dimension, and the right-end step machining dimension are matched to the corresponding step removal areas respectively, and the double-end step machining constraint result is expressed as follows: ; ; ;in, This indicates the machining constraint result for the double-ended step. This indicates the machining constraints for the left-end step. This indicates the machining constraints for the right-end step. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. Indicates the processing posture of the right end. Indicates the machining angle at the left end. Indicates the machining angle at the right end. Indicates the machining dimensions of the left end step. This indicates the machining dimensions of the right-end step.

[0010] Furthermore, based on the double-ended step machining constraint results and the preset compensation relationship between the saw blade axis and the rotation axis center, horizontal feed compensation is performed on the left-end machining angle and the right-end machining angle respectively to obtain the left-end horizontal cutting distance and the right-end horizontal cutting distance. Based on the left-end horizontal cutting distance, the left-end angle cutting trajectory is generated as follows: The right-end angle cutting trajectory generated based on the right-end horizontal cutting distance is represented as follows: ; Based on the left-end and right-end angle cutting trajectories, corresponding left-end and right-end step compensation cutting trajectories are generated respectively. This ensures that the angle cutting trajectories and step compensation cutting trajectories within the same processing head intersect within the corresponding step removal areas, resulting in a combined double-end cutting trajectory. ;in, This indicates the combined cutting trajectory at both ends. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This indicates the cutting trajectory of the left-end step. This indicates the tangent trajectory of the step on the right. Indicates the intersection area at the left end. This indicates the intersection area at the right end.

[0011] Preferably, based on the dual-end combined cutting trajectory and the length of the profile to be processed, the length closure relationship between the left-end angle cutting trajectory and the right-end angle cutting trajectory is determined, and the relative positioning distance between the left and right processing heads is constrained according to the length closure relationship, generating a dual-end linkage processing trajectory as follows: ;in: Indicates the dual-end linkage processing trajectory. This indicates the combined cutting trajectory at both ends. Indicates a length closure relationship. This indicates the positioning constraints of the left machining head. This indicates the positioning constraints of the right machining head.

[0012] As a further technical solution of the present invention, based on the dual-end linkage machining trajectory, the positioning constraints of the left machining head, the positioning constraints of the right machining head, the left-end angle cutting trajectory, the right-end angle cutting trajectory, the left-end step compensation cutting trajectory, and the right-end step compensation cutting trajectory are analyzed, and the dual-end synchronous machining timing is generated according to the machining sequence of clamping, flush cutting, angle cutting, and step compensation cutting as follows: ;in, Indicates the timing sequence of synchronous processing at both ends. This indicates the sequence arrangement according to the order of clamping, flush cutting, angled cutting, and step-by-step trimming; Based on the dual-end synchronous processing sequence, the left and right processing heads are controlled to move to their corresponding positioning positions, and the middle clamping mechanism and the left and right end clamping mechanisms are controlled to simultaneously clamp the profile to be processed. Then, the end-cutting saw blade is controlled to perform vertical end-cutting on the left and right ends of the profile respectively, resulting in a dual-end end-cutting positioning state, as shown below: , , , , ;in, This indicates a position where both ends are aligned. This indicates that the clamping is confirmed. This indicates the state of the end face after a vertical, flush cut at the left end. This indicates the state of the end face after a vertical, flush cut on the right side.

[0013] Furthermore, based on the double-end alignment positioning, the left and right processing heads are controlled to perform angle cutting and step compensation cutting according to the left-end angle cutting trajectory, the right-end angle cutting trajectory, the left-end step compensation cutting trajectory, and the right-end step compensation cutting trajectory, respectively. During the execution, the intersection state of the angle cutting trajectory and the step compensation cutting trajectory in the corresponding step removal area is monitored, and the double-end intersection state judgment result is expressed as follows: ;in, This indicates the result of the two-end intersection state judgment. Indicates the state of intersection at the left end; This indicates the state of intersection at the right end.

[0014] Preferably, based on the result of the double-end intersection state judgment, when the intersection is insufficient, the feed endpoint of the corresponding processing head is extended; when the intersection is excessive, the feed endpoint of the corresponding processing head is shortened; when the intersection is normal, the feed endpoint of the corresponding processing head remains unchanged, and the processing of the left and right end step surfaces is completed according to the corrected feed endpoint. The resulting profile with consistent double-end step surfaces is expressed as follows: ;in, This indicates the processing result of a profile with consistent stepped surfaces at both ends. This indicates the machining result of the left end step surface. This indicates the machining result of the right-end step surface. Indicates the state of intersection at the left end. Indicates the right-end intersection state. Indicates the feed endpoint after left-hand correction. This indicates the feed endpoint after the correction on the right.

[0015] An automated dual-head end milling control system based on deep learning includes the following modules: The end face image acquisition module is used to acquire the end face image of the profile to be processed after it enters the double-head end face milling area. The dual-end common reference positioning module is used to input the end face image into a pre-trained deep learning model, identify the lowest point inside the profile section, the upper step boundary, and the processing posture of the left and right ends, and establish a common processing reference for dual-head end milling based on the lowest point inside the profile section to obtain the dual-end common reference profile positioning result. The dual-end linkage trajectory generation module is used to generate the left-end angle cutting trajectory, the right-end angle cutting trajectory, the corresponding step compensation trajectory and the dual-end linkage processing trajectory based on the positioning results of the dual-end reference profile, the processing angles of the left and right ends, the processing dimensions of the steps of the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis. The dual-end synchronous processing control module is used to control the left and right processing heads to clamp the profile synchronously based on the dual-end linkage processing trajectory and perform end cutting, angle cutting and step compensation cutting. It also corrects the feed endpoint of the corresponding processing head according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory to obtain a profile processing result with consistent step surfaces at both ends.

[0016] The beneficial effects of this invention are: This invention uses a deep learning model to identify the lowest point inside the profile cross-section, the upper step boundary, and the processing posture of the left and right ends from the end face image. The lowest point inside is used as the common processing reference for the left and right processing heads, so that the processing of the left and right ends no longer depends on their own independent zero points. This solves the problems of length error, step misalignment, and angle deviation caused by inconsistent references in the processing of long profiles at both ends.

[0017] This invention generates left and right end angle cutting trajectories and step compensation cutting trajectories based on the compensation relationship between the saw blade axis and the center of the rotation axis. It also constrains the relative positioning of the left and right processing heads through the length closure relationship. During the processing, the feed endpoint is dynamically corrected according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory to avoid step residue or overcutting. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the method flow of Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the system modules in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram comparing the effects of Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the architecture of Embodiment 3 of the present invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0021] Example 1 like Figure 1 As shown, an automated dual-head end milling control method based on deep learning includes the following steps: S1. Based on the end face image of the profile to be processed, the lowest point inside the profile section, the upper step boundary, and the processing posture of the left and right ends are identified by a deep learning model. The common processing reference for the double-head end milling is established with the lowest point inside the profile section, and the positioning result of the double-headed profile with the same reference is obtained.

[0022] S11, when acquiring the end face image of the profile to be processed after entering the double-head end milling processing area, end face imaging components are respectively set at the feeding positioning station of the double-head end milling equipment or near the left and right processing heads. The end face imaging components include an industrial camera, a directional light source, and a fixed mounting bracket. After the feeding mechanism feeds the profile to be processed into the processing area, it first confirms that the profile has reached the preset shooting position through the positioning block, the machine head alignment position, or photoelectric detection signal, and controls the middle clamping mechanism or auxiliary limiting component to temporarily limit the profile, so that the end face of the profile remains stable relative to the camera. Then, the directional light source is turned on to image the profile. The profile end face is supplemented with light by an industrial camera taking pictures of the left and right end faces of the profile, respectively, to obtain end face images that reflect the outer contour, inner lower boundary, bottom boundary of the end face, upper step boundary, and end orientation of the profile. After acquisition, the end face images are subjected to distortion correction, brightness equalization, and processing area cropping to establish a correspondence between the image coordinates of the end face images and the processing coordinates of the dual-head end milling machine. The processed end face images are then used as input to a deep learning model for subsequent identification of the lowest point inside the profile section, the upper step boundary, and the processing posture of the left and right ends.

[0023] The end face image is input into a pre-trained deep learning model, which identifies the profile's outer contour, inner lower boundary, upper step boundary, and end orientation in the end face image to obtain the end face contour recognition result.

[0024] The pre-trained deep learning model is an end-face recognition model that shares a feature extraction structure and a multi-task recognition structure. Specifically, it first collects end-face images of different profile types after entering the double-end milling processing area, and labels the profile's outer contour, inner lower boundary, end-face bottom boundary, upper step boundary, lowest point of the profile cross-section, and the orientation of the left or right end in the images. In the deep learning model structure, an image correction layer is set at the front end to perform size normalization, tilt correction, and processing area cropping on the end-face image. A convolutional feature extraction layer is set in the middle to extract the edge, step, bottom edge, and inner contour features of the profile end face. The back end has three parallel output branches. The first output branch is a contour segmentation branch, used to identify the profile's outer contour, inner lower boundary, and upper step boundary. The second output branch is a key point localization branch, used to locate the intersection of the inner lower boundary and the end-face bottom boundary and output the profile cross-section. The lowest point inside the face, the third output branch is the posture classification branch, used to determine whether the current end face belongs to the left end processing posture or the right end processing posture. During training, the labeled end face images are used as training samples to input into the deep learning model, so that the output results of the deep learning model are compared with the manually labeled results, and the contour recognition error, key point positioning error and end orientation judgment error are jointly optimized. At the same time, enhanced samples such as brightness change, slight deflection, wood chip occlusion, end face burrs and profile placement offset are added to make the deep learning model adapt to the image fluctuations of the actual processing site. When the deep learning model stably reaches the preset accuracy in recognizing the outer contour of the profile, the inner lower boundary, the upper step boundary, the lowest point inside the profile section and the left and right end processing postures, it is solidified into a pre-trained deep learning model, which is used to output the end face contour recognition result based on the end face image of the profile to be processed, and further form the profile end face reference recognition result.

[0025] The deep learning model combines end-face contour segmentation with key point recognition. Its input is an end-face image, and its output is the end-face contour recognition result, which is represented as follows: ; in, This indicates the end face contour recognition result. This refers to a pre-trained deep learning model. This represents an image of the end face of the profile to be processed. Indicates the outer contour of the profile. Indicates the inner lower boundary. Indicates the bottom boundary of the end face. Indicates the boundary of the upper step. Indicates the direction the end faces.

[0026] Deep learning models do not only identify the presence of profiles, but also further identify the inner lower boundary, upper step boundary, and end orientation that are directly related to subsequent end milling control, so that the end profile recognition results can serve as the basis for subsequently determining the lowest point inside the profile section, the step removal area, and the machining posture of the left and right ends.

[0027] S12, based on the end face contour recognition results, determine the intersection position of the inner lower boundary and the bottom boundary of the end face, and take the intersection position as the lowest point inside the profile section. That is, instead of taking the outer edge of the profile or the default zero point of the equipment as the processing reference, the lowest point inside the profile section is taken as the reference point for subsequent double-head end face milling positioning and processing trajectory generation.

[0028] The lowest point inside the profile section is represented as: ;in, Indicates the lowest point inside the profile section. Indicates the inner lower boundary. Indicates the bottom boundary of the end face. It indicates the intersection of two boundaries.

[0029] After determining the lowest point inside the profile section, the step removal area is determined based on the upper step boundary. Specifically, the area to be removed between the upper step boundary and the outer contour of the profile is determined as the step removal area, so that the subsequent angle cutting trajectory and step repair cutting trajectory can be generated around the step removal area.

[0030] The step-cut area is represented as follows: ;in, Indicates the area to be cut off from the step. This indicates the region construction relationship based on the profile's outer contour, the upper step boundary, and the lowest point inside the profile section to determine the step removal area. Indicates the outer contour of the profile. Indicates the boundary of the upper step. This indicates the lowest point inside the profile cross-section.

[0031] The left and right end processing postures are determined based on the end orientation. These postures characterize whether the current end face belongs to the left or right end processing side, as well as the placement direction, angle cutting direction, and step compensation cutting direction of the corresponding end in the processing area. The profile end face reference identification result is expressed as follows: ; in, This indicates the profile end face reference identification result. Indicates the lowest point inside the profile section. Indicates the area to be cut off from the step. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

[0032] The end face contour recognition results are further converted into profile end face reference recognition results. The profile end face reference recognition results not only clarify the common reference point, but also clarify the step removal area and the processing posture of the left and right ends, providing a basis for the subsequent establishment of a common processing reference for double-end face iron.

[0033] S13, based on the profile end face reference recognition results, the lowest point inside the profile section is converted into a common processing reference point for the dual-head end face milling equipment. Specifically, the lowest point inside the profile section in the end face image coordinate system is mapped to the equipment processing coordinate system, so that both the left and right processing heads use the same lowest point inside the profile section as the processing zero point reference, thereby avoiding length errors, step misalignment or angle deviations caused by the left and right processing heads using independent references respectively.

[0034] First, a standard profile or calibration block with known dimensions and known reference points is placed in the processing area, so that its inner lowest point, bottom edge of the end face, and upper step boundary, etc., are imaged in the end face image. At the same time, the actual mechanical coordinates of these feature points in the equipment processing coordinate system are recorded. Then, based on the pixel coordinates of the feature points in the image and the actual coordinates in the equipment, the proportional relationship between the pixel size and the actual millimeter size, the rotation correction relationship caused by the camera installation tilt, the translation relationship between the image origin and the equipment mechanical origin, and the mirror direction relationship that may exist during the processing of the left and right ends are determined, thereby forming a calibration mapping relationship.

[0035] The common processing reference point is represented as: ;in, This indicates the common machining reference point in the equipment's machining coordinate system. This indicates the calibration mapping relationship between the end face image coordinate system and the equipment processing coordinate system. This represents the lowest point inside the profile section in the end face image coordinate system.

[0036] After obtaining the common machining reference point, the step removal area and the machining postures of the left and right ends are mapped to the coordinate system of the left machining head and the coordinate system of the right machining head, respectively, so as to obtain the positioning basis of the left and right machining heads under the same common machining reference.

[0037] The correspondence can be represented as: , ; , ; in, This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. This describes the mapping relationship from the end face image coordinate system to the left machining head coordinate system. First, through the calibration mapping relationship from the end face image coordinate system to the equipment machining coordinate system, the lowest point inside the profile section, the upper step boundary, and the step removal area identified in the end face image are transformed into the overall equipment machining coordinate system. Then, using the mechanical zero point of the left machining head, the center of the left rotation axis, the direction of the left saw blade feed axis, and the left end clamping and positioning position as references, the above-mentioned overall equipment machining coordinate system is further converted into local coordinates that the left machining head can execute. This describes the mapping relationship from the end face image coordinate system to the right machining head coordinate system. First, the lowest point inside the profile section, the upper step boundary, and the step removal area in the end face image are converted to the overall machining coordinate system of the equipment. However, when further converting to the right machining head coordinate system, it is necessary to consider the mechanical zero point of the right machining head, the center of the right rotation axis, the direction of the right saw blade feed axis, the right end clamping and positioning position, and the length positioning relationship between the left and right machining heads. Since the right and left machining heads are usually located at opposite ends of the profile, their machining directions are mirror images or opposite to the left end. Therefore, the mapping of the right machining head coordinate system requires corresponding corrections to the end face orientation, positive feed direction, and angle cutting direction. This represents the step-cut area in the end-face image coordinate system. This indicates the lowest point inside the profile cross-section.

[0038] Finally, based on the common machining reference point, the step removal area in the left machining head coordinate system, the step removal area in the right machining head coordinate system, the left end machining posture, and the right end machining posture, the positioning result of the profile with the same reference at both ends is obtained.

[0039] The positioning results of the double-ended same-reference profile are expressed as follows: ; in, This indicates the positioning results of the profiles at both ends using the same reference. This indicates the common machining reference point in the equipment's machining coordinate system. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

[0040] The deep learning model first identifies the profile's outer contour, upper step position, inner lowest point, and end orientation in the end face image. Then, based on the end face orientation, it determines whether the end face belongs to the left or right end processing side. Combined with the corresponding machine head position, feed direction, rotation axis direction, and cutting angle direction, it determines the left or right end processing posture.

[0041] When generating the left-end angle cutting trajectory, the right-end angle cutting trajectory, and the corresponding step compensation cutting trajectory, the left and right processing heads are based on the same common processing reference point to achieve consistency of the processing reference at both ends. The positioning result of the reference profile at both ends is used as subsequent input to combine the processing angles at the left and right ends, the processing dimensions of the steps at the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis to further generate the double-end linkage processing trajectory.

[0042] S2, based on the positioning results of the double-end reference profile, the processing angles of the left and right ends, the processing dimensions of the steps at the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis, generates the cutting trajectory of the left end angle, the cutting trajectory of the right end angle, and the corresponding step supplementary cutting trajectory, and generates the double-end linkage processing trajectory according to the length closure relationship between the cutting trajectory of the left end angle and the cutting trajectory of the right end angle.

[0043] S21. Based on the obtained positioning results of the double-ended reference profile, the step removal area in the left machining head coordinate system, the step removal area in the right machining head coordinate system, the left-end machining posture, and the right-end machining posture are read respectively. The step removal area in the left machining head coordinate system is used to limit the range of step material to be removed by the left machining head, and the step removal area in the right machining head coordinate system is used to limit the range of step material to be removed by the right machining head. The left-end machining posture and the right-end machining posture are used to determine the feed direction, rotation direction, and positive and negative angle of the corresponding machining head.

[0044] The machining angle and machining dimension of the left end are matched to the step removal area in the coordinate system of the left machining head, and the machining angle and machining dimension of the right end are matched to the step removal area in the coordinate system of the right machining head, thus obtaining the machining constraint result of the double-end step.

[0045] The machining constraint result for the double-ended step is expressed as follows: ; ; ; in, This indicates the machining constraint result for the double-ended step. This indicates the machining constraints for the left-end step. This indicates the machining constraints for the right-end step. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. Indicates the processing posture of the right end. Indicates the machining angle at the left end. Indicates the machining angle at the right end. Indicates the machining dimensions of the left end step. This indicates the machining dimensions of the right-end step.

[0046] The output of the double-end positioning results of the reference profile is established in correspondence with the actual input processing angles and step processing dimensions of the left and right ends, so that the subsequent trajectory generation is no longer based on a single end face, but on the step cutting areas and processing postures of the left and right ends respectively.

[0047] S22, based on the double-end step processing constraint results and the preset saw blade axis to rotation axis center compensation relationship, perform horizontal feed compensation for the left end processing angle and the right end processing angle respectively; the saw blade axis to rotation axis center compensation relationship is used to represent the horizontal position change relationship caused by the distance between the saw blade axis and the rotation axis center when the angle cutting saw blade performs angle cutting around the rotation axis.

[0048] Angle-cutting saw blades do not cut directly in a straight line according to the input step machining dimensions. Instead, they are mounted on a machining component that can rotate around a rotation axis. When the machining angle at the left or right end changes, the saw blade axis will generate an additional horizontal displacement relative to the center of the rotation axis. Therefore, the actual horizontal cutting distance at the left and right ends that need to be controlled cannot be equal to the machining dimensions at the left or right end of the step. Instead, the horizontal compensation amount determined by the distance from the saw blade axis to the center of the rotation axis and the machining angle needs to be added or subtracted.

[0049] First, calculate the horizontal compensation amount at the left end and the horizontal compensation amount at the right end based on the machining angles at the left and right ends, respectively. These are expressed as follows: ; ; in, This indicates the horizontal compensation amount at the left end. This indicates the horizontal compensation amount at the right end. This indicates the distance from the center of the saw blade axis to the center of the rotation axis in the left machining head. This indicates the distance from the saw blade axis to the center of the rotation axis in the right machining head. Indicates the machining angle at the left end. Indicates the machining angle at the right end. Represents pi (π). This represents the sine function.

[0050] Based on the horizontal compensation amounts at the left and right ends, the machining dimensions of the left and right steps are corrected to obtain the horizontal cutting distances at the left and right ends, which are expressed as follows: ; ; in, Indicates the horizontal cutting distance at the left end. Indicates the horizontal cutting distance at the right end. Indicates the machining dimensions of the left end step. This indicates the machining dimensions of the right-end step. This represents the left-end compensation direction coefficient determined by the left-end machining posture. First, it's determined which side the left-end step removal area is located on relative to the lowest point inside the profile cross-section. Then, the feed direction and rotation direction of the angle-cutting saw blade in the left machining head are determined. When the horizontal offset direction caused by the left-end machining angle after the saw blade rotation is consistent with the increasing direction of the left-end step machining dimension, the left-end compensation direction coefficient is positive. When this horizontal offset direction is opposite to the increasing direction of the left-end step machining dimension, the left-end compensation direction coefficient is negative. This represents the right-end compensation direction coefficient determined by the right-end machining posture. Based on the right-end machining posture, the position of the right-end step-cutting area relative to the common machining reference point is determined, along with the horizontal offset direction after the right-end angle cutting saw blade rotates. When this offset direction increases the actual cutting distance, the right-end compensation direction coefficient takes a positive value; when this offset direction decreases the actual cutting distance, the right-end compensation direction coefficient takes a negative value. This indicates the horizontal compensation amount at the left end. This indicates the horizontal compensation amount at the right end.

[0051] The left-end compensation direction coefficient and the right-end compensation direction coefficient are used to characterize whether the angle is positive or negative, and whether the saw blade feeds from the inside to the outside or from the outside to the inside. The horizontal compensation amount should be increased or decreased relative to the step processing size, so that the left-end horizontal cutting distance and the right-end horizontal cutting distance can be consistent with the actual angle cutting requirements.

[0052] The left-end angle cutting trajectory is generated based on the left-end horizontal cutting distance, left-end machining angle, and left-end machining posture. The right-end angle cutting trajectory is generated based on the right-end horizontal cutting distance, right-end machining angle, and right-end machining posture. This is represented as follows: ; ; in: Indicates the cutting trajectory at the left end angle; Indicates the cutting trajectory at the right end angle. This indicates the relationship for generating the cutting trajectory at the left end angle. This indicates the relationship for generating the cutting trajectory at the right end angle. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. Indicates the processing posture of the right end; Indicates the horizontal cutting distance at the left end. Indicates the horizontal cutting distance at the right end. Indicates the machining angle at the left end. This indicates the machining angle at the right end.

[0053] The double-end step machining constraint results are further converted into left-end angle cutting trajectory and right-end angle cutting trajectory, so that the angle cutting at the left and right ends can be performed according to the step removal area, machining posture, horizontal cutting distance and machining angle of the corresponding machining head respectively.

[0054] S23, based on the left-end angle cutting trajectory and the right-end angle cutting trajectory, generate the corresponding left-end step compensation cutting trajectory and right-end step compensation cutting trajectory respectively; the left-end step compensation cutting trajectory is used to compensate cutting the residual step material in the step removal area still located in the left processing head coordinate system after the left-end angle cutting; the right-end step compensation cutting trajectory is used to compensate cutting the residual step material in the step removal area still located in the right processing head coordinate system after the right-end angle cutting.

[0055] The left-end step-filling trajectories and the right-end step-filling trajectories are represented as follows: ; ; in, This indicates the cutting trajectory of the left-end step. This indicates the tangent trajectory of the step on the right. This indicates the generation relationship of the left-end step recutting trajectory. After the left-end angle cutting trajectory has been generated, the step removal area under the left processing head coordinate system is used as the limit to determine the location of the step material that may still remain after the left-end angle cutting trajectory is removed. Then, based on the left-end processing posture, the feed direction, recutting height, and recutting endpoint of the step recutting saw blade in the left processing head are determined so that the left-end step recutting trajectory can cover the residual area and form an intersection area with the left-end angle cutting trajectory within the step removal area. This indicates the generation relationship of the right-end step recutting trajectory. After the right-end angle cutting trajectory has been generated, the step removal area in the right processing head coordinate system is used as the limited range. The position of the remaining step material after the right-end angle cutting is determined by combining the right-end processing posture. The motion path of the right-end step recutting saw blade is generated according to the feed direction, rotation direction and left-right mirror relationship of the right processing head, so that the right-end step recutting trajectory and the right-end angle cutting trajectory intersect within the right-end step removal area. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

[0056] The left-end intersection region and the right-end intersection region can be represented as: ; ; in, Indicates the intersection area at the left end. Indicates the right-hand intersection area. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This indicates the cutting trajectory of the left-end step. This indicates the tangent trajectory of the step on the right. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system.

[0057] When both the left and right intersection areas are located within the corresponding step cutting area, it indicates that the angle cutting trajectory and the step supplementary cutting trajectory can be effectively connected at the corresponding step surface, thereby avoiding incomplete cutting areas or excessive repeated cutting areas on the step surface.

[0058] The combined cutting trajectory formed by the left-end angle cutting trajectory, the right-end angle cutting trajectory, the left-end step-complementary cutting trajectory, the right-end step-complementary cutting trajectory, the left-end intersection region, and the right-end intersection region is represented as follows: ; in, This indicates the combined cutting trajectory at both ends. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This indicates the cutting trajectory of the left-end step. This indicates the tangent trajectory of the step on the right. Indicates the intersection area at the left end. This indicates the intersection area at the right end.

[0059] The left-end angle cutting trajectory and the right-end angle cutting trace are extended into a double-end combined cutting trajectory with step-completion cutting capability, so that the inclined cutting surface processing and step residual removal at each end can be completed in the same step removal area.

[0060] S24. Based on the double-ended combined cutting trajectory and the length of the profile to be processed, determine the length closure relationship between the left-end angle cutting trajectory and the right-end angle cutting trajectory. The length closure relationship is used to constrain the relative positioning distance between the left and right processing heads in the length direction of the profile, so that the processing positions corresponding to the left-end angle cutting trajectory and the right-end angle cutting trajectory are consistent with the length of the profile to be processed.

[0061] Let the projection of the left-end angle cutting trajectory onto the profile length direction be denoted as the left-end trajectory length reference, and let the projection of the right-end angle cutting trajectory onto the profile length direction be denoted as the right-end trajectory length reference. Then, the length closure relationship between the left-end angle cutting trajectory and the right-end angle cutting trajectory can be expressed as: ; in, This represents the positioning reference for the left-end angle cutting trajectory along the length of the profile. This indicates the positioning reference for the right-end angle cutting trajectory along the length of the profile. Indicates the length of the profile to be processed. This indicates the distance between the left-end angle cutting trajectory and the right-end angle cutting trajectory along the length of the profile.

[0062] When the actual distance between the left-end angle cutting trajectory and the right-end angle cutting trajectory does not meet the length of the profile to be processed, the relative positioning distance between the left and right processing heads can be corrected according to the length closure deviation, as expressed as: ; in, Indicates length closure deviation. Indicates the length of the profile to be processed. This represents the positioning reference for the left-end angle cutting trajectory along the length of the profile. This indicates the positioning reference of the right-end angle cutting trajectory along the length of the profile.

[0063] Based on the length closure deviation, the relative positioning distance between the left and right processing heads is constrained so that the left and right processing heads meet the length closure relationship before performing dual-end synchronous processing. One of the processing heads is kept as the positioning reference, and the position of the other processing head is corrected along the length of the profile. Alternatively, the length closure deviation can be distributed to the left and right processing heads so that the left and right processing heads are adjusted to face each other or back to back to the position that meets the length of the profile to be processed.

[0064] By integrating the left-end angle cutting trajectory, right-end angle cutting trajectory, left-end step compensation cutting trajectory, right-end step compensation cutting trajectory, and the relative positioning constraints of the two processing heads that satisfy the length closure relationship, a dual-end linkage processing trajectory is generated as follows: ; in: Indicates the dual-end linkage processing trajectory. This indicates the combined cutting trajectory at both ends. Indicates a length closure relationship. This indicates the positioning constraints of the left machining head. This indicates the positioning constraints of the right machining head.

[0065] The dual-end combined cutting trajectory no longer only represents the cutting paths of the left and right ends, but also establishes a closed constraint with the length of the profile to be processed, so that the left processing head and the right processing head form an interconnected linkage processing relationship at both ends of the same profile.

[0066] S3, based on the dual-end linkage processing trajectory control, the left and right processing heads synchronously clamp the profile and perform end cutting, angle cutting and step compensation cutting. During the processing, the feed endpoint of the corresponding processing head is corrected according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory, so as to obtain the profile processing result with consistent double-end step surfaces.

[0067] S31, based on the obtained dual-end linkage machining trajectory, analyze the positioning constraints of the left machining head, the positioning constraints of the right machining head, the left-end angle cutting trajectory, the right-end angle cutting trajectory, the left-end step compensation cutting trajectory, and the right-end step compensation cutting trajectory; the positioning constraints of the left machining head are used to limit the clamping position, the flush cutting position, and the step removal position of the left machining head at the left end of the profile; the positioning constraints of the right machining head are used to limit the clamping position, the flush cutting position, and the step removal position of the right machining head at the right end of the profile.

[0068] The dual-end linkage processing trajectory is represented as follows: ; in, Indicates the dual-end linkage processing trajectory. This indicates the positioning constraints of the left machining head. This indicates the positioning constraints of the right machining head. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This indicates the cutting trajectory of the left-end step. This indicates the tangent trajectory of the step on the right. It indicates a length closure relationship.

[0069] After the analysis is completed, the above trajectory and positioning constraints are converted into an action sequence that the equipment can execute, according to the processing sequence of clamping, flush cutting, angle cutting and step compensation cutting, and a dual-end synchronous processing sequence is generated.

[0070] The timing sequence for dual-end synchronous processing is represented as follows: ; in, Indicates the timing sequence of synchronous processing at both ends. This indicates the sequence arrangement according to the order of clamping, flush cutting, angled cutting, and step-by-step finishing. This indicates the positioning constraints of the left machining head. This indicates the positioning constraints of the right machining head. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This indicates the cutting trajectory of the left-end step. This represents the cutting trajectory of the step on the right.

[0071] The dual-end linkage machining trajectory is further converted into a sequence of actions that the left and right machining heads can execute synchronously, so that the two machining heads do not act independently, but complete positioning, clamping, end cutting, angle cutting and step compensation cutting under the same machining sequence.

[0072] S32, based on the dual-end synchronous processing timing, controls the left processing head and the right processing head to move to their respective positioning positions. Specifically, the left processing head moves to the left end processing position of the profile according to the positioning constraint of the left processing head, and the right processing head moves to the right end processing position of the profile according to the positioning constraint of the right processing head, and makes the relative distance between the left processing head and the right processing head satisfy the aforementioned length closure relationship.

[0073] After the two processing heads on the left and right reach their corresponding positioning positions, the central clamping mechanism and the left and right end clamping mechanisms are controlled to clamp the profile to be processed simultaneously. The central clamping mechanism is used to limit the swing of the middle part of the profile during the processing, and the left and right end clamping mechanisms are used to limit the displacement of the two ends of the profile during the end-cutting and step-cutting processes.

[0074] The clamped state can be represented as: ; in, This indicates that the clamping is confirmed. This indicates the clamping status of the central clamping mechanism. This indicates the clamping status of the left-end clamping mechanism. This indicates the clamping status of the clamping mechanism at the right end. This indicates that the relationship is satisfied simultaneously.

[0075] Once the clamping confirmation state meets the processing requirements, control the saw blade to perform vertical flush cuts on the left and right ends of the profile respectively; the flush cut is used to first form the vertical end faces of the left and right ends, so that the subsequent angle cuts and step fill cuts are based on the stable end face position.

[0076] After completing the flush cut, the double-end flush positioning state is obtained, which is represented as follows: , , , , ;in, This indicates a position where both ends are aligned. This indicates the positioning constraints of the left machining head. This indicates the positioning constraints of the right machining head. This indicates that the clamping is confirmed. This indicates the state of the end face after a vertical, flush cut at the left end. This indicates the state of the end face after a vertical, flush cut on the right side.

[0077] The two processing heads on the left and right sides complete the vertical and flush cutting at both ends on the basis of the profile being stably clamped, ensuring that the subsequent angle cutting trajectory and step compensation cutting trajectory have a stable starting end face.

[0078] S33, based on the double-end alignment positioning state, controls the left processing head to perform left-end angle cutting and left-end step supplementary cutting according to the left-end angle cutting trajectory and the left-end step supplementary cutting trajectory, and controls the right processing head to perform right-end angle cutting and right-end step supplementary cutting according to the right-end angle cutting trajectory and the right-end step supplementary cutting trajectory.

[0079] During the processing, the intersection status of the angle cutting trajectory and the step compensation trajectory in the corresponding step removal area is monitored. Specifically, the left end monitors whether the left-end angle cutting trajectory and the left-end step compensation trajectory form an effective intersection in the step removal area under the left processing head coordinate system; the right end monitors whether the right-end angle cutting trajectory and the right-end step compensation trajectory form an effective intersection in the step removal area under the right processing head coordinate system.

[0080] The actual intersection states at the left and right ends are represented as follows: ; ; in, This indicates the actual intersection state at the left end. This indicates the actual intersection status on the right. This indicates the actual cutting trajectory at the left end angle during the processing. This indicates the actual right-end angle cutting trajectory executed during the processing. This represents the actual left-end step-complementary cutting trajectory executed during the processing. This indicates the actual right-end step-filling trajectory executed during the processing. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. This indicates the intersection relationship between the trajectory and the region.

[0081] Determine the intersection status, convert the actual intersection status into an intersection quantity, and compare it with the preset effective intersection range. This is represented as follows: During the left-end processing, first record or calculate the actual termination position of the left-end angle cutting trajectory, then record the actual starting position or the compensation position of the left-end step cutting trajectory. Then, along the main feed direction of the left-end step cutting area, determine if there is an overlapping segment between the two trajectories. If the termination position of the left-end angle cutting trajectory has already passed the starting position of the left-end step compensation trajectory, then an overlap is formed between them, and the length of the overlapping segment is taken as the left-end intersection amount. If the two are exactly connected, the left-end intersection amount is close to zero. If there is a gap between them, it indicates that the left-end intersection is insufficient; in this case, the left-end intersection amount can be recorded as a negative distance or as an insufficient state. First, determine the actual termination position of the right-end angle cutting trajectory in the right machining head coordinate system. Then, determine the actual starting position or coverage position of the right-end step recutting trajectory in the right machining head coordinate system. Next, calculate the overlap length of the two within the right-end step removal area along the actual feed direction of the right end. Since the right machining head is usually located at the other end of the profile, its feed direction and angle may be opposite to the left end. Therefore, the left-end direction judgment rule cannot be used directly. Instead, the cutting depth direction and recutting continuation direction should be determined by the right-end machining posture. When there is an effective overlap between the right-end angle cutting trajectory and the right-end step recutting trajectory, the overlap distance is the right-end intersection amount. When there is an uncovered gap between the two, it is judged as insufficient right-end intersection. When the overlap distance is too long, it is judged as excessive right-end intersection. in, Indicates the left-hand intersection quantity. Indicates the intersection quantity at the right end. This represents a quantitative relationship concerning the actual intersection state, expressed as the intersection overlap distance. This indicates the actual intersection state at the left end. This indicates the actual intersection status at the right end.

[0082] First, select a sample that is the same as or similar to the actual processed profile, and perform multiple sets of trial cuts according to different processing angles, different step processing dimensions, and different left and right end processing postures. Record the actual intersection amount of the angle cutting trajectory and the step compensation cutting trajectory in each set, and check whether there are any residues, overcuts, obvious burrs, or uneven cut surfaces on the processed step surface. Then, the minimum intersection amount that can obtain a qualified step surface is taken as the preset minimum intersection amount, and the maximum intersection amount that will not cause overcuts or damage to the step surface is taken as the preset maximum intersection amount. The preset minimum intersection amount and the preset maximum intersection amount together form the preset effective intersection range.

[0083] When the left-end intersection amount is less than the preset minimum intersection amount, the left-end intersection is determined to be insufficient; when the left-end intersection amount is greater than the preset maximum intersection amount, the left-end intersection is determined to be excessive; when the left-end intersection amount is between the preset minimum intersection amount and the preset maximum intersection amount, the left-end intersection is determined to be normal; the right-end intersection status is determined in the same way.

[0084] The result of the two-end intersection state judgment is expressed as follows: ;in, This indicates the result of the two-end intersection state judgment. This indicates the intersection status on the left, with values ​​of insufficient intersection, excessive intersection, or normal intersection. This indicates the intersection status on the right end, with values ​​of insufficient intersection, excessive intersection, or normal intersection.

[0085] During the execution of angle cutting and step compensation cutting, it is determined in real time or in segments whether the two trajectories form an effective connection within the corresponding step cutting area, so as to provide a basis for subsequent correction of the feed endpoint of the corresponding processing head.

[0086] S34, based on the result of the intersection state judgment at both ends, the feed endpoint of the corresponding processing head is corrected; specifically, when the intersection state at one end is insufficient, it indicates that the angle cutting trajectory and the step compensation cutting trajectory at that end are not sufficiently connected, which may result in residual material in the step removal area, so the feed endpoint of the corresponding processing head is extended; when the intersection state at one end is excessive, it indicates that the angle cutting trajectory and the step compensation cutting trajectory at that end overlap too much, which may cause overcutting or damage to the step surface, so the feed endpoint of the corresponding processing head is shortened; when the intersection state at one end is normal, it indicates that the angle cutting trajectory and the step compensation cutting trajectory can be effectively connected, so the feed endpoint of the corresponding processing head is kept unchanged.

[0087] The corrected feed endpoint for the corresponding machining head can be expressed as: ; in, Indicates the first The feed endpoint after end correction Indicates the first The feed endpoint before end correction. Indicates the first End feed correction amount Indicates the processing end, with a value of or , Indicates the left end. Indicates the right end.

[0088] No. The end feed correction amount can be determined based on the rendezvous status as follows: when hour; It is a positive correction amount; when hour; It is a negative correction amount; when hour; It is zero; in, Indicates the first End-to-end meeting volume This indicates the preset minimum intersection quantity. This indicates the preset maximum intersection quantity. Indicates the first End feed correction amount.

[0089] After obtaining the corrected feed endpoint, the left and right machining heads are controlled to complete the machining of the left and right step surfaces according to the corrected feed endpoint. After the machining is completed, the angle cutting trajectory and the step compensation trajectory reach effective intersection at both ends, so that the cutting range, cutting surface continuity and machining datum of the left and right step surfaces are consistent.

[0090] The final profile processing result with consistent stepped surfaces at both ends is represented as follows: ; in, This indicates the processing result of a profile with consistent stepped surfaces at both ends. This indicates the machining result of the left end step surface. This indicates the machining result of the right-end step surface. Indicates the state of intersection at the left end. Indicates the right-end intersection state. Indicates the feed endpoint after left-hand correction. This indicates the feed endpoint after the correction on the right.

[0091] Based on the actual intersection of the angle cutting trajectory and the step compensation cutting trajectory, the feed endpoint of the corresponding processing head can be adjusted to avoid insufficient or excessive step removal, so that the step surfaces at both ends are processed under the same processing datum and the same length closure relationship, thereby obtaining a profile processing result with consistent step surfaces at both ends.

[0092] Example 2 like Figure 2 As shown, an automated dual-head end milling control system based on deep learning includes the following modules: The end face image acquisition module is used to acquire the end face image of the profile to be processed after it enters the double-head end face milling area. The dual-end common reference positioning module is used to input the end face image into a pre-trained deep learning model, identify the lowest point inside the profile section, the upper step boundary, and the processing posture of the left and right ends, and establish a common processing reference for dual-head end milling based on the lowest point inside the profile section to obtain the dual-end common reference profile positioning result. The dual-end linkage trajectory generation module is used to generate the left-end angle cutting trajectory, the right-end angle cutting trajectory, the corresponding step compensation trajectory and the dual-end linkage processing trajectory based on the positioning results of the dual-end reference profile, the processing angles of the left and right ends, the processing dimensions of the steps of the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis. The dual-end synchronous processing control module is used to control the left and right processing heads to clamp the profile synchronously based on the dual-end linkage processing trajectory and perform end cutting, angle cutting and step compensation cutting. It also corrects the feed endpoint of the corresponding processing head according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory to obtain a profile processing result with consistent step surfaces at both ends.

[0093] Example 3 Verification objects and scenarios: Typical long profile processing scenarios were selected for simulation verification. Profile type: Aluminum alloy stepped profile, available in three lengths: 3000mm, 4500mm, and 6000mm; Processing requirements: Left end processing angle: 15°; Right end processing angle: -10°; Left end step processing dimension: 25mm; Right end step processing dimension: 30mm; The lowest point inside the profile section is used as a common processing reference. Comparison: Existing technology: Traditional dual-head end milling equipment, which uses independent zero-point calibration of the left and right heads, manual measurement and parameter input, and static angle compensation; This invention: An automated dual-head end milling control method based on deep learning, namely end face image recognition, inner lowest point with reference, length closure constraint, and dynamic correction of intersection state.

[0094] Simulation environment and test sample size: Simulation platform: Matlab / Simulink and CNC machining simulation module; Visual recognition simulation: 2000 labeled end face images, including normal, offset, burr, sawdust occlusion and other working conditions, are used for forward inference of deep learning model; Machining simulation: 200 pieces of each length specification of profile are processed, totaling 600 pieces; Evaluation indicators: length processing error, misalignment of steps at both ends, angle cutting deviation, overcut / residual occurrence rate of step surface, single-piece processing cycle time and batch processing yield.

[0095] Table 1 Comparison of Simulation Results Data The table shows the results of simulation verification using 600 profiles, three length specifications, various end face postures, and different combinations of double-end angles. This invention uses the lowest point on the inner side as a common machining reference, which solves the problems of length error and step misalignment caused by the independent zero points of the left and right heads of the traditional double-head end milling machine. The length error is reduced by more than 85%.

[0096] The angle cutting trajectory is generated based on the compensation relationship between the saw blade axis and the center of the rotation axis. The angle deviation is reduced from ±0.75° to ±0.12°, and the accuracy is improved by 84%.

[0097] The length closure constraint restricts the relative positioning of the two ends, enabling the left and right processing heads to move in tandem at both ends of the profile, increasing the batch processing CPK from 0.92 to 1.38.

[0098] Dynamically correcting the feed endpoint during the intersection reduces the overcut / residual rate from 12.3% to 1.1%, decreasing defects by more than 90%.

[0099] The overall yield rate increased from 84.2% to 98.5%, and the processing efficiency improved by 12.5%.

[0100] In summary, the present invention is significantly superior to the prior art in terms of processing accuracy, step consistency, angle accuracy and batch stability, and can effectively solve the technical problems of inconsistent datum, step misalignment and angle deviation in the processing of long profiles at both ends.

[0101] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0102] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A deep learning-based automated dual-head end milling control method, characterized in that, Includes the following steps: Based on the end face image of the profile to be processed, a deep learning model is used to identify the lowest point inside the profile section, the upper step boundary, and the processing posture of the left and right ends. A common machining datum for double-head end milling is established using the lowest point inside the profile section to obtain the positioning result of the profile with the same datum at both ends; Based on the positioning results of the double-end reference profile, the processing angles of the left and right ends, the processing dimensions of the steps at the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis; The left-end angle cutting trajectory, the right-end angle cutting trajectory, and the corresponding step compensation cutting trajectory are generated respectively. Based on the length closure relationship between the left-end angle cutting trajectory and the right-end angle cutting trajectory, a double-end linkage machining trajectory is generated. Based on the dual-end linkage processing trajectory control, the left and right processing heads synchronously clamp the profile and perform flush cutting, angle cutting and step compensation cutting; During the processing, the feed endpoint of the corresponding processing head is corrected according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory to obtain a profile processing result with consistent double-end step surfaces.

2. The automated dual-head end milling control method based on deep learning according to claim 1, characterized in that, Acquire the end face image of the profile to be processed after entering the double-head end milling area. Input the end face image into a pre-trained deep learning model. The deep learning model identifies the outer contour, inner lower boundary, upper step boundary, and end orientation of the profile in the end face image. The end face contour recognition result is represented as follows: ;in, This indicates the end face contour recognition result. This refers to a pre-trained deep learning model. This represents an image of the end face of the profile to be processed. Indicates the outer contour of the profile. Indicates the inner lower boundary. Indicates the bottom boundary of the end face. Indicates the boundary of the upper step. Indicates the direction of the end; Based on the end face contour recognition result, the intersection position of the inner lower boundary and the bottom boundary of the end face is determined, and the intersection position is taken as the lowest point inside the profile section. At the same time, the step removal area is determined according to the upper step boundary, and the processing posture of the left and right ends is determined according to the end orientation. The profile end face reference recognition result is expressed as follows: ;in, This indicates the profile end face reference identification result. Indicates the lowest point inside the profile section. Indicates the area to be cut off from the step. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

3. The automated dual-head end milling control method based on deep learning according to claim 2, characterized in that, Based on the profile end face reference identification result, the lowest point inside the profile section is converted into a common machining reference point for the dual-head end milling machine, and the step removal area and the left and right end machining postures are mapped to the left machining head coordinate system and the right machining head coordinate system, respectively. The resulting positioning result of the dual-end common reference profile is expressed as follows: ;in, This indicates the positioning results of the profiles at both ends using the same reference. This indicates the common machining reference point in the equipment's machining coordinate system. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. This indicates the processing posture of the right end.

4. The automated dual-head end milling control method based on deep learning according to claim 1, characterized in that, Based on the positioning results of the dual-end same-reference profile, the step removal area in the left machining head coordinate system, the step removal area in the right machining head coordinate system, the left-end machining posture, and the right-end machining posture are read respectively. The left-end machining angle, the right-end machining angle, the left-end step machining dimension, and the right-end step machining dimension are matched to the corresponding step removal areas respectively. The resulting dual-end step machining constraint is expressed as follows: ; ; ;in, This indicates the machining constraint result for the double-ended step. This indicates the machining constraints for the left-end step. This indicates the machining constraints for the right-end step. This represents the step removal area in the left machining head coordinate system. This represents the step removal area in the right machining head coordinate system. Indicates the processing posture of the left end. Indicates the processing posture of the right end. Indicates the machining angle at the left end. Indicates the machining angle at the right end. Indicates the machining dimensions of the left end step. This indicates the machining dimensions of the right-end step.

5. The automated dual-head end milling control method based on deep learning according to claim 4, characterized in that, Based on the double-ended step machining constraint results and the preset compensation relationship between the saw blade axis and the rotation axis center, horizontal feed compensation is performed on the left-end machining angle and the right-end machining angle respectively to obtain the left-end horizontal cutting distance and the right-end horizontal cutting distance. Based on the left-end horizontal cutting distance, the left-end angle cutting trajectory is generated as follows: The right-end angular cutting trajectory generated based on the right-end horizontal cutting distance is expressed as follows: ; Based on the left-end angle cutting trajectory and the right-end angle cutting trajectory, corresponding left-end step compensation cutting trajectory and right-end step compensation cutting trajectory are generated respectively, so that the angle cutting trajectory and the step compensation cutting trajectory in the same processing head form an intersection area in the corresponding step removal area, and the resulting double-end combined cutting trajectory is represented as follows: ;in, This indicates the combined cutting trajectory at both ends. Indicates the cutting trajectory at the left end angle. Indicates the cutting trajectory at the right end angle. This indicates the cutting trajectory of the left-end step. This indicates the tangent trajectory of the step on the right. Indicates the intersection area at the left end. This indicates the intersection area at the right end.

6. The automated dual-head end milling control method based on deep learning according to claim 5, characterized in that, Based on the dual-end combined cutting trajectory and the length of the profile to be processed, the length closure relationship between the left-end angle cutting trajectory and the right-end angle cutting trajectory is determined, and the relative positioning distance between the left and right processing heads is constrained according to the length closure relationship, generating a dual-end linkage processing trajectory as follows: ;in: Indicates the dual-end linkage processing trajectory. This indicates the combined cutting trajectory at both ends. Indicates a length closure relationship. This indicates the positioning constraints of the left machining head. This indicates the positioning constraints of the right machining head.

7. The automated dual-head end milling control method based on deep learning according to claim 1, characterized in that, Based on the aforementioned dual-end linkage machining trajectory, the positioning constraints of the left machining head, the positioning constraints of the right machining head, the left-end angle cutting trajectory, the right-end angle cutting trajectory, the left-end step compensation cutting trajectory, and the right-end step compensation cutting trajectory are analyzed. The dual-end synchronous machining timing sequence is then generated according to the machining order of clamping, flush cutting, angle cutting, and step compensation cutting as follows: ;in, Indicates the timing sequence of synchronous processing at both ends. This indicates the sequence arrangement according to the order of clamping, flush cutting, angled cutting, and step-by-step trimming; Based on the aforementioned dual-end synchronous processing sequence, the left and right processing heads are controlled to move to their corresponding positioning positions, and the central clamping mechanism and the left and right end clamping mechanisms are controlled to simultaneously clamp the profile to be processed. Then, the end-cutting saw blade is controlled to perform vertical end-cutting on the left and right ends of the profile, respectively, resulting in a dual-end end-cutting positioning state, as shown below: , , , , ;in, This indicates a position where both ends are aligned. This indicates that the clamping is confirmed. This indicates the state of the end face after a vertical, flush cut at the left end. This indicates the state of the end face after a vertical, flush cut on the right side.

8. The automated dual-head end milling control method based on deep learning according to claim 7, characterized in that, Based on the aforementioned double-end alignment positioning state, the left and right processing heads are controlled to perform angle cutting and step compensation cutting according to the left-end angle cutting trajectory, the right-end angle cutting trajectory, the left-end step compensation cutting trajectory, and the right-end step compensation cutting trajectory, respectively. During the execution, the intersection state of the angle cutting trajectory and the step compensation cutting trajectory within the corresponding step removal area is monitored, and the double-end intersection state judgment result is expressed as follows: ;in, This indicates the result of the two-end intersection state judgment. Indicates the state of intersection at the left end; This indicates the state of intersection at the right end.

9. The automated dual-head end milling control method based on deep learning according to claim 8, characterized in that, Based on the results of the double-end intersection state judgment, when the intersection is insufficient, the feed endpoint of the corresponding processing head is extended; when the intersection is excessive, the feed endpoint of the corresponding processing head is shortened; when the intersection is normal, the feed endpoint of the corresponding processing head remains unchanged, and the processing of the left and right end step surfaces is completed according to the corrected feed endpoint. The resulting profile with consistent double-end step surfaces is expressed as follows: ;in, This indicates the processing result of a profile with consistent stepped surfaces at both ends. This indicates the machining result of the left end step surface. This indicates the machining result of the right-end step surface. Indicates the state of intersection at the left end. Indicates the right-end intersection state. Indicates the feed endpoint after left-hand correction. This indicates the feed endpoint after the correction on the right.

10. A deep learning-based automated dual-head end milling control system, used to implement the deep learning-based automated dual-head end milling control method as described in any one of claims 1-9, characterized in that, Includes the following modules: An end face image acquisition module is used to acquire the end face image of the profile to be processed after it enters the double-head end face milling area. The dual-end common reference positioning module is used to input the end face image into a pre-trained deep learning model, identify the lowest point inside the profile section, the upper step boundary and the processing posture of the left and right ends, and establish a common processing reference for dual-head end face milling based on the lowest point inside the profile section to obtain the dual-end common reference profile positioning result. The dual-end linkage trajectory generation module is used to generate the left-end angle cutting trajectory, the right-end angle cutting trajectory, the corresponding step compensation trajectory and the dual-end linkage processing trajectory based on the positioning results of the dual-end reference profile, the processing angles of the left and right ends, the processing dimensions of the steps of the left and right ends, and the preset compensation relationship between the saw blade axis and the center of the rotation axis. The dual-end synchronous processing control module is used to control the left and right processing heads to simultaneously clamp the profile and perform flush cutting, angle cutting and step compensation cutting based on the dual-end linkage processing trajectory. The module also corrects the feed endpoint of the corresponding processing head according to the intersection state of the angle cutting trajectory and the step compensation cutting trajectory to obtain a profile processing result with consistent step surfaces at both ends.

Images