Grinding system, grinding method, and cast product manufacturing method
The grinding system addresses installation variations in castings by using three-dimensional measurement and adaptive trajectory generation, ensuring precise and efficient grinding of castings with defects.
Patent Information
- Application Number
- PCT/JP2025/021217
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2024-07-24
- Filing Date
- 2025-06-11
- Publication Date
- 2026-01-29
AI Technical Summary
Existing grinding systems for castings face challenges in accommodating variations in the installation state of workpieces, leading to improper installation and inefficiencies due to defects such as burrs, sand marks, and feeder marks, which conventional methods like teaching playback for articulated robots struggle to address.
A grinding system that includes a mounting section, a grinding device with a three-dimensional measurement unit, a trajectory generation processing section, and a grinding control section, which uses three-dimensional surface imaging and coordinate transformation to generate a trajectory for the grinding tool, allowing for adaptive grinding without presetting operations.
Enables efficient and precise grinding of castings with defects by accommodating variations in installation, ensuring accurate grinding paths and reaction force control, thereby improving the quality and efficiency of the grinding process.
Smart Images

Figure JP2025021217_29012026_PF_FP_ABST
Abstract
Description
Grinding system, grinding method, and manufacturing method of cast product
[0001] The present invention relates to a grinding system, a grinding method, and a manufacturing method for a cast product, for removing defects present on the surface of a workpiece such as a casting.
[0002] A known grinding system removes burrs from a casting placed on a work table by using an articulated robot equipped with a grinder to bring a grinding tool into contact with the burr while appropriately changing its posture (see, for example, Patent Document 1).
[0003] Japanese Patent Application Laid-Open No. 2019-34396
[0004] In grinding systems such as those described above, the operation of the articulated robot is generally preset to match the casting using a teaching playback method or the like. In such cases, a floating mechanism may be provided to adjust the reaction force from the workpiece to the grinding tool, as described in Patent Document 1, in order to accommodate variations in the expansion rate of the casting that occur during casting.
[0005] However, if the casting is not properly installed on a mounting portion such as a work table and shifts or tilts from the proper position, there is a risk that the floating mechanism described above will not be able to cope with the problem. In particular, the casting may have defects that occur during casting, such as not only burrs on the edges but also sand marks caused by part of the mold (sand mold) getting caught, steps or convex bulges caused by part of the mold being misaligned, and feeder marks that are marks where the feeder is connected to the feeder that controls the flow of molten metal and suppresses internal casting defects, and these defects are likely to cause improper installation.
[0006] Therefore, an object of the present invention is to provide a grinding system, a grinding method, and a manufacturing method for cast products that can accommodate variations in the installation state of workpieces such as castings on the installation area, and that do not require the operation of grinding devices such as articulated robots to be preset using a teaching playback method.
[0007] One aspect of the present invention is as follows.
[0008] [1] A grinding system comprising: a mounting section on which a workpiece is mounted; a grinding device having a grinding tool for grinding the workpiece; a three-dimensional measurement section that measures the surface of the workpiece mounted on the mounting section to obtain a three-dimensional surface image; a trajectory generation processing section that generates a trajectory that plans the path along which the grinding tool will move when performing grinding based on the three-dimensional surface image before performing the grinding; and a grinding control section that controls the grinding device so that the workpiece mounted on the mounting section is ground based on the trajectory.
[0009] [2] The grinding system described in [1], wherein the three-dimensional measurement unit has an imaging unit that images the surface of the workpiece placed on the placement unit from multiple viewpoints to obtain a primary three-dimensional surface image, and an image processing unit that synthesizes the primary three-dimensional surface images obtained from the multiple viewpoints to obtain the three-dimensional surface image.
[0010] [3] The grinding system according to [2], wherein the image processing unit obtains the three-dimensional surface image by synthesizing the primary three-dimensional surface images obtained from the plurality of viewpoints by coordinate transformation using a homogeneous transformation matrix.
[0011] [4] The grinding system described in [3], wherein the three-dimensional measurement unit images the measurement standard from each of the multiple viewpoints to obtain a three-dimensional surface image of the measurement standard, the surface of the measurement standard having a first plane facing upward, a second plane and a third plane facing each other in a first horizontal direction, and a fourth plane and a fifth plane facing each other in a second horizontal direction different from the first horizontal direction, and the image processing unit uses the three-dimensional surface image of the measurement standard obtained from a specific viewpoint included in the multiple viewpoints and the three-dimensional surface image of the measurement standard obtained from the other viewpoints to calculate the homogeneous transformation matrix used to combine the primary three-dimensional surface image obtained from a specific viewpoint included in the multiple viewpoints with the primary three-dimensional surface image obtained from the other viewpoints included in the multiple viewpoints.
[0012] [5] The grinding system according to [4], wherein the measurement standard is a cube or a rectangular parallelepiped.
[0013] [6] The grinding system according to any one of [1] to [5], wherein the trajectory has a plurality of grinding paths arranged in a vertical direction, each of which extends along the three-dimensional surface image at a constant height, and the plurality of grinding paths are used for the grinding in the order in which they are arranged upward or downward.
[0014] [7] The grinding system according to any one of [1] to [6], wherein the grinding device has a reaction force measuring unit that measures a reaction force from the workpiece to the grinding tool, and the grinding control unit controls the grinding device so that the reaction force measured by the reaction force measuring unit becomes a target value when grinding the workpiece placed on the placement unit based on the trajectory.
[0015] [8] The grinding system according to any one of [1] to [7], wherein the installation unit has a mounting surface on which the workpiece is placed and a clamp that fixes the workpiece placed on the mounting surface.
[0016] [9] The grinding system according to any one of [1] to [8], wherein the workpiece is a casting.
[0017]
[10] A grinding method comprising: an installation step of installing the workpiece on the installation section; a three-dimensional measurement step of measuring the surface of the workpiece installed on the installation section to obtain a three-dimensional surface image; a trajectory generation processing step of generating a trajectory, which is a planned path along which a grinding tool will move when performing grinding, based on the three-dimensional surface image before performing the grinding; and a grinding control step of controlling a grinding device having the grinding tool so that the workpiece installed on the installation section is ground based on the trajectory.
[0018]
[11] A method for manufacturing a cast product, comprising grinding a cast product as the workpiece by the grinding method according to
[10] .
[0019] According to the present invention, it is possible to provide a grinding system, a grinding method, and a manufacturing method for cast products that can accommodate variations in the installation state of workpieces such as castings on installation sections, and that do not require the operation of grinding devices such as articulated robots to be preset using a teaching playback method.
[0020] FIG. 5 is a system overview diagram showing a grinding system according to one embodiment of the present invention. FIG. 6 is an explanatory diagram illustrating defects in a workpiece. (a) shows a proper installation state of a workpiece without defects, (b) shows an improper installation state due to burrs, and (c) shows an improper installation state due to feeder marks. FIG. 7 is an explanatory diagram illustrating the processing content of a three-dimensional shape measuring unit in the grinding system of FIG. 1. FIG. 8 is an external view showing an example of a marker used in the three-dimensional shape measuring unit of FIG. 4. FIG. 9 is an explanatory diagram illustrating an example of coordinate transformation using the marker shown in FIG. 5. (a) shows a height Σh generated for an example of a workpiece. n 1 is a conceptual diagram showing a trajectory (grinding path) in (a), (b) is a conceptual diagram showing an extracted image of a three-dimensional surface image extracted to generate the trajectory, and (c) is a conceptual diagram showing a grinding path generated by arranging control points on the extracted image. n8(b) is a conceptual diagram showing a trajectory (grinding path) in (a), and (b) is a conceptual diagram showing a extracted image of a three-dimensional surface image extracted to generate the trajectory. (a) is a conceptual diagram showing a state in which normal estimation is applied to the extracted image of FIG. 8(b) in order to arrange control points on the extracted image, (b) is a conceptual diagram showing a state in which clustering is applied after applying normal estimation in (a), (c) is a conceptual diagram showing a state in which linear fitting is applied after applying clustering in (b), (d) is a conceptual diagram showing a state in which a machining start point is determined after applying linear fitting in (c), and (e) is a conceptual diagram showing a state in which control points are arranged in machining order from the machining start point determined in (d). An explanatory diagram explaining imaging by a three-dimensional measuring unit when a measurement standard is used in the grinding system of FIG. 1. An external view showing an example of the measurement standard of FIG. 10. (a) is a conceptual diagram showing a three-dimensional surface image obtained by imaging the measurement standard from viewpoint 1, and (b) is an explanatory diagram explaining normal estimation of the first plane in (a). 12(a) is a conceptual diagram showing the rotation of the three-dimensional surface image of FIG. 12(a) to a reference coordinate system, and FIG. 12(b) is a conceptual diagram showing the rotation of the three-dimensional surface image captured from viewpoint 2 to a reference coordinate system. (a) is a conceptual diagram showing an outline of z-direction alignment using three-dimensional surface images captured from viewpoint 1 and viewpoint 2, (b) is a conceptual diagram showing an outline of x-direction alignment using three-dimensional surface images captured from viewpoint 1 and viewpoint 2, and (c) is a conceptual diagram showing an outline of y-direction alignment using three-dimensional surface images captured from viewpoint 1 and viewpoint 2. A graph comparing the synthesis accuracy of primary three-dimensional surface images when markers are used and when measurement standards are used.
[0021] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
[0022] As shown in FIG. 1 , in one embodiment of the present invention, the grinding system 1 includes a mounting unit 3 on which a workpiece 2 is mounted, a grinding device 4 having a grinding tool 4a for grinding the workpiece 2, a 3D measurement unit 5 that measures the surface of the workpiece 2 mounted on the mounting unit 3 to obtain a 3D surface image, a trajectory generation processing unit 6 that generates a trajectory 7 based on the 3D surface image before grinding, which is a planned path along which the grinding tool 4a will move when grinding, and a grinding control unit 8 that controls the grinding device 4 to grind the workpiece 2 mounted on the mounting unit 3 based on the trajectory 7. In this embodiment, the trajectory 7 is a planned path along which the contact point of the grinding tool 4a, which is the part that comes into contact with and grinds the workpiece 2, moves when grinding. Note that the trajectory 7 is not limited to the trajectory related to the contact point of the grinding tool 4a.
[0023] In this embodiment, the workpiece 2 is a casting. Castings may have defects that arise during casting, such as burrs 2a, irregularities 2b (sand marks, steps, convex bulges, etc.), and feeder marks 2c, as shown in FIG. 2 . Therefore, as shown in FIG. 3 , these defects can easily cause improper installation on the installation section 3. For example, compared to the proper installation state shown in FIG. 3( a), misalignment as shown in FIG. 3( b) or tilt as shown in FIG. 3( c) can easily occur. The grinding system 1 of this embodiment can also be applied to workpieces 2 other than castings.
[0024] In this embodiment, the installation section 3 has a mounting surface 3a on which the workpiece 2 is placed, and a clamp 3b that clamps and fixes the workpiece 2 placed on the mounting surface 3a. In this case, for example, as shown in Figure 3(b), a burr 2a may get caught in the clamp 3b, causing a misalignment of the central axis O1 of the workpiece 2 with respect to the central axis O2 of the clamp 3b, or as shown in Figure 3(c), when a surface with a feeder mark 2c is placed on the mounting surface 3a, causing an inclination of the central axis O1 of the workpiece 2 with respect to the central axis O2 of the clamp 3b. The structure of the clamp 3b is not particularly limited.
[0025] In this embodiment, the setting unit 3 keeps the workpiece 2 set on the setting unit 3 stationary during grinding. However, the setting is not limited to this, and for example, the setting unit 3 may have a rotation drive unit that rotates the workpiece 2 set on the setting unit 3 around an axis along the vertical direction, and the grinding control unit 8 may be configured to control the timing, direction, and angle of rotation by the rotation drive unit.
[0026] 1, the grinding device 4 has a reaction force measuring unit 4b that measures the reaction force from the workpiece 2 to the grinding tool 4a, and the grinding control unit 8 controls the grinding device 4 so that the reaction force measured by the reaction force measuring unit 4b becomes a target value when grinding the workpiece 2 placed on the placement unit 3 based on the trajectory 7. The target value may be a predetermined numerical value or may be within a predetermined numerical range.
[0027] In this embodiment, the grinding device 4 includes a grinder 4c that holds the grinding tool 4a, an articulated arm 4d, and a reaction force adjustment actuator 4e that is held at the tip of the articulated arm 4d and holds the grinder 4c. The reaction force adjustment actuator 4e adjusts the reaction force from the workpiece 2 to the grinding tool 4a in accordance with the reaction force measured by the reaction force measurement unit 4b. Also in this embodiment, the grinding control unit 8 includes a grinder arm control unit 8a that controls the grinder 4c and the articulated arm 4d, and a reaction force control unit 8b that controls the reaction force adjustment actuator 4e. The grinder arm control unit 8a controls the grinder 4c and the articulated arm 4d based on inverse kinematics calculations or the like so that the grinding tool 4a moves along the generated trajectory 7. The reaction force control unit 8b controls the reaction force adjustment actuator 4e so that the reaction force measured by the reaction force measurement unit 4b becomes a target value. With regard to the movement of the grinding tool 4a in the direction of pressing against the workpiece 2, control by the reaction force control unit 8b takes precedence over control by the grinder arm control unit 8a. In this embodiment, the grinder arm control unit 8a and the reaction force control unit 8b are configured by separate computers, but they may also be configured by a common computer. In other words, the grinding control unit 8 may be configured by a single computer. The trajectory generation processing unit 6 is configured by a computer separate from the computer that configures the grinding control unit 8, but they may also be configured by a common computer.
[0028] 1 and 4, the three-dimensional measuring unit 5 has an imaging unit 5a that captures images of the surface of the workpiece 2 placed on the installation unit 3 from multiple viewpoints (four in this embodiment) to obtain a primary three-dimensional surface image, and an image processing unit 5b that synthesizes the primary three-dimensional surface images obtained from the multiple viewpoints to obtain a three-dimensional surface image. The number of viewpoints is four in this embodiment, but can be set appropriately depending on the workpiece 2, etc. In this embodiment, the imaging unit 5a is made up of multiple imaging devices 5c. The imaging devices 5c are composed of cameras, laser scanners, etc. The image processing unit 5b is composed of a computer separate from the computer that constitutes the grinding control unit 8, but may also be composed of a common computer.
[0029] In this embodiment, the image processing unit 5b synthesizes the primary three-dimensional surface images obtained from the plurality of viewpoints by coordinate transformation using a homogeneous transformation matrix to obtain a three-dimensional surface image. However, the synthesis method for obtaining the three-dimensional surface image is not limited to this.
[0030] As shown in Fig. 10, the three-dimensional measurement unit 5 captures images of the measurement standard 12 from multiple viewpoints to obtain three-dimensional surface images of the measurement standard 12. As shown in Fig. 11, the surface of the measurement standard 12 has a first plane 12a facing upward, a second plane 12b and a third plane 12c facing each other in a first horizontal direction, and a fourth plane 12d and a fifth plane 12e facing each other in a second horizontal direction different from the first horizontal direction. The image processing unit 5b uses the three-dimensional surface image of the measurement standard 12 obtained from a specific viewpoint included in the multiple viewpoints and the three-dimensional surface images of the measurement standard 12 obtained from the other viewpoints to calculate a homogeneous transformation matrix used to combine a primary three-dimensional surface image obtained from the specific viewpoint included in the multiple viewpoints with a primary three-dimensional surface image obtained from another viewpoint included in the multiple viewpoints.
[0031] In this embodiment, the measurement standard 12 is a cube or a rectangular parallelepiped, but is not limited to this.
[0032] The trajectory 7 has a plurality of grinding paths 7a arranged in a vertical direction, each extending at a constant height along the three-dimensional surface image, and the plurality of grinding paths 7a are used for grinding in the order in which they are arranged upward or downward. Note that the grinding system 1 is not limited to a configuration in which a plurality of grinding paths 7a are used as described above.
[0033] According to the above configuration, a 3D surface image is acquired and a trajectory 7 is generated each time the workpiece 2 is placed on the placement unit 3. Grinding can then be performed based on the trajectory 7. This allows for automatic removal of defects by grinding, corresponding to variations in placement conditions resulting from defects in the workpiece 2. Furthermore, measurement of the workpiece 2 placed on the placement unit 3 can be rapidly achieved by capturing and synthesizing primary 3D surface images. In this case, if homogeneous transformation matrices for each viewpoint are prepared in advance and then used to capture and synthesize images each time the workpiece 2 is placed, measurement can be achieved more quickly. Furthermore, grinding can be made more efficient by configuring the trajectory 7 to include multiple grinding paths 7a that are used in ascending or descending order. The reaction force measurement unit 4b enables higher precision and quality in grinding. While the above configuration is particularly useful when a casting as the workpiece 2 is placed using the clamp 3b, it is not limited thereto.
[0034] According to the grinding system 1 of this embodiment, the grinding method of this embodiment can be implemented as follows. However, the grinding method using the grinding system 1 of this embodiment is not limited to this. The grinding method of this embodiment produces a cast product by grinding a cast product as the workpiece 2, but is not limited to this, and may also be applied to workpieces other than cast products.
[0035] The grinding method of this embodiment includes an installation step of installing the workpiece 2 on the installation section 3, a three-dimensional measurement step of measuring the surface of the workpiece 2 installed on the installation section 3 using the three-dimensional measurement section 5 to obtain a three-dimensional surface image, a trajectory generation processing step of generating a trajectory 7, which is a planned path along which the grinding tool 4a will move when grinding, based on the three-dimensional surface image using the trajectory generation processing section 6 before grinding is performed, and a grinding control step of controlling the grinding device 4 having the grinding tool 4a using the grinding control section 8 so as to grind the workpiece 2 installed on the installation section 3 based on the trajectory 7.
[0036] The three-dimensional measurement process includes an imaging process in which the surface of the workpiece 2 placed on the installation section 3 is imaged from multiple viewpoints using the imaging section 5a to obtain a primary three-dimensional surface image, and an image processing process in which the primary three-dimensional surface images obtained from the multiple viewpoints are synthesized using the image processing section 5b to obtain a three-dimensional surface image.
[0037] In the image processing step, primary three-dimensional surface images obtained from a plurality of viewpoints are synthesized by the image processing unit 5b through coordinate transformation using a homogeneous transformation matrix to obtain a three-dimensional surface image.
[0038] The grinding method includes a homogeneous transformation matrix calculation step in which the three-dimensional measurement unit 5 captures images of the measurement standard 12 from a plurality of viewpoints to obtain three-dimensional surface images of the measurement standard 12, and the image processing unit 5b uses the three-dimensional surface image of the measurement standard 12 obtained from the predetermined viewpoint and the three-dimensional surface image of the measurement standard 12 obtained from the other viewpoint to calculate a homogeneous transformation matrix used to combine the primary three-dimensional surface image obtained from the predetermined viewpoint included in the plurality of viewpoints with a primary three-dimensional surface image obtained from another viewpoint included in the plurality of viewpoints.
[0039] The grinding method includes a reaction force measurement step in which a reaction force from the workpiece 2 to the grinding tool 4a is measured by a reaction force measurement unit 4b, and a grinding control step in which, when grinding is performed on the workpiece 2 placed on the placement unit 3 based on a trajectory 7, the grinding device 4 is controlled by a grinding control unit 8 so that the reaction force measured by the reaction force measurement step becomes a target value.
[0040] The imaging process, image processing process, path generation process, and grinding control process of this embodiment will be described in detail below. Note that each process is not limited to the following content.
[0041] <Imaging Step> In this embodiment, as shown in the left diagram of Fig. 4, the surface of the workpiece 2 is imaged from a plurality of viewpoints by the imaging unit 5a. The imaging unit 5a has an imaging device 5c at each viewpoint. In this embodiment, the number of viewpoints is four, but is not limited to this. In the drawing, a reference coordinate system x 0 , y 0 , z 0 and the coordinate system x at viewpoint 1 1 , y 1 , z1 and the coordinate system x at viewpoint 2 2 , y 2 , z 2 and the coordinate system x at viewpoint 3 3 , y 3 , z 3 and the coordinate system x at viewpoint 4 4 , y 4 , z 4 The reference coordinate system x 0 , y 0 , z 0 is a coordinate system that the grinding control unit 8 refers to when controlling the grinding device 4. The imaging device 5c can be configured with a 3D camera, and acquires three-dimensional point cloud (solid) information and color information.
[0042] Methods for capturing images from multiple viewpoints include using a tool changer to equip the articulated arm 4d with an imaging device 5c instead of the grinder 4c and reaction force adjustment actuator 4e, and capturing images while changing the posture of the articulated arm 4d, or preparing imaging devices 5c for multiple viewpoints in advance and placing them at each viewpoint position to capture images. The former method requires time for changing tools and changing the posture of the articulated arm 4d to the imaging position, but only one imaging device 5c is required, reducing equipment installation costs. The latter method requires imaging devices 5c for multiple viewpoints, which increases equipment installation costs, but shortens the work time, including imaging. These methods can be selected appropriately taking into account the cost and capacity required for the entire production system.
[0043] When using a measurement standard 12, which will be described later, each viewpoint is provided so that both the workpiece 2 and the measurement standard 12 can be imaged simultaneously. As shown in Fig. 10, in this embodiment, the measurement standard 12 is installed side by side near the workpiece 2, but the installation position of the measurement standard 12 is not particularly limited as long as it is within the range of the imaging device 5c that includes both the workpiece 2 and the measurement standard 12 from each viewpoint. Furthermore, it is preferable that the measurement standard 12 be installed in an orientation that allows images of three or more of the planes that make up the measurement standard 12 to be captured from each viewpoint.
[0044] Furthermore, when capturing images from each viewpoint, it is not necessary to simultaneously capture images of both the workpiece 2 and the measurement standard 12. For example, if the variation in the images captured at each viewpoint is small, such as by preparing imaging devices 5c for multiple viewpoints in advance and placing them at each viewpoint, it is also possible to capture an image of only the measurement standard 12, obtain a coordinate transformation matrix (homogeneous transformation matrix) for superposition (described later), and apply this to the image capture of the workpiece 2.
[0045] <Image Processing Step> As shown in Figure 4, a three-dimensional surface image of the entire workpiece 2 is obtained by combining the primary three-dimensional surface images captured from each of the viewpoints 1 to 4. Generally, a three-dimensional image is expressed as a series of three-dimensional coordinate data obtained by dividing the surface shape of the measured object using the resolution of the measuring device. Common formats include point cloud or polygon data. In this embodiment, the primary three-dimensional surface image is expressed in a point cloud format, which is a series of coordinate values for each point.
[0046] First, we will explain the process for obtaining a "homogeneous transformation matrix" for converting each point cloud from viewpoints 2 to 4 into the coordinate system of viewpoint 1. This process is basically performed once when the system is started up. Thereafter, the obtained homogeneous transformation matrix is repeatedly used to synthesize primary 3D surface images for each viewpoint.
[0047] FIG. 5 shows a marker 9 for synthesizing a primary 3D surface image and its imaging. To synthesize a primary 3D surface image from each viewpoint, the marker 9 is imaged from each viewpoint. The marker 9 is composed of a base surface 9a and two or more (three in this embodiment) feature portions 9b arranged on the base surface 9a. In this embodiment, the base surface 9a coincides with the placement surface 3a, but this is not limited to this. In this embodiment, the feature portions 9b are two rectangular parallelepipeds arranged in a staggered pattern so that their edges are in contact with the normal direction of the base surface 9a, and the color of their top surfaces is different from that of the base surface 9a. While this embodiment uses color information to identify and detect the marker 9, a configuration that does not use color information is also possible. In this case, the imaging device 5c may be configured not to acquire color information.
[0048] Figures 6(a) to 6(e) show an overview of point cloud synthesis using markers 9. Here, we explain the logic for calculating the representative points, representative normal vectors, and representative reference vectors of markers 9 at each viewpoint and using them to obtain a homogeneous transformation matrix for coordinate transformation of 3D point cloud data from one viewpoint to another. As shown in Figures 6(a) and 6(b), the "top surface" of the feature portion 9b is extracted from the marker 9 measurement point clouds at each viewpoint using an HSV color extraction filter. As shown in Figure 6(c), the average x, y, and z coordinate values of each point cloud on the top surface are used as the "reference point" for each top surface. The center of gravity of the triangle connecting the three reference points common to each imaging viewpoint is used as the "representative point" for each viewpoint. The normal vector of the plane formed by the three reference points is used as the "representative normal vector."
[0049] Here, the coordinates of the representative point of viewpoint n are (x n , y n , z n ), the representative normal vector component is (n xn , n yn , n zn ) the translation component t M (See FIG. 6(e)) is expressed by equations (1) to (4).
[0050] The angle formed by the representative normal vector n1 of viewpoint 1 and the representative normal vector n2 of viewpoint 2 is θ (see FIG. 6(d)), and the rotation axis vector is n = (n x , n y , n z ), the rotation axis vector n and angle θ are expressed by the cross product and inner product of the vectors using equations (5) and (6).
[0051] Since the angle θ and the rotation axis vector n have been determined, the rotation component R for matching the representative normal vector n2 with the representative normal vector n1 is calculated from the Rodrigues rotation formula. θM is expressed by equation (7).
[0052] The translation component t of the point group P2 of the viewpoint 2 relative to the point group P1 of the viewpoint 1 is Mand the rotation component R of the representative normal vector θM Since the homogeneous transformation matrix T M is expressed by equation (8).
[0053] In this embodiment, the viewpoint 1 is used as a reference, and the homogeneous transformation matrices T M and calculate the homogeneous transformation matrix T M The three-dimensional surface image is synthesized by transforming the points at viewpoints 2, 3, and 4 into coordinates at viewpoint 1 and superimposing them. M In order to synthesize a three-dimensional surface image after superimposing the images by the above method, a coordinate transformation is performed to eliminate deviations in the coordinate system in the rotation direction around the normal vector. This coordinate transformation is performed by referring to the relative positions of the image capturing device 5c between viewpoint 1 and viewpoints 2, 3, and 4. In this embodiment, the homogeneous transformation matrix T M To find the center of gravity and the normal vector, but not limited to this.
[0054] Furthermore, if higher accuracy is required for synthesis, improvement in synthesis accuracy can be expected by using an ICP algorithm or the like after point cloud alignment using the above-mentioned marker 9. However, since this increases the calculation load and parameter tuning, whether to apply an additional point cloud synthesis algorithm should be determined on a case-by-case basis, taking into account the cost burden and the required synthesis accuracy.
[0055] The coordinate conversion from the coordinate system of the viewpoint 1 after the synthesis to the reference coordinate system referred to by the grinding control unit 8 is performed by calculation with reference to the position and orientation of the image capture device 5c of the viewpoint 1 in the reference coordinate system. The coordinate system of the viewpoint 1 may be configured to coincide with the reference coordinate system. Alternatively, an image capture device 5c having a coordinate system that coincides with the reference coordinate system may be installed, and the coordinate systems of the viewpoints may be converted to the reference coordinate system using a marker 9 before synthesis.
[0056] Next, the image processing process when using the measurement standard 12 will be described. Figure 10 shows the measurement standard 12 and its imaging for synthesizing a primary three-dimensional surface image. To synthesize a primary three-dimensional surface image for each viewpoint, the measurement standard 12 is imaged from each viewpoint. Figure 11 shows the appearance of the measurement standard 12 used in this embodiment. The measurement standard 12 is preferably a cube or a rectangular parallelepiped, and the first plane 12a, second plane 12b, third plane 12c, fourth plane 12d, and fifth plane 12e are used in the measurement and overlay calculations. When the measurement standard 12 is a cube or a rectangular parallelepiped, the squareness and parallelism of the measurement standard 12 depend on the machining accuracy T th That is, the perpendicularity between the first plane 12a and the second plane 12b, the third plane 12c, the fourth plane 12d, and the fifth plane 12e is limited to the machining accuracy T th The parallelism between the second plane 12b and the third plane 12c is equal to or less than the machining accuracy T th The parallelism between the fourth plane 12d and the fifth plane 12e is equal to or less than the machining accuracy T th The processing accuracy T th can be set taking into consideration the depth direction measurement accuracy of the three-dimensional measurement device. Each dimension of the measurement reference 12 is an arbitrary value represented by L1, L2, and L3. L1 is the distance between the second plane 12b and the third plane 12c, L2 is the distance between the fourth plane 12d and the fifth plane 12e, and L3 is the distance between the first plane 12a and the bottom surface 12f.
[0057] When photographing the measurement standard 12 simultaneously with the workpiece 2, the dimensions L1, L2, and L3 of the measurement standard 12 are set so that each face of the measurement standard 12 can be measured from all viewpoints. Increasing the dimensions L1, L2, and L3 of the measurement standard 12 so that the number of points of the measurement standard 12 in the captured primary 3D surface image increases generally improves synthesis accuracy, but if the measurement standard 12 is significantly larger than the expected dimensions of the workpiece 2, the final synthesis accuracy may deteriorate. Furthermore, it is necessary to fit the workpiece 2 and the measurement standard 12 within the imaging range at each viewpoint. Therefore, when a representative dimension of the workpiece 2 is L, the dimensions L1, L2, and L3 of the measurement standard 12 are preferably set in a range greater than 0.05L and less than 2L. The typical dimension L of the workpiece 2 is the longest outer dimension of the workpiece 2, but is not limited to this. For example, the length of the longest side when the shape of the workpiece 2 is approximated by a rectangular parallelepiped, or the largest value of the outer dimension when a three-view drawing of the workpiece 2 exists, may be used.
[0058] 12A and 12B show an example of calculation of the surface parameters of the measurement standard 12 at the viewpoint 1. As shown in FIG. 12A, each surface of the measurement standard 12 captured at the viewpoint 1 is designated as a surface a 1 , surface b1, surface e 1 Let us consider the surface a 1 The calculation of the surface parameters is shown below.
[0059] First, as shown in FIG. 12(b), 1 An arbitrary point on the surface a is selected and the normal of that point is estimated. 1 The radius of the proximity circle when estimating the normal is the point located near the center of the surface a 1 The normal at any point is set to the surface a 1 The virtual normal vector n a1 Let n a1 = (n xa1 ,n ya1 ,n za1 ) The normal vector n of each point of the measurement standard 12 n to n n = (n xn ,n yn ,n zn), the cross product of the provisional normal and the normal of each point is calculated using equation (9).
[0060] Here, the threshold value for determining whether the virtual normal is parallel to each normal is CP th Then, CP a1n is CP th The smaller point is the point on the surface a 1 CP is defined as the point that th is preferably 0.1 or less. 1 Apply RANSAC to the group of points to obtain the parameters of the surface. 1 The parameters of are given by equation (10). Similarly, parameters are calculated for the other surfaces of viewpoint 1 and each surface of the other viewpoints.
[0061] 13-14 show an overview of the registration using metric 12. The method is divided into several steps.
[0062] Step 1: As shown in FIG. 13(a), the measurement reference point group of viewpoint 1 (hereinafter referred to as viewpoint 1 point group) is rotated to the reference coordinate system. First, the z-axis z 1 ' and face a 1 Normal vector n of a1 The z-axis of the reference coordinate system z 1 'direction representative vector n 0 and face a 1 Normal vector n of a1 The angle θ a1 is expressed by the inner product of vectors as shown in equation (11).
[0063] Next, the rotation axis vector n is expressed as n = (n x , n y , n z ), the rotation axis vector n is expressed by the cross product of the vectors as shown in equation (12).
[0064] Rotation angle θ a1 Since the rotation axis vector n is obtained, from the Rodrigues rotation formula, the surface a 1 Normal vector n of a1 is the z-axis of the reference coordinate system 1The rotation component R to match θa1 is expressed by equation (13).
[0065] From the above, surface a 1 Coordinate transformation matrix T a1 is expressed by equation (14).
[0066] Step 2: As shown in FIG. 13(b), the measurement reference point cloud of viewpoint 2 (hereinafter referred to as viewpoint 2 point cloud) is rotated to the reference coordinate system. 1 ' and face a 2 Normal vector n of a2 = (a a2 ,b a2 ,c a2 ) are rotated so that they overlap. As with viewpoint 1, the z-axis z of the reference coordinate system 1 'direction representative vector n 0 and face a 2 Normal vector n of a2 From the inner product, cross product and Rodrigues' rotation formula, the surface a 2 Normal vector n of a2 is the z-axis of the reference coordinate system 1 The rotation component R to match θa2 By finding 2 Coordinate transformation matrix T a2 is expressed by equation (15).
[0067] Next, as shown in FIG. 14(a), the viewpoint 2 point cloud is translated to obtain the z coordinates of the viewpoint 1 point cloud and the viewpoint 2 point cloud. 1 'Calculate the amount of parallel movement required for directional alignment. 1 z of point cloud 1 'The average value of the direction position is z a1ave , surface a 2 z of point cloud 1 'The average value of the direction position is z a2ave When this is the case, z 1 'direction parallel movement amount t z12 is expressed by equation (16).
[0068] Step 3: Next, as shown in FIG. 14(b), the viewpoint 2 point cloud is translated to obtain the x coordinates of the viewpoint 1 point cloud and the viewpoint 2 point cloud.1 'Calculate the amount of parallel movement required for directional alignment. Surface b 1 x of the point group 1 'The average value of the direction position is x b1ave , face c 2 x of the point group 1 'The average value of the direction position is x c2ave When this is done, x 1 'direction parallel movement amount t x12 is expressed by equation (17).
[0069] Step 4: Next, as shown in FIG. 14(c), the viewpoint 2 point cloud is translated to obtain the y coordinates of the viewpoint 1 point cloud and the viewpoint 2 point cloud. 1 'Calculate the amount of parallel movement required for directional alignment. Surface e 1 y of point cloud 1 'The average value of the y direction position e1ave , surface e 2 y of point cloud 1 'The average value of the y direction position e2ave When this is the case, y 1 'direction parallel movement amount t y12 is expressed by equation (18).
[0070] Step 5: The two viewpoint point clouds are translated by the translation amount calculated using equations (16) to (18).
[0071] Step 6: For the other viewpoints, rotation and translation are performed with respect to viewpoint 1 in the order of upward (surface a), first horizontal direction (surfaces b and c), and second horizontal direction (surfaces d and e), to obtain coordinate transformation matrices of the rotation and translation components, and the primary three-dimensional surface images of the workpiece 2 are superimposed and synthesized to obtain a three-dimensional surface image of the workpiece 2. In this embodiment, rotation and translation are performed in the order of upward, first horizontal, and second horizontal directions, but the order is not limited to this.
[0072] The measurement standard 12 is preferably a cube or a rectangular parallelepiped, but is not particularly limited as long as it has a first plane, a second plane, a third plane, a fourth plane, and a fifth plane, and may be, for example, a polyhedron such as a regular square pyramid or a regular octahedron. Furthermore, the arrangement and orientation of the measurement standard 12 are not particularly limited.
[0073] According to this image processing process, the primary three-dimensional surface images can be aligned and synthesized based on the distances of the surfaces that make up the measurement standard 12, thereby enabling the synthesis of three-dimensional surface images with high accuracy.
[0074] The results of a comparison of synthesis accuracy between the case where marker 9 was used and the case where measurement standard 12 was used are shown in Figure 15. Synthesis accuracy was evaluated by calculating the total amount of deviation of the synthesized 3D image relative to the 3D surface image acquired by the 3D measurement device without synthesis. As a result, if the total amount of deviation when marker 9 was used was set to 100%, the amount was 38% when measurement standard 12 was used, confirming a significant improvement in accuracy.
[0075] <Trajectory Generation Processing Step> A method for generating a trajectory 7 will be described with reference to Fig. 7. Here, a tilted cylindrical workpiece 2 is assumed to have a riser mark 2c at the bottom. Fig. 7(a) shows an outline of the trajectory 7.
[0076] (i) First, when the height of the mounting surface 3a is set to 0, the z value of the contact point where the grinding tool 4a comes into contact according to the number of passes is 0 Axial height h 1 ~h N Set z for each path. 0 The axial height h can be set to any value based on the effective grinding width calculated from a model formula based on experiments, setting conditions such as the pressing force, rotation speed, and feed rate of the grinding tool 4a.
[0077] Here, N represents the number of grinding passes. A plurality of control points 10 for grinding are arranged on the grinding pass 7a in the n-th pass of any variable that satisfies the formula (19).
[0078] (ii) Determine the 3D point cloud data to be referenced in generating the nth pass trajectory 7. Fig. 7(b) shows an overview of the 3D point cloud data to be referenced in generating the trajectory 7. When generating the nth pass grinding pass 7a, reference points among the point clouds whose z coordinate values satisfy the formula (20) are referenced. Here, Δh is a variable that can be set arbitrarily.
[0079] (iii) The control points 10 are arranged based on the reference point group. The control points 10 are arranged as shown in FIG. 7(c). First, the reference point group of the n-th grinding pass 7a obtained in FIG. 7(b) is 0 -y 0 The points are plotted on a plane, and a virtual circle is fitted using the x and y coordinate values of each point group using the least squares method.
[0080] If the center coordinates of a circle are (a, b) and the radius is r, the general formula of a circle is expressed by formula (21).
[0081] For each x and y coordinate value of each point in the point cloud, the squared error of the difference between the square of the distance from the center of the circle and the square of the radius is calculated, and the center coordinates (a, b) and radius r of the virtual circle are found by minimizing the sum of these squared errors.
[0082]
[0083] 8A shows an example of a trajectory 7 generated for a workpiece 2 that is a plane, which is an example of a workpiece 2 that is not cylindrical. In this case, as in the cylindrical case, grinding passes 7a are set at arbitrarily set height positions of each pass.
[0084] Fig. 8(b) shows the three-dimensional point cloud data referenced in generating the trajectory 7. As in Fig. 7(b), when generating the nth grinding pass 7a, among the point clouds, the point cloud whose z coordinate values satisfy the formula (10) is used as the reference point cloud for the nth grinding pass 7a.
[0085] The logic for generating the trajectory 7 is shown in Figures 9(a) to 9(e). First, the normals of each point cloud are estimated. As shown in Figure 9(a), the point clouds are roughly classified according to the direction of the normals. In this embodiment, the point clouds are roughly classified into four directions: x+, x-, y+, and y-. Next, as shown in Figure 9(b), the roughly classified point clouds are further classified by clustering. A common method for clustering point clouds is used, such as DBSCAN. As shown in Figure 9(c), a line fitting is performed using the least squares method on each subdivided point cloud to obtain a virtual line 11. Next, as shown in Figure 9(d), the intersections of the virtual lines 11 of all the subdivided point clouds on the xy plane are found. Here, the distance of each intersection from an appropriately set machining reference point is found, and the farthest point is set as the machining start point. The order of machining is determined in advance as either counterclockwise or clockwise, and the control points 10 are placed on all virtual straight lines 11 in the machining order at intervals of the control points 10 that can be arbitrarily set in the predetermined direction (counterclockwise in this embodiment), as shown in Figure 9(e). In this way, the trajectory 7 can be generated automatically without using the teaching playback method.
[0086] <Grinding Control Process> The grinder / arm control unit 8a sets the operation of pressing the tip point of the grinding tool 4a against each control point 10 of the generated trajectory 7 using inverse kinematics calculation. The grinder 4c and the articulated arm 4d are moved based on this set operation, allowing automatic grinding of the workpiece 2. During grinding, the reaction force generated from the workpiece 2 to the grinding tool 4a is measured by the reaction force measurement unit 4b, and the reaction force control unit 8b controls the operation of the reaction force adjustment actuator 4e so as to press or release the grinding tool 4a to achieve the target grinding reaction force. This reaction force control absorbs errors between the actual three-dimensional surface image and the trajectory 7 due to the spacing between the control points 10, enabling higher-precision, higher-quality machining.
[0087] According to this embodiment, defects occurring in the workpiece 2 such as a casting can be automatically ground and removed. Therefore, the time required for manually grinding and removing defects can be reduced, and safety can also be improved.
[0088] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and the above-described embodiments can be modified in various ways without departing from the gist of the present invention.
[0089] REFERENCE SIGNS LIST 1 Grinding system 2 Workpiece 2a Burr 2b Unevenness 2c Feeder mark 3 Installation section 3a Placement surface 3b Clamp 4 Grinding device 4a Grinding tool 4b Reaction force measurement section 4c Grinder 4d Articulated arm 4e Reaction force adjustment actuator 5 Three-dimensional measurement section 5a Imaging section 5b Image processing section 5c Imaging device 6 Trajectory generation processing section 7 Trajectory 7a Grinding path 8 Grinding control section 8a Grinder arm control section 8b Reaction force control section 9 Marker 9a Base surface 9b Feature section 10 Control point 11 Virtual straight line 12 Measurement reference 12a First plane 12b Second plane 12c Third plane 12d Fourth plane 12e Fifth plane 12f Bottom surface O1 to O2 Central axis
Claims
1. A grinding system comprising: a mounting section on which a workpiece is mounted; a grinding device having a grinding tool for grinding the workpiece; a three-dimensional measurement section that measures the surface of the workpiece mounted on the mounting section to obtain a three-dimensional surface image; a trajectory generation processing section that generates a trajectory that plans the path along which the grinding tool will move when performing grinding based on the three-dimensional surface image before the grinding is performed; and a grinding control section that controls the grinding device to perform grinding on the workpiece mounted on the mounting section based on the trajectory.
2. The grinding system described in claim 1, wherein the three-dimensional measurement unit has an imaging unit that captures images of the surface of the workpiece placed on the placement unit from multiple viewpoints to obtain a primary three-dimensional surface image, and an image processing unit that synthesizes the primary three-dimensional surface images obtained from the multiple viewpoints to obtain the three-dimensional surface image.
3. The grinding system according to claim 2, wherein the image processing unit obtains the three-dimensional surface image by synthesizing the primary three-dimensional surface images obtained from the plurality of viewpoints by coordinate transformation using a homogeneous transformation matrix.
4. The grinding system described in claim 3, wherein the three-dimensional measurement unit images the measurement standard from each of the multiple viewpoints to obtain a three-dimensional surface image of the measurement standard, the surface of the measurement standard having a first plane facing upward, second and third planes facing each other in a first horizontal direction, and fourth and fifth planes facing each other in a second horizontal direction different from the first horizontal direction, and the image processing unit uses the three-dimensional surface image of the measurement standard obtained from the specified viewpoint and the three-dimensional surface image of the measurement standard obtained from the other viewpoint to calculate the homogeneous transformation matrix used to combine the primary three-dimensional surface image obtained from a specified viewpoint included in the multiple viewpoints with the primary three-dimensional surface image obtained from the other viewpoint included in the multiple viewpoints.
5. The grinding system of claim 4, wherein the measurement standard is a cube or a rectangular parallelepiped.
6. A grinding system according to any one of claims 1 to 5, wherein the trajectory has a plurality of grinding paths arranged in a vertical direction, each of which extends along the three-dimensional surface image at a constant height, and the plurality of grinding paths are used for the grinding in the order in which they are arranged upward or downward.
7. A grinding system according to any one of claims 1 to 6, wherein the grinding device has a reaction force measuring unit that measures a reaction force from the workpiece to the grinding tool, and the grinding control unit controls the grinding device so that the reaction force measured by the reaction force measuring unit becomes a target value when grinding the workpiece placed on the placement unit based on the trajectory.
8. A grinding system according to any one of claims 1 to 7, wherein the installation section has a mounting surface on which the workpiece is placed, and a clamp for fixing the workpiece placed on the mounting surface.
9. A grinding system according to any one of claims 1 to 8, wherein the workpiece is a casting.
10. A grinding method comprising: an installation step of installing the workpiece on the installation section; a three-dimensional measurement step of measuring the surface of the workpiece installed on the installation section to obtain a three-dimensional surface image; a trajectory generation processing step of generating a trajectory that plans the path along which a grinding tool will move when performing grinding based on the three-dimensional surface image before performing the grinding; and a grinding control step of controlling a grinding device having the grinding tool so that the workpiece installed on the installation section is ground based on the trajectory.
11. A method for manufacturing a cast product, comprising grinding a cast product as the workpiece using the grinding method according to claim 10 to manufacture the cast product.
Citation Information
Patent Citations
Control method for grinder working robot
JP1994312345A
Automatic machining method for workpiece and automatic machining system for workpiece
JP2007021634A
Processing tool and processing device including the same
JP2020104180A
Suction device and conveyance system
JP2023179123A
Camera calibration method, camera calibration system, and computer program
JP2024066817A