Surface finishing device

Abstract

The present invention provides a surface finishing device, comprising: a robot arm equipped with a tool; a force sensor for detecting the force applied to the tool; a visual sensor for photographing a plane formed by processing on a metal component; a storage device for storing data representing a target state of the plane; and a control device for performing the following processing: a removal position determination processing for determining a plurality of removal positions on the plane of the component using at least unfinished photographic data obtained by the visual sensor and data representing the target state, the plurality of removal positions being separated from each other; and a robot arm control processing for controlling the robot arm to perform surface removal on the plurality of determined removal positions in sequence using the tool, apply a surface inspection agent on the plane photographed by the visual sensor, and rub a flat surface of the metal on the plane, thereby distributing the surface inspection agent on the plane according to the state of the plane.

Description

Technical Field

[0001] The invention relates to a surface finishing device. Background Art

[0002] Conventionally, there is known a technique in which a scraping tool is attached to the tip of a robot or a robotic arm of a processing machine, and the robotic arm is operated to scrape the upper surface of a plate-like member used as a machine bed for a machine tool, etc. For example, see Patent Documents 1 to 4.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 2016-137551

[0006] Patent Document 2: Japanese Patent Application Laid-Open No. 2010-240809

[0007] Patent Document 3: Japanese Patent Application Laid-Open No. 07-136843

[0008] Patent Document 4: Japanese Patent Application Laid-Open No. 05-123921 Summary of the Invention

[0009] Problems to be solved by the invention

[0010] As mentioned above, attempts are underway to automate scraping. However, in one example, scraping involves adjusting the flatness of a surface that has been precision-machined using milling, grinding, or other methods, and has gently undulating surfaces with an irregularity of less than 10 μm, with a precision exceeding that of mechanical processing. Furthermore, scraping is a specialized process that can only be properly performed by craftsmen with specialized knowledge and skills based on experience. In one example, each craftsman uses a special scraping tool customized based on their own experience, relying on the feel of the scraping tool to cut multiple parts of the surface. Based on experience and intuition, they vary the force applied to the scraping tool, the cutting speed, and other factors depending on the location to be cut. Furthermore, craftsmen vary the cutting position, type of scraping tool, the force applied to the scraping tool, and the cutting speed based on experience and intuition, depending on the state, size, and purpose of the surface. Craftsmen with this knowledge and skills are gradually aging, yet no successors are being trained.

[0011] Furthermore, when manufacturing multiple plate-like components for machine tool bases, the top surfaces of these components often differ. Therefore, scraping technicians must determine the desired location for each plate-like component and apply the appropriate force and speed for each location. As mentioned above, scraping requires skilled craftsmen with specialized experience and advanced skills, making it extremely difficult to automate with high precision in practice.

[0012] In view of the above-mentioned actual situation, a surface finishing device capable of realizing high-precision scraping processing by a robot is desired.

[0013] Solutions for solving problems

[0014] According to one aspect of the present application, a surface finishing apparatus includes: a robot arm; a tool mounted on a front end of the robot arm; a force sensor for detecting a force applied to the tool; a vision sensor for capturing an image of a plane formed by machining on a metal member; a storage device for storing data representing a target state of the plane; and a control device for executing the following processing: a removal position determination process for determining a plurality of removal positions on the plane of the member using at least unfinished image data obtained by the vision sensor and data representing the target state, the plurality of removal positions being separated from one another; and a robot arm control process for controlling the robot arm to sequentially perform surface removal at the plurality of determined removal positions using the tool, apply a surface inspection agent to the plane captured by the vision sensor, and rub a flat metal surface against the plane to distribute the surface inspection agent on the plane according to the state of the plane, wherein the control device uses the detection result of the force sensor to control the force applied to the tool during the surface removal. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a perspective view of a surface finishing device according to one embodiment of the present invention.

[0016] Figure 2 It is a side view of the main parts of the surface finishing device of this embodiment.

[0017] Figure 3 This is a block diagram of a control device for a robot of a surface finishing device according to the present embodiment.

[0018] Figure 4 This is a diagram showing a state in which a metal member is rubbed against a plate-like member whose surface has been removed in this embodiment.

[0019] Figure 5This is an example of unfinished image data regarding the surface of the plate-like member on which surface removal is performed in this embodiment.

[0020] Figure 6 This is a flowchart showing an example of processing by the robot control device according to this embodiment.

[0021] Figure 7 3 is a diagram showing an example of a removal position determined in this embodiment.

[0022] Figure 8 This is an example of observation data on the surface of the plate-like member subjected to surface removal in this embodiment. DETAILED DESCRIPTION

[0023] Hereinafter, a surface finishing device 1 according to an embodiment of the present invention will be described with reference to the drawings.

[0024] A surface finishing apparatus 1 according to one embodiment includes a robot 2 and a control device 20 for controlling a robot arm 10 of the robot 2. The surface finishing apparatus 1 also includes a force sensor 30 mounted on the front end of the robot arm 10 of the robot 2; a tool 50 mounted on the front end of the robot arm 10 of the robot 2 across the force sensor 30; and a vision sensor 60.

[0025] In this embodiment, the surface finishing device 1 is Figure 1 The surface of the plate-like member P shown in the figure, one side in the thickness direction, i.e., plane S, is subjected to surface removal in multiple locations. The plate-like member P is used, for example, as a machine base. Plane S of the plate-like member P is precision-machined by milling, grinding, or other processes to achieve a completely flat surface. However, even after such precision machining, plane S often has gentle undulations less than 10 μm in size, and sometimes a portion of plane S is slightly tilted. To improve the machining accuracy of the machine tool, a plane without these undulations and inclinations is preferred.

[0026] Therefore, in the past, after applying the surface inspection agent to almost the entire surface of the plane S, Figure 4 As shown, the flat surface S0 of the metal member is rubbed against the plane S to inspect whether the unevenness or inclination exists on the plane S. That is, the portion where the surface inspection agent is removed by rubbing against the flat surface S0 becomes a portion that protrudes relative to other portions.

[0027] For example, based on the observed unevenness or inclination, the craftsman presses a chisel-shaped tool or a tool with a flat tip against multiple removal locations on plane S. At each removal location, the craftsman moves the tool a distance of a few centimeters or less, for example, 2 cm or less. This scrapes the tool against each removal location, removing the surface at each location. This surface removal is performed by removing a few micrometers, typically 3 μm or less, from plane S. This surface removal reduces or eliminates the undulations in plane S, which is advantageous for improving machining accuracy.

[0028] On the other hand, when the mounting surface on which the ball screw bracket is mounted, such as the plane S, is a completely mirror-like plane, and the gap between it and the corresponding side member completely disappears, there is no lubricating oil between the plane S and the corresponding side member, and between the mounting surface and the corresponding side member. This causes work defects such as sticking, and is not preferred. In order to eliminate such work defects, the craftsman pushes a chisel-shaped tool or a tool with a flat front end, for example, against multiple removal positions on the mounting surface, and at each removal position, the craftsman moves the tool a distance of less than a few centimeters, for example, less than 2 cm. This surface removal is performed in a manner that removes a thickness of several micrometers, typically less than 3 μm, from the plane S. By performing this surface removal, a recessed portion that functions as an oil pit is formed on the mounting surface, which helps to reduce or prevent work defects such as sticking.

[0029] like Figure 2 As shown, the tool 50 of this embodiment includes a portion to be fixed 51, which is fixed to the front end of the robot arm 10 of the robot 2 via the force sensor 30; a plate-shaped extension portion 52 extending from the portion to be fixed 51; and a flat plate-shaped front end portion 53 fixed to the front end of the extension portion 52. In one example, the portion to be fixed 51 and the extension portion 52 are made of metal, while the front end portion 53 is made of high-hardness steel such as tool steel. In this embodiment, the robot 2 presses the front end portion 53 against the flat surface S of the plate-shaped member P to perform the surface removal.

[0030] In this embodiment, the visual sensor 60 is a two-dimensional camera, but a three-dimensional camera may also be used.

[0031] The robot arm 10 of the robot 2 includes a plurality of robot arm members and a plurality of joints. In addition, the robot arm 10 includes a plurality of servo motors 11 (see Figure 3As each servo motor 11, various servo motors such as rotary motors and linear motors can be used. Each servo motor 11 has a working position detection device for detecting its working position and working speed. As an example, the working position detection device is an encoder. The detection value of the working position detection device is transmitted to the control device 20.

[0032] The force sensor 30 is a well-known six-axis force sensor. Figure 1 As shown, the force sensor 30 is fixed to the wrist flange 12 of the robot arm 10. In addition, the extension direction of the Z axis of the force sensor 30 is parallel to the extension direction of the central axis CL of the wrist flange 12 of the robot arm 10. In this embodiment, the central axis of the force sensor 30 is consistent with the central axis CL of the wrist flange 12. In the following description, Figure 1 The X-axis direction, the Y-axis direction, and the Z-axis direction of the force sensor 30 shown are simply referred to as the X-axis direction, the Y-axis direction, and the Z-axis direction.

[0033] The force sensor 30 detects the Z-axis force, the X-axis force, and the Y-axis force applied to the tool 50. Furthermore, the force sensor 30 detects the moment about the Z-axis, the X-axis force, and the Y-axis force applied to the tool 50. In this embodiment, a six-axis sensor is used as the force sensor 30, but a three-axis force sensor, a two-axis force sensor, a one-axis force sensor, or the like may also be used.

[0034] like Figure 3 As shown, the control device 20 includes a processor 21 such as a CPU; a display device 22; a storage device 23 including nonvolatile memory, ROM, RAM, etc.; an input device 24 such as a keyboard, touch panel, or operation panel; and a transceiver 25 for transmitting and receiving signals. The input device 24 and the transceiver 25 function as input devices. The control device 20 is connected to the force sensor 30 and each servo motor 11.

[0035] In the present embodiment, the control device 20 is a robot control device provided in the robot 2. Alternatively, the control device 20 may be a computer having the above-described structure provided inside or outside the robot control device.

[0036] The storage device 23 stores a system program 23a, which assumes the basic functions of the control device 20. The storage device 23 also stores an action program 23b. The action program 23b is created based on the reference coordinate system of the robot 2 and is used to sequentially position the tool 50 attached to the tip of the robot arm 10 at a plurality of removal positions and in a plurality of postures within the reference coordinate system.

[0037] The storage device 23 also stores a surface removal program 23c. The surface removal program 23c controls the force of the tool 50 disposed at each removal position while pressing it down by a predetermined distance, for example, a distance of several centimeters or less (a distance of 2 cm or less in this embodiment), thereby scraping the flat surface S at each removal position with the tool 50.

[0038] The storage device 23 further stores a removal position determination program 23d. The removal position determination program 23d performs image processing on the image data captured by the visual sensor 60 and determines the plurality of removal positions in the processed image.

[0039] The storage device 23 further stores a learning program 23e. In the present embodiment, the control device 20 operating based on the learning program 23e functions as the learning unit, but another computer operating based on the learning program 23e may function as the learning unit.

[0040] The control device 20 performs the following processing based on the operation program 23b, the surface removal program 23c, the removal position determination program 23d, and the learning program 23e, for example: Figure 6 ).

[0041] First, with the plate-shaped member P placed or fixed on a predetermined placement portion 70, the control device 20 sends a capture command to the vision sensor 60 based on the removal position determination program 23d (step S1-1). The control device 20 then receives unfinished image data obtained by the vision sensor 60. In this embodiment, the entire plane S of the plate-shaped member P is within the field of view of the vision sensor 60. If only a portion of the plane S of the plate-shaped member P is within the field of view of the vision sensor 60, the control device 20 captures the entire plane S of the plate-shaped member P while moving the vision sensor 60. In this case, the vision sensor 60 can be moved by a moving device such as the robot arm 10 of the robot 2.

[0042] In addition, before the image is taken by the visual sensor 60, a surface inspection agent is applied on the plane S, and then Figure 4 As shown, the flat surface S0 of the metal member is rubbed against the plane S. This operation is referred to as inspection treatment in this embodiment. By rubbing, the surface inspection agent is removed from portions of the plane S that protrude (higher portions) relative to other portions. In one example, the surface inspection agent is a colored powder, such as a bright powder.

[0043] Next, the control device 20 performs image processing on the obtained unfinished photographic data based on the removal position determination program 23d as needed, and detects the distribution state of the surface inspection agent in the processed image (step S1-2). Figure 5 As shown, an area AR where the surface inspection agent is absent is detected on plane S of the plate-like member P. Alternatively, multiple areas may be detected on plane S based on the color density of the surface inspection agent. In this case, a first area lighter than a first color and a second area lighter than a second color darker than the first color are detected on plane S. Furthermore, the image representing the distribution obtained in step S1-2 is also an aspect of the unprocessed captured data.

[0044] Next, the control device 20 performs an operation according to the removal position determination program 23d, based on the distribution of the range AR, the first range, the second range, etc. on the plane S, as shown in FIG. Figure 7 As shown in FIG. 1 , a plurality of removal positions RP (S1-3) are determined for surface removal. The plurality of removal positions RP are separated from each other. In addition, at this time, the control device 20 also Figure 7 As shown by the arrows, the removal direction of each removal position RP is determined. In addition, when the removal direction is fixed, the control device 20 does not determine the removal direction.

[0045] Furthermore, the plurality of post-processing image data obtained by capturing the state of the plane S after surface removal using the vision sensor 60 or another vision sensor can also be stored in the storage device 23 of the control device 20. In this embodiment, the plurality of post-processing image data relates to the plane S of the same type of plate-like member P, but can also relate to the planes of different types of plate-like members or the planes of other members. Furthermore, the plurality of post-processing image data is stored for the plane S in a state that is not problematic for use.

[0046] When a plurality of plate-like members P are manufactured, the distribution of the range AR, the first range, the second range, etc. on the plane S differs among the plurality of plate-like members P. Therefore, the surface removal positions and amounts differ among the plurality of post-processing imaging data.

[0047] When executing step S1-3, the control device 20 also uses, among the plurality of post-processing imaging data (data representing the target state), the post-processing imaging data suitable for detecting the distribution state of the surface inspection agent detected in step S1-2 to determine the plurality of removal positions RP where surface removal should be performed. The plurality of post-processing imaging data may each include data on the distribution state of the surface inspection agent before surface removal. In this case, the removal positions RP can be determined more accurately.

[0048] When executing step S1-3, the operator may also input a target into the input device 24 as data representing the target state. In one example, the operator inputs the purpose of surface removal as the first target. In addition, as the second target, the operator performs the inspection process on the plane S after the surface removal, and when the plane S is photographed by the visual sensor 60, the operator inputs the range occupied by the range AR on the plane S in percentage. In addition, the operator inputs the range where the surface removal should be focused as the third target. For example, when focusing on Figure 5 When removing a portion of plane S, such as the upper half of plane S or the middle portion in the vertical direction, input is performed to specify the portion. If the purpose of surface removal in surface finishing apparatus 1 is determined, the first target is not necessary. The first to third targets are stored in storage device 23.

[0049] When executing step S1 - 3 , the control device 20 may also use a plurality of post-finishing imaging data, one of the first target, the second target, and the third target, or a combination thereof to determine a plurality of removal positions RP where surface removal should be performed.

[0050] Next, the control device 20 determines which tool 50 to use based on the surface removal program 23c (step S1-4). The surface finishing apparatus 1 includes a tool storage unit 80 such as a tool placement table and a tool box. The tool storage unit 80 stores a plurality of tools 50. The plurality of tools 50 differ in shape, material, etc., of the tip portion 53. When determining the tool in step S1-4, the control device 20 uses, for example, a plurality of post-finishing imaging data, one or a combination of the first target, the second target, and the third target, and the distribution of the surface inspection agent obtained based on the unfinished imaging data in step S1-2.

[0051] Next, the control device 20 controls the robot arm 10 of the robot 2 based on the surface removal program 23c to attach the tool 50 determined in step S1-4 to the robot 2 (step S1-5). To perform this attachment, in this embodiment, a male component of a known automatic tool changer is fixed to the wrist flange 12 of the robot arm 10, and a female component of the automatic tool changer is fixed to the portion to be attached 51 of each tool 50.

[0052] Next, the control device 20 controls the robot arm 10 based on the action program 23b to sequentially position the front end of the tool 50 at the multiple removal positions RP determined in step S1-3 (step S1-6). At this time, the control device 20 uses the detection results of the force sensor 30 to detect the contact between the front end of the tool 50 and the plane S at each removal position RP. When contact is detected, it is determined that the tool 50 is configured at the removal position RP. In step S1-6, the control device 20 controls the posture of the front end of the robot arm 10 so that the front end of the tool 50 faces Figure 7 Direction of arrow shown.

[0053] Next, the control device 20 controls the force applied to the tool 50 using the detection results of the force sensor 30 based on the surface removal program 23c, while moving the tool 50 a distance of 2 cm or less in the direction of its tip (step S1-7). In step S1-7, the control device 20 may also control the movement speed of the tool 50. For example, the control device 20 controls the movement speed of the tool 50 so that it is within a predetermined speed range.

[0054] The control device 20 repeats steps S1-6 and S1-7 a number of times corresponding to the number of removal positions RP (step S1-8), and the control device 20 sends an image capture instruction to the visual sensor 60 (step S1-9) based on input from the input device 24. Before the input to the input device 24 is performed, the inspection process is performed on the plane S.

[0055] In addition, the control device 20 implements image processing to the shooting data (observation data) obtained by step S1-9 as needed, and detects the distribution state of the surface inspection agent in the processed image (step S1-10). The image representing the distribution state obtained in step S1-10 is also an example of observation data. In addition, the image representing the distribution state obtained in step S1-10 is used as post-finishing shooting data in subsequent surface removal. The processing of the control device 20 in step S1-10 is the same as the processing in step S1-2. In addition, the control device 20 can use the observation data obtained in step S1-10 to evaluate the flatness of the plane S that has been surface removed.

[0056] Figure 8 FIG2 shows an example of a plane S after surface removal. The surface inspection agent is likely to remain in the surface removal marks RM formed by surface removal at each removal position RP. Therefore, in step S1-10, the control device 20 can also detect the distribution of the surface inspection agent while ignoring the surface inspection agent in the surface removal marks RM.

[0057] In addition, Figure 8The surface removal marks RM are separated from each other, but the surface removal marks RM may also overlap with each other.

[0058] The control device 20 stores the image of the distribution state of the surface inspection agent obtained as unfinished imaging data in step S1-2 and the image of the distribution state of the surface inspection agent obtained as observation data in step S1-10 in a state in which the two are associated in the storage device 23 (step S1-11). The stored observation data is used as the post-finishing imaging data in step S1-3 in the subsequent surface removal.

[0059] The control device 20 operates according to the learning program 23e and learns the plurality of removal positions RP for determining the next surface removal step S1-3 (step S1-12). At this time, the control device 20 uses the unfinished image data and observation data stored in the storage device 23, the plurality of finished image data, the first target, the second target, and the third target, or a combination thereof. As described above, the plurality of finished image data, the first target, the second target, and the third target, or a combination thereof, is data representing the target state.

[0060] For example, the distribution state of the surface inspection agent in the observation data is evaluated for the data representing the target state. Figure 5 The smaller area AR in the lower left is fully enlarged by surface removal, but Figure 5 If the larger area AR above and below the right of the image is not sufficiently expanded through surface removal, the surface removal associated with the larger area AR above and below the right of the image will be evaluated as insufficient. In this evaluation, as a result of learning, the control device 20 will increase the number of removal positions RP and / or expand the range used to configure the removal positions RP when the area AR is large. Based on the learning results, the distance between the areas AR must also be considered. In addition, the shape of each area AR, the position of each area AR, the color concentration of the surface inspection agent surrounding each area AR, and other factors must also be considered.

[0061] Furthermore, in step S1 - 12 , the control device 20 may evaluate only the distribution state of the surface inspection agent in the observation data.

[0062] Furthermore, the control device 20 operates according to the learning program 23e and can also use the unfinished image data and the observation data to perform learning to optimize the force applied to the tool during surface removal. The observation data may indicate that a large amount of surface inspection agent remains at the end of the surface removal mark RM. One reason for this large amount of surface inspection agent remaining at the end of the surface removal mark RM is the large height difference formed at this end. This height difference is related to the amount of scraping during surface removal.

[0063] Therefore, as an example of the optimization, the control device 20 can increase or decrease the force applied to the tool 50 when performing surface removal using the unfinished image data and the observation data.

[0064] Furthermore, in step S1-12, the size of the height difference may be estimated based on the amount of the surface inspection agent retained, and the pass / fail status of each surface removal mark (surface removal mark RM) may be determined using the estimation result.

[0065] Furthermore, the control device 20 operates according to the learning program 23e and can use the unfinished image data and the observation data to perform learning to optimize the movement speed of the tool 50 during surface removal. One of the reasons for the increased height difference is believed to be the slow movement speed of the tool 50. Therefore, as an example of such optimization, the control device 20 can use the unfinished image data and the observation data to speed up or slow down the movement of the tool 50 during surface removal.

[0066] In addition, the control device 20 performs operations according to the learning program 23e, and can use the unfinished shooting data and the observation data to learn about the best tool 50 corresponding to the situation. One of the reasons for the increase in the height difference may be that the tool 50 used for the surface removal this time is not suitable for the plane S. For example, there may be a tendency for the height difference to become too large, or the surface removal mark RM may become an undesirable shape. The above situation may affect the processing roughness of the plane S, etc. The processing roughness of the plane S will change due to the deterioration of the tool used to process the plane S or the tool used to grind it. In addition, the extension direction of the processing mark is different depending on the position of the plane S. Since the above-mentioned surface state appears in the unfinished shooting data, the control device 20 can perform the following learning: for the surface state appearing in the unfinished shooting data this time, it is judged whether the tool 50 used is appropriate.

[0067] In addition, the operator can input information related to the observation data from the input device 24. For example, the operator is a craftsman who performs scraping processing or a person who has sufficient experience and knowledge in scraping processing, and can accurately evaluate the state of the plane S after scraping processing. The operator inputs the judgment of whether it is qualified or unqualified, the reason for the unqualified situation, etc. into the input device 24 for the observation data obtained in step S1-10. The control device 20 associates the input information related to the observation data with the observation data and stores it in the storage device 23. The information related to the observation data is an example of learning data, which is used for learning related to the number and position of the removal position RP, learning related to the force applied to the tool 50, and learning related to the moving speed of the tool 50.

[0068] Furthermore, the plurality of post-finishing imaging data stored in the storage device 23 may include post-finishing imaging data of the plane S processed by the craftsman performing the scraping process.

[0069] In this embodiment, unfinished image data obtained by the vision sensor 60 and data indicating the target state, such as the first target, are used to determine multiple removal positions RP on the plane S. Surface removal is then performed sequentially at the multiple removal positions RP using the tool 50 at the tip of the robot arm 10. Furthermore, the control device 20 uses the detection results of the force sensor 30 to control the force applied to the tool 50 during surface removal.

[0070] According to this configuration, the position where the surface is to be removed can be automatically determined, so even without a skilled scraping craftsman, the position where the surface is to be removed can be determined. In addition, since the force applied to the tool 50 is controlled, surface removal can be accurately performed to a depth of, for example, 3 μm or less.

[0071] In addition, in this embodiment, the acceptance of the surface S after the surface removal is performed and / or the acceptance of the state of the surface removal mark RM is determined based on the observation data. Therefore, even if there is no person who is proficient in the scraping process, such as a craftsman, it can be determined whether the surface S after the surface removal is usable.

[0072] Furthermore, in this embodiment, the observation data is captured using the visual sensor 60 or another visual sensor to capture the surface-removed plane S. As described above, the state of the plane S of multiple plate-like members P varies, and the distribution and number of removal positions RP also differ for each plate-like member P. In other words, even if two planes S have the same total area of ​​the surface-removed areas AR, the shapes and number of the areas AR appearing on the two planes S will differ. Therefore, using the captured data of the surface-removed plane S as observation data helps accurately determine the state of the surface-removed plane S.

[0073] Furthermore, the observation data may be data obtained by measuring the surface shape of the plane S through surface roughness measurement or the like.

[0074] Furthermore, in this embodiment, at least the unfinished image data, the data representing the target state, and the observation data are used as learning data to perform learning for determining multiple removal positions RP. For example, using the distribution of the surface inspection agent on the plane S after surface removal as observation data, it is possible to determine during learning which portion of the plane S has experienced excessive or insufficient surface removal.

[0075] In addition, as described above, the states of the planes S of the plurality of plate-like components P are different from each other, and the distribution and number of removal positions RP also differ for each plate-like component P. Therefore, even if the same surface removal is performed at the same location, the flatness of the planes S of the plurality of plate-like components P cannot be made uniform. In addition, the craftsman who performs the scraping process cannot accurately convey his or her knowledge and feeling to others. One of the reasons for this situation is that the technology and feeling applicable to each plate-like component P are different, and the tools used by each craftsman for the scraping process are also different. When the distribution state of the surface inspection agent on the plane S after surface removal is used as learning data for machine learning, the control device 20 can obtain data corresponding to the knowledge and feeling of the craftsman who performs the scraping process by repeated learning. This is extremely useful for realizing the following technology, that is, using a machine to process an area that previously exceeded the processing accuracy limit of the machine and had to rely on the craftsman's technology.

[0076] Furthermore, in this embodiment, at least the unfinished image data and observation data are used as learning data to optimize the force applied to the tool 50 during surface removal. As mentioned above, scraping craftsmen cannot accurately convey their knowledge and intuition to others. The acceptance of the surface removal mark RM varies depending on the area being removed. For example, areas much higher than other areas require deeper surface removal, while areas slightly higher require shallower surface removal. This adjustment is difficult to express numerically, which is one of the reasons for the aforementioned situation.

[0077] For example, when unfinished image data and the distribution state of the surface inspection agent on the plane S after surface removal are used as learning data, the control device 20 can obtain data corresponding to the knowledge and feeling of the scraping craftsman through repeated learning.

[0078] Similarly, in this embodiment, at least the unfinished image data and observation data are used as learning data to optimize the movement speed of the tool 50 during surface removal. Regarding the movement speed, as well as the force applied to the tool 50, the control device 20 can acquire data corresponding to the knowledge and intuition of the scraping craftsman through repeated learning.

[0079] In this embodiment, the control device 20 uses at least the unfinished image data and data indicating the target state to determine the tool 50 to be mounted on the tip of the robot arm 10. The tool 50 has a relatively long and slender extension portion 52 and a relatively thin tip portion 53. Therefore, during surface removal, a portion of the tool 50 may be slightly bent. This bending is sometimes necessary for accurate surface removal, and the characteristics of this bending vary from tool to tool 50. Furthermore, other characteristics also vary from tool to tool 50.

[0080] As described above, according to the configuration of automatically selecting the tool 50 , even without a person who is well versed in the scraping process, such as a craftsman, the position where the surface is to be removed can be determined.

[0081] Alternatively, the learning process may be performed by another computer. For example, a learning program 23e may be stored in a host computer connected to a plurality of control devices 20. In this case, unfinished image data, observation data, post-finished image data, and data related to the movement of the robot arm 10 during surface removal are transmitted from each control device 20 to the host computer, and the host computer performs the learning process using the received data.

[0082] In addition, instead of attaching the tool 50 to the robot arm 10 of the robot 2, the tool 50 may be attached to the robot arm of a processing machine. In this case as well, the same operational effects as described above can be achieved.

[0083] Description of Reference Numerals

[0084] 1: Surface finishing device

[0085] 2: Robot

[0086] 10: Robotic Arm

[0087] 11: Servo motor

[0088] 20: Control device

[0089] 23: Storage device

[0090] 23c: Surface Removal Procedure

[0091] 23d: Remove location determination program

[0092] 23e: Learning Program

[0093] 30: Force sensor

[0094] 50: Tools

[0095] 51: To be fixed

[0096] 52: Extension setting part

[0097] 53: Front end

[0098] 60: Vision Sensor

[0099] P: Plate-like member

[0100] S: Flat

Claims

1. A surface finishing device comprising: Robotic arm; A tool, the tool being mounted on the front end of the robot arm; a force sensor configured to detect a force applied to the tool; a visual sensor that images a plane formed by processing on a metal member; a storage device for storing data representing a target state of the plane; as well as a control device that executes the following processing: a removal position determination processing that determines a plurality of removal positions on the plane of the component using at least the unfinished photographic data obtained by the vision sensor and the data representing the target state, the plurality of removal positions being separated from each other; and a robot arm control process for controlling the robot arm to sequentially perform surface removal on the plurality of determined removal locations using the tool, A surface inspection agent is applied to the plane imaged by the vision sensor, and a metal flat surface is rubbed on the plane to distribute the surface inspection agent on the plane according to the state of the plane. The control device controls the force applied to the tool during the surface removal using the detection result of the force sensor. The control device determines whether a state of a trace at each of the plurality of removal locations where the surface removal of the tool has been performed is acceptable based on observation data of a state of the plane where the surface removal of the tool has been performed.

2. The surface finishing device according to claim 1, wherein: The observation data is photographed data obtained by photographing the plane after the surface removal using the visual sensor or other visual sensors.

3. The surface finishing device according to claim 1, wherein: The surface finishing device includes a learning unit configured to perform learning for determining the plurality of removal positions using at least the unfinished imaged data, the data indicating the target state, and the observation data as learning data.

4. The surface finishing device according to claim 1, wherein The surface finishing device includes a learning unit configured to perform learning for optimizing a force applied to the tool when performing the surface removal, using at least the unfinished imaging data and the observation data as learning data.

5. The surface finishing device according to claim 1, wherein The surface finishing device includes a learning unit configured to perform learning to optimize a moving speed of the tool when performing the surface removal, using at least the unfinished imaging data and the observation data as learning data.

6. The surface finishing device according to any one of claims 1 to 5, wherein: The surface finishing device further includes a tool storage portion, wherein the tool storage portion holds a plurality of tools. The control device performs the following processing: a tool determination processing of determining a tool to be mounted on the front end of the robot arm using at least the unfinished imaged data and data indicating the target state; and a tool exchange process to control the robot arm to install the determined tool on the front end of the robot arm.

Images