Surface finishing device

Abstract

Surface finishing device (1), comprising: an arm (10); a tool (50) attached to a distal end of the arm (10); a force sensor (30) that detects a force exerted on the tool (50); a visual sensor (60) that captures an image of a flat surface (S) of a metal member, the flat surface (S) being formed by machining; a storage device (23) which stores data indicating a target state of the flat surface (S); and a controller (20) that performs a removal position determination process that, at least using image data of a non-finish surface acquired with the visual sensor (60) and the data indicating the target state, determines a plurality of removal positions (RP) that are located on the flat surface (S) of the member and that are separated from each other, and performs an arm control process that controls the arm (10) to sequentially perform surface removal at the plurality of determined removal positions (RP) by means of the tool (50), wherein a surface inspection agent is applied to the flat surface (S) whose image is to be detected by the visual sensor (60), wherein a metallic flat surface is rubbed against the flat surface (S), and thereby distributing the surface inspection agent over the flat surface (S) in accordance with the condition of the flat surface (S), and wherein the controller (20), based on a detection result of the force sensor (30), controls the force exerted on the tool (50) when performing the surface removal.

Description

Technical FieldThe present invention relates to a surface finish machining apparatus.Prior ArtIn the related art, a scraping method is known which is performed on an upper flat surface or the like of a plate-like member for a bed of a machine tool by attaching a scraper to a distal end of an arm of a robot or a machine tool and operating the arm. See, for example, references PTLs 1-4 of the patent literature:PTL 1: JP 2016-137 551 A,PTL 2: JP 2010-240 809 A,PTL 3: JP H07-136 843 A,Document PTL 4: JP H05-123 921 A.DE 10 2017 128 757 A1 describes a deburring device, by means of which the deburring precision is to be increased and the time for deburring is to be shortened. Here, the deburring device includes a robot that uses a deburring tool to perform a deburring process for an object that is supported and cut by a support in a machine tool. The deburring device further comprises a visual sensor and a relative movement means for causing a relative movement between the visual sensor and the object supported by the support.Furthermore, DE 10 2005 011 330 B4 describes a method for position detection of a molded part. In this case, the molded part is illuminated by means of a light source in a point or line shape at a first angle. A camera image is recorded at an angle different from this. The camera image represents a height profile which is provided as a gray scale image with different gray scale values for different molded part heights.An apparatus for predicting finishing amounts at respective parts of a component to be mounted on a machine is described in EP 3 372 342 A1. The apparatus for predicting the finishing amounts includes a machine learning means.EP 2 624 087 B1 describes a method for numerically controlled scraping of a workpiece. Deviations of an upper surface of a workpiece are scanned by means of a laser distance measuring device. A step of automatic scraping will be described. Moreover, after the automatic scraping step, the upper surface of the workpiece is scanned in an analysis step.U.S. Pat. No. 10,131,033 B2 describes a system for finishing a three-dimensional surface, having a movable arm which is coupled to a finishing tool which has a finishing surface which can apply a variable contact force against the three-dimensional surface during a finishing operation. A sensor unit for measuring an actual force vector associated with the application of the variable contact force between the finishing surface and the three-dimensional surface is provided.Furthermore, US 2006 / 0 048 364 A1 describes applying a control method in a system with a robot and a tool for machining a workpiece, which control method comprises obtaining a signal representative of a force exerted on the workpiece by the tool. It will be described to control the relative movement between the workpiece and the tool and the relative position between the tool and the workpiece using this signal.SUMMARY OF THE INVENTIONAs described above, there has been a tendency to automate scraping. However, in an example, scraping is applied to a planar surface subjected to precision machining such as milling or polishing, and scraping is performed to improve flatness of a moderately uneven surface having irregularities of 10 μm or less beyond precision of machining. Moreover, scraping is a special machining that could be performed adequately only by specialists with special knowledge and techniques based on experience. In one example, the respective specialists scrape off a plurality of points of a planar surface by using special scrapers adapted by their own experience and by relying on sensations transferred to their hands, and in addition, the specialists change the forces to be applied to the scrapers, the scraping speed, etc. according to the scraping positions by their experience and intuition. Moreover, owing to their experience and intuition, specialists change scraping positions, the type of scrapers, the forces to be applied to the scrapers, the scraping speed, etc. according to the state, size, usage, etc. of the planar surface. Those specialists having such knowledge and techniques will be older and there are also no successors.Moreover, in the manufacture of a plurality of plate-shaped members serving as beds for machine tools or the like, the states of the upper planar surfaces of the plurality of plate-shaped members are different from each other. For this reason, scraping specialists determine the scraping positions for each of the plate-shaped members manufactured and perform scraping with forces and speeds suitable for the respective positions. As described above, scraping is performed by scraping specialists who have advanced techniques based on specific experiences, and therefore it is extremely difficult to automate this process in fact in a precise manner.In view of the above-described circumstances, there is a need for a surface fine machining apparatus that enables highly precise scraping by a robot.The technical problem to be solved by the invention is thus to provide a proposal which takes into account the aforementioned need.According to the present invention, the above-described technical problem is solved by a surface finish machining apparatus having the features of claim 1.A surface finish apparatus according to an aspect of the present disclosure includes: an arm; a tool attached to a distal end of the arm; a force sensor that detects a force applied to the tool; a visual sensor that detects an image of a planar surface of a metal member, the planar surface being formed by machining; a storage device that stores data indicating a target state of the planar surface; and a controller that performs a removal position determination process that determines, at least using image data of a non-finish surface detected by the visual sensor and the data indicating the target state, a plurality of removal positions that are located on the planar surface of the member and that are separated from each other, and performs an arm control process that controls the arm to perform, by means of the tool to successively perform surface ablation at the plurality of determined ablation positions, wherein a surface inspection means is applied to the planar surface whose image is to be captured by the visual sensor, wherein a metallic planar surface is rubbed against the planar surface, and thereby the surface inspection means is distributed over the planar surface in accordance with the state of the planar surface, and wherein the controller controls, based on a detection result of the force sensor, the force applied to the tool when performing the surface ablation.Brief Description of the DrawingsFIG. 1 is a perspective view of a surface finish apparatus according to an embodiment of the present invention. FIG. 2 is a side view of an essential portion of the surface finish apparatus according to this embodiment. FIG. 3 is a block diagram of a controller of a robot of the surface finish apparatus according to this embodiment. FIG. 4 is a diagram showing a state in which a metal member is rubbed against a plate-shaped member to be subjected to surface abrasion in this embodiment. FIG. 5 shows an example of image data of non-finished surfaces, the image data relating to a surface of the plate-shaped member to be subjected to surface removal in this embodiment. FIG. 6 is a flowchart showing an example of processing by the robot controller according to this embodiment. FIG. 7 is a diagram showing examples of ablation positions determined in this embodiment. FIG. 8 shows exemplary observation data relating to the surface of the plate-shaped member subjected to surface ablation in this embodiment.DESCRIPTION OF THE EMBODIMENTA surface finish machining apparatus 1 according to an embodiment of the present invention will be described below with reference to the drawings.The surface finish device 1 according to the embodiment includes a robot 2 and a controller 20 for controlling an arm 10 of the robot 2. moreover, the surface finish device 1 includes: a force sensor 30 attached to a distal end of the arm 10 of the robot 2; a tool 50 attached to the distal end of the arm 10 of the robot 2 via the force sensor 30; and a visual sensor 60.In this embodiment, the surface finish apparatus 1 applies surface removal at multiple locations on a planar surface S, which is one of the faces in the thickness direction of a plate-like member P, as shown in FIG. 1. The plate-shaped member P is used as, for example, a bed of a machine tool. The planar surface S of the plate-shaped element P is subjected to precision machining, the purpose of which is to form a completely planar surface by milling, polishing, etc. However, even after performing such precision processing, there are many cases where the planar surface S is slightly undulated with irregularities of 10 μm or less or where a portion of the planar surface S is slightly inclined. It is preferable that said irregularities and slopes are removed to increase machining precision of the machine tool.Accordingly, a planar surface S0 in a metal member is rubbed against the planar surface S in a conventional manner as shown in Fig. 4 after a surface inspection agent is applied to almost the entire planar surface S, thereby checking the presence / absence of the irregularities or the inclination on the planar surface S. In other words, portions where the surface inspection agent is removed by rubbing on the planar surface S 0 are portions protruding from other portions.Based on the determined irregularities or inclination, for example, a specialist presses an instrument, which is a chisel-like instrument or an instrument having a flat, plate-like distal end, against the plurality of ablation positions in the planar surface S, wherein the specialist moves the instrument at each of the ablation positions by a distance of several centimeters or less, for example, only by a distance of 2 cm or less. In this case, the instrument scratches at the respective ablation positions, so that the surface ablation is carried out at the respective ablation positions. The surface removal is performed to remove a thickness of several micrometers, typically 3 μm or less, from the planar surface S. By the surface removal, the corrugated state of the planar surface S is reduced or eliminated, which is preferable for increasing the machining precision of the machine tool.On the other hand, when a mounting surface to which a ball screw holder is mounted, the flat surface S, etc. becomes completely mirror-like flat surfaces and a gap between such a surface and a counterpart of this surface is completely eliminated, no lubricating oil is present between the flat surface S or the mounting surface and the counterpart. This is not preferred because such a lack of lubricating oil becomes a cause of an operation failure such as "seizure". To prevent such an operational failure, the specialist presses, for example, an instrument which is a chisel-like instrument or an instrument having a flat, plate-like distal end, against a plurality of ablation positions in the fastening surface, wherein the specialist moves the instrument at each of the ablation positions by a distance of several centimeters or less, for example by a distance of 2 cm or less. The surface removal is performed to remove a thickness of several micrometers, typically 3 μm or less, from the planar surface S. As a result, depressions are formed in the mounting surface to act as oil sumps, which helps to reduce or prevent failure of operation such as "seizure.".As illustrated in FIG. 2, the tool 50 of this embodiment includes: a fixed portion 51 fixed to a distal end portion of the arm 10 of the robot 2 via the force sensor 30; a plate-like extension portion 52 extending from the fixed portion 51; and a flat-plate-like distal end portion 53 fixed to a distal end of the extension portion 52. In one example, the fixed portion 51 and the extension portion 52 are formed of a metal, and the distal end portion 53 is formed of a high hardness steel such as tool steel. In this embodiment, the robot 2 performs the surface removal by pressing the distal end portion 53 against the flat surface S of the plate-like member P.Although the visual sensor 60 is a 2D camera in this embodiment, it is possible to use a 3D camera.The arm 10 of the robot 2 includes a plurality of arm members and a plurality of joints. Moreover, the arm 10 includes a plurality of servomotors 11 that individually drive the plurality of joints (see FIG. 3 ). Various kinds of servomotors such as rotary or linear motors could be employed as the respective servomotors 11. The individual servomotors 11 have operation position detection means for detecting their operation positions and operation speeds, an encoder being an example of the operation position detection means. The detection values of the operation position detectors are transmitted to the controller 20.The force sensor 30 is a conventional 6-axis force sensor. The force sensor 30 is attached to a wrist flange 12 of the arm 10, as illustrated in FIG. 1. Moreover, the direction in which the Z axis of the force sensor 30 extends is parallel to the direction in which the central axis CL of the wrist flange 12 of the arm 10 extends. In this embodiment, the central axis of the force sensor 30 is aligned with the central axis CL of the wrist flange 12. In the following description, an X-axis direction, a Y-axis direction, and a Z-axis direction of the force sensor 30 shown in FIG. 1 are simply referred to as an X-axis direction, a Y-axis direction, and a Z-axis direction in some cases.The force sensor 30 detects a Z-axis direction force, an X-axis direction force, and a Y-axis direction force that act on the tool 50. In addition, the force sensor 30 also detects a torque about the Z axis, a torque about the X axis, and a torque about the Y axis that act on the tool 50. In this embodiment, a 6-axis sensor is used as the force sensor 30, but it is also possible to use a 3-axis force sensor, a 2-axis force sensor, a 1-axis force sensor, or the like.As illustrated in FIG. 3, the controller 20 includes: a processor 21 such as a CPU; a display device 22; a storage device 23 including a nonvolatile memory, a ROM, a RAM, etc.; an input device 24 that is a keyboard, a touch panel, a control panel, or the like; and a transmission / reception unit 25 for transmitting / receiving signals. The input device 24 and the transmission / reception unit 25 serve as input units.In this embodiment, the controller 20 is a robot controller provided in the robot 2. However, the controller 20 may be a computer provided in the robot controller or outside the robot controller and having the configuration described above.The storage device 23 stores a system program 23 a, wherein the system program 23 atakes over the basic functions of the controller 20. The storage device 23 also stores an operation program 23 b. The operation program 23 bis created based on a reference coordinate system of the robot 2, and serves to sequentially arrange the tool 50 attached to the distal end portion of the arm 10 at a plurality of cutting positions and orientations in the reference coordinate system.The storage device 23 also stores a surface removal program 23 c. The surface removal program 23 ccauses the tool 50 disposed at each of the scraping positions to be slid by a prescribed distance, e.g., a distance of several centimeters or less (a distance of 2 cm or less in this embodiment) using the force control, whereby the surface removal program 23 ccauses the tool 50 at each of the scraping positions to scratch the planar surface S.The storage device 23 also stores a removal position determination program 23 d. The ablation position determination program 23 dapplies image processing to the captured image data of the visual sensor 60, and determines the plurality of ablation positions in the processed image.The storage device 23 also stores a learning program 23 e. In this embodiment, the controller 20 that operates based on the learning program 23 efunctions as a learning unit; however, another computer based on the learning program 23 emay also function as a learning unit.For example, the controller 20 performs processing explained below on the basis of the operation program 23 b, the area removal program 23 c, the removal position determination program 23 d, and the learning program 23 e(FIG. 6 ).First, in the state where the plate-shaped member P is mounted or fixed on a prescribed mounting portion 70, the controller 20 transmits the image acquisition instruction to the visual sensor 60 based on the ablation position determination program 23 d(step S 1- 1). Accordingly, the controller 20 receives the image acquisition data of the non-fine machined surface acquired from the visual sensor 60. In this embodiment, the entire planar surface S of the plate-shaped member P is located in the field of view of the visual sensor 60. in the case where only a portion of the planar surface S of the plate-shaped member P is located in the field of view of the visual sensor 60, the controller 20 causes the visual sensor 60 to acquire an image of the entire planar surface S of the plate-shaped member P while the visual sensor 60 is moved. In this case, the visual sensor 60 can be moved by using a moving means of the arm 10 or the like of the robot 2.Note that the surface inspection agent is applied to the planar surface S by the visual sensor 60 before imaging, and then a planar surface S 0 of the metal portion is rubbed against the planar surface S as illustrated in FIG. 4. This work is referred to as inspection preparation in this embodiment. Due to the rubbing, the surface inspection means is removed from portions (high portions) that protrude more than other portions in the planar surface S. An example of the surface inspection agent is colored powder, such powder being referred to as red lead primer.Next, the controller 20 applies image processing to the obtained non-fine-machined surface image data as needed based on the ablation position determination program 23 d, and determines a distribution state of the surface inspection agent in the processed image (step S 1- 2). For example, as illustrated in FIG. 5, regions AR in which the surface inspection agent is absent from the planar surface S of the plate-shaped member P are detected. Note that, according to the color concentration due to the surface inspection means, a plurality of kinds of regions in the planar surface S can be recognized. In this case, in the planar surface S, a first region in which the color is lighter than a first color, a second region in which the color is lighter than a second color darker than the first color, and so on are recognized. Note that an image indicating the distribution state obtained in step S 1- 2 is also a shape of the image data of the non-fine-machined surface.Subsequently, the controller operates according to the ablation position determination program 23 d, and determines a plurality of ablation positions RP to be subjected to the surface ablation as illustrated in FIG. 7 based on the distributions of the areas AR, the first areas, the second areas, etc. in the planar surface S (S 1- 3). The plurality of ablation positions RP are separated from each other. Moreover, at this time, the controller determines the removal directions of the respective removal positions RP as indicated by arrows in FIG. 7. Note that the controller does not determine the removal directions when the removal directions are set.Note that a plurality of fine surface image data may be stored in the storage device 23 of the controller 20, the fine surface image data being obtained by the visual sensor 60 or another visual sensor by acquiring images of the state of the planar surfaces S after applying the surface removal thereto. In this embodiment, the plurality of the image data of the fine machined surface refers to the planar surfaces S of the same type as the plate-like elements P; however, the data may refer to planar surfaces of different types of plate-like elements, or the data may refer to planar surfaces of other elements. Moreover, the plurality of fine surface image data are stored with respect to the planar surfaces S which are in a good or proper ready-to-use state.Moreover, in the case of manufacturing a plurality of plate-shaped members P, the distributions of the areas AR, the first areas, the second areas, etc. in the planar surfaces S in the plurality of plate-shaped members P differ. Accordingly, the plurality of the image data of the fine-machined surface differ in the positions and the number of places where the surface removal is performed.In performing step S 1- 3, the controller 20 determines the plurality of removal positions RP to be subjected to surface removal by using, from the plurality of fine surface image data (data indicating the target state), fine surface image data that matches the distribution state of the surface inspection means acquired in step S 1- 2. The plurality of fine surface image data may individually have data relating to the distribution state of the surface inspection agent before performing the surface removal. In this case, the ablation positions RP are determined more accurately.In executing step S 1- 3, a user may input the destinations to the input device 24 as data indicating the destination state. In one example, the user enters the purpose of surface ablation as the first target. Moreover, as a second target, the user inputs a percentage of the area occupied by the areas AR in the planar surface S when the inspection preparation is applied to the planar surface S after the surface ablation and images of the planar surface S are captured using the visual sensor 60. In addition, the user inputs regions as a third target, onto which the focusing is to be carried out when the surface ablation is carried out. For example, when portions of the planar surface S such as an upper half of the planar surface S, a central portion thereof in the up-down direction, etc. in FIG. 5 are to be focused in performing the surface ablation, inputs are made to specify areas of these portions. The first target is not required when the purpose of surface removal is set in the surface finish apparatus 1. The first to third targets are stored in the storage device 23.In performing step S 1- 3, the controller 20 may determine the plurality of ablation positions RP to be subjected to the surface ablation by using the plurality of the fine machined surface image data, the first target, the second target, or the third target, or a combination thereof.Next, the controller 20 selects the specific tool 50 to be used based on the surface removal program 23 c(step S 1- 4). The surface finish machining apparatus 1 includes a tool storage unit 80 such as a tool stand, a tool cassette, or the like, and a plurality of tools 50 are stored in the tool storage unit 80. The plurality of tools 50 are different from each other in shapes, materials, etc. of the distal end portions 53 thereof. In the determination of the type of the tool 50 in step S 1- 4, the control unit 20 uses, for example, the plurality of the fine machined surface image data, the first target, the second target, or the third target, or a combination thereof, and the distribution state of the surface inspection means obtained in step S 1- 2 on the basis of the non-fine machined surface image data.Subsequently, the controller 20 controls the arm 10 of the robot 2 to mount the tool 50 selected in step S 1- 4 on the basis of the surface removal program 23 c(step S 1- 5). In order to perform this assembly, in this embodiment, a male component of a conventional auto tool changer is attached to the wrist flange 12 of the arm 10, and a female component of the auto tool changer is attached to the fixed portion 51 of each tool 50.Next, the controller 20 controls the arm 10 to sequentially arrange the distal end of the tool 50 at the plurality of ablation positions RP determined in step S 1- 3 based on the operation program 23 b(step S 1- 6). At this time, the controller 20 uses the detection results of the force sensor 30 to recognize that the distal end of the tool 50 is in contact with the planar surface S at each removal position RP, and determines that the tool 50 is disposed at each removal position RP when the contact is recognized. In step S 1- 6, the controller 20 controls the orientation of the distal end portion of the arm 10 to point the distal end of the tool 50 in the direction indicated by the arrows in FIG. 7.Next, the controller 20 makes the tool 50 move by a distance of 2 cm or less in the direction in which its distal end faces, while the controller controls the force applied to the tool 50 by using the detection results of the force sensor 30 based on the surface removal program 23 c(step S 1- 7). In step S 1- 7, the controller 20 may control the moving speed at which the tool 50 is moved. For example, the controller 20 controls the moving speed of the tool 50 to fall within a predetermined speed range.The controller 20 repeats steps S 1- 6 and S 1- 7 corresponding to the number of ablation positions RP a plurality of times (step S 1- 8), and then the controller 20 transmits the image pickup instruction to the visual sensor 60 based on the inputs to the input device 24 or the like (step S 1- 9). The inspection preparation has been applied to the planar surface S before the input into the input device 24 takes place.In addition, the controller 20 applies image processing to the image data (observation data) acquired in step S 1- 9, as necessary, and recognizes the distribution state of the surface inspection agent in the processed image (step S 1- 10). The image indicating the distribution state obtained in step S 1- 10 is also an example of the observation data. Note that the image indicating the distribution state obtained in step S 1- 10 is used in the next surface removal and thereafter as the image data of the fine machined surface. The processing performed by the controller 20 in step S 1- 10 is the same as the processing in step S 1- 2. Note that the controller 20 may evaluate the flatness of the planar surface S subjected to the surface removal using the observation data obtained in step S 1- 10.FIG. 8 shows an example of the planar surface S on which the surface removal has been performed. The surface inspection means tends to accumulate in surface ablation marks RM formed due to performing the surface ablation at the respective ablation positions RP. Accordingly, in step S 1- 10, it is also possible for the controller 20 to recognize the distribution state of the surface inspection agent by ignoring the surface inspection agent in the surface removal marks RM.Note that the surface removal marks RM in FIG. 8 are separated from each other; however, the surface removal marks RM may overlap each other.The controller 20 stores, in the storage device 23, the image of the distribution state of the surface inspection agent obtained as the non-finished surface image data in step S 1- 2 and the image of the distribution state of the surface inspection agent obtained as the observation data in step S 1- 10 in a state where the two images are associated with each other (step S 1- 11).The stored observation data is used as the fine machined surface image data in step S1-3 at the next surface removal and thereafter.The controller 20 operates according to the learning program 23 e, and performs the learning to determine the plurality of ablation positions RP in step S 1- 3 in performing the next surface ablation (step S 1- 12). At this time, the controller 20 uses the non-fine machined surface image data and the observation data stored in the storage device 23, as well as one of the plurality of fine machined surface image data, the first target, the second target, and the third target, or a combination thereof. The plurality of the fine machined surface image data, the first target, the second target, or the third target, or a combination thereof, are data indicating the target states as described above.For example, the distribution state of the surface inspection agent in the observation data is evaluated with respect to the data indicating the target state. For example, in the case where the small area AR in a lower left portion of FIG. 5 has sufficiently expanded by the surface removal, but the large areas AR in the upper and lower right portions of FIG. 5 have not sufficiently expanded by the surface removal, the surface removal with respect to the large areas AR in the upper and lower right portions is judged to be insufficient. In the case where this evaluation is made, the controller increases, as a result of learning, the number of the ablation positions RP in the case where the area AR is large and / or increases the area where the ablation positions RP are arranged. To obtain this learning result, the separation distances between the regions AR are also taken into account. Moreover, the shapes of the respective surfaces AR, the positions of the respective surfaces AR, the color depth of the surface inspection agent in the areas surrounding the respective surfaces AR, etc. could also be taken into account.Note that, in step S 1- 12, the controller 20 can easily evaluate the distribution state of the surface inspection agent in the observation data.Note that the controller 20 may also operate according to the learning program 23 e, and perform learning to optimize the force applied to the tool when performing the surface removal using the non-fine machined surface image data and the observation data. In the observation data, there are cases where large amounts of the surface inspection agent have accumulated in end portions of the surface removal marks RM. Large level differences formed in these end portions act as a cause of large amounts of the surface inspection agent accumulated in the end portions of the surface removal marks RM in this manner. The level differences relate to the quantity which has to be scraped off during the surface removal.Accordingly, as an example of the optimization using the non-fine machined surface image data and the observation data, the controller may increase or decrease the force applied to the tool 50 when the surface is removed.Note that, in step S 1- 12, it is also possible to estimate the dimensions of the level differences based on the accumulated amounts of the surface inspection agent, and judge whether or not the individual marks of the surface removal (surface removal marks RM) are appropriate from the estimation results.In addition, the controller 20 may also operate according to the learning program 23 e, and perform learning to optimize the moving speed of the tool 50 when performing surface removal using the non-fine-machined image data and the observation data. A slow speed of movement of the tool 50 is a conceivable cause of an increase in the dimensions of the height differences. Accordingly, as an example of the optimization, the controller 20 may increase or decrease the speed at which the tool 50 is moved in performing the surface removal using the non-fine machined surface image data and the observation data.Moreover, the controller 20 may also operate in accordance with the learning program 23 e, and perform learning regarding optimum tools 50 in accordance with the situation, from the non-fine-machined surface image data and the observation data. There are cases where the tool 50 used in the actual passage of surface removal, which is unsuitable for the flat surface S, is one of the causes for an increase in the dimensions of the level differences. For example, there are cases where the dimensions of the level differences excessively increase, and therefore, there are cases where the ablation marks RM have unintended shapes. Such situations may result from the influence of the machining roughness or the like of the planar surface S. The machining roughness of the planar surface S changes due to wear or the like of an instrument used for machining the planar surface S or an instrument used for polishing the planar surface S. Since these surface states manifest themselves in the image data of the non-finish surface, the controller 20 can recognize whether or not the tool 50 is appropriate with respect to the surface states shown in the image data of the non-finish surface of the current pass.Note that the user can input information on the observation data via the input device 24. The user is, for example, a scraping specialist or a person with sufficient experience and knowledge about scraping, and can accurately judge the state of the planar surface S after scraping. The user inputs, to the input device 24, an acceptable / unacceptable determination, causes of unacceptable cases, etc. with respect to the observation data obtained in step S 1- 10. The controller 20 stores the input information about the observation data and this observation data in association with each other in the storage device 23.Note that the image data of the planar surface S processed by a scraping specialist may be included in the plurality of the image data of the finely machined surface stored in the storage device 23.In this embodiment, using the image data of the non-fine machined surface obtained from the visual sensor 60 and the data indicating the target states such as the first target, the plurality of ablation positions RP in the planar surface S are determined, and the surface ablation is sequentially performed at the plurality of ablation positions RP by the tool 50 at the distal end of the arm 10. In addition, using the detection results of the force sensor 30, the controller 20 controls the force applied to the tool 50 in performing the surface removal.With this configuration, since the positions at which the surface removal is performed are automatically determined, it is possible to determine the positions at which the surface removal is performed even when a person familiar with this processing, such as a scraping specialist, is not present. Moreover, since the force applied to the tool 50 is controlled, it is possible to accurately perform the surface removal at, for example, a depth of 3 μm or less.Moreover, in this embodiment, it is determined whether or not the state of the planar surface S after performing the surface removal is appropriate and / or whether or not the states of the surface removal marks RM after performing the surface removal are appropriate based on the observation data. Accordingly, it can be determined whether or not the planar surface S is usable after the surface removal even if a person familiar with this processing, such as a scraping specialist, is not present.Moreover, the observation data in this embodiment is image data obtained by capturing images of the planar surfaces S after surface ablation using the visual sensor 60 or another visual sensor. As described above, the states of the planar surfaces S of the plurality of plate-shaped members P are different from each other, and the distributions and numbers of the removal positions RP in the respective plate-shaped members P are also different. In other words, even if there are two planar surfaces S in which the total surfaces of the regions AR after the surface removal are the same, the shapes, the numbers, etc. of the regions AR appearing in the two planar surfaces S are different from each other. Accordingly, using the image data of the planar surfaces S after being subjected to the surface removal as observation data results in accurate determination of the states of the planar surfaces S after being subjected to the surface removal.Note that the observation data may be data obtained by measuring the surface shapes of the planar surfaces S by a surface roughness measurement or the like.Moreover, in this embodiment, the learning for determining the plurality of ablation positions RP is performed by using at least the non-fine machined surface image data, the data indicating the target states, and the observation data as learning data. For example, the use of distribution states of the surface inspection agent in the planar surfaces S after the surface ablation as observation data during the teaching makes it possible to determine specific portions of the planar surfaces S in which there is an excess / insufficient surface ablation.Moreover, as described above, the states of the planar surfaces S of the plurality of plate-shaped members P are different from each other, and also the distributions and numbers of the removal positions RP are different in the respective plate-shaped members P. For this reason, it is not possible to achieve uniform flatness in the planar surfaces S of the plurality of plate-shaped members P even if the same surface removal is performed at the same locations. Moreover, it is not possible for the scraping specialists to pass the knowledge and feel that the scraping specialists themselves possess to other persons in detail. One of the causes that such a situation creates is that the techniques and sensory experiences to be applied differ in the respective plate-shaped elements P, and thus the scraping tools are also different in the respective specialists. Using the distribution states of the surface inspection agent in the planar surfaces S after the surface removal as learning data for the machine learning enables the controller 20 to obtain data corresponding to the knowledge and the senses possessed by the scraping specialists by repeating the learning. This is extremely useful in order to be able to process areas with the aid of a machine in which the limit of the processing precision of a machine is exceeded and in which the techniques of the specialist had to be relied on up to now.Moreover, in this embodiment, the learning for optimizing the force to be applied to the tool 50 in performing the surface removal is performed by using at least the image data of the non-fine machined surface and the observation data as learning data. As described above, it is not possible for the scraping specialists to give the experience and feel of the scraping specialists to other persons in detail. Whether or not the surface removal marks RM are appropriate depends on the locations where the surface removal is performed; for example, the surface removal needs to be applied more to areas much higher than the other areas, and the surface removal needs to be applied less to areas slightly higher than the other areas. It is difficult to express such adjustments by numerical values, which is also one of the causes causing the above-described situation.For example, when the image data of the non-fine machined surface and the distribution states of the surface inspection agent in the planar surfaces S after the surface removal are used as learning data, it is possible for the controller 20 to obtain data corresponding to the experience and the feeling of the scraping specialists by repeating the learning.Similarly, in this embodiment, the learning for optimizing the moving speed of the tool 50 is performed in performing the surface removal by using at least the image data of the non-fine machined surface and the observation data as learning data. Also, with respect to the moving speed, it is possible for the controller 20 to obtain data corresponding to the experience and the feeling of the scraping specialists by repeating the teaching as in the case of the force to be applied to the tool 50.In this embodiment, the controller 20 selects the tool 50 to be attached to the distal end of the arm 10 by using at least the image data of the non-fine machined surface and the data indicating the target states. Since the extension portion 52 of the tool 50 has a relatively long, thin shape and the distal end portion 53 is relatively thin, a portion of the tool 50 sometimes easily distorts when the surface removal is performed. There are cases where this warping is also necessary to accurately perform the surface removal, and the warping characteristics are different in the respective tools 50. Note that other characteristics also differ among the respective tools 50.The configuration in which the tool 50 is automatically selected as described above makes it possible to determine the positions at which the surface removal is performed even when a person familiar with this machining, such as a scraping specialist, is not present.Note that another computer can perform the learning. For example, a host computer connected to a plurality of controllers 20 may store the learning program 23 e. In this case, the non-fine machined surface image data, the observation data, the fine machined surface image data, the data related to the operation of the arms 10 in performing the surface removal, etc. are transmitted from the controllers 20 to the host computer, and the host computer performs the above-described learning using the received data.Moreover, instead of attaching the tool 50 to the arm 10 of the robot 2, the tool 50 may be attached to an arm of a machining tool. In this case, too, similar functional effects to those described above are obtained.List of reference characters1 Surface finish machining apparatus 2 Robot 10 Arm 11 Servomotor 20 Controller 23 Storage 23 cSurface removal program 23 d Abtrag position determination program 23 e Lern program 30 Force sensor 50 Tool 51 Fixed portion 52 Extension portion 53 Distal end portion 60 Visual sensor P Plate-shaped member S Planar surface

Claims

A surface finish apparatus (1) comprising: an arm (10); a tool (50) attached to a distal end of the arm (10); a force sensor (30) that detects a force applied to the tool (50); a visual sensor (60) that detects an image of a planar surface (S) of a metal member, the planar surface (S) being formed by machining; a storage device (23) that stores data indicating a target state of the planar surface (S); and a controller (20) that performs a removal position determination process that determines a plurality of removal positions (RP) using at least image data of a non-finish surface detected by the visual sensor (60) and the data indicating the target state, The beams which are located on the planar surface (S) of the element and which are separated from each other, and perform an arm control operation which controls the arm (10) to sequentially perform, by the tool (50), surface ablation at the plurality of predetermined ablation positions (RP), wherein a surface inspection agent is applied to the planar surface (S) whose image is to be captured by the visual sensor (60), wherein a metallic planar surface is rubbed against the planar surface (S), and thereby the surface inspection agent is distributed over the planar surface (S) in accordance with the state of the planar surface (S), and wherein the controller (20) controls, based on a detection result of the force sensor (30), the force applied to the tool (50) in performing the surface ablation.The surface finish apparatus according to claim 1, wherein, based on observation data of a state of the planar surface (S) on which the surface removal has been performed by the tool (50), at least one of the following cases is determined: whether or not the planar surface (S) on which the surface removal has been performed is in an appropriate state; and whether or not a mark (RM) formed as a result of the surface removal is in an appropriate state.The surface finish apparatus according to claim 2, wherein the observation data is image data obtained by acquiring an image of the planar surface (S) on which the surface removal has been performed with the aid of the visual sensor (60) or another visual sensor (60).The surface finish apparatus according to claim 2 or 3, further comprising a learning unit that performs learning for determining the plurality of ablation positions (RP) by using, as the learning data: at least the image data of the non-finish surface; the data indicating the target state; and the observation data.The surface finish machining apparatus according to claim 2 or 3, further comprising a learning unit that performs learning for optimizing the force to be applied to the tool (50) in performing the surface removal by using, as the learning data, at least the image data of the non-finish surface and the observation data.The surface finish machining apparatus according to claim 2 or 3, further comprising a learning unit that performs learning for optimizing a moving speed of the tool (50) in performing the surface removal by using, as the learning data, at least the image data of the non-finish surface and the observation data.The surface finish apparatus according to any one of claims 1 to 6, further comprising a tool storage (80) that stores a plurality of tools (50), wherein the controller (20) performs, using at least the image data of the non-finish surface and the data indicating the target state, a tool selection process of selecting a tool (50) to be attached to the distal end of the arm (10), and a tool replacement process of controlling the arm (10) to attach the selected tool (50) to the distal end of the arm (10).

Images