JP2020019087A - Grinding tool abrasive plane evaluation device and learning equipment thereof, evaluation program and evaluation method - Google Patents

Abstract

To make it possible to evaluate surface quality of a grinding tool work plane with good efficiency and high accuracy.SOLUTION: In a learning phase, abrasive plane image data before processing of a grinding tool is acquired and stored, abrasive plane image data after use for processing of the grinding tool is acquired, a deterioration part such as clogging, damage and so on is detected from the image data, and an image before processing, of which a position corresponds to the deterioration part, is extracted from stored image data before processing. The extracted image before processing, a feature amount of individual abrasive grains and a feature amount which indicates a distribution state thereof are inputted to CNN as information which represents grinding performance reduction factor, and information, which represents an evaluation result for information which represents the grinding performance reduction factor, is given to the CNN as teacher data, thereby causing the CNN to learn and generating CNN for abrasive plane evaluation.SELECTED DRAWING: Figure 3

Description

  One embodiment of the present invention relates to an apparatus for evaluating a grinding surface of a grinding tool, a learning device, an evaluation program, and an evaluation method used in the apparatus.

  In the grinding process, the finish quality of the machined surface of the workpiece depends on the state of the grinding tool working surface (also referred to as a grinding surface) of the grindstone. The state of the abrasive surface is mainly determined by the shape and size of the surface portion of the abrasive grain that functions as a cutting edge, and the amount of protrusion of the abrasive grain. Therefore, various techniques for measuring and evaluating the fixed coordinates, the area, the shape, and the like of the individual abrasive grains dispersed on the abrasive surface of the grinding tool have been studied.

  For example, by acquiring the three-dimensional image of the working surface of the grinding tool while managing the position of the industrial camera by servo control and performing image processing on the acquired three-dimensional image, the number of abrasive grains involved in grinding and the like Techniques for numerically measuring shapes and the like have been proposed.

  Using this technique, for example, when measuring the surface of a grinding tool 10 mm in width and 250 mm in diameter to which diamond abrasive grains having an average particle diameter of 100 μm are fixed, about 70,000 abrasive grains are present in the entire area of the tool working surface. The results of analysis of the shape of the individual abrasive grains, which are fixed, can be output as CSV data. The time required from the setting of the grinding tool to the measuring device to the end of the measuring process is about 5 minutes.

  However, in the grinding process, in which the workpiece is finely and tinyly abraded by the abrasive grains scattered on the grinding surface of the grinding tool, the workpiece has a desired shape in a form in which the distribution state of the abrasive grains is transferred to the grinding surface of the workpiece. Finished. For this reason, in order to perform high-precision machining in actual grinding, it is necessary to appropriately manage the distribution of abrasive grains on the working surface of the grinding tool (hereinafter referred to as surface texture) rather than the state of individual abrasive grains. is important.

  Therefore, conventionally, a method of evaluating the surface properties of the working surface of the grinding tool using, for example, an electron microscope or a laser has been proposed (for example, see Non-Patent Documents 1, 2, or 3).

Sakamoto, Sasaki, Kobayashi, Shimizu, "Identification of Specific Grinding Force Based on On-machine Measured Wheel Work Surface Profile", Proc. Of JSAP Spring Meeting 2013, pp. 539-540 Kakino, Matsubara, Yamaji, Matsuda, Nakagawa, Hirogaki, Kida, "Research on On-the-Machine Measurement of Grinding Stone Work Surface Topography (1st Report)", Journal of Precision Engineering, vol.63, No.2, pp. 228-232 (1997) Shoji and Zhou, "Study on Truing and Dressing of Diamond Wheel (2nd Report)", Journal of the Japan Society of Precision Engineering, Vol.55, No.12, pp. 2267-2272 (1989)

  However, in the method using an electron microscope, the surface properties of the grinding tool working surface can be clearly observed, but since the observation range for one time is very narrow, it takes a long time to observe the surface properties of the entire grinding tool working surface. It costs. In addition, the method using a laser can relatively easily and accurately measure the surface properties of the working surface of the grinding tool. It is difficult to distinguish irregularities from the surface of the base material to which the abrasive grains are fixed. Further, in order to increase the measurement accuracy, it is necessary to reduce the laser spot diameter, so that only one width of about 1 mm can be measured by a single evaluation, and a long time is required for the evaluation.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a technology that enables efficient and highly accurate evaluation of a surface property of a working surface of a grinding tool.

  In order to solve the above-mentioned problems, a first aspect of a grinding tool grinding surface evaluation device according to the present invention is a state in which a plurality of abrasive grains of the grinding tool are dispersed while moving the grinding tool in a predetermined direction. And an image acquisition unit that acquires image data obtained by imaging the grinding surface fixed by the binder, and stores the image data acquired before the start of machining of the grinding tool by the image acquisition unit in a storage medium. A storage control unit, a detection unit that detects information indicating a deteriorated portion on the grinding surface of the grinding tool from image data acquired after the processing of the grinding tool by the image acquisition unit, and a storage unit that is stored in the storage medium. An extraction unit configured to extract an image of the unprocessed part corresponding to the position of the degraded part from the image data before the processing is started as information indicating a factor of deterioration in grinding performance.

  According to a second aspect of the grinding tool grinding surface evaluation device according to the present invention, the detection unit detects clogging of the abrasive grains on the grinding surface, An image representing at least one of the sites of occurrence of upward slip where the amount of protrusion of the abrasive grains from the binder is less than or equal to a threshold value is detected as information representing the deteriorated site.

  According to the first and second aspects, for example, a clogged abrasive grain, a damaged abrasive grain, a deteriorated portion such as a portion where an upper slip occurs on the abrasive surface are detected from the image data of the abrasive surface captured after the start of the grinding process. Then, an image before processing corresponding to the deteriorated portion and the position is extracted as information indicating a factor of reduction in grinding performance. Therefore, it is possible to detect factors that cause the deterioration of grinding performance such as clogging or damage of abrasive grains and the occurrence of upper slip, as reflecting not only the state of individual abrasive grains but also the distribution state of abrasive grains. Becomes

  In a third aspect of the grinding tool grinding surface evaluation device according to the present invention, the extracting unit extracts, from the image of the unprocessed part corresponding to the deteriorated part and the position, extracted as information indicating the grinding performance reduction factor. A feature amount reflecting the moving direction of the abrasive tool contained in the image and the grinding tool is calculated, and the calculated feature amount is included in the information representing the grinding performance reduction factor.

  According to the third aspect, a characteristic amount reflecting the moving direction of the grinding tool of the abrasive grain, which is a factor of lowering the grinding performance, is calculated, and this characteristic amount is added to the information indicating the above-mentioned factor of lowering the grinding performance. For this reason, the distribution state of the abrasive grains, which is a cause of the decrease in the grinding performance, can be represented not only by the image but also by the feature amount thereof, whereby the cause of the decrease in the grinding performance can be represented with higher accuracy.

  A fourth aspect of the grinding tool grinding surface evaluation device according to the present invention is to provide a learning device with information representing the grinding performance reduction factor as learning data and information representing a deteriorated portion of the grinding surface as teacher data. The apparatus further includes a learning device generation unit for learning the learning device.

  According to the fourth aspect, the learning process is performed by the learning device using the information representing the grinding performance reduction factor as learning data (explanatory variables) and the information representing the state of the deteriorated portion of the grinding surface as teacher data (objective variables). Done. For this reason, when the information indicating the grinding performance deterioration factor is input, a learning device that outputs information indicating the state of the deteriorated portion of the grinding surface can be generated.

In the present invention, for example, the following fifth, sixth, and seventh aspects can be considered as the learning unit generator.
In a fifth aspect, the learning device is provided with, as the learning data, an image including an unprocessed part corresponding to a deteriorated part and a position of the grinding surface, and the feature amount of abrasive grains included in the image, and The learning device is made to learn by giving an image representing the deteriorated part as the teacher data.

  According to the fifth aspect, when the image including the unprocessed portion whose position corresponds to the deteriorated portion of the grinding surface and the feature amount of the abrasive grains included in the image are input, the state in which the unprocessed portion is deteriorated is input. It is possible to generate a learning device that outputs the image shown.

  In a sixth aspect, the learning device is configured to provide, as the learning data, an image including a part before processing whose position corresponds to the deteriorated part and the feature amount of abrasive grains included in the image, and The learning device is made to learn by giving, as the teacher data, an image representing a part before processing corresponding to a position.

  According to the sixth aspect, when the image including the unprocessed portion corresponding to the deteriorated portion and the position of the abrasive surface and the feature amount of the abrasive grains included in the image are input, the unprocessed portion corresponds to the deteriorated portion and the position. Can be generated.

  In a seventh aspect, the learning device is provided with, as the learning data, an image including a part before processing whose position corresponds to the deteriorated part and the feature amount of abrasive grains included in the image, and The learning device is made to learn by giving, as the teacher data, numerical data indicating the degree to which the portion before processing corresponding to the position causes the grinding performance to be reduced.

  According to the seventh aspect, when the image including the unprocessed part corresponding to the deteriorated part and the position and the feature amount of the abrasive grains included in the image are input as the learning data, the part before the processing becomes the grinding performance. It is possible to generate a learning device that outputs numerical data indicating the degree of reduction.

  An eighth aspect of the grinding tool grinding surface evaluation device according to the present invention is configured to input image data acquired by the image acquisition unit to the learning device before or after machining of the grinding tool, and The information processing apparatus further includes an evaluation unit that outputs information representing a deteriorated portion of the grinding surface output from the learning device as information representing an evaluation result of the grinding surface of the grinding tool.

  According to the eighth aspect, just by inputting image data obtained before or after the processing of the grinding tool to the learning device, the information indicating the deteriorated portion of the grinding surface is obtained from the learning device by the grinding device. The information is output as information indicating the evaluation result of the grinding surface of the tool. In other words, by using the learning device, it is possible to evaluate the surface texture that is a cause of a decrease in the grinding performance of the grinding surface before and after the processing is started in a short time and with high accuracy. For this reason, the person in charge of evaluation can perform quality control before shipping or after starting use of the grinding tool based on the output evaluation result.

  A ninth aspect of the grinding tool grinding surface evaluation device according to the present invention is the evaluation unit, wherein, based on the information output from the learning unit and indicating the deteriorated surface of the grinding surface, the grinding unit A visualized image representing the distribution of the moving direction of the grinding tool in the entire area of the grinding surface of the tool is generated and output.

  According to the ninth aspect, a visualized image representing the distribution state of the movement direction in the entire area of the grinding surface of the portion expected to be a cause of deterioration in grinding performance is generated and output. For this reason, the production manager of the grinding tool or the use manager of the grinding tool, for each grinding tool, the distribution state of the portion that causes a reduction in the grinding performance of the grinding surface over the entire area of the grinding surface and movement of the grinding tool It is possible to easily and quickly grasp along the direction.

  In a tenth aspect of the grinding tool grinding surface evaluation device according to the present invention, the dressing time of the grinding tool is determined by the evaluation unit based on information indicating a deteriorated portion of the grinding surface output from the learning device. Is estimated, and information representing the estimation result is output.

  According to the tenth aspect, the estimation unit outputs the estimation information of the dressing time of the grinding tool. For this reason, for example, the manufacturing manager of the grinding tool or the manager of the use of the grinding tool can perform the shipping verification of the grinding tool and manage the dressing time of the grinding tool based on the estimated information.

  That is, according to each aspect of the present invention, it is possible to provide a technology that enables efficient and highly accurate evaluation of the surface property of the working surface of the grinding tool.

FIG. 1 is a block diagram showing an overall configuration of an inspection system using a grinding tool abrasive surface evaluation device according to an embodiment of the present invention. FIG. 2 is a diagram showing an example of a relationship between a ring-type grindstone, which is an example of a grinding tool to which the present invention can be applied, and a camera. FIG. 3 is a block diagram showing a functional configuration of the grinding tool grinding surface evaluation device according to one embodiment of the present invention. FIG. 4 is a flowchart showing a procedure and contents of a learning phase executed for learning the learning device in the grinding tool grinding surface evaluation device shown in FIG. FIG. 5 is a flowchart showing the procedure and contents of an image data acquisition routine in the learning phase shown in FIG. FIG. 6 is a flowchart showing the procedure and contents of a learning device generation routine in the learning phase shown in FIG. FIG. 7 is a flowchart illustrating a procedure and contents of an evaluation phase of evaluating a grinding tool using a learning device in the grinding tool grinding surface evaluation device illustrated in FIG. 3. FIG. 8 is a flowchart showing the procedure and contents of the evaluation routine in the evaluation phase shown in FIG. FIG. 9 is a diagram showing an example of image data for explaining the deteriorated portion detection processing in the learning phase shown in FIG. FIG. 10 is a diagram illustrating an example of a configuration of the learning device. FIG. 11 is a diagram illustrating an example of image processing performed by the feature amount calculation unit illustrated in FIG. FIG. 12 is a diagram illustrating an example of feature amount calculation data obtained by the feature amount calculation unit illustrated in FIG. FIG. 13 is a diagram illustrating an example of a learning device generation process performed by the learning device generation unit illustrated in FIG. FIG. 14 is a diagram illustrating an example of an evaluation process using a learned learning device by the grinding performance reduction factor evaluation unit illustrated in FIG. 3. FIG. 15 is a diagram illustrating an example of a visualized image of the grinding performance reduction factor over the entire circumference of the grinding surface, generated by the grinding performance reduction factor evaluation unit illustrated in FIG. 3. FIG. 16 is a diagram showing an example of a grinding surface image of a diamond wire which is another example of a grinding tool to which the present invention can be applied. FIG. 17 is a diagram showing an example of a relationship between a cup-type grindstone, which is another example of a grinding tool to which the present invention can be applied, and a camera. FIG. 18 is a diagram for specifically explaining one effect according to one embodiment of the present invention. FIG. 19 is a diagram for specifically explaining another effect according to the embodiment of the present invention. FIG. 20 is a diagram for specifically explaining another effect according to the embodiment of the present invention.

An embodiment according to the present invention will be described below with reference to the drawings.
[One embodiment]
(Configuration example)
(1) System FIG. 1 is a diagram showing an overall configuration of an inspection system including a grinding tool grinding surface evaluation device according to an embodiment of the present invention.
In FIG. 1, 1A indicates a grinding tool. The grinding tool 1A is made of, for example, a disk-shaped grindstone having a ring shape as shown in FIG. 2, and a number of abrasive grains 12, 12,. Is a polished surface. The tops of the abrasive grains 12, 12, ... are processed so as to form a flat surface.

  The grinding tool 1A is mounted on a rotating shaft of the rotating mechanism 2. The rotation mechanism 2 is connected to, for example, a θ-axis drive unit 21 using a servomotor. The θ-axis drive unit 21 rotates the grinding tool 1A in a stepwise manner at a predetermined angle interval via the rotation mechanism 2 under the control of a grinding tool grinding surface evaluation device 8 described later. The rotation direction is set to the same direction as the rotation direction of the grinding tool 1A when actually performing machining.

  XY tables 3 and 4 are provided on the side of the rotation mechanism 2. A column 5 is erected on the XY tables 3 and 4, and a camera 7 is attached to the column 5 via an arm 6 in a downward direction. The XY tables 3 and 4 have an X-axis drive unit 22 and a Y-axis drive unit 23. The X-axis drive unit 22 and the Y-axis drive unit 23 are driven under the control of the grinding tool grinding surface evaluation device 8, and move the camera 7 stepwise in the XY directions shown in FIG. 2, that is, in the horizontal direction.

  The arm 6 is configured to be movable in a stepwise manner in the Z direction shown in FIG. 2, that is, in the vertical direction with respect to the column 5 by the Z-axis driving unit 24. Can be changed. The drive of the Z-axis drive unit 24 is also controlled under the control of the grinding tool grinding surface evaluation device 8, similarly to the X-axis drive unit 22 and the Y-axis drive unit 23 described above.

  The camera 7 is, for example, a digital microscope mounted on a high-magnification metallographic microscope. The camera 7 captures an image of the grinding surface of the grinding tool 1A at a predetermined magnification in accordance with a driving instruction from the grinding tool grinding surface evaluation device 8, and the captured image data Is output to the grinding tool grinding surface evaluation device 8. An illuminator (not shown) is arranged above the grinding tool 1A for the above-mentioned imaging, and illuminates the grinding surface of the grinding tool 1A.

(2) Apparatus By the way, the grinding tool grinding surface evaluation apparatus 8 is composed of, for example, a personal computer and is configured as follows. FIG. 3 is a block diagram showing the functional configuration.
That is, the grinding tool grinding surface evaluation device 8 includes a control unit 81, a storage unit 82, and an input / output interface unit 83. The input / output interface unit 83 performs input / output processing of control signals and image data between the drive units 21 to 24 of the XYZ θ axes and the camera 7 described above, and performs input / output processing between the input device 25 and the display device 26. Performs input / output processing of input data and display data.

  The storage unit 82 uses a non-volatile memory that can be written and read at any time, such as a hard disk drive (HDD) or a solid state drive (SSD), as a storage medium. The storage unit 82 includes a program storage area, and image data storage section 821, deterioration information storage section 822, reduction factor information storage section 823, teacher data storage as storage areas necessary for implementing one embodiment. A section 824, a learned data storage section 825, and an evaluation data storage section 826 are provided.

  The control unit 81 has a central processing unit (CPU), and includes a position coordinate control unit 811, an image data acquisition unit 812, and a deteriorated portion detection unit as control functions necessary for implementing one embodiment. 813, a grinding performance reduction factor extraction unit 814, a feature value calculation unit 815, a teacher data generation unit 816, a learning device generation unit 817, a grinding performance reduction factor evaluation unit 818, and an evaluation data output unit 819. ing. Each of these functional units 811 to 819 is realized by causing the CPU to execute a program stored in a program storage area in the storage unit 82.

  When the camera 7 captures an image of the grinding surface of the grinding tool 1A, the position coordinate control unit 811 sets the XYθ coordinates for stepwise moving and setting the imaging target area of the grinding surface of the grinding tool 1A in units of a predetermined number of pixels. It is generated and output to the XYθ driving units 21, 22, 23. At the same time, the position coordinate control unit 811 generates Z coordinates for variably setting the focal length of the camera 7 in the Z direction with respect to the grinding surface in a stepwise manner, and outputs the generated Z coordinates to the drive unit 24.

  The image data acquisition unit 812 outputs an imaging control signal to the camera 7 every time the imaging target area is set to move stepwise in synchronization with the position coordinate control unit 811, and the grinding surface captured by the camera 7 Then, a process of taking in the image data and storing the image data in the image data storage unit 821 is performed. The image data acquisition unit 812 stores the XYZθ coordinates generated by the position coordinate control unit 811 in the image data storage unit 821 in association with the image data. The image data acquisition unit 812 performs the acquisition and storage processing of the image data on the grinding tool 1A before being used for machining and the grinding tool 1A after using the grinding tool 1A for machining.

  The deteriorated portion detection unit 813 reads out image data (image data after processing) of the grinding tool 1A after use for processing from the image data storage unit 821 and performs image processing on the image data after processing. Detects the deteriorated part of the grinding surface. The deteriorated portions include, for example, clogging in which chips are clogged between the abrasive grains, and flaws in which the abrasive grains 12 peel off from the binder 11 of the base material, but are not limited thereto. For example, a portion where the protrusion height of the abrasive grains 12 is substantially the same as the surface of the binder 11 and the workpiece surface slides on the binder 11 during processing, which causes a so-called upward slip phenomenon, is detected as a deteriorated portion. It may be.

  The feature amount calculation unit 815 performs a predetermined image process on each of the image data before and after processing stored in the image data storage unit 821, and calculates a feature amount representing each abrasive grain and its distribution state. calculate.

  The grinding performance reduction factor extraction unit 814 reads from the image data storage unit 821 an image before processing having the same position coordinates as the detected deteriorated portion of the grinding surface. Then, the unprocessed image data and its position coordinates are stored in the deterioration factor information storage unit 823 as information indicating the grinding performance reduction factor. The grinding performance reduction factor extraction unit 814 performs the processing extracted as the grinding performance reduction factor from the abrasive grain and the feature quantity between the abrasive grains in the entire area of the grinding surface of the grinding tool 1A calculated by the feature quantity calculation unit 815. The feature amount corresponding to the previous image is selected, and the selected feature amount is included in the information indicating the factor of reduction in grinding performance.

  When using the information indicating the above-described factor of lowering the grinding performance as the learning data, the teacher data generator 816 generates information indicating the state of the deteriorated portion of the grinding surface that is paired with the information indicating the lowering factor of the grinding performance, and Information indicating the state of the deteriorated portion of the surface is stored in the teacher data storage unit 824 as teacher data. Examples of the information indicating the state of the deteriorated portion of the grinding surface include, for example, an image indicating a state in which clogging, clogging or slippage occurs, an image indicating a portion in which clogging, clogging or slippage occurs, or clogging. In addition, information indicating the degree of occurrence of the eye-catching or sliding-up phenomenon (for example, the presence or absence and probability of occurrence) can be considered.

  The learning device generation unit 817 uses a convolutional neural network (CNN) as a learning device. For example, as shown in FIG. 10, the CNN has a plurality of convolutional layers C1, C2 and pooling layers P1, P2 alternately arranged in an intermediate layer, and a fully connected layer K and an output layer O arranged in a subsequent stage. It is. Then, in the learning phase, the learning device generation unit 817 gives the CNN the information representing each grinding performance reduction factor stored in the reduction factor information storage unit 823 as learning data (explanatory variables) and stores the teacher data. The CNN is learned by giving, as teacher data (objective variable), information indicating the state of the deteriorated portion of the grinding surface, which is paired with the information indicating the above-mentioned respective factors for reducing the grinding performance, stored in the section 824. Then, the learned CNN is stored in the learned data storage unit 825 as a learned learning device or a learned model.

  In the evaluation phase, the grinding performance reduction factor evaluating unit 818 determines the image data of the grinding surface of the grinding tool before shipping or the grinding tool 1A used or used for machining, and the abrasive grains included in the image data or the distribution thereof. The feature quantity indicating the state is input to the learned learning device (learned CNN). The grinding performance deterioration factor evaluation unit 818 outputs the information indicating the state of the deteriorated portion of the grinding surface output from the learned CNN in response to the input, together with the position coordinates, as evaluation data of the grinding performance reduction factor as the evaluation data storage unit 826. To memorize.

  The evaluation data output unit 819 reads out the evaluation data of the grinding performance reduction factor from the evaluation data storage unit 826. Then, display data for visualizing the evaluation data of the cause of the reduction in grinding performance is generated, output to the display device 26, and displayed.

(Operation example)
Next, the operation of the grinding tool grinding surface evaluation device 8 configured as described above will be described with reference to the flowcharts shown in FIGS.

(1) Learning Phase First, the processing procedure and processing contents of the learning phase will be described with reference to FIG.
(1-1) Initial setting of learning parameters In the grinding tool grinding surface evaluation device 8, the control unit 81 first takes in the learning parameters input by the administrator operating the input device 25 via the input / output interface unit 83. The learning parameters are stored in the learned data storage unit 825, for example. The learning parameters include a configuration of the learning device, an initial value of a connection weight between a plurality of neurons constituting each layer of the learning device, and an initial value of a threshold value of each neuron. The learning parameters can be given by template data.

  In addition, the control unit 81 specifies the size of one imaging region when the imaging surface of the grinding tool 1 is imaged by the camera 7 and the step width when the imaging region is moved, which are input by the input device 25 by the administrator. The received data is received via the input / output interface unit 83 and stored in a control memory (not shown) in the control unit 81.

(1-2) Acquisition of Image Data of Grinding Surface Before Processing of Grinding Tool 1A First, the administrator attaches the grinding tool 1A to the rotating mechanism 2 before use for processing. Then, an image acquisition start instruction is input through the input device 25. Then, under the control of the position coordinate control unit 811 and the image data acquisition unit 812, the control unit 81 first performs the process of acquiring the image data of the grinding surface of the grinding tool 1A before machining in step S10 as follows. Execute.

  That is, the position coordinate control unit 811 first sets the initial position coordinates of the imaging region with respect to the grinding surface of the grinding tool 1A in step S101. Then, in step S102, the position of the imaging region with respect to the grinding surface is moved stepwise by a preset step width by changing the XYZθ coordinates. Of these, θ is changed by rotating the grinding tool 1A in the same direction as that used for machining by operating the rotation mechanism 2 by the θ-axis drive unit 21. Of XYZ, the XY tables 3 and 4 are moved in the horizontal direction by the X-axis drive unit 22 and the Y-axis drive unit 23 for XY, and the camera 7 is moved in the vertical direction by the Z-axis drive unit 24 for Z. Change by moving.

  The image data acquisition unit 812 operates the camera 7 in step S103 in synchronization with the movement control of the position of the imaging region, and outputs the image data of the imaging region of the grinding surface captured by the camera 7 to the input / output interface unit 83 To get through. Then, the acquired image data is stored in the image data storage unit 821 in association with the XYZθ coordinates of the imaging region set by the position coordinate control unit 811. The image data acquiring unit 812 determines in step S105 that the acquisition of the image data over the entire area of the grinding surface along the rotation direction of the grinding tool 1A is completed, until the image data acquisition unit 812 determines that the image data has been acquired in steps S102 to S104. Repeat the acquisition process.

  Thus, in the image data storage unit 821, image data of the entire grinding surface of the grinding tool 1A before being used for processing is stored together with the XYZθ coordinates of the imaging position in the rotational direction of the grinding tool 1A.

(1-3) Acquisition of Image Data of Grinding Surface after Processing of Grinding Tool 1A Next, the administrator rotates the grinding tool 1A after using the grinding tool 1A having acquired the pre-processing image for actual processing. Attach to mechanism 2. Then, an image acquisition start instruction is input through the input device 25.

  Then, under the control of the position coordinate control unit 811 and the image data acquisition unit 812, the control unit 81 changes the position of XYZθ in step S11 in the same manner as when acquiring the image data of the grinding surface before machining. Then, a process of acquiring captured image data of the ground surface after processing of the grinding tool 1A is executed. The process of acquiring the image data of the ground surface after the processing is also executed according to the flowchart of FIG. 5, similarly to the process of obtaining the image data of the ground surface before the processing described above.

  Thus, the image data storage unit 821 stores the image data of the entire area of the grinding surface of the grinding tool 1A after processing along with the XYZθ coordinates of the imaging position in the rotation direction of the grinding tool 1A.

(1-4) Detection of Deteriorated Area on Grinding Surface Next, under the control of the deteriorated part detecting unit 813, the control unit 81 captures image data of the ground surface after processing from the image data storage unit 821 in step S12. Each area is read out, and predetermined image processing is performed on the image data to perform clogging in which chips are clogged between the abrasive grains 12 and 12, and that the abrasive grains 12 are peeled off from the binder 11 of the base material. The binder 11 detects a site where an upward sliding phenomenon occurs on the surface of the workpiece.

  Clogging is likely to occur when a relatively soft material such as aluminum or copper is ground, for example, and the surface of the clogged chip becomes a flat surface having almost the same height as the uppermost surface of the abrasive grains 12. In many cases, the surface of the abrasive grains 12 is larger than the area of the uppermost surface of the abrasive grains 12 because the particles are often stuck between the grains 12.

  Therefore, in the image processing, an image corresponding to the uppermost surface of the abrasive grains 12 is detected using a binarization process or the like, and an image having an area equal to or larger than a threshold is detected from among a large number of images detected thereby. This is regarded as an image of a site where clogging has occurred. FIG. 9A shows an example of a detection image of clogging. FIG. 9B shows a detection image of abrasive grains before clogging occurs.

  On the other hand, the fouling is likely to occur when a material having a high hardness is searched, and a hole is formed in the surface of the binder due to the drop of the abrasive grains 12. Therefore, the bottom surface of the hole becomes lower than the position of the surface of the binder as well as the uppermost surface of the abrasive grains 12, and the image obtained by focusing on the uppermost surface of the abrasive grains 12 becomes a blurred image without being focused. .

  Therefore, when detecting eye damage, an in-focus area is extracted from a captured image, and the area other than the area is extracted as a candidate for abrasive grains falling off. At this time, the extracted candidate includes the uneven portion of the binder. However, since the area of the concave and convex portions of the binder is smaller than the area of the falling marks of the abrasive grains 12 and the distance between the concave and convex portions is short, the area or the distance is determined by using a threshold value. It becomes possible to detect a dropout mark.

  Further, since the uppermost surface of the abrasive grains 12 and 12 and the surface of the binder 11 are substantially flush with each other, the upper slip occurrence site is imaged, for example, with the focal length adjusted to the uppermost surface of the abrasive grains 12 and 12. From the image, the area of the region where the surface of the binder 11 is clearly detected is detected, and when the area of the binder 11 is equal to or larger than the threshold value, it can be detected by regarding the site as an upper slip occurrence site.

  Then, the deteriorated part detection unit 813 correlates the detected image of the part where the clogging, the damage, and the upper slip occurred, detected as described above, with the position coordinates (XYZθ coordinate values) of the imaging region in which the image is detected. The state is stored in the deterioration information storage unit 822.

(1-5) Calculation of feature amount of abrasive grains and their distribution state The control unit 81 reads out image data before processing from the image data storage unit 821 in step S13 under the control of the feature amount calculation unit 815, A process is performed to calculate a feature amount for each of the abrasive grains 12, 12 included in the image data before processing and its distribution state.

  For example, the feature amount calculation unit 815 performs binarization processing or labeling processing on the image data before processing for each imaging region, and thereby detects the abrasive grains 12, 12 from the image of each imaging region. FIGS. 11B and 11C show an example of a binarized image and a labeling image obtained by performing the binarizing process and the labeling process on the original image shown in FIG. 11A, respectively.

  Then, the feature amount calculation unit 815 obtains the barycentric coordinates for each of the detected abrasive grains 12. The barycenter coordinates can be calculated by using the XYZθ coordinates stored in association with the imaging region. In addition, the feature amount calculation unit 815 calculates the area and circularity of each of the abrasive grains 12. Further, the feature amount calculation unit 815 calculates the distance between the adjacent abrasive grains 12 and 12 using the barycentric coordinates of the respective abrasive grains 12 and 12, and calculates the area using the barycentric coordinates of the respective abrasive grains 12 and 12. Is Voronoi-divided, and the area of each divided region is obtained.

  The feature amount calculation unit 815 calculates the various feature amounts of the individual abrasive grains 12, 12 calculated as described above, and the distribution state of the abrasive grains 12, 12, such as the distance between the abrasive grains 12, 12 and the area of the divided region. Is stored in the memory in the control unit 81. FIG. 12 shows an example of the detection result of the feature amount.

  It should be noted that specific examples of the above-described detection processing of the abrasive grains 12 and 12 and the processing of calculating the analysis parameters such as the barycentric coordinates of the detected abrasive grains 12 and 12 are described in, for example, Japanese Patent No. 5569883 proposed by the present inventor. It is described in detail in the gazette. Regarding the method of analyzing the distribution of abrasive grains using Voronoi diagrams, see, for example, Kawashita, Sakaguchi, Kawaguchi, Matsui, Maeda, Matsuo, "Study on Three-Dimensional Measurement of Grinding Wheel Work Surface Topography by Image Processing", 4th Report, Line High-speed image acquisition with a camera and analysis of abrasive grain cutting edge distribution using Voronoi diagram, detailed description in Journal of Japan Society of Abrasive Technology, Vol.60, No.8, pp.442-447 (2016).

(1-6) Extraction of Grinding Performance Deterioration Factor Next, under the control of the grinding performance deteriorating factor extraction unit 814, the control unit 81 determines in step S14 that the clogging is detected from the deterioration information storage unit 822 as a deteriorated part. The position coordinates of the imaging region including the site where the eyebrows and the upper slip occur are read. Then, the image of the imaging region before processing associated with the same position coordinates as the read position coordinates is selectively read from the image data storage unit 821. Then, the image of the imaging region before the processing is stored in the deterioration factor information storage unit 823 as information representing the grinding performance reduction factor.

  The grinding performance reduction factor extraction unit 814 indicates the characteristic amount of each of the abrasive grains 12 included in the image of the imaging region before processing selected as the grinding performance reduction factor and the distribution state of the abrasive grains 12. The feature amount is read out from the memory of the feature amount calculation unit 815 based on the barycenter coordinates of the abrasive grains 12, 12, and stored in the reduction factor information storage unit 823 together with the information indicating the grinding performance reduction factor.

  Thus, the deterioration factor information storage unit 823 stores an image of a part of the abrasive grains 12, 12 existing on the grinding surface of the grinding tool 1 </ b> A before machining, which may be a factor that may cause a clogging, damage, or an upper slip occurrence part. And the characteristic amount of each of the abrasive grains 12 and 12 and the characteristic amount indicating the distribution state of the abrasive grains 12 and 12 included in the image are stored as information indicating a factor of reduction in grinding performance. The information indicating the grinding performance reduction factor is used as learning data (explanatory variables) for generating a learning device for evaluating a grinding surface, which will be described later.

(1-7) Generation of Teacher Data Under control of the teacher data generation unit 816, the control unit 81 generates, in step S15, teacher data used when generating a learning device for grinding surface evaluation described later. The teacher data generation unit 816 links the position coordinates of the imaging region including the abrasive grains 12, 12 which may be the grinding performance reduction factor stored in the reduction factor information storage unit 823 as information indicating the grinding performance reduction factor, for example. The image of the obtained deteriorated part is read from the deterioration information storage unit 822.

  For example, taking the image shown in FIG. 9 as an example, a state in which clogging, clogging or slippage occurs corresponding to the image (b) before clogging, clogging or slippage occurs. Is read from the deterioration information storage unit 822. Then, the read image representing the state in which the clogging, blinding, or upward slip phenomenon has occurred is stored in the teacher data storage unit 824 as it is as the teacher data in association with the position coordinates of the imaging region. FIG. 13A is a diagram illustrating an example of a case where an image representing a state in which the clogging, blinding, or slip phenomenon occurs is used as teacher data.

  In addition, the teacher data generation unit 816 uses, for example, images representing the abrasive grains 12 and 12 that can be a grinding performance reduction factor stored in the reduction factor information storage unit 823 as information indicating the grinding performance reduction factor, as teacher data. The teacher data is stored in the teacher data storage unit 824 in association with the position coordinates of the imaging region. FIG. 13B is a diagram showing an example in which an image representing only the abrasive grains 12, 12 which can cause the above-described clogging, damage, or slippage phenomenon is used as teacher data.

  Further, the teacher data generation unit 816 may store, for example, each of the images of the imaging regions including the abrasive grains 12, 12 that may be the grinding performance reduction factor, stored in the reduction factor information storage unit 823 as information representing the grinding performance reduction factor. Then, the deterioration factors included in the imaging region are classified into classes according to their types. Specifically, as shown in FIG. 13 (c), the clogging factor is CLASS1, the clogging factor is CLASS2, the clogging and clogging factor is CLASS3, and the clogging and clogging factor is CLASS1. The case where neither factor is included is referred to as CLASS4. Then, for each of the imaging regions, the CLASS is set in accordance with the presence or absence of the abrasive grains 12, 12 which may be a cause of clogging and damage, and numerical data representing the degree of occurrence of a reduction factor for each CLASS is set. Are stored in the teacher data storage unit 824 in association with the position coordinates of the imaging area.

  Note that the teacher data is not limited to the above three examples, and various other data can be considered. Further, the teacher data may be set by an administrator, and the setting data may be manually input from the input device 25 or read using a storage medium to be stored in the teacher data storage unit 824.

(1-8) Generation of a learning device for grinding surface evaluation The control unit 81 generates a learning device for grinding surface evaluation in step S16 under the control of the learning device generation unit 817 as follows. FIG. 6 is a flowchart showing an example of the processing procedure and processing contents.

  That is, the learning device generation unit 817 first specifies one address of the deterioration factor information storage unit 823 in step S161, and in step S162, determines the image of the imaging region including the grinding performance reduction factor and the grinder included in the image. The individual characteristic amounts of the grains 12 and the characteristic amounts indicating the distribution state thereof are read as learning data. At the same time, the teacher data corresponding to the imaging area is read from the teacher data storage unit 824.

  Next, in step S134, the learning device generation unit 817 inputs the read learning data to the CNN prepared in advance and provides the teacher data to make the CNN perform learning. In step S165, learning parameters such as the connection weight between neurons and the threshold value of each neuron are updated based on the result of the learning. The learning device generation unit 817 repeats the above learning process by sequentially providing all the deterioration factor information stored in the deterioration factor information storage unit 823 and the corresponding teacher data to the VNN. Then, when it is determined in step S166 that the learning processing for all the deterioration factor information has been completed, the learning processing is completed. The learned learning device (learned model) including the learning parameters after the learning is completed is stored in the learned data storage unit 825.

  The learning process of the CNN may be repeatedly performed by giving not only the information of the grinding performance deterioration factor obtained for one grinding tool 1A but also the information of the grinding performance deterioration factor obtained for each of a plurality of grinding tools. It may be. The more the number of input points of the information on the factor of the decrease in the grinding performance, the more the evaluation performance by the CNN can be improved.

(2) Evaluation phase The grinding tool manufacturer or the user of the grinding tool uses the grinding tool grinding surface evaluation device 8 to evaluate the grinding tool 1A before shipping, or grinding during or after processing. The evaluation of the tool 1A is performed as follows. FIG. 7 is a flowchart showing an example of the processing procedure and processing contents of the evaluation phase in the grinding tool grinding surface evaluation device 8.

(2-1) Acquisition of Image Data of Grinding Surface of Grinding Tool 1A The evaluator mounts the grinding tool 1A to be evaluated on the rotating mechanism 2. Then, an evaluation start instruction is input through the input device 25. Then, under the control of the position coordinate control unit 811 and the image data acquisition unit 812, the control unit 81 executes a process of acquiring captured image data of the grinding surface of the grinding tool 1A in step S20. The image data acquisition process at this time is also executed in accordance with the flowchart in FIG. 5, similarly to the image data acquisition process in the learning phase described above. Thus, the image data storage unit 821 stores the image data in the entire area of the grinding surface of the grinding tool 1A to be evaluated together with the XYZθ coordinates of the imaging area in the order of the rotation direction of the grinding tool 1A.

(2-2) Calculation of Feature Amount Under control of the feature amount calculation unit 815, the control unit 81 reads out image data from the image data storage unit 821 in step S21, and reads each abrasive grain 12 included in the image data. , 12 and the distribution state thereof are calculated. This feature amount calculation process can be performed in the same manner as in the case where the feature amount is calculated from the image data of the grinding surface before machining of the grinding tool 1A in the learning phase. As a result, with respect to the grinding tool 1A to be evaluated, the characteristic amount of each of the abrasive grains 12, 12 on the abrasive surface and the characteristic amount indicating the distribution state of the abrasive grains 12, 12 are obtained.

(2-3) Evaluation Routine Using Learned Learning Unit (CNN) The control unit 81 performs the evaluation of the grinding surface of the grinding tool 1A in step S22 as described below under the control of the grinding performance reduction factor evaluation unit 818. Do. FIG. 8 is a flowchart showing the processing procedure and processing contents.

  The grinding performance reduction factor evaluation unit 818 first reads the learned CNN from the learned data storage unit 825 in step S221 and expands the learned CNN on the RAM of the control unit 81. Then, in step S222, an initial value of the XYZθ coordinates of the grinding surface imaging region of the grinding tool 1A is designated, and in step S223, the image data of the corresponding imaging region is read from the image data storage unit 821, and the image and the feature amount calculation are performed. The corresponding feature amount calculated by the unit 815 is input to the learned CNN as evaluation target data. Then, the grinding performance reduction factor evaluation unit 818 stores the output data of the learned CNN in the evaluation data storage unit 826 in association with the XYZθ coordinates as evaluation result data for the evaluation target data.

  The grinding performance reduction factor evaluation unit 818 repeats the evaluation processing for each imaging region using the learned CNN in steps S222 to S224 until the end of the designation of the XYZθ coordinates of all the imaging regions is determined in step S225.

  FIG. 14 shows an example of an evaluation processing operation using the learned CNN. FIG. 14A shows the case where the clogging or the clogging occurs when the image and the feature amount of the imaging region including the abrasive grains 12 and 12 that may cause the clogging or the clogging are input to the learned CNN. An example in which a state estimation image is output from a learned CNN is shown.

  FIG. 14B shows an example in which, when similar images and feature amounts are input to the learned CNN, only the images of the abrasive grains 12 and 12 that may cause the clogging or damage are output. Further, FIG. 14C shows that when the same image and feature amount are input to the learned CNN, the above-described reduction is performed for each CLASS according to the presence of the abrasive grains 12 that can cause clogging or damage. An example is shown in which numerical data indicating the degree of occurrence of a factor is output from a learned CNN. In addition, it is possible to similarly output from the learned CNN the upper slip occurrence portion.

  When the evaluation results for the entire area of the grinding surface of the grinding tool 1A are obtained, the control unit 81 performs a process of outputting the evaluation results in step S23 under the control of the evaluation data output unit 819. For example, as shown in FIG. 8, in step S226, based on the above-described evaluation result and the position coordinates associated with the evaluation result, the position of a part that can cause clogging, damage, or an upper slip phenomenon is determined based on the grinding surface. Generates visualization data displayed over the entire area. Then, the visualization data is output from the input / output interface unit 83 to the display device 26 and displayed. FIG. 15 shows an example of the visualization data.

  Further, in step S227, the evaluation data output unit 819 calculates, for example, the area or the number of the parts that cause the deterioration of the grinding performance in the entire area of the grinding surface of the grinding tool 1A. Then, the calculated area or number of the portion that causes the reduction of the grinding performance is output from the input / output interface unit 83 to the display device 26 and displayed.

  Further, in step S228, the evaluation data output unit 819 compares the area or the number of the parts, which are factors causing a decrease in the grinding performance, calculated in step S227 with reference to the dressing timing determination table stored in advance to determine the grinding performance. The dressing time corresponding to the area or the number of the parts that may be the cause of the decrease is determined. Then, information representing the determined dressing time is output from the input / output interface unit 83 to the display device 26 and displayed.

  Finally, in step S229, the evaluation data output unit 819 displays the generated visualization data indicating the position of the part that can be a factor of lowering the grinding performance in the entire area of the grinding surface, the area of the calculated part that is a factor of lowering the grinding performance or The number and the information indicating the determined dressing time are stored in the evaluation data storage unit 826 in association with the product number or the like of the grinding tool 1A.

  In the above description, the evaluation data output unit 819 outputs the visualization data indicating the position of the part that may cause the grinding performance deterioration, the information indicating the area or the number of the part that may cause the grinding performance deterioration, and the information indicating the dressing time by the evaluation data output unit 819. Generated or calculated. However, the present invention is not limited thereto, and the above-described data and information may be generated or calculated by the grinding performance reduction factor evaluation unit 818.

(effect)
As described above in detail, in one embodiment, in the learning phase, while acquiring and storing the grinding surface image data before machining of the grinding tool 1A, the image data of the grinding surface after being used for machining of the grinding tool 1A is obtained. From the image data to detect a deteriorated portion such as a portion in which clogging, clogging, and an upper sliding phenomenon occurs, and to store the unprocessed image corresponding to the deteriorated portion and position in the stored unprocessed image. Extract from data. Then, the extracted unprocessed image and the characteristic amount of each of the abrasive grains 12 and 12 included in the image and the characteristic amount indicating the distribution state thereof are input to the CNN as information indicating a factor for reducing the grinding performance. In addition, information indicating an evaluation result with respect to the information indicating the cause of the decrease in the grinding performance is given to the CNN as teacher data, whereby the CNN is learned to generate a CNN for grinding surface evaluation.

  Therefore, information representing the cause of grinding performance deterioration such as clogging or damage of abrasive grains, the occurrence of slippage, etc., obtained by actual processing is used as learning data (explanatory variable), and the evaluation result corresponding to the above-described grinding performance deterioration factor Can be made to learn CNN by using the information representing as the teacher data (objective variable). Therefore, a learned CNN with high evaluation performance based on actual machining data can be generated without performing detailed analysis processing on the grinding surface.

  In the evaluation phase, the learned CNN is used to show image data of the grinding surface of the grinding tool 1A to be evaluated, individual feature amounts of the abrasive grains 12 included in the image data, and a distribution state thereof. The feature amount is input to the learned CNN as learning data, and output data of the learned CNN is output as evaluation data of the grinding surface of the grinding tool 1A.

  Therefore, simply by inputting the image data acquired before or after the processing of the grinding tool 1A to the learned CNN, information representing the evaluation result of the grinding surface is output. That is, by using the learned CNN, it is possible to preliminarily evaluate, in a short time and with high accuracy, a surface property which is a cause of a decrease in the grinding performance of the grinding surface before or after the processing of the grinding tool 1A. . For this reason, the evaluator can perform quality control before shipping or after starting use of the grinding tool 1A based on the output evaluation data.

  Further, in one embodiment, as learning data to be input to the CNN, in addition to an image of the abrasive surface before machining which may be a factor of reducing the grinding performance, individual feature amounts of the abrasive grains 12 and 12 included in the image and the feature amount thereof are included. A feature quantity indicating a distribution state is input, and CNN is learned. For this reason, compared with the case where only the grinding surface image of the grinding tool to be evaluated is used as the learning data, it is possible to generate a CNN having higher classification or classification performance. Further, a more accurate evaluation result can be obtained by evaluating the grinding surface using the learned CNN.

  Further, the evaluation data output unit 819 outputs visualization data indicating the position of the part that can be a factor of reducing the grinding performance in the entire area of the grinding surface, information indicating the area or number of the part that causes the reduction of the grinding performance, and information indicating the dressing time. Generate or calculate and output. Therefore, the evaluator can grasp the evaluation result of the grinding tool 1A visually or accurately as a numerical value.

  In particular, by evaluating the grinding surface of the grinding tool before shipping, factors that lead to a decrease in grinding performance (for example, clogging or damage, a site where slippage occurs, an exposed state of the outermost peripheral surface of the binder 11 of the base material) , The size of the flank of the cutting edge of the abrasive grains, etc.) can be predicted before shipment, thereby improving the reliability of quality assurance of the grinding tool. In addition, it is possible to establish a control guideline for an appropriate abrasive grain distribution and shape that does not cause performance degradation in fixed abrasive processing. As a result, by accumulating the above data and using the data as basic data, it can be utilized for development of a more accurate grinding tool and development of a grinding tool from a new viewpoint.

  When the function of the grinding tool grinding surface evaluation device 8 of the present embodiment is added to the NC machine tool to enable on-machine measurement, it is necessary to estimate the dressing time of the grinding tool during or after machining. Thus, dressing and whetstone replacement at the optimal timing can be realized, and a significant improvement in productivity can be expected.

  Further, since the know-how of the user can be added to the AI by opening the processing procedure and processing contents of the learning phase of the CNN to the user, it is possible to develop an NC machine tool employing the AI processing technology unique to the user. Become.

  In general, the distribution of abrasive grains in which clogging is likely to occur on the abrasive surface of the grinding tool 1A is, for example, such that a plurality of abrasive grains 12, 12 are close to each other in the rotation direction M of the grinding tool 1A as shown in FIG. It is in a dense state. In this state, the chips by the first abrasive grains positioned at the head with respect to the rotation direction M of the tool overlap with the chips by the second and third abrasive grains located close to the rear in the rotation direction M. It is presumed that clogging occurs as a result.

  On the other hand, even in a state where the plurality of abrasive grains 12 and 12 are densely close to each other in the rotation direction M of the grinding tool, for example, as shown in FIG. If the 0-th abrasive grain is present within a certain distance upstream of the second abrasive grain, the first abrasive grain hardly acts as a cutting edge, so that chips from the first abrasive grain are second and third. No longer overlap with the chips of the abrasive grains, and as a result, clogging does not occur at the dense portion of the abrasive grains.

  That is, even in a region where the plurality of abrasive grains 12 and 12 are densely dispersed in the same dispersion state, clogging does not occur depending on the positional relationship with other abrasive grains existing on the upstream side in the rotation direction of the tool. Sometimes. As described above, the relationship between the dispersion state of the abrasive grains 12 and 12 on the grinding surface and the occurrence of clogging is complicated, and it is necessary to grasp all the abrasive grains that cause clogging even when analyzing individual abrasive grains. It is difficult.

  On the other hand, in one embodiment of the present invention, a portion where clogging actually occurs is detected, and the distribution state of the abrasive grains 12 and 12 corresponding thereto is extracted as a factor for lowering the grinding performance. The learning device learns the abrasive grains 12, 12 contained in the image and the feature amount of the dispersed state as explanatory variables. Therefore, when evaluating the grinding surface before or during machining of the grinding tool using this learning device, it is possible to easily and completely detect the distribution state of the abrasive grains 12, 12 which causes clogging. Becomes

  On the other hand, as one of the factors for reducing the grinding performance, there is a portion where the protrusion height of the abrasive grains becomes almost the same as the surface of the binder and the surface of the workpiece slides on the binder at the time of processing, which causes a so-called upward sliding phenomenon. FIGS. 19A and 19B show an example thereof, wherein FIG. 19A is a plan view showing a dispersion state of the abrasive grains 12, 12, and FIG. 19B is a side view thereof.

  In general, on the grinding surface of the grinding tool, since the surface of the binder 11 is scraped off by shavings or the like in a machining process, the protruding height of the abrasive grains with respect to the binder surface increases. Also, the protruding height of the abrasive grains is increased by shaving the surface of the binder by dressing. However, for example, as shown in FIG. 19 (a), the binder 11 is hard to be scraped off at a portion where the abrasive grain distribution on the abrasive surface is dense. For this reason, even if the height from the tool center or the reference plane to the uppermost surface of the abrasive grains is kept constant by tooling or the like, the abrasive grains 12 from the surface of the binder 11 as shown in FIG. , 12 are small. In this state, further processing is continued, and as the wear of the abrasive grains 12, 12 progresses, the height of the uppermost surface of the abrasive grains 12, 12 and the surface of the binder 11 become substantially the same as shown in FIG. 20, for example. If processing is performed in this state, the binder 11 slips on the surface of the workpiece, causing a slip phenomenon, and the processing accuracy is significantly reduced.

  On the other hand, in the embodiment of the present invention, in the learning phase, the upper slip occurrence portion of the grinding surfaces 12 and 12 is detected as a deteriorated portion, and an image of a portion before machining corresponding to the deteriorated portion is extracted. The image before processing, the abrasive grains 12, 12 contained in the image, and the feature amount of the dispersed state are given to the learning device as one of explanatory variables, and learning is performed. Therefore, by evaluating the grinding surface of the grinding tool before or during machining by using the learning device, it is possible to detect the occurrence of slippage of the grinding tool on the grinding surface without omission.

[Other embodiments]
In the above-described embodiment, a case where a ring-shaped disc-type grindstone is applied as a grinding tool has been described as an example. However, as shown in FIG. The present invention is also applicable to grinding wheels. Further, as another example of the grinding tool, the present invention may be applied to a wire-type grindstone as shown in FIG. The wire-type grindstone illustrated in FIG. 16 is an example of a diamond wire, in which (a) shows a wire surface with coarse abrasive grains dispersed, and (b) shows a wire surface with fine abrasive grains dispersed. When acquiring the image data of the grinding surface of the wire grindstone, the camera captures an image of the grinding surface while moving the wire grindstone linearly. In this case, it is also possible to image the grinding surface by fixing the wire-type grindstone and moving the position of the camera.

  In addition to the convolutional neural network (CNN), a general forward-propagation type neural network having a multilayer structure, a support vector machine, a self-organizing map, a learning device for learning by reinforcement learning, It is possible to use a recurrent neural network. In any learning device, the number of hidden layers, the number of neurons in each layer, the connection relationship between neurons, and the transfer function of neurons may be appropriately determined according to the embodiment.

  In one embodiment, a function for executing learning of the learning device and a function of evaluating the grinding surface of the grinding tool using the learned learning device are provided in one grinding tool grinding surface evaluation device. The description has been made by taking the case of providing as an example. However, the invention is not limited to this.For example, a function for performing learning of a learning device is provided in a computer on a cloud, and a function of evaluating a grinding surface of a grinding tool is provided in a control device or the like disposed on a site such as a factory, The learning parameter of the learning device learned by the computer on the cloud may be transmitted to the control device and set in the learning device.

  In addition, the type and configuration of data to be provided as learning data to the learning device, and the characteristics of the abrasive surface image and abrasive grains to be provided to the learning device as learning data can be variously modified without departing from the gist of the present invention. .

  As described above, the embodiments of the present invention have been described in detail. However, the above description is merely an example of the present invention in every respect. It goes without saying that various improvements and modifications can be made without departing from the scope of the invention. That is, in implementing the present invention, a specific configuration according to the embodiment may be appropriately adopted.

  In short, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements in an implementation stage without departing from the scope of the invention. Various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiment. Further, components of different embodiments may be appropriately combined.

  DESCRIPTION OF SYMBOLS 1 ... Grinding tool, 2 ... Rotating mechanism, 3, 4 ... XY table, 5 ... Support, 6 ... Arm, 7 ... Camera, 8 ... Grinding tool grinding surface evaluation device, 11 ... Binder, 12 ... Abrasives, 21 ... θ-axis drive unit, 22: X-axis drive unit, 23: Y-axis drive unit, 24: Z-axis drive unit, 25: input device, 26: display device, 81: control unit, 82: storage unit, 83: input / output Interface unit, 811: Position coordinate control unit, 812: Image data acquisition unit, 813: Degraded site detection unit, 814: Grinding performance reduction factor extraction unit, 815: Feature amount calculation unit, 816: Teacher data generation unit, 817: Learning Generator generation unit, 818: grinding performance reduction factor evaluation unit, 819 ... evaluation data output unit, 821 ... image data storage unit, 822 ... deterioration information storage unit, 823 ... reduction factor information storage unit, 824 ... teacher data storage unit, 82 ... learned data storage unit, 826 ... evaluation data storage unit.

In order to solve the above-mentioned problems, a first aspect of a grinding tool grinding surface evaluation device according to the present invention is a state in which a plurality of abrasive grains of the grinding tool are dispersed while moving the grinding tool in a predetermined direction. And an image acquisition unit that acquires image data obtained by imaging the grinding surface fixed by the binder, and stores the image data acquired before the start of machining of the grinding tool by the image acquisition unit in a storage medium. A storage control unit, a detection unit that detects information indicating a deteriorated portion on the grinding surface of the grinding tool from image data acquired after the processing of the grinding tool by the image acquisition unit, and a storage unit that is stored in the storage medium. a extraction unit for extracting an image of a site before machining position and the deteriorated portion from image data before processing starts corresponding as information representing the grinding performance degradation factor in the extraction section, the grindability From the image of the unprocessed part corresponding to the deteriorated part and the position extracted as information representing the cause of the decrease, a feature amount in which the moving direction of the grinding tool of the abrasive contained in the image is reflected is calculated and calculated. The obtained feature amount is included in the information representing the grinding performance reduction factor .

Further , according to the first and second aspects, a feature amount reflecting the moving direction of the grinding tool of the abrasive grain, which is a factor of reducing the grinding performance, is calculated. Will be added. For this reason, the distribution state of the abrasive grains, which is a cause of the decrease in the grinding performance, can be represented not only by the image but also by the feature amount thereof, whereby the cause of the decrease in the grinding performance can be represented with higher accuracy.

A third aspect of the grinding tool grinding surface evaluation device according to the present invention is to provide a learning device with information representing the grinding performance degradation factor as learning data and information representing a deteriorated portion of the grinding surface as teacher data. The apparatus further includes a learning device generation unit for learning the learning device.

According to the third aspect, the learning processing is performed by the learning device using the information indicating the grinding performance deterioration factor as learning data (explanatory variable) and the information indicating the state of the deteriorated portion of the grinding surface as teacher data (objective variable). Done. For this reason, when the information indicating the grinding performance deterioration factor is input, a learning device that outputs information indicating the state of the deteriorated portion of the grinding surface can be generated.

In the present invention, for example, the following fourth, fifth, and sixth aspects can be considered as the learning unit generation unit.
A fourth aspect is to provide the learning device with, as the learning data, an image including an unprocessed part corresponding to a deteriorated part and a position of the grinding surface, and the feature amount of abrasive grains included in the image, and The learning device is made to learn by giving an image representing the deteriorated part as the teacher data.

According to the fourth aspect, when the image including the unprocessed portion whose position corresponds to the deteriorated portion of the grinding surface and the feature amount of the abrasive grains included in the image are input, the state in which the unprocessed portion is deteriorated is input. It is possible to generate a learning device that outputs the image shown.

In a fifth aspect, the learning device is configured to provide, as the learning data, an image including a part before processing whose position corresponds to the deteriorated part and the feature amount of abrasive grains included in the image, and The learning device is made to learn by giving, as the teacher data, an image representing a part before processing corresponding to a position.

According to the fifth aspect, when the image including the unprocessed portion corresponding to the deteriorated portion and the position of the abrasive surface and the feature amount of the abrasive grains included in the image are input, the unprocessed portion corresponds to the deteriorated portion and the position. Can be generated.

In a sixth aspect, the learning device is configured to provide, as the learning data, an image including a part before processing whose position corresponds to the deteriorated part and the feature amount of abrasive grains included in the image, and The learning device is made to learn by giving, as the teacher data, numerical data indicating the degree to which the portion before processing corresponding to the position causes the grinding performance to be reduced.

According to the sixth aspect, when the image including the unprocessed part corresponding to the deteriorated part and the position and the feature amount of the abrasive grains included in the image are input as the learning data, the part before the processing becomes the grinding performance. It is possible to generate a learning device that outputs numerical data indicating the degree of reduction.

In a seventh aspect of the grinding tool grinding surface evaluation device according to the present invention, the image data acquired by the image acquisition unit is input to the learning device before or after the machining of the grinding tool is started. The information processing apparatus further includes an evaluation unit that outputs information representing a deteriorated portion of the grinding surface output from the learning device as information representing an evaluation result of the grinding surface of the grinding tool.

According to the seventh aspect, just by inputting the image data acquired before or after the processing of the grinding tool to the learning device, the information indicating the deteriorated portion of the grinding surface is obtained from the learning device by the grinding device. The information is output as information indicating the evaluation result of the grinding surface of the tool. In other words, by using the learning device, it is possible to evaluate the surface texture that is a cause of a decrease in the grinding performance of the grinding surface before and after the processing is started in a short time and with high accuracy. For this reason, the person in charge of evaluation can perform quality control before shipping or after starting use of the grinding tool based on the output evaluation result.

An eighth aspect of the grinding tool grinding surface evaluation apparatus according to the present invention, in the evaluation unit, based on the information representing the deteriorated portion of the grinding surface output from the learning device, the grinding portion of the deteriorated portion, A visualized image representing the distribution of the moving direction of the grinding tool in the entire area of the grinding surface of the tool is generated and output.

According to the eighth aspect, a visualized image representing the distribution state of the movement direction in the entire area of the grinding surface of the portion expected to be a cause of deterioration in grinding performance is generated and output. For this reason, the production manager of the grinding tool or the manager of the use of the grinding tool determines, for each grinding tool, the distribution state of the portion that causes a reduction in the grinding performance of the grinding surface over the entire area of the grinding surface and moves the grinding tool. It is possible to easily and quickly grasp along the direction.

In a ninth aspect of the grinding tool grinding surface evaluation device according to the present invention, in the evaluation unit, the dressing time of the grinding tool is determined based on information indicating a deteriorated portion of the grinding surface output from the learning device. Is estimated, and information representing the estimation result is output.

According to the ninth aspect, the estimation unit outputs the estimation information of the dressing time of the grinding tool. For this reason, for example, the manufacturing manager of the grinding tool or the manager of the use of the grinding tool can perform the shipping verification of the grinding tool and manage the dressing time of the grinding tool based on the estimated information.

Claims (13)

While moving the grinding tool in a predetermined direction, an image acquisition unit that acquires image data obtained by imaging a grinding surface fixed by a binder in a state where a plurality of abrasive grains of the grinding tool are dispersed. ,
A storage control unit that stores image data acquired before the grinding tool is processed by the image acquisition unit in a storage medium,
From the image data acquired after the processing of the grinding tool by the image acquisition unit, a detection unit that detects information indicating a state of a deteriorated portion on the grinding surface of the grinding tool,
An extraction unit configured to extract, from the image data before processing stored in the storage medium, an image before processing corresponding to the deteriorated portion and the position as information indicating a factor of a decrease in grinding performance, a grinding tool grinding surface evaluation apparatus.   The detection unit detects, from the image data obtained after the processing of the grinding tool, clogging of the abrasive grains on the grinding surface, damage of the abrasive grains, and slippage of the projection of the abrasive grains from the binder below a threshold value. The grinding tool grinding surface evaluation device according to claim 1, wherein an image representing at least one of the occurrence sites is detected as information indicating a state of the deteriorated site.   The extraction unit, extracted as information representing the grinding performance reduction factor, from the image before processing corresponding to the deteriorated site and position, the moving direction of the grinding tool of the abrasive grains included in the image before processing is The grinding tool grinding surface evaluation device according to claim 1, wherein the reflected feature amount is calculated, and the calculated feature amount is included in the information indicating the grinding performance reduction factor.   The learning device further includes a learning device generation unit that learns the learning device by providing information representing the grinding performance deterioration factor as learning data and providing information representing the state of the deteriorated portion of the grinding surface as teacher data to the learning device. The grinding tool grinding surface evaluation device according to any one of claims 1 to 3.   The learning device generation unit, to the learning device, at least one of an unprocessed image corresponding to the position of the degraded portion and the position and the feature amount of abrasive grains included in the image are provided as the learning data, and the degraded portion is provided. The grinding tool grinding surface evaluation device according to claim 4, wherein the learning device is made to learn by giving an image representing the following as the teacher data.   The learning device generation unit, as the learning data, the learning device is provided with at least one of the image including the unprocessed part corresponding to the position of the deteriorated part and the position and the feature amount of the abrasive grains included in the image, and The grinding tool grinding surface evaluation device according to claim 4, wherein the learning device is made to learn by giving, as the teacher data, an image representing a part before processing in which the deteriorated part and the position correspond to each other.   The learning device generation unit, as the learning data, the learning device is provided with at least one of the image including the unprocessed part corresponding to the position of the deteriorated part and the position and the feature amount of the abrasive grains included in the image, and The grinding tool grinding surface according to claim 4, wherein the learning device is made to learn by giving, as the teacher data, numerical data indicating a degree at which a part before processing corresponding to the deteriorated part and the position corresponds to a grinding performance reduction factor. Evaluation device.   Input the image data acquired by the image acquisition unit before or after machining of the grinding tool to the learning device, and at this time, information indicating the deteriorated portion of the grinding surface output from the learning device, The grinding tool grinding surface evaluation device according to claim 4, further comprising: an evaluation unit that outputs the information indicating the evaluation result of the grinding surface of the grinding tool.   The evaluation unit is based on information representing a deteriorated portion of the grinding surface output from the learning device, and the distribution of the deteriorated portion in the moving direction of the grinding tool in the entire area of the grinding surface of the grinding tool. The grinding tool grinding surface evaluation device according to claim 8, wherein a visualization image representing the is generated and output.   The said evaluation part estimates the time of the dressing of the said grinding tool based on the information which shows the deterioration part of the said grinding surface output from the said learning device, and outputs the information showing the estimation result. A grinding tool abrasive surface evaluation device according to item 1.   A learned learning device generated by a learning device generation unit included in the grinding tool grinding surface evaluation device according to claim 4.   A grinding tool grinding surface evaluation program for causing a processor of the grinding tool grinding surface evaluation device to execute the processing of each of the units included in the grinding tool grinding surface evaluation device according to any one of claims 1 to 10. A grinding tool grinding surface evaluation method executed by an information processing device including a processor and a storage medium,
While moving the grinding tool in a predetermined first direction before the start of machining, a first surface obtained by imaging a grinding surface fixed by a binder in a state in which a plurality of abrasive grains of the grinding tool are dispersed is obtained. A first obtaining step of obtaining the image data of the first type and storing the first image data in the storage medium;
A second acquisition step of acquiring second image data obtained by imaging a grinding surface of the grinding tool while moving the grinding tool in the first direction after processing is started;
From the acquired second image data, a detection step of detecting information indicating a deteriorated portion on the grinding surface,
From the first image data stored in the storage medium, an extraction step of extracting an image of a part before processing corresponding to the detected deteriorated part and the position as information representing a grinding performance reduction factor,
A learning device generation step of learning a learning device using information representing the grinding performance reduction factor as learning data, and information indicating a deteriorated portion of the grinding surface as teacher data,
Image data obtained before or after processing of the grinding tool is input to the learning device, and at this time, information indicating a deteriorated portion of the grinding surface output from the learning device is used as the grinding tool of the grinding tool. An evaluation process of outputting as information representing the evaluation result of the surface.