Grinding wheel determination device, determination program, and grinding wheel determination method

Abstract

To determine the deterioration state of a grindstone with high accuracy.SOLUTION: An automatic grinding device is provided with a vibration sensor that detects the vibration of a table. An electronic control device of a grindstone determination device performs frequency analysis of detection data detected by the vibration sensor (step S11). Analysis data obtained from the frequency analysis are stored in a storage part of the electronic control device as determination data (step S12). The storage part of the electronic control device stores a learning machine for vibration that has been learned by using the analysis data as teacher data in a case where grinding processing is performed by using a grindstone in a specific deterioration state. The electronic control device outputs a value according to a deterioration degree of the grindstone from the learned learning machine for vibration stored in the storage part with the determination data as input data (step S13).SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a grinding wheel determining apparatus for determining the state of a grinding wheel ,same a determination program used in the apparatus, and a grinding wheel determining method for determining the state of a grinding wheel. Background Art

[0002] Conventionally, grinding machines that grind the surface of a work as a processing object using a grinding wheel have been widely used (see, for example, Patent Document 1). In grinding performed by a grinding machine, the finishing quality of the processed surface of the work depends on the state of the grinding wheel serving as a grinding tool. Therefore, when performing grinding with a grinding machine, maintenance of the processing surface of the grinding wheel (the so-called grinding surface) is regularly performed. This maintains the processing accuracy of grinding performed by the grinding machine. Prior Art Documents Patent Documents

[0003] Patent Document 1 Japanese Unexamined Patent Publication No. 2002-127006 Summary of the Invention Problems to be Solved by the Invention

[0004] In a grinding machine, in order to perform maintenance of the grinding wheel at an appropriate timing matching the state of the grinding wheel, it is desired to determine the state of the grinding wheel with high accuracy. Means for Solving the Problems

[0005] A grinding wheel determination device for solving the above problems is applied to a processing machine that performs grinding using a grinding wheel, and is a grinding wheel determination device for determining the condition of the grinding wheel, comprising: a vibration detection unit provided in the processing machine that detects data related to the vibration of the processing machine; a vibration analysis unit that performs frequency analysis of the detection data detected by the vibration detection unit; a vibration input unit that stores the analysis data obtained as a result of the frequency analysis by the vibration analysis unit as determination data; a vibration storage unit that stores a vibration learner that has been trained using the analysis data from when grinding is performed using a grinding wheel in a specific deteriorated state as training data; and a vibration output unit that takes the determination data as input data and outputs a value corresponding to the degree of deterioration of the grinding wheel from the trained vibration learner stored in the vibration storage unit.

[0006] The learning device for solving the above problem is the learned vibration learning device provided in the grinding wheel determination device described above. The determination program for solving the above problem is a determination program that causes the electronic control device equipped in the grinding wheel determination device to execute the processing of the vibration detection unit, the vibration analysis unit, the vibration input unit, the vibration storage unit, and the vibration output unit, all of which are provided in the grinding wheel determination device.

[0007] A grinding wheel determination method for solving the above problems is applied to a processing machine that performs grinding with a grinding wheel, and is a grinding wheel determination method for determining the condition of the grinding wheel, comprising: a detection step of detecting data related to the vibration of the processing machine by a vibration detection unit provided in the processing machine; an analysis step of performing frequency analysis on the detection data detected by the vibration detection unit; an input step of storing the analysis data obtained as a result of the frequency analysis as determination data in the vibration memory unit of the grinding wheel determination device; a storage step of storing in the vibration memory unit a vibration learner that has been trained using the analysis data in the case of grinding using a grinding wheel in a specific deteriorated state as training data; and an output step of outputting a value corresponding to the degree of deterioration of the grinding wheel from the trained vibration learner stored in the vibration memory unit, using the determination data as input data.

[0008] When grinding a workpiece with a grinding wheel, the machine vibrates. As the grinding wheel deteriorates, the grinding action of the workpiece changes, and consequently, the vibration pattern of the machine also changes.

[0009] According to the grinding wheel determination device, learning device, determination program, and grinding wheel determination method described above, data (training data) corresponding to the analysis data when grinding is performed using a grinding wheel in a specific deteriorated state is prepared. Then, a pre-trained vibration learning device is prepared using this training data. Therefore, by using the pre-trained vibration learning device to capture the change in the vibration pattern of the processing machine due to the deterioration of the grinding wheel, the deterioration state of the grinding wheel can be determined with high accuracy. [Effects of the Invention]

[0010] According to the present invention, the deterioration state of a grinding wheel can be determined with high accuracy. [Brief explanation of the drawing]

[0011] [Figure 1] This is a schematic diagram showing the general configuration of an automatic grinding apparatus to which a grinding wheel determination device of one embodiment is applied. [Figure 2] This is a cross-sectional view showing the structure of the grinding wheel cover and its surroundings in the open state. [Figure 3] This is a diagram showing the schematic configuration of a grinding wheel determination device. [Figure 4] This is a flowchart showing the procedure for executing the vibration detection process. [Figure 5] This is an explanatory diagram illustrating the learning mechanism of a learning device. [Figure 6] This is an explanatory diagram to illustrate the generation error output from the learning model. [Figure 7] This is a flowchart showing the procedure for executing the image recognition process. [Figure 8] This is a flowchart showing the procedure for executing the imaging process. [Figure 9] This is a flowchart showing the execution procedure for the comprehensive judgment process. [Figure 10]It is a graph showing an example of the relationship between the grinding amount of a grinding wheel and the generated error ΔP1.

Mode for Carrying Out the Invention

[0012] Hereinafter, an embodiment of a grinding wheel determination device, a learner, a determination program, and a grinding wheel determination method will be described. (Automatic grinding device as processing machine) As shown in FIG. 1, the grinding wheel determination device 30 of the present embodiment is applied to a numerically controlled (NC) automatic grinding device 20.

[0013] The automatic grinding device 20 includes a substantially disk-shaped grinding wheel 21 as a grinding tool. The grinding wheel 21 has a structure in which a large number of abrasive grains dispersed on the outer circumferential surface of a base material are fixed by a binder. The grinding wheel 21 is rotatably supported. For example, a rotation drive unit 22 having a servo motor is connected to the grinding wheel 21. The rotation drive unit 22 includes a position sensor 45 for detecting the rotational position of the grinding wheel 21.

[0014] The automatic grinding device 20 includes a support base 201. A table 202 is provided on the support base 201 in a relatively movable state. The table 202 is configured to allow a workpiece (hereinafter referred to as workpiece W) to be attached and detached. The table 202 is provided with a vibration sensor 46 for detecting vibration thereof. In the present embodiment, a vibration sensor 46 capable of detecting vibration in a predetermined frequency range (for example, 10 kHz to 1 MHz) is employed. Specifically, an AE (Acoustic Emission) sensor is employed as the vibration sensor 46. It is preferable to employ a vibration sensor 46 having a detectable frequency range of "several tens of kHz to several MHz". In the present embodiment, the vibration sensor 46 corresponds to a vibration detection unit.

[0015] The automatic grinding apparatus 20 is configured to be capable of relatively moving a grinding wheel 21 and a table 202. When performing grinding, the automatic grinding apparatus 20 relatively moves the grinding wheel 21 and the table 202 while rotationally driving the grinding wheel 21, with a workpiece W fixed to the table 202. At this time, by bringing the grinding wheel 21 into contact with the workpiece W, an upper surface of the workpiece W is ground by a grinding surface 23 of the grinding wheel 21.

[0016] The automatic grinding apparatus 20 includes a supply device 203 that supplies grinding fluid to a processed portion by the grinding wheel 21 (specifically, a portion where the grinding wheel 21 and the workpiece W are in contact with each other). The supply device 203 includes a reservoir (not shown) that stores the grinding fluid, a pump (not shown) that pressure-feeds the grinding fluid in the reservoir, a nozzle 204 that injects the grinding fluid toward the processed portion, and the like. When grinding is performed by the automatic grinding apparatus 20, the grinding fluid is supplied to the processed portion by the supply device 203.

[0017] As shown in Fig. 1 and Fig. 2, the automatic grinding apparatus 20 includes a grinding wheel cover 25. The grinding wheel cover 25 is formed in a rectangular box shape with an open lower portion. The grinding wheel 21 is housed inside the grinding wheel cover 25 in a state where a lower portion of the grinding wheel 21 is exposed to the outside.

[0018] An upper wall portion 26, which forms an upper wall of the grinding wheel cover 25, is provided with an opening 27 that communicates the inside and outside of the grinding wheel cover 25. The grinding wheel cover 25 is provided with a plate-shaped shutter portion 28 that opens and closes the opening 27, and an actuator (hereinafter, shutter operation portion 29) for reciprocating the shutter portion 28. By moving the shutter portion 28 in one direction through operation control of the shutter operation portion 29, a closed state (the state shown in Fig. 1) is achieved in which the opening 27 of the grinding wheel cover 25 is blocked by the shutter portion 28. On the other hand, by moving the shutter portion 28 in the other direction through operation control of the shutter operation portion 29, an open state (the state shown in Fig. 2) is achieved in which the shutter portion 28 is retracted from the opening 27 of the grinding wheel cover 25. In the present embodiment, switching between the closed state where the opening 27 of the grinding wheel cover 25 is closed and the open state where the opening 27 is opened can be performed through operation control of the shutter operation portion 29.

[0019] A camera 31 is mounted on the grinding wheel cover 25 as an imaging unit. The camera 31 is mounted in such a manner that it can image the grinding surface 23 of the grinding wheel 21 housed inside the grinding wheel cover 25 through an opening 27 in the grinding wheel cover 25. An area scan camera is used as the camera 31. In this embodiment, the camera 31 images the grinding surface 23 of the grinding wheel 21 at a predetermined magnification, and the captured image (more specifically, image data) is output to an electronic control device 40, which will be described later. In this embodiment, an illuminator (not shown) for illuminating the grinding surface 23 of the grinding wheel 21 when imaging is performed by the camera 31 is integrally provided with the camera 31.

[0020] The camera 31 is attached to the grinding wheel cover 25 via an X-axis drive unit 32, a Y-axis drive unit 33, and a Z-axis drive unit 34. These X-axis drive unit 32, Y-axis drive unit 33, and Z-axis drive unit 34 each have a sliding mechanism and an actuator (a servo motor in this embodiment) that operates the sliding mechanism. In this embodiment, the camera 31 can be moved relative to the grinding wheel cover 25 in the X direction (left-right direction in Figure 1) through the operation control of the X-axis drive unit 32. The camera 31 can also be moved relative to the grinding wheel cover 25 in the Y direction (up-down direction in Figure 1) through the operation control of the Y-axis drive unit 33. Furthermore, the camera 31 can be moved relative to the grinding wheel cover 25 in the Z direction (depth direction in Figure 1) through the operation control of the Z-axis drive unit 34.

[0021] (Electronic control unit) The automatic grinding machine 20 has an electronic control unit 40. As shown in Figure 3, the electronic control unit 40 has a CPU 41, a ROM 42, a RAM 43, and a storage unit 44 for storing various programs and data. The storage unit 44 is composed of a non-volatile memory that can be written to and read at any time, such as a hard disk drive or a solid-state drive. The electronic control unit 40 is configured to perform various calculation processes based on various programs and data. Based on the calculation results, the electronic control unit 40 performs various controls related to numerical control of the grinding process and determination of the state of the grinding wheel 21. In this embodiment, the electronic control unit 40 corresponds to a vibration analysis unit, an image output unit, a vibration output unit, and an execution unit.

[0022] The automatic grinding device 20 has a display device 35. This display device 35 is connected to an electronic control device 40. The display device 35 displays the result of determining the condition of the grinding wheel 21.

[0023] The grinding wheel determination device 30 of this embodiment is a device that performs two types of determination to determine the state of the grinding wheel 21: a determination based on the vibration of the automatic grinding device 20 (specifically, its table 202) (hereinafter referred to as vibration determination) and a determination based on an image of the grinding surface 23 (hereinafter referred to as image determination). In vibration determination, the state of the grinding wheel 21 is determined by a pre-trained vibration learner 51A using determination data 55, which is data obtained by processing the detection signal of the vibration sensor 46, as input data. In image determination, the state of the grinding wheel 21 is determined by a pre-trained image learner 51B using an image (determination image 50) of the grinding surface 23 of the grinding wheel 21 captured by the camera 31 as input data.

[0024] The memory unit 44 of the electronic control unit 40 has a memory area for storing various execution programs 52 and a memory area for storing training data 53A and 53B used for machine learning of the learners 51A and 51B. The training data 53A is data used to perform machine learning on the vibration learner 51A so that it can acquire the ability to determine the state of the grinding wheel 21. The training data 53B is data used to perform machine learning on the image learner 51B so that it can acquire the ability to determine the state of the grinding wheel 21.

[0025] The memory unit 44 has a memory area that functions as a vibration memory unit for storing a learned vibration learner 51A (more specifically, training data created through the execution of machine learning). The memory unit 44 also has a memory area that functions as an image memory unit for storing a learned image learner 51B (more specifically, training data created through the execution of machine learning). Furthermore, the memory unit 44 has a memory area that functions as a vibration input unit for storing judgment data 55, a memory area that functions as an image input unit for storing judgment images 50, and a memory area that stores judgment result data 54 indicating the result of determining the state of the grinding wheel 21.

[0026] The execution program 52 includes a program for executing machine learning on the learners 51A and 51B and for generating training data, which is data showing the results of the machine learning. The execution program 52 also includes a determination program for executing various processes related to determining the state of the grinding wheel 21 and for generating determination result data 54 that shows the determination result.

[0027] (Vibration detection) The following provides a detailed explanation of vibration detection. Figure 4 is a flowchart showing the execution procedure for vibration detection (vibration detection process). Note that the series of processes shown in this flowchart conceptually represents the execution procedure for vibration detection; the actual process is executed by the electronic control device 40 at predetermined cycle intervals.

[0028] As shown in Figure 4, this process first executes a process to detect vibrations of the table 202 (vibration detection process) (step S11). In this embodiment, the process in step S11 corresponds to the detection step.

[0029] In the vibration detection process, while grinding is being performed by the automatic grinding device 20, the vibration waveform of the table 202 over a predetermined period of time (specifically, the processing period for processing one workpiece W) is detected by the vibration sensor 46. In this embodiment, vibration detection of the table 202 is performed for multiple passes (for example, 8 passes) that are executed at the midpoint of the processing period. More specifically, vibration detection of the table 202 is performed for multiple up-cuts (for example, 3 passes) that are executed around the midpoint of each pass. Then, in the vibration detection process, a dataset is created consisting of each vibration waveform detected by the vibration sensor 46. The dataset created is time-series data with "time" on the horizontal axis and "amplitude" on the vertical axis.

[0030] Subsequently, a process for analyzing the time-series data (vibration analysis process) is executed (step S12). In this embodiment, the process in step S12 corresponds to the analysis step. In the vibration analysis process, the time-series data is converted into a spectrum by performing a short-time Fourier transform, thereby creating judgment data 55. More specifically, in the vibration analysis process, the time-series data is extracted using a window function (Hanning window in this embodiment) while changing the extraction start time for each data point within a predetermined time window TW, and then shaped into judgment data 55. In this embodiment, the range of window TW is set to a range corresponding to the length of the portion in contact between the workpiece W and the grinding surface 23 during grinding (for example, the range in which the grinding wheel 21 rotates by 35 degrees). This makes it possible to obtain frequency characteristics in units of the contact range in which the workpiece W and the grinding surface 23 come into contact during grinding. Therefore, the deterioration state of the grinding wheel 21 can be analyzed in units of such contact ranges.

[0031] In the vibration analysis process, analysis data (spectrogram) is generated as judgment data 55, where the vertical axis is "frequency" and the horizontal axis is "time". In this spectrogram, the intensity of vibration is represented by color. Specifically, in the spectrogram, areas with lower vibration intensity are represented by darker blue colors, and areas with higher intensity are represented by brighter red colors. In this embodiment, vibration data included in one dataset becomes one spectrogram (judgment data 55). This judgment data 55 is then stored in the storage unit 44. In this embodiment, the processing in step S12 corresponds to the input process.

[0032] (Vibration learning device) In this embodiment, an efficient generative adversarial network (GAN) is used as the vibration learner 51A. In this embodiment, machine learning is performed on the learner 51A based on the training data 53A stored in the memory unit 44. Specifically, a learning process is repeatedly performed for each image data that makes up the training data 53A, in which one of the multiple image data that make up the training data 53A is extracted and the learner 51A is trained using this image data as input data. The learning data created through the execution of machine learning is then stored in the memory unit 44. In this embodiment, an execution program 52 is constructed and stored in the memory unit 44 so that this series of processes related to the machine learning of the learner 51A is executed automatically. In this embodiment, the process of executing machine learning of the learner 51A in this way corresponds to the storage process.

[0033] When acquiring training data 53A to be used for machine learning of the vibration learning device 51A, a grinding wheel 21[A] in a non-degraded state, such as an unused grinding wheel or a grinding wheel that has been properly dressed, is prepared. This grinding wheel 21[A] is attached to the automatic grinding device 20. Then, vibration detection processing (processing in step S11) and vibration analysis processing (processing in step S12) are performed by the electronic control device 40. The image data (specifically, the spectrogram described above) acquired through the vibration detection processing and vibration analysis processing is stored in the storage unit 44 as training data 53A to be used for machine learning of the learning device 51A. Thus, in this embodiment, the analysis data (spectrogram) when grinding is performed using a grinding wheel 21[A] in a non-degraded state is defined as training data 53A. In this embodiment, image data for multiple grinding wheels 21[A] is acquired, and training data 53A is composed of these image data. In this embodiment, the state in which the grinding wheel 21 is not deteriorated corresponds to a "specific deterioration state".

[0034] As conceptually shown in Figure 5, the machine learning of the vibration learner 51A is performed in a manner that the input data and the generated image data generated by the learner 51A based on the input data are made to represent the same image (image data A). Then, the trained learner 51A outputs a value equivalent to the difference between the input data and the generated image data as the generation error ΔP1. As a result, if the learner 51A has been properly trained, when the process for determining the state of the grinding wheel 21 to be judged (vibration judgment process) is executed, the learner 51A outputs the following value as the generation error ΔP1.

[0035] If the grinding wheel 21 being judged is not deteriorated, the judgment data 55, which is the input data, and the generated image data generated by the learner 51A will be approximately identical. In this case, the generation error ΔP1 output from the learner 51A will be a very small value (for example, "0").

[0036] On the other hand, if the grinding wheel 21 being judged is in a deteriorated state, the judgment data 55, which is the input data, and the generated image data will no longer match, as shown in Figure 6. In this case, the generation error ΔP1 output from the learning device 51A will be a "positive value" corresponding to the difference between the input data and the generated image data. Moreover, in this case, the greater the degree of deterioration of the grinding wheel 21, the larger the difference between the input data and the generated image data, and therefore the larger the generation error ΔP1 output from the learning device 51A will be.

[0037] Therefore, in the judgment process using the vibration learning device 51A, the larger the generation error ΔP1, the more advanced the deterioration of the grinding wheel 21 being judged. (Vibration detection process) As shown in Figure 4, in the vibration determination process, after the vibration analysis process is executed (step S12), a first output process is executed to output a generation error ΔP1 (step S13). In this embodiment, the process in step S13 corresponds to the output process. In the first output process, the generation error ΔP1 is output from the learned vibration learner 51A using the determination data 55 as input data. In the grinding wheel determination device 30 of this embodiment, the degree of deterioration of the grinding wheel 21 is determined based on this generation error ΔP1.

[0038] (Image recognition) The image recognition method mentioned above will be explained in more detail below. Figure 7 is a flowchart showing the execution procedure for the image determination process. Note that the series of processes shown in this flowchart conceptually represents the execution procedure for the image determination process; the actual process is executed by the electronic control unit 40 at predetermined cycles.

[0039] As shown in Figure 7, in this process, first, an imaging process is performed in which the camera 31 images the grinding surface 23 of the grinding wheel 21 in order to acquire the judgment image 50 (step S21). The procedure for executing this imaging process will be explained with reference to Figure 8. Figure 8 conceptually illustrates the procedure for executing the imaging process. The series of processes shown in the flowchart of Figure 8 are executed by the electronic control unit 40.

[0040] As shown in Figure 8, in the imaging process, the shutter unit 28 is first opened through the operation control of the shutter operation unit 29 (step S31). Subsequently, the grinding wheel 21 is rotated in a manner suitable for imaging the grinding surface 23, and imaging of the grinding surface 23 is performed by the camera 31 (step S32). In this embodiment, the rotation drive pattern of the grinding wheel 21, the position control pattern of the camera 31, and the imaging pattern of the camera 31 are predetermined and stored in the electronic control unit 40 so as to enable the camera 31 to efficiently image the entire surface of the grinding surface 23. The following methods (Method A) and (Method B) are assumed as methods for storing the above patterns in the storage unit 44. (Method A) After the user of the automatic grinding device 20 has attached the grinding wheel 21 to be used, the automatic grinding device 20 is operated to determine the above patterns suitable for the grinding wheel 21 and store them in the electronic control unit 40. (Method B) The manufacturer of the automatic grinding device 20 has predetermined the above patterns that enable the camera 31 to efficiently image the entire surface of the grinding surface 23 of the grinding wheel 21 based on the results of various experiments and simulations. Then, the data related to these patterns is stored in the automatic grinding machine 20 before shipment, or in the automatic grinding machine 20 installed in the user's factory.

[0041] In step S32, the operation control of the rotation drive unit 22 is performed based on the rotation drive pattern of the grinding wheel 21, and the operation control of the X-axis drive unit 32, Y-axis drive unit 33, and Z-axis drive unit 34 is performed based on the position control pattern of the camera 31. In addition, the camera 31 takes images of the grinding surface 23 based on the imaging pattern of the camera 31. The electronic control device 40 then takes the image data of the grinding surface 23 captured by the camera 31 and stores it in the storage unit 44. Specifically, the camera 31 takes images of the grinding surface 23 in multiple stages. The multiple captured images are then combined in a predetermined order, arranged vertically and horizontally, to generate a single image (judgment image 50) that shows the entire grinding surface 23. The electronic control device 40 stores the image generated in this way in the storage unit 44.

[0042] Once the camera 31 has finished imaging the grinding surface 23, the shutter unit 28 is closed through the operation control of the shutter operation unit 29 (step S33). The imaging process shown in Figure 8 is performed on the condition that grinding of the workpiece W by the grinding wheel 21 has not been performed. Prior to performing the imaging process, if necessary, the cutting oil on the surface of the grinding surface 23 is removed by blowing air onto the grinding surface 23 to be imaged.

[0043] (Image learning machine) In this embodiment, an efficient generative adversarial network (Efficient-GAN) is used as the image learning model 51B. In this embodiment, machine learning is performed on the learning model 51B based on the training data 53B stored in the memory unit 44. Specifically, a learning process is repeatedly performed for each image data that makes up the training data 53B, in which one of the multiple image data that make up the training data 53B is extracted and the learning model 51B is trained using that image data as input data. The learning data created through the execution of machine learning is then stored in the memory unit 44. In this embodiment, an execution program 52 is constructed and stored in the memory unit 44 so that this series of processes related to the machine learning of the learning model 51B is executed automatically.

[0044] When acquiring training data 53B to be used for machine learning of the learning device 51B, a grinding wheel 21[A] in an undegraded state is prepared. This grinding wheel 21[A] is attached to the automatic grinding device 20. Then, imaging processing (see Figure 8) is performed by the electronic control device 40 in that state. The image data acquired through the imaging processing is then stored in the storage unit 44 as training data 53B to be used for machine learning of the learning device 51B. In this embodiment, image data for multiple grinding wheels 21[A] is acquired, and the training data 53B is composed of these image data.

[0045] As conceptually shown in Figure 5, the machine learning of the learner 51B is performed in a manner in which the input data and the generated image data generated by the learner 51B based on the same input data are made to represent the same image (image data A). Then, the trained learner 51B outputs a value equivalent to the difference between the input data and the generated image data as the generation error ΔP2. As a result, if the learner 51B has been properly trained, when the process of determining the state of the grinding wheel 21 to be judged (image judgment process) is executed, the learner 51B outputs the following value as the generation error ΔP2.

[0046] If the grinding wheel 21 being judged is not deteriorated, the judgment image 50, which is the input data, and the generated image data generated by the learning device 51B will be approximately identical. In this case, the generation error ΔP2 output from the learning device 51B will be a very small value (for example, "0").

[0047] On the other hand, if the grinding wheel 21 being judged is in a deteriorated state, the input data and the generated image data will no longer match, as shown in Figure 6. In this case, the generation error ΔP2 output from the learning device 51B will be a "positive value" corresponding to the difference between the input data and the generated image data. Moreover, in this case, the greater the degree of deterioration of the grinding wheel 21, the larger the difference between the input data and the generated image data, and therefore the larger the generation error ΔP2 output from the learning device 51B will be.

[0048] Therefore, in the judgment process using the image learning unit 51B, the larger the generation error ΔP2, the more advanced the deterioration of the grinding wheel 21 being judged. Here, when the grinding wheel 21 is not deteriorated, the outermost surface of the grinding surface 23 is composed only of the tip surfaces of the abrasive grains. When grinding is performed by the grinding wheel 21 and the grinding wheel 21 (more specifically, the grinding surface 23) deteriorates, parts other than abrasive grains (mainly parts composed of binders) appear on the outermost surface of the grinding wheel 21. In this way, the degree of deterioration of the grinding wheel 21 is reflected on the outermost surface of the grinding surface 23. In this embodiment, images of the grinding surface 23 are used as image data to constitute the judgment image 50 and the training data 53B. Therefore, the image data to constitute the judgment image 50 and the training data 53B can be said to contain information about the outermost surface of the grinding surface 23, more specifically information about characteristic parts of the grinding surface 23, as information representing the deterioration state of the grinding surface 23. Accordingly, in this embodiment, when performing machine learning on the image learning unit 51B, the state of the outermost surface of the grinding surface 23 can be extracted as a feature, and the machine learning of the learning unit 51B can be advanced based on this feature.

[0049] As shown in Figure 7, in the image determination process, after the imaging process is performed (step S21), a second output process is performed to output the generation error ΔP2 (step S22). In the second output process, the determination image 50 is used as input data, and the generation error ΔP2 is output from the trained image learner 51B.

[0050] Subsequently, it is determined whether the generation error ΔP2 is greater than or equal to the second determination value J2 (step S23). In this embodiment, the generation error ΔP2 at the time when dressing should be performed on the grinding wheel 21 to be judged is predetermined based on the results of various experiments and simulations conducted by the inventors. The value corresponding to this generation error ΔP2 is then defined as the second determination value J2 and stored in the storage unit 44.

[0051] Then, if the generation error ΔP2 is greater than or equal to the second determination value J2 (step S23: YES), it is determined that the deterioration of the grinding wheel 21 has progressed beyond a predetermined level, and that it is time to perform dressing on the grinding wheel 21 to be judged (step S24). In this embodiment, the judgment result from the image judgment process is stored in the storage unit 44 as judgment result data 54 and is also displayed on the display device 35. If the generation error ΔP2 is greater than or equal to the determination value J2, the display device 35 displays text information (dressing timing) indicating that it is time to perform dressing. The user can understand that it is time to perform dressing based on the judgment result data 54 stored in the storage unit 44 and the judgment result displayed on the display device 35.

[0052] On the other hand, if the generation error ΔP2 is less than the second judgment value J2 (step S23: NO), the deterioration of the grinding wheel 21 is considered to be within the acceptable range, and the process in step S24 is skipped. (Overall judgment process) In this embodiment, the state of the grinding wheel 21 is determined by combining vibration detection processing (see Figure 4) and image detection processing (see Figure 7).

[0053] The following describes the procedure for executing such processing (overall judgment processing) with reference to Figure 9. Figure 9 is a flowchart showing the execution procedure for the comprehensive judgment process. The series of processes shown in this flowchart conceptually illustrates the execution procedure for the comprehensive judgment process; the actual processes are executed by the electronic control unit 40 at predetermined cycles.

[0054] As shown in Figure 9, this process first performs a vibration detection process (see Figure 4) (step S41). In step S41, the vibration learning device 51A outputs a generation error ΔP1 based on the detection data 55.

[0055] Thereafter, it is determined whether or not the generation error ΔP1 is equal to or greater than a first determination value J1 (step S42). When the generation error ΔP1 is less than the first determination value J1 (step S42: NO), the image determination process is not executed (the process of step S43 is skipped).

[0056] FIG. 10 shows an example of the relationship between the grinding amount of the grinding wheel 21 and the generation error ΔP1. As shown in FIG. 10, when no grinding process by the grinding wheel 21 is performed after dressing, that is, when the grinding wheel 21 is not deteriorated (time t11), the generation error ΔP1 output from the learner 51A in the vibration determination process has a small value. When grinding by the grinding wheel 21 is performed and deterioration of the grinding wheel 21 progresses, the generation error ΔP1 gradually increases.

[0057] In the present embodiment, during a period (from time t11 to t12) until the generation error ΔP1 becomes equal to or greater than the first determination value J1, the image determination process is not executed, and only the vibration determination process is executed. That is, during a period from when the grinding wheel 21 is not deteriorated until the degree of deterioration of the grinding wheel 21 reaches a predetermined level, the image determination process is not executed, and only the vibration determination process is executed.

[0058] As shown in FIG. 9, when the present process is repeatedly performed thereafter and the generation error ΔP1 becomes equal to or greater than the first determination value J1 (step S42: YES), the image determination process (see FIG. 7) is executed (step S43). In the present embodiment, when the degree of deterioration of the grinding wheel 21 reaches a predetermined level, the image determination process and the vibration determination process are executed thereafter.

[0059] As described above, in the present embodiment, when the generation error ΔP1 output in the vibration determination process (see FIG. 4) is a value (ΔP1<J1) indicating that the degree of deterioration of the grinding wheel 21 is less than a predetermined level, the image determination process (see FIG. 7) is not executed. When the generation error ΔP1 becomes a value (ΔP1≧J1) indicating that the degree of deterioration of the grinding wheel 21 is equal to or higher than a predetermined level, the image determination process is executed.

[0060] (Effects and Benefits) The following explains the effects and benefits of performing this comprehensive judgment process. In the automatic grinding device 20, grinding fluid is supplied to the workpiece during grinding with the grinding wheel 21. In the automatic grinding device 20, a camera 31 for imaging the grinding wheel 21 is provided near the grinding surface 23. Imaging of the grinding surface 23 by the camera 31 is not successful if grinding fluid adheres to the camera 31. Therefore, in this embodiment, imaging by the camera 31 is performed when grinding of the workpiece W by the grinding wheel 21 is not being performed, that is, when grinding fluid is not being supplied. When imaging by the camera 31 is not being performed, the shutter unit 28 is closed (as shown in Figure 2) to prevent grinding fluid from coming into contact with the camera 31. Thus, in the grinding wheel determination device 30, imaging by the camera 31 and determination based on the captured images are easily affected by grinding fluid. Therefore, it can be said that the grinding wheel determination device 30 has many constraints on imaging by the camera 31 and determination based on the captured images.

[0061] In contrast, the vibration sensor 46, due to its structure, can accurately detect vibrations even if grinding fluid adheres to it. Furthermore, the vibration sensor 46 can be installed at a distance from the grinding wheel 21. Therefore, in the grinding wheel determination device 30, the detection of vibrations by the vibration sensor 46 and the determination based on the detection signal from the vibration sensor 46 are less affected by the grinding fluid. Thus, the grinding wheel determination device 30 has fewer constraints on detection by the vibration sensor 46 and determination based on the detection signal compared to determination based on images.

[0062] Furthermore, from the results of various experiments and simulations conducted by the inventors, it was found that judgment based on images of the grinding surface 23 can determine the deterioration of the grinding wheel 21 with higher accuracy compared to judgment based on vibration of the table 202.

[0063] According to the comprehensive judgment process of this embodiment, when the deterioration of the grinding wheel 21 is not very advanced, a vibration judgment based on the vibration of the table 202 (see Figure 4) can be performed, that is, a judgment that is less affected by the grinding fluid. In this case, through the vibration judgment process, the vibration of the table 202 can be easily detected without considering the effects of the grinding fluid even during the grinding process. Therefore, the degree of deterioration of the grinding wheel 21 can be determined in a timely manner based on the vibration of the table 202 detected in this way.

[0064] Furthermore, once the deterioration of the grinding wheel 21 has progressed to a certain extent, the degree of deterioration of the grinding wheel 21 can be determined by image judgment based on an image of the grinding surface 23 (see Figure 7), in other words, by a judgment method that is susceptible to the influence of the grinding fluid but is expected to have high judgment accuracy. According to this embodiment, the execution period of image judgment can be shortened because image judgment is not performed when the deterioration of the grinding wheel 21 has not progressed to a certain extent. Then, when the deterioration of the grinding wheel 21 has progressed to a certain extent, the deterioration of the grinding wheel 21 can be determined with high accuracy by performing image judgment thereafter.

[0065] Thus, according to this embodiment, the deterioration of the grinding wheel 21 can be determined with a high degree of freedom by utilizing vibration determination based on the vibration of the table 202 and image determination based on an image of the grinding surface 23.

[0066] (effect) According to this embodiment, the following effects can be obtained. (1) A vibration sensor 46 is provided on the table 202 to detect vibrations of the table 202. The electronic control unit 40 performs frequency analysis on the detection data detected by the vibration sensor 46. The analysis data obtained as a result of the frequency analysis (specifically, a spectrogram) is stored in the memory unit 44 of the electronic control unit 40 as judgment data 55. The analysis data when grinding is performed using a grinding wheel 21[A] that is not deteriorated is used as training data 53A. A vibration learner 51A that has learned from this training data 53A is stored in the memory unit 44 of the electronic control unit 40. The electronic control unit 40 takes the judgment data 55 as input data and outputs a value (generation error ΔP1) corresponding to the degree of deterioration of the grinding wheel 21 from the learned vibration learner 51A stored in the memory unit 44.

[0067] The automatic grinding device 20 vibrates when grinding the workpiece W with the grinding wheel 21. When the grinding wheel 21 deteriorates, the grinding pattern of the workpiece W by the grinding wheel 21 changes, and consequently, the vibration pattern of the automatic grinding device 20 also changes.

[0068] According to this embodiment, training data 53A is prepared that corresponds to the analysis data (specifically, the spectrogram) obtained when grinding is performed using a grinding wheel 21[A] that is not deteriorated. Then, a pre-trained vibration learner 51A is prepared, which has been trained in advance using this training data 53A. Therefore, by using the pre-trained vibration learner 51A to capture the change in the vibration pattern of the table 202 due to the deterioration of the grinding wheel 21, the deterioration state of the grinding wheel 21 can be determined with high accuracy.

[0069] (2) The analysis data was obtained by performing a short-time Fourier transform on time-series data consisting of detection values ​​from the vibration sensor 46 over a predetermined period of time. According to this embodiment, a "spectrogram" can be obtained by analyzing time-series data consisting of detection values ​​from the vibration sensor 46. Then, by using the image data showing this "spectrogram" as training data 53A for the vibration learner 51A, or as judgment data 55 in judgment based on the learner 51A, the deterioration of the grinding wheel 21 can be determined with high accuracy.

[0070] (3) The storage unit 44 of the electronic control unit 40 stores an image of the grinding surface 23 of the grinding wheel 21 to be judged as the judgment image 50. The storage unit 44 of the electronic control unit 40 stores a trained image learner 51B that has been trained using training data 53B consisting of images of the grinding surface 23 of the grinding wheel 21 in an undegraded state. The electronic control unit 40 takes the judgment image 50 as input data and outputs a value (generation error ΔP2) from the trained image learner 51B corresponding to the degree of deterioration of the grinding wheel 21. The electronic control unit 40 does not perform the image judgment process if the generation error ΔP1 output from the vibration learner 51A is less than the first judgment value J1, and performs the process if the generation error ΔP1 is equal to or greater than the first judgment value J1.

[0071] According to this embodiment, an image of the grinding surface 23 of the grinding wheel 21[A] in an undeteriorated state is prepared, and a trained image learner 51B is prepared, which has been pre-trained using this image as training data 53B. Therefore, the degree of deterioration of the grinding wheel 21 to be judged can be determined with high accuracy using the trained image learner 51B. Furthermore, according to this embodiment, the deterioration of the grinding wheel 21 can be judged with a high degree of freedom by using vibration judgment based on the vibration of the table 202 and image judgment based on the image of the grinding surface 23.

[0072] (4) The automatic grinding device 20 is equipped with a camera 31 that captures images of the grinding surface 23 of the grinding wheel 21. The images captured by the camera 31 are stored as judgment images 50 in the storage unit 44 of the electronic control device 40. The automatic grinding device 20 has a shutter unit 28 that covers the grinding surface 23 of the grinding wheel 21 and extends along the grinding surface 23, and a shutter operation unit 29 that opens and closes the shutter unit 28. When the grinding surface 23 is being imaged, the shutter operation unit 29 opens so that the space between the grinding surface 23 and the camera 31 is not separated by the shutter unit 28, while when the grinding surface 23 is not being imaged, the shutter operation unit 29 closes so that the space between the grinding surface 23 and the camera 31 is separated by the shutter unit 28.

[0073] This allows the operation of imaging the grinding surface 23 and the operation of storing the image of the grinding surface 23 in the electronic control device 40 to be easily performed with the grinding wheel 21 attached to the automatic grinding device 20, without having to attach or detach the camera 31 to the automatic grinding device 20 in accordance with these operations. Furthermore, when imaging of the grinding surface 23 by the camera 31 is not performed, the shutter unit 28 can be closed to create a closed state where the camera 31 and the grinding wheel 21 are separated. This prevents foreign matter (grinding fluid, chips, oil, etc.) from adhering to the camera 31. When imaging is performed to capture the grinding surface 23 by the camera 31, the shutter unit 28 can be opened to create an open state where the camera 31 and the grinding wheel 21 are not separated, enabling imaging of the grinding surface 23 by the camera 31.

[0074] (Example of change) The above embodiment can be implemented with the following modifications. The above embodiment and the following modifications can be combined with each other to the extent that they do not contradict each other technically.

[0075] The device does not only display the result of determining the state of the grinding surface 23 on the display device 35, but may also sound a warning buzzer or emit audio information from a speaker depending on the determination result.

[0076] The configuration for displaying the determination result of the condition of the grinding surface 23 on the display device 35 may be omitted. The vibration sensor 46 can be any sensor, not limited to an AE sensor, as long as it can detect vibrations within a predetermined frequency range (10kHz to 1MHz, or several tens of kHz to several MHz). The mounting location of the vibration sensor 46 can be changed to any location, not limited to the table 202, as long as it can detect vibrations generated during the grinding process of the workpiece W by the grinding wheel 21.

[0077] In the vibration detection process, the method for creating the dataset (time-series data) can be arbitrarily changed, as long as it is a method that forms the data based on vibration waveforms detected by the vibration sensor 46 over a predetermined period of time. For example, vibration detection in table 202 in the vibration detection process may be performed for all of the multiple passes during the processing period, for all of the multiple up-cuts performed in each pass, or for the down-cuts in each pass.

[0078] The image captured by a camera mounted on a device other than the automatic grinding device 20 on which the grinding wheel 21 to be judged is installed (such as a dedicated imaging device or another automatic grinding device) may be stored in the electronic control device 40 and used as training data 53B.

[0079] As the judgment image 50 and training data 53B, it is not limited to using a single image showing the entire grinding surface 23, but it is also possible to use a single image showing a part (for example, half) of the grinding surface 23, or to use multiple images obtained by dividing the image showing the entire grinding surface 23 into multiple parts.

[0080] The process of storing the learned learners 51A and 51B in the electronic control unit 40 can be carried out in any manner as shown in (Process A) to (Process C) below. (Task A) Using the automatic grinding device 20 equipped with the grinding wheel 21 to be judged, training data 53A and 53B are acquired, and the learners 51A and 51B are trained using the training data 53A and 53B.

[0081] (Task B) The manufacturer of the automatic grinding machine 20 or the manufacturer of the grinding wheel 21 provides the training data 53A and 53B corresponding to the grinding wheel 21 that will actually be used. Then, the training data 53A and 53B are stored in the electronic control unit 40 of the grinding wheel determination device 30 used by the user, thereby allowing the learners 51A and 51B to learn from machine learning.

[0082] (Operation C) During the manufacturing of the automatic grinding machine 20, the learned learners 51A and 51B are stored in the electronic control unit 40. The camera 31 may be mounted on a part of the automatic grinding device 20 other than the grinding wheel cover 25.

[0083] The imaging unit, which has a camera 31, may be configured separately from the automatic grinding device 20. In this configuration, when determining the state of the grinding surface 23, the imaging unit is temporarily attached to the automatic grinding device 20 and the grinding surface 23 is imaged by the imaging unit, thereby obtaining an image (determination image 50) of the grinding surface 23 of the grinding wheel 21 to be determined. According to the above configuration, when the grinding surface 23 is imaged by the imaging unit, the relative position between the grinding wheel 21 (specifically, its grinding surface 23) and the imaging unit can be controlled by using various drive units that make up the automatic grinding device 20. Examples of various drive units include a moving device for moving the grinding wheel 21 and a moving device for moving the table 202 on which the workpiece W is fixed.

[0084] - Immediately after the dressing of the grinding wheel 21 to be judged is completed, image judgment based on an image of the grinding surface 23 may be performed once (or multiple times). In this image judgment, a process should be performed to determine whether the generation error ΔP2 output from the learning device 51B is less than or equal to a predetermined value (specifically, a small value other than "0"). As a result, if the generation error ΔP2 output from the learning device 51B is less than or equal to the predetermined value, it can be determined that the generation error ΔP2 is approximately "0", and that the dressing of the grinding wheel 21 to be judged has been performed properly.

[0085] • As training data 53A and 53B, it is not limited to using image data corresponding to a grinding wheel 21 in an undeteriorated state; image data corresponding to a grinding wheel 21 in any deteriorated state can be used.

[0086] As such training data 53A and 53B, for example, image data corresponding to the grinding wheel 21 in the state most suitable for starting dressing can be used. In this configuration, the state most suitable for starting dressing corresponds to a specific deterioration state. In this case, the acquisition of training data 53A and 53B can be performed as follows: Prepare a grinding wheel 21[B] in the state most suitable for starting dressing, and attach the grinding wheel 21[B] to the automatic grinding device 20. Then, in that state, generate image data (spectrum) based on the detection value of the vibration sensor 46, and store this image data as training data 53A in the storage unit 44. Also, with the grinding wheel 21[B] attached to the automatic grinding device 20, generate image data based on the image captured by the camera 31, and store this image data as training data 53B in the storage unit 44.

[0087] In this configuration, when the grinding wheel 21 being evaluated is not deteriorated, the absolute value of the generation error output from the learning device is large. Then, as the deterioration of the grinding wheel 21 progresses, the absolute value of the generation error gradually decreases. When the grinding wheel 21 is in the state most suitable for starting dressing, the generation error becomes a very small value (for example, "0"). With the above configuration, the state of the grinding wheel 21 can be determined based on this generation error.

[0088] • The learning devices 51A and 51B are not limited to employing generative adversarial networks; they can also employ autoencoders or the like. • As learners 51A and 51B, it is also possible to employ convolutional neural networks, general feedforward neural networks with a multilayer structure, support vector machines, etc. If convolutional neural networks are employed as learners 51A and 51B, the state of the grinding wheel 21 can be determined as follows.

[0089] The learners 51A and 51B, which are composed of convolutional neural networks, are trained using training data 53A and 53B that correspond to three types of deterioration states of the grinding surface 23: "immediately after dressing," "processing period," and "dressing timing." "Immediately after dressing" is the state after optimal dressing has been performed, "processing period" is the state of the grinding surface 23 that is suitable for grinding, and after grinding by the grinding wheel 21 has been performed for a predetermined period after the completion of dressing, and "dressing timing" is the state that is most suitable for starting dressing. In this configuration, "immediately after dressing," "processing period," and "dressing timing" correspond to specific deterioration states.

[0090] The trained learners 51A and 51B, trained using the training data 53A and 53B, output the result of classifying the judgment image 50 (or judgment data 55) as input data into three classification classes (classes 0, 1, and 2). If the judgment image 50 (or judgment data 55) is image data corresponding to "immediately after dressing," the image data is classified into "class 0." If the judgment image 50 (or judgment data 55) is image data corresponding to the "processing period," the image data is classified into "class 1." Furthermore, if the judgment image 50 (or judgment data 55) is image data corresponding to the "dressing timing," the image data is classified into "class 2." Specifically, the learners 51A and 51B output the probability that a single image data (judgment image 50 or judgment data 55) belongs to each of the three classification classes. In other words, the learning devices 51A and 51B output the probability P0 of belonging to "Class 0", the probability P1 of belonging to "Class 1", and the probability P2 of belonging to "Class 2" for a single judgment image 50 (or judgment data 55). The grinding wheel judgment device 30 then determines the deterioration state of the grinding surface 23 on the grinding wheel 21 based on the probabilities P0, P1, and P2 for each classification class output from the learning devices 51A and 51B.

[0091] In the above configuration, the process for determining whether or not to execute the image judgment process in the overall judgment process (Figure 9) (see the process in step S42) can be, for example, the process for determining whether or not the following (condition A) is met.

[0092] (Condition A) The probability P1 of belonging to "Class 1" is greater than or equal to a predetermined value (e.g., 45%), and the probability P2 of belonging to "Class 2" is greater than or equal to a predetermined value (e.g., 5%). Furthermore, in the above configuration, as a process to determine whether or not it is the "dress timing" in the vibration determination process (Figure 7) (see the process in step S23), for example, a process to determine whether or not the following (condition B) is met can be executed.

[0093] (Condition B) The probability P2 of belonging to "Class 2" is greater than or equal to a predetermined value (e.g., 95%). Furthermore, in the above configuration, it is possible to determine that the dressing of the grinding wheel 21 being judged has been properly performed if the probability P0 of belonging to "Class 0" is 95% or higher.

[0094] As a process for determining the state of the grinding wheel 21, it is also possible to perform only the vibration determination (see Figure 4) and the image determination (see Figure 7) based on an image of the grinding surface 23, as opposed to vibration determination (see Figure 4) based on vibration of the automatic grinding device 20.

[0095] In this case, if the generation error ΔP1 output from the vibration learning device 51A is greater than or equal to a predetermined third determination value J3, it should be determined that it is time to perform dressing on the grinding wheel 21 to be judged (so-called dressing timing). If the generation error ΔP1 is less than the third determination value J3, it should not be determined that it is dressing timing. In the above configuration, the generation error ΔP1 when dressing timing occurs should be determined in advance based on the results of various experiments and simulations by the inventors, and a value corresponding to this generation error ΔP1 should be defined as the third determination value J3 and stored in the storage unit 44.

[0096] With the above configuration, it is possible to prepare training data 53A corresponding to the analysis data (spectrogram) when grinding is performed using a grinding wheel 21[A] that is not deteriorated. Furthermore, it is possible to prepare a pre-trained vibration learner 51A that has been trained using this training data 53A. Therefore, by using the pre-trained vibration learner 51A to capture the change in the vibration pattern of the table 202 due to the deterioration of the grinding wheel 21, the deterioration state of the grinding wheel 21 can be determined with high accuracy.

[0097] • As a method for performing frequency analysis on time-series data, it is not limited to using a Short-Time Fourier Transform (STF) with a Hanning window as the window function. Any frequency analysis method can be used, such as using a Short-Time Fourier Transform with a Hann window as the window function. Such methods are not limited to Short-Time Fourier Transforms; any frequency analysis method that can produce some kind of image data, such as a spectrogram, can be used.

[0098] The grinding wheel determination device according to the above embodiment can also be applied to grinding wheels having a structure in which a large number of abrasive grains are fixed to the bottom surface of a cylindrical base material with a binder. (Note) The above embodiment includes the configuration described in the following appendix.

[0099] [Note 1] A grinding wheel determination device applied to a processing machine that performs grinding using a grinding wheel, for determining the condition of the grinding wheel, comprising: a vibration detection unit provided on the processing machine that detects data relating to the vibration of the processing machine; a vibration analysis unit that performs frequency analysis of the detection data detected by the vibration detection unit; a vibration input unit that stores the analysis data obtained as a result of the frequency analysis by the vibration analysis unit as determination data; a vibration storage unit that stores a vibration learner that has been trained using the analysis data in the case of grinding using a grinding wheel in a specific deteriorated state as training data; and a vibration output unit that takes the determination data as input data and outputs a value corresponding to the degree of deterioration of the grinding wheel from the trained vibration learner stored in the vibration storage unit.

[0100] [Note 2] The analysis data is data obtained by performing a short-time Fourier transform on time-series data consisting of data relating to the vibration of the processing machine detected by the vibration detection unit over a predetermined period of time, as described in [Note 1].

[0101] [Note 3] The grinding wheel determination device according to [Note 1] or [Note 2], comprising: an image input unit that stores an image of the grinding surface of a grinding wheel to be determined as a determination image; an image storage unit that stores a trained image learner that has been trained using training data including an image of the grinding surface of a grinding wheel in a specific state of deterioration; an image output unit that takes the determination image as input data and outputs a value corresponding to the degree of deterioration of the grinding wheel from the trained image learner; and an execution unit that does not perform the storage of the determination image by the image input unit and the output of a value corresponding to the degree of deterioration of the grinding wheel by the image output unit if the output value of the vibration output unit is a value indicating that the degree of deterioration of the grinding wheel is less than a predetermined level, and performs the storage of the determination image by the image input unit and the output of a value corresponding to the degree of deterioration of the grinding wheel by the image output unit if the output value of the vibration output unit is a value indicating that the degree of deterioration of the grinding wheel is at or above the predetermined level.

[0102] [Note 4] The grinding wheel determination device according to [Note 3], comprising: an imaging unit for imaging the grinding surface of the grinding wheel; a shutter unit extending along the grinding surface in a manner that covers the grinding surface of the grinding wheel; and a shutter operation unit for opening and closing the shutter unit, wherein when imaging the grinding surface, the space between the grinding surface and the imaging unit is not separated by the shutter unit, and when the grinding surface is not imaging, the space between the grinding surface and the imaging unit is separated by the shutter unit, thus creating a closed state.

[0103] [Note 5] The learned vibration learning device provided in the grinding wheel determination device described in any one of the above items [Note 1] to [Note 4]. [Appendix 6] A determination program that causes the electronic control device of the grinding wheel determination device to execute the processing of the vibration detection unit, the vibration analysis unit, the vibration input unit, the vibration memory unit, and the vibration output unit, which are all included in the grinding wheel determination device described in any one of the above [Appendix 1] to [Appendix 4].

[0104] [Note 7] A grinding wheel determination method applicable to a machine that performs grinding using a grinding wheel, for determining the condition of the grinding wheel, comprising: a detection step of detecting data relating to the vibration of the machine using a vibration detection unit provided in the machine; an analysis step of performing frequency analysis on the detection data detected by the vibration detection unit; an input step of storing the analysis data obtained as a result of the frequency analysis as determination data in the vibration memory unit of the grinding wheel determination device; a storage step of storing in the vibration memory unit a vibration learner that has been trained using the analysis data in the case of grinding using a grinding wheel in a specific deteriorated state as training data; and an output step of outputting a value corresponding to the degree of deterioration of the grinding wheel from the trained vibration learner stored in the vibration memory unit, using the determination data as input data. [Explanation of symbols]

[0105] 20 Automatic grinding machine 201 Support stand 202 Table 203 Feeding device 204 nozzles 21 Sharpening stones 22 Rotary drive unit 23 Grinding surface 25 Grinding wheel cover 26 Upper wall 27 Opening 28 Shutter section 29 Shutter operation section 30. Grinding wheel determination device 31 Camera 32 X-axis drive unit 33 Y-axis drive unit 34 Z-axis drive unit 35 Display device 40 Electronic control unit 41 CPU 42 ROM 43 RAM 44 Memory section 45 Position Sensor 46. ​​Vibration Sensor 50 Judgment Images 51A, 51B Learning Device 52 Executable Program 53A, 53B Training Data 54. Judgment Result Data 55 Data for Judgment

Claims

1. In a grinding wheel determination device applied to a processing machine that performs grinding using a grinding wheel, the condition of the grinding wheel is determined, The processing machine is provided with a vibration detection unit that detects data related to the vibration of the processing machine, A vibration analysis unit performs frequency analysis of the detection data detected by the vibration detection unit, A vibration input unit stores the analysis data obtained as a result of frequency analysis by the vibration analysis unit as data for determination, A vibration memory unit stores a vibration learning device that has been trained using the analysis data obtained when grinding is performed using a grinding wheel in a specific deteriorated state as training data, A vibration output unit that takes the aforementioned determination data as input data and outputs a value corresponding to the degree of deterioration of the grinding wheel from the learned vibration learner stored in the vibration memory unit, An image input unit that stores an image of the grinding surface of the grinding wheel to be judged as the judgment image, An image storage unit that stores a trained image learner that has been trained using training data including images of the abrasive surface of a grinding wheel in a specific state of deterioration, An image output unit that takes the aforementioned judgment image as input data and outputs a value corresponding to the degree of deterioration of the grinding wheel from the previously trained image learning device, An execution unit which stores the determination image by the image input unit and outputs a value corresponding to the degree of deterioration of the grinding wheel by the image output unit, does not perform this operation if the output value of the vibration output unit is a value indicating that the degree of deterioration of the grinding wheel is below a predetermined level, and performs this operation if the output value of the vibration output unit is a value indicating that the degree of deterioration of the grinding wheel is at or above the predetermined level, A grinding wheel determination device equipped with the following features.

2. The aforementioned analysis data is obtained by performing a short-time Fourier transform on time-series data consisting of data relating to the vibration of the processing machine detected by the vibration detection unit over a predetermined period of time. The grinding wheel determination device according to claim 1.

3. An imaging unit for imaging the abrasive surface of the grinding wheel, A shutter portion extends along the grinding surface in such a manner that it covers the grinding surface of the grinding wheel, The shutter unit is operated to open and close the shutter unit, and when the grinding surface is being imaged, the shutter unit opens so that the space between the grinding surface and the imaging unit is not separated by the shutter unit, while when the grinding surface is not being imaged, the shutter unit closes so that the space between the grinding surface and the imaging unit is separated by the shutter unit. The grinding wheel determination device according to claim 1 or 2.

4. A determination program that causes an electronic control device provided in the grinding wheel determination device to execute the processing of the vibration detection unit, the vibration analysis unit, the vibration input unit, the vibration storage unit, the vibration output unit, the image input unit, the image storage unit, the image output unit, and the execution unit, all of which are provided in the grinding wheel determination device according to claim 1.

5. A grinding wheel determination method applied to a machine that performs grinding using a grinding wheel, for determining the condition of the grinding wheel, A detection step in which a vibration detection unit provided in the processing machine detects data related to the vibration of the processing machine, An analysis step is performed on the detection data detected by the vibration detection unit, and The input step involves storing the analysis data obtained as a result of the frequency analysis as judgment data in the vibration memory unit of the grinding wheel judgment device, A storage step involves storing in the vibration memory unit a vibration learning device that has been trained using the analysis data obtained when grinding is performed using a grinding wheel in a specific deteriorated state as training data. An output step in which the aforementioned determination data is used as input data, and a value corresponding to the degree of deterioration of the grinding wheel is output from the learned vibration learner stored in the vibration memory unit, An image input step involves storing an image of the grinding surface of the grinding wheel to be judged as a judgment image in the image input unit of the grinding wheel judgment device, Image storage step: Store in the image storage unit of the grinding wheel determination device a trained image learner, which has been trained using training data including images of the grinding surface of a grinding wheel in a specific state of deterioration; Image output step: Using the judgment image as input data, output a value corresponding to the degree of deterioration of the grinding wheel from the learned image learner stored in the image storage unit. An execution step which is performed such that the image input step and the image output step are not performed if the output value of the output step is a value indicating that the degree of deterioration of the grinding wheel is below a predetermined level, and are performed if the output value of the output step is a value indicating that the degree of deterioration of the grinding wheel is at or above the predetermined level, A grinding wheel determination method comprising the following:

Images