JP2023024706A - Grinding quality estimation model generator, grinding quality estimation device, defective quality occurrence factor presumption device, and operation command data update device of grinder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a grinding quality estimation model generator which can acquire the grinding quality of a workpiece.

SOLUTION: A grinding quality estimation model generator includes: an actual measurement data acquisition part 120 which acquires actual measurement data, which is created when a workpiece W is ground by an abrasive wheel 16 in a grinder 1 and is at least one of first actual measurement data indicating states of structural members 12, 13, 14, 15 of the grinder 1 and second actual measurement data regarding a ground portion, for each workpiece W for a predetermined time; and a first learning model generation part 150 which generates a first learning model for estimating the grinding quality of the workpiece W by machine learning using the actual data regarding the multiple workpieces W as input data for first learning.

SELECTED DRAWING: Figure 3

COPYRIGHT: (C)2023,JPO&INPIT

Description

The present invention relates to a grinding quality estimation model generation device, a grinding quality estimation device, a defective quality occurrence factor estimation device, an operation command data adjustment model generation device for a grinding machine, and an operation command data update device.

2. Description of the Related Art When grinding a workpiece with a grinding wheel on a grinder, it is required that the grinding quality of the workpiece satisfies a predetermined condition. For example, it is required to prevent the formation of a process-affected layer on the workpiece, to ensure that the surface properties (for example, surface roughness) of the workpiece are within a predetermined value, and to prevent the occurrence of chatter patterns on the workpiece. be done.

By inspecting the workpiece after grinding, an operator confirms whether or not the grinding quality satisfies a predetermined condition, and if the predetermined condition is satisfied, the workpiece is regarded as a non-defective product. Further, Patent Document 1 describes determining whether or not a work-affected layer is generated in a workpiece based on the grinding load measured during grinding.

By the way, in recent years, along with the improvement of computer processing speed, artificial intelligence has developed rapidly. For example, Patent Document 2 describes generation of laser processing condition data by machine learning.

JP 2013-129028 A JP 2017-164801 A

However, as described in Patent Document 1, the presence or absence of a work-affected layer cannot be determined with high accuracy only by using the grinding load. This is because there are various factors that cause a work-affected layer. Therefore, it is required to obtain the grinding quality of the workpiece, such as the presence or absence of a work-affected layer, in consideration of various factors. In addition, it is required to obtain grinding conditions such that the grinding quality of the workpiece is good.

An object of the present invention is to provide a grinding quality estimation model generation device and a grinding quality estimation device that can acquire the grinding quality of a workpiece. Another object of the present invention is to provide a defective quality generation factor estimating device capable of estimating the factors that cause defective quality in a workpiece that has been determined to be defective. In addition, the present invention provides an operation command data adjustment model generation device for a grinder and an operation of the grinder to obtain operation command data for the grinder capable of improving the grinding quality by using the grinding quality of the workpiece. An object of the present invention is to provide a command data update device.

(1. Grinding quality estimation model generation device)
The grinding quality estimation model generation device is actually measured data when grinding a workpiece with a grinding wheel in a grinder, which is first measured data representing the state of a structural member of the grinder, and a grinding part. an actual measurement data acquisition unit that acquires the actual measurement data, which is at least one of the second actual measurement data, for each of the workpieces for a predetermined period; a first learning model generation unit that generates a first learning model for estimating the grinding quality of the workpiece by machine learning;

The first learning model is generated by machine learning using measured data as first learning input data. The measured data is at least one of first measured data representing the state of the structural member of the grinder and second measured data relating to the grinding portion. The actual measurement data is data acquired for a predetermined period for each workpiece. For example, the actual measurement data includes data from the initial stage of grinding to the final stage of grinding, data from the initial stage of rough grinding to the final stage of rough grinding, and the like. Therefore, even the actual measurement data of one workpiece is a large amount of data. Furthermore, when it comes to actual measurement data for a plurality of workpieces, the amount of data is extremely large. However, by using machine learning, the first learning model can be easily generated even with a large amount of measured data on multiple workpieces.

Therefore, the grinding quality of the workpiece can be obtained as a result by generating the first learning model considering a large amount of actual measurement data that affects the grinding quality of the workpiece. The first measured data representing the state of the structural member of the grinder is, for example, the vibration of the structural member, the amount of deformation of the structural member, and the like. The second measured data on the grinding portion is, for example, the dimensions of the workpiece that change due to grinding, the temperature of the grinding point, and the like.

(2. Grinding quality estimation device)
The grinding quality estimation device uses the above-described grinding quality estimation model generation device, the estimation input data that is the actual measurement data for a predetermined period when grinding a new workpiece, and the first learning model. and a grinding quality estimator for estimating the grinding quality of the new workpiece. By using the first learning model generated by machine learning, it is possible to estimate the grinding quality of the new workpiece based on the input data for estimation, which is a large amount of measured data regarding the new workpiece.

(3. Defective Quality Occurrence Factor Estimating Device)
The defective quality occurrence factor estimating device is the grinding quality estimating device described above, and includes a determination unit that determines the quality of the workpiece based on the grinding quality of the workpiece estimated by the grinding quality estimating unit. Machining of non-defective products created based on the actual operation data of the device and the driving device controlled by the control device of the grinding machine, which is obtained in advance and is related to non-defective products, or the actual measurement data related to non-defective products acquired in advance. Comparing a non-defective product processing data storage unit that stores data, the non-defective product processing data, and the defective product processing data that is the actual operation data or actual measurement data regarding the workpiece that is estimated to be a defective product by the discrimination unit, and a difference information extraction unit for extracting processing data difference information for specifying the cause of defective quality.

The defective quality occurrence factor estimating device can use the processing data difference information extracted by the difference information extracting unit to estimate the factors that have caused the defective quality with respect to the workpiece that is estimated to be defective by the determining unit. .

(4. Grinding Machine Operation Command Data Adjustment Model Generating Device)
The operation command data adjustment model generation device for a grinding machine includes an operation command data acquisition unit for acquiring operation command data for a control device of the grinding machine when grinding a workpiece with a grinding wheel in the grinding machine; a grinding quality data acquisition unit that acquires grinding quality data of a workpiece for each workpiece; a reward determination unit that determines a reward for the operation command data according to the grinding quality data for each workpiece; a second learning model generation unit for generating a second learning model for adjusting the motion command data so as to increase the reward by machine learning using the motion command data and the reward related to the workpiece. .

The generator uses machine learning to generate a second learning model for adjusting the operation command data of the grinding machine. Machine learning uses motion command data and rewards for a plurality of workpieces. Therefore, even if a large amount of data is used, the second learning model can be easily generated by applying machine learning. Additionally, in machine learning, grinder motion command data is adjusted to increase rewards determined using workpiece grinding quality data. Therefore, it is possible to generate operation command data that will improve the grinding quality.

(5. Grinding Machine Operation Command Data Update Device)
The operation command data update device for the grinder includes the above-described operation command data adjustment model generation device for the grinder, the operation command data regarding the grinding of the new workpiece, the grinding quality data of the new workpiece, the reward, and an action command data adjustment unit that adjusts the action command data using the second learning model. That is, the motion command data is updated using the second learning model generated by machine learning. Therefore, even if the grinding state changes, the operation command data is updated according to the current grinding state. By updating the operation command data in this way, it is possible to improve the grinding quality of the workpiece.

It is a top view which shows a grinder. 1 is a functional block diagram showing a schematic configuration of a machine learning device according to a first embodiment; FIG. 3 is a functional block diagram showing the detailed configuration of the learning phase of the machine learning device of the first embodiment; FIG. It is a functional block diagram which shows the detailed structure of the estimation phase of the machine learning apparatus of 1st embodiment. It is a functional block diagram which shows a schematic structure of the machine learning apparatus of 2nd embodiment. FIG. 11 is a functional block diagram showing the detailed configuration of the learning phase of the machine learning device of the second embodiment; It is a functional block diagram which shows the detailed structure of the estimation phase of the machine-learning apparatus of 2nd embodiment. It is a functional block diagram which shows a schematic structure of the machine-learning apparatus of 3rd embodiment. FIG. 13 is a functional block diagram showing the detailed configuration of the learning phase of the machine learning device of the third embodiment; FIG. 12 is a functional block diagram showing the detailed configuration of the estimation phase of the machine learning device of the third embodiment; It is a functional block diagram which shows the schematic structure of the machine-learning apparatus of 4th embodiment. FIG. 14 is a functional block diagram showing the detailed configuration of the learning phase of the machine learning device of the fourth embodiment; FIG. 12 is a functional block diagram showing the detailed configuration of the estimation phase of the machine learning device of the fourth embodiment;

(1. First embodiment)
(1-1. Configuration of grinding machine)
The configuration of the grinder 1 will be described with reference to FIG. The grinder 1 is a machine for grinding a workpiece W. As the grinder 1, grinders having various configurations such as a cylindrical grinder and a cam grinder can be applied. In the present embodiment, the grinding machine 1 is an example of a wheelhead traverse type cylindrical grinding machine. However, the grinding machine 1 can also apply a table traverse type.

The grinding machine 1 mainly includes a bed 11, a headstock 12, a tailstock 13, a traverse base 14, a grinding wheel head 15, a grinding wheel 16, a sizing device 17, a grinding wheel correction device 18, a coolant device 19, and a control device. 20.

The bed 11 is fixed on the installation surface. The headstock 12 is provided on the upper surface of the bed 11 on the front side in the X-axis direction (lower side in FIG. 1) and one end side in the Z-axis direction (left side in FIG. 1). The headstock 12 supports the workpiece W so as to be rotatable around the Z axis. The workpiece W is rotated by driving a motor 12 a provided on the headstock 12 . The tailstock 13 is located on the upper surface of the bed 11 at a position facing the headstock 12 in the Z-axis direction, i.e., on the front side in the X-axis direction (lower side in FIG. 1). That is, the headstock 12 and the tailstock 13 rotatably support the workpiece W at both ends.

The traverse base 14 is provided on the upper surface of the bed 11 so as to be movable in the Z-axis direction. The traverse base 14 is moved by driving a motor 14 a provided on the bed 11 . The wheelhead 15 is provided on the upper surface of the traverse base 14 so as to be movable in the X-axis direction. The wheel head 15 is moved by driving a motor 15 a provided on the traverse base 14 . The grinding wheel 16 is rotatably supported on the grinding wheel head 15 . The grinding wheel 16 is rotated by driving a motor 16 a provided on the grinding wheel head 15 . The grinding wheel 16 is configured by fixing a plurality of abrasive grains with a bond material.

The sizing device 17 measures the dimension (diameter) of the workpiece W. As shown in FIG. The grinding wheel modifying device 18 modifies the shape of the grinding wheel 16 . The grinding wheel correction device 18 is a device for truing the grinding wheel 16 . The grinding wheel correction device 18 may be a device for dressing the grinding wheel 16 in addition to or instead of truing. Further, the grinding wheel correction device 18 also has the function of measuring the dimension (diameter) of the grinding wheel 16 .

Here, the truing is a reshaping work, such as a work of shaping the grinding wheel 16 in accordance with the shape of the workpiece W when the grinding wheel 16 is worn by grinding, a work of removing runout of the grinding wheel 16 due to uneven wear, and the like. is. Dressing is a dressing (sharpening) work, and is a work of adjusting the protrusion amount of abrasive grains and creating a cutting edge of abrasive grains. Dressing is an operation for correcting blindness, clogging, spillage, etc., and is usually performed after truing.

The coolant device 19 supplies coolant to the grinding point of the workpiece W by the grinding wheel 16 . The coolant device 19 cools the collected coolant to a predetermined temperature and supplies it to the grinding point again.

The control device 20 controls each driving device based on an NC program generated based on operation command data such as the shape of the workpiece W, machining conditions, the shape of the grinding wheel 16, coolant supply timing information, and the like. That is, the control device 20 receives operation command data, generates an NC program based on the operation command data, and controls the motors 12a, 14a, 15a, 16a, the coolant device 19, etc. based on the NC program. The workpiece W is ground. In particular, the control device 20 grinds the workpiece W based on the diameter of the workpiece W measured by the sizing device 17 until the workpiece W has a finished shape. Further, the control device 20 corrects (truing and dressing) the grinding wheel 16 by controlling the motors 14a, 15a, 16a, the grinding wheel correction device 18, etc. at the timing of correcting the grinding wheel 16. .

1, the grinder 1 includes various sensors 21, 22, 23 (shown in FIG. 3, etc.), which will be described later. For example, the grinder 1 includes actual operation data of each motor and the like, data indicating the state of structural members constituting the grinder 1, a sizing device 17, a grindstone diameter sensor, a temperature sensor, and the like. Details of various sensors and the like will be described later.

(1-2. Overview of machine learning device 100)
Next, an overview of the machine learning device 100 of the first embodiment will be described with reference to FIG. The machine learning device 100 (a) generates a first learning model for estimating the grinding quality of the workpiece W, and (b) estimates the grinding quality of the workpiece W using the first learning model. The machine learning device 100 can be configured as a device separate from the grinding machine 1, or can be configured as a device incorporated in the control device 20 of the grinding machine 1 or the like. In this embodiment, the machine learning device 100 is connected to the grinder 1 via a network line, and transmits and receives various data.

The machine learning device 100 has elements 101a, 101b, 101c that function in a first learning phase 101 that generates a first learning model, and an estimation phase 102 that estimates the grinding quality (also commonly referred to as an "inference phase"). It comprises elements 102a, 102b. The machine learning device 100 has, as elements that function in the first learning phase 101, an element 101a that acquires first learning input data, an element 101b that acquires first teacher data, and an element that generates a first learning model. 101c.

The first learning input data acquired by the element 101a is input data used for machine learning, and includes, for example, motion command data, actual motion data, first measured data (data representing the state of structural members), first 2) actual measurement data (data relating to the ground portion), and the like.

The first teacher data acquired in element 101b is teacher data used for machine learning in supervised learning. The first teaching data is grinding quality data of the workpiece W, such as work-affected layer data of the workpiece W, surface texture data of the workpiece W, chatter pattern data of the workpiece W, and the like.

The first learning model generated in the element 101c is a model for estimating the grinding quality of the workpiece W by performing supervised learning among machine learning based on the first learning input data and the first teacher data. (function). However, the first learned model can also be generated by applying unsupervised learning for the purpose of classifying the grinding quality. However, when supervised learning is applied, grinding quality can be obtained with high accuracy.

The machine learning apparatus 100 includes, as elements that function in the estimation phase 102, an element 102a that acquires input data for estimation, and an element 102b that estimates the grinding quality and determines whether the workpiece W is good or bad. The estimation input data acquired by the element 102a is the same type of data as the first learning input data, and is data acquired for a workpiece W different from the workpiece W used for learning (a new workpiece W). is. The element 102b estimates the grinding quality using the input data for estimation and the first learning model, and determines the quality of the workpiece W based on the estimated grinding quality. The first learning model used in element 102 b is the first learning model generated by machine learning in the first learning phase 101 .

(1-3. Configuration of grinding machine 1 related to machine learning device 100)
A configuration of the grinder 1 related to the machine learning device 100 will be described with reference to FIG. As shown in FIG. 3 , the grinding machine 1 has a control device 20 . The control device 20 is a so-called CNC (Computerized Numerical Control) device. As described above, the control device 20 generates an NC program based on the operation command data, and based on the NC program, each drive device 12a, 14a, 15a, 16a, 17, 18 ("12a etc." in FIG. 3). ”).

Structural members 12, 13, 14, and 15 (indicated by "15, etc." in FIG. 3) are operated by driving the driving devices 12a, 14a, 15a, 16a, 17, and 18, respectively. The movement of the structural members 12 , 13 , 14 , 15 causes the workpiece W to be ground by the grinding wheel 16 . In FIG. 3, the portion of the workpiece W to be ground by the grinding wheel 16 is shown as the grinding portion.

The grinder 1 further includes a sensor 21 for detecting actual operation data of the driving device 12a and the like, a sensor 22 for detecting the state of the structural member 15 and the like (data representing the state of the structural member), and a grinding portion W that changes due to grinding. A sensor 23 for detecting data (grinding part data) is provided. The sensor 21 is, for example, a current sensor that detects the driving current of the motor 12a, a position sensor that detects the current position (rotational angle) of the motor 12a, or the like. The sensor 21 detects similar information for the other drives 14a, 15a, 16a, 17, 18 as well. The sensor 22 is a vibration sensor that detects vibration of the structural member 15 or the like, a strain sensor that detects the amount of deformation of the structural member 15 or the like, or the like. As the vibration sensor, a sensor that detects acceleration corresponding to vibration, a sensor that detects sound waves corresponding to vibration, or the like can be applied. The sensor 23 is a sizing device that detects the dimension (diameter) of the workpiece W that changes due to grinding, and a temperature sensor that detects the grinding point temperature during grinding.

(1-4. Configuration of external device 2 related to machine learning device 100)
A configuration of the external device 2 related to the machine learning device 100 will be described with reference to FIG. The external device 2 detects the grinding quality data of each workpiece W ground by the grinding wheel 16 in the grinding machine 1 . The grinding quality data includes, for example, work-affected layer data (grinding burn data, etc.), surface texture data (surface roughness data, etc.), chatter pattern data, and the like.

That is, the external device 2 includes a work-affected layer detector that acquires work-affected layer data (data relating to grinding burns, softened layer due to grinding, etc.), and a surface texture measuring instrument that acquires surface texture data (data relating to surface roughness, etc.). , a chatter detector for acquiring chatter pattern data, and the like. Note that the external device 2 may be a device that directly acquires the data. Further, the external device 2 may be a device that obtains target data by obtaining another data having correlation and performing calculation using the another data, that is, a device that indirectly obtains the target data.

The work-affected layer data may be data relating to the presence or absence of a work-affected layer, or may be a score relating to the extent of the work-affected layer. The surface texture data may be, for example, the surface roughness value itself, or may be a score relating to the degree of surface roughness. Further, the chatter pattern data may be data regarding the presence or absence of the chatter pattern, or may be a score regarding the degree of the chatter pattern. Each score is represented by, for example, a multi-level score.

(1-5. Detailed configuration of first learning phase 101 of machine learning device 100)
A detailed configuration of the first learning phase 101 of the machine learning device 100 will be described with reference to FIG. Here, the configuration of the first learning phase 101 corresponds to the grinding quality estimation model generation device.

The configuration of the first learning phase 101 includes a first input data acquisition unit 130 that acquires first input data, a grinding quality data acquisition unit 140 that acquires grinding quality data, a first learning model generation unit 150, and a first A learning model storage unit 160 is provided.

The first input data acquisition unit 130 acquires first input data regarding a plurality of workpieces W as first learning input data for machine learning. In addition, the grinding quality data acquisition unit 140 acquires grinding quality data regarding the plurality of workpieces W as first teacher data for machine learning. Here, the first learning input data and the first teacher data are as shown in Table 1. As shown in Table 1, the first learning input data are a large number of data, but it is not necessary to use all the data shown in Table 1, and only part of the data may be used.

The first input data acquisition section 130 includes a motion-related data acquisition section 110 and an actual measurement data acquisition section 120 . The motion-related data acquisition unit 110 acquires the motion command data for the control device 20, and the actual motion data for the drive device 12a controlled by the control device 20 from the sensor 21. An acquisition unit 112 is provided.

As shown in Table 1, the operation command data in the operation-related data includes the command cutting speed for each process, the command positions of the moving bodies 14 and 15 at the time of process switching, the command rotation speed of the grinding wheel 16, the workpiece W command rotation speed, coolant supply information, and the like. Here, the workpiece W is ground by a plurality of grinding processes such as coarse grinding, fine grinding, fine grinding, and spark out. As shown in Table 1, the actual operation data in the operation-related data includes the driving current of the motor 12a and the like, the actual position of the motor 12a and the like. The actual operation data acquisition unit 112 acquires actual operation data for each workpiece W for a predetermined period. The predetermined period is, for example, from the beginning of grinding to the end of grinding, from the beginning of rough grinding to the end of rough grinding, and the like. It should be noted that when grinding is in an unsteady state, it is unstable, so it is possible to obtain data only for the steady state.

The measured data acquisition unit 120 includes a first measured data acquisition unit 121 that acquires first measured data from the sensor 22 and a second measured data acquisition unit 122 that acquires second measured data from the sensor 23 . The first actual measurement data is actual measurement data when the workpiece W is ground by the grinding wheel 16, and includes vibration of the structural member 15 and the like, deformation amount of the structural member 15 and the like. The second actual measurement data is actual measurement data when the workpiece W is ground by the grinding wheel 16, and includes the dimensions (diameter) of the workpiece W, the grinding point temperature, and the like.

The first measured data acquisition unit 121 acquires the first measured data for each workpiece W for a predetermined period. The second measured data acquisition unit 122 also acquires the second measured data for each workpiece W for a predetermined period. The first measured data and the second measured data are acquired for the same predetermined period as the actual operation data described above. As described above, the predetermined period is, for example, from the beginning of grinding to the end of grinding, or from the beginning of rough grinding to the end of rough grinding.

The grinding quality data acquisition unit 140 acquires the grinding quality data regarding the plurality of workpieces W acquired by the external device 2 as first teacher data for supervised learning. That is, the grinding quality data acquisition unit 140, for example, acquires process-affected layer data (data related to grinding burn, softened layer due to grinding, etc.), surface texture data (data such as surface roughness), chatter pattern data, etc., as the first teacher. Get it as data.

The first learning model generation unit 150 performs supervised learning to generate a first learning model. Specifically, the first learning model generation unit 150 uses the first input data regarding the plurality of workpieces W acquired by the first input data acquisition unit 130 as the first learning input data, and the grinding quality data acquisition unit 140 A first learning model for estimating the grinding quality of the workpiece W is generated by machine learning using the acquired grinding quality data on the plurality of workpieces W as the first training data.

That is, the first learning model generation unit 150 uses the motion command data, the actual motion data, the first actual measurement data, and the second actual measurement data as the first input data for learning, and the grinding quality data as the first teaching data. Learning generates a first learning model. The first learning model is a model that indicates the relationship between the first learning input data and the first teacher data.

Here, among the first learning input data, at least the actual operation data, the first actual measurement data, and the second actual measurement data are data for each workpiece W for a predetermined period. Therefore, even the first learning input data for one workpiece W is a large amount of data. Furthermore, if it becomes the first learning input data for a plurality of workpieces W, it will be an extremely large amount of data. However, by using machine learning, the first learning model can be easily generated even with a large amount of first learning input data regarding a plurality of workpieces W. FIG. Therefore, as will be described later, the grinding quality of the workpiece W is obtained by generating the first learning model in consideration of a large amount of first learning input data that affects the grinding quality of the workpiece W. be able to.

The first learning model is a model for estimating the state of the work-affected layer of the workpiece W, the surface texture of the workpiece W, and the chatter pattern state of the workpiece W as the grinding quality of the workpiece W. However, the first learning model is not limited to estimating the grinding qualities of all of them, and may estimate the grinding qualities of only some of them. The first learning model generated by the first learning model generation unit 150 is stored in the first learning model storage unit 160 .

Moreover, when the predetermined period to be acquired is from the initial stage of grinding to the final stage of grinding, the first learning model is a model that considers all grinding processes. On the other hand, if the predetermined period to be acquired is, for example, from the beginning of rough polishing to the end of rough polishing, the first learning model is a learning model that considers only the rough polishing process. If it is desired to specify the process that affects the grinding quality, the first learning model may be obtained for each process.

(1-6. Detailed configuration of estimation phase 102 of machine learning device 100)
A detailed configuration of the estimation phase 102 of the machine learning device 100 will be described with reference to FIG. Here, the configuration of the first learning phase 101 and the configuration of the estimation phase 102 correspond to the grinding quality estimation device. The configuration of the first learning phase 101 is as described above.

The configuration of the estimation phase 102 includes a first input data acquisition unit 130 that acquires first input data, a first learning model storage unit 160, a grinding quality estimation unit 170, and a discrimination unit 180. The first input data acquisition unit 130 acquires the first input data for the grinding of the new workpiece W for a predetermined period of time, and the content is substantially the same as that described in the first learning phase 101 . The predetermined period here is the same as the predetermined period in the first learning phase 101 . The first learning model storage unit 160 stores the first learning model generated by the first learning model generation unit 150 as described in the first learning phase 101 .

The grinding quality estimating unit 170 uses the first input data for a predetermined period during the grinding of the new workpiece W as input data for estimation, and uses the first learning model stored in the first learning model storage unit 160 as input data for estimation. Using the model, the grinding quality of the new workpiece W is estimated. Here, the first learning model is a model that indicates the relationship between the first learning input data and the first teacher data, as described above. The first learning model is a model relating to the condition of the work-affected layer of the workpiece W, the surface texture of the workpiece W, and the chatter pattern condition of the workpiece W as the grinding quality data, which is the first teaching data.

Therefore, the grinding quality estimator 170 estimates the state of the work-affected layer of the workpiece W, the surface texture of the workpiece W, and the chatter pattern state of the workpiece W as the grinding quality. However, the grinding quality estimator 170 may estimate only part of the grinding quality instead of estimating the entire grinding quality. For example, the grinding quality estimator 170 may estimate only the state of the work-affected layer. In this case, the first learning model is generated as a model that estimates only the state of the work-affected layer.

The grinding quality estimator 170 estimates a plurality of targets as described above. By using the first learning model generated by applying machine learning, the grinding quality estimator 170 can easily estimate a plurality of targets. In this way, machine learning device 100 can estimate a complicated target at once.

The determining unit 180 determines whether the workpiece W is good or bad based on the grinding quality of the workpiece W estimated by the grinding quality estimating unit 170 . For example, when it is determined that a work-affected layer exists in the workpiece W (does not satisfy a predetermined condition) based on the estimated state of the work-affected layer, the determination unit 180 determines that the work-affected layer W is determined to be defective. Further, when it is determined that the estimated surface texture does not satisfy the predetermined condition, the determination unit 180 determines that the workpiece W is defective. Further, when it is determined that a chatter pattern exists (does not satisfy a predetermined condition) based on the estimated chatter pattern state, the determining unit 180 determines that the workpiece W is defective. .

On the other hand, the discrimination unit 180 discriminates the workpiece W as a non-defective product when each condition regarding the state of the work-affected layer, the surface texture, and the chatter pattern state is satisfied. By using the first learning model generated by applying machine learning in this way, it is possible to easily determine a plurality of conditions.

(2. Second embodiment)
(2-1. Overview of machine learning device 200)
An overview of the machine learning device 200 of the second embodiment will be described with reference to FIG. Like the machine learning device 100 of the first embodiment, the machine learning device 200 (a) generates a first learning model for estimating the grinding quality of the workpiece W, and (b) generates the first learning model. is used to estimate the grinding quality of the workpiece W. Furthermore, the machine learning device 200 (c) generates a second learning model for adjusting the operation command data of the grinder 1 so that the grinding quality is good, and (d) uses the second learning model The operation command data of the grinder 1 is updated so that the grinding quality is improved.

The machine learning device 200 comprises elements 101a, 101b, 101c functioning in a first learning phase 101 for generating a first learning model, and elements 102a, 102b functioning in an estimation phase 102 for estimating grinding quality. The first learning phase 101 and the estimation phase 102 have the same configuration as the corresponding phases of the first embodiment.

In addition, the machine learning device 200 includes, as elements functioning in the second learning phase 203 that generates the second learning model, an element 203a that acquires the second learning input data, an element 203b that acquires the first evaluation result data, and , an element 203c for generating a second learning model.

The second learning input data acquired by the element 203a is input data used for machine learning, and is, for example, operation command data. As shown in Table 1 in the first embodiment, the operation command data includes command cutting speed for each process, command positions of moving bodies 14 and 15 at the time of process switching, command rotational speed of grinding wheel 16, and workpiece W These are command rotation speed, coolant supply information, and the like. The motion command data is data for generating an NC program to be executed by the control device 20. FIG.

The first evaluation result data acquired in element 203b is evaluation result data for deriving a reward used for machine learning in reinforcement learning. The first evaluation result data is grinding quality data of the workpiece W, such as work-affected layer data of the workpiece W, surface texture data of the workpiece W, chatter pattern data of the workpiece W, and the like. The second learning model generated in the element 203c is for adjusting the operation command data of the grinding machine 1 by performing reinforcement learning among machine learning based on the second learning input data and the first evaluation result data. It is a model (function).

The machine learning device 200 includes an element 204a that acquires update input data and an element 204b that updates the action command data as elements that function in an update phase 204 that updates the action command data. The update input data acquired in the element 204a is the same type of data as the second learning input data, and is data acquired for a workpiece W different from the workpiece W used for learning (new workpiece W). is. Element 204b updates the motion command data using the update input data, the second learning model, and the estimated grinding quality. The second learning model used in element 204 b is the second learning model generated by machine learning in the second learning phase 203 . Also, the estimated grinding quality is the grinding quality estimated in the estimation phase 102 .

(2-2. Detailed configuration of first learning phase 101 of machine learning device 200)
The detailed configuration of the first learning phase 101 of the machine learning device 200 is the same as that of the first embodiment.

(2-3. Detailed configuration of second learning phase 203 of machine learning device 200)
A detailed configuration of the second learning phase 203 of the machine learning device 200 will be described with reference to FIG. Here, the configuration of the second learning phase 203 corresponds to the operation command data adjustment model generation device of the grinding machine.

The configuration of the second learning phase 203 includes an operation command data acquisition unit 111, a grinding quality data acquisition unit 140, a reward determination unit 210, a second learning model generation unit 220, and a second learning model storage unit 230. consists of

The motion command data acquisition unit 111 acquires motion command data for the control device 20 of the grinder 1 when grinding the workpiece W with the grinding wheel 16 in the grinder 1 . The motion command data acquisition unit 111 acquires motion command data regarding a plurality of workpieces W as second learning input data for machine learning. In addition, the grinding quality data acquisition unit 140 acquires grinding quality data regarding a plurality of workpieces W as first evaluation result data of machine learning. Here, the second learning input data and the first evaluation result data are as shown in Table 2. Here, as shown in Table 2, the second learning input data is a large number of data, but it is not necessary to use all the data shown in Table 2, and only a part of the data may be used. .

The remuneration determination unit 210 acquires the motion command data, which is the second learning input data, and the grinding quality data, which is the first evaluation result data, and determines a remuneration for the motion command data according to the grinding quality data. Here, the reward is a reward for a combination of action command data in reinforcement learning. Then, when the grinding quality data corresponding to the operation command data derives a desired result, a large reward is given to the operation command data, and when an undesirable result is derived, a small reward is given to the operation command data. Rewards (including negative rewards) are given.

For example, the remuneration determination unit 210 increases the remuneration when there is no work-affected layer in the work-affected layer data of the workpiece W, and decreases the remuneration when there is a work-affected layer. Further, the remuneration determination unit 210 increases the remuneration when the surface texture data of the workpiece W is equal to or less than the predetermined threshold, and decreases the remuneration when the surface texture data is greater than the predetermined threshold. Furthermore, in the chatter pattern data of the workpiece W, the reward determining unit 210 increases the reward when there is no chatter pattern, and decreases the reward when there is a chatter pattern. Note that the remuneration determining unit 210 may determine the remuneration based on all of the work-affected layer data, the surface texture data, and the chatter pattern data, or may determine the remuneration based on only one of these. good.

The second learning model generation unit 220 generates a second learning model for adjusting the action command data so as to increase the reward by machine learning. In the second learning model generation unit 220, for example, Q-learning, Sarsa, Monte Carlo method, etc. are applied as reinforcement learning.

Here, it is assumed that the operation command data before adjustment is data relating to the first workpiece W, and the operation command data after adjustment is data relating to the second workpiece W. A relationship between the operation command data relating to the first workpiece W and the grinding quality data of the first workpiece W is defined as a first data relationship. A relationship between the operation command data of the second workpiece W and the grinding quality data of the second workpiece W is defined as a second data relationship.

The second learning model is a model representing the interrelationship between the first data relationship before adjustment and the second data relationship after adjustment. The second learning model generation unit 220 makes the grinding quality data of the second workpiece W after adjustment better than the grinding quality data of the first workpiece W before adjustment, that is, the reward increases. , the method of adjusting the motion command data of the first workpiece W before adjustment to the motion command data of the second workpiece W after adjustment is learned.

However, the adjusted operation command data can be adjusted with respect to the unadjusted operation command data within the limits set in advance. For example, for the command cutting speed, which is one of the adjustable parameters, the command cutting speed after adjustment is limited to within a predetermined ratio (eg ±3%) of the command cutting speed before adjustment. there is The predetermined ratio can be set arbitrarily. The same is true for other adjustable parameters. It is also possible to set adjustable parameters. The generated second learning model is stored in the second learning model storage unit 230 .

Note that the second learning model generation unit 220 can also learn the second learning model in the update phase 204 described later. In this case, the grinding quality data obtained by the estimation phase 102 (described in the first embodiment) is used as the grinding quality data, which is the first evaluation result data.

(2-4. Detailed configuration of estimation phase 102 of machine learning device 200)
The estimation phase 102 of the machine learning device 200 is the same as the estimation phase 102 of the first embodiment.

(2-5. Detailed configuration of update phase 204 of machine learning device 200)
A detailed configuration of the update phase 204 of the machine learning device 200 will be described with reference to FIG. Here, the configuration of the second learning phase 203 and the configuration of the update phase 204 correspond to the operation command data update device for the grinding machine. The configuration of the second learning phase 203 is as described above.

The configuration of the update phase 204 includes an operation command data acquisition unit 111, a grinding quality data acquisition unit 140, a reward determination unit 210, a second learning model storage unit 230, and an operation command data adjustment unit 240. be.

The operation command data acquisition unit 111 and the grinding quality data acquisition unit 140 acquire each data for grinding the new workpiece W, and the contents are substantially the same as those described in the second learning phase 203 . The remuneration determination unit 210 determines remuneration using the motion command data and the grinding quality data acquired in grinding the new workpiece W. That is, the remuneration determination unit 210 determines a remuneration for the operation command data for grinding the new workpiece W according to the grinding quality data. The second learning model storage unit 230 stores the second learning model generated by the second learning model generation unit 220 as described in the second learning phase 203 .

The motion command data adjustment unit 240 determines a motion command data adjustment method using motion command data related to grinding of the new workpiece W, grinding quality data of the new workpiece W, remuneration, and the second learning model. Then, the operation command data is adjusted based on the determined adjustment method. Here, the second learning model is a model generated by learning an adjustment method from the action command data before adjustment to the action command data after adjustment so as to increase the reward.

Specifically, the action command data adjuster 240 acquires the current action command data as action command data before adjustment, and acquires the reward at that time. In this case, the action command data adjuster 240 uses the current action command data, the reward for the current action command data, and the second learning model to determine the action command data to be adjusted. That is, the action command data to be adjusted is action command data that provides a larger reward than the current action command data.

In the process of the action command data adjustment unit 240, there are cases where a plurality of candidates with equal rewards are output as action command data to be adjusted. In this case, for example, the action command data adjustment unit 240 can rank the plurality of candidates by setting the priority of parameters to be adjusted. For example, the order of priority of the parameters to be adjusted is such that the commanded cutting speed is first and the command rotation speed of the workpiece W is second.

Then, the action command data adjuster 240 determines the first candidate as the action command data to be adjusted, and updates the current action command data to the action command data to be adjusted. Then, the grinder 1 grinds the workpiece W based on the updated operation command data. Then, in the update phase 204 of the machine learning device 200, the operation command data for the next grinding is again adjusted based on each data in the grinding of the workpiece W. It is also possible to set the frequency of adjustment of the operation command data. For example, after grinding a set number of workpieces W, the operation command data may be adjusted.

That is, the motion command data is updated using the second learning model generated by machine learning of the machine learning device 200 . Therefore, even if the grinding state changes, the operation command data is updated according to the current grinding state. By updating the operation command data in this manner, the grinding quality of the workpiece W can be improved.

(3. Third embodiment)
(3-1. Overview of machine learning device 300)
An overview of the machine learning device 300 of the third embodiment will be described with reference to FIG. The machine learning device 300 (a) generates a first learning model for estimating the grinding quality of the workpiece W, and (b) estimates the grinding quality of the workpiece W using the first learning model. Furthermore, the machine learning device 300 (e) generates a third learning model for estimating the surface state of the grinding wheel 16, and (f) estimates the surface state of the grinding wheel 16 using the third learning model. . Furthermore, the machine learning device 300 (g) adjusts the operation command data of the grinder 1 so as to improve the grinding quality and reduce the frequency of repair or replacement of the grinding wheel 16. Generating two learning models and (h) using the second learning model to operate the grinding machine 1 for better grinding quality and less frequent modification or replacement of the grinding wheel 16. Update command data.

The machine learning device 300 comprises elements 101a, 101b, 101c, 305d, 305e that function in a first learning phase 305 to generate a first learning model and a third learning model. The machine learning device 300 includes, as elements that function in the first learning phase 305, an element 101a that acquires first learning input data, an element 101b that acquires first teacher data, an element 101c that generates a first learning model, It has an element 305d for acquiring the second training data and an element 305e for generating the third learning model. Here, the elements 101a, 101b, 101c have the same configuration as the corresponding elements of the first embodiment.

The second teacher data acquired in element 305d is teacher data used for machine learning in supervised learning. The second teaching data is data representing the surface condition of the grinding wheel 16 (surface condition data of the grinding wheel 16). The surface condition data of the grinding wheel 16 includes, for example, data on the condition of crushing, clogging, spillage, etc., and data on the condition of over-grinding.

The surface of the grinding wheel 16 affects the grinding quality of the workpiece W. In other words, the surface condition of the grinding wheel 16 is the extent to which the grinding quality of the workpiece W is affected. The surface state of the grinding wheel 16 is, for example, a state where the grain is crushed, clogged, or loose, or excessively sharpened. If the surface condition of the grinding wheel 16 is not good, the grinding quality of the workpiece W may deteriorate. Therefore, it is required to grasp the surface condition of the grinding wheel 16 .

If the surface condition of the grinding wheel 16 is broken, clogged, loose, etc., it is necessary to perform dressing, or perform dressing after shaping by truing. Also, if the surface condition of the grinding wheel 16 is too sharp, it is necessary to perform truing. Dressing is usually performed after truing. When the number of times of truing reaches a predetermined number, or when a predetermined amount of truing is formed, the grinding wheel 16 needs to be replaced.

In order to extend the life of the grinding wheel 16, it is required to reduce the number of truing and dressing operations. In addition, the time required for truing, dressing and changing the grinding wheel 16 increases the grinding cycle time. Naturally, there is a desire to shorten the grinding cycle time. From this point of view as well, it is required to grasp the surface condition of the grinding wheel 16 . Therefore, the element 305d acquires the surface condition data of the grinding wheel 16 as the second teacher data. The surface condition data of the grinding wheel 16 is data indicating the degree of influence on the grinding quality of the workpiece.

The third learning model generated in element 305e is a model for estimating the surface state of the grinding wheel 16 by performing supervised learning among machine learning based on the first learning input data and the second teacher data. (function). However, the third learning model can also be generated by applying unsupervised learning for the purpose of classifying the surface condition of the grinding wheel 16 . However, when supervised learning is applied, the surface state of the grinding wheel 16 can be obtained with high accuracy.

The machine learning device 300 comprises elements 203a, 203b, 306d, 306e that function in a second learning phase 306 to generate a second learning model. As elements that function in the second learning phase 306, the machine learning device 300 includes an element 203a that acquires second learning input data, an element 203b that acquires first evaluation result data, and an element 306d that acquires second evaluation result data. , and an element 306e for generating a second learning model. Here, the elements 203a, 203b have the same configuration as the corresponding elements of the second embodiment.

The second evaluation result data acquired in element 306d is evaluation result data for deriving a reward used for machine learning in reinforcement learning. The second evaluation result data is surface condition data of the grinding wheel 16 . The second learning model generated in the element 306e is based on the second learning input data, the first evaluation result data, and the second evaluation result data, and performs reinforcement learning of machine learning to generate an operation command for the grinding machine 1. It is a model (function) for adjusting the data.

The machine learning device 300 functions in an estimation phase 307 for estimating the grinding quality and the surface condition of the grinding wheel 16, and includes an element 102a for acquiring input data for estimation, an estimation of the grinding quality, and an evaluation of the quality of the workpiece W. It has an element 102b that makes a decision. Here, the elements 102a, 102b have the same configuration as the corresponding elements of the first embodiment.

Furthermore, the machine learning device 300, as elements functioning in the estimation phase 307, estimates the surface state of the grinding wheel 16, executes truing of the grinding wheel 16, executes dressing of the grinding wheel 16, and It has an element 307c that determines whether an exchange is to be performed. The element 307c estimates the surface state of the grinding wheel 16 using the input data for estimation and the third learning model, and performs truing of the grinding wheel 16 based on the estimated surface state of the grinding wheel 16. Determines whether the dressing of the wheel 16 is to be performed and the replacement of the grinding wheel 16 is to be performed. The third learning model used in element 307 c is the third learning model generated by machine learning in the first learning phase 305 .

The machine learning device 300 includes an element 204a that acquires update input data and an element 308c that updates the action command data as elements that function in an update phase 308 that updates the action command data. Here, element 204a has the same configuration as the corresponding element of the second embodiment. Element 308c updates the motion command data using the update input data, the second learning model, the estimated grinding quality, and the estimated grinding wheel 16 surface condition. The second learning model used in element 308 c is the second learning model generated by machine learning in the second learning phase 306 . Also, the estimated grinding quality is the grinding quality estimated in the estimation phase 307 . The estimated surface condition of the grinding wheel 16 is the surface condition of the grinding wheel 16 estimated in the estimation phase 307 .

(3-2. Detailed configuration of first learning phase 305 of machine learning device 300)
A detailed configuration of the first learning phase 305 of the machine learning device 300 will be described with reference to FIG. The configuration of the first learning phase 305 constitutes a grinding quality estimation model generation device and a grinding wheel surface state estimation model generation device. The configuration of the first learning phase 305 includes a first input data acquisition unit 130, a quality data acquisition unit 310, a first learning model generation unit 150, a third learning model generation unit 320, and a first learning model storage unit 160. , and a third learning model storage unit 330 .

The first input data acquisition unit 130 acquires first input data regarding a plurality of workpieces W as first learning input data for machine learning. The quality data acquisition unit 310 includes a grinding quality data acquisition unit 140 that acquires grinding quality data, and a grinding wheel surface condition data acquisition unit 311 that acquires surface condition data of the grinding wheel 16 . The grinding quality data acquisition unit 140 acquires grinding quality data regarding a plurality of workpieces W as first teacher data for machine learning. Further, the grinding wheel surface state data acquisition unit 311 acquires the surface state of the grinding wheel 16 after grinding for each workpiece W as second training data for machine learning. Here, the first learning input data, the first teacher data and the second teacher data are as shown in Table 3.

The first input data acquisition unit 130 and the grinding quality data acquisition unit 140 have the same configuration as the corresponding configuration in the first embodiment. The grinding wheel surface condition data acquisition unit 311 acquires the surface condition data of the grinding wheel 16 corresponding to the grinding quality data of the workpiece W acquired by the external device 2 as second teacher data for supervised learning.

The surface condition data of the grinding wheel 16 includes first surface condition data corresponding to the condition of the work-affected layer of the workpiece W, second surface condition data corresponding to the surface condition of the workpiece W, and chatter patterns of the workpiece W. A third surface condition data corresponding to the condition is included. However, the first surface state data may be the work-affected layer data itself, or may be data calculated based on the work-affected layer data. The second surface condition data may be the surface condition data of the workpiece W itself, or may be data calculated based on the surface condition data. The third surface state data may be the chatter pattern data itself, or may be data calculated based on the chatter pattern data.

The first learning model generation unit 150 learns the first learning model and has the same configuration as the corresponding configuration of the first embodiment. The first learning model storage unit 160 stores the first learning model generated by the first learning model generation unit 150 .

The third learning model generation unit 320 performs supervised learning to generate a third learning model. Specifically, the third learning model generation unit 320 uses the first input data acquired by the first input data acquisition unit 130 as the first learning input data, and the grinding wheel surface state data acquisition unit 311 acquires the machining A third learning model for estimating the surface condition of the grinding wheel 16 is generated by machine learning using the surface condition data of the grinding wheel 16 for each object W as second teaching data.

That is, the third learning model generation unit 320 uses the motion command data, the actual motion data, the first actual measurement data, and the second actual measurement data as the first input data for learning, and the grinding wheel surface condition data as the second teacher data. A third learning model is generated by machine learning. The third learning model is a model that indicates the relationship between the first learning input data and the second teacher data. Even if there are many kinds of input data for first learning, the third learning model can be generated by applying machine learning.

The third learning model is a model for estimating the extent to which the surface state of the grinding wheel 16 affects the grinding quality of the workpiece. For example, the first learning model is a model for estimating, as the surface state of the grinding wheel 16, a state in which the grinding wheel 16 is dented, clogged, or missing, and a state in which the grinding wheel 16 is over-dressed. is.

For example, the first learning model includes, as the surface conditions of the grinding wheel 16, a first surface condition corresponding to the condition of the work-affected layer of the workpiece W, a second surface condition corresponding to the surface properties of the workpiece W, and a machining This is a model for estimating the third surface state corresponding to the chatter pattern state of the object W. FIG. However, the third learning model is not limited to estimating all of these surface states, and may estimate only some of these surface states. The third learning model generated by the third learning model generation section 320 is stored in the third learning model storage section 330 .

(3-3. Detailed configuration of second learning phase 306 of machine learning device 300)
A detailed configuration of the second learning phase 306 of the machine learning device 300 will be described with reference to FIG. The configuration of the second learning phase 306 constitutes a grinding machine operating command data adjustment model generator.

The configuration of the second learning phase 306 includes an operation command data acquisition unit 111, a grinding quality data acquisition unit 140, a grinding wheel surface condition data acquisition unit 311, a grinding cycle time calculation unit 340, and a grinding wheel shape information acquisition unit 350. , a reward determination unit 210 , a second learning model generation unit 220 , and a second learning model storage unit 230 .

The motion command data acquisition unit 111 acquires motion command data regarding a plurality of workpieces W as second learning input data for machine learning. In addition, the grinding quality data acquisition unit 140 acquires grinding quality data regarding the plurality of workpieces W as first evaluation result data of machine learning. The grinding wheel surface condition data acquisition unit 311 acquires surface condition data of the grinding wheel 16 after grinding of each workpiece W as second evaluation result data of machine learning. Here, the second learning input data, the first evaluation result data, and the second evaluation result data are as shown in Table 4. Here, as shown in Table 4, the second learning input data is a large number of data, but it is not necessary to use all the data shown in Table 4, and only a part of the data may be used. .

The grinding cycle time calculator 340 calculates the grinding cycle time per workpiece W. FIG. The grinding cycle time includes the time required for grinding a plurality of workpieces W, the time required for replacing the grinding wheel 16 in the grinding, the time required for dressing the grinding wheel 16 in the grinding, and the time required for dressing the grinding wheel 16 in the grinding. It is a value obtained by dividing the total time required for truing by the number of workpieces W. That is, the less the grinding wheel 16 is replaced, the less the grinding wheel 16 is dressed, and the less the grinding wheel 16 is trued, the shorter the grinding cycle time.

The grinding wheel shape information acquisition unit 350 acquires shape information of the grinding wheel 16 . The grinding wheel shape information acquisition unit 350 acquires the dimension (diameter) of the grinding wheel 16 measured by the grinding wheel correction device 18 . That is, the grinding wheel shape information acquisition unit 350 is acquired when the grinding wheel correction device 18 performs truing or dressing of the grinding wheel 16 . Further, the grinding wheel shape information acquisition section 350 can acquire the dimensional change of the grinding wheel 16 and the deformation of the grinding wheel 16 as the shape information of the grinding wheel 16 .

The remuneration determination unit 210 acquires operation command data as second learning input data, grinding quality data as first evaluation result data, and surface condition data of the grinding wheel 16 as second evaluation result data, A reward for the motion command data is determined according to the grinding quality data and the surface condition data. Here, the reward is a reward for a combination of action command data in reinforcement learning.

As in the second embodiment, when the grinding quality data corresponding to the operation command data leads to a desired result, a large reward is given to the operation command data, and when an undesirable result is derived, the operation Small rewards (including negative rewards) are given for command data.

Furthermore, when the surface state data corresponding to the action command data yields a desirable result, a large reward is given to the motion command data, and when an undesirable result is derived, a small reward is given to the motion command data. reward is given.

For example, the remuneration determining unit 210 increases the remuneration when there is no work-affected layer corresponding to the first surface state data, and decreases the remuneration when there is a work-affected layer. Further, the remuneration determination unit 210 increases the remuneration when the surface texture data of the workpiece W corresponding to the second surface condition data is equal to or less than the predetermined threshold, and decreases the remuneration when the surface texture is greater than the predetermined threshold. Further, the remuneration determination unit 210 increases the remuneration when there is no chatter pattern corresponding to the third surface state data, and decreases the remuneration when there is a chatter pattern.

Further, the remuneration determination unit 210 acquires the grinding cycle time calculated by the grinding cycle time calculation unit 340, and determines remuneration for the operation command data according to the grinding cycle time. Specifically, the remuneration determination unit 210 increases the remuneration as the grinding cycle time becomes shorter. That is, the remuneration determination unit 210 increases the remuneration as at least one of the time required for exchanging the grinding wheel 16, the time required for dressing the grinding wheel 16, and the time required for truing the grinding wheel 16 is shortened.

Further, the remuneration determination section 210 determines remuneration based on the shape information of the grinding wheel 16 acquired by the grinding wheel shape information acquisition section 350 . Specifically, the remuneration determination unit 210 increases the remuneration as the dimensional change of the grinding wheel 16 becomes smaller and the shape of the grinding wheel 16 becomes smaller.

The second learning model generation unit 220 generates a second learning model for adjusting the action command data so as to increase the reward by machine learning. The generated second learning model is stored in the second learning model storage unit 230 .

(3-4. Detailed configuration of estimation phase 307 of machine learning device 300)
A detailed configuration of the estimation phase 307 of the machine learning device 300 will be described with reference to FIG. The estimation phase 307 includes a first input data acquisition unit 130, a first learning model storage unit 160, a third learning model storage unit 330, a grinding quality estimation unit 170, a grinding wheel surface condition estimation unit 360, and a determination unit 370 .

The grinding wheel surface state estimating unit 360 uses the first input data for a predetermined period during grinding of a new workpiece W as input data for estimation, and uses the first input data stored in the third learning model storage unit 330 as input data for estimation. By using the three learning models, the surface condition of the grinding wheel 16 when a new workpiece W is ground is estimated. Here, the third learning model is a model showing the relationship between the first learning input data and the second teacher data, as described above.

Therefore, the grinding wheel surface condition estimating section 360 estimates the extent to which the grinding quality of the workpiece W is affected as the surface condition of the grinding wheel 16 . For example, the grinding wheel surface state estimating unit 360 selects, as the surface state of the grinding wheel 16, a first surface state corresponding to the state of the work-affected layer of the workpiece W, a second surface state corresponding to the surface property of the workpiece W, Then, a third surface state corresponding to the chatter pattern state of the workpiece W is estimated. However, the grinding wheel surface condition estimator 360 may estimate only a part of the surface condition of the grinding wheel 16 instead of estimating the surface condition of all the grinding wheels 16 . For example, the grinding wheel surface condition estimator 360 may estimate only the first surface condition. In this case, the first learning model is generated as a model that estimates only the first surface state.

As described above, the grinding wheel surface state estimation unit 360 estimates a plurality of targets as the surface state. By using the third learning model generated by applying machine learning, the grinding wheel surface state estimator 360 can easily estimate a plurality of targets. In this way, machine learning device 300 can estimate a complicated target at once.

The determination unit 370 determines whether the workpiece W is good or bad based on the grinding quality of the workpiece W estimated by the grinding quality estimation unit 170 . Furthermore, based on the surface condition of the grinding wheel 16 estimated by the grinding wheel surface condition estimating unit 360, the determination unit 370 performs truing of the grinding wheel 16, dressing of the grinding wheel 16, and operation of the grinding wheel 16. Determine at least one of the exchange executions.

(3-5. Detailed configuration of update phase 308 of machine learning device 300)
A detailed configuration of the update phase 308 of the machine learning device 300 will be described with reference to FIG. The configuration of the update phase 308 includes an operation command data acquisition unit 111, a grinding quality data acquisition unit 140, a grinding wheel surface condition data acquisition unit 311, a grinding cycle time calculation unit 340, a grinding wheel shape information acquisition unit 350, It comprises a remuneration determination unit 210 , a second learning model storage unit 230 , and an action command data adjustment unit 240 .

The operation command data acquisition unit 111, the grinding quality data acquisition unit 140, and the grinding wheel surface condition data acquisition unit 311 acquire each data in grinding the new workpiece W, and the contents explained in the second learning phase 306 and are substantially identical. Also, the grinding cycle time calculation section 340 and the grinding wheel shape information acquisition section 350 are substantially the same as those described in the second learning phase 306 .

The remuneration determination unit 210 determines remuneration using the motion command data, the grinding quality data, and the surface condition data of the grinding wheel 16 acquired in grinding the new workpiece W. That is, the remuneration determination unit 210 determines remuneration for the operation command data for grinding the new workpiece W according to the grinding quality data and the surface condition data of the grinding wheel 16 . Further, the remuneration determination unit 210 determines a remuneration for operation command data according to the grinding cycle time and the shape information of the grinding wheel 16 . The second learning model storage unit 230 stores the second learning model generated by the second learning model generation unit 220 as described in the second learning phase 306 .

The motion command data adjustment unit 240 provides motion command data relating to grinding of the new workpiece W, grinding quality data of the new workpiece W, surface condition data of the grinding wheel 16 when grinding the new workpiece W, Using the reward and the second learning model, a method for adjusting the action command data is determined, and the action command data is adjusted based on the determined adjustment method. Here, the second learning model is a model generated by learning an adjustment method from the action command data before adjustment to the action command data after adjustment so as to increase the reward. The action command data adjuster 240 is substantially the same as the action command data adjuster 240 described in the second embodiment.

Then, the motion command data is updated using the second learning model generated by machine learning of the machine learning device 200 . Therefore, even if the grinding state changes, the operation command data is updated according to the current grinding state. By updating the operation command data in this manner, the grinding quality of the workpiece W can be improved.

Furthermore, by updating the operation command data, it is possible to perform grinding according to the surface condition of the grinding wheel. That is, the surface condition of the grinding wheel 16 is improved by updating the operation command data. By improving the surface condition of the grinding wheel 16, the grinding quality of the workpiece W can be improved. Furthermore, by updating the operation command data, the time required for exchanging the grinding wheel 16, the time required for dressing the grinding wheel 16, and the time required for truing the grinding wheel 16 are reduced. As a result, the grinding cycle time is shortened. Furthermore, by updating the operation command data, the dimensional change of the grinding wheel 16 is reduced, and the deformation of the grinding wheel 16 is reduced.

(4. Fourth embodiment)
(4-1. Overview of machine learning device 400)
An overview of the machine learning device 400 of the fourth embodiment will be described with reference to FIG. The machine learning device 400 (a) generates a relational information learning model for estimating the defective quality occurrence factor of the workpiece W determined to be defective, and (b) using the relational information learning model, A defective quality occurrence factor of the workpiece W determined to be defective is estimated. The configuration of the machine learning device 400 constitutes a defective quality occurrence factor estimating device.

Machine learning device 400 includes, as elements functioning in relational information learning phase 401 for generating a relational information learning model, element 401a for obtaining input data for relational information learning, element 401b for obtaining relational information teacher data, and relational information It has an element 401c for generating a learning model.

The relational information learning input data acquired in the element 401a is input data used for machine learning, such as actual operation data, first measured data, and second measured data. The relational information teacher data acquired in the element 401b is teacher data used for machine learning in supervised learning. The relational information teaching data is information about factors causing defective quality of the workpiece W, such as information about processing conditions (feed rate, etc.) of the workpiece W by the grinding machine 1, information about sharpness of the grinding wheel 16, and the like. . The relational information learning model generated in element 401c is a model ( function).

The machine learning device 400 includes an element 402a for acquiring input data for estimation and an element 402b for estimating the cause of defective quality, as elements functioning in an estimation phase 402 for estimating the cause of defective quality. The input data for estimation acquired in the element 402a is the same type of data as the input data for learning related information, and is data acquired for a workpiece W different from the workpiece W used for learning (a new workpiece W). is. The element 402b estimates the cause of defective quality using the input data for estimation and the relational information learning model. The relational information learning model used in element 402b is the relational information learning model generated by machine learning in the relational information learning phase 401 .

(4-2. Detailed configuration of relational information learning phase 401 of machine learning device 400)
A detailed configuration of the relational information learning phase 401 of the machine learning device 400 will be described with reference to FIG. The configuration of the relationship information learning phase 401 includes a defective product processing data storage unit 410, a good product processing data storage unit 420, a difference information extraction unit 430, a defective quality occurrence factor data storage unit 440, and a relationship information learning model generation unit 450. , and a relational information learning model storage unit 460 .

The defective product processing data storage unit 410 acquires and stores in advance a plurality of types of processing data (defective product processing data) regarding a plurality of defective workpieces W as input data for machine learning related information learning. In addition, the defective product processing data includes information on the defective quality of the workpiece W, which is a defective product. The defective quality includes the grinding quality that can be estimated by the grinding quality estimation unit 170, the surface condition of the grinding wheel 16 that can be estimated by the grinding wheel surface condition estimation unit 360, and the like.

The non-defective product processing data storage unit 420 preliminarily acquires and stores a plurality of types of processing data (non-defective product processing data) relating to a plurality of non-defective workpieces W as relational information teacher data for machine learning. The type of non-defective product processing data corresponds to the type of defective product processing data. The types of defective product processing data and non-defective product processing data include the actual operation data acquired from the sensor 21 by the actual operation data acquiring unit 112, the first measured data acquired from the sensors 22 and 23 by the measured data acquiring unit 120, and the Data corresponding to two actual measurement data and the like are exemplified.

In this embodiment, the non-defective product processing data storage unit 420 stores a plurality of types of non-defective product processing data. Just do it.

The difference information extraction unit 430 acquires the defective product processing data and the non-defective product processing data, and compares the defective product processing data and the non-defective product processing data. Then, the difference information extraction unit 430 extracts the processed data that is different between the two as processed data difference information. The defective quality occurrence factor data storage unit 440 acquires and stores in advance information (defective quality occurrence factor data) regarding the occurrence factors of the defective quality of the workpiece W. The defective quality occurrence factor data stored in the defective quality occurrence factor data storage unit 440 includes the machining conditions (feed rate) of the workpiece W by the grinding machine 1, the sharpness of the grinding wheel 16, the temperature of the machining point, the grinding machine 1 Vibration of the component parts of is exemplified.

The relational information learning model generation unit 450 performs supervised learning to generate a relational information learning model. Specifically, the relationship information learning model generation unit 450 uses the processed data difference information extracted by the difference information extraction unit 430 as input data for relationship information learning, and determines the defective quality stored in the defective quality occurrence factor data storage unit 440. A relational information learning model is generated for estimating the factor (defective factor) that caused the poor quality of the workpiece W by machine learning using the occurrence factor data as the relational information teacher data.

The relationship information learning model storage unit 460 stores the relationship information learning model generated by the relationship information learning model generation unit 450 . In addition, in the relational information learning model storage unit 460, the relational learning model stores a plurality of types of defective qualities (grinding quality estimated by the grinding quality estimating unit 170, or grindstone The surface condition of the grinding wheel 16 estimated by the vehicle surface condition estimation unit 360, etc.) are stored in association with each other. In the relational information learning model storage unit 460, association between the relational learning model and the defective quality can be omitted.

In this way, the relational information learning model generating unit 450 generates a learning model (relational information learning model) regarding the occurrence factor relational information indicating the relationship between the processed data difference information and the cause of defective quality. The relational information learning model is a model for estimating factors that have caused defective quality in the workpiece W that has been determined to be defective. By using the relational information learning model, the machine learning device 400 can further clarify the relation between the processed data difference information and the cause of defective quality.

In the present embodiment, the relational information learning model storage unit 460 stores a plurality of types of relational information learning models indicating the relationship between the processed data difference information and a plurality of types of defective quality occurrence factors. Only part of the relational information learning model may be stored.

(4-3. Detailed configuration of estimation phase 402 of machine learning device 400)
A detailed configuration of the estimation phase 402 of the machine learning device 400 will be described with reference to FIG. 13 . The estimation phase 402 includes a defective product processing data storage unit 410, a non-defective product processing data storage unit 420, a difference information extraction unit 430, a relationship information learning model storage unit 460, and a defect factor estimation unit 470. be done.

The defective product machining data storage unit 410 stores the actual operation data acquisition unit 112 and the first actual measurement data acquisition unit 121 when the determination units 180 and 370 determine that the newly ground workpiece W is defective. , from the second measured data acquisition unit 122, the actual operation data, the first measured data, and the second measured data are acquired and stored.

The difference information extraction unit 430 compares the defective product processing data newly acquired and stored by the defective product processing data storage unit 410 with the non-defective product processing data stored in the non-defective product processing data storage unit 420, thereby identifying the defective product. New processed data difference information is extracted from the difference between the processed data and non-defective product processed data.

Next, the defect factor estimation unit 470 uses the newly extracted processing data difference information as input data for estimation, and uses the relationship information learning model stored in the relationship information learning model storage unit 460 to newly grind the In the workpiece W, the factor that caused the defective quality is estimated. As a result, the machine learning device 400 can estimate the cause of the poor quality of the workpiece W that has been determined to be defective by the determination units 180 and 370 . Further, the defect factor estimation unit 470 estimates factors causing defective quality based on the processed data difference information extracted by the difference information extraction unit 430 . Therefore, the defective factor estimating section 470 can easily estimate the defective quality occurrence factor.

In this embodiment, the non-defective product processing data storage unit 420 stores a plurality of types of non-defective product processing data, and the difference information extraction unit 430 compares the plurality of types of non-defective product processing data with the defective product processing data, respectively. Extract processing data difference information of species. Therefore, the defect factor estimating section 470 can select one defective quality occurrence factor from among a plurality of types of defective quality occurrence factors, so that the accuracy of estimation by the defect factor estimating section 470 can be improved.

Moreover, the non-defective product processing data stored in the non-defective product processing data storage unit 420 is created based on the actual operation data or actual measurement data regarding a plurality of non-defective products obtained in advance, so that the quality of the non-defective product processing data can be improved. Accordingly, the difference information extraction unit 430 can accurately extract the processed data difference information.

In the present embodiment, the machine learning device 400 estimates the cause of defective quality using a learning model related to the cause-related information. does not have to be That is, the machine learning device 400 compares the defective product machining data obtained from at least one defective workpiece W with the non-defective product machining data obtained from at least one non-defective workpiece W. Alternatively, information in which one piece of processing data difference information is associated with information about a specific defective quality possessed by one defective workpiece W may be included as occurrence factor relation information. In this case as well, the defect factor estimating unit 470, based on the processing data difference information based on the defective product processing data and the non-defective product processing data obtained from the workpiece W newly determined to be defective, and the occurrence factor relationship information, An inference can be made as to whether or not it has a particular bad quality.

In addition, in the update phases 204 and 308 in the second and third embodiments, the action command data adjusting section 240 may adjust the action command data based on the estimation result by the defect factor estimating section 470 . In this case, the machine learning devices 200 and 300 can improve the grinding quality of the workpiece.

1: grinder, 2: external device, 11: bed, 12: headstock (structural member), 12a, 14a, 15a, 16a: motor (driving device), 13: tailstock (structural member), 14: traverse Base (structural member, moving body), 15: Wheelhead (structural member, moving body), 16: Grinding wheel, 17: Sizing device (driving device), 18: Grinding wheel correcting device (driving device), 19: Coolant Apparatus 20: Control device 21, 22, 23: Sensor 100: Machine learning device 101: First learning phase 102: Estimation phase 110: Action-related data acquisition unit 111: Action command data acquisition unit 112 : actual operation data acquisition unit 120: actual measurement data acquisition unit 121: first actual measurement data acquisition unit 122: second actual measurement data acquisition unit 130: first input data acquisition unit 140: grinding quality data acquisition unit 150 : first learning model generation unit, 160: first learning model storage unit, 170: grinding quality estimation unit, 180: determination unit, 200: machine learning device, 203: second learning phase, 204: update phase, 210: reward Determination unit 220: Second learning model generation unit 230: Second learning model storage unit 240: Operation command data adjustment unit 300: Machine learning device 305: First learning phase 306: Second learning phase 307 : Estimation phase 308: Update phase 310: Quality data acquisition unit 311: Grinding wheel surface condition data acquisition unit 320: Third learning model generation unit 330: Third learning model storage unit 340: Grinding cycle time calculation Section 350: Grinding wheel shape information acquisition section 360: Grinding wheel surface state estimation section 370: Discrimination section 400: Machine learning device (defective quality occurrence factor estimation device) 420: Non-defective product processing data storage section 430: Difference Information extraction unit 460: relational information learning model storage unit (relationship information storage unit) 470: defect factor estimation unit W: workpiece (grinding part)

The present invention relates to a grinding quality estimation model generation device, a grinding quality estimation device, a defective quality occurrence factor estimation device, and an operation command data updating device for a grinding machine.

An object of the present invention is to provide a grinding quality estimation model generation device and a grinding quality estimation device that can acquire the grinding quality of a workpiece. Another object of the present invention is to provide a defective quality generation factor estimating device capable of estimating the factors that cause defective quality in a workpiece that has been determined to be defective. Another object of the present invention is to provide an apparatus for updating operation command data of a grinding machine in order to acquire operation command data of the grinding machine that can improve the grinding quality by using the grinding quality of the workpiece. do.

(1. Grinding quality estimation model generation device)
In one aspect of the present invention, the workpiece is ground by the grinding wheel in a grinder in which the workpiece is ground by the grinding wheel in the order of rough grinding, fine grinding, fine grinding, and spark-out . The actual measurement data, which is at least one of the first actual measurement data representing the state of the structural member of the grinding machine and the second actual measurement data related to the grinding portion, is measured data for each workpiece and an actual measurement data acquisition unit that acquires each grinding step ;
First learning model generation for generating a first learning model for estimating the grinding quality of the workpiece for each grinding process by machine learning using the measured data on the plurality of workpieces as input data for first learning. Department and
A grinding quality estimation model generation device comprising :

The first learning model is generated by machine learning using measured data as first learning input data. The measured data is at least one of first measured data representing the state of the structural member of the grinder and second measured data relating to the grinding portion. The measured data are data acquired for each workpiece and for each grinding process . For example, the actual measurement data includes data from the initial stage of grinding to the final stage of grinding, data from the initial stage of rough grinding to the final stage of rough grinding, and the like. Therefore, even the actual measurement data of one workpiece is a large amount of data. Furthermore, when it comes to actual measurement data for a plurality of workpieces, the amount of data is extremely large. However, by using machine learning, the first learning model can be easily generated even with a large amount of measured data on multiple workpieces.

(2. Grinding quality estimation device)
Another aspect of the present invention is the grinding quality estimation model generation device described above,
In the estimation phase, the grinding quality of the new workpiece is estimated using the input data for estimation, which is the measured data for each grinding process and the first learning model when grinding the new workpiece. a grinding quality estimation unit that estimates
A grinding quality estimation device comprising :
By using the first learning model generated by machine learning, it is possible to estimate the grinding quality of the new workpiece based on the input data for estimation, which is a large amount of measured data regarding the new workpiece.

(3. Defective Quality Occurrence Factor Estimating Device)
Another aspect of the present invention is the grinding quality estimation device described above,
Stores the actual operation data of the driving device controlled by the control device of the grinding machine, which is the actual operation data related to the non-defective product acquired in advance, or the non-defective product processing data created based on the actual measurement data related to the non-defective product acquired in advance. a non-defective product processing data storage unit for
Processing for identifying a cause of defective quality by comparing the non-defective product processing data with the defective product processing data, which is the actual operation data or actual measurement data regarding the workpiece estimated to be defective by the discrimination unit. a difference information extraction unit for extracting data difference information;
A defective quality occurrence factor estimating device comprising :

(4. Grinding Machine Operation Command Data Update Device )
Another aspect of the present invention is the grinding quality estimation device described above,
an operation command data acquiring unit for acquiring operation command data for a control device of the grinding machine for each workpiece when grinding the workpiece with the grinding wheel in the grinding machine ;
a grinding quality data acquisition unit that acquires grinding quality data of the workpiece for each workpiece;
a remuneration determination unit that determines a remuneration for the operation command data according to the grinding quality data for each workpiece;
a second learning model generation unit for generating a second learning model for adjusting the motion command data so as to increase the reward by machine learning using the motion command data and the reward for the plurality of workpieces; ,
In an update phase, using the motion command data for grinding a new workpiece, the grinding quality data estimated by the grinding quality estimation unit for the new workpiece, the reward, and the second learning model, an operation command data adjustment unit that adjusts the operation command data;
An apparatus for updating operation command data for a grinding machine, comprising:

The updating device uses machine learning to generate a second learning model for adjusting the operation command data of the grinding machine. Machine learning uses motion command data and rewards for a plurality of workpieces. Therefore, even if a large amount of data is used, the second learning model can be easily generated by applying machine learning. Additionally, in machine learning, grinder motion command data is adjusted to increase rewards determined using workpiece grinding quality data. Therefore, it is possible to generate operation command data that will improve the grinding quality.

Further , the motion command data is updated using the second learning model generated by machine learning. Therefore, even if the grinding state changes, the operation command data is updated according to the current grinding state. By updating the operation command data in this way, it is possible to improve the grinding quality of the workpiece.

Claims (27)

At least one of first measured data representing the state of a structural member of the grinder and second measured data relating to a grinding portion, which is measured data when a workpiece is ground by a grinding wheel in a grinder. an actual measurement data acquisition unit that acquires the actual measurement data for each workpiece for a predetermined period;
a first learning model generation unit that generates a first learning model for estimating the grinding quality of the workpiece by machine learning using the measured data on the plurality of workpieces as input data for first learning;
A grinding quality estimation model generation device. The actual measurement data acquisition unit obtains the first actual measurement data, which is at least one of vibration of the structural member of the grinding machine and deformation amount of the structural member of the grinding machine, and Acquiring the second measured data, which is at least one of dimensions and grinding point temperature, as the measured data;
The first learning model generation unit generates the first learning model by machine learning using the first measured data and the second measured data regarding the plurality of workpieces as the first learning input data. Item 1. The grinding quality estimation model generation device according to item 1. The grinding quality estimation model generation device further comprises a grinding quality data acquisition unit that acquires grinding quality data of the workpiece for each workpiece,
3. The grinding quality estimation model generation device according to claim 1, wherein said first learning model generation unit generates said first learning model by said machine learning using said grinding quality data as teacher data. 4. Grinding according to claim 3, wherein the grinding quality data of the workpiece is at least one of process-affected layer data of the workpiece, surface texture data of the workpiece, and chatter pattern data of the workpiece. Quality estimation model generator. The grinding quality estimation model generation device further generates the motion-related data, which is at least one of motion command data to the control device of the grinding machine and actual motion data of the driving device controlled by the control device. An operation-related data acquisition unit that acquires data for a predetermined period for each workpiece,
2. The first learning model generation unit generates the first learning model by the machine learning using the measured data and the motion-related data relating to the plurality of workpieces as the first learning input data. -4, the grinding quality estimation model generation device according to any one of items. A grinding quality estimation model generation device according to any one of claims 1 to 5;
Grinding quality for estimating the grinding quality of the new workpiece using the input data for estimation, which is the actual measurement data for a predetermined period while grinding the new workpiece, and the first learning model an estimation unit;
A grinding quality estimation device. The first learning model generation unit estimates, as the grinding quality of the workpiece, at least one of a state of a work-affected layer of the workpiece, a surface texture of the workpiece, and a chatter pattern state of the workpiece. generating said first learning model for
The grinding quality estimation unit estimates at least one of a state of a work-affected layer of the workpiece, a surface texture of the workpiece, and a chatter pattern state of the workpiece as the grinding quality of the new workpiece. The grinding quality estimating device according to claim 6, wherein 8. Grinding according to claim 6 or 7, wherein said grinding quality estimating device further comprises a discriminating section for discriminating quality of said workpiece based on said grinding quality of said workpiece estimated by said grinding quality estimating section. quality estimator. A grinding quality estimation device according to claim 8;
Stores the actual operation data of the driving device controlled by the control device of the grinding machine, which is the actual operation data related to the non-defective product acquired in advance, or the non-defective product processing data created based on the actual measurement data related to the non-defective product acquired in advance. a non-defective product processing data storage unit for
Processing for identifying a cause of defective quality by comparing the non-defective product processing data with the defective product processing data, which is the actual operation data or actual measurement data regarding the workpiece estimated to be defective by the discrimination unit. a difference information extraction unit for extracting data difference information;
A defective quality occurrence factor estimation device. The defective quality occurrence factor estimating device further comprises:
a relationship information storage unit for storing occurrence factor relationship information indicating the relationship between the processed data difference information and the cause of the defective quality;
a defect factor estimating unit for estimating the cause of the defective quality based on the relationship between the processed data difference information and the occurrence factor relationship information;
The defective quality occurrence factor estimating device according to claim 9, comprising: 11. The defective quality occurrence factor estimation according to claim 10, wherein said relationship information storage unit stores a plurality of types of said occurrence factor relationship information indicating a relationship between said processed data difference information and said multiple types of occurrence factors of said defective quality. Device. 12. The defective quality occurrence factor estimation according to claim 10 or 11, wherein said occurrence factor related information is a learning model generated by said machine learning using said processed data difference information and said occurrence factors of defective quality as learning data. Device. 13. Any one of claims 9 to 12, wherein the cause of the defective quality is at least one of conditions for processing the workpiece by the grinding machine, sharpness of the grinding wheel, and vibration of components of the grinding machine. 1. The apparatus for estimating factors causing defective quality according to item 1. 14. The defective quality occurrence factor estimation according to any one of claims 9 to 13, wherein said difference information extraction unit extracts a difference between said non-defective product processed data and said defective product processed data as said processed data difference information. Device. 15. The defective quality occurrence factor estimating device according to claim 9, wherein said non-defective product processing data is created based on said actual measurement data or said actual operation data relating to a plurality of non-defective products obtained in advance. The non-defective product processing data storage unit stores a plurality of types of non-defective product processing data,
16. The difference information extraction unit according to any one of claims 9 to 15, wherein the plurality of types of non-defective product processing data and the defective product processing data are compared with each other to extract a plurality of types of processing data difference information. 3. The device for estimating factors causing defective quality according to 1. an operation command data acquiring unit for acquiring operation command data for a control device of the grinder for each workpiece when grinding the workpiece with a grinding wheel in the grinder;
a grinding quality data acquisition unit that acquires grinding quality data of the workpiece for each workpiece;
a remuneration determination unit that determines a remuneration for the operation command data according to the grinding quality data for each workpiece;
a second learning model generation unit for generating a second learning model for adjusting the motion command data so as to increase the reward by machine learning using the motion command data and the reward for the plurality of workpieces; ,
A grinding machine operation command data adjustment model generation device. 18. Grinding according to claim 17, wherein the grinding quality data of the workpiece is at least one of process-affected layer data of the workpiece, surface texture data of the workpiece, and chatter pattern data of the workpiece. Panel operation command data adjustment model generator. 19. Grinding according to claim 18, wherein, in the work-affected layer data of the workpiece, the reward determination unit increases the reward when there is no work-affected layer, and decreases the reward when the work-affected layer is present. Panel operation command data adjustment model generator. 19. The operation of the grinding machine according to claim 18, wherein the remuneration determining unit increases the remuneration when the surface texture data of the workpiece is equal to or less than a predetermined threshold, and decreases the remuneration when the surface texture data of the workpiece is greater than the predetermined threshold. Command data adjustment model generator. 19. The grinder according to claim 18, wherein in the chatter pattern data of the workpiece, the reward determining unit increases the reward when there is no chatter pattern, and decreases the reward when there is a chatter pattern. motion command data adjustment model generator. The grinding machine operation command data adjustment model generating device further comprises a surface condition data acquisition unit for acquiring surface condition data of the grinding wheel for each workpiece,
The grinder according to any one of claims 17 to 21, wherein the remuneration determination unit determines a remuneration for the operation command data according to the grinding quality data and the surface condition data for each workpiece. Operation command data adjustment model generator. 23. The motion command data adjustment model generation device for a grinder according to claim 22, wherein said surface condition data of said grinding wheel is data affecting grinding quality of said workpiece. The surface condition data of the grinding wheel includes first surface condition data corresponding to the condition of a work-affected layer of the workpiece, second surface condition data corresponding to the surface texture of the workpiece, and chatter of the workpiece. 24. The operation command data adjustment model generation device for a grinding machine according to claim 23, wherein the operation command data adjustment model generation device for a grinding machine is at least one of the third surface condition data corresponding to the pattern condition. The grinding quality data of the workpiece according to any one of claims 17-24, wherein the grinding quality is estimated by the grinding quality estimation device according to any one of claims 6-8. Grinding machine operation command data adjustment model generator. A grinding machine operation command data adjustment model generation device according to any one of claims 17 to 25,
a motion command data adjustment unit that adjusts the motion command data using the motion command data for grinding a new workpiece, the grinding quality data for the new workpiece, the reward, and the second learning model;
A grinding machine operation command data update device. The operation command data update device for the grinder further comprises:
Equipped with the defective quality occurrence factor estimation device according to any one of claims 10 to 16,
27. The operation command data update device for a grinding machine according to claim 26, wherein said operation command data adjusting unit further adjusts said operation command data using said occurrence factor relation information.

Images