JP2021077205A - Quality prediction system - Google Patents

Abstract

To provide a quality prediction system that is superb in versatility.SOLUTION: A quality prediction system 1 includes a model production unit 73 that regards as an explanatory variable a feature quantity of condition data detected by a first detector 13 at every processing step and acquired via a control device 12, regards as an objective variable quality related data relevant to the processing quality of a workpiece subjected to grinding processing, performs machine learning using a training data set including the explanatory variable and objective variable, and thus produces a learned model, a model memory unit 81 that stores the learned model produced at every grinding step, and a quality prediction unit 83 that predicts the quality of the workpiece using the learned model and condition data.SELECTED DRAWING: Figure 3

Description

The present invention relates to a quality prediction system.

Patent Document 1 discloses an abnormality detection device capable of appropriately detecting an abnormality in tool change based on a state in which a tool is attached to the tool change device. In the abnormality detection device, machine learning is performed to relate the tool weight data of the tool attached to the tool changer, the tool balance data related to the tool balance, and the tool change state data related to the state at the time of tool change. ..

Japanese Unexamined Patent Publication No. 2019-38074

However, the abnormality detection device of Patent Document 1 is a machine that associates tool weight data of a tool attached to a tool changer, tool balance data related to tool balance, and tool change state data related to a state at the time of tool change. Only learn. That is, the abnormality detection device of Patent Document 1 is specialized in determining the necessity of tool replacement.

By the way, when machining a workpiece, it is important to determine the necessity of tool replacement because it affects the quality of the workpiece, but there are various other events that affect the quality. To do. Therefore, when it is assumed that the abnormality detection device of Patent Document 1 is applied to various events, for example, the operator determines the correlation of various data for each event that occurs and uses the machine as the abnormality detection device. Learning needs to be done, which is time consuming and extremely complicated. Therefore, especially when predicting quality, it is desired to have versatility so that it can respond to various events that occur.

An object of the present invention is to provide a quality prediction system having excellent versatility.

The quality prediction system detects observable state data in machining by a processing machine that performs machining of a workpiece through multiple machining processes under the control of a control device, and uses the detected state data as a control device. The feature amount of the output detector and the state data detected by the detector and acquired via the control device in each machining process is used as the explanatory variable, and the quality-related data related to the machining quality of the workpiece by machining is used as the objective variable. Then, the trained model storage unit that stores the trained model generated by performing machine learning using the training data set including the explanatory variables and the objective variables for each machining process, and the trained model and the state data are stored. It is equipped with a quality prediction unit that predicts the quality of the workpiece using it.

According to this, the quality prediction system can extract the feature amount from the state data detected in each machining process of a plurality of machining steps and generate a trained model for each machining process using the extracted feature amount. it can. In this way, by extracting the feature quantity for each machining process and generating the trained model, it is possible to predict the quality of the workpiece objectively and accurately with versatility.

It is a figure which shows the structure of the quality prediction system. It is a figure which shows an example of a processing machine. It is a figure which shows the functional block composition of a quality prediction system. It is a figure explaining the data format output from a control device. It is a figure explaining the empty grinding period of the rough grinding process which constitutes a grinding process (machining process). It is a graph which shows the time change of the drive current value which rotates a grindstone, and the surface roughness of a workpiece in the fine grinding process which constitutes a grinding process (machining process). It is a graph which shows the time change of the drive current value for each grinding process (machining process). It is a histogram explaining the contribution degree (contribution rate) of a feature amount. It is a figure which shows the functional block composition of the quality prediction system according to another example.

(1. Processing machine to which the quality prediction system is applied)
The quality prediction system predicts the processing quality of a processing machine that performs machining of a workpiece through a plurality of processing processes. Here, the machining quality includes a score for a preset quality standard, the presence or absence of quality abnormality, the life of a tool that affects the machining quality, and the like. The life of the tool includes not only a state in which the tool is completely unusable, but also a state in which the tool is re-sharpened, a grindstone wheeling, a dressing, and the like are modified.

The processing machine includes a grinding machine that performs grinding. The grinding process by the grinding machine includes rough grinding, fine grinding, fine grinding, spark out and the like. Further, the processing machine includes a processing machine that performs cutting such as a machining center, a lathe, and a gear processing device.

(2. Outline of the configuration of the quality prediction system 1)
The outline of the configuration of the quality prediction system 1 will be described with reference to FIG. The quality prediction system 1 includes at least one processing machine 10 and one arithmetic unit (20, 30). The processing machine 10 may be targeted at one machine, or may be targeted at a plurality of machines as shown in FIG. In this example, the quality prediction system 1 includes a case where a plurality of processing machines 10 are provided as an example.

The processing machine 10 is at least a processing machine main body 11 that performs machining of the workpiece W using the tool T, and a first detector that detects state data observable in the processing machine main body 11 during machining of the workpiece W. 13 and.

The arithmetic unit (20, 30) predicts the processing quality of the workpiece W machined by the processing machine 10 by applying machine learning using the state data detected by the first detector 13. In FIG. 1, the arithmetic unit (20, 30) is composed of the learning processing device 20 and the prediction arithmetic unit 30, and the learning processing unit 20 and the prediction arithmetic unit 30 are shown as independent configurations, but one device. It can also be. Further, a part or all of the arithmetic units (20, 30) can be incorporated into the processing machine 10.

In this example, the case where the learning processing device 20 and the prediction calculation device 30 have independent configurations will be taken as an example. Further, the learning processing device 20 has a so-called server function, and is connected to a plurality of processing machines 10 in a communicable manner. On the other hand, the prediction arithmetic unit 30 is provided on each processing machine 10 on a one-to-one basis and is communicably connected to each processing machine 10. That is, the plurality of prediction arithmetic units 30 function as so-called edge computers, and can realize high-speed arithmetic processing.

(3. Details of the configuration of the quality prediction system 1)
The configuration of the quality prediction system 1 will be described in more detail with reference to FIG. The quality prediction system 1 includes a plurality of processing machines 10, one learning processing device 20 that functions as a part of an arithmetic unit, and a plurality of prediction calculation devices 30 that function as another part of the arithmetic unit.

As described above, various processing machines can be applied to each processing machine 10. The processing machine 10 includes a processing machine main body 11 that performs machining of a workpiece W using a tool T, a control device 12 that controls the processing machine main body 11, a first detector 13, a second detector 14, and the like. It mainly includes an interface 15.

The processing machine main body 11 has a tool T, supports the workpiece W, and has a configuration in which the tool T and the workpiece W are relatively moved. That is, the processing machine main body 11 includes a structure and a driving device for driving the structure. The control device 12 includes a CNC device, a PLC device, and the like. The control device 12 controls a drive device or the like in the processing machine main body 11. The interface 15 is a device capable of communicating with the outside with the processing machine main body 11, the control device 12, the first detector 13 and the second detector 14. The outside includes a learning processing device 20 and an inspection device 2 that inspects the workpiece W machined by the processing machine 10.

The first detector 13 detects state data related to the observable machining operation of the workpiece W (machining load, drive load of the drive device, etc.) in the machining machine main body 11 during machining of the workpiece W. The first detector 13 is, for example, a displacement sensor that detects relative position (or distance) data between the tool T and the workpiece W, a current sensor that detects drive current data of a motor as a drive device, and a processing machine main body. A vibration sensor for detecting vibration data of 11 constituent members, a microphone for detecting sound data during processing, and the like. That is, the state data detected by the first detector 13 is, for example, position data, drive current data, vibration data, sound data, and the like.

The second detector 14 detects quality-related data related to the observable processing quality of the workpiece W in the processing machine main body 11 during or after the machining of the workpiece W. The second detector 14 is, for example, an image sensor that detects color data on the surface of the work W, a roughness sensor that detects surface roughness data on the machined surface of the work W, and hardness of the machine W on the machined surface. A hardness sensor or the like that detects data.

Further, the second detector 14 is used as a sensor for detecting a state affecting the machining quality of the workpiece W, for example, a sensor (image sensor) for detecting wear state data of the tool T or an operating state of the machining machine main body 11. It is a sensor or the like that detects anomaly occurrence data indicating the occurrence of an anomaly that affects it. That is, the quality-related data detected by the second detector 14 is, for example, color data, roughness data, hardness data, and the like, wear state data, and abnormality occurrence data related to the processing quality of the workpiece W. The inspection device 2 detects the quality-related data related to the processing quality of the machined workpiece W as well as the quality-related data detected by the second detector 14.

The learning processing device 20 includes a processor 21, a storage device 22, an interface 23, and the like. Further, the learning processing device 20 has a server function, and is communicably connected to a plurality of processing machines 10 (including, for example, processing machines 10 installed in other factories separated from each other).

The learning processing device 20 applies machine learning to the workpiece based on the state data detected by the first detector 13, the second detector 14 and / and the quality-related data output from the inspection device 2. Generate a trained model for predicting the quality of W. In particular, the learning processing device 20 divides the state data output in time series from the first detector 13 for each processing process, and uses, for example, the feature amount extracted by statistical calculation for each state data. A trained model is generated for each processing process. That is, the learning processing device 20 generates a plurality of trained models for each processing process.

For example, in the learning processing device 20, a plurality of trained models are generated by generating at least one trained model for one machining process in a plurality of machining steps. That is, the learning processing device 20 processes the trained model in a plurality of processing processes by generating a learned model for each of the individual processing processes, assuming that the processing qualities in the individual processing processes are different processing qualities. Generate for each quality.

Each prediction arithmetic unit 30 includes a processor 31, a storage device 32, an interface 33, and the like. Further, the prediction calculation device 30 is communicably connected to the learning processing device 20 as a server and the corresponding processing machine 10.

Each prediction arithmetic unit 30 is arranged at a position close to each processing machine 10, and functions as a so-called edge computer. The prediction arithmetic unit 30 uses the trained model generated for each machining process by the learning processing device 20, and is based on the state data detected by the first detector 13 during the machining of the work W. Predict the processing quality of. Then, in this example, as will be described later, the prediction arithmetic unit 30 presents a feature amount having a high degree of contribution that causes the machining state of the workpiece W by predicting the machining quality of the workpiece W, and the tool. The correction of T is also presented.

The quality prediction system 1 further includes a common display device 40 and a plurality of individual display devices 50. However, the quality prediction system 1 may be configured not to include the common display device 40, or may be configured not to include the individual display device 50. The common display device 40 is arranged corresponding to the learning processing device 20. Further, the individual display device 50 is arranged corresponding to each of the processing machines 10.

(4. Example of processing machine main body 11)
As an example of the processing machine main body 11, a grinding machine for performing grinding will be described with reference to FIG. As described above, the processing machine main body 11 is an example of a grinding machine, and may be applied to other processing.

The grinding machine 60, which is the processing machine main body 11, is a processing machine for grinding the workpiece W. As the grinding machine 60, a grinding machine having various configurations such as a cylindrical grinding machine and a cam grinding machine can be applied. In this example, the grinding machine 60 is a grindstone traverse type cylindrical grinding machine as an example. However, a table traverse type can be applied to the grinding machine 60.

The grinding machine 60 mainly includes a bed 61, a headstock 62, a tailstock 63, a traverse base 64, a grindstone base 65, a grindstone 66 (corresponding to a tool T), a sizing device 67, a grindstone correction device 68, and A coolant device 69 is provided.

The bed 61 is fixed on the installation surface. The headstock 62 is provided on the upper surface of the bed 61 on the front side in the X-axis direction (lower side in FIG. 2) and one end side in the Z-axis direction (left side in FIG. 2). The headstock 62 rotatably supports the workpiece W around the Z axis. The workpiece W is rotated by driving a motor 62a provided on the headstock 62. The tailstock 63 is located on the upper surface of the bed 61 at a position facing the headstock 62 in the Z-axis direction, that is, the front side in the X-axis direction (lower side in FIG. It is provided on the right side of FIG. 2). That is, the spindle base 62 and the tailstock 63 rotatably support both ends of the workpiece W.

The traverse base 64 is provided on the upper surface of the bed 61 so as to be movable in the Z-axis direction. The traverse base 64 is moved by driving a motor 64a provided on the bed 61. The grindstone base 65 is provided on the upper surface of the traverse base 64 so as to be movable in the X-axis direction. The grindstone base 65 is moved by driving a motor 65a provided on the traverse base 64. The grindstone wheel 66 is rotatably maintained on the grindstone stand 65. The grindstone wheel 66 is rotated by driving a motor 66a provided on the grindstone base 65. The grindstone 66 is configured by fixing a plurality of abrasive grains with a bond material.

The sizing device 67 measures the dimension (diameter) of the workpiece W. The grindstone correction device 68 corrects the shape of the grindstone 66. The grindstone correction device 68 is a device for performing the growing of the grindstone 66. The grindstone correction device 68 may be a device for dressing the grindstone 66 in addition to or in place of the trueing. Further, the grindstone correction device 68 also has a function of measuring the dimensions (diameter) of the grindstone 66.

Here, the trueing is a reshaping work, such as a work of forming the grindstone 66 according to the shape of the workpiece W when the grindstone 66 is worn by grinding, a work of removing the runout of the grindstone 66 due to uneven wear, and the like. Is. Dressing is a work of sharpening (sharpening), adjusting the amount of protrusion of abrasive grains, and creating a cutting edge of abrasive grains. Dressing is the work of correcting fouling, clogging, spillage, etc., and is usually performed after truing.

The coolant device 69 supplies the coolant to the grinding point of the workpiece W by the grindstone 66. The coolant device 69 cools the recovered coolant to a predetermined temperature and supplies it to the grinding point again.

The control device 12 provided in the grinding machine 60 is driven by each drive based on an NC program generated based on operation command data such as the shape of the workpiece W, the machining conditions, the shape of the grindstone 66, and the coolant supply timing information. Control the device. That is, the control device 12 inputs the operation command data, generates an NC program based on the operation command data, and controls the motors 62a, 64a, 65a, 66a, the coolant device 69, and the like based on the NC program. Grind the workpiece W. In particular, the control device 12 grinds the workpiece W until it has a finished shape based on the diameter of the workpiece W measured by the sizing device 67. Further, the control device 12 corrects the grindstone 66 (truing and dressing) by controlling the motors 64a, 65a, 66a, the grindstone repair device 68, and the like at the timing of modifying the grindstone 66. ..

(5. Functional block configuration of quality prediction system 1)
The functional blocks of the quality prediction system 1 will be described with reference to FIG. The quality prediction system 1 includes a control device 12, a first detector 13, a second detector 14, a learning processing device 20, a prediction calculation device 30, and display devices 40 and 50.

As described above, the first detector 13 detects the observable state data in the processing machine main body 11 during the processing of the workpiece W. The state data includes, for example, drive load data in the motor 66a that rotationally drives the grindstone 66 (tool T). The state data further includes drive load data in the motor 62a of the spindle 62 that rotationally drives the workpiece W. The drive load data corresponds to the drive current data of the motors 62a and 66a. The state data may include position data, vibration data, processing sound data, and the like.

As described above, the second detector 14 detects quality-related data related to the observable processing quality of the workpiece W in the processing machine main body 11 during the machining of the workpiece W. The quality-related data includes color data, roughness data, hardness data, and the like on the processed surface of the workpiece W during processing, that is, quality data. The state data and quality-related data are time-series data from the start of machining to the end of machining in one workpiece W. In this example, a case where the grinding machine 60 grinds the workpiece W will be illustrated. Therefore, "from the start of processing to the end of processing" includes a rough grinding process, a fine grinding process, a fine grinding process, a spark-out process, and the like.

Further, the control device 12 of the grinding machine 60 (processing machine 10) includes a counter. The counter counts the number of workpieces W ground (machined) by the grinding machine 60 (machining machine 10). In this example, the grindstone repairing device 68 counts the number of machining of the workpiece W from the time when the grindstone 66 (tool T) is trued or dressed. In addition to the control device 12, the counter may be provided by the first detector 13 itself or by the inspection device 2.

The learning processing device 20 operates based on the state data detected by the first detector 13 and the quality-related data related to the processing quality of the workpiece W detected by the second detector 14 (or the inspection device 2). Generate a trained model for quality prediction of object W. The learning processing device 20 includes a training data set acquisition unit 71, a training data set storage unit 72, and a model generation unit 73.

The training data set acquisition unit 71 acquires a training data set for performing machine learning for each processing process. The training data set acquisition unit 71 includes a state data acquisition unit 71a, a state data division unit 71b, a feature amount extraction unit 71c, and a quality-related data acquisition unit 71d.

Here, the control device 12 outputs the state data output from the first detector 13 in time series (continuously) to the learning processing device 20. Further, the control device 12 adds (associates) process information for identifying the current processing process, which is output in time series based on the NC program generated according to the operation command data, to the state data. Then, the control device 12 outputs the state data and the process information to the learning processing device 20 according to the data format illustrated in FIG.

As shown in FIG. 4, the data format is such that, for example, when the grindstone correction device 68 corrects the grindstone 66 (tool T) by truing, dressing, or the like, or the grindstone 16 (tool T) is set in the header portion. Includes the number of times information indicating the number of times since the time of replacement is set to zero (corresponding to the number of times of processing). Here, the number of times information is output by a counter provided in the control device 12. Further, the data format includes, for example, time information of the time when the state data is acquired (or the time when the state data is output), time-series state data, and process information.

The state data acquisition unit 71a acquires the state data detected by the first detector 13 via the control device 12 during the machining of the workpiece W. That is, the state data acquisition unit 71a acquires a plurality of state data corresponding to the number of minutes of the first detector 13 based on the above-mentioned data format.

The state data dividing unit 71b divides the state data acquired by the state data acquisition unit 71a from the control device 12 in time series (continuously) for each grinding process which is a processing process. That is, the state data dividing unit 71b obtains the state data continuously acquired by the state data acquisition unit 71a based on the process information associated with the state data according to the data format, for example, in the rough grinding process and the fine grinding process. , Divide into each micro-grinding process and spark-out process.

Here, as shown in FIG. 5, when the grinding machine 60 grinds the workpiece W, particularly in the rough grinding process, the rough grinding process is performed under the control of the control device 12, even though the rough grinding process is performed. In reality, there is a period during which the grindstone 66 is approaching the workpiece W, that is, a period during which the grindstone is not substantially rough-ground. Therefore, in the rough grinding process, the state data dividing unit 71b estimates, for example, whether or not the rough grinding process is substantially based on the position of the grindstone base 65 (grindstone wheel 66) included in the state data. To do.

Alternatively, in the rough grinding process, the state data dividing unit 71b changes to, for example, the drive current data of the motor 66a included in the state data, that is, the drive current data is changed when the grindstone 66 and the workpiece W come into contact with each other. Based on the increase, the period of Kuken is determined and it is estimated whether or not it is a rough grinding process. In addition, the state data dividing unit 71b divides the state data based on the process information after acquiring the state data with the progress of all the grinding processes, instead of sequentially dividing the state data for each grinding process. It is also possible.

Further, the state data dividing unit 71b modifies, for example, the grindstone 16 (tool T) or grindstone 16 (tool T) for each grinding process based on the number of times information included in the data format. It is also possible to divide the state data when the number of times after the exchange reaches a predetermined number of times. As a result, it is possible to easily grasp the correction time or replacement time of the grindstone 16 (tool T).

The feature amount extraction unit 71c extracts the feature amount from the state data of each grinding process (rough grinding process, fine grinding process, fine grinding process, spark-out process, etc.) divided by the state data dividing unit 71b. In this example, the feature amount extraction unit 71c uses the basic statistical amount obtained by performing statistical calculation on each state data divided for each grinding process as the feature amount. Here, as the basic statistics, for example, the maximum value, the minimum value, the average value, the dispersion value, the standard deviation, the intermediate value, the skewness, the kurtosis, etc. in the state data of one grinding process can be exemplified.

Further, as the feature quantity, a basic statistic for the data obtained by differentiating the state data and a basic statistic for the data obtained by frequency analysis of the state data can be included. In addition to extracting a plurality of features in each grinding process, the feature extraction unit 71c can also extract a plurality of features in the total state data acquired with the progress of all the grinding processes. It is possible.

The feature amount extraction unit 71c performs machine learning to extract a feature amount having a large contribution (contribution rate) to the quality of the work W, that is, the quality-related data, from among the plurality of feature amounts. The machine learning performed by the feature amount extraction unit 71c is performed by using the variable selection method. Examples of the variable selection method include linear regression, ridge regression, lasso regression, principal component analysis, elastic net, and random forest. By using such a variable selection method, it is possible to grasp the degree of contribution (contribution rate) for a plurality of features. Thereby, for example, the feature amount having a high contribution degree (contribution rate) can be preferentially extracted (selected). The machine learning performed by the feature amount extraction unit 71c is not limited to the variable selection method using regression, and other variable selection methods can be used.

The quality-related data acquisition unit 71d acquires the quality-related data detected by the second detector 14 and / and the inspection device 2. As for the quality-related data, the second detector 14 or the inspection device 2 outputs the quality-related data via the control device 12, so that the process information can be linked in the same manner as the state data. ..

The training data set storage unit 72 stores the training data set acquired by the training data set acquisition unit 71. Specifically, the training data set storage unit 72 associates the feature amount of the state data extracted by the feature amount extraction unit 71c with the quality-related data acquired by the quality-related data acquisition unit 71d, and performs a grinding process ( Memorize each processing process). That is, the training data set storage unit 72 stores the training data set for each of the grinding processes (machining processes).

The model generation unit 73 performs machine learning using the training data set stored in the training data set storage unit 72. Specifically, the model generation unit 73 uses the feature amount with a large contribution (contribution rate) extracted by the feature amount extraction unit 71c as an explanatory variable for each grinding process (machining process), and sets the quality-related data as an objective variable. Machine learning is performed. Then, the model generation unit 73 learns to predict the quality of the workpiece W, the wear state of the grindstone 66 that affects the quality of the workpiece W, and the presence or absence of an abnormality in the grinding machine 60 (processing machine main body 11). Generate a finished model.

The prediction calculation device 30 predicts the quality of the workpiece W that has been ground (machined) based on the state data during machining and the quality-related data on the corresponding grinding machine 60 (machining machine 10). The prediction calculation device 30 includes a model storage unit 81, a prediction data acquisition unit 82, a quality prediction unit 83, and an output unit 84.

The model storage unit 81 stores a plurality of learned models generated by the model generation unit 73 for each grinding process (machining process). The prediction data acquisition unit 82 acquires prediction data during machining in the grinding process (machining process) to be predicted. The prediction data acquisition unit 82 includes a state data acquisition unit 82a, a state data division unit 82b, and a feature amount extraction unit 82c.

The state data acquisition unit 82a acquires the state data detected by the first detector 13 and output via the control device 12 during processing in the grinding process (machining process) to be predicted. Specifically, the state data acquisition unit 82a acquires the state data associated with the process information according to the above-mentioned data format for each of the rough grinding process, the fine grinding process, the fine grinding process, and the spark-out process to be predicted.

The state data dividing unit 82b divides the state data acquired by the state data acquiring unit 82a into each grinding process (machining process). Specifically, the state data dividing unit 82b divides the state data for each grinding process (machining process) based on the process information associated with the state data.

The feature amount extraction unit 82c extracts the feature amount of the state data divided for each grinding process (machining process) by the state data dividing unit 82b. Here, as the feature amount, various basic statistics and the like in the state data are used. The feature amount is the same as the feature amount extracted by the feature amount extraction unit 71c of the training data set acquisition unit 71.

Here, the state data acquisition unit 82a, the state data division unit 82b, and the feature amount extraction unit 82c perform the same processing as the state data acquisition unit 71a, the state data division unit 71b, and the feature amount extraction unit 71c of the training data set acquisition unit 71. I do. In this example, the state data acquisition unit 82a, the state data division unit 82b, and the feature amount extraction unit 82c are the state data acquisition unit 71a, the state data division unit 71b, and the feature amount extraction unit 71c of the training data set acquisition unit 71. It was explained as a separate element from. However, it is also possible to use the elements 71a, 71b, 71c of the training data set acquisition unit 71 together with the elements 82a, 82b, 82c of the prediction data acquisition unit 82. That is, the functions of the elements 71a, 71b, and 71c in the learning processing device 20 are also used as a part of the functions of the prediction calculation device 30.

The quality prediction unit 83 selects a trained model corresponding to the grinding process (machining process) to be predicted from among the plurality of trained models stored in the model storage unit 81, and the selected trained model and prediction data. Based on the data acquired by the acquisition unit 82, the quality of the workpiece W or the presence or absence of an abnormality or failure of the grindstone 66 (tool T) and the grinding machine 60 (processing machine main body 11) that affect the quality. Predict.

The output unit 84 outputs information on the processing quality in the grinding process (machining process) of the prediction target predicted by the quality prediction unit 83 to the display devices 40 and 50. Here, the output unit 84 is a plurality of feature quantities extracted by the feature quantity extraction unit 82c (feature quantity extraction unit 71c), such as the quality of the workpiece W, for example, the surface texture on the ground surface, and the grindstone 66 (tool T). ), And various information regarding the operation of the grinding machine 60 (processing machine main body 11) can be output.

The display devices 40 and 50 display the information output from the output unit 84 of the prediction calculation device 30. Specifically, as shown in FIG. 6, the display devices 40 and 50 display information on the processing quality of the workpiece W, for example, the surface roughness which is the surface texture of the ground surface in chronological order. In this case, the display devices 40 and 50 supply the feature amount extracted by the feature amount extraction unit 82c (feature amount extraction unit 71c), for example, to the motor 66a in the fine grinding step as shown by a star in FIG. The minimum value of the drive current value to be generated is displayed in chronological order in correspondence with the surface roughness.

As a result, the operator or the control device 12 of the grinding machine 60 (processing machine main body 11) has the surface roughness (indicated by a solid line in FIG. 6) and the minimum value of the drive current value (indicated by a broken line in FIG. 6). Correlation can be grasped. Therefore, the operator or the control device 12 can grasp the correlated surface roughness by, for example, managing the minimum value of the drive current value that can be continuously detected by the first detector 13. ..

When the minimum value of the drive current value becomes the preset drive current value I, the allowable upper limit value U of the surface roughness is obtained. In other words, the wear state of the grindstone 66 (tool T) deteriorates. ing. Therefore, when the drive current value I is reached, the operator or the control device 12 operates the grindstone correction device 68 to perform, for example, growing the grindstone 66 or replace the grindstone 66. To do.

Further, by grasping the correlation between the surface roughness and the minimum value of the drive current value, for example, in order to reduce the frequency of growing, it is necessary to improve so that the change of the minimum value of the drive current value becomes gradual. Is possible. In this case, as shown in FIG. 8, the display devices 40 and 50 can display the contribution degree (contribution ratio) of the extracted plurality of feature quantities as, for example, a histogram. As a result, the operator can moderate the change of the feature amount A (for example, the minimum value of the drive current value) having the highest contribution (contribution rate), so that the feature amount having the next highest contribution (contribution rate) can be moderated. Focusing on B (for example, the position (feed amount)), the feed amount of the grindstone 66 (tool T) can be reduced to advance the improvement.

In addition, the display devices 40 and 50 can also display, for example, feature quantities for all the state data detected in the grinding process (machining process) in order for the operator to grasp the improvement points. In this case, as described above, since the process information is associated with the state data, all the feature quantities can be divided and displayed for each grinding process (machining process). This makes it easier for the display devices 40 and 50 to display the feature amount, and also makes it easier for the operator to find the feature amount useful for improvement.

(6. Another example of the learning processing device 20)
In this example described above, the model generation unit 73 generates one trained model for each grinding process (machining process) using the training data set stored in the training data set storage unit 72. By the way, the model generation unit 73 is not limited to performing machine learning using one algorithm, and can perform machine learning using a plurality of different algorithms. Further, the model generation unit 73 is not limited to using one feature amount or a feature amount of a specific combination as an explanatory variable, and performs machine learning using all the feature amounts or a plurality of different combinations of feature amounts as explanatory variables. It is possible.

In these cases, the model generation unit 73 generates a plurality of learned models for each grinding process (machining process). Therefore, in this alternative example, as shown in FIG. 9, the learning processing device 20 includes the model evaluation unit 74, and evaluates the quality of the prediction accuracy of the plurality of trained models generated by the model generation unit 73, and most of all. Identify trained models with high prediction accuracy. Then, the prediction calculation device 30 stores the learned model with the highest prediction accuracy in the model storage unit 81. Hereinafter, this alternative example will be specifically described.

In this alternative example, the training data set storage unit 72 stores a plurality of feature quantities and quality-related data extracted by the feature quantity extraction unit 71c as a training data set. The model generation unit 73 performs machine learning using a plurality of feature quantities, a combination of a plurality of feature quantities, and quality-related data stored in the training data set storage unit 72, and a plurality of features per grinding process (machining process). Generate a trained model. In this case, the model generation unit 73 stores a part of the training data sets stored in the training data set storage unit 72, for example, 80 training data sets when 100 training data sets are stored. Use to perform machine learning. Then, the model generation unit 73 supplies the generated plurality of trained models to the model evaluation unit 74.

The model evaluation unit 74 evaluates the accuracy of the predicted value for each trained model acquired from the model generation unit 73. Specifically, the model evaluation unit 74 is the other part of the training data set stored in the training data set storage unit 72, for example, 20 training data sets when 100 training data sets are stored. Is used.

In this case, the model evaluation unit 74 is an explanatory variable of the training data set, inputs the same kind of features as when the trained model was generated into each trained model, and obtains a predicted value and a purpose of the training data set. Compare with variables, i.e. measured values. In this comparison, the model evaluation unit 74 determines, for example, the trained model having the smallest difference (difference, etc.) between the predicted value and the measured value as the trained model with the highest prediction accuracy. Here, by determining the trained model with the highest prediction accuracy, the feature quantity used for machine learning by the trained model is determined as the feature quantity having a high contribution (contribution rate) to the quality-related data.

Then, the model storage unit 81 of the prediction calculation device 30 stores the learned model with the highest prediction accuracy, so that the quality prediction unit 83 can use the learned model with the highest prediction accuracy. As a result, the quality prediction unit 83 can more accurately predict the quality of the workpiece W (other factors that affect the quality), and the output unit 84 displays extremely useful information on the display devices 40 and 50. can do.

(7. Others)
In this example and another example described above, the learning processing device 20 includes a feature amount extraction unit 71c and a model generation unit 73. By the way, the machine learning method (algorithm that realizes the variable selection method) used by the feature extraction unit 71c in machine learning may be the same as the machine learning method (algorithm) used by the model generation unit 73 in machine learning. In this case, for example, the feature amount extraction unit 71c is omitted, and the model generation unit 73 extracts the feature amount from the state data division unit 71b using a predetermined algorithm for the state data stored in the training data set storage unit 72. be able to.

Then, the model generation unit 73 can generate a trained model by performing machine learning according to the same algorithm using a training data set in which the extracted features are used as explanatory variables and quality-related data is used as the objective variable. it can. That is, in this case, the model generation unit 73 simultaneously extracts the feature amount of the state data and generates the trained model. In this case as well, the learning processing device 20 can generate a trained model in the same manner as in this example and another example described above.

(8. Effect)
According to the quality prediction system 1, the state data detected in time series (continuously) can be divided into each grinding process (machining process) based on the process information. Then, the quality prediction system 1 extracts a feature amount from the state data divided for each grinding process (machining process), and generates a learned model for each grinding process (machining process) using the extracted feature amount. Can be done. In this way, by extracting the features for each grinding process (machining process) and generating a trained model, it is possible to objectively and accurately predict the quality of the workpiece W with versatility. ..

Further, various information related to the processing quality of the workpiece W and a feature amount having a high contribution (contribution rate) to the quality can be displayed on the display devices 40 and 50. Therefore, the operator can easily grasp the improvement points in the grinding (machining) of the workpiece W using the grinding machine 60 (machining machine main body 11) by checking the display devices 40 and 50. ..

1 ... Quality prediction system, 2 ... Inspection device, 10 ... Processing machine, 11 ... Processing machine body, 12 ... Control device, 13 ... First detector, 14 ... Second detector, 15 ... Interface, 20 ... Learning processing device , 21 ... Processor, 22 ... Storage device, 23 ... Interface, 30 ... Prediction calculation device, 31 ... Processor, 32 ... Storage device, 33 ... Interface, 40 ... Common display device, 50 ... Individual display device, 60 ... Grinding machine, 61 ... bed, 62 ... headstock, 62a ... motor, 63 ... tailstock, 64 ... traverse base, 64a ... motor, 65 ... grinding table, 65a ... motor, 66 ... grinding wheel, 66a ... motor, 67 ... fixed size Device, 68 ... Grinding wheel correction device, 69 ... Coolant device, 71 ... Training data set acquisition unit, 71a ... State data acquisition unit, 71b ... State data division unit, 71c ... Feature quantity extraction unit, 71d ... Quality-related data acquisition unit , 72 ... Training data set storage unit, 73 ... Model generation unit, 74 ... Model evaluation unit, 81 ... Model storage unit, 82 ... Prediction data acquisition unit, 82a ... State data acquisition unit, 82b ... State data division unit, 82c ... feature amount extraction unit, 83 ... quality prediction unit, 84 ... output unit, U ... allowable upper limit value, I ... drive current value, T ... tool, W ... workpiece

Claims (15)

A processing machine that performs machining of a workpiece through multiple processing processes under the control of a control device,
A detector that detects observable state data in the machining by the processing machine and outputs the detected state data to the control device.
The feature amount of the state data detected by the detector and acquired through the control device in each machining process is used as an explanatory variable, and the quality-related data related to the machining quality of the workpiece by the machining is used as an objective variable. A trained model storage unit that stores a trained model generated by performing machine learning using a training data set including the explanatory variables and the objective variable for each processing process.
A quality prediction unit that predicts the quality of the workpiece using the trained model and the state data,
A quality prediction system equipped with. The state data is added with process information for identifying the processing process by the control device.
The quality prediction system according to claim 1, further comprising a state data dividing unit that divides the state data for each processing process based on the process information. The state data division unit
The quality prediction system according to claim 2, wherein the state data is divided into each processing process based on the process information with respect to the state data acquired with the progress of all the processing processes. A feature amount extraction unit for extracting a plurality of feature amounts from the state data acquired in the processing step is provided.
The trained model storage unit
Claim 1-3 for storing the learned model generated by using at least one of the plurality of feature quantities extracted by the feature quantity extraction unit as the explanatory variable for each processing step. The quality prediction system according to any one of the items. The feature amount extraction unit
The quality prediction system according to claim 4, wherein the feature amount having a large contribution to the quality-related data is extracted from a plurality of the feature amounts in the state data by machine learning using a predetermined variable selection method. The feature amount extraction unit
The quality prediction system according to claim 4 or 5, wherein a basic statistic obtained by performing a statistical calculation on the state data is extracted as the feature amount. The trained model is obtained by performing machine learning using the explanatory variable and the training data set including the objective variable, with the feature amount in each processing step as the explanatory variable and the quality-related data as the objective variable. The quality prediction system according to any one of claims 1 to 6, further comprising a trained model generator to be generated. Of the plurality of training data sets, the plurality of trained models generated by machine learning using different algorithms depending on the trained model generation unit using a part of the plurality of the training data sets. Learning to evaluate a plurality of the trained models based on the difference between the predicted value predicted when the explanatory variables are given to the plurality of trained models and the actually measured value which is the objective variable using another part. Equipped with a completed model evaluation department
The trained model storage unit
The quality prediction system according to claim 7, wherein the trained model evaluation unit stores the trained model having the smallest difference. The quality prediction system according to any one of claims 1 to 8, further comprising a display device for displaying the feature amount having a large contribution to the quality of the predicted work. The display device is
The quality prediction system according to claim 9, which displays information on modification or replacement of a tool provided on the processing machine to perform the machining on the workpiece. The processing machine is a grinding machine that grinds the workpiece using a grindstone.
The quality prediction system according to any one of claims 1-10, wherein the state data can be divided for each grinding process of the workpiece by the grinding machine. The quality prediction system according to claim 11, wherein the state data is divided based on the number of times the grinding process of the workpiece is performed for each grinding process after the grindstone is modified. The quality prediction system according to claim 11, wherein the state data is divided based on the number of times the grinding process of the workpiece is performed for each grinding process after the grindstone is replaced. The quality prediction system according to any one of claims 11 to 13, wherein the grinding step includes at least a rough grinding step, a fine grinding step, a fine grinding step, and a spark-out step. The state data in the rough grinding process is
In addition to the process information for identifying the rough grinding process added to the state data by the control device, the rough grinding process includes changes in the drive current data supplied to the motor for rotationally driving the grindstone. The quality prediction system according to claim 14, wherein the period during which the grindstone is brought close to the workpiece and the period during which the grindstone is not coarsely grinding the workpiece is determined and divided. ..

Images