JP7697235B2 - Cutting process monitoring system - Google Patents

Description

The present invention relates to a cutting processing monitoring system that is used, for example, to prevent problems in cutting processing.

In lathe machining, it is generally necessary to quickly detect wear or chipping of the tool cutting edge and take action such as replacing the tool. For this reason, a method has been known in the past in which the tool cutting edge is photographed with a camera and the condition of the cutting edge is examined from the image. For example, in Patent Document 1, the tool cutting edge is photographed with a camera in a standby state before and after cutting, and the amount of wear is determined from the binary image. In Patent Document 2, the tool cutting edge during cutting is captured with an infrared camera and the image is learned by a neural network to determine abnormalities.

Japanese Unexamined Patent Publication No. 6-114694 Japanese Patent Application Publication No. 11-267949

In the method of Patent Document 1, the tool cutting edge is photographed with a camera while the tool is in a standby state (stationary state) other than during cutting, so there is a problem in that it is not possible to immediately respond to problems that occur during cutting. In the method of Patent Document 2, the amount of cutting edge wear is indirectly inferred from the temperature distribution, so quantitative evaluation of the amount of wear is difficult, and a lot of learning is required to improve the accuracy of the judgment.

One of the objectives of the present invention is to provide a cutting process monitoring system that can analyze the condition of at least one of the tool cutting edge and the machined surface of the workpiece in real time with high accuracy during cutting.

One embodiment of the cutting process monitoring system of the present invention comprises a camera that is positioned with respect to a turning tool so that the tool cutting edge is in the same position within the imaging field of view, and is capable of continuously photographing the tool cutting edge and the machined surface of a workpiece during turning, a classification unit that classifies a plurality of images taken by the camera into one of a plurality of patterns based on information contained in each of the images, and an analysis unit that analyzes the condition of at least one of the tool cutting edge and the machined surface of the workpiece from the images classified into a predetermined pattern from the plurality of patterns, wherein the plurality of images include images that show the machined surface and images that do not show the machined surface, and the plurality of patterns includes a pattern into which images that show the machined surface of the workpiece are classified.

In the present invention, the tool cutting edge and the machined surface of the workpiece during turning (hereinafter sometimes simply referred to as the tool cutting edge, etc.) are continuously photographed, i.e., in real time, by a camera. Note that "continuously photographing" in the present invention refers to taking images at least 10 frames per second. In addition, to position and fix the camera relative to the turning tool so that the tool cutting edge is in the same position within the imaging field of view, for example, the camera may be attached via an arm or the like to a tool rest that supports the turning tool.
The classification unit then classifies each image into one of a plurality of patterns by, for example, comparing the obtained image information with the learning results of the storage unit. The analysis unit analyzes the state of at least one of the tool cutting edge and the machined surface of the workpiece from each classified image.

Specifically, the images continuously captured by the camera include images in which the cutting edge of the tool is hidden by cutting chips, and images in which the cutting edge has a large amount of adhesion and the wear state cannot be seen. Therefore, from the many images continuously captured, images in which, for example, the outline of the cutting edge (edge line of the cutting edge), adhesion to the rake face, or the machined surface of the workpiece is visible are selected (classified), and image processing is used to analyze the wear and damage state of the cutting edge, the state of adhesion, and the machining state of the machined surface. Deep learning can be used for the above selection, and since the judgment is relatively simple, a high level of judgment accuracy can be obtained with a small amount of learning.

As described above, according to the present invention, the state of at least one of the tool cutting edge and the machined surface of the workpiece during cutting can be analyzed in real time with high accuracy. In other words, users can grasp the state of the tool cutting edge and other parts during cutting in real time. This enhances the practicality of the system as, for example, an anomaly detection system or an in-process (during cutting) machining dimensional accuracy measurement system.

In the above cutting process monitoring system, it is preferable that the classification unit classifies the image in which a cutting edge ridge having a predetermined cutting edge length or more is captured into a first pattern among the multiple patterns, and the analysis unit analyzes the damage state of the tool cutting edge based on the shape of the cutting edge ridge of the image classified into the first pattern.

In this case, the analysis unit analyzes the images classified into the first pattern by the classification unit, and the state and amount of damage to the tool cutting edge, such as wear and chipping, can be measured. Specifically, for example, the cutting edge ridge (edge) is detected by image processing such as a canny filter, and the change (increase) in the number of pixels of the edge is measured to analyze the state of damage to the cutting edge.

In the above cutting process monitoring system, it is preferable that the classification unit classifies the images showing welding on the tool cutting edge into a second pattern among the multiple patterns, and the analysis unit analyzes the damage state of the tool cutting edge based on a change in the frequency of appearance of the images classified into the second pattern.

In general, when damage occurs to the cutting edge of a tool, welding tends to occur due to the damage. That is, welding to the cutting edge of the tool and its peeling (falling off) are randomly repeated during cutting, but when damage occurs to the cutting edge, welding often adheres strongly to the damaged part and is difficult to remove, so that the frequency of welding in images increases. That is, for example, when the continuous shooting speed (shooting interval) of a camera is constant, the number of images classified into the second pattern per unit time increases.
According to the above-mentioned configuration of the present invention, even if the tool cutting edge is hidden by welding and cannot be seen or is difficult to see, the damage to the cutting edge can be indirectly detected. That is, the analysis unit can infer the state of damage to the tool cutting edge based on the change in the appearance frequency of the images classified into the second pattern by the classification unit.

In the above cutting process monitoring system, it is preferable that the classification unit classifies the image showing the machined surface of the workpiece into a third pattern from among the multiple patterns, and the analysis unit analyzes the machining state of the machined surface from the image classified into the third pattern.

In this case, the analysis unit analyzes the images classified into the third pattern by the classification unit, thereby making it possible to determine the machining dimensional accuracy, the machining surface quality, the occurrence of burrs, the discharge state of chips, etc. of the machined surface of the workpiece. That is, the "machined state of the machined surface of the workpiece" in the present invention refers to any one or more of the machining dimensional accuracy, the machining surface quality, the occurrence of burrs, and the discharge state of chips of the machined surface of the workpiece.
In particular, with the above-described configuration of the present invention, the user can grasp in real time the machining accuracy and quality of the machined surface immediately after cutting, and can quickly respond, for example, if a problem occurs in the machining state.

In the above cutting processing monitoring system, it is preferable that the classification unit performs image classification using deep learning.

In this case, the accuracy of classification judgment by the classification unit is steadily improved through deep learning.

According to one embodiment of the cutting process monitoring system of the present invention, the condition of at least one of the tool cutting edge and the machined surface of the workpiece during cutting can be analyzed in real time with high accuracy.

FIG. 1 is a side view showing an example of a turning device according to the present embodiment. FIG. 2 is a functional block diagram showing an example of the cutting process monitoring system of the present embodiment. FIG. 3 shows an example of an image classified into the first pattern by the classification unit. FIG. 4 shows an example of an image classified into the second pattern by the classification unit. FIG. 5 shows an example of an image classified into the third pattern by the classification unit. FIG. 6 shows an example of analysis (image processing) of an image classified into the first pattern. FIG. 7 shows an example of analysis (image processing) of an image classified into the first pattern. FIG. 8 is a diagram showing an example of a change in appearance frequency of images classified into the second pattern. FIG. 9 is a flowchart showing an example of the process of the cutting monitoring device of this embodiment.

A cutting process monitoring system 10 according to one embodiment of the present invention will be described with reference to the drawings. The cutting process monitoring system 10 according to this embodiment includes a turning device 20 and a cutting process monitoring device 30.

As shown in Figures 1 and 2, the turning device 20 is a machine tool such as an NC lathe. The turning device 20 is a device that turns a workpiece W, such as a metal workpiece. In other words, the cutting process monitoring system 10 is used for turning (lathe processing) using a machine tool such as an NC lathe. Turning refers to cutting using a turning tool such as a cutting tool.

The turning device 20 includes a turning tool 21, a tool rest 22, a camera 23, an arm 24, and a chuck (not shown). In other words, the cutting process monitoring system 10 includes the camera 23.

As shown in FIG. 1, the turning tool 21 is, for example, a replaceable cutting tool in which a cutting insert 26 is removably attached to the tip of a holder 25. That is, the turning tool 21 has a holder 25 and a cutting insert 26. However, the turning tool 21 is not limited to this configuration, and may be, for example, a solid-type cutting tool in which the cutting edge is integrally formed with the holder.

The holder 25 is columnar and extends in one direction. In the example shown in FIG. 1, the direction in which the holder 25 extends is inclined at an angle θ with respect to the horizontal plane H. The angle θ is, for example, 30°. The rear end of the holder 25 is supported by the tool rest 22. The turning tool 21 and the tool rest 22 are fixed together.

The cutting insert 26 is made of, for example, a cemented carbide alloy, etc. The cutting insert 26 is, for example, in the shape of a polygonal plate such as a quadrangular plate, or in the shape of a disk, etc. In this embodiment, the cutting insert 26 is, for example, in the shape of a rhombic plate.
3, the cutting insert 26 has a rake face 26a, a clearance face (not shown), and a cutting edge 26b disposed on a ridge between the rake face 26a and the clearance face. In this embodiment, at least a portion of the cutting insert 26, specifically, a portion including at least the rake face 26a and the cutting edge 26b, may be simply referred to as a "tool cutting edge" or a "cutting edge".

The rake surface 26a is disposed on at least one of a pair of plate surfaces facing the plate thickness direction of the cutting insert 26. As shown in FIG. 1, the rake surface 26a is disposed on one plate surface facing the upper side of the cutting insert 26, i.e., the upper surface. In the example shown in FIG. 1, the rake surface 26a is inclined with respect to the horizontal plane H at approximately the same angle as the angle θ at which the holder 25 is inclined with respect to the horizontal plane H.

As shown in FIG. 5, the rake face 26a of this embodiment has a feature point 26c. The feature point 26c constitutes part of the pattern provided on the rake face 26a. The feature point 26c is provided on the rake face 26a for the purposes of, for example, smooth discharge of the chips C, improving the external design, and improving the distinguishability. In this embodiment, the feature point 26c is, for example, a protruding chip breaker.

As shown in FIG. 1, the cutting edge 26b is disposed at the tip of the turning tool 21. The cutting edge 26b is brought into contact with the workpiece W rotating around the central axis O, whereby the workpiece W is turned to form a machined surface Wa as shown in FIG. 5.

In this embodiment, as shown in FIG. 3, when the upper surface, i.e., the rake surface 26a, of the cutting insert 26 is viewed from the front, the cutting edge 26b is substantially V-shaped. Specifically, this cutting edge 26b has one corner edge that extends in a curved line and a pair of straight edges that are connected to both ends of this corner edge and extend in a straight line.

As shown in FIG. 1, the tool rest 22 supports the turning tool 21. The tool rest 22 moves the turning tool 21 relative to the workpiece W at least in the direction in which the horizontal plane extends, i.e., along the surface direction of the horizontal plane. The tool rest 22 may also move the turning tool 21 in the vertical direction relative to the workpiece W.

The camera 23 is disposed above the turning tool 21 and the workpiece W. The camera 23 is, for example, a global shutter type camera. The camera 23 continuously captures the tool cutting edge. Note that in this embodiment, "continuously captures images" refers to capturing images at least 10 frames per second. Specifically, in this embodiment, the camera 23 continuously captures the tool cutting edge at a speed of, for example, 30 frames per second or more. During cutting, the camera 23 captures the machined surface Wa of the workpiece W as well as the tool cutting edge.

In this embodiment, the camera 23 photographs the tool cutting edge from a direction perpendicular to the cutting surface 26a. Specifically, the camera 23 photographs the tool cutting edge from a direction inclined at a predetermined angle from the vertical direction. This predetermined angle is approximately the same as the angle at which the cutting surface 26a is inclined with respect to the horizontal plane H (corresponding to angle θ). The distance between the camera 23 and the tool cutting edge is, for example, 300 mm or more.

The camera 23 is fixed to the tool rest 22 via the arm 24. Therefore, the camera 23 can move in accordance with the movement of the tool rest 22. The camera 23 is positioned relative to the turning tool 21 so that the tool tip is in the same position within the imaging field of view, as shown in, for example, Figures 3 to 5.

As shown in FIG. 1, the arm 24 connects the camera 23 and the tool post 22. The arm 24 has a tip connected to the camera 23 and a rear end connected to the tool post 22. The arm 24 has at least two shafts 24a, at least one joint 24b connecting the ends of adjacent shafts 24a, and a fixed base 24c that fixes the rear end of the arm 24 to the tool post 22.

The shaft portion 24a is, for example, a shaft or a pipe.
The joint 24b can be switched between a lock mode in which the ends of the shaft 24a are fixed so as not to rotate and a free mode in which the ends of the shaft 24a are rotatably connected to each other by operating a knob (not shown). The provision of the joint 24b allows the arm 24 to be deformed. By changing the shape of the arm 24, the position of the camera 23 can be adjusted so as to face the cutting edge of the tool, regardless of the tool shape or type of the turning tool 21, for example. During cutting, the joint 24b is in the lock mode.

The fixed base 24c fixes the arm 24 to the tool rest 22, for example, by magnetic force. Therefore, the mounting position of the arm 24 can be adjusted with respect to the tool rest 22. That is, in this embodiment, the position of the camera 23 can be adjusted so that it faces the tool tip by adjusting the mounting position of the fixed base 24c to the tool rest 22.

Although not specifically shown, the chuck removably holds the workpiece W. The chuck rotates the workpiece W around its central axis O.

Although not specifically shown, the turning device 20 may also be equipped with a backlight for highlighting the contours of the cutting edge 26b and the machining surface Wa, and a frontlight for illuminating the rake face 26a. This allows the shutter speed of the camera 23 to be increased.

As shown in FIG. 2, the cutting process monitoring device 30 includes an image acquisition unit 31, a memory unit 32, a classification unit 33, an analysis unit 34, and a display unit 35. In other words, the cutting process monitoring system 10 includes a classification unit 33 and an analysis unit 34.

The image acquisition unit 31 acquires the image captured by the camera 23, i.e., image data. The image acquisition unit 31 stores the acquired image information in the storage unit 32. The cutting process monitoring device 30 may, for example, associate the acquired image information with a tool ID, a device ID, a user ID, etc., and store the information in the storage unit 32.

The memory unit 32 stores various information used by the cutting process monitoring device 30. The memory unit 32 stores acquired image information. The memory unit 32 stores, for example, deep learning teacher data and learning results used by the classification unit 33 to classify images.

The classification unit 33 performs image classification using deep learning, for example, based on the learning results stored in the memory unit 32. That is, the classification unit 33 classifies multiple images captured by the camera 23 into one of multiple patterns based on the information captured in each image. In this embodiment, the multiple patterns include a first pattern (cutting edge line), a second pattern (welding), and a third pattern (machined surface of the workpiece).

Specifically, as shown in FIG. 3, the classification unit 33 classifies an image in which the cutting edge ridge of the cutting edge 26b having a predetermined cutting edge length or more is captured within the imaging field of view into a first pattern among multiple patterns. In the example shown in FIG. 3, the cutting edge 26b is not in contact with the workpiece W, and the entire cutting edge ridge is clearly captured within the imaging field of view.

The classification unit 33 also classifies an image in which the welded TA is reflected on the tool cutting edge into a second pattern among the multiple patterns, as shown in FIG. 4. In the example shown in FIG. 4, at least a portion of the rake face 26a and at least a portion of the cutting edge 26b are hidden by the welded TA attached to the tool cutting edge.

Further, the classification unit 33 classifies the image showing the machined surface Wa of the workpiece W into a third pattern among the multiple patterns, as shown in Fig. 5. In the example shown in Fig. 5, the surface and the contour of the machined surface Wa immediately after being cut by the cutting blade 26b are clearly shown.
3 to 5 are images of the cutting edge of a tool taken when a plate-shaped workpiece W was being turned as shown in Fig. 1. The workpiece W was made of stainless steel, the rotation speed of the workpiece W was 100 revolutions per minute, and the cutting was performed in a dry environment.

The classification of images by the classification unit 33 is performed, for example, by the following method.
First, a plurality of images are taken by the camera 23 during a test cutting process performed before the main cutting process of the workpiece W. The plurality of images thus obtained are classified and stored (teacher data) into a first pattern (cutting edge ridge) in which the cutting edge 26b is not in contact with the workpiece W and the cutting edge ridge is clearly visible, a second pattern (welding) in which the cutting edge 26b is not in contact with the workpiece W but the welded TA is on the rake face 26a, and a third pattern (machined surface of the workpiece) capturing the moment when the cutting edge 26b comes into contact with the workpiece W, and the classification method is learned by deep learning. That is, machine learning is performed to estimate the classification of the first to third patterns based on the teacher data in the memory unit 32, and the learning result is fed back to the memory unit 32, thereby improving the judgment accuracy of the image classification.
It should be noted that this classification is merely an example, and the appropriate classification method varies depending on the cutting conditions, the type and shape of the workpiece W, etc.
Furthermore, automatic image recognition software such as HALCON (manufactured by Links Corporation) can be used for the classification (sorting) of the image data described above.

The analysis unit 34 analyzes the state of at least one of the tool cutting edge and the machining surface Wa of the workpiece W from the images classified into a predetermined pattern from among the multiple patterns. For example, an image processing program or the like can be used as the analysis unit 34.

Specifically, the analysis unit 34 analyzes the damage state of the tool cutting edge based on the shape of the cutting edge ridge of the cutting edge 26b in the image classified into the first pattern (hereinafter, this may be referred to as the first analysis). In this embodiment, the analysis unit 34 detects the cutting edge ridge (edge) of the cutting edge 26b by image processing such as a canny filter, and measures the change (increase) in the number of pixels of the edge to analyze the damage state of the cutting edge.

6 and 7 show an example of image processing of the cutting edge ridge of the cutting edge 26b by the analysis unit 34. In a given cutting process, the image (original image) of FIG. 7 was taken later than the image of FIG. 6. Compared to FIG. 6, in FIG. 7, convex portions due to welding TA and concave portions due to chipping CP are formed on the cutting edge ridge, and the number of pixels of the edge, i.e., the contour length of the cutting edge ridge, is longer due to the unevenness. In other words, the damage state of the cutting edge can be estimated by comparing the number of pixels of the edge.
In this manner, the analysis unit 34 analyzes the state of damage to the cutting edge of the tool. Note that the above analysis is an example, and other analysis methods may be used to analyze the first pattern.

Furthermore, the analysis unit 34 analyzes the state of damage to the tool cutting edge based on the change in the appearance frequency of the images classified into the second pattern (hereinafter, may be referred to as the second analysis).
Fig. 8 shows an example of a change in the appearance frequency of images classified into the second pattern. The hatched areas in the graph of Fig. 8 indicate that the images have been classified into the second pattern (welding) by the classification unit 33.

Specifically, during cutting, adhesion to the cutting edge of the tool and its peeling (falling off) are repeated randomly, but when damage occurs to the cutting edge, adhesion often becomes strong on the damaged area and is difficult to remove, which increases the frequency of adhesion. From this, it is inferred that the cutting edge damage occurred near the symbol P on the horizontal axis (cutting time) in Figure 8.
In this manner, the analysis unit 34 analyzes the state of damage to the cutting edge of the tool. Note that the above analysis is an example, and other analysis methods may be used to analyze the second pattern.

The analysis unit 34 also analyzes the machining state of the machined surface Wa of the workpiece W from the images classified into the third pattern (hereinafter, this may be referred to as the third analysis). Note that in this embodiment, the "machined state of the machined surface Wa of the workpiece W" refers to one or more of the machining dimensional accuracy, machined surface quality, burr generation state, and chip C discharge state of the machined surface Wa of the workpiece W.

In Fig. 5, the analysis unit 34 judges the machining dimension accuracy by, for example, measuring the distance L between the feature point 26c on the rake face 26a and the machined surface Wa through image processing. That is, in this embodiment, even if the camera 23 shakes (shifts the origin position) due to vibration during cutting, the analysis unit 34 can measure the machining dimension with high accuracy. The analysis unit 34 also judges the machining surface quality by, for example, measuring the dimensions of the uneven shape appearing on the surface and contour of the machined surface Wa. The analysis unit 34 may also analyze, for example, the presence or absence of burrs on the machined surface Wa, the length and curl state of the chips C, and the discharge direction of the chips C.
In this manner, the analysis unit 34 analyzes the machined state of the machined surface Wa of the workpiece W. Note that the above analysis is an example, and other analysis methods may be used to analyze the third pattern.

The display unit 35 displays various information of the cutting processing monitoring system 10. For example, the display unit 35 displays the analysis results of the analysis unit 34 described above.

FIG. 9 is a flowchart showing an example of the process of the cutting process monitoring device 30.
As shown in FIG. 9, first, the image acquisition unit 31 acquires image information output from the camera 23 (step S1).
Next, the classification unit 33 classifies each image into a first to third pattern based on the learning results stored in the storage unit 32 (step S2). That is, the classification unit 33 classifies each image into one of the first to third patterns by comparing the acquired image information with the learning results.

Next, the analysis unit 34 analyzes the state of at least one of the tool cutting edge and the machined surface Wa of the workpiece W from the images classified into the first to third patterns (step S3). Specifically, the analysis unit 34 performs the first analysis as described above on the images classified into the first pattern (cutting edge ridge). The analysis unit 34 also performs the second analysis as described above on the images classified into the second pattern (welding). The analysis unit 34 also performs the third analysis as described above on the images classified into the third pattern (machined surface of the workpiece).
Next, the display unit 35 displays the analysis results output from the analysis unit 34 (step S4). After the process of step S4, the cutting process monitoring device 30 ends the process of the above example.

In the cutting process monitoring system 10 of the present embodiment described above, the tool cutting edge and the machined surface Wa of the workpiece W during turning (hereinafter, may be simply abbreviated as the tool cutting edge, etc.) are continuously photographed, i.e., in real time, by the camera 23. The camera 23 is positioned and fixed relative to the turning tool 21 so that the tool cutting edge is in the same position within the imaging field of view.
The classification unit 33 classifies each image into one of a plurality of patterns by comparing the obtained image information with the learning results in the storage unit 32. The analysis unit 34 analyzes the state of at least one of the tool cutting edge and the machined surface Wa of the workpiece W from each classified image.

Specifically, the images continuously captured by the camera 23 include images in which the tool cutting edge is hidden by chips C, and images in which a lot of welded TA is attached to the cutting edge, making the wear state invisible. Therefore, in this embodiment, from the many images continuously captured, images in which the outline (edge line) of the cutting edge 26b, welded TA to the rake face 26a, etc., or the machined surface Wa of the workpiece W are visible are selected (classified), and the wear and damage state of the cutting edge 26b, the state of welded TA, and the machined state of the machined surface Wa are analyzed by image processing. Deep learning can be used for the above selection, and since the judgment is relatively simple, high judgment accuracy can be obtained with a small amount of learning.

As described above, according to this embodiment, the state of at least one of the tool cutting edge and the machined surface Wa of the workpiece W during cutting can be analyzed in real time with high accuracy. In other words, the user can grasp the state of the tool cutting edge and the like during cutting in real time. This enhances the practicality of the system as, for example, an anomaly detection system or an in-process (during cutting) machining dimensional accuracy measurement system.

Furthermore, in this embodiment, the classification unit 33 classifies images in which a cutting edge ridge having a predetermined cutting edge length or more is captured within the imaging field of view into a first pattern among a plurality of patterns, and the analysis unit 34 analyzes the damage state of the tool cutting edge based on the shape of the cutting edge ridge of the image classified into the first pattern.
In this case, the image classified into the first pattern by the classification unit 33 is analyzed by the analysis unit 34, so that the state and amount of damage such as wear and chipping of the tool cutting edge can be measured.

In addition, in this embodiment, the classification unit 33 classifies images in which the welded TA is visible on the tool cutting edge into a second pattern among multiple patterns, and the analysis unit 34 analyzes the damage state of the tool cutting edge based on changes in the appearance frequency of images classified into the second pattern.
In general, when damage occurs to the cutting edge of a tool, there is a tendency for the damage to cause welded TA. That is, when damage occurs to the cutting edge, the welded TA adheres strongly to the damaged portion and becomes difficult to remove, so that the frequency with which the welded TA appears in images increases. That is, for example, when the continuous shooting speed (shooting interval) of the camera 23 is constant, the number of images classified into the second pattern per unit time increases.
According to this embodiment, even if the tool cutting edge is hidden by the welded TA and cannot be seen or is difficult to see, the damage to the cutting edge can be indirectly detected. That is, the analysis unit 34 can infer the state of damage to the tool cutting edge based on the change in the appearance frequency of the images classified into the second pattern by the classification unit 33.

In this embodiment, the classification unit 33 classifies an image showing the machined surface Wa of the workpiece W into a third pattern among multiple patterns, and the analysis unit 34 analyzes the machining state of the machined surface Wa from the image classified into the third pattern.
In this case, the analysis unit 34 analyzes the images classified into the third pattern by the classification unit 33, and the machining dimensional accuracy, machined surface quality, burr occurrence state, and chip C discharge state of the machined surface Wa of the workpiece W can be determined.
In particular, in this embodiment, the user can grasp in real time the machining accuracy and quality of the machined surface Wa immediately after cutting, and can quickly respond if a problem occurs in the machining state, for example.

Furthermore, in this embodiment, the classification unit 33 performs image classification using deep learning.
In this case, the accuracy of classification judgment by the classification unit 33 is stably improved by deep learning.

In this embodiment, the camera 23 captures an image of the tool cutting edge from a direction perpendicular to the rake face 26a, that is, from a direction directly facing the rake face 26a.
In this case, the camera 23 can be easily focused on the cutting edge of the tool, and the effects of the present embodiment described above can be obtained more stably.

In this embodiment, the distance between the camera 23 and the tool tip is 300 mm or more.
If the distance is 300 mm or more, for example, when replacing the cutting insert 26 with a new one or when indexing the cutting edge 26b (changing the corner used), that is, when replacing the cutting edge, the camera 23 is unlikely to get in the way of the work. Also, even if the chips C extend or scatter during cutting, they are unlikely to come into contact with the camera 23, so that effective images can be stably obtained.

The present invention is not limited to the above-described embodiment, and the configuration may be modified without departing from the spirit of the present invention, for example as described below.

In the above embodiment, an example was given in which the camera 23 was fixed to the tool post 22 via the arm 24 that can be switched between a locked mode and a free mode, but this is not limited to the above. That is, the camera 23 may be fixed to the tool post 22 via an undeformable rigid member or the like. The camera 23 may also be fixed directly to the tool post 22. Note that the camera 23 may be fixed to a member other than the tool post 22, such as the holder 25 (turning tool 21), as long as it is positioned relative to the turning tool 21 so that the tool cutting edge is in the same position within the imaging field of view.

In addition, the configurations described in the above embodiments and modifications may be combined without departing from the spirit of the present invention, and additions, omissions, substitutions, and other modifications of configurations are possible. Furthermore, the present invention is not limited to the above-described embodiments, but is limited only by the claims.

The cutting process monitoring system of the present invention can analyze the state of at least one of the tool cutting edge and the machined surface of the workpiece during cutting in real time with high accuracy. Therefore, it has industrial applicability.

10... Cutting processing monitoring system, 21... Turning tool, 23... Camera, 33... Classification unit, 34... Analysis unit, TA... Welding, W... Workpiece, Wa... Machining surface

Claims (5)

a camera that is positioned relative to the turning tool so that the tool tip is in the same position within the imaging field of view and that can continuously capture images of the tool tip and the machined surface of the workpiece during turning;
a classification unit that classifies a plurality of images captured by the camera into one of a plurality of patterns based on information captured in each of the images;
and an analysis unit that analyzes a state of at least one of the tool cutting edge and the machined surface of the workpiece from the image classified into a predetermined pattern from among the plurality of patterns,
The plurality of images include an image showing the processing surface and an image not showing the processing surface,
The plurality of patterns includes a pattern in which an image showing a machined surface of the workpiece is classified.
Cutting process monitoring system. The classification unit classifies the image in which a cutting edge ridgeline having a predetermined cutting edge length or more is captured into a first pattern among the plurality of patterns,
The analysis unit analyzes a damage state of the tool cutting edge based on a shape of the cutting edge ridge of the image classified into the first pattern.
The cutting process monitoring system according to claim 1 . The classification unit classifies the image in which welding is shown on the tool cutting edge into a second pattern among the plurality of patterns,
The analysis unit analyzes a damage state of the tool tip based on a change in an appearance frequency of the images classified into the second pattern.
3. The cutting processing monitoring system according to claim 1 or 2. The classification unit classifies the image showing the machined surface of the workpiece into a third pattern among the plurality of patterns,
The analysis unit analyzes a processing state of the processing surface from the images classified into the third pattern.
The cutting process monitoring system according to any one of claims 1 to 3. The classification unit performs image classification using deep learning.
The cutting process monitoring system according to any one of claims 1 to 4.

Images