Method and system for predicting tool wear amount by using on-machine measurement

Abstract

A method for predicting a tool wear amount by using on-machine measurement, related to the present invention, comprises the steps of: receiving an instruction to capture an image of a cutting tool mounted on a main shaft of a machine tool; issuing a command for the correct position of the main shaft of the machine tool; initializing the direction of a camera provided in the machine tool; checking the position of a clearance surface of the cutting tool; capturing a wear image of the clearance surface when the position of the clearance surface of the cutting tool is perpendicular to the direction of the camera; correcting the position of the camera and rotating the camera such that the direction of the camera faces the clearance surface if the position of the clearance surface of the cutting tool is not perpendicular to the direction of the camera; constructing a dataset for the image captured by the camera and performing deep learning; and predicting a wear amount through an image of a wear region according to the performance of the deep learning.

Description

Method and system for predicting tool wear using meteorological measurements

[0001] The present invention relates to a method and system for predicting tool wear using meteorological measurements.

[0002] Cutting tools continuously wear out while cutting workpieces, and when a certain amount of wear is exceeded, the tool life ends. According to ISO standards (ISO 3685:1993, ISO 8688-1:1989, ISO 8688-2:1989), the tool life is determined by a uniform flank wear width (VB) of 0.3 mm for both turning and milling tools. However, cutting tools vary in size and depending on the workpiece, chipping or breakage frequently occurs before the flank wear width reaches 0.3 mm, ending the tool life. In actual manufacturing settings, tool life standards vary depending on the working conditions, so technology for accurately measuring tool wear is crucial.

[0003] A common method for measuring tool wear involves removing the tool from its holder at predetermined intervals during the cutting process, photographing the wear with an optical measuring device such as a tool microscope, and then manually measuring the wear on the flank. This method requires the tool to be removed at each wear measurement interval, which is time-consuming. Furthermore, since wear is measured manually, the subjective judgment of the measurer can lead to deviations and, consequently, lower reliability.

[0004] * Prior art

[0005] Japanese Patent Publication No. 2022-014559 (published on January 20, 2022)

[0006] Japanese Patent Publication No. 2022-131366 (published on September 7, 2022)

[0007] Japanese Patent Publication No. 1997-192983 (published on July 29, 1997)

[0008] Korean Patent Publication No. 10-2022-0102700 (published on July 21, 2022)

[0009] Japanese Patent Publication No. 2021-070114 (published on May 6, 2021)

[0010] The purpose of the present invention is to provide a method and device capable of accurately measuring the condition of a tool in real time through meteorological measurements and predicting the amount of tool wear with high reliability using artificial intelligence technology.

[0011] The tasks of the present invention are not limited to those mentioned above, and other technical tasks can be clearly understood by those skilled in the art from the description below.

[0012] A method for predicting tool wear using meteorological measurements related to the present invention comprises the steps of: receiving an instruction to take an image of a cutting tool mounted on a spindle of a machine tool; instructing the spindle of the machine tool to a fixed position; initializing the direction of a camera installed in the machine tool; confirming the position of a clearance surface of the cutting tool; taking an image of a wear of the clearance surface if the position of the clearance surface of the cutting tool is in a direction perpendicular to the direction of the camera; correcting the position of the camera and rotating the direction of the camera so that it faces the clearance surface if the position of the clearance surface of the cutting tool is not in a direction perpendicular to the direction of the camera; constructing a dataset and performing deep learning on images captured by the camera; and predicting the wear amount through an image of the wear area according to the performance of the deep learning.

[0013] As an example according to the present invention, the step of receiving an instruction to take an image of a cutting tool mounted on a main shaft of the machine tool may be started automatically by directly receiving a control command from a user or by achieving a pre-stored condition.

[0014] As an example according to the present invention, the stored condition may include the elapsed time of a reference cutting time.

[0015] As an example according to the present invention, in the step of constructing a dataset and performing deep learning on an image captured by the camera, the deep learning may use a model including object detection and image segmentation for the captured image.

[0016] As an example according to the present invention, in the step of constructing a dataset and performing deep learning on an image captured by the camera, the deep learning may use a Mask R-CNN (Regional Convolutional Neural Network) model.

[0017] As an example according to the present invention, when the position of the clearance surface of the cutting tool is not perpendicular to the direction of the camera, the step of correcting the position of the camera and rotating the direction of the camera so that it faces the clearance surface may include the step of moving the bed on which the camera is installed; and the step of rotating the camera by a servo motor.

[0018] As an example according to the present invention, the step of moving the bed on which the camera is installed is performed by adjusting the radius (R) of the cutting tool, the focal length (lF) of the camera, and the length (l) of the camera. C ) may be moved in the horizontal direction (X) and vertical direction (Y) so that the sum of the two is maintained constant.

[0019] As an example according to the present invention, the method for predicting tool wear amount using the above meteorological measurement may further include a step of controlling an air blower to remove foreign substances from the surface of the cutting tool before taking an image using the camera.

[0020] As an example according to the present invention, the step of confirming the position of the clearance surface of the cutting tool can identify the position of the clearance surface of the cutting tool and derive the misalignment angle between the direction of the clearance surface and the optical axis of the camera.

[0021] The present invention can also be implemented as a computer-readable storage medium storing a program configured to execute a method for predicting tool wear amount using the above-described meteorological measurement.

[0022] A wear prediction system for a tool using weather measurement according to the present invention comprises: a camera arranged on one side of a bed of a machine tool and configured to acquire an image of a cutting tool mounted on a main shaft, and rotatably installed so as to change direction according to a position of a clearance surface of the cutting tool; a data processing unit configured to receive and process an image of the cutting tool acquired by the camera; a deep learning model configured to perform object detection and image segmentation on the image of the cutting tool; And a control unit configured to command the main axis of the machine tool to be in a fixed position according to an instruction for image capturing for the cutting tool, initialize the direction of the camera, check the position of the clearance surface of the cutting tool, and if the position of the clearance surface of the cutting tool is in a direction perpendicular to the direction of the camera, capture an image of the wear of the clearance surface, and if the position of the clearance surface of the cutting tool is not in a direction perpendicular to the direction of the camera, move the bed to correct the position of the camera and rotate the direction of the camera so that it faces the clearance surface, and perform deep learning according to the deep learning model to predict the amount of wear for the image of the captured wear area.

[0023] According to the tool prediction method and system using weather measurement related to the present invention, a camera system is installed on a machine tool table to automatically acquire tool wear images and these data are applied to deep learning training. After training is completed, it is possible to diagnose the tool condition and predict the amount of wear in real time when taking tool wear images. In addition, by replacing the conventional manual method of measuring wear amount with artificial intelligence technology, unmanned operation is possible and the reliability of tool wear measurement can be significantly improved. This method can be applied to various cutting tools with a single training session, and it can enable tool replacement at an appropriate time, thereby reducing tool costs and improving the processing quality of products in manufacturing sites.

[0024] FIG. 1 is an exemplary flowchart for explaining a method for predicting tool wear amount using meteorological measurements according to an example of the present invention.

[0025] Figure 2 is a plan view conceptually showing a state in which a camera is installed on a machine tool in relation to the present invention.

[0026] Figure 3 conceptually shows the alignment status of the camera and the measurement target location, which is the free surface, when taking a wear image of a rotating tool.

[0027] Figure 4 conceptually illustrates a correction algorithm for camera position coordinates according to the position of the tool clearance surface.

[0028] Figure 5 shows examples of the diagnosis results for wear status of captured images of tools using a CNN (Convolutional Neural Networks) model.

[0029] Figure 6 shows examples showing the diagnosis results of wear status for captured images of tools when using the Mask R-CNN (Regional Convolutional Neural Network) model.

[0030] Figure 7 shows examples showing the results of diagnosing tool conditions using the instance segmentation technique among the image segmentation algorithms of Mask R-CNN.

[0031] Figure 8 is an example showing the results of detecting tool wear areas on a pixel-by-pixel basis using the instance segmentation technique of Mask R-CNN.

[0032] Figures 9 and 10 are examples comparing the predicted tool wear amount using image segmentation deep learning for two different captured images with actual measured values.

[0033] Figure 11 is a diagram comparing the tool wear prediction results of 30 verification data not included in the learning data with actual measured values ​​in relation to the present invention.

[0034] Fig. 12 is a flowchart for explaining a method for predicting tool wear amount using meteorological measurement according to another example of the present invention.

[0035] Figure 13 is a series of images arranged in time to explain the process for calculating a correction angle corresponding to a reference position for measuring the free surface from images taken in relation to Figure 12.

[0036]

[0037] Hereinafter, a method and system for predicting tool wear using meteorological measurements related to the present invention will be described in detail with reference to the attached drawings. The terms used in this specification and claims are not limited to their dictionary meanings, but are appropriately defined and understood to best describe the present invention.

[0038] The main idea of ​​the present invention is to install a camera on a machine tool to directly acquire wear images of tools in use on the machine tool and predict the amount of wear through deep learning, utilizing on-machine measurement.

[0039] Figure 1 is an exemplary flowchart illustrating a method for predicting tool wear using meteorological measurements according to the present invention. The target tool for predicting wear is mounted on a machine tool. To this end, the cutting tool is first fixed to a tool holder and mounted on the spindle of the machine tool to machine the workpiece (S10).

[0040] FIG. 2 is a plan view conceptually showing a state in which a measuring device is installed on a machine tool to configure a wear prediction system of a tool using meteorological measurements related to the present invention. A cutting tool (30) mounted on a main shaft processes a workpiece (20) placed on a bed (20) of the machine tool according to set conditions. A camera (40) is installed on one side of the bed (20) to obtain a wear image of the cutting tool (30). The camera (40) is configured so that the rotation angle can be controlled by a servo motor (not shown) so that it can rotate according to the position of the cutting tool (30).

[0041] The control unit of the tool wear prediction system using weather measurements maintains a check status to determine whether it is time to capture a wear image of the tool (S20). At this stage, it is checked whether there is an instruction to capture an image of the cutting tool (30). The instruction to capture can be determined by directly receiving a control command from the user, or an operation related to capturing can be automatically initiated upon the achievement of a pre-stored condition. An exemplary condition may be whether a reference cutting time set for each tool or each processing step has elapsed.

[0042] When the time to capture the tool image arrives, the control unit commands the main axis of the machine tool to stop rotating (M19) (S30) and initializes the direction of the camera (40) (S40).

[0043] Accordingly, the spindle moves to the relative origin for measurement by the camera (40), and then the position of the flank surface of the cutting tool (30) is confirmed (S50). In the case of a turning tool, the position and direction of the tool can be maintained the same, but in the milling process, the probability that the flank surface, which is the main area of ​​wear measurement, is in the ideal position is very low. That is, in most cases, the flank surface of the tool is out of the direction perpendicular to the direction of the camera (40), so correction of the position of the camera (40) is required to obtain an ideal wear image. Fig. 3 shows the alignment state with the camera (40) according to the rotational position of the flank surface (31) of the cutting tool (30). FIG. 3(a) shows a case where the clearance surface (31) of the milling tool (30) is arranged vertically with respect to the camera (40), FIG. 3(b) shows a case where the clearance surface (31) is misaligned at an angle of θ1 in the counterclockwise direction with respect to the optical axis of the camera (40), and FIG. 3(c) shows a case where the clearance surface (31) is misaligned at an angle of θ2 in the clockwise direction. If the clearance surface (31) is aligned vertically with respect to the camera (40) as in FIG. 3(a), the image can be acquired by taking a picture as is, but if the clearance surface (31) is not vertical with respect to the camera (40) as in FIG. 3(b) and FIG. 3(c), the camera (40) must be moved and rotated to form a direction perpendicular to the clearance surface (31).

[0044] Figure 4 conceptually shows a correction algorithm for the camera position coordinates according to the position of the tool clearance surface (31). In order to correct the position and direction of the camera (40), the radius (R) of the cutting tool (30) and the focal length (l) of the camera (40) are used as in (a). F ) and the length of the camera (l C ) of the sum (R+l) F + l C) is measured and stored. If, as in Fig. 4(b), the clearance surface is misaligned at an angle of θ1 with respect to the camera, as the main axis of the machine tool is fixed, the camera (40) is moved and rotated to be corrected so as to be perpendicular to the clearance surface. At this time, cosθ = L' / L = (L-X') / L, and the relationship that X' = LL cosθ, sinθ = Y1 / L from L cosθ = L-X' is used. If the rotation center of the camera (40) is P, and the coordinates of the camera origin are (X0, Y0), when the rotational position of the tool is θ1, the position correction coordinates of the camera become (X0-X', Y0-Y1). The control unit corrects the position of the bed (20) so that the origin of the camera (40) is placed at the position correction coordinates, and rotates the camera (40) by θ1 so as to be aligned perpendicular to the clearance surface (31). In the same way, when the clearance surface is misaligned at an angle of θ2 with respect to the camera, as in Fig. 4(c), the position correction coordinates of the camera become (X0-X', Y0+Y1), and the control unit corrects the position of the bed (20) so that the origin of the camera (40) is placed at the position correction coordinates, and rotates the camera (40) by θ2 so that it is aligned vertically with respect to the clearance surface (31). By performing this procedure, it becomes possible to always capture tool wear images at a constant position thereafter, so that automation is possible.

[0045] In this way, referring back to FIG. 1, it is checked whether the position of the free surface (31) is in a direction perpendicular to the camera (40) (S60), and if it is in a vertical direction, a wear image of the free surface (31) is immediately taken (S61), and if it is not in a vertical direction, the camera (40) is rotated and its position is corrected (S65), and then a wear image of the free surface (31) is taken (S66).

[0046] The wear image of the cutting tool (30) captured by the camera (40) can be configured to be transmitted to a PC located outside the machining center via a wireless transmission system connected to the camera (S70). The transmitted image data is automatically stored in a designated location, and an image of the tool's clearance surface is continuously generated for each machining pass.

[0047] Tool wear can be predicted by automating image data acquisition via a camera system, measuring tool flank wear, constructing an image dataset, and then training the system using deep learning, an artificial intelligence technology (S80). Tool wear is measured and recorded using mean flank wear and maximum flank wear. Once a certain amount of image data is acquired, a dataset is constructed and used as training and validation data for tool wear prediction.

[0048] Deep learning encompasses a variety of algorithms, but among them, the Mask R-CNN model, capable of detecting and extracting desired regions at the pixel level, is particularly useful. Mask R-CNN is an image segmentation algorithm capable of instance segmentation. Even if identical objects are detected, they can be classified as distinct objects, and regions of interest can be extracted at the pixel level.

[0049] Figure 5 is an example showing the diagnosis results of the wear condition for a photographed image when using a general-purpose CNN (Convolutional Neural Networks) model, and Figure 6 is an example showing the diagnosis results of the wear condition for a photographed image of a tool when using a Mask R-CNN (Regional Convolutional Neural Network) model. When using a general-purpose CNN model (Fig. 5), it only presents the diagnosis result of whether there is wear (classification) and where the area (the part indicated by the dotted rectangle in the right picture) is (localization) from the captured image, but when using Mask R-CNN as a model (Fig. 6), the area with wear (indicated by the dense dotted rectangle in the left picture) and the area with damage (indicated by the long dotted rectangle in the left picture) can be detected (object detection), and through instance segmentation, the area with wear can be extracted pixel by pixel (the part indicated by the long dotted rectangle in the right picture) and the area with damage can also be extracted pixel by pixel (the part indicated by the dark black w shape within the long dotted rectangle in the right picture).

[0050] Tool wear is measured and recorded by measuring the mean flank wear and maximum flank wear. When a certain amount of image data is secured, a dataset is constructed and used as training and verification data for tool wear prediction (S80). The results of tool condition diagnosis using the instance segmentation technique of Mask R-CNN are exemplified in Figs. 7 and 8. In other words, Figs. 7 and 8 are examples showing the results of tool condition diagnosis using the instance segmentation technique among the image segmentation algorithms of Mask R-CNN. These are the results of performing tool condition diagnosis when a new tool wear image is input as input data after training the Mask R-CNN model. As shown in Fig. 7(b), the wear area and the damage area are classified with an accuracy of 99.8%. In this way, the wear area, which is the region of interest, can be extracted on a pixel-by-pixel basis, so wear can be predicted by utilizing the detected pixel data. Fig. 8 shows an example of the results of detecting tool wear areas on a pixel-by-pixel basis using the deep learning model proposed in the present invention. As shown in Fig. 8, if the average value of the pixels in the vertical direction is converted into length, it becomes the average free surface wear width, and if the maximum pixel value is converted into length, it becomes the maximum free surface wear width. The method of converting the pixel unit into length can be converted using the pixel number of the scale bar on the camera image.

[0051] FIGS. 9 and 10 are examples showing the results of prediction of tool wear amount using image segmentation deep learning for two different captured images compared with actual measured values, and FIG. 11 is a diagram showing the results of prediction of tool wear amount for 30 verification data sets not included in the learning data in relation to the present invention compared with actual measured values.

[0052] According to Fig. 9, the predicted values ​​based on the results of detecting the tool wear area in pixels (top of the left image) were VBmax = 0.071 mm and VBmean = 0.058 mm (right image), and compared to the results actually measured using a microscope (bottom of the left image), VBmax = 0.070 mm (measured value) and VBmean = 0.057 mm (measured value), it was found that it showed excellent performance within about 5% in terms of mean absolute percentage error (MAPE). The comparative result presented in Fig. 9 corresponds to No. 10 based on Val_data in the list of Fig. 11. Figure 10 is a comparison result corresponding to No. 15 based on Val_data in the list of Figure 11, showing that VBmax = 0.079 mm (predicted value) and VBmean = 0.063 mm (predicted value) are almost identical to VBmax = 0.078 mm (measured value) and VBmean = 0.061 mm (measured value).

[0053] Unlike the previous embodiment, where the camera was installed rotatably on the machine tool table, wear can be predicted even when the camera is fixedly installed on the machine tool table. FIGS. 12 and 13 illustrate other embodiments according to the present invention and are provided as examples to explain a method for doing so.

[0054] When the cutting tool is fixed to the tool holder and mounted on the spindle of the machine tool to process the workpiece (S110), and the time to capture the tool image has arrived (S120), a position command (M19 command) is transmitted to the spindle, and the spindle is positioned as the relative origin for measurement by the camera (S130).

[0055] In order to align the cutting tool mounted on the main shaft so that the clearance surface is perpendicular to the camera, the main shaft is rotated and images are continuously captured for a certain period of time using a motion detection algorithm (S140). Among the acquired images, as shown in Fig. 13, the nth image where the reference line, which is a solid line, and the clearance surface blade coincide, is the image captured at the origin correction time t from the image capture cycle. n Since it is possible to produce the origin correction angle (θ) from the following mathematical expression 1 (S150). Once the origin correction angle is calculated, the indexing function of the main axis can be used to rotate it by that angle to determine the free surface measurement position (S160).

[0056]

[0057] In the above equation, ω is the angular velocity (rad / sec) and N is the spindle rotation speed (rpm).

[0058] Thereafter, the process of taking a wear image of the free surface according to the determined measurement position (S170), constructing an image dataset based on the transmitted captured image (S180) and performing deep learning (S190), and segmenting the image of the wear area and predicting the amount of wear (S200) is common to the previous embodiment, and a detailed description thereof will be replaced with the previous description.

[0059] Using this algorithm, the reference position for measuring the clearance surface can be found even when the camera head cannot rotate, and the wear of the clearance surface of all cutting tool blades can be predicted by rotating the cutting tool blades by a certain angle from the reference position according to the number of cutting tool blades.

[0060] Thus, the tool wear measurement system and deep learning-based tool wear prediction method possess the versatility to be applied to a variety of tools with a single training session. Because this technology utilizes machine vision and deep learning to quickly and easily acquire tool wear images and predict wear in real time, it can be applied to a wide range of equipment and tools.

[0061] The method and system for predicting tool wear using meteorological measurements described above are not limited to the configuration and method of the described embodiments. The embodiments may be selectively combined in whole or in part to allow for various modifications. The scope of the present invention should be interpreted to include such embodiments or technical concepts within an equivalent scope.

Claims

1. A step of receiving an instruction to take an image of a cutting tool mounted on the main spindle of a machine tool; A step of commanding a fixed position on the main axis of the above machine tool; A step of initializing the direction of a camera installed in the above machine tool; A step of checking the position of the clearance surface of the above cutting tool; A step of capturing a wear image of the clearance surface when the position of the clearance surface of the cutting tool is in a direction perpendicular to the direction of the camera; A step of correcting the position of the camera and rotating the camera so that the direction of the camera faces the clearance surface when the position of the clearance surface of the cutting tool is not perpendicular to the direction of the camera; A step of constructing a dataset and performing deep learning on images captured by the above camera; and A method for predicting wear amount of a tool using meteorological measurements, comprising a step of predicting wear amount through an image of the wear area according to the performance of the above deep learning.

2. In paragraph 1, The step of receiving an instruction to take an image of a cutting tool mounted on the main axis of the above machine tool is: A method for predicting tool wear using meteorological measurements, which is started automatically by directly receiving a control command from a user or by achieving a pre-stored condition.

3. In paragraph 2, A method for predicting tool wear using meteorological measurements, wherein the above-mentioned stored conditions include the elapsed time of a standard cutting time.

4. In paragraph 1, In the step of constructing a dataset and performing deep learning on images captured by the above camera, The above deep learning is a method for predicting tool wear using meteorological measurements, using a model that includes object detection and image segmentation for the above-mentioned photographed image.

5. In paragraph 1, In the step of constructing a dataset and performing deep learning on images captured by the above camera, The above deep learning is a method for predicting tool wear using weather measurements, using the Mask R-CNN (Regional Convolutional Neural Network) model.

6. In paragraph 1, If the position of the clearance surface of the cutting tool is not perpendicular to the direction of the camera, the step of correcting the position of the camera and rotating the direction of the camera so that it faces the clearance surface is: A step of moving the bed on which the above camera is installed; and A method for predicting tool wear using meteorological measurements, comprising a step of rotating the above camera by a servo motor.

7. In paragraph 2, The step of moving the bed on which the above camera is installed is: The radius (R) of the above cutting tool and the focal length (l) of the above camera F ) and the length of the above camera (l C ) to maintain a constant sum, wherein the bed is moved in the horizontal direction (X) and vertical direction (Y) to predict the wear amount of the tool using meteorological measurements.

8. In paragraph 1, A method for predicting tool wear using meteorological measurements, further comprising a step of controlling an air blower to remove foreign substances from the surface of the cutting tool before taking an image using the camera.

9. In paragraph 1, The step of checking the position of the clearance surface of the above cutting tool is: A method for predicting tool wear using meteorological measurements, which identifies the location of the clearance surface of the cutting tool and derives the misalignment angle between the direction of the clearance surface and the optical axis of the camera.

10. Step of receiving instructions to take an image of a cutting tool mounted on the spindle of a machine tool; A step of moving the main axis of the machine tool to a measuring position according to a position command; A step of rotating the main axis of the above machine tool and filming it for a certain period of time using a camera; A step of calculating a correction angle corresponding to a reference position for measuring the clearance surface of a cutting tool from images captured by the above camera; A step of rotating the main axis of the cutting tool to reflect the above correction angle; A step of taking an image of the wear of a relief surface of a cutting tool fixed to a rotating spindle; A step of constructing a dataset and performing deep learning on images captured by the above camera; and A method for predicting wear amount of a tool using meteorological measurements, comprising a step of predicting wear amount through an image of the wear area according to the performance of the above deep learning.

11. A computer-readable storage medium storing a program configured to execute a method for predicting tool wear amount using meteorological measurements according to any one of claims 1 to 10.

12. A camera arranged on one side of a bed of a machine tool and configured to acquire an image of a cutting tool mounted on a main shaft, and installed rotatably so as to change direction according to the position of a clearance surface of the cutting tool; A data processing unit configured to receive and process an image of the cutting tool obtained by the camera; A deep learning model configured to enable object detection and image segmentation of the image of the above cutting tool; and A control unit configured to command a main axis of the machine tool to be in a fixed position according to an instruction for image capturing for the cutting tool, initialize the direction of the camera, check the position of the clearance surface of the cutting tool, and if the position of the clearance surface of the cutting tool is in a direction perpendicular to the direction of the camera, capture an image of the wear of the clearance surface, and if the position of the clearance surface of the cutting tool is not in a direction perpendicular to the direction of the camera, move the bed to correct the position of the camera and rotate the direction of the camera so that it faces the clearance surface, and perform deep learning according to the deep learning model to predict the amount of wear for the image of the captured wear area; A tool wear prediction system using meteorological measurements, including:

13. In paragraph 12, The above deep learning model is a tool wear prediction system using meteorological measurements, including Mask R-CNN (Regional Convolutional Neural Network).

14. A camera arranged on one side of the bed of the machine tool and configured to acquire an image of a cutting tool mounted on the spindle; A data processing unit configured to receive and process an image of the cutting tool obtained by the camera; A deep learning model configured to enable object detection and image segmentation of the image of the above cutting tool; and A wear prediction system for a tool using weather measurement, configured to command the machine tool to move the spindle to a measurement position, rotate the spindle of the machine tool and photograph the image for a predetermined period of time using a camera, calculate a correction angle corresponding to a reference position for measuring a clearance surface of a cutting tool from images photographed by the camera, rotate the spindle of the cutting tool by reflecting the correction angle, photograph a wear image of the clearance surface of the cutting tool fixed to the rotated spindle, construct a dataset and perform deep learning on the images photographed by the camera, and predict the wear amount through the image of the wear area according to the performance of the deep learning.

Images