An online detection method for end mill wear parameters

Abstract

The present invention provides an online detection method for end mill wear parameters. The method uses a laser displacement sensor measuring device installed at a corresponding position to return distance data. The method then uses the displacement capacity of the machine tool spindle to control the movement of the milling cutter, manually controls the small reciprocating rotation of the tool, collects point data of the defective part of the milling cutter, fits the point data, and obtains a two-dimensional defect profile of the milling cutter. Secondly, according to the camera shooting method along the spiral line of the milling cutter, a group of images are taken to splice the three-dimensional milling cutter blade into a two-dimensional milling cutter blade image, and the trained VGG16-Unet model is used to identify the wear area. Then, the defect area and the wear area of the milling cutter are fused to obtain the overall wear profile of the milling cutter. The present invention provides a more accurate detection method for milling cutter wear, and the installation of the camera and laser displacement sensor is simple and convenient, which improves the detection accuracy while also improving the detection efficiency.

Description

Technical Field

[0001] The present invention relates to the interdisciplinary field of advanced manufacturing technology in mechanical engineering, artificial intelligence and image processing based on machine vision, and in particular to an online detection method for wear parameters of an end mill. Background Art

[0002] During the workpiece machining process, the selection, installation, control and wear of the tool are all important links that need to be paid attention to during the use of the tool. During the working process of the tool, the tool material, workpiece material, machining parameters and other major factors directly affect the tool wear state. While the tool is wearing, it will also produce defects under long-term operation. Tool wear and defects directly affect the machining quality of the workpiece. The detection of tool wear is a crucial part of the machining process. For most wear detection schemes, only the wear of the front cutting edge of the end mill is often detected, thereby ignoring the defective part of the milling cutter. This is incomplete for the wear detection of the milling cutter blade, and the description of the wear of the entire milling cutter blade is inaccurate. In response to this situation, this article mainly solves the quantitative detection of the wear and defect amount of the spatial spiral blade of the end mill.

[0003] With the continuous advancement of computer vision theory and the rapid development of modern industrialization, machine vision can achieve intelligent, flexible, fast, and low-cost detection and classification during the machining process. Machine vision systems typically consist of cameras, lenses, and light sources to capture images and perform detection and classification in various industrial environments. Currently, most milling cutter wear inspections focus on the worn areas of the cutter, while cameras cannot detect defective areas, resulting in a decrease in the overall wear accuracy of the cutter.

[0004] Online detection of side edge wear and damage at high depths of cut. Most cutter wear detection solutions for high depths of cut often only detect scratches on the flank of the cutter's side edge, ignoring any damage. Furthermore, due to the large depth of cut and the three-dimensional curved surface of the side edge, a camera's field of view cannot capture all wear information on the end mill, making it impossible to obtain a complete picture of the cutter's wear. This results in incomplete detection of cutter edge wear and an inaccurate representation of the wear condition of the entire cutter edge. Summary of the Invention

[0005] Purpose of the invention: In response to the deficiencies in the prior art, the present invention provides an online detection method for the wear parameters of an end mill. A laser displacement sensor measuring device installed at a corresponding position returns distance data, and then the displacement capacity of the machine tool spindle is used to control the movement of the milling cutter. The tool is manually controlled to rotate back and forth in a small amplitude, and point data of the defective part of the milling cutter is collected and fitted to obtain a two-dimensional defect morphology of the milling cutter. Secondly, a group of images are taken according to a method of shooting along the spiral line of the milling cutter with a camera to splice the three-dimensional milling cutter blade into a two-dimensional milling cutter blade image, and the trained VGG16-Unet model is used to identify the wear area. Then, the defect area and the wear area of the milling cutter are fused to obtain the overall wear morphology of the milling cutter.

[0006] Technical solution: An online detection method for end mill wear parameters includes the following steps:

[0007] Step 1: Install the camera and laser displacement sensor on the machine tool and calibrate the camera pixel size;

[0008] Step 2: Use a camera to capture images of the milling cutter and stitch the captured images together;

[0009] Step 3: Train the VGG16-Unet model to be used, import the spliced image data into the VGG16-Unet model to obtain the tool wear binary map, and calculate the wear width value in the wear binary map;

[0010] Step 4: Adjust the laser displacement sensor to determine the tool acquisition position;

[0011] Step 5: Collect the data of the tool defect area through the laser displacement sensor and process the data of the defect area;

[0012] Step 6: Fit the milling cutter wear data and the milling cutter defect data.

[0013] In view of the situation that the milling cutter has a large cutting depth, the present invention proposes a shooting method along the spiral line of the milling cutter to ensure that the shooting angle and position of multiple groups of photos relative to the camera are consistent; and combines semantic segmentation to express the three-dimensional spatial surface of the flank face of the milling cutter on a two-dimensional plane, combined with the defect morphology fitted by laser sensor data, and finally, quantizes the three-dimensional surface of the flank face of the milling cutter on a two-dimensional plane and represents it through one-dimensional data, which more intuitively reflects the wear and defect status of the milling cutter, and facilitates the rapid judgment of the milling cutter status.

[0014] Preferably, the step 1 is specifically as follows:

[0015] The camera is fixedly mounted on the side wall near the machine tool spindle through a card holder. The camera is installed and debugged and the pixel size of the camera is calibrated to obtain the pixel equivalent value. The machine tool spindle is controlled to reach the camera shooting range position to capture the tool image. The calculation method used for the pixel size calibration is:

[0016] Width in pixels:

[0017]

[0018] Where: K is the width in pixels; L is the actual width of the ruler; N is the number of pixels used to represent the ruler width in the image; the ruler is a standard object of known size.

[0019] Preferably, the step 2 is specifically as follows:

[0020] The tool edges are numbered and the relative distance between the tool and the camera is adjusted to present a clear image of the tool side edge in the camera field of view. A complete image of the tool side edge is acquired along the tool spiral edge, and the acquired images are stitched together to stitch the captured fragmented images into a complete milling cutter image.

[0021] Preferably, the step 3 is specifically as follows:

[0022] After the images are stitched together, the processed images are input into the pre-trained convolutional neural network VGG16-Unet model.

[0023] The wear binary map of the input image can be obtained through the prediction of the model; the minimum enclosing rectangle of the tool wear binary map is drawn, the width of the minimum enclosing rectangle is calculated, and the obtained width is multiplied by the width pixel equivalent to obtain the actual width of the tool wear area.

[0024] Preferably, the step 4 is specifically as follows:

[0025] Fix the laser sensor on the other side wall close to the machine tool spindle, use the holder and fixture to achieve a close connection between the laser sensor and the machine tool, install and debug the laser sensor, and move the tool through the machine tool so that the tool axis and the center line of the laser sensor's laser beam are in the same plane and perpendicular to each other to complete the centering work;

[0026] Open the laser sensor acquisition software, move the tool through the machine tool, ensure that the laser sensor beam has a reading when it is at the tool tip position, and start the acquisition work.

[0027] Preferably, the step 5 is specifically as follows:

[0028] At the tool tip position, rotate the machine tool spindle to obtain point data at the tool tip position. Move the tool through the machine tool to obtain point data at the next position. The movement direction is that the Z-axis tool holder of the machine tool points to the tool tip. Each time the tool moves along the Z axis, the distance h = 0.2 mm is used. Start collecting data from the tool tip until the unworn area of the milling cutter is collected. The number of collections is n:

[0029]

[0030] Where: H = cutting depth ap / Z axis moving distance h,

[0031] And collect data again at the unworn position and record the point data of its position;

[0032] The data collected by the laser sensor is processed and a line graph of the points is drawn. According to the acquisition principle of the laser sensor, the required defective edge data is the point data closest to the laser sensor. The data of each edge corresponding to each collected sample is extracted and the average is calculated, that is, the data of each collection position corresponding to the radial direction of the tool;

[0033] The defect in the radial direction of the milling cutter is converted into the defect on the back face of the milling cutter. The formula is as follows:

[0034] b 2 =a 2 +c 2 -2ac cosθ

[0035] b=ae

[0036] Where: a is the radius of the milling cutter; b is the radius of the milling cutter after the defect occurs; e is the radial defect value; θ is the acute angle between the rake face and the flank face of the milling cutter radial section; c is the desired flank face defect value of the milling cutter.

[0037] Preferably, the step 6 is specifically as follows:

[0038] The number of images n taken to completely capture a milling cutter blade, where the maximum scratch value of the i-th image is W i , the corresponding maximum defect value is V i (where i = 1, 2, ..., n), the following method is proposed to calculate the tool surface wear and defect results, formula:

[0039]

[0040] Preferably, the image acquisition method in step 2 is specifically as follows:

[0041] The camera does not move, and the tool rotates and moves. The moving distance and the rotation angle can be calculated. That is, the Z-axis distance of the moving tool is 1mm, and the rotation angle is Ф. The formula is as follows:

[0042]

[0043] Where: β is the helix angle of the milling cutter, and D is the diameter of the milling cutter.

[0044] Beneficial effects: The present invention proposes a method for shooting along the spiral line of the milling cutter when the milling cutter has a large cutting depth, ensuring that the shooting angles and positions of multiple groups of photos relative to the camera are consistent; and combining semantic segmentation to express the three-dimensional spatial surface of the back face of the milling cutter side edge on a two-dimensional plane, combined with the defect morphology fitted by the laser sensor data, and finally, quantifying the three-dimensional surface of the back face of the milling cutter side edge to a two-dimensional plane, and representing it through one-dimensional data, which more intuitively reflects the wear and defect status of the milling cutter, and facilitates the rapid judgment of the milling cutter status. By obtaining the overall picture of the milling cutter wear, the degree of milling cutter wear can be effectively observed, and whether it is in a uniform wear state. For workpieces with relatively high processing precision, the cutter can be replaced in time, avoiding the processing deviation caused by tool problems affecting product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0046] Figure 1 It is a flow chart of the steps of the present invention.

[0047] Figure 2 The wear and tear of the end mill and the schematic diagram of the present invention are shown in FIG.

[0048] Figure 3 It is a schematic diagram of the measurement position of the camera and laser displacement sensor of the present invention.

[0049] Figure 4 It is a schematic diagram of image acquisition by the camera shooting method of the present invention.

[0050] Figure 5 It is the U-net structure diagram of the present invention.

[0051] Figure 6 This is a diagram of the VGG16 structure of the present invention.

[0052] Figure 7 This is a real wear picture taken by the camera of the present invention along the spiral line of the milling cutter.

[0053] Figure 8 The camera of the present invention takes a mosaic of images along the spiral line of the milling cutter.

[0054] Figure 9It is a radial diagram of the milling cutter of the present invention.

[0055] 1 is the main spindle, 2 is the milling cutter, 3 is the camera, 4 is the camera support structure, 5 is the bolt structure, 6 is the laser displacement sensor, 7 is the laser displacement sensor support structure, 8 is the bolt structure, and 9 is the machine tool bed. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] In the description of the present invention, it should be understood that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0058] In the present invention, unless otherwise expressly specified or limited, a first feature being "above" or "below" a second feature may include the first and second features being in direct contact, or may include the first and second features being in contact not directly but through another feature between them. Furthermore, a first feature being "above," "above," and "above" a second feature may include the first feature being directly above or obliquely above the second feature, or may simply mean that the first feature is higher in level than the second feature. A first feature being "below," "below," and "below" a second feature may include the first feature being directly below or obliquely below the second feature, or may simply mean that the first feature is lower in level than the second feature.

[0059] like Figures 1 to 9 As shown, an online detection method for end mill wear parameters includes the following steps:

[0060] Step 1: Install the camera and laser displacement sensor on the machine tool and calibrate the camera pixel size;

[0061] The image acquisition device is installed on the side wall near the machine tool spindle. The card holder is used to achieve a close connection between the acquisition device and the machine tool. The visual system is installed and debugged, and the camera pixel size is calibrated to obtain the pixel equivalent value. The machine tool spindle is controlled to reach a position near the visual system to prepare for tool image acquisition. The pixel size calibration algorithm used is:

[0062] Width in pixels:

[0063]

[0064] Where: K is the width in pixels; L is the actual width of the ruler; N is the number of pixels used to represent the ruler width in the image; the ruler is a standard object of known size;

[0065] Step 2: Use a camera to capture images of the milling cutter and stitch the captured images together;

[0066] The image acquisition method is: the camera does not move, the tool rotates and moves, the moving distance and the rotation angle can be calculated, that is, the Z-axis distance of the moving tool is H2, and the rotation angle is Ф2. For every 1mm increase, the rotation angle is:

[0067]

[0068] Where: β is the helix angle of the milling cutter, and D is the diameter of the milling cutter.

[0069] The tool edges are numbered and the relative distance between the tool and the camera is adjusted to provide a clear image of the tool side edge in the camera field of view. Three complete images of the tool side edge are collected along the tool spiral edge, and the collected images are preprocessed.

[0070] Step 3: Train the VGG16-Unet model to be used, import the spliced image data into the VGG16-Unet model to obtain the tool wear binary map, and calculate the wear width value in the wear binary map;

[0071] After preprocessing the image, the processed image is input into the pre-trained convolutional neural network VGG16-Unet model.

[0072] The VGG16-Unet model is a combination of U-net and VGG16.

[0073] The image training process is: collect images and create image labels - training - verification - testing.

[0074] The backbone feature extraction part of U-net consists of convolution + maximum pooling, and the overall structure is similar to VGG. The backbone feature extraction network of U-net used in this paper is replaced with VGG16.

[0075] The U-net network structure is symmetrical and looks like the English letter U, so it is called U-net. The original architecture of U-net can be found in Figure 5The entire image is composed of gray / white boxes and arrows of various colors. The gray / white rectangles represent feature maps; arrow ① (conv3x3, ReLU) represents 3x3 convolution for feature extraction; arrow ② (copyand crop) represents skip-connection for feature fusion; arrow ③ (max pool 2x2) represents pooling for dimensionality reduction; arrow ④ (up-conv 2x2) represents upsampling for dimensionality recovery; arrow ⑤ (conv 1x1) represents 1x1 convolution for output.

[0076] The wear binary map of the input image can be obtained through the prediction of the model;

[0077] The backbone feature extraction part of U-Net used in this invention is replaced with VGG16, and its name is VGG16-UNet. The VGG16-UNet structure is divided into three parts. The first part is the backbone feature extraction part. This part adopts the first 4 pooling and 13 convolution kernels of VGG16 as the feature extraction network structure, and uses the stacking operation of convolution plus maximum pooling for downsampling. The original 512 pixels × 512 pixels × 3 pixels is converted to 32 pixels × 32 pixels × 512 pixels after 4 downsampling to complete the feature extraction. Using the backbone feature extraction part, we can To obtain four preliminary effective feature layers, in the second part, these four effective feature layers are used for feature fusion; the second part is the enhanced feature extraction part, which is upsampling and convolution stacking to complete the restoration of the milling cutter wear feature map, and perform feature fusion to obtain a final effective feature layer that integrates all features, and restore the final output feature map to 512 pixels × x512 pixels in size; the third part is the prediction part, which uses the last effective feature layer obtained to classify each feature point, which is equivalent to classifying each pixel point, and outputting a binary map of milling cutter wear.

[0078] Draw the minimum circumscribed rectangle of the tool wear binary image, calculate the width of the minimum circumscribed rectangle, and multiply the width by the width pixel equivalent to obtain the actual width of the tool wear area, where N1 is the number of pixels occupied by the wear area in the width direction of the first image, and L2 is the corresponding actual width; where N2 is the number of pixels occupied by the wear area in the width direction of the second image, and L1 is the corresponding actual width; where N3 is the number of pixels occupied by the wear area in the width direction of the third image, and L3 is the corresponding actual width;

[0079] L1=K×N1=0.002232×14.16=0.0316mm

[0080] L2=K×N2=0.002232×14.95=0.0334mm

[0081] L3=K×N3=0.002232×145.33=0.0342mm

[0082] Installation and alignment of laser displacement sensor:

[0083] Install the laser sensor device on the side wall near the machine tool spindle, close to the vision system, and without affecting the normal processing of the machine tool. Use the holder and fixture to achieve a close connection between the laser sensor device and the machine tool. Install and debug the laser sensor device and move the tool through the machine tool so that the tool axis and the center line of the laser sensor's laser beam are in the same plane and perpendicular to each other to complete the centering work;

[0084] Step 4: Adjust the laser displacement sensor to determine the tool acquisition position;

[0085] Open the laser sensor acquisition software, move the tool through the machine tool, ensure that the laser sensor beam is reading when it is at the tool tip position, and start the acquisition work;

[0086] Step 5: Collect the data of the tool defect area through the laser displacement sensor and process the data of the defect area;

[0087] At the tool tip position, rotate the machine tool spindle to obtain point data at the tool tip position. Move the tool through the machine tool to obtain point data at the next position. The movement direction is that the Z-axis tool holder of the machine tool points to the tool tip. Each time the tool moves along the Z axis, the distance h = 0.2 mm is used. Start collecting data from the tool tip until the unworn area of the milling cutter is collected. The number of collections is n:

[0088]

[0089] Where: H = cutting depth ap / Z axis moving distance h,

[0090] Then collect data at the unworn position and record the point data of its position, collecting a total of 17 sets of data;

[0091] The data collected by the laser sensor is processed and a line graph of the points is drawn. According to the acquisition principle of the laser sensor, the required defective blade data is the point data closest to the laser sensor. Five groups of data values of the collected blade in this example are extracted, and the maximum and minimum values are removed to calculate their average.

[0092] The defect in the radial direction of the milling cutter is converted into the defect on the back face of the milling cutter. The formula is as follows:

[0093] b 2 =a 2 +c 2 -2ac cosθ

[0094] b=ae

[0095] Where: a is the radius of the milling cutter; b is the radius of the milling cutter after the defect occurs; e is the radial defect value; θ is the acute angle between the rake face and the flank face of the milling cutter radial section; c is the desired flank face defect value of the milling cutter. Positions 15 and 16 are the tip of the milling cutter. Due to the special chamfer of the tip, they are not considered in this paper.

[0096] Table 1 Data processing results

[0097]

[0098] Step 6: Fit the milling cutter wear data and the milling cutter defect data.

[0099] The number of images taken to fully capture a milling cutter blade is 3. V1 = 0.0316 mm, V2 = 0.0334 mm, and V3 = 0.0342 mm are the maximum wear values corresponding to each image. The corresponding maximum defect values are W1 = 0.0563 mm, W = 20.0639 mm, and W3 = 0.0611 mm. W is obtained by the following formula: mean and V mean :

[0100]

[0101] W mean =0.0604mm

[0102] V mean =0.0331mm

[0103] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0104] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An online detection method for end mill wear parameters, characterized in that: It includes the following steps: Step 1: Install the camera and laser displacement sensor on the machine tool and calibrate the camera pixel size; Step 2: Use a camera to capture images of the milling cutter and stitch the captured images together; The step 2 is specifically as follows: The tool edges are numbered and the relative distance between the tool and the camera is adjusted to provide a clear image of the tool side edge in the camera field of view. A complete image of the tool side edge is acquired along the tool spiral edge, and the acquired images are stitched together to form a complete milling cutter image. Step 3: Train the VGG16-Unet model to be used, import the spliced image data into the VGG16-Unet model to obtain the tool wear binary map, and calculate the wear width value in the wear binary map; The step 3 is specifically as follows: After the images are stitched together, the processed images are input into the pre-trained convolutional neural network VGG16-Unet model. The wear binary map of the input image can be obtained through the prediction of the model; the minimum enclosing rectangle of the tool wear binary map is drawn, and the width of the minimum enclosing rectangle is calculated. The obtained width is multiplied by the pixel equivalent to obtain the actual width of the tool wear area; Step 4: Adjust the laser displacement sensor to determine the tool acquisition position; Step 5: Collect the data of the tool defect area through the laser displacement sensor and process the data of the defect area; The step 5 is specifically as follows: At the tool tip position, rotate the machine tool spindle to obtain point data at the tool tip position. Move the tool through the machine tool to obtain point data at the next position. The movement direction is that the Z-axis tool holder of the machine tool points to the tool tip. Each time the tool moves along the Z axis, the distance h = 0.2 mm is used. Start collecting data from the tool tip until the unworn area of the milling cutter is collected. The number of collections is n: Where: H = cutting depth ap / Z axis moving distance h, And collect data again at the unworn position and record the point data of its position; The data collected by the laser sensor is processed and a line graph of the points is drawn. According to the acquisition principle of the laser sensor, the required defective edge data is the point data closest to the laser sensor. The data of each edge corresponding to each collected sample is extracted and the average is calculated, that is, the data of each collection position corresponding to the radial direction of the tool; The defect in the radial direction of the milling cutter is converted into the defect on the back face of the milling cutter. The formula is as follows: b 2 =a 2 +c 2 -2ac cosθ b=ae Where: a is the radius of the milling cutter; b is the radius of the milling cutter after the defect occurs; e is the radial defect value; θ is the acute angle between the rake face and the flank face of the milling cutter radial section; c is the desired flank face defect value of the milling cutter; Step 6: Fit the milling cutter wear data and the milling cutter defect data; The step 6 is specifically as follows: The number of images n taken to completely capture a milling cutter blade, where the maximum scratch value of the i-th image is W i , the corresponding maximum defect value is V i (where i = 1, 2, ..., n), the following method is proposed to calculate the tool surface wear and defect results, formula:

2. The online detection method for end mill wear parameters according to claim 1, characterized in that: The step 1 is specifically as follows: The camera is fixedly mounted on the side wall near the machine tool spindle through a card holder. The camera is installed and debugged and the pixel size of the camera is calibrated to obtain the pixel equivalent value. The machine tool spindle is controlled to reach the camera shooting range position to capture the tool image. The calculation method used for the pixel size calibration is: Width in pixels: Where: K is the width in pixels; L is the actual width of the ruler; N is the number of pixels used to represent the ruler width in the image; the ruler is a standard object of known size.

3. The online detection method for end mill wear parameters according to claim 2, characterized in that: The step 4 is specifically as follows: Fix the laser sensor on the other side wall close to the machine tool spindle, use the holder and fixture to achieve a close connection between the laser sensor and the machine tool, install and debug the laser sensor, and move the tool through the machine tool so that the tool axis and the center line of the laser sensor's laser beam are in the same plane and perpendicular to each other to complete the centering work; Open the laser sensor acquisition software, move the tool through the machine tool, ensure that the laser sensor beam has a reading when it is at the tool tip position, and start the acquisition work.

4. The online detection method for end mill wear parameters according to claim 2, characterized in that: The image acquisition method in step 2 is specifically as follows: The tool does not move, and the camera moves along the spiral line of the tool to take pictures, obtaining a set of pictures with the same shooting angle.

5. The online detection method for end mill wear parameters according to claim 2, characterized in that: The image acquisition method in step 2 is specifically as follows: The camera does not move, and the tool rotates and moves. The moving distance and the rotation angle can be calculated. That is, the Z-axis distance of the moving tool is 1mm, and the rotation angle is Ф. The formula is as follows: Where: β is the helix angle of the milling cutter, and D is the diameter of the milling cutter.

Images