Measurements of Tool Wear Parameters Using Machine Vision System

Monitoring tool wear is very important in machining industry as it may result in loss of dimensional accuracy and quality of ﬁnished product. This work includes the development of machine vision system for the direct measurement of ﬂank wear of carbide cutting tool inserts. This system consists of a digital camera to capture the tool wear image, a good light source to illuminate the tool, and a computer for image processing. A new approach of inline automatic calibration of a pixel is proposed in this work. The captured images of carbide inserts are processed, and the segmented tool wear zone has been obtained by image processing. The vision system extracts tool wear parameters such as average tool wear width, tool wear area, and tool wear perimeter. The results of the average tool wear width obtained from the vision system are experimentally validated with those obtained from the digital microscope. An average error of 3% was found for measurements of all 12 carbide inserts. Scanning electron micrographs of the wear zone indicate the severe abrasion marks and damage to the cutting edge for higher machining time. This study indicates that the eﬃcient and reliable vision system can be developed to measure the tool wear parameters.


Introduction
Measurement of tool wear is extremely important to predict the useful life of tool inserts. is will be helpful to monitor and to study the effects of the tool wear on quality of machined workpiece and economy of manufacturing process.
ere are two main methods to measure tool wear: indirect and direct methods. In the indirect method, tool wear is estimated with the signals coming from different types of sensors such as surface texture of machined workpiece, acoustics, vibration, feed forces, and current consumption [1][2][3][4][5]. e tool wear prediction model is prepared based on the magnitude of collected signals. Other method for the measurement of tool wear is the direct measurement over the tool wear zone. ere are two main tool wear types: flank wear and crater wear. e flank wear is widely used to quantify the severity of tool wear. Characteristics of qualitative and quantitative morphology of tool wear are of great concern for researchers nowadays. More morphological features other than commonly considered parameter, i.e., average tool wear width, are required for better evaluation of the actual condition of tool which can affect machining process and quality of machined workpieces [2,5]. e study shows that there is prominent effect of these new tool wear parameters on producing quality workpieces and also has an economic advantage by making strategies for timely changing tool inserts.
Tool wear generally results in loss in dimensional accuracy of finished products, possible damage to workpiece, and decrease in surface integrity and amplification of chatter. Detailed review for the tool condition monitoring indicates that the machine vision system can be extremely useful for the direct measurement of various types of tool wear [6]. Some statistical approaches are also useful in conjunction with machine vision system to find tool wear [1,2,5,7,8]. Some researchers developed their own algorithm for the edge detection and segmentation of tool zone [1,8]. White light interferometry [9][10][11] and stereo vision technique [12] are used for the measurement of volumetric wear in crater as well as flank wear region. Teti et al.
presented detailed review of the sensor technologies, signal processing, and decision-making strategies for the efficient machining systems [13].
Various methods are suggested for the online and offline condition monitoring of the machining tool. Danesh and Khalili measured tool wear in terms of surface texture of the workpiece during the turning process using undecimated wavelet transform and statistical features of the surface irregularities [5]. Yu et al. used morphological component analysis and edge detection techniques to detect the wear edges under carrying working conditions [7]. D'Addona and Teti used artificial neural network for the automatic and realtime evaluation of the crater wear depth during quasiorthogonal cutting tests on AISI 1045 steel using tungsten carbide insert [8]. Xiong et al. developed an image processing algorithm using Matlab to measure the tool wear area. e image acquisition system consists of highresolution CCD camera, fluorescent high-frequency linear lights, and data acquisition module [11]. Schmitt et al. developed an automatic tool wear monitoring system based on the active contour algorithm and neural networks for the flank wear measurement [14]. Fernández-Robles et al. developed an algorithm to measure the defects in cutting edges of milling inserts online without disturbing the machining operation. A three-stage algorithm consists of edge preserving smoothing filter, computation of image gradient, and assessment of damage of cutting edge using geometrical properties [15].
Some assumptions are also made to quantify the volume of wear region approximately. e wear at the tool nose was measured by assuming the part of cutting edge as a disk of radius equal to a tool nose radius [9]. Various geometrical parameters are determined for flank wear region by standardizing the wear region as an ellipse [2]. Researchers suggested various tool wear parameters such as maximum wear land width, wear land area, wear land perimeter [16] and compactness [3], length of major axis, length of minor axis, eccentricity, orientation, equivalent diameter, solidity and extent [2,4], end wear length [17], and nose radius and flank wear width [18].
Current work is focused on the measurement of flank wear using the machine vision system. A new approach of inline automatic calibration is proposed here. Average tool wear width, tool wear area, and tool wear perimeter are measured using the machine vision system. All these tool wear parameters are correlated with the machining time.

Methodology
A schematic diagram of the tool wear measurement system is shown in Figure 1. e digital camera (SONY Cyber-shot DSC-HX400 V) was used for capturing the image of worn out tool inserts. LED was used for illumination purpose of tool insert and calibration square. "Image processing toolbox" of the software MATLAB ® was used for the image processing. e worn out carbide tool inserts (SEMT 1304 PETR-M TT8020, TagueTec) were used for conducting the experiments in this work. e fixture has been developed for proper positioning of tool insert and the calibration square. Figure 2(a) shows the designed fixture for this purpose. e calibration square (10 × 10 mm 2 radium paper) was pasted on a vertical plate of fixture. e fixture was covered with a black paper to avoid the reflection of light. is square is also used for defining the origin of the virtual coordinate system developed for the measurement of tool wear volume. e vertical plate has provision to move forward and backward so that the plane of calibration square and tip of tool insert remains same. e position of the camera, light source, fixture, and insert is indicated in Figure 2(b). e image has been captured such that it contains both the calibration square and tool insert as indicated in Figure 3(a). Figure 3(b) shows the processed image of the calibration square.
Equations (1) and (2) give the calibration factor of pixel in horizontal and vertical directions. Figure 4 shows a flowchart of an algorithm for image processing to calculate tool wear parameters. e images of calibration square and tool wear zone are cropped and processed separately: horizontal calibration distance of a pixel, vertical calibration distance of a pixel, P y (mm) � side of the square(mm) number of pixels in the y direction .
(2) Figure 5 indicates the results of various stages of the image processing algorithm. Figure 5(a) indicates the grayscale image of the tool wear zone. Figure 5(b) indicates the binary image of the segmented tool wear zone obtained by using Otsu's thresholding method. A threshold value of 0.5216 has been computed by using the inbuilt function of software Matlab. Noises from the image are removed using the median filter. Figure 5(c) indicates the image of the wear zone obtained after applying the median filter. Figures 5(d) and 5(e) are the images of the wear zone obtained after the application of dilation and erosion operations. Finally, the canny edge detection algorithm is used to characterize the boundary of the wear zone. Figure 5(f ) is finally used to measure the various tool wear parameters. e algorithm is used to calculate three tool wear parameters, i.e., average tool wear width, tool wear area, and wear perimeter.
(1) Average tool wear width. At different locations in the segmented tool wear zone, the pixels were counted vertically. e average of these readings was multiplied by vertical calibration factor to get average tool wear width. (2) Tool wear area. e number of pixels in segmented tool wear zone was counted. e counted number of pixels was multiplied by the area of single pixel to get the tool wear area. (3) Tool wear perimeter. After skeletonization operation, only one pixel remains at the boundary of the tool wear zone. ese pixels were counted and multiplied by the horizontal calibrating factor to get the tool wear perimeter.

Results
In this section, the final results of measurements of three tool wear parameters measured by the vision system have been presented. For the measurement of the wear of tool insert, the turning experiments are conducted on low alloy steel which is widely used for the production of bearing cover of an IC (internal combustion) engine. Turning operations are conducted on a CNC turning machine (Seimens control; DX200, Jyothi Company) using carbide inserts (TNMG-16-04-04-QM J13 A) for the machining of internal diameter. e details of the machining parameters are indicated in Table 1.
For every turning operation, fresh carbide insert was used in which the machining and machining parameters were kept same. Machining time required for every specimen was around 5 minutes. Table 2 indicates the details of machining times for all the inserts. After the machining operation, the tool wear parameters of all the inserts were measured by the vision system. In order to validate the accuracy of the developed system, the average tool wear width for all inserts was also measured using a digital microscope (ViTiny UM05). Table 3 indicates the images of the wear zones of all the carbide inserts. ese images are obtained after the processing of actual images of tool wear by the machine vision  system. ese images are used to automatically determine the wear width of the carbide inserts. Wear width is measured at five locations, and the average value of the wear width is determined and compared with the readings of the digital microscope. e magnitudes of average wear width obtained by the vision system and the digital microscope are fairly close for all the inserts. An average error of approximately 3% between both the readings indicates that the machine vision system can correctly estimate the magnitude of tool wear. Figure 6 indicates the comparison of the average wear width obtained by both the techniques. Tool wear width is seen increasing with the machining time in almost linear fashion. When the insert is used for the machining of multiple workpieces, it undergoes the abrasion marks due to friction between the workpiece surface and the cutting edge.
And it increases with the number of machined workpieces or machining time. Figure 7 indicates the scanning electron micrographs of the wear zone of carbide inserts. e magnitudes of other tool wear parameters determined by the vision system for all twelve inserts are indicated in Table 4. Similarly, Figures 8(a)-8(c) indicate the variation of wear width, wear area, and wear perimeter, respectively, with the machining time. It indicates almost linear increase of wear parameters with the machining time. Interestingly the cutting speed, feed rate, and depth of cut were kept constant during the turning process.
e study indicates the application of the machine vision system for the measurement of flank wear in terms of average wear width, wear area, and wear perimeter. e machine vision system can be preferred over the conventional approach of measurement of only one parameter such as average wear width. e machine vision system can give exhaustive picture of the wear zone and thus can indicate the severity of tool wear more efficiently. Scanning electron micrographs indicate severe abrasion marks and damage to the cutting edge in the case of higher machining time. Previous researchers also indicate the more exhaustive approach for the measurement of tool wear and its effect on the machining accuracy. ree geometric descriptors, i.e., eccentricity, extent, and solidity, were found to contribute significantly to characterize the tool wear severity [2]. A new feature of flank wear, i.e., end wear length, was proposed in addition to flank wear area, average wear width, and maximum wear width for microdrill to predict the tool life [12]. is indicates that the use of only one parameter is inadequate to decide the extent of tool wear qualitatively and quantitatively. Hence, this vision system is designed to measure tool wear parameters such as tool wear area and tool wear perimeter in addition to the commonly used tool wear parameter, i.e., average tool wear width.

Conclusion
e inline automatic calibration system was successfully implemented for the measurement of tool wear parameters. With this calibration system, there is no need for separate calibration of the vision system. e measurements of an average tool wear width with the present vision system are found to be in close agreement with that with the digital microscope.
e average absolute error in measuring average tool wear width for all the twelve inserts was found to be 3.08%. Average wear width, wear area, and wear perimeter were seen increasing with the machining time. e scanning electron micrographs indicate severe abrasion     Insert number  1  2  3  4  5  6  7  8  9  10  11  12  Number of workpieces machined  1  2  3  4  5  6  7  8  9  10  11  12  Machining time (min)  5  10  15  20  25  30  35  40  45  50  55  60   marks and damage to the cutting edge in the case of higher machining time. is study shows that the machine vision system can be effectively used to measure all tool wear parameters and thus presents the correct and exhaustive picture of the tool wear. is methodology will be extremely useful for manufacturing industry to monitor tool wear effectively rather than relying only one parameter.
is will be extremely useful to study the effects of tool wear on quality of machined surface and economy of machining process.

Data Availability
Required data are already provided in the manuscript.

Disclosure
e work is carried out in V. N. I. T, Nagpur, as a part of dissertation work carried out by for the postgraduate students in Mechanical Engineering department under the guidance of Dr. A. A. akre.